University of Southern California Dissertations and Theses

Absolute glare and relative glare factors: predicting and quantifying levels of interior glare and exterior glare caused by sunlight and daylight

(USC Thesis Other)

Absolute glare and relative glare factors: predicting and quantifying levels of interior glare and exterior glare caused by sunlight and daylight

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content i

Absolute Glare and Relative Glare Factors:
Predicting and Quantifying Levels of Interior Glare and Exterior Glare Caused by
Sunlight and Daylight

by
Jae Yong Suk

A dissertation
submitted in partial fulfillment of the
requirements for the degree of

Doctor of Philosophy in Architecture

University of Southern California

Committee:
Professor Marc Schiler, Chair
Professor Karen Kensek
Professor Bosco Tjan

June 2014
ii
Absolute Glare and Relative Glare Factors: Predicting and Quantifying Levels of
Interior Glare and Exterior Glare Caused by Sunlight and Daylight

©Copyright 2014

Jae Yong Suk
iii
ABSTRACT

Absolute Glare and Relative Glare Factors: Predicting and Quantifying Levels of
Interior Glare and Exterior Glare Caused by Sunlight and Daylight

Jae Yong Suk
Doctor of Philosophy in Architecture
University of Southern California
Professor Marc Schiler, Chair

Building facades comprise the first layer between the occupant and the outside world.
They are largely responsible for conserving building energy, encouraging occupant
comfort, providing views, and allowing for the expression of design aesthetics. The
trend toward sustainable architecture has inspired an increasing number of
transparent building facades. Transparent facades allow more natural light into the
interior spaces and are more thermally reflective, which lessens the solar heat gain of
the building. The use of transparent materials to harvest daylight can result in
occupant discomfort from glare or veiling reflections, while the use of reflective and
specular materials can cause momentary blindness and produce excessive thermal
discomfort for those outside the building. Complex building geometry, including
curved or faceted facades, have made it even more difficult to predict when or where
iv
sunlight will cause discomfort glare. Even though many research groups have
developed their own analysis metrics and tools to address interior glare issues,
existing glare metrics remain too inconsistent, inaccurate, and complicated to be
incorporated into daylighting practice. Hence, the existing metrics are used by only a
few researchers.

Many absolute luminance thresholds and specific contrast ratio values have also been
developed in various lighting codes and standards, such as the Swedish energy
authority NUTEK, ISO Standard 9241-6, and ANSI/IESNA RP-1 VDT Lighting
Standard. These thresholds are inconsistent with each other. Due to these problems
and limitations, this study aimed to develop a new daylight glare analysis method
that is practical enough to be utilized in practice without compromising evaluation
accuracy. Human subject studies have been performed inside and outside buildings
with the help of high dynamic range (HDR) imaging techniques and light sensors.
An HDR image analysis tool was developed in MATLAB to analyze digitally
captured glare scenes, to calculate luminance values, and to visualize glare sources in
the field of view. Along with the MATLAB tool, a glare analysis program called
Evalglare was used to calculate the five existing glare index scores of the HDR
images. Collected subjective evaluation data and the visual scenes have been
analyzed statistically to examine the advantages and disadvantages of the existing
glare indices and then to develop a new method based upon these revelations.

v
Absolute glare factor (AGF) and relative glare factor (RGF) were used to explain the
different causes of daylight glare problems inside and outside the buildings. Every
glare scene has both AGF and RGF, although one of them can be more dominant in
causing discomfort glare than the other. AGF causes discomfort glare with excessive
glare source brightness, while RGF causes glare with high contrast ratios between
task luminance and glare source luminance. Based on the human subject study
results, the categories were successfully defined within absolute luminance ranges
and glare ratios for different office task activities. These categories were:
imperceptible, perceptible, disturbing, and intolerable. When no task is being
performed inside the building by the occupant, an imperceptible glare has up to
2,752 cd/m
2
, a perceptible glare occurs between 2,752 cd/m
2
and 7,000 cd/m
2
, a
disturbing glare is dominant from 7,000 cd/m
2
to 12,522 cd/m
2
, and an intolerable
glare occurs from 12,522 cd/m
2
and up. When a computer-based typing task is being
performed, an imperceptible glare occurs up to 1,920 cd/m
2
, a perceptible glare
occurs between 1,920 cd/m
2
and 5,000 cd/m
2
, a disturbing glare is dominant from
5,000 cd/m
2
to 11,718 cd/m
2
, and an intolerable glare occurs from 11,718 cd/m
2
and
up. When a paper-based writing task is being performed, there is an imperceptible
glare up to 1,696 cd/m
2
, a perceptible glare between 1,696 cd/m
2
and 5,263 cd/m
2
,
and a disturbing or intolerable glare occurs from 5,263 cd/m
2
and up. These defined
glare categories with and without task activity clearly show that occupants have
different visual sensitivities depending on whether and what kind of office task
activity is involved. Vertical illuminance and glare ratio ranges were also
vi
successfully defined for different glare categories. Unlike the luminance and vertical
illuminance ranges, glare ratios were defined only for typing task scenes, since the
no-task and writing task scenes were not RGF dominant in the interior glare research
setting. When a computer-based typing task is being performed, imperceptible glare
is dominant up to 12, the glare ratio between task luminance and glare source
luminance while disturbing glare is dominant from 22 to 32, and intolerable glare
occurs beyond the glare ratio of 32. Perceptible glare occurs between 12 and 22 glare
ratios.

Finally, AGF- and RGF-based glare equations were developed to evaluate perceived
glare categories. AGF and RGF zones were also displayed on a scatter plot graph to
visually explain which glare factor is more dominant to cause discomfort glare. After
validating the existing glare indices using human subject study data, the study
presents several ideas to future users to assist their understanding and application of
existing glare metrics. Even though an exterior glare study was not conclusive
enough to develop absolute and relative glare factors, we hope that the study will
initiate new exterior glare human subject studies to prove the existence of exterior
visual discomfort issues caused by strong sunlight reflections on specular building
envelopes. The findings suggest that such an exterior glare study needs a different
research method from that of the interior glare research method used in this study.
The glare factor analysis method can provide the causes and results of discomfort
glare; it may help daylight glare analysis procedures to be more widely used in
vii
practice with the over-arching goal that daylight glare problems can be prevented
prior to building construction or properly remediated after building completion.
viii
TABLE OF CONTENTS
LIST OF FIGURES ................................................................................................. xiv
LIST OF TABLES .................................................................................................. xxii
ACKNOWLEDGEMENTS .................................................................................. xxiii
Chapter 1 Introduction: Daylighting and Glare ....................................................... 1
1.1 Benefits of Daylighting in Buildings ...................................................... 1
1.2 Disadvantages of Daylighting Inside Buildings ..................................... 6
1.3 Disadvantages of Reflected Sunlight Outside Buildings ...................... 11
1.4 Glare Categories ................................................................................... 19
1.4.1 Glare Categories Based on the Results of Glare .................................. 20
1.4.2 Glare Categories Based on the Perceived Degree of Glare.................. 22
1.4.3 Glare Factors Based on Process ........................................................... 22
1.5 Problem Statement ................................................................................ 25
1.6 Research Objectives .............................................................................. 29
Chapter 2 Literature Review: Daylight Glare and Building Envelopes ................ 31
2.1 Material Characteristics of Building Envelopes ................................... 31
2.1.1 Transparent Materials (Glass Facade) .................................................. 33
2.1.2 Reflective and Specular Materials ....................................................... 36
2.2 Lighting Fundamentals for Glare Analysis ........................................... 39
2.2.1 Luminance and Illuminance ................................................................. 40
2.2.2 Contrast (Brightness Difference) ......................................................... 45
2.2.3 Visual Adaptation................................................................................. 47
2.3 Existing Discomfort Glare Indices ....................................................... 49
ix
2.3.1 Daylight Glare Probability (DGP) and Simplified DGP (DGPs) ........ 50
2.3.2 Daylight Glare Index (DGI) ................................................................. 53
2.3.3 British Glare Index (BGI) .................................................................... 55
2.3.4 Visual Comfort Probability .................................................................. 55
2.3.5 Unified Glare Rating ............................................................................ 56
2.3.6 CIE Glare Index ................................................................................... 58
2.3.7 Relative Visual Performance................................................................ 58
2.3.8 Video Photometry Method ................................................................... 60
2.4 Existing Discomfort Glare Thresholds ................................................. 60
2.4.1 Luminance Thresholds ......................................................................... 60
2.4.2 Luminance Ratios (Contrast Ratios) .................................................... 63
2.4.3 Illuminance Thresholds ........................................................................ 65
2.5 Existing Discomfort Glare Analysis Tools ........................................... 66
2.5.1 Findglare (Radiance Visual Comfort Calculation) .............................. 66
2.5.2 Evalglare .............................................................................................. 67
2.5.3 Per-pixel Lighting Data Analysis ......................................................... 69
2.6 High Dynamic Range (HDR) Imaging Technique ............................... 75
2.6.1 HDR Photographing ............................................................................. 79
2.6.2 Luminance Mapping ............................................................................ 83
2.6.3 Limitations of HDR Imaging ............................................................... 85
Chapter 3 Preliminary Study: Issues within Existing Daylight Glare Analysis
Tools and Methodologies ......................................................................................... 87
3.1 Interior Daylight Glare Analysis on USC Watt Hall third floor ........... 87
x
3.1.1 HDR Photography of Interior Glare Issues .......................................... 89
3.1.2 Ecotect-Radiance Simulation of Interior Glare Issues ......................... 93
3.1.3 Inconsistent Evaluations of Five Glare Indices .................................... 97
3.1.4 DGP Score Comparisons of Simulated PIC files ............................... 108
3.2 Investigation of Evalglare ................................................................... 112
3.2.1 Research Method ................................................................................ 112
3.2.2 Results ................................................................................................ 115
3.3 Problems of Existing Glare Analysis Methods and Tools .................. 123
3.4 Field of View Issue ............................................................................. 123
3.5 Ideas for a New Daylight Glare Analysis Methodology ..................... 125
3.5.1 Preliminary Findings—Luminance Ranges ....................................... 127
3.5.2 Hypothetical Findings from the Watt Hall third floor Glare Study ... 129
3.5.3 Discussion .......................................................................................... 133
Chapter 4 Research Methods: Interior and Exterior Human Subject Studies ..... 134
4.1 Human Subject Study: Participants .................................................... 135
4.2 Human Subject Study: Equipment ...................................................... 137
4.3 Human Subject Study: Interior Glare ................................................. 140
4.4 Lighting Conditions and Tasks ........................................................... 142
4.4.1 Detailed Interior Glare Research Procedure ...................................... 144
4.4.2 Visual Maps and Questionnaires ........................................................ 146
4.5 Human Subject Study: Exterior Glare ................................................ 150
4.5.1 Lighting Conditions and Tasks .......................................................... 152
4.5.2 Detailed Exterior Glare Research Procedure ..................................... 153
xi
4.5.3 Visual Maps and Questionnaires ........................................................ 155
4.6 MATLAB Code Development ............................................................ 156
Chapter 5 Validation Study: Human Subject Data Analysis Using Existing Glare
Indices……………… ............................................................................................. 164
5.1 Analysis of Interior Glare Human Subject Study Data ....................... 164
5.2 Analysis based on Glare Category (Sensation) ................................... 171
5.2.1 Imperceptible Glare Scenes Only ...................................................... 172
5.2.2 Perceptible Glare Scenes Only ........................................................... 174
5.2.3 Disturbing/Intolerable Glare Scenes Only ......................................... 177
5.3 Glare Score Ranges from Human Subject Study ................................ 180
Chapter 6 Results and Analysis: Development of the AGF and RGF Method
Based on Human Subject Study Results and Analysis ........................................... 193
6.1 Interior Glare Study Results ............................................................... 193
6.1.1 Percentages of Subject Processed in MATLAB ................................ 196
6.1.2 Analysis of Variance (ANOVA) Test: Subjective Response
Consistency ....................................................................................................... 202
6.1.3 Field of View: Full Fisheye vs. Human Eye ...................................... 208
6.1.4 Glare Source Detection in Visual Map and HDR Image ................... 213
6.1.5 ANOVA Test: Glare Source Minimum Luminance .......................... 221
6.1.6 Absolute Luminance Ranges for Glare Category .............................. 224
6.1.7 Glare Source Detection with Absolute Luminance Thresholds ......... 236
6.1.8 Absolute Illuminance Ranges for Glare Category ............................. 246
6.1.9 Absolute Glare Factor (AGF) and Relative Glare Factor (RGF) ....... 256
xii
6.2 Exterior Glare Study Results .............................................................. 271
6.2.1 Percentages of Subject Glare Sensations ........................................... 273
6.2.2 ANOVA Test: Subjective Response Consistency.............................. 278
6.2.3 Field of View: Full Fisheye vs. Human Eye ...................................... 285
6.2.4 Glare Source Detection in Visual Map and HDR Image ................... 289
6.2.5 ANOVA Test: Glare Source Minimum Luminance .......................... 294
6.2.6 Absolute Luminance Ranges for Glare Category .............................. 296
6.3 Analysis Summary .............................................................................. 301
Chapter 7 Conclusions and Future Work: AGF- and RGF-Based Daylight Glare
Analysis Method ..................................................................................................... 304
7.1 Conclusions ......................................................................................... 304
7.1.1 Interior Glare: Visual Sensitivity for Different Office Activities ...... 309
7.1.2 Interior Glare: Total FOV vs. Human Eye FOV ................................ 310
7.1.3 Interior Glare: The Meaning of Perceptible Glare Sensation............. 311
7.1.4 Interior Glare: Validations of Existing Glare Indices ........................ 312
7.1.5 Exterior Glare: Investigation of the Exterior Glare Issue from Specular
Building Envelopes ........................................................................................... 314
7.1.6 Research Method: Expert Group and Visual Map ............................. 316
7.2 Limitations .......................................................................................... 317
7.3 Future Work ........................................................................................ 319
7.3.1 Exterior Glare Research ..................................................................... 319
7.3.2 Relative Glare Factor Focused Study ................................................. 320
7.3.3 Perceptible Glare Focused Study ....................................................... 320
xiii
7.3.4 Validation Study on AGF and RGF Method...................................... 321
References………………………………………………………………………...322
Appendix A. USC Institutional Review Boards Approvals ................................... 332
Appendix B. Participant Recruitment Email .......................................................... 334
Appendix C. Information Sheet for Non-Medical Research .................................. 335
Appendix D. Human Subjects Certificate: Human Subjects Education Program
(CITI)……………….. ............................................................................................ 339
Appendix E. UTA/HOBO connector ...................................................................... 344

xiv
LIST OF FIGURES
Figure 1-1. A breakdown of common office building costs on an annual basis. ......... 4
Figure 1-2. Components of heat transfer and solar heat gain through a window.
Source: Carmody et al. 2004, p.23 and 26. .................................................................. 7
Figure 1-3. Example of daylight glare through a building envelope. .......................... 9
Figure 1-4. Examples of reflected sunlight pollution in downtown Los Angeles: a) a
building with a highly reflective glass façade; b) reflected sunlight on a neighboring
building; and c) a shadow created by reflected sunlight on the street. ...................... 13
Figure 1-5. Reflected sunlight from 20 Fenchurch Street in London. ....................... 15
Figure 1-6. Walt Disney Concert Hall and reflected light onto Promenade
condominium. ............................................................................................................ 17
Figure 1-7. Glare categories. ...................................................................................... 20
Figure 1-8. Three categories of glare based on the results of glare: a) discomfort
glare, b) disability glare, and c) veiling reflection. .................................................... 22
Figure 1-9. Two initial conceptual diagrams of AGF and RGF definitions in relation
to absolute luminance and glare ratio......................................................................... 24
Figure 1-10. Final conceptual diagram of AGF and RGF definitions in relation to
absolute luminance and glare ratio. ............................................................................ 30
Figure 2-1. The glass facade: Requirements and physical standards. ........................ 32
Figure 2-2. Specular, spread, and diffuse reflections. ................................................ 38
Figure 2-3. Sunlight reflections on stainless steel panels of Walt Disney Concert Hall
(left) and glass facades of Bonaventure Tower (right) in downtown Los Angeles. ... 39
Figure 2-4. Example of luminance histogram from an LDR image. ......................... 44
Figure 2-5. Brightness differences. ............................................................................ 45
Figure 2-6. Simultaneous contrast effect. .................................................................. 46
Figure 2-7. Demonstration of the human eye’s field of view. .................................... 47
Figure 2-8. The resulting effect of visual adaptation. ................................................ 48
Figure 2-9. Example of visual adaptation. ................................................................. 49
Figure 2-10. Per-pixel data extraction from physically based renderings and HDR
xv
photography . ............................................................................................................... 70
Figure 2-11. Glare analysis of HDR photographs from a workstation in the NY Times
mockup building. ....................................................................................................... 71
Figure 2-12. IES’ potential glare source detection function. ..................................... 73
Figure 2-13. High dynamic range of luminance values. ............................................ 77
Figure 2-14. Conventional low dynamic range images and a high dynamic range
image of hanging art pieces below skylights. ............................................................ 79
Figure 2-15. Camera response function recovery from five exposures. .................... 82
Figure 2-16. Captured sky images: cloudy, partly cloudy, and mostly clear sky. ...... 85
Figure 3-1. Watt Hall at University of Southern California. ...................................... 88
Figure 3-2. Watt Hall’s third floor plan and photograph locations. ........................... 92
Figure 3-3. Ten fields of view captured by HDR imaging technique. ....................... 93
Figure 3-4. Ten field of views simulated from Ecotect-Radiance simulation. ........... 95
Figure 3-5. Captured HDR image from Photosphere and simulated HDR scene from
Radiance at view 01. .................................................................................................. 98
Figure 3-6. Glare score comparison of the scene in Figure 3-5. .............................. 101
Figure 3-7. Captured HDR image from Photosphere and simulated HDR scene from
Radiance at view 06. ................................................................................................ 102
Figure 3-8. Glare score comparison of the scene in Figure 3-7. .............................. 103
Figure 3-9. Captured HDR image from Photosphere and simulated HDR scene from
Radiance at view 09. ................................................................................................ 104
Figure 3-10. Glare score comparison of the scene in Figure 3-9. ............................ 105
Figure 3-11. Comparison of captured HDR image and simulated PIC scene. ......... 107
Figure 3-12. DGP score results from view 01 to view 04. ....................................... 109
Figure 3-13. DGP glare score results from view 05 to view 07. .............................. 110
Figure 3-14. DGP glare score results from view 08 to view 10. .............................. 111
Figure 3-15. Four different glare scenes in perspective and angular fisheye views. 114
Figure 3-16. DGI score comparisons between perspective and fisheye views. ....... 117
Figure 3-17. DGP score comparisons between perspective and fisheye views with
and without the measured vertical illuminance values. ........................................... 119
xvi
Figure 3-18. Human eye’s field of view. .................................................................. 125
Figure 3-19. HDR image, glare source detection, and log luminance histogram .... 128
Figure 3-20. Watt Hall third floor glare analysis results. ......................................... 130
Figure 3-21. The three zones: no glare, relative glare, and absolute glare. ............. 132
Figure 4-1. The camera and fisheye lens mounted on a tripod, the illuminance
sensors mounted on a monitor, the illuminance sensors mounted on a camera, and a
luminance meter. ...................................................................................................... 139
Figure 4-2. Interior glare study research setting: 1) Camera; 2, 3, 4, 5) Li-Cor sensors
for vertical illuminance; 4, 5) Li-Cor sensors for horizontal illuminance; 2, 3, 4, 5)
HOBO sensors; 6) Tripod; 7) Cooke luminance meter. ........................................... 141
Figure 4-3. Three tasks under fully open blind setting. ........................................... 142
Figure 4-4. Three tasks under the roller blind setting. ............................................. 143
Figure 4-5. Three tasks under venetian blind setting. .............................................. 144
Figure 4-6. Diagram of interior study procedure for each subject. .......................... 146
Figure 4-7. Visual map of FOV looking straight ahead for each of the three tasks. 149
Figure 4-8. Interior glare questionnaire example. .................................................... 150
Figure 4-9. Reflected sunlight in FOV . .................................................................... 152
Figure 4-10. Diagram of exterior study procedure for each subject. ....................... 154
Figure 4-11. Visual maps for exterior glare study. ................................................... 155
Figure 4-12. Exterior glare questionnaire example. ................................................. 156
Figure 4-13. A screenshot of the MATLAB code. ................................................... 157
Figure 4-14. Logarithmic luminance histogram created from MATLAB code. ...... 160
Figure 4-15. Images created from MATLAB code: Full fisheye FOV vs. human eye
FOV . ......................................................................................................................... 161
Figure 4-16. Interior glare example: Visible sun in FOV . ........................................ 162
Figure 4-17. Exterior glare example: Glare sources with extreme brightness. ........ 163
Figure 5-1. No-task condition scenes processed in Evalglare, compared to
participants’ subjective evaluation. .......................................................................... 166
Figure 5-2. Typing task scenes processed in Evalglare. ........................................... 167
Figure 5-3. Writing task scenes processed in Evalglare. .......................................... 168
xvii
Figure 5-4. All glare scenes: Full fisheye FOV (top) and the human eye FOV
(bottom). ................................................................................................................... 170
Figure 5-5. Imperceptible glare scenes only: Full fisheye FOV (top) and human eye
FOV (bottom). .......................................................................................................... 173
Figure 5-6. Perceptible glare scenes only: Full fisheye FOV (top) and the human
eye’s FOV (bottom). ................................................................................................. 176
Figure 5-7. Disturbing and intolerable glare scenes only: Full fisheye FOV (top) and
the human eye’s FOV (bottom). ............................................................................... 178
Figure 5-8. Interval plot of CGI score ranges with full fisheye FOV (top) and human
eye FOV (bottom). ................................................................................................... 183
Figure 5-9. Interval plot of UGR score ranges with full fisheye FOV (top) and human
eye FOV (bottom). ................................................................................................... 185
Figure 5-10. Interval plot of DGI score ranges with full fisheye FOV (top) and
human eye FOV (bottom). ....................................................................................... 187
Figure 5-11. Interval plot of VCP score ranges with full fisheye FOV (top) and
human eye FOV (bottom). ....................................................................................... 189
Figure 5-12. Interval plot of DGP score ranges with full fisheye FOV (top) and
human eye FOV (bottom). ....................................................................................... 191
Figure 6-1. Nine different HDR images captured from a single test. ...................... 194
Figure 6-2. Nine different HDR images processed in MATLAB, wherein red fields
indicate potential glare sources as calculated in MATLAB. .................................... 196
Figure 6-3. Percentages of glare sensation experienced by subjects. ...................... 198
Figure 6-4. Percentages of subjects’ visual satisfaction levels. ............................... 199
Figure 6-5. Percentages of subjects’ visual comfort levels. ..................................... 200
Figure 6-6. Percentages of subjects’ visual comfort levels when the daily office
condition was assumed. ............................................................................................ 201
Figure 6-7. One-way ANOV A: Glare level vs. glare sensation category. ................ 203
Figure 6-8. One-way ANOV A: Visual satisfaction vs. glare sensation .................... 205
Figure 6-9. One-way ANOV A: Visual comfort level vs. glare category .................. 206
Figure 6-10. One-way ANOVA: Visual comfort level with daily workplace
xviii
assumption vs. glare category .................................................................................. 207
Figure 6-11. Total frequency of glare source detection outside human eye FOV for
the no-task condition. ............................................................................................... 210
Figure 6-12. Total frequency of glare source detection outside human eye FOV in the
typing task condition. ............................................................................................... 211
Figure 6-13. Total frequency of glare source detection outside the human eye FOV
during the writing task. ............................................................................................ 212
Figure 6-14. Glare source detection example under the no-task condition. ............. 215
Figure 6-15. Glare source detection example under the typing task condition. ....... 217
Figure 6-16. Glare source detection example under the writing task condition. ..... 219
Figure 6-17. One-way ANOV A: Glare source minimum luminance vs. glare level 222
Figure 6-18. One-way ANOVA: Glare source minimum luminance vs. glare
sensation category .................................................................................................... 223
Figure 6-19. Glare source minimum luminance vs. glare category for all task
conditions. ................................................................................................................ 225
Figure 6-20. Glare source minimum luminance vs. glare category under the no-task
condition. .................................................................................................................. 227
Figure 6-21. Glare source minimum luminance vs. glare category under the typing
task condition. .......................................................................................................... 228
Figure 6-22. Glare source min luminance vs. glare category for the writing task
condition. .................................................................................................................. 230
Figure 6-23. Scatter plot of glare ratio and glare source minimum luminance for no-
task scenes. ............................................................................................................... 233
Figure 6-24. Scatter plot of glare ratio and glare source minimum luminance for
typing task. ............................................................................................................... 234
Figure 6-25. Scatter plot of glare ratio and glare source minimum luminance for
writing task. .............................................................................................................. 236
Figure 6-26. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the typing task condition. ............................. 238
Figure 6-27. Comparison of visual map and HDR image glare detections using
xix
different luminance thresholds under the typing task condition. ............................. 240
Figure 6-28. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the no-task condition. ................................... 242
Figure 6-29. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the writing task. ............................................ 245
Figure 6-30. Vertical illuminance vs. glare category under the no-task condition. . 248
Figure 6-31. Vertical illuminance vs. glare category within the typing task condition.
.................................................................................................................................. 249
Figure 6-32. Vertical illuminance vs. glare category under the writing task condition.
.................................................................................................................................. 250
Figure 6-33. Scatter plot of glare ratio and vertical illuminance at human eyes for no-
task condition. .......................................................................................................... 251
Figure 6-34. Enlarged scatter plot from Figure 6-33 shows glare ratio and vertical
illuminance at human eyes for no-task condition. ................................................... 252
Figure 6-35. Scatter plot of glare ratio and vertical illuminance at human eyes for the
typing task condition. ............................................................................................... 253
Figure 6-36. Enlarged scatter plot from Figure 6-35 shows glare ratio and vertical
illuminance at human eyes for the typing task condition. ........................................ 254
Figure 6-37. Scatter plot of glare ratio and vertical illuminance at human eyes for the
writing task condition. .............................................................................................. 255
Figure 6-38. Glare ratio of task luminance to glare source minimum luminance vs.
glare category, within the typing task condition. ..................................................... 259
Figure 6-39. Scatter plot of glare ratio and glare source minimum luminance for the
typing task condition. ............................................................................................... 260
Figure 6-40. Imperceptible (white), perceptible (blue), disturbing (orange), and
intolerable (red) glare zones on the scatter plot of glare ratio and glare source
luminance for the typing task condition. .................................................................. 261
Figure 6-41. AGF and RGF dominant zones for typing task. .................................. 263
Figure 6-42. Scatter plot of glare ratio and vertical illuminance for typing task. .... 265
Figure 6-43. Imperceptible (white), perceptible (blue), disturbing (orange), and
xx
intolerable (red) glare zones on the scatter plot of glare ratio and vertical illuminance
for the typing task condition. ................................................................................... 266
Figure 6-44. AGF and RGF dominant zones for typing task. .................................. 267
Figure 6-45. Six different scenes captured from a single test. ................................. 272
Figure 6-46. Six different HDR images processed in MATLAB. ............................ 273
Figure 6-47. Percentages of exterior glare sensation by subjects. ........................... 275
Figure 6-48. Percentages of subjects’ visual satisfaction levels. ............................. 276
Figure 6-49. Percentages of subjects’ visual comfort levels. ................................... 277
Figure 6-50. Percentages of subjects’ visual comfort levels when they assumed the
area was a daily resting place. .................................................................................. 278
Figure 6-51. One-way ANOVA comparing glare level and glare sensation category
.................................................................................................................................. 280
Figure 6-52. One-way ANOV A: Visual satisfaction vs. glare sensation .................. 282
Figure 6-53. One-way ANOV A: Visual comfort level vs. glare category ................ 283
Figure 6-54. One-way ANOVA: Visual comfort level with daily rest place
assumption vs. glare category .................................................................................. 284
Figure 6-55. Total frequency of field of view matching glare source detection during
the no-task condition. ............................................................................................... 286
Figure 6-56. Total frequency of glare source detection outside human eye FOV under
the iPad reading task condition. ............................................................................... 287
Figure 6-57. Total frequency of glare source detection outside human eye FOV under
the paper reading task condition. ............................................................................. 288
Figure 6-58. Glare source detection example in the no-task condition.................... 290
Figure 6-59. Glare source detection example under the iPad reading task condition.
.................................................................................................................................. 292
Figure 6-60. Glare source detection example under the paper reading task condition.
.................................................................................................................................. 293
Figure 6-61. One-way ANOVA: Glare source minimum luminance vs. glare level.
.................................................................................................................................. 295
Figure 6-62. One-way ANOVA: Glare source minimum luminance vs. glare
xxi
sensation category. ................................................................................................... 296
Figure 6-63. Glare source minimum luminance vs. glare category under all task
conditions. ................................................................................................................ 298
Figure 6-64. Glare source minimum luminance vs. glare category under the no-task
condition. .................................................................................................................. 299
Figure 6-65. Glare source minimum luminance vs. glare category under the iPad
reading task condition. ............................................................................................. 300
Figure 6-66. Glare source minimum luminance vs. glare category under the paper
reading task condition. ............................................................................................. 301
Figure 7-1. AGF and RGF dominant zones. ............................................................ 307

xxii
LIST OF TABLES
Table 2-1. U-values and light transmittance values for various glazing types. .......... 35
Table 2-2. Transmittance values of materials. ............................................................ 35
Table 2-3. Reflectance of materials. ........................................................................... 37
Table 2-4. Luminance and illuminance with conversion factors. .............................. 40
Table 2-5. Ambient luminance levels for some common lighting environments. ..... 42
Table 2-6. Luminance values above perceptible range. ............................................. 43
Table 2-7. Glare criterion for daylight glare probability. ........................................... 52
Table 2-8. Glare criterion for glare index and daylight glare index. .......................... 54
Table 2-9. Unified glare rating index. ........................................................................ 57
Table 2-10. Degree of perceived glare in different glare indices. .............................. 68
Table 3-1. Comparisons of calculated and measured vertical illuminance values. .. 121
Table 6-1. ANOV A test results of glare ratio and glare category. ............................ 257
Table 6-2. Absolute luminance and illuminance thresholds. .................................... 270

xxiii
ACKNOWLEDGEMENTS

I would like to acknowledge the contributions of my committee members and all the
people who supported me during the long journey of my PhD study at the University
of Southern California. My exceeding gratitude goes to the chair of my dissertation
committee, Prof. Marc Schiler, for his endless trust and support for my success.
There is no doubt that I was fortunate to have him as an advisor for both my Master’s
and PhD studies. Since I first met him in 2005, his profound knowledge in building
science and dedicated teaching attitude has inspired me to follow my scientific
interests.

I would like to acknowledge the contributions of the rest of my committee members,
as well. As a passionate educator, Professor Karen Kensek has never hesitated to
share her research ideas and provide me with a different point of view. Professor
Bosco Tjan, in the psychology department, has also contributed a great deal to my
dissertation research, especially on the special considerations for studying human
subjects. His profound experiences with human subject research gave me a clear
direction for my research.

I am also grateful to Professor Doug Noble, who has led the PhD program in USC
Architecture for many years. He never rested from encouraging every single PhD
student to pursue academic excellence. It would not have been possible for me to
begin this study without him. Special thanks also go to all the volunteers who
xxiv
participated in the human subject study. Their sincere attitudes toward my research
topic became my motivation to complete my research.

I would also like to thank Professor Joon-ho Choi for his guidance in statistical
analysis, and Joseph Pingree and Shitian Shen for their help developing the
MATLAB code to analyze HDR images. I would also like to extend my gratitude to
the USC School of Architecture and Southern California Edison for their financial
support of my tuition and research.

I feel privileged for the opportunity to study and work with the unique individuals in
USC’s School of Architecture PhD program. Although we all have different
specialties, we never stopped encouraging each other during our journeys. PhD
Powwow is and will always be an unforgettable memory for me. Mic, Jeff, Yara, Ed,
Lizzie, Andrea, Eve, Myoboon, and Simon: I hope every single one of us will
successfully complete this long journey.

I also would like to thank all of my colleagues at Horton Lees Brogden Lighting
Design. I have truly enjoyed the past seven years there, among the best architectural
lighting designers in the world. Especially, I would like to thank Teal Brogden and
Tina Aghassian, for their consistent trust and generous gestures that allowed me to
balance my family, work, and school. I would also like to thank my other coworkers
for their gracious encouragement.

xxv
I am also deeply grateful to my parents and parents-in-law for their infinite support,
faith, and love from the opposite side of the earth.

Last and certainly not least, my pleasure in completing my PhD degree must be
shared with my lovely wife and daughters, who have always been on my side no
matter what. You provided the only place where I could rest, and were the primary
reasons I never gave up.
xxvi
DEDICATION
To my loves,
Kyung Mi, Gianna, and Julia

1

Chapter 1 Introduction
Daylighting and Glare

1.1 Benefits of Daylighting in Buildings
Daylighting has recently emerged as a promising design strategy to save the energy
consumption of electrical lighting (daylight harvesting) and increase human comfort
and productivity for the occupants of buildings. Daylighting is considered to be a
new concept, which resulted from serious concerns about energy and sustainability.
However, daylight has been used throughout history to provide interior lighting for
both workplaces and religious structures. It had numerous advantages over the
technology of the past (fire, candles, oil lamps, gas lamps, etc.), although obviously
it was not used at night. Still, daylight was cheap, abundant, and could also be used
for heating. Architectural forms often tried to harvest this resource passively. The
reliance on daylighting and natural ventilation, especially in commercial buildings,
faded with the development of technologies such as electrical lamps, heating systems,
air conditioning, and power electrical grid systems (Baker and Steemers 2002). This
caused the creation of common typologies for commercial buildings that did not
require or even necessarily desire to use daylighting as a primary approach.

After serious energy concerns emerged from the 1970s oil crisis, daylighting has
again become a popular design feature among those who wish to reduce energy
consumption and achieve sustainable architecture. Design features to introduce
2
natural light into indoor spaces were considered for commercial buildings. The
increase in windows and openings on building facades have improved the
daylighting performance, but have also created serious human discomfort issues
from glare and solar heat gain, both inside and outside of buildings. Overhangs, light
shelves, louvers, fins, and filtering systems have been utilized to control these issues.

Building energy savings is one way to quantify the benefits of daylighting.
According to the Energy Information Administration’s 2003 commercial buildings
energy consumption survey, the largest single consumer of electricity in commercial
buildings is lighting, at 38% of total consumption (Energy Information
Administration 2009). Since electrical lighting takes the biggest portion of building
electricity consumption, daylight harvesting is the obvious solution to increase
energy savings in buildings. Daylighting also has the potential to decrease peak
demand and reduce the cooling loads caused by heat released into the space by
electrical lighting fixtures (Boubekri 2008). With more efficient electrical lighting
fixtures and daylight harvesting control systems, electric energy usage can be
expected to decrease while sufficient light levels are maintained for building
occupants. However, effective daylight harvesting can be only achieved with
properly designed daylighting systems.

Daylighting is also known for its positive effects on human health and productivity.
A number of researchers have studied the benefits of daylighting for human health
3
and well-being (Stone 1999; Rangi and Osterhaus 1999; Küller and Lindsten 1992).
The biggest health advantage of daylighting is the human body’s production of
vitamin D by way of photosynthesis. Vitamin D is essential for proper bone
development and growth. It enables bones and teeth to harden by increasing the
deposition of calcium and assists in the migration of calcium across body cell
membranes. Of the vitamin D produced in the human body, 80 to 100% occurs
through photolysis of solar ultraviolet type B radiation (UV-B). This is why vitamin
D is called the “sunshine vitamin.” Some studies have shown that people in low
latitude countries have significantly less vitamin D in their blood in winter than in
summer. Other studies have shown that exposure to daylight alleviates bone disease,
heart disease, cancers, diabetes, and Seasonal Affective Disorder (SAD), among
other health concerns. Because a much higher proportion of UV energy passes
through glass from daylight radiation than from artificial light sources such as
incandescent, halogen, or fluorescent light sources (Boubekri 2008), daylighting
provides a better opportunity to provide vitamin D and other daylight benefits than
artificial lighting.

According to the General Service Administration, “Since people are the most
important resource and the greatest expense of any organization, the long-term cost
benefits of a properly designed, user-friendly work environment should be factored
into any initial cost consideration” (Boubekri 2008, p.103). Figure 1-1, below, shows
common office building cost data based on Department of Labor (2000) data,
4
Building Owner and Managers Association (2000) data, and International Facility
Management Association (1997) data. As shown in the chart, the cost of salaries is
the biggest portion of the office building expenses. The annual cost of electricity,
including lighting, is around 1% of total cost. This breakdown of costs demonstrates
how important most companies consider their human resources to be.

Figure 1-1. A breakdown of common office building costs on an annual basis.
Source: Steffy 2008, p.2.

Productivity in the workplace can be increased with an increase in daylight, since
daylight regulates the levels of melatonin and serotonin in the human body.
Melatonin levels affect the energy and activity levels in human bodies. At low light
levels, melatonin levels increase, and drowsiness occurs. Daylight suppresses the
production of melatonin and fosters an alert state of mind by increasing the level of
5
serotonin (Stone 1999; Boubekri 2008). A number of studies have shown compelling
correlations between daylight and student performance in classrooms (Küller and
Lindsten 1992)—particularly that daylight can help to improve student learning
performance (Boubekri 2008). The studies showed that students exhibited decreased
attention and greater hyperactivity, fatigue, and irritability under cool-white standard
fluorescent lights than did students under daylight. Students under daylight showed
lower stress, decreased absenteeism, and improved overall achievements in
classrooms than students under standard fluorescent lights.

Despite the substantial benefits of daylighting over artificial lighting, there seems to
be a reliance on artificial lighting among building designers. This reliance could be
overcome with more education on the effective use of daylighting, as there is more to
daylighting strategies than the simple provision of more windows. For example,
every building and site presents unique conditions of orientation, shading, façade
geometry, and material. The complexity of the daylighting design and analysis
process makes it difficult for designers to accurately predict daylighting performance.

Furthermore, daylighting strategies must help account for the possibility of
daylighting glare and thermal discomfort. The sun’s location and sky conditions
change constantly throughout the day and year, which creates a challenge for
predicting how to best address these issues. Daylight glare analysis presents research
groups and professional designers a more complex web of issues to consider than do
6
the glare issues presented by artificial light sources, and this significantly
complicates the process of evaluation and prediction.

1.2 Disadvantages of Daylighting Inside Buildings
Daylighting has its downsides, as well. It can cause problems such as visual and
thermal discomfort inside building facades. In office buildings, the fully glazed
transparent building facade increases daylight levels around the perimeter of the
indoor space. However, this space is often also affected by increased glare, excessive
solar gains and heat losses, and a lack of privacy (Baker and Steemers 2002).

Daylighting was used as a heating source for buildings historically when there was
no energy-driven heating equipment, and there exist a number of architectural
designs that include passive solar heating systems. These passive solar heating
strategies can be seen more often in northern countries, where the heating is useful
during parts of the year due to the cold weather. As wider and more transparent glass
became more common in the building envelope, it became easier to allow for the
capture of heat and light energy from the sun. This transparent material helps to
maximize the performance of daylighting and provide exterior views to building
occupants, but it can also cause overheating issues. Figure 1-2 shows how heat flows
from the sun to occupants inside a building. Energy always moves from a hot object
to a cold object through three different mechanisms; radiation, convection, and
conduction. In winter, strong sunlight through windows works as a good heating
7
source, although windows can also contribute to increased heat loss, which becomes
a matter of concern. In summer, on the other hand, the building envelope becomes
hot, and heat moves from the envelope to occupants. Usually, heat gain happens both
internally and externally due to lighting fixtures, equipment, and the sun. Solar heat
gain analysis is critically important to daylighting design, as direct sunlight
penetration into indoor space can increase the energy used by the air conditioning
system, which can end up increasing energy consumption and negating the benefits
of daylighting. If a consistent, comfortable temperature is not maintained inside the
building, then the discomfort caused by thermal radiation to the building occupants
can circumvent the health benefits of daylighting.

Heat Transfer Solar Heat Gain
Figure 1-2. Components of heat transfer and solar heat gain through a window.
Source: Carmody et al. 2004, p.23 and 26.

Not only can excessive solar heat gain can cause thermal discomfort, but also
8
excessive natural illumination through building envelopes can cause visual
discomfort to occupants. Visual discomfort is a common subjective phenomenon
both outside and inside buildings, primarily caused by glare from natural and
artificial light sources. For many decades, discomfort glare sources were categorized
into one of two types: artificial light sources and the sun. Since the invention of
electric light sources, poorly-designed light fixtures and poor lighting designs have
caused discomfort glare. This created the need to predict the level and type of
potential discomfort glare in order to avoid it, which was the beginning of discomfort
glare analysis.

Most of the glare issues were caused by the severe contrast between the dark indoor
space and bright artificial light sources. When this type of discomfort glare was
analyzed, daylight was not considered or included in the analysis procedures, since
most ambient indoor lighting was provided by electric light sources. Discomfort
glare analysis methods for static artificial light sources are quite straightforward.

Unfortunately, analysis methods for artificial discomfort glare do not work for
daylighting. Glare from strong sunlight changes as the sun moves and experiences
altered intensity during the day. The biggest differences between daylight glare and
electric light glare are the size and the brightness of the glare sources. Compared to
glare from electric light sources, daylight glare can be much larger, since the entire
area of the windows can become a glare source, and the glare can be much brighter,
9
especially when direct sunlight hits highly specular interior surfaces.

Hand calculations and computer-based analysis tools can be used to predict glare
from sunlight, but it is not yet possible to anticipate all instances of daylight glare.
Therefore, it is still challenging to effectively control daylight glare while also trying
to maximize daylighting performance.

Figure 1-3. Example of daylight glare through a building envelope.

Discomfort glare is often the result of excessive contrast between bright and dark
areas in the field of view. The human eye has difficulty adjusting to such extreme
differences in brightness. The experience of discomfort glare occurs when the eyes
are adapted to an overall low light level, and the iris aperture is wide open to capture
more light. If there is a point light source whose brightness is substantially greater
10
than the overall brightness within the human visual field of view, that light point
source will effectively burn a hole or cause sensor damage at the retina where it is
focused. There is visual discomfort before the source burns the hole. Thus, the eye
attempts to protect itself from this phenomenon by signaling visual discomfort and
reacting to the bright light.

As a result of this discomfort, people will make an adjustment by squinting and
turning away, in order to correct the environment. Being exposed to discomfort glare
for some period of time can reduce the work performance and productivity of
building occupants. In some extreme cases, discomfort glare might cause physical
and psychological damage (Tedjakusuma 2003). There are some interesting reports
showing different tolerance levels of discomfort glare from daylighting and electric
light sources. Because of its dynamic characteristics and full spectrum distribution,
daylight glare tends to be more easily tolerable than glare from artificial light sources.
Thus, the Daylight Glare Index differs from the glare index used for electric light
sources not only because daylight sources tend to be large and require different
quantification methods, but also because the level of tolerance for daylighting glare
is higher than that for electric lighting (Boubekri 2008).

To avoid this drawback of daylighting, extensive research has been conducted to
predict and prevent offensive sunlight from causing glare through transparent
building facades in order to circumvent potential issues prior to building completion.
11
Both quantitative and qualitative research approaches have been used to examine
how to accurately evaluate the issue and how to prevent and control glare for
building occupants. The following section introduces exterior glare issues and
references several examples to help define the issue.

1.3 Disadvantages of Reflected Sunlight Outside Buildings
While the drawbacks of offensive sunlight inside the building envelope have been
addressed by many researchers and professionals, sunlight reflected by specular
building facades has not received as much attention. It can be difficult to define
discomfort glare outside buildings, but most people experience it daily. Glare can
come from a strong light source, such as car headlights at night, or from a strong
glint from a shiny surface, like a highly polished car or a sparkling water feature.
Direct view of the bright sun in the clear sky can cause the same glare condition.

Such glare can obscure the vision after exposure. Exterior glare at night is
particularly an issue in road safety, as bright or badly shielded light sources around
roads may partially blind drivers or pedestrians unexpectedly and contribute to traffic
accidents.

People can easily experience exterior glare problems from highly specular materials,
such as the glass or polished stainless steel of exterior building facades. Buildings
that are designed for the purpose of decreasing solar heat gain are built with highly
12
reflective exterior skins with unique geometries that reflect a high percentage of
incident sunlight toward the outdoor environment; these are much more likely than
traditional buildings to make neighbors and public spaces visually and thermally
uncomfortable. Figure 1-4 shows an example of exterior glare. In Figure 1-4, the left
top picture is a view of a north-facing façade, while the left bottom picture is a view
of a west-facing facade in the late afternoon. Reflected sunlight such as that shown in
Figure 1-4 often comes within the pedestrian or driver’s field of view from
unexpected directions. In a city, there is also the possibility that one will experience
multiple or intensified glare effects, especially with unique building facade
geometries such as concave shapes. Concave façade surfaces with specular materials
can cause perceptible hotspots in the streetscape and even result in sunlight
concentrations that blind pedestrians. It can pose a serious threat to the drivers of
automobiles, as even a brief distraction can result in serious consequences. There is
certainly growing concern about the problems caused by reflected sunlight. Thus, it
is necessary to utilize a validated evaluation methodology when considering the
effects of building façade on the exterior environment.

13
a) a building with a highly reflective
glass façade
b) reflected sunlight on a neighboring
building
c) a shadow created by reflected sunlight
on the street
Figure 1-4. Examples of reflected sunlight pollution in downtown Los Angeles: a) a
building with a highly reflective glass façade; b) reflected sunlight on a neighboring
building; and c) a shadow created by reflected sunlight on the street.

Unlike discomfort glare occurring inside building envelopes, exterior glare caused by
specular building envelopes has not yet been well defined. Therefore, it is first
required to make a clear definition of exterior glare. Exterior glare is unwanted
sunlight reflection and concentration from specular building facades (“reflected
sunlight trespass”) that cause thermal and visual discomfort. This is actually similar
to the concept of light trespass, which is an unwanted additional artificial
14
illumination on neighboring sites. While light trespass affects visual comfort at night,
exterior glare affects thermal comfort, visual comfort, or both during the day. While
light trespass has been well addressed and incorporated in current building codes,
standards, and design guidelines, the exterior glare issue has not been covered due to
the lack of architectural research on the topic.

There are several studies and reports that address the potential danger of reflected
sunlight from highly reflective or specular building facades. A concave building
façade geometry can act like a light concentrator that focuses reflected sunlight into a
single spot around the building. The focused sunlight drastically increases radiant
temperature. Figure 1-5 shows a recent exterior glare case in London, UK. This
newly constructed skyscraper, 20 Fenchurch Street, has a concave specular façade
that faces south. It reflects and concentrates strong sunlight onto the ground and
neighboring buildings.

15
Figure 1-5. Reflected sunlight from 20 Fenchurch Street in London.
Sources: Duell and Webb, 2013 (Left); Neal, 2013 (Right).

As shown in this example, exterior glare occurs mostly from vertical or tilted
building facades with high specularity and reflectivity. Once direct sunlight
illumination hits specular building facades at an incident angle, recipients perceive
the beam at the reflected angle to be a high luminance glare source. This illumination
occurs only during the day, particularly under clear sky conditions. It often causes
thermal and visual discomfort simultaneously.

So far, most building energy and occupant comfort studies have focused on interior
spaces only, and many architectural professionals have ignored the consequences to
adjoining areas when building highly specular building facades. For example, the
Vdara Hotel in Las Vegas allows for sunlight reflection from the south facades of the
building, which has become a serious concern to the health and comfort of hotel
customers. The building’s concave geometry concentrates the reflected sunlight into
16
its own swimming pool deck areas, and one hotel guest was almost burned from the
focused reflected sunlight from the south façade of the building. The hotel had
anticipated a strong sun reflection and placed a thin film over the window that
reduced the reflection by 70 percent. However, the treatment was not enough to
avoid the problem. Finally, the hotel decided to put larger and thicker umbrellas on
the pool deck to protect their guests from the strong sunlight reflection until they
could find a permanent solution (Mayerowitz 2010).

Another example is the Walt Disney Concert Hall (WDCH), which is located in
downtown Los Angeles. After its completion in 2004, residents in a neighboring
condominium started complaining about excessive thermal discomfort caused by
some of the specular facades of the WDCH. The WDCH is shown in Figure 1-6,
below. An extensive study examined how serious were the WDCH’s thermal and
visual discomfort glare issues. Ground temperature and mid-air radiant temperatures
were recorded by using data loggers and an infrared thermometer. The maximum
reading of a piece of black-colored foam core was in excess of 300° F (150° C) at the
curbside around the building. After a thorough evaluation study, a glare remediation
method was employed on the Founder’s Room and REDCAT marquee, which are
located near the Promenade Condominiums (Schiler 2004; 2005). In 2007, another
investigation was performed to re-examine these issues after the surface treatment. It
was found that thermal discomfort was not relieved, although the remediation had
relieved the visual glare issue quite a bit (Suk et al. 2007).

Figure 1-6
condomini
Source: Sc

For the pu
sunlight c
several un
a discomfo
terms of w
Most refle
building f
completion
visual and

Second, re
tools. Inte
simple to p
simulation
6. Walt Disne
ium.
chiler 2004,
urpose of g
can be cons
ique charact
fort glare sou
when and wh
ected sunligh
facades, and
n of a new
thermal disc
eflected sun
erior discom
predict, eithe
n software,
ey Concert H
p.2.
glare analysi
idered the s
teristics of re
urce. First, s
here it occur
ht occurs by
d people wh
building wi
comfort from
nlight is diffi
mfort glare c
er, but vario
and a sund
Hall and refle
is research,
same as gla
eflected sunl
sunlight refl
rs, whereas
y a new build
ho did not
ill be stuck
m the strong
ficult to pred
aused by th
ous methods
dial make it
ected light o
discomfort
are from the
light that dif
lection can m
the sun itsel
ding develop
have proble
experiencin
sunlight ref
dict in adva
he sun throu
and tools su
possible to
onto Promen
glare cause
e sun. How
fferentiate it
more often b
lf is a bit m
pment with
ems with g
ng unwanted
flection.
ance using c
ugh building
uch as a heli
o study the
ade
ed by reflec
wever, there
from the sun
be surprising
ore predicta
highly specu
lare before
d or unexpec
current softw
g facades is
odon, compu
likelihood
17
cted
are
n as
g in
able.
ular
the
cted
ware
not
uter
that
18
discomfort glare will occur. Discomfort glare from reflected sunlight is much more
complicated, however, since it interacts with the surrounding buildings. To locate
potential glare sources, one must check neighboring buildings with high reflectance
facades, considering the sun and building facades at the same time. Neighboring
building locations and building geometry become crucial factors, and the analysis
procedure is very intense.

Third, reflected sunlight can be concentrated by a unique building geometry. For
example, a concave building façade can focus sunlight reflection into a small area
and cause excessive increases of light and temperature intensity that seriously affect
human health and comfort. It can also damage aspects of neighboring properties,
such as landscaping.

Last, reflected sunlight glare source locations are generally lower to recipients than
the sun. When strong sunlight hits specular building facades, the illuminated building
facades become a potential glare source with very high luminance levels. Depending
on the angle of the facade, the reflected light from a vertical building façade can
increase the chances or levels of discomfort glare, since the incoming light angle of
glare source is much lower than that of the sun.

Recently, there have been growing concerns about reflected sunlight problems,
especially in dense cities with lots of glazed, high-rise buildings. It is difficult to
19
prove whether reflected sunlight can seriously deteriorate human comfort in
neighboring buildings or public outdoor spaces. Yet, it is acknowledged that building
facades have huge impacts on an occupants’ visual and thermal comfort, so it is not
difficult to imagine that they would also cause huge impacts on human comfort for
those who are outside of a building. Therefore, it is important for designers to be
aware of the consequences of highly specular building façade materials in new
building construction.

With regards to the light trespass issue, building codes and standards require a new
building project to limit the amount of additional light that extends beyond the
property boundary. It may also be necessary that exterior glare be addressed in
building codes and standards much like light trespass, to avoid potential discomfort
issues.

1.4 Glare Categories
Although many people cannot accurately define the causes and types of glare, most
people experience it daily inside and outside of buildings. Considerable research has
been done to examine how to evaluate glare and how to mitigate the problem for
building users and the general public from a design standpoint. Confounding the
problem is the fact that glare is a subjective phenomenon, and people do not always
agree on what constitutes glare. For more accurate glare evaluation, it is critical to
make a clear definition of glare and determine the most appropriate way to analyze it.

Researche
brightness
field. Glar
created the
according
three poten

Figure 1-7

1.4.1 Gla
Three cate
and veiling
rs agree that
and some a
re can be des
e glare, acco
to the result
ntial describ
7. Glare categ
are Categori
egories can b
g reflection.
t glare occur
annoying, dis
scribed in on
ording to an i
ts of the gla
ers of glare.
gories.
ies Based on
be used to d
They are de
rs when the
stracting, or
ne of three m
individual’s
are. Figure 1

n the Result
describe the
efined as foll
eyes have a
blinding lig
main ways: a
perceived d
-7 describes
ts of Glare
results of g
lows.
adjusted to a
ht appears w
according to
egree of glar
s the categor
glare: discom
a certain gen
within the vis
the process
re intensity,
rization of th
mfort, disabi
20
neral
sual
that
and
hese

lity,
21
1. Discomfort glare can cause strain and pain to eyes, and it can also reduce
productivity and comfort level. This kind of problem is most often found
inside buildings or at night, when the background lighting is at a low level
and an individual is exposed to a bright lighting source. Direct afternoon
sunlight through windows inside a building and the headlights of oncoming
cars at night are common examples of discomfort glare.
2. Disability glare occurs when the eyes are looking at a very bright light source
that prevents people from completing a task, as the glare is strong enough
that people look away to avoid physical damage to the eyes. This can happen
either when looking at an extremely bright glare source or when exposed to
the extreme contrast between the glare source and background.
3. A veiling reflection is a reflected light that veils or masks some of the
information to be seen. This type of glare often happens when people read a
specular magazine page or a computer monitor and some part of the page is
veiled or masked by the unwanted reflection of light that comes to the eyes.
This type of glare can sometimes be avoided by changing the angle of
incidence or light source placement.

22
a) discomfort glare b) disability glare c) veiling reflection
Figure 1-8. Three categories of glare based on the results of glare: a) discomfort
glare, b) disability glare, and c) veiling reflection.
Source: b) Suk 2007, p.144; c) Suk and Schiler 2007, p.111

1.4.2 Glare Categories Based on the Perceived Degree of Glare
Another way to define glare is by the perceived degree of glare. Since the perceived
degree provides many different levels describing the discomfort glare problem and
helps to evaluate how serious a specific glare issue is for occupants, it has been
widely used in many existing glare analysis metrics. Perceived degree of glare is
generally described into the following four categories: imperceptible, perceptible,
disturbing, and intolerable glare. However, these can be subdivided into up to eight
categories if necessary.

1.4.3 Glare Factors Based on Process
For more accurate glare evaluation, it is critical to define the process and cause of the
glare issue and find the most appropriate way to analyze it. The process of glare can
be described using two factors: absolute glare factor and relative glare factor (Schiler
2009).
23

Absolute glare factor (AGF) is a process that occurs as the result of an excessively
bright glare source, such as the sun or extremely high luminance light sources. When
a glare scene is associated with AGF, the cause of glare is not the level of contrast
between the glare source and its surroundings, but the exceeding brightness of the
glare source.

Relative glare factor (RGF) is a process that occurs as the result of a glare source to
which the human eye could adapt, given proper background conditions. The glare is
created by the contrast between the brightness of the glare source and its
surroundings. This type of glare can be measured and analyzed by calculating the
contrast between glare source and background.

Either AGF alone or RGF alone could be dominant to cause either kind of result
based glare, but most glare cases are caused by a combination of these two factors.
Figure 1-9 shows two initial conceptual diagrams of AGF and RFG definitions in
relation to absolute luminance values and contrast ratios.

24

Figure 1-9. Two initial conceptual diagrams of AGF and RGF definitions in relation
to absolute luminance and glare ratio.

Further study has shown that these two diagrams do not accurately define AGF and
RGF, but they provide a valuable indication of how the definitions of AGF and RGF
have been refined and redefined. The final development of AGF and RGF definitions
is presented in section 1.6.

Glare ratio
Luminance

Relative glare factor

Absolute glare factor
Glare ratio
Luminance
Relative
glare
factor
Absolute glare factor
25
1.5 Problem Statement
Many analytical methods have been developed to evaluate and quantify discomfort
glare problems based on both subjective and objective analysis methods. These
methods have most often focused on discomfort glare caused by electrical lighting
sources inside an office space. After much study, architects and lighting professionals
have improved their understanding and developed certain techniques to control light
sources to avoid glare.

Daylighting strategies have been successfully incorporated into architectural design
practice due to increasing attention on building energy savings and sustainability.
Building codes have included mandatory requirements regarding daylighting
performance, and building standards provide even stricter requirements for
maximizing energy savings and occupant comfort by using daylight harvesting.
Recent revisions of ASHARE 90.1 and Title 24 require that daylight sensors and
sophisticated lighting control systems harvest more daylight. However, none of these
codes and standards outline visual comfort levels for daylit spaces.

There are several guidelines and recommendations for luminance values and contrast
ratios of the occupant’s field of view to avoid potential glare issues in a daylighting
context, but these have not yet been fully validated. For example, the U.S. Facilities
Standard for the Public Buildings Service (P100) stipulates guidelines for occupants’
visual comfort within buildings, and aspects of comfort covered under these
26
guidelines include luminance, illuminance, and contrast ratio thresholds for both
electrical lighting and daylighting.

With the recent growth of daylighting projects in the architecture field, professionals
have had many chances to analyze daylighting performance in various building types
and designs. However, they still have only a limited ability to anticipate actual light
levels (illuminance levels) or daylight factors contributed by natural light.

While architectural professionals have invested much effort to anticipate light levels
throughout the year, they have overlooked the visual comfort of daylit spaces and
have simply assumed that daylit spaces would provide higher visual quality to
occupants. However, there is growing concern that designers need to understand
exactly when and where occupants experience visual discomfort from natural light,
since such discomfort impedes successful daylight harvesting.

Under clear sky conditions, direct daylight through a window or skylight often
causes discomfort or disability glare problems for building occupants. One recent
study shows that occupants simply close blinds when they experience glare issues
and do not bother to open them again even when discomfort glare disappears
(Saxena et al. 2010). This occupant behavior impairs the architecture’s ability to
optimize the anticipated daylighting performance.

27
Daylight glare analysis is more complicated and difficult to perform than glare
analysis for electrical light sources, due to the following two factors: 1) the sun’s
location and the sky condition are changing constantly and 2) every building and site
have unique conditions, including sun angle, facade material, orientation, geometry,
and context. In addition, architectural professionals have repeatedly had trouble
dealing with daylight glare issues in architectural projects, even though computer-
based glare analysis software has been developed to assist the practice. AGI32,
Daysim for Ecotect, and DIVA for Rhino are currently available software that assist
the analysis of potential discomfort glare issues. AGI32 provides a unified glare
rating (UGR) for artificial interior lighting applications and a CIE glare rating (GR)
for the artificial lighting of outdoor and sports lighting applications. Both Daysim for
Ecotect and DIVA for Rhino, on the other hand, have a capability to analyze glare
from daylighting. It utilizes Evalglare code to calculate five glare indices: daylight
glare probability (DGP), daylight glare index (DGI), unified glare rating (UGR),
visual comfort probability (VCP), and CIE glare index (CGI).

New glare metrics such as DGI and DGP were developed by Chauvel (1982) and
Wienold (2005) to address daylight glare issues, but none of them have successfully
evaluated the actual levels of daylight glare in a building, especially when the glare
source has extremely high luminance levels. As luminance values of glare sources
increase, the impact of the absolute luminance value begins to dominate the impact
of the contrast ratio. This type of glare is caused mainly by AGF, but the existing
28
methods are not able to differentiate between AGF and RGF.

Furthermore, designers, lighting consultants, and daylighting experts often cannot
figure out what actually causes discomfort glare, even after having received glare
evaluation results from the existing methods. This is the primary reason that many
are still reluctant to utilize the methods in practice.

Many absolute luminance thresholds and specific contrast ratio values have also been
developed to evaluate daylight glare issues in various lighting codes and standards,
such as Swedish energy authority NUTEK, ISO Standard 9241-6, and ANSI/IESNA
RP-1 VDT lighting standard. These thresholds are not consistent with each other.
Furthermore, the developed luminance ranges and contrast ratios are not correlated
to determine different glare levels. Due to these problems and limitations of the
existing methods, different groups of lighting and daylighting experts usually follow
their own practices and rules based on their personal experiences, instead of using
the methods based on the findings of academic research.

Most previous daylight glare studies focused on indoor office spaces to understand
the degree to which occupants’ comfort and productivity could be deteriorated by
discomfort glare. Even though people have become more aware of exterior glare
issues, it is still difficult to evaluate the problem, since there is no validated
evaluation method. Therefore, a new daylight glare methodology that covers both
29
interior and exterior glare issues should be developed to overcome the problems and
limitations of current methods.

1.6 Research Objectives
The goal of this research is to develop a practical daylight glare analysis method for
both interior and exterior glare without compromising evaluation accuracy. To
achieve this, it is necessary to examine existing methods and define glare factors.
The three primary research objectives are thus as follows.
1. Examine existing discomfort glare metrics and tools in terms of consistency,
accuracy, and practicality to understand why they are barely used in
daylighting practice.
2. Define absolute glare factor (AGF) and relative glare factor (RGF) based on
human subject study results in daylit spaces inside and outside building
envelopes.
3. Develop a consistent and accurate daylight glare analysis method using AGF
and RGF that can overcome the issues of the existing glare indices.

Interior and exterior glare issues caused by direct sunlight through transparent
building envelopes or reflected sunlight from specular building envelopes were
studied. Even though interior and exterior glare issues are subject to different
analysis approaches, it is expected that a consistent logic can be used to assess the
effects of glare on human eyes both inside and outside a building. A new glare
30
analysis methodology was developed that can differentiate AGF from RGF (see
Figure 1-10, below). Furthermore, existing daylighting and glare analysis methods
were investigated in order to identify the variables and typical errors that result from
these methods. With the help of the high dynamic range imaging technique,
subjective tests were performed for both interior and exterior glare to help develop a
new methodology for anticipating glare and to confirm the methodology’s evaluation
accuracy. This methodology was intended to investigate the visual effects of daylight
glare, and did not include any assessment or prediction concerning the thermal
effects of daylighting.

Figure 1-10. Final conceptual diagram of AGF and RGF definitions in relation to
absolute luminance and glare ratio.

Glare ratio
Luminance
Relative glare factor
Absolute glare factor
31
Chapter 2 Literature Review
Daylight Glare and Building Envelopes

For a better understanding of daylight glare in relation to the building envelope, it is
necessary to understand the material properties of building envelopes. These include
transmittance, reflectance, and specularity. Lighting fundamentals are also crucial to
explain the causes of interior and exterior glare issues in relation to the building
envelope. This chapter explains the material characteristics of glass, lighting
fundamentals for glare analysis, and existing discomfort glare indices, glare
thresholds, glare analysis tools, and high dynamic range imaging techniques.

2.1 Material Characteristics of Building Envelopes
Building envelopes directly influence environmental conditions such as temperature,
humidity, light, sound, view, air flow, and air quality. These environmental
conditions directly affect human comfort (Lovell 2010). It is important to understand
what types of building envelopes can cause discomfort glare from natural light.

The use of transparent materials on building envelopes can cause interior and
exterior glare issues while allowing more natural light into the space. Köster (2004)
made simple glass property recommendations for different seasons with regards to
daylighting through glass facades (Figure 2-1). Natural light becomes a heating and
lighting source inside and outside the building. Properly designed glass envelopes
32
can provide good daylighting illumination, improve human thermal and visual
comfort, and lower building energy consumption.

Figure 2-1. The glass facade: Requirements and physical standards.
Source: Köster 2004, p.81.

On a sunny day, a building envelope receives around 100,000 lux illuminance. If the
efficiency of the daylighting strategy were 100%, that would be enough luminous
33
energy to illuminate 100 m
2
at 1,000 lux (Boubekri 2008). Of course 100%
conversion is unrealistic, but the numbers indicate that there is plenty of lighting
outside a building that can be used to supplement artificial lighting for interior spaces
if the building envelopes are well designed and their materials are carefully chosen.
This high amount of natural light can cause serious thermal discomfort and visual
glare if it is not successfully controlled, however.

2.1.1 Transparent Materials (Glass Facade)
Glass windows were first used by the Romans as architectural elements. They used
glass to capture and trap solar energy, including heat and light to warm their homes,
baths, and greenhouses (Baker and Steemers 2002). Due to its transparency and light
weight, glass has been widely used in commercial and residential buildings around
the world, especially since the cost of glass material has plummeted since ancient
times. Currently, glass is one of the most important architectural elements; it is a
critical element of any daylight harvesting strategy, and a main player in the daylight
glare issue.

There are a number of glass types and technologies currently available for
architectural use, including the following:
 Clear single glazing
 Insulated glazing, e.g. insulating glass units (IGUs)
 Tinted glazing
34
 Reflective coatings
 Low-E coatings
 Surface coatings or treatments: frit, acid-etched, and sandblasted glass
 Laminated glass
 Insulation-filled glazing
 Evacuated windows
 Smart windows such as photochromics, thermochromics, liquid crystal
device windows, suspended particle device (SPD) windows, electrochromic
windows, gasochromic windows, and motorized shading (Carmody et al.
2004)

When sunlight strikes a building surface, solar energy is either transmitted through
the surface, absorbed by it, or reflected away from it. Each glass type has different
thermal and visual properties, including solar heat gain coefficient, R-value, U-value,
transmittance, absorption, and reflectance (Table 2-1).
35
Table 2-1. U-values and light transmittance values for various glazing types.
Glazing specification U-value (W/m
2
K) Light transmittance (%)
Single 5.4 87
Double 2.8 75
Triple 1.9 65
Double low-e 1.8 74
Double low-e argon 1.5 74
Triple low-e argon 0.8 63
Source: Baker and Steemers 2002, p.107.

Transmittance is the percentage of incident light that passes through a surface. For
example, when sunlight reaches glass, which normally has a high transmittance value,
some light is absorbed by the glass, some is reflected back into the atmosphere, and
some is transmitted through the glass. Clear glass with high visual transmittance
values may cause interior glare problems from daylight. Transmittance values vary
across types of different materials, as defined in Table 2-2.

Table 2-2. Transmittance values of materials.
Transmission Material Transmittance (%)
Direct
Clear glass or plastic/
Transparent colored glass
80–94
Diffuse
Glass block 40–75
Marble 5–40
Plastic 30–65
Spread
Etched glass, toward source 82–88
Etched glass, away from source 63–78
Source: Egan 2002.

Higher light transmittance values admit more daylight into a space, while lower U-
36
values admit less solar heat. A glass type with high visual light transmittance usually
allows more heat transmittance. However, a number of new glass products have been
developed to increase visual transmittance values while decreasing thermal
transmittance values. Window-films can also be used to reduce the solar heat gain
coefficient and the visual transmittance that can cause thermal and visual discomfort
inside a building. These films are often used to retrofit the facade in a cost effective
way.

2.1.2 Reflective and Specular Materials
Solar reflectance is a measure of the ability of a surface material to reflect sunlight,
including the visible, infrared, and ultraviolet wavelengths. It is measured on a scale
of 0 to 1, with 1 describing 100% transmittance. If a building is constructed with
high solar reflectance materials, the building can reduce heat gain from sunlight and
save cooling energy during hot summer months. Of course, the reflected sunlight
might accumulate in the surrounding area and increase temperatures outside the
building. Table 2-3, below, defines the reflectance of various materials.

37
Table 2-3. Reflectance of materials.
Material Reflectance (%) Material Reflectance (%)
Aluminum, brushed 55–58 Clear or tinted Glass 5–10
Aluminum, etched 70–85 Reflective Glass 20–30
Aluminum, polished 60–70 Asphalt 5–10
Stainless steel 50–60 Concrete 40
Tin 67–72 Grass 5–30
Brick, red 10–20 Snow 60–75
Limestone 35–60 White Paint 70–90
White Plaster 90–92 Mahogany 6–12
Source: Egan 2002, p.58.

When reflectance occurs, the remaining light is either absorbed, transmitted, or both.
The total amount of light reflected from a surface includes all reflections: diffuse,
spread, and specular (Egan 2002). The reflectance of any given material affects the
perceived brightness and measured brightness on the surface of that material. Light
consists of photons, which normally travel in a straight line until they hit a surface.
When photons are reflected from a surface, the distribution of photons reflected in
certain directions given an angle of incidence gives a surface its appearance. Matte
surfaces such as plaster distribute light almost equally in all directions, whereas
glossy and shiny surfaces reflect light in a preferred direction (Reinhard and Ward
2010). Light striking a polished surface such as that of a mirror will reflect specularly,
which means that the angle of the incoming incident light will be equal to the angle
of the outgoing reflected light, as seen in Figure 2-2, below (Egan 2002). Some
discomfort glare problems can be avoided by simply changing the angle of the
reflective surface or changing the beam spread of the reflected light from a specular
38
reflection to a diffuse reflection.

Figure 2-2. Specular, spread, and diffuse reflections.
Source: Egan 2002, p.58.

Specular materials are usually polished, giving incident light and reflected light the
same angle. Common specular materials include glass windows and mirrors. Other
specular materials for building envelopes include the stainless steel panels on Frank
Gehry’s buildings, as shown in Figure 2-3. The Guggenheim Museum in Bilbao,
Spain, the Walt Disney Concert Hall in Los Angeles, USA, and the Experience
Music Project and Science Fiction Museum and Hall of Fame in Seattle, USA,
exhibit shiny, specular building facades with unique geometries. The titanium and
stainless steel panels on Gehry’s buildings create very unique and attractive
aesthetics, but also reflect a great amount of incident sunlight into their surrounding
39
spaces.

Figure 2-3. Sunlight reflections on stainless steel panels of Walt Disney Concert Hall
(left) and glass facades of Bonaventure Tower (right) in downtown Los Angeles.

Highly specular materials on a building facade are more likely to cause exterior glare
problems from sunlight. When the building has vertical walls with concave shapes,
the walls can focus sunlight reflection into one point, and the glare can seriously
damage human vision or increase the surrounding temperature beyond acceptable
levels.

2.2 Lighting Fundamentals for Glare Analysis
Luminance, illuminance, and contrast are the key elements of glare analysis, since
40
these three elements are important components in most existing glare indices. Visual
adaptation is also an important factor, as it contributes to our understanding of the
different effects of low and high ambient light environments, which are not
addressed in existing glare indices.

2.2.1 Luminance and Illuminance
Luminance is a parameter of light that describes the intensity of visible brightness of
a source or surface in the direction of the observer divided by the area of the source
or the surface. Luminance values change depending on surface material properties
and the angle of the observer’s view. The units of luminance are candelas per square
foot (cd/sf). As an international standard unit, candelas per square meter (cd/m
2
) is
more broadly used. The conversion factors between international and American
systems of units are shown in Table 2-4, below.

Table 2-4. Luminance and illuminance with conversion factors.
Non-international
system of units
International
system of units
Conversion factors
Luminance
candela
per square foot
(cd/sf)
candela
per square meter
(cd/m
2
)
1 cd/sf = 10.76 cd/m
2

Illuminance footcandle (fc) lux (lx) 1 fc = 10.76 lx

Illuminance is the density of luminous flux incident on a surface. Illuminance is
measured in lumens per square foot, which are commonly referred to as footcandles.
One footcandle is a quantity of light on 1 square foot of surface area 1 foot away
41
from a light source, of the intensity of 1 cd (one standard candle). The lux (lx) is an
international standard unit of illuminance. The eye does not see footcandles
(illuminance); when a surface is illuminated, the eye sees the light leaving the
surface, or its luminance (Egan 2002). When electrical lighting or daylighting
systems are designed for indoor or outdoor spaces, illuminance level calculations are
required to satisfy standards or codes concerning required light levels. Most existing
glare indices consider the (reflected) luminance values of illuminated surfaces in a
field of view. Daylight glare probability (DGP), for example, accounts for vertical
illuminance values when evaluating discomfort glare.

The luminance value is an important factor in glare analysis, since most glare
analysis considers the luminance of glare sources and backgrounds in a field of view.
The human eye has the capability to adjust to any portion of a huge range of
illumination on a daily basis. The dynamic range capacity of the human eye with
respect to luminance is from 0.0000001 cd/m
2
to 1,000,000 cd/m
2
, which is a range
of nearly 12 orders of magnitude. This is the reason people can see in strong sunlight
during daytime and also by starlight at night. Within a single scene, the adapted
visual range of the human eye becomes around 1 to 100,000, depending on the
brightness and contrast of the scene. The luminance levels of common lighting
conditions around us are listed in Table 2-5. This table shows that the sun’s
luminance value is extremely high. Thus, we may assume that the sun is the most
likely light source to cause discomfort glare during the daytime, regardless of glare
42
ratios.

Table 2-5. Ambient luminance levels for some common lighting environments.
Condition luminance (cd/m
2
)
Weakest visible light 10
-6

White paper in starlight 10
-4

Starlight 10
-3

White paper in moonlight 10
-2

Moonlight 10
-1

Comfortable reading 10
Indoor lighting 10
2

White paper under an overcast sky 10
3

White paper in sunlight 10
4

Sunlight 10
5

Source: Reinhard et al. 2010, p.6.

Egan claims that luminance values higher than 7,000 cd/m
2
(or 650 cd/sf) cause
human eyes to blink or squint (Egan 2002). This is a somewhat suspicious claim,
since white paper in sunlight has around 10,000 cd/m
2
, but one does not necessarily
blink or squint when looking at white paper.

There are luminance values that exist beyond the capacity of human eyes to see, as
well. Table 2-6 describes these.

43

Table 2-6. Luminance values above perceptible range.
Condition luminance (cd/m
2
)
Electronic flash 10
10

The sun surface by clear sky 10
9

Filament of a halogen lamp 10
7

Filament of an incandescent light bulb 10
6

Source: Baker and Steemers 2002, p.169.

Luminance histograms describe the perceived brightness distribution within an
image. Brightness is not the same as luminance—it is simply an attribute of human
visual perception, and cannot be used for quantitative references. Luminance, on the
other hand, is adjusted brightness to indicate what the human eye really sees.
Luminance takes into account that the human eye is more sensitive to green light
than red or blue light. In an image, each pixel is converted so that it represents a
luminosity based on a weighted average of the three colors (RGB) at that pixel. A
luminance histogram can be plotted using the following formula:

RGB Luminance value (cd/m
2
) = k*(0.2127 *R + 0.7151* G + 0.0722* B)

Where, k is a constant value that can be determined either for a camera or a scene
with a physical measurement (Inanici 2005 and 2006; Kumaragurubaran 2012).

A luminance histogram plotted from a single low dynamic range (LDR) digital
44
image is shown in Figure 2-4. The X-axis represents the luminance values of each
pixel, and the Y-axis represents the total number of pixels with the same luminance
value in an image. Reading the plotted luminance histogram allows one to
understand the luminance pattern of the image, such as the luminance of the entire
image and the luminance of glare sources. With conventional low dynamic range
images, luminance values are divided into 256 relative bins, which start from 0 for
black to 255 for white. With high dynamic range images, luminance value range can
exceed 0 to 100,000.

Figure 2-4. Example of luminance histogram from an LDR image.

In HDR imaging, luminance provides a natural boundary of visible wavelengths.
Any wavelength outside the visible range does not need to be recorded, stored, or
manipulated, since human vision is not capable of detecting those wavelengths.
(Drago et al. 2002; Reinhard et al. 2010).

45
2.2.2 Contrast (Brightness Difference)
Contrast is the brightness difference between the object being viewed and the
immediate surroundings. The magnitude of contrast in a space influences subjective
mood and affects productivity. Up to a certain point, higher contrast actually helps us
gain better visual performance. However, a glare problem potentially occurs when
there is too much contrast between a light source and the background. For
comfortable contrast in most situations, the brightness difference should be within
the allowable range depicted in Figure 2-5 (Egan 2002).

Figure 2-5. Brightness differences.
Source: Egan 2002, p.25.

The simultaneous contrast effect demonstrates an interesting concept of lightness
perception and lightness illusions, as shown in Figure 2-6 (Adelson 2000). Two small
squares located inside two different big squares have the same reflectance. However,
the small square inside the darker background looks brighter than the one inside the
46
lighter square. This example shows that it is necessary to check the relationship
between glare source brightness and background brightness in the field of view when
analyzing discomfort glare that is caused by contrast.

Figure 2-6. Simultaneous contrast effect.
Source: Adelson 2000, p.340.

The binocular visual field among humans extends vertically 130° and horizontally
more than 120° when both eyes are focused on a fixed object (Egan 1983). The range
of visual ability is not uniform across a field of view, however. The human visual
system can be quite insensitive to large luminance differences in the total field of
view, but it is very sensitive to small luminance differences in the foveal region, as
shown in Figure 2-7.

47

Figure 2-7. Demonstration of the human eye’s field of view.
Source: Inanici 2005, p.25.

2.2.3 Visual Adaptation
Visual adaptation is the ability to accommodate different brightness levels. The
human eye’s adaptation to brightness is influenced by the entire field of view. For
instance, when looking into a building from outdoors, the room seen through the
windows usually appears barely visible. This happens because the outdoor average
luminance level exceeds 1,000 cd/m
2
and the visual adaptation accordingly adjusts
the shadow limit between 10 and 100 cd/m
2
, which is the luminance level of a
typical room that is lighted only by daylight, as described in Figure 2-8 (Baker and
Steemers 2002).

48

Figure 2-8. The resulting effect of visual adaptation.
Source: Baker and Steemers 2002, p.170.

Visual adaptation does not instantaneously react to a rapid luminance change. Any
sudden and drastic change of luminance levels causes the visual adaptation system to
momentarily lose its normal functionality. This decreased functionality contributes to
the occurrence of discomfort glare. Following a momentary loss in sensitivity, the
visual system gradually adapts to the prevailing illumination and recovers its
sensitivity (Reinhard et al. 2010).

There are other examples of visual adaptation. Car headlights that look very bright at
night do not look bright at all during the day (Figure 2-9). The full moon is not very
bright during the day, but it looks much brighter at night. The actual luminance
values of car headlights and moon are the same, regardless of whether it is day or
night. However, the adaptation level of the human eye makes these items look
brighter or darker depending on the average luminance values in the field of view.
49

Figure 2-9. Example of visual adaptation.
Source: Reinhard et al. 2010, p.238.

2.3 Existing Discomfort Glare Indices
Many glare studies have been conducted inside a room with subjective research
methods. The surveys were performed to find out how many people would feel
uncomfortable with different levels of glare. The results of these tests are somewhat
subjectively dependent on each person’s eye sensitivity.

There are also several objective research methodologies to evaluate glare. Each of
them uses different variables and formulas, but the following four elements of light
are critical in all of these methods: the size of the glare source, the luminance of the
glare source, the position of the glare source, and the mean luminance of the
background.

There are at least nine significant existing glare analysis indices:
 Daylight glare probability
50
 Daylight glare index
 British glare index
 Unified glare rating
 Visual comfort probability
 CIE glare index
 Relative visual performance
 Video photometry

2.3.1 Daylight Glare Probability (DGP) and Simplified DGP (DGPs)
Most of the existing glare analysis methods focus on the contrast ratio between
background mean luminance and glare source luminance, or between task area
luminance and glare source luminance. Unlike most of the other indices, daylight
glare probability (DGP) (Wienold and Christoffersen 2005) weighs the vertical
illuminance values that human eyes receive. Since glare can be caused by very high
illuminances, the illuminance at the eye can be a direct proxy to judge whether there
is discomfort glare or not. Thus, DGP and simplified daylight glare probability
(DGPs) are the most appropriate metrics by which to analyze glare caused by
extremely bright glare sources. According to Jakubiec and Reinhart (2011), DGP is
the most robust metric with the least chance of getting inaccurate glare predictions
under various sky conditions.

Kleindienst and Andersen (2011) explained that DGP performs well for glare
51
situations based on high vertical illuminance, but is less accurate for luminance
contrast based glare. The following explains how DGP and DGPs evaluate
discomfort glare and the limitations of these methodologies.

In order to determine glare, the new DGP formula uses the vertical eye illumination
as a glare measure and combines it with the central term of existing glare indexes
(Wienold 2006). It also uses the part of the CIE-glare index that describes the
influence of the glare source. The DGP formula is as follows:

16 . 0 1 log 10 18 . 9 10 87 . 5
2 87 . 1
,
2
, 2 5











      

 
i
i v
i s i s
v
P E
L
E DGP


Where E
v
is vertical illumination at eye level [lux], L
s
is luminance of the source
[cd/m2], ω
s
is solid angle of the source [sr], and P is Guth position index [-].

The calculated DGP scores can be used to determine the degree of perceived glare, as
shown in Table 2-7.

52
Table 2-7. Glare criterion for daylight glare probability.
Degree of Perceived Glare DGP
Imperceptible < 0.35
Perceptible 0.35–0.40
Disturbing 0.40–0.45
Intolerable > 0.45
Source: Jakubiec and Reinhart, 2010, p.156.

The luminance distribution within a field of view can be recorded using CCD
camera-based luminance mapping technology. Also developed by Wienold (2006),
Evalglare uses a luminance picture in the radiance picture format that enables the
user to apply the picture also to simulated lighting scenes. This program
automatically detects glare sources within the view of a 180-degree fish-eye lens.

Wienold (2009) also developed a simplified glare index that can correlate only the
vertical illuminance to the levels of glare. The vertical illuminance level can be
easily calculated in Daysim by locating a calculation grid at the human eye. It can
also be measured using an illuminance meter at the human eye point. The DGPs
formula is as follows:

184 . 0 10 22 . 6
5
   

v
E DGPs

Where Ev is vertical illumination at eye level [lux].

53
Furthermore, Wienold (2009) noted the limitation of this simplified formula: the
equation neglects the influence of individual glare sources. Therefore, the DGPs can
be applied only if no direct sun or specular reflection hits the eye of the observer
(Wienold 2009). This means that DGPs cannot be used for absolute glare factor
conditions that usually include a direct view of glare sources in the field of view.

2.3.2 Daylight Glare Index (DGI)
The glare index (GI) developed by Hopkinson (1957) was the first formula to
calculate discomfort glare specifically from a large glare source. Extensive
experiments were conducted with an illuminated diffusing screen that had provided a
uniform luminance value (Inanici 2005). Observers were exposed to the large glare
source and asked to rate the degree of the disturbance on a subjective scale. The field
of view and glare source were defined in terms of steradians. This method seemed to
address the complex phenomena of glare in terms of contrast between a high
intensity glare source and background luminance (Japee and Schiler 1995).
Hopkinson then discovered that the DGI’s estimated glare levels are higher than the
actual perceived discomfort glare levels (Velds 2002; Kim 2012).

By replacing the luminance of the source with the luminance of the window, Chauvel
(1982) developed the daylight glare index (DGI), which accounts for the differences
between the glare experienced from a real window and the glare calculated using GI.
The DGI formula is shown below (Bellia et al. 2008).
54

100.478 . ∙
.
0.07 ∙ .
∙

Where L
s
is the luminance of each part of the source [cd/m
2
], L
b
is the average
luminance of surfaces in the environment within the field of view [cd/m
2
], L
w
is the
weighted average luminance of the window in function of the relative areas of sky,
obstruction, and ground [cd/m
2
], ω is the solid angle of the window [sr], and Ω is the
solid angle of the source, modified in function of the line of sight

The glare criterions for the glare index and daylight glare index are shown in Table
2-8.

Table 2-8. Glare criterion for glare index and daylight glare index.
Degree of Perceived Glare GI DGI
Just Imperceptible 10 16
Noticeable 13 18
Just Acceptable 16 20
Acceptable 18.5 22
Just Uncomfortable 22 24
Uncomfortable 25 26
Just Intolerable 28 28
Intolerable N/A 30
Source: Bellia et al. 2008, p.2.

55
2.3.3 British Glare Index (BGI)
The British glare index (BGI) was developed by the British Research Establishment
(BRE). The BGI formula is similar to DGI, but it considers the Guth position index,
as well. Originally developed by Petherbridge and Hopkinson, this formula cannot
consider a large glare source. As stated above, Hopkinson later modified this formula
to allow it to assess discomfort glare from large glare sources (Toshie and Ken-ichi
1990). The BGI formula is provided below.

100.478 ∑
.
∙
.
∙
.

Where L
s
is the luminance of each part of the source [cd/m
2
], L
b
is the average
luminance of surfaces in the environment within the field of view [cd/m
2
], P is the
Guth position index, and Ω is the solid angle of the source, modified in function of
the line of sight.

2.3.4 Visual Comfort Probability
Visual comfort probability (VCP) is the discomfort glare evaluation rating originally
developed by Guth (1966) and later completed by Dilaura (1976). It has been used
for decades by the Illuminating Engineering Society of North America (IESNA) to
evaluate discomfort glare caused by artificial lighting systems. The VCP method
estimates how many people out of 100 would be comfortable within the space
56
(Schiler 1995). Higher VCP values represent lower discomfort glare within a given
lighting system. The experimental work predicts discomfort glare ratings (DGR)
from luminous conditions such as glare source luminance, luminances in the field of
view, the visual size of the glare source, and the location of the glare source in the
field of view. The field luminance or the background luminance is calculated as the
sum of the luminance of each luminaire and its contribution to the background by
interaction with the reflective surfaces of the room. The DGRs are directly
convertible to the probability that a population of viewers will consider that
sensation acceptable. In a given geometry along a specified line of sight, a luminaire
with a VCP of 80 will produce acceptable glare sensations in 80% of the population.
Veitch’s (2006) study adopted standard conditions for the calculation of VCP; in it, a
consensus determined that a luminaire with a VCP of 70 is an acceptable glare
threshold (Veitch 2006). However, it is not advised to apply VCP to the context of
discomfort glare from daylighting, since the size of glare sources in VCP is limited to
standard fluorescent lighting systems.

2.3.5 Unified Glare Rating
This method was developed by the Commission Internationale de L’eclairage (CIE)
in 1995, in order to establish an international standard for visual glare analysis. This
method uses a combined formula of Hopkinson’s glare constant formula and the
Guth position index. It uses a limited range of glare source sizes, and thus is effective
only for normal-sized glare sources within a field of view. The value for the unified
57
glare rating (UGR) can be found using the following formula:









2
2
10
25 . 0
log 8
P
L
L
UGR
s
b


Where L
b
is the average background luminance [cd/m
2
], L
s
is the luminaire
luminance [cd/m
2
], ω is the solid angle of the luminaire [sr], and P is the Guth
position index.

Depending on the UGR rating, the level of glare can be determined as shown in the
UGR index shown in Table 2-9. Much as with the visual comfort probability method,
UGR was developed to analyze glare sources with limited sizes and luminance.
Therefore, it cannot be used for daylight glare analysis.

Table 2-9. Unified glare rating index.
Degree of Perceived Glare UGR
Just Imperceptible 10
Perceptible 16
Just Acceptable 19
Unacceptable 22
Just Uncomfortable 25
Uncomfortable 28
Just Intolerable 31
Source: Hwang and Kim 2002, p.120.

58
2.3.6 CIE Glare Index
The CIE glare index (CGI) was originally developed by Einhorn in 1969, and then
was later modified and adopted by the Commission Internationale de l’Eclairage
(CIE). Much as with DGP, CGI also considers the direct and indirect vertical
illuminance values that human eyes receive. This index was originally developed for
artificial lighting sources, not for daylight glare (Jakubiec and Reinhart 2010).
However, several recent studies show that CGI gives the best agreement with reports
of discomfort glare in open plan daylit spaces among the existing glare indices
(Isoardi et al. 2012).

82
1 500 ⁄

∙

Where L
s
is the luminance of each part of the source [cd/m
2
], ω is the solid angle of
the window [sr], E
d
is the direct vertical illuminance, E
i
is the diffuse vertical
illuminance, and P is the Guth position index

2.3.7 Relative Visual Performance
The relative visual performance (RVP) method was developed in order to overcome
deficiencies in previous models of visual performance, including the visibility level
model, for the purpose of improving illuminance level recommendations. The RVP is
a measure of human visual performance with regards to discomfort glare. The
59
measurement was created based on human speed and accuracy of performing a task
under given conditions. A number of people were asked to perform a task in a room
lit by a particular light source and background lighting. Both the background and the
light source illuminance level were changed while the group worked on the task, and
the researcher recorded the time it took the occupants to finish the task and the
number of errors that occurred at different combinations of light levels. This method
is best used for measuring the productivity of a result of the changing illuminance on
a surface, but is not useful for predicting or evaluating the luminance quality of
discomfort glare in a space (Tedjakusuma 2003).

Where RVP is relative visual performance, ∆T
vis
is visual performance time, and
∆T
vis
,Ƭ is the estimated value of ∆T
vis
at the readability contrast threshold

Thus, RVP
RT
= 0 when ∆T
vis
= ∆T
vis
,Ƭ and RVP
RT
= 0.998 when ∆T
vis
= ∆T
vis
(Rea
1991).

Different lighting systems can be compared by calculating the RVP for each with the
same task and age of individual specified. When the RVP percentage is high, it
indicates that a better lighting environment for the given task is possible.









  
  



,
'
, '
vis vis
vis vis
RT
T T
T T
RVP RVP
60
2.3.8 Video Photometry Method
Mark S. Rea (1986) developed an image analysis system to measure luminance
levels using video photometry systems. The primary benefit of using video cameras
is their ability to record different solar positions, shadows, and so on over time and to
be correlate action with each variable. Video data can be digitized and used for later
analysis in different algorithms to evaluate the luminous environment, and then be
correlated with the recorded occupant behavior. However, extensive calibrations are
required to offset camera gains, settings, and spectral responsiveness (Schiler 1995).

Many promising discomfort glare methods have been developed, but the existing
metrics are not in the final stage of development when it comes to evaluating
daylight glare problems (Velds 2002). Chapter 3 will explain problems of the
existing glare metrics in detail.

2.4 Existing Discomfort Glare Thresholds
It is also possible to utilize absolute thresholds to detect discomfort glare in a field of
view. There are three existing threshold methods using luminance, glare ratio
(contrast), or illuminance. Building codes recommend using these thresholds when
considering glare from a design standpoint, as the thresholds are simple use.

2.4.1 Luminance Thresholds
In order to define and detect an absolute glare factor in a particular glare scene, one
61
must know the luminance value of glare sources. Compared to the discomfort glare
indices explained above, it is much easier and simpler for users to utilize luminance
thresholds to determine the existence of discomfort glare problems. Therefore, the
following luminance thresholds have been developed to avoid discomfort glare
issues inside an office area. As with many existing glare indices, research groups
have not found agreement on what can be considered to be an absolute luminance
threshold.

There are three luminance thresholds for discomfort glare issues based on the
Swedish National Board for Industrial and Technical Development (NUTEK)
guideline for energy efficient offices (Dubois 2001):
 Luminance higher than 2,000 cd/m
2
in any point in a room.
 Luminance between 1,000 cd/m
2
and 2,000 cd/m
2
in a field of view
(peripheral).
 Luminance between 500 cd/m
2
and 1,000 cd/m
2
in a field of view
(peripheral).

Osterhaus (2002) explains that the maximum luminance in the field of view should
not exceed 1,500 cd/m
2
, since that amount is approximately ten times brighter than
an average CRT or LCD screen (Osterhaus 2002). He later claims that 2,500 cd/m
2
is
the absolute luminance threshold for daylit indoor spaces (Osterhaus 2009). Linney
(2008) performed a discomfort glare test on 48 different subjects to see if the existing
62
luminance threshold for office environment is accurate. His study shows that the
existing luminance threshold of 1,500 cd/m
2
can be increased to 2,740 cd/m
2
when a
window is positioned to the side of a workstation, and that the threshold can be
increased to 2,160 cd/m
2
when a window is positioned in front of a user (Linney
2008).

Bülow-Hübe (2008) claimed that 2,000 cd/m
2
is too strict for window surfaces,
which may have a sky luminance range from 5,000 cd/m
2
(overcast sky) to 100,000
cd/m
2
(sun in clear sky). Shin et al.’s (2012) study shows that the glare ratings are
acceptable glare at the luminance values of 1,000, 1,800, and 3,200 cd/m
2
, just
uncomfortable glare at a luminance of 5,600 cd/m
2
, and uncomfortable or intolerable
glare at a luminance of 10,000 cd/m
2
. While developing the daylight glare
probability method, Wienold and Christoffersen (2005) discovered luminance
thresholds with more detailed glare level categories: 2,000 cd/m
2
for perceptible
glare, 4,000 cd/m
2
for acceptable glare, 6,000 cd/m
2
for uncomfortable glare, and
8,000 cd/m
2
for intolerable glare, when a person’s field of view is parallel to
windows. Based on the latest findings for luminance thresholds, 4,000 cd/m
2
might
be a more reasonable luminance threshold for absolute glare factor. Further
investigation is required to determine whether or not these values can be utilized for
daylight glare issues.

The aforementioned guidelines do not seem to apply to exterior situations that have
63
much brighter luminous environments than a closed office environment. For exterior
glare evaluation, Schiler (2009) set a luminance value of 12,000 cd/m
2
as absolute
glare luminance threshold for his study on Walt Disney Concert Hall glare evaluation.
This luminance threshold has been verified through human subject study.

2.4.2 Luminance Ratios (Contrast Ratios)
Much like the existing glare luminance thresholds, contrast ratios between
backgrounds and glare source (or task area) are used for glare thresholds. The
Illuminating Engineering Society of North America (IESNA) and NUTEK guidelines
recommend less than a 1:3 ratio between the surroundings and the task at-hand for
visual display terminal office lighting conditions. Lukiesh has claimed that
luminance ratios smaller than 1:5 are desirable, that luminance ratios greater than
1:10 should be avoided, and that a luminance ratio of 1:100 is not tolerable
(Wymelenberg 2012; Osterhaus 2002). Egan (1983) claims that a 1:40 luminance
ratio should not be exceeded anywhere within the human field of view. Egan also
provided the following recommendations for visual comfort: 3:1 between task and
adjacent darker surroundings 10:1 between task and remote darker surfaces, and 20:1
between lighting fixtures (or windows) and sizable adjacent surfaces (Osterhaus
2002). Veitch and Newsham (2000) claim that a 1:10 contrast ratio between
background and glare source luminance is the upper limit of visual comfort. Wienold
and Christoffersen (2005) also claimed that a 1:10 luminous ratio between
backgrounds and visual display terminal is the upper limit of acceptable glare
64
threshold. Osterhaus (2009) claims more detailed contrast ratios as follows: 1:3 for
visual task and immediate surroundings 1:10 for visual task and near surfaces, 1:20
for task and more distance surfaces, and 1:40 for task and any surfaces in the field of
view (Osterhaus 2009). Then again, Schiler (2000) claims that a contrast threshold of
1:3 between the mean background value and the glare source is ideal. This is
different from the contrast ratio between the brightest and darkest spots within a field
of view, which might attain 100:1 with no ill effects, when viewing dark ink on a
bright white page.

The recommended ratio values provided by different research groups are clearly not
consistent with one another. It is also difficult to know whether or not these values
can be utilized for different daylighting situations. Linney (2008) found that the
currently recommended luminance thresholds and ratios might be too low for a daylit
environment. Also, he noted many current recommendations were developed based
on completely different luminous environments, such as incandescent sources and
smaller windows. It would be worthwhile to determine what value is proper for
absolute luminance thresholds in various daylit spaces. First of all, the luminance
ratios of interior environments should be verified by thorough study. Then, the
luminance ratio of exterior environments should also be developed for exterior glare
analysis.

Much as with the luminance threshold issue, we must see if these contrast thresholds
can be used to detect potential glare issues without correlating them with glare
65
source luminance thresholds. The existing recommendations for contrast thresholds
are not consistent, but provide valuable information for developing a new method
that will lack some of the problems from the existing analysis methods.

2.4.3 Illuminance Thresholds
Illuminance values have been also used to determine the existence of discomfort
glare. The conventional lighting criteria in the first stage (1898–1945) were based on
uniform illumination over a horizontal plane, which is still widely used today (Cai
2013). In the second stage (1945–present), the objective was to provide illuminance
suited to human need based on visual performance, which has not yet been fully
achieved because the predominant illuminance-based metrics currently used to
specify, measure, and calculate light levels are inappropriate for the evaluation of
visual perception (Cai 2013). Researchers have developed various illuminance
thresholds for horizontal illuminance at task height and vertical illuminance value
reaching the human eye. Wienold and Christoffersen (2005) claim that 500 lux of
work plane illuminance is ideal for paper work, while 300 to 500 lux is ideal for
VDT work. Wymelenberg’s (2012) recent study found that an upper horizontal
illuminance threshold should be set between 2,000 and 4,300 lux, even if some
individuals may be comfortable with values as high as 5,000 lux. Since illuminance
values can be easily calculated or measured on site, this approach is considered to be
much more convenient than luminance or contrast methods. However, it is important
to note that current illuminance thresholds have not shown strong correlations to
66
occupants’ visual comfort.

2.5 Existing Discomfort Glare Analysis Tools
There are several computer-based tools to evaluate a simulated or photographed
glare scene. The following glare evaluation tools work with either artificial lighting,
daylighting, or both:
 Findglare
 Evalglare
 Per-pixel Lighting Data Analysis

2.5.1 Findglare (Radiance Visual Comfort Calculation)
Developed by Greg Ward, the Findglare tool utilizes the lighting/daylighting analysis
software Radiance to pinpoint discomfort glare. It has implemented two existing
glare analysis formulas: the Guth visual comfort probability (VCP) and the CIE glare
index (CGI). Similar to other methods, Findglare considers four key elements: the
direction, solid angle, and average luminance of glare sources, as well as the
background luminance for a particular viewpoint. The calculation procedure samples
the visual field for bright areas, designates these as light sources, and uses the rest of
the samples to compute the indirect illuminance (i.e. background level). Then, it
computes the glare index for a given field of view.

Findglare takes a Radiance rendering image, locates the glare sources, and calculates
67
the background levels (indirect vertical illuminances) for a specified view field. This
method has limitations, but may give the most reliable results for daylight glare
analysis (Ward 1992).

2.5.2 Evalglare
Developed by Wienold, Evalglare is used to evaluate glare originating from daylight
solutions. It can analyze glare scenes captured by HDR imaging techniques or
simulated in Radiance. Several computer-based daylighting simulation programs
such as Daysim, Diva, and HDRscope utilize Evalglare code to calculate glare scores.
In Evalglare, glare sources are automatically detected based on a threshold value,
which can be a fixed luminance value, a luminance value that is x-times higher than
the average luminance of the entire picture, or a luminance value that is x-times
higher than the calculated average luminance of a given task area (Wienold and
Christoffersen 2006; Inanici 2004).

Evalglare software not only calculates DGP, but also calculates the following four
glare indices: DGI, UGR, VCP, and CGI. It can analyze a digital image in either
HDR format or PIC format. It can create an HDR format image from conventional
digital images with different exposure settings using a digital camera and HDR
software. It can create PIC format images from Ecotect-Radiance simulations. Even
though it is recommended by Wienold (2009; 2012) to use a 180-degree fisheye
image to obtain a more accurate DGP evaluation, the software can analyze a normal
68
perspective image in HDR format. By typing Evalglare commands in a DOS window,
the HDR (or PIC) image can be resized for calculation or recreated to a color-coded
image that identifies potential glare sources. Once Evalglare analyzes an image, it
calculates the luminance value and location of each pixel of the image and then uses
the information to calculate the crucial values, including background mean
luminance, glare source luminance, glare source position, solid angle of glare
sources, vertical illuminance, and direct vertical illuminance. Since Evalglare is an
open source software program, users are able to check how the code calculates each
glare index. Degrees of perceived glare were defined in different glare score ranges
for DGP, DGI, UGR, VCP, and CGI (Table 2-10). Unlike the other glare methods,
the score represents probability of comfort, so a higher score indicates better visual
comfort in VCP.

Table 2-10. Degree of perceived glare in different glare indices.
Degree of Perceived Glare DGP DGI UGR VCP CGI
Imperceptible < 0.35 < 18 < 13 80-100 < 13
Perceptible 0.35-0.40 18-24 13-22 60-80 13-22
Disturbing 0.40-0.45 24-31 22-28 40-60 22-28
Intolerable > 0.45 > 31 > 28 < 40 > 28
Source: Jakubiec and Reinhart 2012, p.156.

Jakubiec and Reinhart (2010; 2012) have actively performed glare evaluation studies
using Evalglare and have found the following issues with the software’s five glare
indices. Based on their research, DGP shows the most robust results for most
daylight situations. VCP is not appropriate for daylight based glare, and CGI tends to
69
show much higher glare levels than the other indices. DGI and UGR can be used for
daylight glare evaluation, but they work only when there is no direct sunlight
(Jakubiec 2010). In their research summary, Jakubiec and Reinhart (2010) claim that
DGP works best for most daylight situations among those five indices. Unfortunately,
DGP does not have a capability to analyze glare issues that include extremely high
luminance glare sources such as direct sun penetration. Although the software has
great potential to analyze glare caused by sunlight, it is still in the experimental stage
and has yet to be validated.

2.5.3 Per-pixel Lighting Data Analysis
Lawrence Berkeley National Laboratory (LBNL) has developed a lighting
measurement and a simulation toolbox called Per-pixel Lighting Data Analysis,
which provides several techniques for analyzing luminance distribution patterns,
luminance ratios, and adaptation luminance and glare assessment (Figure 2-10). This
toolbox is available in both computer-generated renderings and digitally captured
HDR images. In this toolbox, the Glare Module is an interactive script for locating
glare sources and computing glare indices. Glare sources are identified based on a
threshold value, which can be specified manually by the user as a fixed luminance
value, or can be computationally determined from the average luminance in the field
of view. The adaptation level is computed using indirect vertical illuminance as the
background level. The following glare indices and quantities can be calculated in the
toolbox (Inanici 2005); these are the major visual comfort metrics used in different
70
countries around the world:
 Guth visual comfort probability (VCP)
 CIE glare index (CGI)
 Unified glare index (UGI)
 BRS glare index
 Daylight glare index (DGI)
 Guth disability glare rating

Figure 2-10. Per-pixel data extraction from physically based renderings and HDR
photography.
Source: Inanici 2005, p.4.

The glare module provides automated calculations of DGI and VCP to determine
71
glare sources from daylighting and fluorescent lighting fixtures. Figure 2-11 shows a
glare analysis from a study performed by Inanici (Inanici 2005). Visual comfort was
determined using various controlled shading and lighting systems through sets of
images captured at the New York Times (NY Times) Headquarters Mockup Building.
A total of 150 High Dynamic Range (HDR) photographs were assembled from
multiple exposure sequences captured in November of 2004 (Inanici 2005).

Figure 2-11. Glare analysis of HDR photographs from a workstation in the NY Times
mockup building.
Source: Inanici 2005, p.11.

When using the Per-pixel Lighting Data Analysis tool, there are several different
72
approaches to evaluating luminance distribution. The following list provides a quick
overview of the evaluation options:
 Region of interest—luminance of the elements or region of interest are
isolated from the rest of the scene.
 Image subtraction—different luminance distribution of two or more scenes.
 Luminance ratios—luminance ratios provided for task and architectural
elements such as wall, ceiling, and surround.
 Luminance contrast—contrast between a target and background in luminance.
 Visual field of view—luminance in foveal vision, binocular vision, and
peripheral vision.

The maximum, minimum, and average values are calculated from a matrix that can
correspond to the whole scene, a surface, or a region of interest (Inanici 2005). By
considering these different applications, the tool has the significant potential to create
different glare analysis methods with the use of HDR imaging. IES also provides a
potential glare detection function that indicates what surfaces have high luminance
values in a field of view (Figure 2-12).

73

Figure 2-12. IES’ potential glare source detection function.
Source: Hardin et al. 2008, p.10.

Beyond the glare analysis tools explained above, there are several different
approaches to evaluating discomfort glare issues. Neural networks have been
developed to understand or mimic the functions of the human brain. This has been
used to gain an understanding of perceived discomfort glare. A neural network can
be provided with the source luminance (L
s
), the background luminance (L
b
), and sets
of training data made by a subject. Each set of data consisted of three items: L
s
, L
b
,
and the value of a subject’s assessment according to DGI. Then, the neural network
was trained to approximate a subject’s assessment via the input of L
s
and L
b
. Once
the connection weights of the network had been established, the network was ready
to start predicting on the basis of new, unseen, input data. Thus, the network was
provided with new sets of data, and this time the set consisted only of L
s
and L
b
. The
network then attempted to predict the subject’s assessment based on the relevant
74
values of L
s
and L
b
. It used a common network paradigm and looked at a very simple
architecture with simple node properties. In principle, it should be straightforward to
add many more factors that are likely to influence discomfort glare, such as time of
day, complexity of task, and so on. It seems possible that such extra parameters will
improve the performance of the network’s analysis of discomfort glare (Osterhaus
2005).

Researchers at City University in London have studied the objective measurement of
light scatter in human eyes. New equipment has been developed to enable a direct,
objective estimation of light scatter in the eye. This new technique provides a more
rapid (~4 minute test) and accurate estimation than conventional clinical and visual
psychophysical tests (Spencer et al. 1995). The equipment has not been fully
validated yet, but there is on-going research with this new technique. In fact, another
study is underway to discern the human eye’s threshold for discomfort glare and
disability glare. Subjects judged whether or not different sizes and intensities of glare
sources were in their comfort level. The results can help to establish subjects’
perceived discomfort glare in relation to the measurement result of light scatter in
their eyes.

Functional magnetic resonance imaging (fMRI) is an objective assessment technique
of neural activity in the human brain. Recently, fMRI has been used for glare
analysis. Several studies discovered the effect of luminance contrast on the blood
75
oxygenation level-dependent (BOLD) signal in response to human primary visual
areas. By using fMRI, other studies also have shown an increase in the activation
levels within the visual cortex (area V1) with increasing stimulus luminance contrast
(Goodyear 1998). The visual cortex has a very well-defined map of the spatial
information in vision. Areas V1 and V2/V3 have also been shown to respond reliably
and strongly to changes in the luminance of uniform surfaces (Osterhaus et al. 2005).
These studies have not yet been validated, but results from these studies have great
potential to improve discomfort glare indices.

2.6 High Dynamic Range (HDR) Imaging Technique
HDR images store a much larger range of luminance information in a digital image
than do conventional low dynamic range (LDR) photographs. They have been widely
used for graphical and artistic purposes in both photography and movie industries.
The popularity of HDR encouraged some scientists to use the technique for their
scientific research. There have been a few studies on HDR imaging for research
purposes, particularly in the context of luminance evaluation, glare evaluation, and
daylighting analysis (Inanici 2004; 2005; 2009). In photography, the dynamic range
represents the ratio of maximum and minimum luminance of a picture. The dynamic
range capacity of the human eye with respect to luminance ranges from 0.0000001
cd/m
2
to 1,000,000 cd/m
2
, which is a range of nearly twelve orders of magnitude.
Within a single scene, the adapted visual range of the human eye becomes around 1
to 100,000, depending on the brightness and contrast of the scene (Reinhard et al.
76
2010). However, conventional 24-bit digital images have a dynamic range of only 0
to 255.

LDR has 8 bits each for red, green, and blue (RGB) colors per pixel. This yields a
total of 16.7 million different color possibilities, but the images will store only a 256
luminance range (a value of 0 is black and 255 is white) per each color channel
(Reinhard et al. 2010). These values enable us to judge the different luminance
intensities in a picture. However, the luminance shown in a conventional digital
image is different from real world luminance values. Instead of giving accurate
luminance values, conventional images only show the ratio between darker or
brighter pixels in an image. This happens because conventional images must
compress the real world’s luminance range into a smaller range of 0 to 255
luminance intensity. Therefore, these conventional images and display devices are
not sufficient to record, let alone represent, the dynamic range of luminance in the
real world.

High dynamic range (HDR) imaging, on the other hand, enables a digital image to
store a much larger dynamic range than conventional digital photographs. Depending
on exposure settings, HDR imaging can capture luminance intensity ranges greater
than 0 to 100,000. As shown in Figure 2-13, the real world produces a much larger
dynamic range than a conventional digital camera and display device can represent,
and HDR spans even larger than that. In fact, Yin and Schiler (2011a; 2011b)
77
demonstrated that the clearest way to study such ranges is by using logarithmic
histograms.

Figure 2-13. High dynamic range of luminance values.
Source: Suk and Schiler 2013, p.115.

Currently, there are two different ways to create HDR images: they can be rendered
using 3D computer graphics techniques such as radiosity and ray tracing, or they can
be captured from real scenes through photography. The first method is widely used in
the movie and animation industries, and utilizes rendering algorithms and other
computer graphics techniques. The second method utilizes inexpensive conventional
digital cameras to create HDR images (Reinhard et al. 2010). Due to the limitations
inherent in most digital image sensors, it is not generally possible to capture the full
dynamic range of a scene in a single exposure. However, using the manual capability
of a conventional digital camera, multiple sequences of a static scene can be taken
with different exposure values specifically controlled by the shutter speed. Within the
78
multiple exposures, some parts of the image will be properly exposed and other parts
will be under or overexposed. It is possible to ascertain the relationship of the
different luminances of different pixels to determine the most accurate luminances by
identifying the portions of each image with the largest clarity. Following that, an
HDR image can be created by combining these multiple sequences of LDR images
using specifically developed HDR software, as shown in Figure 2-14. Many useful
techniques, such as alignment and ghost removal, can be used to get an almost
perfect HDR image, so long as the scene stays static and the camera stays in the
exact same place.
79

Figure 2-14. Conventional low dynamic range images and a high dynamic range
image of hanging art pieces below skylights.

HDR imaging has great potential in the analysis of daylighting and daylight glare,
which may lead to future possibilities of utilizing HDR imaging. HDR presents the
opportunity to accurately record much higher luminance levels than were previously
available in luminance mappings.

2.6.1 HDR Photographing
HDR imaging technique requires several pieces of equipment and, in most cases, this
equipment is not expensive or difficult to obtain. Any digital camera with manual
control capability can capture multiple sequences of LDR images to create an HDR
image. A DSLR with an auto-bracketing function, a tripod, and HDR software are all
80
that are needed for HDR imaging.

The following list provides step-by-step directions for capturing LDR images for the
purpose of combining them with HDR software.
1. Capture a good calibration sequence first in order to create a camera
response curve. A daylit space is best for this purpose.
2. Pick a scene with a high dynamic range and smooth transitions (an indoor
scene with a view outdoors, but no direct sun, is ideal).
3. Use a fixed film speed (preferably ISO 100).
4. Fix the aperture to eliminate vignetting variance (i.e. reduction of
brightness at an image’s periphery compared to the center).
5. Fix the camera’s white balance on a specific setting (daylight setting
preferred).
6. Use a tripod for optimal alignment.
7. Use multiple exposures. Bracket with shutter speed only.
8. Take as many exposures as possible. Take more sequences for outdoor
pictures than interior pictures
9. Calibrate with a luminance meter if available.

It is necessary to calibrate a camera’s response the first time you use that camera, and
then this calibration can be reused for later exposure sequences and for other scenes,
so long as it is for the same camera and lens combination. The camera response
81
curve can be automatically generated using HDR software.

To merge several LDR exposures into a single HDR image, one must understand the
camera response function. The camera response function shows the relationship
between the amount of incoming light and image pixel values of a digital camera.
There is almost no camera that has linear response curve, since most camera makers
boost their image contrast beyond the standard sRGB gamma to allow for the
creation of better and unique images. However, it is possible to have a linear camera
response function with proper image sequence, as long as the response function stays
the same between exposures in a single camera. There are several proposed methods
to make a linear input data, including Debevec and Malik’s enumerated table and
relative exposure ratios.

Debevec and Malik presented a simple, practical, and accurate technique for
recovering the camera response function directly from a set of images taken with
known exposure values. The method is shown in Figure 2-15. By capturing different
exposures of a static scene, which is the same as the HDR imaging procedure, it is
possible to obtain the pixel values and exposure time of the captured LDR images.
Based on these values, the following graphs are plotted in the left chart (Figure 2-15).
Three separate image positions were sampled at five different exposures. The relative
exposure ratios at each of the three positions are given by the speed settings on the
camera; the linear optimization process is used to achieve a smooth and monotonic
82
curve.

Figure 2-15. Camera response function recovery from five exposures.
Source: Debevec and Malik 1997, p.373 and 375.

For scientific research, it is necessary to record the luminance information of a real
scene with a digital luminance meter during the HDR image capture process. This
procedure is called absolute calibration. Uniformly illuminated surfaces are good
sources for actual luminance measurements. Measured luminance values from the
luminance meter are needed to compare luminance values from an HDR image. If
the luminance values match, the absolute calibration procedure is complete. If the
83
luminance values do not match, then it is necessary to calculate a calibration factor
using the ratio of the HDR luminance over the real luminance with the following
equation.

CF = Luminance Real / Luminance HDR

This calibration factor should be close to 1.0. If the calibration factor is not close to
1.0, the entire procedure needs to be re-examined in order to locate the error. Placing
a known luminance in an image can provide an alternative approach to calibrating
luminance information. This approach was used to calibrate an LDR image and
would also work perfectly in an HDR image.

2.6.2 Luminance Mapping
Conventional digital photography is not an adequate technology to capture the
radiance distribution of a scene, since it cannot capture the full spectral content and
present dynamic range perceived by the human eye. Digital single-lens reflex
cameras (DSLRs) are pushing the boundary of what can be captured in a single
digital exposure, but it is still not enough.

Using a traditional measuring device such as a luminance meter, it is possible to
measure the photometric information point by point. However, these measurements
are time consuming and prone to error, due to measurement uncertainties in the field.
84
Data obtained by luminance meters may also be too coarse for use in analyzing
lighting distribution and variation. Clearly, there is a need for a tool that can capture
the luminance values within a large field of view at high resolution, in a quick and
inexpensive manner (Inanici 2004).

The HDR imaging technique satisfies these requirements. HDR imaging was not
specifically developed for lighting measurement purposes, but there have been a
number of studies to evaluate its potentials as a luminance measurement tool. The
main focus of these studies has been to evaluate whether the pixel values in an HDR
image correspond to the physical quantity of luminance within a reasonable accuracy
and repeatability. With the help of currently available HDR software, it is also
possible to create false/pseudo color images or iso-contour lines to provide better
visual presentations of the luminance map of an image. Unlike the information
contained within LDR images, this luminance information is not compressed or lost
in HDR images.

Even sky conditions can be captured by the HDR imaging technique, as shown in
Figure 2-16. As a potential daylighting analysis tool, Image Based Lighting (IBL) is
a technique that uses an HDR image as a global illumination light source in
computer rendering. Using the IBL technique, Inanici performed a study to improve
the accuracy of lighting and daylighting simulations with site specific, virtual sky
models. HDR images of the sky dome and a daylit indoor space were captured for a
85
specific site at different times under different sky conditions. Physically based
rendering (PBR) and image based rendering (IBR) techniques were then used to
create a simulation of the same indoor space. The results from HDR images and
simulated renderings were compared to evaluate image based sky models. The study
found that the image based sky models provided a more accurate and efficient
method for defining the sky luminance distributions and the impact of surrounding
urban fabric and vegetation than did the generic CIE sky models.

Figure 2-16. Captured sky images: cloudy, partly cloudy, and mostly clear sky.
Source: Inanici 2009, p.269.

2.6.3 Limitations of HDR Imaging
There are two major limitations of HDR imaging with regards to glare analysis
studies. First, HDR imaging requires reasonably stable conditions over the period of
86
measurement. Unlike in the movie or photography industries, high accuracy is a
critical objective of using HDR imaging technique in glare analysis research (Inanici
2004). Dynamic lighting conditions resulting in significant light changes between
differently exposed photographs can compromise the accuracy of the end result.
Second, the glare indices being used today were developed long ago, when
researchers did not have the measuring capabilities that are available today. At that
time, the glare studies were predominantly based on oversimplifications and
impractical assumptions. Utilization of these glare indices with HDR images invites
oversimplification of the results. Moreover, the DGI and VCP were developed to
determine glare from large area sources and fluorescent lamps, respectively; neither
of them are capable of determining glare from integrated daylighting and modern
electric lighting solutions. There is a need for the development of new glare indices
utilizing the advanced measuring capability of modern technology (Inanici 2004).

This chapter defines and describes light transmission and reflection in relation to the
building envelope. It also outlines the existing glare indices and tools. It is possible
to assume that many variables in the glare indices make it difficult to differentiate
absolute glare factor and relative glare factor when evaluating discomfort glare. The
existing luminance thresholds and contrast ratios are not consistent, but they can
provide a clue to the essence of absolute luminance and contrast values that cause
discomfort glare. Chapter 3 investigates the existing glare indices and tools with
particular emphasis on any issues of inaccuracy or inconsistency.
87

Chapter 3 Preliminary Study
Issues within Existing Daylight Glare Analysis Tools and Methodologies

This chapter describes preliminary research performed to prove or disprove the
stated problems of existing glare tools and methods. It also examines why those
existing methods have not been widely utilized by professionals or students.
Specifically, this chapter will discuss an interior daylight glare analysis case study, an
investigation on Evalglare, a summary of the problems of existing glare analysis
methods and tools, and a description of the field of view problem. It will conclude
with a summary of ideas for the creation of a new daylight glare analysis
methodology.

3.1 Interior Daylight Glare Analysis on USC Watt Hall third floor
Prior to the main study, it was necessary to fully understand the pros and cons of the
various existing glare analysis tools and methods. Therefore, several case studies
have been performed with the existing tools and methods. The research projects were
performed on the spaces in USC Watt Hall’s third floor, since the main research on
human subjects was planned to be performed in the same building. There was no
preliminary study for exterior glare, since none of the existing tools and methods are
designed for exterior glare issues. The findings from the interior glare study were
also used to develop the exterior glare analysis method.
88

Ray and Nadine Watt Hall was completed on May 14, 1974, for use by USC’s School
of Architecture. Watt Hall initially consisted of two stories above grade and one
basement floor. The building is a bush hammered, exposed concrete structure with a
significant percentage of glazing. Studios and offices of the School of Fine Arts are
located at the ground level. The second floor is occupied by undergraduate
architecture studios and the School of Architecture offices. The new third floor,
which was built in 2006, holds five graduate program studios and all the faculty
offices. The new expansion is well lit by large windows in offices, with window
walls between studios and sky gardens, and with a clerestory around the entire studio.

Figure 3-1. Watt Hall at University of Southern California.
Source: Suk and Schiler 2007, p.110.
89

Because of the big windows to the sky gardens and clerestory, the studios get
sufficient light throughout daylight hours. There are certain periods of each day that
cause glare problems inside the building. Sometimes, the sunlight is too bright to
work in the studio spaces.

In 2007, discomfort glare and the veiling reflection issues of Watt Hall’s third floor
were studied using the histogram glare method developed by Schiler (Suk and
Schiler 2007). The histogram glare score was calculated after processing luminance
histograms with photographs. The score is decided based on the ratio between the
median brightness of the background and the median brightness of the glare source.
If the score is over 3.0, it is concluded that there is a discomfort glare problem.
RASCAL and CULPLITE software were used to create the luminance histograms
from pictures. The 2007 study found that glare scenes with extremely bright glare
sources could not be analyzed by photographic methods.

The research objective of this study was to determine whether the new glare analysis
method with HDRI can evaluate the absolute and relative glare factor issues of Watt
Hall’s third floor. The veiling reflection was not analyzed.

3.1.1 HDR Photography of Interior Glare Issues
To support the idea of a correlated threshold method for absolute and relative glare
90
factors, a glare analysis study using the existing methods (DGP, DGI, UGR, VCP,
and CGI) and tools (HDR imaging technique and radiance simulation) was
performed. The interior space of the existing building was evaluated in terms of
daylight glare issues. A total of ten different views were captured by the high
dynamic range (HDR) imaging technique with the following four pieces of
equipment: a digital SLR camera, a fisheye lens, a tripod, and HDR software. A
Nikon Coolpix 4500 Camera, a Nikon FC-E8 lens, and a tripod were used to capture
conventional digital images. Then, a popular HDR software program called
Photosphere was used to create HDR images. A luminance meter (Minolta 1 degree)
was utilized to measure actual luminance values while capturing the HDR images of
Watt Hall’s third floor. The luminance values obtained by the luminance meter were
used as a known luminance value to calibrate the HDRI luminance information in
Photosphere. Photosphere then “remembers” the response curve of the camera and
lens combination that was used. Thus, it was not necessary to measure luminance
values for every single HDR photograph. This absolute luminance calibration
procedure is required for each camera and lens combination if different cameras or
lenses are used.

The photographs of Watt Hall third floor were taken every thirty minutes from 7:00
AM to 6:30 PM on April 29, 2011. A clear day was selected, as it was assumed that
the clear sky would provide the worst glare conditions. A Nikon Coolpix 4500 digital
camera, a fisheye lens, and a tripod were prepared for this test. The camera was
91
mounted at 1.37 m (4' 6'') above the floor (this height is the author’s seated eye level)
and aimed parallel to the floor. A total of ten different views were photographed at
three different sections of Watt Hall’s third floor, in the locations and directions
shown in Figure 3-2. Location A is at the south corner of the building and Location B
is at the southwest corner of the building. Location C is inside the atrium space. All
three locations are frequently used by a large number of students and faculty
members for classes and presentations. In Figure 3-2, red arrows indicate the
orientations for each view. Location A has views 01 through 04, which are aimed
southeast, southwest, and northwest. Location B has views 05 through 07, which are
aimed northwest, southwest, and southeast. Location C has views 08 through 10,
which are aimed southwest, northeast, and southeast. Using the auto bracketing
function of the camera, five photographs with normal exposure, +- 1 full stop
exposures and +- 2 full stop exposures were taken for each view, in order to capture
the dynamic range of the human eye. All views include big windows or clerestories
that can cause potential glare issues.
92

Figure 3-2. Watt Hall’s third floor plan and photograph locations.
Source: Suk and Schiler 2012, p.453.

Figure 3-3 shows the ten different captured HDR images at 9:00 AM. The exact same
views were simulated under CIE clear sky condition in Ecotect-Radiance to compare
DGP, DGI, UGR, VCP, and CGI scores between captured HDR images from
Photosphere and virtually simulated scenes from Ecotect-Radiance.

93

Figure 3-3. Ten fields of view captured by HDR imaging technique.
Source: Suk et al. 2013, p.116.

3.1.2 Ecotect-Radiance Simulation of Interior Glare Issues
A virtual model of Watt Hall’s third floor was built in Ecotect to run daylighting
simulations in Radiance, and identical views were simulated at the three different
sections of Watt Hall’s third floor. As with the HDR photographs, simulations were
done every 30 minutes from 7:00 AM to 6:30 PM on April 29, 2011. A standard CIE
clear sky model was chosen in the Radiance simulation to match the sunny day
condition of the HDR photography. The reflectance properties of the virtual model
94
were matched to the actual conditions of interior surfaces as much as possible: the
ceiling was set at 75%, the walls at 40%, the floor at 25%, and the furniture at 40%
or 50%, depending on the finish. The transmittance of the window was set to 91%,
although the window actually had a 90% transmittance value. However, it was fully
understood that the surface properties in Radiance are not identical to the actual
space of Watt Hall’s third floor. Also, it was expected that the missing trees and
neighbors in the Ecotect model would cause different results between captured and
simulated scenes. It is important to note that the purpose of this test was not to
compare the HDR imaging technique to the Radiance simulation. These tools were
used to see whether or not the five glare indices provide consistent and accurate glare
evaluations from these existing tools. A total of ten different simulated scenes at 9:00
AM are shown in Figure 3-4.

95

Figure 3-4. Ten field of views simulated from Ecotect-Radiance simulation.
Source: Suk et al. 2013, p.117.

Both captured HDR images and simulated PIC scenes were analyzed in Evalglare
software to analyze glare evaluations using five glare indices (DGP, DGI, UGR, VCP,
and CGI). The purpose of the analysis was to investigate whether or not different
glare indices would provide consistent evaluation results when different tools were
utilized to capture or simulate glare scenes. Evalglare’s default settings for
luminance threshold factor, peak extraction, search radius angle, and disabled
smoothing function were applied. Also, Evalglare’s internally calculated vertical
96
illuminance values were used instead of externally measured vertical illuminance
values, since the vertical illuminance value was measured only once for calibration
purposes.

The analysis shows that the glare evaluations from the five different glare indices in
the captured HDR images and simulated PIC scenes are not consistent at all.
Specifically, DGP scores from captured HDR images are suspiciously low, as they
indicated imperceptible glare in all ten views throughout the day. Through an
additional study to understand the reasons for the suspiciously low DGP scores, Suk
and Schiler (2012) found that Evalglare calculates vertical illuminance from the
entire square canvas of the HDR image, including the four black corners, instead of
calculating values only from the angular fisheye region. This difference in
calculation drastically decreases the vertical illuminance value, which is the
dominant factor of the DGP score. It was also found that externally measured vertical
illuminance values are necessary for different views to make more accurate glare
analyses in Evalglare using captured HDR images (Suk and Schiler 2012).

Since this problem was not detected from the simulated PIC scenes, this phase of the
research only analyzes DGP scores from simulated PIC scenes. Unfortunately, the
glare evaluations from simulated PIC scenes were also inconsistent among the
different glare indices. After obtaining DGP scores in Evalglare, the simulated PIC
images were also analyzed in a MATLAB code that was specifically developed for
97
analyzing captured HDR files or simulated PIC files. The MATLAB code made it
possible to analyze different portions of the glare scene and create a logarithmic
luminance histogram for a more detailed analysis.

3.1.3 Inconsistent Evaluations of Five Glare Indices
This case study showed that there were inconsistent glare evaluations from different
glare indices. One of the options to calculate existing glare indices is to use a High
Dynamic Range Imaging (HDRI) technique and Evalglare glare analysis software.
From the interior glare evaluation study on USC Watt Hall third floor, it was found
that DGP scores are very low compared to the other glare indices when using this
method. In fact, DGP scores showed “imperceptible glare” for the entire day, even
though there were several scenes with direct views of the sun. Thus, DGP might
underestimate glare scenes when a glare scene is captured using a camera.

More than 240 different scenes were investigated to understand how Evalglare
calculated different glare indices for the same scene, with particular emphasis on
comparing the results when the scene was captured by a camera to those when the
scene was simulated from Ecotect-Radiance. Three examples were chosen to
compare glare evaluations from captured HDR images and simulated HDR scenes in
more detail. One glare scene was selected per location to study whether or not
different glare indices provide a consistent glare evaluation on a same glare scene.

98
Figure 3-5 shows view 01, which looks southeast from Location A. The HDR image
was captured at 9:30 AM, and the simulated HDR scene was set to 8:45 AM. As
shown in the scene, direct sun penetration hits portions of the desk. By setting 8:45
AM in Ecotect-Radiance, it was possible to get almost identical sun penetration from
the simulation. It is not clear yet why a different time had to be set in Ecotect-
Radiance to get the same sun penetration, but perhaps it could be attributed to the
difference between the center of the time zone, the current location and the sidereal
equation of time. Different color coded areas are shown in the simulated HDR scene;
these different colors identify potential glare sources. Potential glare sources include
windows, clerestories, the illuminated desk, and the floor in this scene.

Figure 3-5. Captured HDR image from Photosphere and simulated HDR scene from
Radiance at view 01.
Source: Suk and Schiler 2012, p.455.

99
The glare scores of captured HDR images and simulated HDR scenes were
calculated in Evalglare and plotted on a graph. To visually correlate the five glare
index evaluations, the glare scores of each glare index were multiplied by scaling
factors, and then the correlated scores were plotted (Figures 3-6, 3-8, and 3-10). Not
only do the indices have different overall numerical ranges, but their internal
categories range by different fractions of the total index. This makes direct linear
comparison impossible. However, the rating categories can be compared by category.
Glare ratings were grouped into bins of imperceptible, perceptible, disturbing, and
intolerable glare. Each glare rating was subdivided into ten bins, so that the entire
range of four glare ratings were divided into forty different bins, with imperceptible
glare rating from 0 to 10, perceptible glare rating from 10 to 20, disturbing glare
rating from 20 to 30, and intolerable glare rating from 30 to 40. The different glare
scores of each index were scaled to fall into one of the forty bins. For example, a
DGP score of 0.35, a DGI of 18, a UGR/CGI of 13, and a VCP of 80 fall into bin
#10—thus, all five bars of glare indices are shown at the left end of perceptible glare
rating in Figures 3-6, 3-8, and 3-10. In the same way, a DGP score of 0.4, a DGI of
24, a UGR/CGI of 22, and a VCP of 60 fall into bin #20; thus, all five bars are shown
at the left end of disturbing glare rating in the figures. Any scores in between bins
#10 and #20 were assigned to one of ten possible values according to their relative
worth and properly located in the corresponding bin to visualize the scaled glare
scores. For example, a DGP score of 0.37 falls into bin #14 and a DGI score of 20.4
falls into the same bin. This straightforward binning procedure was used to visualize
100
the different scores of the various glare ratings in a single graph. However, it is still
necessary to note that there is no linear scaling factor that resulted in identical glare
assessments. Even though all five glare index scores fall into bin #35, which is an
intolerable glare rating, that does not indicate that the levels of perceived glare for
five different glare indices are identical; rather, the grouping allows one to visually
compare the glare evaluations of the five indices.

The glare evaluations from the five glare indices are not quite consistent, as shown in
Figure 3-6. Glare evaluations of the captured HDR image are imperceptible for DGP,
perceptible for DGI and VCP, or disturbing glare for UGR and CGI. These
inconsistent glare evaluations are also shown with the simulated HDR scene. The
simulated HDR scene was evaluated as having perceptible for DGI, disturbing for
UGR, or intolerable glare for DGP, VCP, and CGI. Glare scores between the
captured HDR image and the Ecotect-Radiance simulation are quite different for
DGP, VCP, and CGI. DGP and VCP scores from the captured HDR image are
extremely low, compared to the simulation. The Ecotect-Radiance simulation shows
intolerable glare for this scene, while the captured HDR image shows imperceptible
glare or perceptible glare. Without a subjective survey, it might be difficult to
determine what glare index and what image type provides more accurate evaluations.

101

Figure 3-6. Glare score comparison of the scene in Figure 3-5.
Source: Suk and Schiler 2012, p.456.

View 06 is looking to the southwest from Location B. The sun at late afternoon is
coming into the space through west-facing windows. Direct sun penetration hits
more than half of the white desk and creates extreme contrast for students sitting at
the desk. To match the sun penetration of the captured HDR image, the Ecotect-
Radiance simulation was set to 3:45 PM. It is not understood yet why the actual
photographing and the simulation have a 1 hour and 15 minute discrepancy. The
simulated HDR scene does not have students sitting on the chairs and trees at the sky
garden area (outside the windows on the left side of the view), such as are shown in
the captured HDR image (Figure 3-7). It is assumed that these missing objects in the
simulated scene may have contributed to the different glare scores from the captured
HDR image. As briefly mentioned above, more detailed 3D modeling is needed to
minimize the discrepancy between the actual space and the simulation. Potential
glare sources include illuminated portions of desk and floor, as well as windows and
clerestories.
102

Figure 3-7. Captured HDR image from Photosphere and simulated HDR scene from
Radiance at view 06.
Source: Suk and Schiler 2012, p.457.

Again, glare evaluations from both the captured HDR image and the simulated scene
are not consistent. Evaluations of the HDR image vary by the entire range of
perceived glare—from imperceptible to intolerable. The simulated HDR scene also
has different glare degree analyses, including perceptible, disturbing, and intolerable.
With the simulated HDR scene, VCP and CGI show intolerable glare while DGP and
UGR show disturbing glare. DGI shows the lowest glare evaluation among the five
indices in the simulated HDR scenes of view 01 and 06, while UGR shows the
second lowest glare evaluation in both cases. Glare scores from the captured HDR
image are quite different from the scores from the simulated HDR scene. Except for
DGI and CGI, all the other glare indices show different glare degrees between the
103
captured HDR image and the simulation. As with the scene shown in Figure 3-7, it is
difficult to judge from the images the seriousness of the glare scene. However, from
personal in-situ experience, it is likely that the glare perceived in view 06 is similar
to or more serious than view 01. It could be perceptible or disturbing glare.

Figure 3-8. Glare score comparison of the scene in Figure 3-7.
Source: Suk and Schiler 2012, p.458.

One of the most serious glare scenes is shown in Figure 3-9. View 09 faces the
northeast inside the atrium space. This scene clearly shows intolerable glare.
Morning sun is visible through the east-facing clerestory of the atrium and is also
hitting large portions of the interior wall. This scene contains a high glare source and
also has high contrast between the lit and non-lit surfaces of the atrium space.

104

Figure 3-9. Captured HDR image from Photosphere and simulated HDR scene from
Radiance at view 09.
Source: Suk and Schiler 2012, p.458.

As expected, most of the glare scores are evaluated as intolerable glare. With the
simulated HDR scene, all five glare indices make a consistent evaluation. The DGI
score is a little lower than the rest, but still falls into the intolerable glare category.
The captured HDR image shows somewhat different evaluation results, however.
The DGP score is still very low, defining the scene to have only imperceptible glare.
The DGI score, on the other hand, finds the scene to have disturbing glare. UGR,
VCP, and CGI scores show intolerable glare, as does the Ecotect-Radiance
simulation. This shows that the simulated HDR scene works better with the DGP
index than the captured HDR image does. This may be because the simulated HDR
has a much greater range of absolute values (cd/m
2
) than the captured HDR. It is also
possible that Evalglare correctly calculates vertical illuminance from Ecotect-
105
Radiance simulations, while it does not correctly calculate from captured HDR
images. The DGP also appears to be much less accurate with the captured HDR
images that are processed by Evalglare. We may assume that the luminance is simply
beyond the range of the camera and the HDRI software is unable to correct for that.

In addition, DGI slightly underestimates the same glare scene of both captured and
simulated HDR, compared to the other indices.

Figure 3-10. Glare score comparison of the scene in Figure 3-9.
Source: Suk and Schiler 2012, p.459.

While photographing Watt Hall’s third floor, several more intolerable glare scenes
were captured at different times of the day to see how the five glare indices in
Evalglare calculate glare scenes with extremely bright glare sources. Similar to the
results of the analysis shown above, DGP scores from the captured HDR images are
extremely low for all cases, and DGI seems to slightly underestimate the glare scenes
while the other glare indices provide somewhat reasonable or higher glare levels. We
106
may assume that the HDR processing of photographs containing very high
luminances does not capture the correct values at the high end. It is possible that the
Evalglare analysis procedure might have caused the inaccurate evaluation of DGP. It
is probable that the glare indices disagree among themselves, and Evalglare is only
reporting that disagreement.

One potential reason for low DGP scores from the captured HDR images would be
the vertical illuminance factor, since it is the only discrepancy between DGP and the
other indices. The following section addresses this and the other issues in Evalglare.

View 05 at Location B has direct sun penetration through the west-facing windows at
5:30 PM onto the table (Figure 3-11). These images have been analyzed through
Evalglare software. The analysis results are quite inconsistent among one another,
but are similar to the examples already discussed. Evalglare evaluates the captured
HDR image with the luminance range from 0.23 cd/m
2
to 5,385.86 cd/m
2
as follows:
DGP: 0.26 (imperceptible), DGI: 19.9 (perceptible), VCP: 24.2 (disturbing), UGR:
60.0 (perceptible), and CGI: 25.8 (disturbing). The simulated PIC image with the
luminance range from 0 cd/m
2
to 6,315.12 cd/m
2
is evaluated as follows: DGP: 0.452
(intolerable), DGI: 21.5 (perceptible), VCP: 8.66 (intolerable), UGR: 26.64
(disturbing), and CGI: 33.4 (intolerable). Although the captured HDR image has a
wider luminance range than the simulated PIC scene, the evaluation results from
both formats are quite different. The researcher expected to see different results
107
between captured HDR images and simulated PIC scenes due to the modeling details,
surface properties, sky conditions, etc. However, they did not expect to see
inconsistent evaluations from the five indices.

The HDR image is evaluated as imperceptible, perceptible, and disturbing, while the
PIC scene is evaluated as perceptible, disturbing, and intolerable.

Figure 3-11. Comparison of captured HDR image and simulated PIC scene.
Source: Suk et al. 2013, p.119.

As shown in these examples, users can get totally different glare evaluations
depending on the different glare indices. Therefore, it is important to determine what
index can provide the most accurate analysis. From the Watt Hall third floor glare
study, this inconsistency issue was found for all ten views for the entire day. It is
apparent that the glare indices disagree and Evalglare is only reporting that
108
disagreement.

3.1.4 DGP Score Comparisons of Simulated PIC files
DGP scores from simulated PIC files were compared to understand the discomfort
glare issues on Watt Hall’s third floor throughout a single day. As explained in
Chapter 2, perceptible glare occurs if DGP score is higher than 0.35, disturbing glare
occurs when DGP score is higher than 0.4, and intolerable glare occurs when DGP
score is higher than 0.45. When the DGP score is lower than 0.35, the glare is
imperceptible. Three different graphs were created to show DGP score changes in
each view every thirty minutes from 7:00 AM to 6:30 PM.

DGP scores from views 01 through 04 are compared in Figure 3-12. Views 01 and 04
show very high glare scores, ranging from disturbing glare to intolerable glare,
between 7:00 AM and 9:30 AM. The morning sunlight penetrated the east-facing
windows and caused serious discomfort glare. View 02 has relatively lower DGP
scores, while view 03 has the lowest scores—even in the morning. From 11:00 AM
through the rest of the day, all views except view 01 show imperceptible glare. As all
four views were taken at the east corner of the building, serious glare issues early in
the morning are only to be expected.

109

Figure 3-12. DGP score results from view 01 to view 04.
Source: Suk et al. 2013, p.117.

DGP scores from view 05 through 07 are compared in Figure 3-13. Unlike views 01
to 04, views 05, 06, and 07 show increasing DGP scores for the entire day. The
perceived glare levels of view 05 continuously increase from 11:00 AM to 6:00 PM.
Since view 05 is looking at west-facing windows, this glare evaluation makes sense.
View 06 also has disturbing glare problems from 3:00 PM to 5:00 PM, as it is
looking at south-facing windows. For view 07, 3:00 PM is the worst time in terms of
discomfort glare during the entire day, but even at 3:00 PM, the glare is only
perceptible.

110

Figure 3-13. DGP glare score results from view 05 to view 07.
Source: Suk et al. 2013, p.118.

The worst glare problems occurred inside the atrium space that is located in the
middle of the building. The atrium is a two-story space, and it has clerestories on its
four orientations. From the early morning to the afternoon, there are direct sun
penetrations through the clerestories, and these are able to cause very serious visual
discomfort. From 8:00 AM to 10:30 AM, all three views in the atrium had intolerable
glare problems.

111

Figure 3-14. DGP glare score results from view 08 to view 10.
Source: Suk et al. 2013, p.119.

As shown above, the glare evaluations from DGP provide a good correlation to the
interior visual conditions of Watt Hall’s third floor through the entire day, but it is
difficult to say whether the evaluation is accurate or not, since the results were not
validated with human subject tests. As explained previously, the glare evaluations
from five different glare indices are not consistent when using either simulated
scenes or captured HDR images. The results brought up lots of questions and
problems in the existing glare analysis methods concerning their inconsistency,
complexity, impracticality, etc. It is easy to simply use one of the existing methods to
analyze or predict glare, but it is very difficult to verify if the analysis results are
accurate, since there is no established, valid baseline for comparison.

Furthermore, it is difficult to understand what causes inconsistent evaluations.
112
Especially for students and lighting/daylighting professionals, it is difficult to look
consider different glare index formulas or Evalglare code to find out potential errors
of their study procedures. Therefore, it is important to use absolute luminance
thresholds and contrast ratios for a more accurate and consistent glare evaluation
metric.

3.2 Investigation of Evalglare
Evalglare algorithms and HDR image capturing procedures were investigated.
Pathological cases, such as closed blinds under overcast sky, fully open blinds under
overcast sky, fully open blinds under clear sky, and direct view of the sun under clear
sky, were set up inside the office on USC Watt Hall’s third floor and tested to see if
the results are as predicted by the algorithms.

3.2.1 Research Method
The research setting was inside the office located in the south corner of USC Watt
Hall’s third floor (Location D in Figure 3-2). The office has two huge glass windows
on the southeast and southwest façades. November 7 and November 18, 2011, were
selected to capture the interior daylight conditions under both clear and overcast sky.
The amount of daylight was controlled using window roller blinds, as shown in
Figure 3-15. Many daylit interior scenes were captured using two different sets of
cameras and lenses: the Nikon D200 with Nikon 18-70mm for a perspective view
and the Nikon Coolpix 4500 with Nikon FC-E8 for an angular fisheye view. Both
113
cameras were mounted at exactly the same location at the height of 1.37 m using a
tripod.

Even though the 18mm lens captures quite a wide angle of view, it is still a much
smaller field of view than an angular fisheye view, and did not capture as much data
as the angular fisheye view. The fisheye image would be more ideal to capture a
human’s field of view than a perspective image since it captures a bigger FOV than
human eye FOV and retains all visual information that can be seen by human eyes.

For each scene, a minimum of eleven Low Dynamic Range (LDR) pictures were
taken with many exposures from five full stop under-exposure to five full stop over-
exposure in order to get a wide range of luminance in the HDR images by capturing
more under/over-exposed LDR images.

114

Figure 3-15. Four different glare scenes in perspective and angular fisheye views.
Source: Suk and Schiler 2012, p.454.

When capturing these images, illuminance and luminance values were also measured
using an illuminance meter (Extech EasyViewTM 30 Light Meter- Model EA30) and
a luminance meter (Cooke Corp. Cal-SPOT 401). The illuminance meter was located
right in front of each lens, to measure vertical illuminance levels at the human eye.
The luminance meter was located right next to each lens, to measure luminance
values. One luminance value and illuminance value were taken for each HDR image.
Externally measured illuminance values were compared to the vertical illuminance
values calculated in Evalglare and also used for the Evalglare ‘-i’ option, which
allows the utilization of externally measured illuminance values instead of internally
calculated values for glare index calculations. Measured luminance values were used
to calibrate the HDR luminance information. A total of four different daylit scenes
were captured from the perspective and angular fields of view (Figure 3-15). Three
115
of the scenes have an identical field of view, with the only differences being the
roller blinds and the sky conditions, while scene #4 shows only an extremely bright
glare source. The reason to capture the same scene in different sky conditions was to
capture scenes from an extremely low possibility of glare to an extremely high
possibility of discomfort glare.

3.2.2 Results
After the first study was performed, the researcher checked how the Evalglare
algorithm retrieves luminance and vertical illuminance information from HDR
images and how each glare index is calculated. One important finding was that
Evalglare does not recognize an angular fisheye view type when the view is captured
by a digital camera, although it will recognize an angular fisheye view simulated
from Radiance. Because the fisheye images taken by the Nikon Coolpix and fisheye
lens have a rectangular canvas with black corners outside the circle in the middle,
Evalglare recognizes the view type as a perspective view. Then, it calculates every
single pixel in the entire image, including the black corners. The areas outside the
circle are not supposed to be considered as parts of a field of view, but Evalglare is
not able to exclude these unwanted portions of the HDR image for discomfort glare
analysis. The black areas certainly affect average luminance and vertical illuminance
values in the Evalglare calculation. This issue explains why there was a warning
message of “vertical illuminance is below 100 lux and out of range for dgp
calculation. Value is set to 100 lux” when the fisheye image was processed in
116
Evalglare.

Possible solutions to this problem include changing the HDR image header
information from a perspective view type to an angular fisheye view type or
cropping the areas outside the circle before combining LDR images into an HDR
image. Another solution would be to write a code to calculate luminance values only
inside the fisheye view. Further investigation is required to see whether or not this
additional procedure can make Evalglare recognize the view type, and to see whether
or not this additional step can eliminate the unwanted pixel information in the
rectangular HDR images.

Four different scenes in HDR format were processed in Evalglare to compare DGI
and DGP scores. The scores are plotted on the two graphs shown in Figure 3-16 and
Figure 3-17. The other glare indices scores are not compared.

117

Figure 3-16. DGI score comparisons between perspective and fisheye views.
Source: Suk and Schiler 2012, p.460.

Figure 3-16 compares DGI scores of angular fisheye and perspective views in four
different daylit conditions. The graph shows that DGI scores gradually increase
under overcast sky when the blinds are more open, and that more direct sunlight
comes into the space under clear sky. The scores also show that there are higher DGI
scores in clear sky than in overcast sky, when the roller blinds are fully open. In the
perspective view, the DGI scores clearly cover the entire range of glare levels, from
imperceptible glare to intolerable glare. However, the DGI scores in the angular
fisheye view show somehow suspicious results in scenes #3 and #4. It is hard to
believe that scenes #3 and #4 have almost the same glare scores (25.1 and 25.3,
respectively). Even though both scenes are evaluated to contain disturbing glare, it
was expected that scene #4 should have a higher glare score than scene #3. A similar
issue was found from the DGP score comparison shown in Figure 3-17. Angular
fisheye and perspective views provide quite similar DGP scores for the first three
scenes. Unlike the DGI scores, the DGP scores of scene #3 indicate imperceptible
118
glare in both views. Scene #4, which has an extremely bright glare source in the
middle of the view, shows quite a big discrepancy between the angular fisheye and
perspective views. As expected, a perspective view from Nikon D200 shows a very
high DGP score as intolerable glare. However, the angular fisheye view from the
Nikon Coolpix shows a very low glare score (imperceptible), even though the score
is still higher than the DGP scores of the other scenes. This result clearly shows that
Evalglare has a problem reading a glare scene with high luminance glare sources,
especially when the image is recorded with an angular fisheye lens. This is despite
the fact that the Evalglare manual allows for the use of externally measured vertical
illuminance values for non-180 degree fisheye images, stating that externally
measured vertical illuminance only affects DGP in the case of non-180 degree
fisheye images when one uses the ‘-i’ option (Wienold 2009).

Vertical illuminance values were measured right before and after taking multiple
LDR pictures with different exposures. Then, the average values were applied for
both angular fisheye and perspective views in Evalglare as externally measured
vertical illuminance values. Figure 3-17 shows that the DGP scores from both view
types are quite close after using the ‘-i’ option. Since the vertical illuminance is the
dominant factor affecting DGP scores, very similar DGP scores for both view types
are only to be expected. Scene #4 with a fisheye view shows a more reasonable DGP
score of 0.502, although it is still different from the DGP score of the perspective
view. Further investigation is required to see why there is still a gap between the two
119
view types, even after using measured illuminance values.

Figure 3-17. DGP score comparisons between perspective and fisheye views with
and without the measured vertical illuminance values.
Source: Suk and Schiler 2012, p.461.

Evalglare has various additional options for users to manipulate manually. There are
more than ten different options outlined in the Evalglare manual, and those described
below might be the most critical ones to test how the manual inputs affect the glare
scores (Wienold 2009). Although the ‘-i’ option affects DGP only, the other options
affect all five glare indices in Evalglare. These options were not tested.
 -s: Smoothing option
Non-glare pixels are surrounded by glare source pixels.
 -y: Peak extraction
Glare source peak luminance value defaults to 50,000 cd/m
2
.
 -b: Luminance threshold factor (constant threshold=100)
Calculates task area’s (which can be defined by option –t or -T)
120
average luminance, then multiplies it by 5 (this default value can be
changed) to find glare sources.

The researcher compared the vertical illuminance values from an illuminance meter
and from Evalglare software to see how accurately Evalglare could calculate
illuminance values based on luminance information in an HDR image. Table 3-1
compares the measured illuminance values to the calculated illuminance values in
Evalglare, along with the DGP scores of each scene before and after applying
measured illuminance values. Since Evalglare provides two illuminance values
(vertical illuminance and direct vertical illuminance), both values are added in the
following table to show the difference between the two. From the four different
daylit scenes, illuminance values were measured from 54 lux to 58,125 lux. The
Evalglare-calculated vertical illuminance values vary depending on view types
(Table 3-1). A huge discrepancy was found between measured and calculated
illuminance values. Furthermore, illuminance values from both view types are much
higher than measured illuminance values, except in one case. The angular fisheye
view for scene #4 shows much lower illuminance values than the actual
measurement.

121
Table 3-1. Comparisons of calculated and measured vertical illuminance values.
Scene View type
DGP score with
calculated vertical
illuminance
DGP score with
measured vertical
illuminance
Calculated vertical
illuminance in
Evalglare
Measured
vertical
illuminance
Scene #1
Perspective
0.17
(Imperceptible)
0.17
(Imperceptible)
387.5 lux
(49.5 lux direct)
53.8 lux
Angular fish
eye
0.16
(Imperceptible)
0.16
(Imperceptible)
119.5 lux
(29.1 lux direct)
54.6 lux
Scene #2
Perspective
0.20
(Imperceptible)
0.22
(Imperceptible)
1655.5 lux
(472.5 lux direct)
269.1 lux
Angular fish
eye
0.22
(Imperceptible)
0.22
(Imperceptible)
631.8 lux
(226.0 lux direct)
366.0 lux
Scene #3
Perspective
0.29
(Imperceptible)
0.34
(Imperceptible)
15207.3 lux
(5956.7 lux direct)
3013.9 lux
Angular fish
eye
0.31
(Imperceptible)
0.34
(Imperceptible)
5186.1 lux
(2654.4 lux direct)
3013.9 lux
Scene #4
Perspective
0.67
(Intolerable)
0.63
(Intolerable)
68274.4 lux
(29356.4 lux direct)
58125.1 lux
Angular fish
eye
0.34
(Imperceptible)
0.50
(Intolerable)
15486.0 lux
(4618.8 lux direct)
58125.1 lux
Source: Suk and Schiler 2012, p.462.

This table also shows that the calculated vertical illuminance values gradually
increased as more daylight entered the room. However, it is difficult to tell if these
calculated illuminance values can be used as a dominant factor of the DGP formula,
even though they are not close to the measured values. Based on this comparison, it
is better to use externally measured illuminance values instead of relying on
automatically calculated illuminance values in Evalglare when capturing HDR
images.
122

The researcher also found that the measured vertical illuminance value in Evalglare’s
detailed report is not shown, even though the ‘-i’ option is used. The Evalglare report
still shows the same calculated vertical illuminance values before and after applying
the ‘-i’ option.

Evalglare is a very powerful glare analysis tool that provides much valuable data,
including average luminance, background luminance, glare source luminance, and
vertical illuminance values. Evalglare can be used for virtual model simulations, thus
helping to avoid potential glare issues prior to building construction. It also uses an
HDR imaging technique that has a great potential, especially for post-occupancy
glare analysis research.

Another value of Evalglare is its ability to evaluate glare issues caused by either
natural or artificial light sources through its use of five different glare indices. Any
disagreement appears to come from the indices themselves, as opposed to the
software program. Evalglare thus may also be a valuable tool for validating the
indices themselves across a range of situations. Various discomfort glare scenes can
be documented and evaluated by using HDRI and Evalglare. The DGP calculation in
Evalglare can be made more reliable through use of externally measured illuminance
values. The proper use of an angular fisheye view in Evalglare analysis is still
unknown, especially when the view is captured by a digital camera. Further study is
123
required on Evalglare options to see how the manual inputs can affect the glare
scores, since it is very important for users to be aware of the impacts of these options.
Even if there are still many questions on the accuracy of Evalglare’s calculations, the
software is certainly a useful and powerful tool for discomfort glare analysis.

3.3 Problems of Existing Glare Analysis Methods and Tools
Through the preliminary studies, four main potential problems of existing glare
analysis methods and tools were discovered:
1. Existing glare indices provide inconsistent glare evaluations.
2. Existing absolute luminance and relative contrast thresholds are not
consistent, and they are not correlated.
3. Neither HDRI captured nor computer generated calculations provided
consistent glare evaluation results between the existing glare indices.
4. Evalglare is powerful software, but it is difficult to troubleshoot mistakes,
even when calculation results are suspicious.

From the reasons above, it has been difficult to incorporate the existing glare
methods and tools into daylighting design practice.

3.4 Field of View Issue
Glare analysis methods have long used photographic methods to capture and analyze
different glare conditions. Before fisheye lenses were introduced, a normal lens with
124
a perspective view was considered good enough to capture glare scenes. However,
researchers have pointed out that a normal perspective view has a missing field of
view that can be captured by a full fisheye lens. Therefore, the fisheye view has been
considered to create a more accurate depiction of glare scenes.

The latest glare metric DGP was developed by using a full fisheye view. Once a glare
scene is captured by a fisheye lens and camera or Radiance simulation, it is possible
to get vertical illuminance from the luminance values of the entire fisheye view.
Then, the calculated vertical illuminance becomes a dominant factor, with other
variables such as glare source position, glare source luminance, and background
luminance. Much as with DGP calculation, Evalglare calculates the other glare
metrics by using luminance and illuminance values from a full fisheye view.

Humans have a 180-degree, forward-facing field of view, but the range of visual
ability is not uniform across that field of view (Figure 3-18). Total human vision can
see 180° horizontally and 130° vertically (Inanici 2004). The binocular visual field is
vertically 130° and horizontally more than 120° when both eyes are focused on a
fixed object (Egan 1983). It is possible that the full fisheye view does not accurately
depict the perspective of a human eye when it comes to a potential glare source
located outside the human eye field of view and inside the boundary of the full
fisheye view. Therefore, it is important to use the actual field of view that can be
seen by human eyes for the purpose of discomfort glare analysis. Inanici (2003) and
Wymelenberg et al. (2012) addressed this concept in their previous glare studies.
125
Further investigation should discover whether or not the difference between full
fisheye and human eye FOVs make a significant difference in glare analysis. This
could help in the development of a new glare analysis method that is more accurate
and more applicable in practice.

Figure 3-18. Human eye’s field of view.
Source: Parker and West 1973, p.631.

3.5 Ideas for a New Daylight Glare Analysis Methodology
The most important role of the new method is to recognize absolute and relative
glare factors before calculating different levels of perceived glare. The absence of
absolute glare factor or relative glare factor is equivalent to no glare. In DGP, vertical
126
illuminance is currently the most important proxy for the absolute glare factor
evaluation method. However, it would be somewhat difficult to recognize absolute
glare factor using only vertical illuminance values, since the total amount of vertical
illuminance at the human eye could be the same from either a small glare source with
extremely high luminance or a very large light source with relatively low luminance.
Since the vertical illuminance level is closely related to the luminance values of the
glare sources, it is possible to use glare source luminance to determine the existence
of the absolute glare factor. Several different research groups have already explored
this idea (Wienold 2005; Dubois 2001; Bulow-Hube 2008; Wymelenberg 2012), and
thus various thresholds have been developed to assess glare source luminance.

The ideal method not only provides a luminance threshold but also provides an
analysis methodology that might help to avoid complicated glare analysis procedures.
Once the new method recognizes glare issues as being the result of absolute glare
factor either by integrating luminance across the entire hemispherical field or at least
across the human eye’s field of view, absolute luminance thresholds can be applied to
determine the presence of glare without considering contrast ratios. If the method
recognizes glare issues as being the result of relative glare factor, the existing glare
methods can calculate the levels of discomfort glare by glare source luminance
ranges and contrast ratios.

Such a glare analysis methodology has yet to be fully developed, but the following
127
study provides a clear step in the development of a new method.

3.5.1 Preliminary Findings—Luminance Ranges
Besides the study on Watt Hall’s third floor, additional studies were performed to
define the luminance ranges for interior daylit conditions. First, a number of interior
glare cases were analyzed using the HDR imaging technique. Using the same set of
equipment, the researcher captured HDR images in Photosphere and analyzed the
images using a MATLAB code to create an image with potential glare mapping and a
logarithmic luminance histogram, as shown in Figure 3-19.

128

Figure 3-19. HDR image, glare source detection, and log luminance histogram

Figure 3-19 shows an interior view inside an office under overcast sky condition
with no visible sun. Roller blinds were fully closed to reduce incoming daylight.
Maximum luminance in this view is 8,515 cd/m
2
—almost four times bigger than the
2,000 cd/m
2
threshold by NUTEK and twice as big as the 4,000 cd/m
2
threshold
claimed by several researchers. It might be difficult to tell whether or not this view
129
has a glare issue solely based on the picture. However, one can probably assume that
there is very little chance of having a glare issue under overcast sky conditions, even
with fully closed roller blinds. This example clearly shows that 2,000 cd/m
2
cannot
guarantee the existence of discomfort glare for a daylit space. As previously
explained, it might be appropriate to set the luminance threshold higher than 4,000
cd/m
2
for daylight glare inside a building. However, human subject testing is
required to validate this preliminary finding.

3.5.2 Hypothetical Findings from the Watt Hall third floor Glare Study
From the Watt Hall third floor glare study, a total of 240 different scenes in PIC
format files were analyzed in Evalglare and a MATLAB code. The MATLAB code
provided two critical values: glare source luminance range and contrast ratio between
background and average glare source luminance values. Evalglare provided glare
evaluations by DGP. After analyzing 240 scenes, each scene was plotted on a single
graph shown to visualize the relations between minimum glare source luminance and
glare ratio based on the glare evaluations by DGP. The plot is shown in Figure 3-20.
The graph shows the following four glare categories by DGP value: imperceptible
glare is represented by black X’s, perceptible glare by green triangles, disturbing
glare by orange diamonds, and intolerable glare by red squares.

130

Figure 3-20. Watt Hall third floor glare analysis results.
Source: Suk et al. 2013, p.121.

The disturbing and intolerable glare scenes are plotted with high glare luminance
values while imperceptible glare scenes are plotted at low glare luminance values
(Figure 3-20). Scenes with a contrast ratio of more than 1:20 are considered to be
imperceptible glare when the glare luminance is relatively low. This example clearly
supports that the existing 1:10 contrast ratio threshold is too low for daylight glare
analysis purposes. This graph also shows that contrast threshold should be correlated
with glare luminance values to see if there is a glare issue or not.

Based on the results of this preliminary study, it is possible to create three different
zones describing absolute glare factor, relative glare factor, and no glare. These zones
are shown as part of the scatterplot in Figure 3-21. The absolute glare factor zone is
shown from the luminance value of 5,500 cd/m
2
upward. This means that visual
131
scenes with luminance levels higher than 5,500 cd/m
2
are considered to be
dominated by the absolute glare factor and assumed to cause discomfort glare no
matter how high or low contrast ratios are. A no-glare zone can be considered to exist
when luminance levels of a scene are lower than 3,000 cd/m
2
. As with the absolute
glare factor, the no-glare zone is not affected by high or low contrast ratios.

From 3,000 cd/m
2
to 5,500 cd/m
2
, there is a grey area that can be defined as relative
glare factor zone. Within this luminance range, glare source luminance should be
correlated with contrast ratios to determine whether or not there is a glare issue.

Based on the results, it is difficult to say whether or not these established values
describing no glare, relative glare factor, and absolute glare factor are accurate, even
though DGP has been validated by many different research groups. Human subject
studies will be used in this study to validate or invalidate these preliminary findings
and see if they support the DGP evaluations. Further investigation is also required to
see why some scenes with lower glare luminance and contrast are evaluated as
having worse glare than scenes with higher luminance and contrast.

132

Figure 3-21. The three zones: no glare, relative glare, and absolute glare.
Source: Suk et al. 2013, p.121.

Another research objective is to define various luminance thresholds that can be
applied to daylit spaces such as a closed office, open office, and exterior space. By
performing human subject tests at various daylit conditions inside and outside a
building, it is expected that the study will attain more accurate luminance and
contrast thresholds for determining the thresholds for absolute glare factor and
relative glare factor. It is likely that using a daylighting test facility (LBNL, EPFL)
with similar glare patterns at successively brighter luminance values with human
subjects would show when the absolute glare factor begins to take over from the
relative glare processes that have already been tested. Through the human subject
tests under various glazing conditions and blind settings, the study will incorporate
the impact of adaptation on these three glare zones into this new methodology. This
approach should be used in further studies to help create a more practical glare
133
evaluation tool for designers, consultants, and students.

3.5.3 Discussion
As discussed in this chapter, it is important to establish clear definitions of AGF and
RGF when developing a more practical analysis method, as the ideal method will
first determine whether a glare scene is dominant with AGF or RGF before making
more thorough evaluations on different levels of discomfort glare. Glare analysis
experts have different opinions on what constitutes AGF and RGF. It is important for
research groups to continue discussions and experiments to develop more simplified
but robust discomfort glare analysis procedures.

One goal of this study is to see the application of the created method in practice,
providing more consistent results for closed offices and open offices. It would be
possible to develop a new chart using logarithmic luminance histograms from HDR
imaging, which could then be incorporated into computer-based software such as
Radiance. Another option would be to update DGP to relate vertical illuminance to
AGF, or replace vertical illuminance with another absolute glare factor. Evalglare
software could then be modified to incorporate this new definition of AGF. Further
investigation is required to determine the best approach.

134
Chapter 4 Research Methods
Interior and Exterior Human Subject Studies

Following the study of existing glare analysis methods and tools, the researcher
developed human subject tests that could provide valuable data for the creation of a
new glare analysis method that can predict and quantify interior and exterior daylight
glare issues. After reviewing a number of precedents (Luckiesh and Guth 1949;
Hopkinson 1957; Ngai and Boyce 2000; Velds 2002; Osterhaus 2005; Linney 2008;
Wymelenberg 2012), two separate human subject research plans were developed
specifically for the interior glare caused by daylight through transparent building
facades and the exterior glare caused by skylight and reflected sunlight from specular
building facades.

The analysis of several different interior glare scenes was used to find hypothetical
absolute luminance thresholds and contrast ratios. These were used as a basis to
analyze data from interior and exterior glare human subject studies. Electrical
lighting sources were completely excluded from both interior and exterior studies.
Therefore, the visual discomfort experienced participants in the research setting is
known to be the sole result of daylight glare, rather than glare from artificial light
sources. The factors of thermal discomfort, dust, and noise were avoided in the
interior research setting, but were not controlled in the exterior research setting.
Future research may use the same proposed protocols and equations to apply to both
natural and artificial lighting, but the current research pursued natural lighting as its
135
first priority.

Human subject studies utilized HDR imaging techniques to capture the visual
information that was experienced by human subjects. A MATLAB code was
specifically developed to analyze the captured HDR images. In addition, two other
strategies were used.
 Use of the same expert group for both interior and exterior glare studies.
 Use of a visual map for subjects to locate glare sources in FOV .

4.1 Human Subject Study: Participants
A total of six subjects (3 male and 3 female) were recruited from the USC School of
Architecture for the interior and exterior human subject studies. The recruiting
requirements for participants were as follows:
 No vision-related illness
 No color blindness
 Aged between twenty and forty years
 English speaking, reading, and writing ability
 Basic typing skill in MS Word and PDF
 Basic knowledge of architecture and architectural engineering

Similar to Hopkinson’s (1957) research, an expert group with a small number of
subjects was chosen instead of a large number of random subjects because of the
136
unique characteristics of daylight glare research. Hopkinson’s study compared the
findings from an experienced group of six people to the findings from fifty random
subjects, and found that the experienced group provided more consistent glare
evaluation data than the random subjects. As the sun and sky conditions constantly
change, it is hard to control daylight glare conditions to be identical for a large
number of subjects. Therefore, it is ideal to have a small number of subjects and have
them all experience many different daylit environments. Testing the same subject
multiple times helps to ensure evaluation accuracy and consistency. Hopkinson’s
(1957) study thereby guided the research method towards an expert group approach.

Recruited applicants were asked to answer personal information regarding their age,
sex, any vision related health issues, work environment, etc. While personal
information was reviewed, applicants were asked to take a color blindness test using
the Ishihara template (24-plate version). This procedure was critical to exclude those
with vision related illness and color blindness from the study. The recruited subjects
passed this test. The recruited subjects were from various races, with two Caucasians
and four Asians, although diversity of race was not a particular goal of the selection
process. The very detailed study procedure was explained to the recruited subjects,
and they agreed to participate in both the interior and exterior glare studies.

The recruited subjects are all future architectural professionals with various
specialties who are students at USC’s school of architecture. Since they are familiar
137
with occupant thermal and visual comfort issues in architecture, they well
understood the study purpose and test procedures. The subjective data from these
future professionals is expected to be more accurate and consistent than that from
random subjects, who have no background knowledge of architecture. It was
anticipated that their experiences from being tested under various daylit conditions
have helped them to become daylight glare evaluation experts who can accurately
evaluate different daylit conditions inside and outside of buildings. An additional
benefit of using future architects for the study was the study’s provision of an
opportunity to personally experience daylight glare issue, which should help to refine
their future architectural practice.

As the same expert group participated in both interior and exterior glare studies, it
was expected that visual evaluations of the various lighting conditions inside and
outside the building would be consistent. The interior and exterior human subject
study plans were reviewed and approved by the USC Institutional Review Board
(IRB) on January 24, 2013, before the first test was performed.

4.2 Human Subject Study: Equipment
Different types of equipment were used to perform the interior and exterior glare
studies. The same equipment was used for the human study as that used for the
preliminary studies described in Chapter 3. The equipment was carefully calibrated
and normalized prior to the main study. After a glare scene was captured using
138
various exposures by a Nikon Coolpix 4500 camera and angular fisheye lens.
Photosphere was used to create an HDR image out of the captured jpeg images. The
luminance values on the captured HDR images were also calibrated with field
measured luminance values by a Cooke luminance meter. Illuminance sensors and
data loggers recorded the vertical illuminance values arriving into the subject’s eyes
and horizontal illuminance at the task height every thirty seconds. Temperature and
relative humidity were also recorded at the same interval.

The following equipment was used:
 Nikon Coolpix 4500
 Nikon FC-E8 lens (Equidistant projection and 183 degree field of view)
 Tripod
 Luminance meter: Cooke cal-SPOT 401
 (4) Li-Cor LI-201SZ photometric sensors
 (4) UTA/HOBO adapters
 (4) HOBO temperature/relative humidity sensors
 Photosphere

139

Figure 4-1. The camera and fisheye lens mounted on a tripod, the illuminance
sensors mounted on a monitor, the illuminance sensors mounted on a camera, and a
luminance meter.
Source: The Cooke corp. 2008, p.1

The interior studies used a desktop computer and a keyboard in order to allow the
subjects to perform typing tasks. The exterior glare study used an iPad for the
reading task. Desktop monitor and iPad screen brightness levels were set to be
consistent throughout the entire study.
140
4.3 Human Subject Study: Interior Glare
The human subject study to assess interior glare issues was performed inside a closed
office space. One of the faculty offices on the third floor of Watt Hall at the
University of Southern California was selected for the interior glare tests. There is no
exterior visual obstruction that is closely located to the office. The room is a corner
office with two clear glass windows facing southwest and southeast. Each window
has two adjustable blinds: a venetian blind and a roller blind. The corner office
condition was selected to allow more natural light inside the space (to be more
exposed to the outside) and to avoid the severe contrast issues that can occur in a
closed room with small aperture windows. The allowance of adequate daylight also
helped to avoid the necessity of electrical lighting usage. It was expected that this
research setting would help the researcher to discover the absolute luminance
thresholds that cause discomfort glare. Another advantage of using the corner office
condition is that it provides more opportunity to experience potential glare sources.
Since the office has front and side windows, it can have potential glare sources from
different directions.

The research setting was used for the subjects and HDR photography, as shown in
Figure 4-2. The room measured 11'-3'' high by 9'-6'' wide by 11'-4'' long. A desk was
located adjacent to the windows facing southwest and southeast; a desktop monitor
was placed on top of the desk, in front of a southwest-facing window. A total of four
illuminance sensors and data loggers were installed as shown in Figure 4-1: one
141
behind the monitor facing the window and one above the camera facing the window.
There was also a sensor on either edge of the desk, facing up. Temperature and
relative humidity sensors were installed at the same locations. Data loggers were
provided thirty-second recording intervals in which to record data from the sensors.

With HDR Photography With Human Subject
Figure 4-2. Interior glare study research setting: 1) Camera; 2, 3, 4, 5) Li-Cor sensors
for vertical illuminance; 4, 5) Li-Cor sensors for horizontal illuminance; 2, 3, 4, 5)
HOBO sensors; 6) Tripod; 7) Cooke luminance meter.

A total of 53 tests were performed inside the office from February 18, 2013, to June
17, 2013. Most of the tests were performed under clear sky conditions, but each
subject performed at least one test under an overcast sky condition. With the various
sky conditions, subjects experienced various levels of discomfort glare, even when
the blinds were fully opened.

142
4.4 Lighting Conditions and Tasks
The office has a fluorescent light fixture hanging from the ceiling, but it was turned
off during the study. Three different lighting conditions were given to each subject.
The first lighting condition was a fully opened blind condition on both front and side
windows (Figure 4-3). Direct sunlight threw sunlight onto the desk through the floor-
to-ceiling windows and cast big shadows. The subjects were asked to perform three
task conditions, including “no-task,” under the fully open blind condition.

No Task Typing Task Writing Task
Figure 4-3. Three tasks under fully open blind setting.

The second lighting condition was created with roller blinds only (Figure 4-4).
Venetian blinds are fully open, while subjects were able to adjust the roller blinds to
suit their comfort. Subjects were able to separately adjust front and side roller blinds
as they preferred. Figure 4-4 shows the setting with the side roller blind fully closed
and the front roller blind closed three-quarters of the way. This roller blind condition
was set by one of the subjects. Direct sunlight was only allowed through the portion
of the front window that was not blocked by the roller blind. It is possible to see a
143
silhouette of the sun and outdoor conditions through the roller blinds.

No Task Typing Task Writing Task
Figure 4-4. Three tasks under the roller blind setting.

The third lighting condition was created with venetian blinds only (Figure 4-5). The
roller blinds were fully open for this condition, and the three different task activities
were performed using only adjustable venetian blinds. Again, subjects were able to
control how they preferred the blinds to be placed to create a more visually
comfortable indoor space. Unlike with the roller blinds, they were able to adjust
blind angles as they preferred. The blind angles allowed the subjects to introduce
more natural light into the room if they wanted higher light levels and to block
incoming natural light if they wanted lower light levels. Figure 4-5 shows that the
side venetian blind was fully closed, while the front blind was closed for the top
three quarters of the window. The blind angles were set to completely block natural
light from the exterior.

144
No Task Typing Task Writing Task
Figure 4-5. Three tasks under venetian blind setting.

The order of exposure to the three lighting conditions was switched in order to avoid
the same chronological lighting condition pattern.

4.4.1 Detailed Interior Glare Research Procedure
Each subject performed three different tasks for each of the three lighting conditions.
Thus, each subject experienced a total of nine different combinations of lighting
condition and task type within a single test. Each test takes around one hour to
complete. The detailed interior glare study procedure is as follows:

1. Explain detailed study procedure to a participant
2. 1
st
Condition: shading devices are fully open
a) Task 1: No task (General impression)
b) Fill in survey 1 and HDR photography
c) Task 2: Reading and typing task on computer
d) Fill in survey 2 and HDR photography
145
e) Task 3: Reading and writing task on paper
f) Fill in survey 3 and HDR photography
3. Adjusting roller blinds and three minute break
4. 2
nd
Condition: Roller blinds as controlled by the subject
a) Repeat the procedure from the 1st condition
5. Adjusting venetian blinds and three minute break
6. 3
rd
Condition: Venetian blinds as controlled by the subject
a) Repeat the procedure from the 1st condition

The order of 1, 3/4, and 5/6 was randomly rearranged for each participant.

A single test chamber was used to perform the human subject tests and the visual
condition documentation. HDR photography and luminance measurements were
conducted right before and after the subject completed each task (Figure 4-6). Once
subjects became more familiar with the test procedure, HDR photography and
human subject tests were performed without interruption.

146

Figure 4-6. Diagram of interior study procedure for each subject.

4.4.2 Visual Maps and Questionnaires
Conventional human subject tests usually ask human subjects verbal or written
questions to collect subjective evaluations while documenting their fields of view
(FOVs). This conventional approach is also effective to understand the existence of
glare issues in the FOV , but it cannot provide accurate information about glare source
locations and sizes within the FOV .

In order to overcome this limitation, line drawings of the subject’s FOV were created,
and subjects were asked to indicate the areas causing their visual discomfort in their
FOVs. This line drawing is called a visual map. The visual maps contain lines of core
components of the space, such as the walls and window frames, as well as the desk,
monitor, keyboard, etc. The visual maps were created only with black lines on a
147
white background, so the subjects were not biased by the visual information
contained in the photograph. The visual map was also intended to confirm whether
subjects could visually indicate annoying areas in their FOVs when they evaluate a
certain level of discomfort glare. Subjects were able to draw circles or lines to
indicate the sources of any visual discomfort. In that way, the subjects indicated their
levels of visual discomfort and visually illustrated the glare source information in the
FOV. By comparing these two responses from the same subject, it was possible to
check the consistency of subjective evaluations. The visual map has a 180-degree,
angular fisheye view, which is bigger somewhat than the human eye’s FOV. It was
expected that participants might indicate glare sources inside and outside the human
eye’s FOV, depending on glare source locations. By checking the glare source
locations marked in visual maps, it will be possible to figure out whether or not the
human eye’s FOV would be a more appropriate visual map for discomfort glare
analysis.

Three different visual maps were created for three different tasks: 1) no-task; 2)
reading and typing texts on a computer monitor; and 3) reading and writing texts on
paper sitting on a desk. The FOV also changes depending on the task type, as shown
in Figure 4-7. The left image in Figure 4-7 shows a FOV when the subjects sit on a
chair and looks straight ahead toward the front window. During this task, subjects did
not perform any activity but were asked to provide their general impression of the
research setting. After subjects gained a general impression of the space for three
148
minutes, a survey form with the visual map was presented on the desk. Subjects were
able to hold the form up to compare it to their FOV or keep it on the desk as they
preferred.

The middle image was created for the reading and typing on a computer task.
Compared to the first visual map, the middle image has a slightly lower view angle
to locate monitor and keyboard in the middle of the FOV. This view angle was
measured for each subject prior to testing. The visual map was shown on the
computer monitor after the subject had completed the typing task for three minutes,
so the subjects could maintain their FOVs during the task and survey.

The right image shows the visual map for the paper-based reading and writing task.
Subjects performed the reading and writing task on paper for three minutes. After the
task was complete, a survey form with the visual map was presented to subjects at
the same task area. Again, subjects did not need to change their FOVs during the
writing task. This was deliberate, as having different FOVs for task and survey could
cause inaccurate and inconsistent evaluations.

149

No Task Typing Task Writing Task
Figure 4-7. Visual map of FOV looking straight ahead for each of the three tasks.

A seven-point Likert scale was used for the written questionnaires (Figure 4-8). In
the written questionnaires, subjects evaluated general light levels, visual satisfaction,
visual comfort, and glare levels inside the research setting. As the sun and sky
conditions are exceedingly changeable, the written questionnaires were designed to
be simple and straightforward. Thus, subjects could answer the questions quickly,
while the sun and sky conditions stayed roughly the same. Figure 4-8 is an example
of the questionnaire provided for the no-task condition. The questionnaires for the
typing and writing tasks are slightly different from this example, but they ask for the
same kind of information about light level, visual satisfaction, visual comfort, and
glare levels. The questions are all related to visual comfort and discomfort issues.
The research setting was designed to minimize thermal and acoustical discomfort
issues. These issues were recorded, but subjects were not asked about them. It was
expected that evaluation consistency could be secured by asking each subject
multiple questions about visual comfort.
150

Figure 4-8. Interior glare questionnaire example.

4.5 Human Subject Study: Exterior Glare
As explained in Chapter 1, the issue of exterior glare has not been thoroughly
investigated for scientific research. Therefore, it was difficult to find previous
research that would indicate how to design a research method for exterior glare
issues. Instead of developing a different research method, therefore, the researcher
slightly adjusted the interior glare research method to apply it to exterior glare
conditions. It was understood that this method might not be the best way to perform
a human subject study for exterior glare issues, but it was expected that any
difficulties or problems from the exterior glare study would help to address the
exterior glare problem.
151

An outdoor patio space located at the University of Southern California was chosen
for the exterior glare study. The outdoor patio is located between two five story
buildings—one on the north side and one on the south. The building on the south
side blocks the sunlight from hitting the outdoor patio, and the building on the north
side reflects sunlight from its glazing facades. It was expected that an extreme
thermal discomfort issue would be avoided in this research setting, since subjects
received only reflected sunlight from the north while receiving no direct sunlight
from the south. Furthermore, this research setting isolated reflected sunlight as the
only potential glare source that subjects could experience during tests.

Twelve exterior tests were performed from April 10, 2013, to May 10, 2013. The
same expert group who joined the interior glare study participated. During the test
period, subjects experienced different daylight conditions. Rainy days were avoided.
Unlike with the interior glare studies, there were several difficulties to overcome:
subjects were able to experience reflected sunlight from the glass facades only for a
limited time as the sun moved; reflected sunlight hit the research setting at different
times of day as the sun’s altitude changed; and by May 10, the summer sun’s altitude
was too high for subjects to see reflected sunlight. Additional difficulties included
lack of privacy and uncomfortable ambient temperature, as well as an excess of noise,
wind, and pollutants. It was not possible to control these factors, which were likely to
affect the subjective evaluations to some extent.
152

4.5.1 Lighting Conditions and Tasks
There were two lighting conditions for the exterior glare study: reflected sunlight in
the FOV and no reflected sunlight in the FOV. These two conditions were achieved
by performing tests at different times of day, as reflected sunlight in the patio occurs
during a limited time every day. Subjects were asked to perform three different tasks
under each lighting condition: no-task, a reading iPad task, and a reading paper book
task (Figure 4-9). The three photographs in Figure 4-9 contain two potential glare
sources: one from the building façade and the other one from the table. As the sun
hits the building façade, its specular windows reflect the sunlight toward the research
setting. Once the reflected sunlight arrives at the table, the specular table reflects it
toward the subjects. The iPad’s screen brightness was set to maximum (410 cd/m
2
for
white) to minimize a veiling reflection in the research setting. The setting was not
altered in any way during the study.

No Task Reading iPad Reading Paper book
Figure 4-9. Reflected sunlight in FOV .

153
4.5.2 Detailed Exterior Glare Research Procedure
The detailed exterior glare study procedure proceeded as follows:
1. Explain detailed study procedure to the subject
2. First condition: Reflected sunlight is in the field of view
a) Task 1: General impression without any task
b) Fill in survey question #1
c) HDR photography and luminance measurement
d) Task 2: Reading text on paper for two minutes
e) Fill in survey question #2
f) HDR photography and luminance measurement
g) Task 3: Reading texts on iPad for two minutes.
h) Fill in survey question #3
i) HDR photography and luminance measurement
3. Second condition: Reflected sunlight is not in the field of view
a) Repeat the procedure for the 1
st
condition, from a) to i)
4. Repeat the procedure of the first and second conditions at different locations on
the patio.

Tables and chairs were set up to ensure that subjects were facing toward the reflected
sunlight from the north building’s façade. The subjects and camera were located at
the same location in order to provide an identical view angle and direction. Camera
heights were also set to individual eye height in a seated position. HDR photography
154
and luminance measurements were made right before and after each task and survey
were performed by a subject. The measured luminance values were used to create
absolute luminance calibration in the HDR images. The three tasks were repeated
under two different lighting conditions (Figure 4-10). The order of lighting
conditions was randomly switched to avoid a same lighting condition transition
pattern.

Figure 4-10. Diagram of exterior study procedure for each subject.

Three illuminance sensors were installed; two on the middle of the table (one facing
up and the other one facing horizontal towards the building façade) and one on the
top of the camera facing towards the building facade. Temperature and relative
humidity sensors were also installed at the same locations. Data loggers recorded the
data from the sensors every thirty seconds.
155
4.5.3 Visual Maps and Questionnaires
Visual maps were utilized for the exterior glare study. The visual maps only
contained depictions of the core components of the outdoor space, such as buildings,
windows, tables, chairs, and trees. Two different visual maps were created for the
exterior glare study: FOV looking straight ahead and FOV looking down at the table
(Figure 4-11). The left image in Figure 4-11 is the visual map for no-task, which has
an FOV looking straight toward the glass façade of the building. The right image is
the visual map for the reading the iPad and paper book tasks, which has an FOV
looking down on the table. As shown in both visual maps, most of the FOV is
occupied by the table and the building. The same visual maps were utilized for both
lighting conditions.

No Task Reading iPad or Paper book
Figure 4-11. Visual maps for exterior glare study.

Figure 4-12 shows the written questionnaire for no-task in the exterior glare test. The
156
questionnaires for the reading iPad task and the reading a paper book task are
slightly different from this, but they ask the same types of information about general
light levels, visual satisfaction, visual comfort, and glare levels at the outdoor open
space. A seven-point Likert scale was used for everything except the glare problem
question. The glare problem question used a bar scale with the four glare categories.

Figure 4-12. Exterior glare questionnaire example.

4.6 MATLAB Code Development
Prior to performing the human subject study, a MATLAB code was developed to
analyze the HDR images. To develop and validate the code, a number of HDR
images were captured inside and outside of buildings in Los Angeles and a
157
luminance meter was used to calibrate luminance values in the HDR images. The
code can analyze captured HDR images using a camera and simulated PIC files from
Radiance software. Based on the R (red), G (green), and B (blue) values, the code
calculates the luminance value of each pixel of an HDR image to create a logarithmic
luminance histogram in either full fisheye FOV or the human eye’s FOV. When a
histogram is created, a processed image is also created to identify potential glare
sources. Figure 4-13 provides a screenshot of the developed MATLAB code.

Figure 4-13. A screenshot of the MATLAB code.

The histogram glare method uses a combination of a luminance ratio (3:1, peak to
mean) and a spike in the histogram to determine whether glare was present (Schiler
2000). This method was appropriate for interior glare cases with electrical lighting
sources. However, it was not proper for the visual conditions with very wide
158
luminance ranges, such as interior space with windows and exterior daylit conditions.
It is hypothesized that the luminance ratio will be larger than 3:1, since it is now
possible to capture wider luminance ranges with HDR techniques.

By default, the code detects glare sources by multiplying the mean background
luminance value by 5. In other words, any pixel with higher than 5 times the mean
background luminance is detected as a potential glare source. The value of 5 is the
same used in the Evalglare default setting. However, it is still possible to adjust this
multiplier value to more accurately detect glare sources in an HDR image. Also, it is
possible to use an absolute luminance threshold for glare source detection instead of
using the multiplier, or use the threshold in conjunction with the multiplier. The
researcher expects to find an absolute luminance threshold from the human subject
study.

As previously explained, a logarithmic luminance histogram can be plotted from an
HDR image. Because human eyes can see luminance ranges from one-millionth to
one million, the x-axis of a plotted luminance histogram is set to this range. The y-
axis plots the frequency of a specific luminance value. This program uses the
existing MATLB function data HDRREAD, which reads HDR images in Radiance
RGBE format. Created by Mathworks, Inc., in 2008, this function helps the code to
read the Red, Green, Blue, and Exponent (RGBE) data of an HDR image. By using
the HDRREAD function, the new code calculates each pixel’s luminance value by
159
using the formula below.

Luminance = 179*(0.2127*R + 0.7151*G + 0.0722*B) / exposure

The primaries for the red, green, and blue values were taken from Inanici’s paper
(Inanici 2005). Inanici created this formula specifically for daylight. For the other
lamp sources or Radiance simulations, these primaries should be adjusted
accordingly. The exposure value is retrieved from the HDR image header.

After the code calculates luminance values for each pixel in the HDR image, it
calculates the following values: minimum luminance, maximum luminance (same as
glare source maximum luminance), background mean luminance, glare source
minimum luminance, glare source mean luminance, glare ratio, and glare size in
percentage. These calculated values are then plotted on a logarithmic luminance
histogram, shown in Figure 4-14).

160

Figure 4-14. Logarithmic luminance histogram created from MATLAB code.

When the code calculates luminance values from the HDR image, it is possible to
define one of two FOVs: that of a full fisheye and that of a human eye. These two
FOVs are shown in Figure 4-15. Potential glare sources are identified in red in both
images. The left image shows the red boundary of full fisheye FOV, while the right
image shows the red boundary of the human eye’s FOV . Only the areas inside the red
boundaries are analyzed, while the pixels outside the boundary are ignored. Both
images show two potential glare sources: one through the front window and the other
one on the desk. It is possible to see that the glare source from the window has
different sizes between the two FOVs. These processed images in both FOVs were
compared to the subjects’ marked visual maps.

161
Angular Fisheye FOV Human Eye FOV
Figure 4-15. Images created from MATLAB code: Full fisheye FOV vs. human eye
FOV .

While the code was developed, interior daylight glare scenes were captured at
various daylit indoor and outdoor spaces. The captured HDR images were processed
in the code to read luminance values, glare ratios, and luminance ranges of
backgrounds and glare sources. The measured luminance values were used to
calibrate luminance values captured in the HDR images. Figure 4-16 shows an
interior glare scene that has the visible sun in the middle of the view. After absolute
calibration, the maximum luminance of this image is 44,165 cd/m
2
, and the glare
ratio is 9.91. The potential glare source luminance range is from 9,980 cd/m
2
to
44,165 cd/m
2
, while glare size is 3.01% of the entire FOV .

Figure 4-1

Exterior g
ratios and
buildings
building th
wall of the
building g
the buildi
direction.
16,412 cd
source is 2

6. Interior g
glare scenes
luminance r
and also to
hat reflects
e building th
geometry, it
ing while s
The code an
d/m
2
to 258,
2.7% of the v
glare exampl
were also c
ranges. Faca
o pedestrians
strong sunli
hat reflects th
sends multip
simultaneous
nalysis inter
000 cd/m
2
,
view.
e: Visible su
captured an
ades throw s
s and drive
ight to its su
he sunlight to
ple reflectio
sly causing
rprets the gla
with a glar
un in FOV .
d processed
sunlight refle
rs. Figure 4
urroundings.
o multiple di
ons to a buil
very serio
are source lu
re ratio of 1
d in the cod
ections to th
4-17 shows
. The image
irections. Du
ding located
ous glare in
uminance ra
14.8. The si
de to read g
heir neighbor
a glass faç
e shows a so
ue to the uni
d to the wes
n the south
ange to be fr
ize of the g
162

glare
ring
çade
outh
ique
st of
east
from
glare
163

Figure 4-17. Exterior glare example: Glare sources with extreme brightness.

As shown in these two examples, the code successfully provides many critical values,
such as glare source luminance range, glare source size, and glare ratio from an HDR
image, but it is still difficult to determine whether these images are really discomfort
glare scenes just by looking at these values or a logarithmic luminance histogram.
The existence of discomfort glare can be determined by the existing glare indices,
but they do not explain the main factor causing discomfort glare. A new
methodology using absolute glare factor (AGF) and relative glare factor (RGF) does
illuminate the main factor, however.

Chapters 5 and 6 analyze the captured HDR images and the subjective glare
evaluation data collected from the human subject study. In addition, Chapter 6
explains the development procedure of a new glare analysis method using AGF and
RGF.
164

Chapter 5 Validation Study
Human Subject Data Analysis Using Existing Glare Indices

The collected human subject study data was analyzed using the five existing glare
indices (DGP, DGI, VCP, UGR, and CGI) prior to being analyzed for the
development of a new analysis method. Interior glare scenes captured in HDR image
format were processed in Evalglare to see the accuracy of each glare index’s
analysis of various daylit conditions and to discover what causes inaccurate
evaluations. Exterior glare scenes were not analyzed by the existing glare indices,
since it was well understood that these glare indices were originally developed to
evaluate interior glare issues, not exterior glare issues. As the existing indices have
an inconsistent evaluation issue on one set of data in Chapter 3, the expanded set was
evaluated to see whether or not they would also provide inaccurate evaluations.
Based on the analysis results and findings, several suggestions were made to help
improve each glare index.

5.1 Analysis of Interior Glare Human Subject Study Data
More than 450 glare scenes captured from this interior glare study were analyzed in
Evalglare code to calculate glare evaluations using the existing glare indices. The
calculated glare scores in Evalglare were transferred to the perceived glare degree
categories based on the glare score ranges that were developed for each glare index.
165
Then, the glare evaluation results from each glare index were compared to the
collected human subject evaluation data to see whether they matched each other or
not. All glare scenes were analyzed in Evalglare twice with two different fields of
view (FOV): full fisheye FOV and the human eye’s FOV. Evalglare indicates
potential glare sources in different colors within a full fisheye FOV while calculating
the glare scores of DGP, DGI, VCP, UGR, and CGI. Because Chapter 3 found that
DGP shows suspiciously low glare scores for all glare scenes with internally
calculated vertical illuminance, this validation study did not rely on the internally
calculated vertical illuminance values. Instead, it applied externally measured
illuminance values from the human subject study to the DGP formula in Evalglare in
order to determine whether or not the internally calculated vertical illuminance was
the source of the inconsistency.

Subjective glare evaluation data was compared to the glare scores calculated by the
existing glare indices to check what glare indices match best to the subjective
evaluation. Three comparison examples are shown in Figure 5-1 for a no-task glare
scene under the sun control conditions of fully open blinds, roller blinds, and
venetian blinds. The glare indices matching subjective glare evaluations are
highlighted in red. A fully open blind scene is correctly evaluated by DGI, while the
other indices overestimate the scene as intolerable glare. For the roller blind scene,
DGI and VCP match to subjective evaluation, as they evaluate perceptible glare.
None of the indices evaluates the venetian blind scene as perceptible glare, and all of
166
them overestimate the glare to be disturbing or intolerable compared with the
subjective evaluation.

Fully open

Subjective evaluation: Disturbing glare

 DGP = 1.000 (Intolerable)
 DGI = 24.807 (Disturbing)
 UGR = 30.464 (Intolerable)
 VCP = 20.28 (Intolerable)
 CGI = 34.58 (Intolerable)

Roller blinds

Subjective evaluation: Perceptible glare

 DGP = 0.420 (Disturbing)
 DGI = 18.792 (Perceptible)
 UGR = 22.373 (Disturbing)
 VCP = 76.564 (Perceptible)
 CGI = 24.823 (Disturbing)

Venetian blinds

Subjective evaluation: Perceptible glare

 DGP = 0.447 (Disturbing)
 DGI = 25.694 (Disturbing)
 UGR = 33.408 (Intolerable)
 VCP = 17.562 (Intolerable)
 CGI = 33.506 (Intolerable)

Figure 5-1. No-task condition scenes processed in Evalglare, compared to
participants’ subjective evaluation.

Glare evaluations between human subjects and the existing glare indices were
compared for typing task glare scenes, and three examples were chosen for fully
open blinds, roller blinds, and venetian blinds. The images for these examples are
shown in Figure 5-2. The fully open blind scene shows that all glare indices except
167
DGI evaluate the scene to have intolerable glare, which matches what the subject
experienced. However, none of the indices matches to the subjective evaluations for
the roller blind scene. Also, only DGP matches to the subjective evaluation for the
venetian blind scene in its evaluation of the scene to have perceptible glare.

Fully open

Subjective evaluation: Intolerable glare

 DGP = 1.0 (Intolerable)
 DGI = 24.511 (Disturbing)
 UGR = 30.336 (Intolerable)
 VCP = 25.443 (Intolerable)
 CGI = 34.218 (Intolerable)

Roller blinds

Subjective evaluation: Imperceptible glare

 DGP = 0.351 (Perceptible)
 DGI = 18.298 (Perceptible)
 UGR = 21.815 (Perceptible)
 VCP = 79.419 (Perceptible)
 CGI = 24.278 (Disturbing)

Venetian blinds

Subjective evaluation: Perceptible glare

 DGP = 0.369 (Perceptible)
 DGI = 25.788 (Disturbing)
 UGR = 33.435 (Intolerable)
 VCP = 16.798 (Intolerable)
 CGI = 33.664 (Intolerable)

Figure 5-2. Typing task scenes processed in Evalglare.

Three examples were chosen for the writing task condition glare scenes under fully
open blinds, roller blinds, and venetian blinds. The images from these examples are
168
shown in Figure 5-3. The fully open blind scene was correctly evaluated by UGR
only, as it evaluates disturbing glare. DGP, VCP, and CGI evaluated the scene as
intolerable glare, while DGI evaluated it as perceptible glare. The roller blind scene
was correctly evaluated by DGP only. The venetian blind scene was not correctly
evaluated by the indices—they all overestimated the scene as disturbing or
intolerable glare.

Fully open

Subjective evaluation: Disturbing glare

 DGP = 0.70 (Intolerable)
 DGI = 21.845 (Perceptible)
 UGR = 26.447 (Disturbing)
 VCP = 35.570 (Intolerable)
 CGI = 30.152 (Intolerable)
Roller blinds

Subjective evaluation: Imperceptible glare

 DGP = 0.218 (Imperceptible)
 DGI = 18.522 (Perceptible)
 UGR = 20.736 (Perceptible)
 VCP = 78.775 (Perceptible)
 CGI = 23.057 (Disturbing)

Venetian blinds

Subjective evaluation: Imperceptible glare

 DGP = 0.419 (Disturbing)
 DGI = 24.668 (Disturbing)
 UGR = 31.990 (Intolerable)
 VCP = 22.991 (Intolerable)
 CGI = 32.476 (Intolerable)

Figure 5-3. Writing task scenes processed in Evalglare.

169
As shown in these examples, only one or none of the five indices correctly matches
to the subject’s evaluation for each scene. This evaluation comparison study supports
the findings that the five glare indices have vast inconsistency issues. Furthermore, it
indicates that the existing glare indices have significant inaccuracies in their
evaluations of glare. After calculating all 450 glare scenes in both full fisheye FOV
and the human eye FOV, the evaluation accuracy and inaccuracy rates of each glare
index were compared. The graphic, bar chart comparison of the data is shown in
Figure 5-4.

170
VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
34.0
66.0
17.3
82.7
55.3
44.7
39.1
60.9
13.1
86.9
T otal FOV in E valglare_Incor (Incorr ect)/Corr (Cor rect)

VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
47.3
52.7
21.3
78.7
55.6
44.4
47.3
52.7
17.6
82.4
FOV seen by both eyes in E valglare_Incor (Incorr ect)/Corr (Cor rect)

Figure 5-4. All glare scenes: Full fisheye FOV (top) and the human eye FOV
(bottom).

171
For the purposes of this study, the subject data is considered to be accurate. The
accuracy rate indicated by the chart shows the extent to which the existing glare
indices match to the subjective evaluation on a same glare scene. In each bar chart,
the dark grey bar shows the percentage of inaccurately evaluated scenes that do not
match to subjective evaluations, while the light grey bar shows the percentage of
accurately evaluated scenes that do match to subjective evaluations. When glare
scenes were analyzed with a full fisheye FOV, DGP shows 55.3% accuracy, which is
the highest among the five. CGI and UGR show the lowest accuracy, as the
percentages for correct evaluation are only 13.1% and 17.3%. DGI and VCP show
better accuracy than CGI and UGR, but still show only 39.1% and 34.0% accuracy.
The human eye’s FOV slightly increased the percentage of accurately evaluated glare
scenes (Figure 5-4). The biggest increase in accuracy was shown by VCP, as it went
up from 34.0% to 47.3%. The second biggest increase was DGI, which showed 39.1%
with full fisheye and 47.3% with the human eye FOV. DGP still shows the highest
percentage of the correctly evaluated glare scenes, even though there is almost no
difference between the two FOVs. It is assumed that DGP is not affected by a
specific FOV, since it puts more weight on vertical illuminance values than
luminance values. CGI and UGR show increased accuracy with the human eye FOV,
but they still show very low accuracy. None of the existing indices exceeded 56%
evaluation accuracy.

5.2 Analysis based on Glare Category (Sensation)
172
Interior glare scenes were grouped into three different glare categories: imperceptible,
perceptible, and disturbing/intolerable. The researches intended to find out whether
or not different glare indices could more accurately evaluate a glare scene with
higher or lower glare issues. Therefore, glare evaluations from each index were
compared to human subject evaluation data for each glare category group.

5.2.1 Imperceptible Glare Scenes Only
Imperceptible glare scenes were analyzed and the results of calculated glare
evaluations and subjective data were compared, as shown in the graphical
representations of Figure 5-5. The subjects determined that a total of 202 interior
glare scenes had imperceptible glare. DGP evaluation accuracy is 89.6% with the full
fisheye FOV, while CGI and UGR recorded 0% evaluation accuracy for
imperceptible glare scenes. This result indicates that CGI and UGR might have
overestimated glare levels for all imperceptible glare scenes captured from the
interior glare study. DGI and VCP show better accuracy percentages than CGI and
UGR, but they are only 27.7% and 37.1%.

173
VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
37.1
62.9
0
100
89.6
10.4
27.7
72.3
0
100
T otal FOV in E valglare_Incor (Incorr ect)/Corr (Cor rect)

VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
59.4
40.6
0
100
89.6
10.4
58.9
41.1
0
100
FOV seen by both eyes in E valglare_Incor (Incorr ect)/Corr (Cor rect)

Figure 5-5. Imperceptible glare scenes only: Full fisheye FOV (top) and human eye
FOV (bottom).

174
The same imperceptible glare scenes were analyzed using the Guth position index in
Evalglare for the human eye’s FOV. The graphical representation of this analysis is
shown in Figure 5-5. CGI and UGR still show 0% evaluation accuracy for all the
imperceptible glare scenes. This means that CGI and UGR most likely analyze the
scene to have perceptible, disturbing, or intolerable glare, even when there is no
glare in a scene. DGP still shows the same high accuracy percentage of 89.1%. FOV
does not affect DGP’s glare evaluation accuracy. Unlike CGI, UGR, and DGP, DGI
and VCP have better improvement in accuracy percentages when calculated with the
Guth position index for the human eye’s FOV. This improved accuracy was also
found in the comparisons shown in Figure 5-4. DGI’s evaluation accuracy increased
from 27.7% to 58.9%, while VCP’s accuracy increased from 37.1% to 59.4%.

5.2.2 Perceptible Glare Scenes Only
After imperceptible glare scenes were analyzed, Evalglare was used to analyze
perceptible glare scenes with the five glare indices. The results were compared to the
results of the human subject evaluation data, and are shown in Figure 5-6. With full
fisheye FOV, CGI, DGP, UGR, and VCP show pretty low evaluation accuracy
percentages, while DGI shows the best accuracy at 64.8%. CGI shows only 2.3%,
which is the lowest evaluation accuracy, while DGP, UGR, and VCP show an
accuracy range from 13.3% to 21.9%. Even though DGP showed the highest
accuracy for imperceptible glare scenes, it still cannot accurately evaluate perceptible
glare scenes.
175

When the same scenes were analyzed in the human eye’s FOV, accuracy percentages
were slightly changed in all five indices. The human eye’s FOV helps to increase
evaluation accuracy for CGI, DGP, UGR, and VCP. CGI and DGP show the same
accuracy level, at 14.8%. UGR shows 29.7% accuracy, and VCP shows 25.0%
accuracy. Even with the improved accuracy levels using human eye FOV, CGI, DGP,
UGR, and VCP show an evaluation accuracy of less than 30%. DGI’s evaluation
accuracy slightly decreased from 64.8% to 62.5%, but it is still the best among the
five.

176
VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
21.9
78.1
17.2
82.8
13.3
86.7
64.8
35.2
2.3
97.7
T otal FOV in E valglare_Incor (Incorr ect)/Corr (Cor rect)

VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
25.0
75.0
29.7
70.3
14.8
85.2
62.5
37.5
14.8
85.2
FOV seen by both eyes in E valglare_Incor (Incorr ect)/Corr (Cor rect)

Figure 5-6. Perceptible glare scenes only: Full fisheye FOV (top) and the human
eye’s FOV (bottom).

177
5.2.3 Disturbing/Intolerable Glare Scenes Only
The evaluation accuracy of the five existing glare index evaluations were compared
to the subjective responses for disturbing and intolerable glare scenes. The graphical
representation of this comparison is shown in Figure 5-7. With a full fisheye FOV , all
five glare indices show an evaluation accuracy of less than 50%. The lowest
accuracy is shown in DGI, which shows only 28.7%. The other four indices show
accuracy ranges from 42.6% to 47.8%.

The same scenes were analyzed with the human eye FOV in Evalglare. The accuracy
of DGI grew much worse with the human eye FOV, decreasing from 28.7% to 9.6%.
It is not clear what caused this sudden drop in DGI’s evaluation accuracy. DGP and
UGR were not affected by the human eye’s FOV, while CGI and VCP show an
accuracy of slightly above 50% with it.

178
VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
42.6
57.4
47.8
52.2
44.3
55.7
28.7
71.3
47.8
52.2
T otal FOV in E valglare_Incor (Incorr ect)/Corr (Cor rect)

VCP UGR DGP DGI CGI
Corr Incor Corr Incor Corr Incor Corr Incor Corr Incor
10 0
80
60
40
20
0
Percent
52.2
47.8
48.7
51.3
43.5
56.5
9.6
90.4
51.3
48.7
FOV seen by both eyes in E valglare_Incor (Incorr ect)/Corr (Cor rect)

Figure 5-7. Disturbing and intolerable glare scenes only: Full fisheye FOV (top) and
the human eye’s FOV (bottom).

179
Based on the data, we can assume that existing glare indices perform differently in
visual scenes with different glare categories:
 DGP is the most reliable index for imperceptible glare scenes, but it shows
very poor accuracy for perceptible glare scenes. FOV does not affect DGP
evaluation performance.
 DGI shows the best accuracy for perceptible glare scenes, but it shows poor
accuracy for disturbing and intolerable glare scenes. Human eye FOV
increases DGI evaluation accuracy only for imperceptible glare scenes.
 VCP shows the best accuracy for disturbing and intolerable glare scenes only
when it utilizes the human eye FOV. The human eye FOV helps to improve
evaluation accuracy.
 CGI and UGR cannot accurately evaluate imperceptible and perceptible glare
scenes, but they show relatively higher accuracy for disturbing and
intolerable glare scenes.

The interior glare scenes were grouped into three different data sets to determine
how the different sun control conditions would affect the accuracy of the glare
indices. The results indicate that different glare indices show different accuracy
performances based on the ambient light levels inside the office as allowed by the
three blind conditions and various blind settings. The analysis results were as
expected. DGP shows the highest accuracy and DGI shows the second highest
accuracy among the five indices. CGI and UGR show the worst accuracy, especially
when ambient light levels inside the office were low. It was again found that the
180
human eye FOV helps improve the accuracy of DGI and VCP.

Another detailed analysis was performed on the same glare scenes, in which the
scenes were organized based on a task type. This analysis checked whether or not
different office task activities would affect the evaluation accuracy of the existing
indices. As the interior glare study considered three task types (no-task, typing task,
and writing task), the glare scenes were grouped into these three tasks. The glare
evaluations for each group were then compared to human subject data. Again, similar
results were found from this analysis. DGP still shows the best accuracy, and DGI
shows the second best accuracy. CGI and UGR show the worst accuracy among the
five. However, no strong correlation was found between office task type and
evaluation accuracy. The different task types do not affect evaluation accuracy.

5.3 Glare Score Ranges from Human Subject Study
After validating the evaluation performance of the existing glare indices, the glare
score ranges for each index were defined using the captured HDR images and
subjective evaluation data. The newly defined glare score ranges of imperceptible,
perceptible, disturbing, and intolerable glare were plotted in interval plots for each
index. The glare scores ranges were calculated twice—once for full fisheye FOV and
once for human eye FOV. Then, newly defined glare score ranges were compared to
the glare category guidelines of the existing glare indices. It was expected that this
glare score range comparison would provide a clue for understanding the low
181
evaluation accuracy issue.

One-way ANOVA tests were performed for CGI to compare subjective glare
evaluations to the calculated glare score ranges in either full fisheye FOV or the
human eye FOV. The ANOVA test of the full fisheye FOV shows a p-value of 0.000,
R-sq of 33.30%, and standard deviation of 2.423. The ANOVA test of the human eye
FOV shows a p-value of 0.000, R-sq of 28.25%, and standard deviation of 2.654.
With a 95% confidence level, two interval plots were created with the newly defines
glare score ranges for four glare categories (Figure 5-8). The calculated glare score
ranges are compared to the existing CGI score ranges, which are shown in red. As
shown in the top graph, the calculated CGI glare score range for imperceptible glare
falls into CGI’s existing disturbing glare range. Also, the calculated perceptible,
disturbing, and intolerable glare score ranges are all within CGI’s intolerable glare
range. With the human eye FOV, the calculated imperceptible and perceptible glare
ranges are within CGI’s disturbing glare range. The bottom half of the calculated
disturbing glare score range is also within CGI’s disturbing glare range. This
explains why the accuracy of CGI increases when glare scenes are analyzed with the
human eye FOV. However, it is clear that CGI’s existing glare ranges are not
accurately defined, and glare scenes cannot be accurately evaluated using the
existing ranges. Specifically, imperceptible and perceptible glare scenes cannot be
evaluated correctly based on the finding here. Also, CGI’s existing score ranges
might be too wide to differentiate different glare categories. It is possible that the
glare score ranges vary from one subject to another, as glare perception is a
182
subjective phenomenon, but it is undeniable that adjusting the existing score ranges
will improve the evaluation performance of these indices. The calculated glare score
ranges can be a good starting point of the adjustment. It is important to address that
there are gaps between the calculated glare score ranges, and it is not clear what
would happen in these gaps based on the current data set. The gaps might be
narrowed or even removed by collecting more human subject study data. It would
also be useful to develop different glare score ranges for full fisheye FOV and human
eye FOV , as glare level analyses are clearly affected by FOV .

183
Intolerable Disturbing Pe r ceptible Imperceptible
36
34
32
30
28
26
24
22
20
Glare category
CGI- FE
22
28
32.8
33.8
29.5
31.3
27.9
29.2
25.2
26.0

Intolerable Disturbing Pe r ceptible Imperceptible
36
34
32
30
28
26
24
22
20
Glare category
CGI- HE
22
28
30.5
31.5
27.3
29.1
26.0
27.3
23.2
24.2

Figure 5-8. Interval plot of CGI score ranges with full fisheye FOV (top) and human
eye FOV (bottom).

Perceptible
Disturbing
Intolerable
Perceptible
Disturbing
Intolerable
184
CGI and UGR show pretty similar evaluation accuracy. A one-way ANOVA test was
also performed for UGR to compare the calculated glare score ranges to subjective
glare evaluations in either full fisheye FOV or human eye FOV. The ANOVA test for
full fisheye FOV shows a p-value of 0.000, R-sq of 23.67%, and standard deviation
of 3.592, while the human eye FOV shows a p-value of 0.000, R-sq of 23.02%, and
standard deviation of 3.689. The calculated UGR score ranges for full fisheye FOV
are 23.7~24.7 for imperceptible glare 26.1~27.4 for perceptible glare 27.1~28.9 for
disturbing glare, and 29.8~30.6 for intolerable glare, as depicted in Figure 5-9. With
a full fisheye FOV, the calculated imperceptible and perceptible glare ranges are
within UGR’s disturbing glare range. The bottom half of the calculated disturbing
glare score range is also within UGR’s disturbing glare range. These score ranges are
different from the human eye FOV, which are 21.1~22.1 for imperceptible glare,
23.4~24.8 for perceptible glare, 24.5~26.3 for disturbing glare, and 27.2~28.2 for
intolerable glare. The calculated glare score ranges with the human eye FOV show
that the imperceptible glare score range is within UGR’s perceptible glare range,
while all the other glare score ranges fall within UGR’s disturbing glare range. Based
on these results, CGI and UGR’s existing glare score ranges must be revised in order
for the indices to provide more accurate daylight glare evaluation.

185
Intolerable Disturbing Pe r ceptible Imperceptible
36
34
32
30
28
26
24
22
20
Glare category
UGR- FE
22
28
29.8
30.6
27.1
28.9
26.1
27.4
23.7
24.7

Intolerable Disturbing Pe r ceptible Imperceptible
36
34
32
30
28
26
24
22
20
Glare category
UGR- HE
22
28
27.2
28.2
24.5
26.3
23.4
24.8
21.1
22.1

Figure 5-9. Interval plot of UGR score ranges with full fisheye FOV (top) and human
eye FOV (bottom).

Perceptible
Disturbing
Intolerable
Perceptible
Disturbing
Intolerable
186
A one-way ANOVA test was performed for DGI to compare the calculated glare
score ranges to subjective glare evaluations. Captured glare scenes were analyzed in
a full fisheye FOV or the human eye FOV. The ANOVA test with full fisheye FOV
shows a p-value of 0.000, R-sq of 33.30%, and standard deviation of 2.423, while the
human eye FOV shows a p-value of 0.000, R-sq of 28.25%, and standard deviation
of 2.645. The calculated DGI glare score ranges for a full fisheye FOV are as follows:
19.3~20.0 for imperceptible glare, 21.3~22.2 for perceptible glare, 22.3~23.8 for
disturbing glare, and 24.5~25.2 for intolerable glare, as shown in Figure 5-10. The
calculated score ranges for imperceptible, perceptible, and disturbing glare are all
within DGI’s perceptible glare range. The calculated DGI glare score ranges for the
human eye FOV are as follows: 17.6~18.4 for imperceptible glare, 19.6~20.6 for
perceptible glare, 20.4~21.8 for disturbing glare, and 22.7~23.5 for intolerable glare.
All four ranges except the bottom half of the calculated imperceptible glare score
range are within DGI’s perceptible glare range. While CGI and UGR have the
calculated glare score ranges under disturbing or intolerable glare categories, DGI’s
calculated glare score ranges mostly fall into the perceptible glare category. This
explains why DGI shows such high evaluation accuracy, especially among
perceptible glare scenes. Unfortunately, the existing DGI score ranges might be too
broad to accurately evaluate glare scenes with different glare levels. It might be
necessary to adjust the existing score ranges to gain better glare evaluation accuracy.

187
Intolerable Disturbing Pe r ceptible Imperceptible
26
24
22
20
18
16
Glare category
DGI- FE
18
24
24.5
25.2
22.3
23.6
21.3
22.2
19.3
20.0

Intolerable Disturbing Pe r ceptible Imperceptible
26
24
22
20
18
16
Glare category
DGI- HE
24
18
22.7
23.5
20.4
21.8
19.6
20.6
17.6
18.4

Figure 5-10. Interval plot of DGI score ranges with full fisheye FOV (top) and
human eye FOV (bottom).

Imperceptible
Perceptible
Disturbing
Imperceptible
Perceptible
Disturbing
188
A one-way ANOVA test was performed for VCP between the calculated glare score
ranges and subjective glare evaluations. It was performed twice: once with full
fisheye FOV and once with human eye FOV. The full fisheye FOV shows a p-value
of 0.000, R-sq of 30.90%, and a standard deviation of 20.51, while the human eye
FOV shows a p-value of 0.000, R-sq of 33.75%, and a standard deviation of 19.23.
With the human eye FOV, the calculated glare score ranges are 74.5~79.3 for
imperceptible glare, 56.7~64.2 for perceptible glare, 44.7~55.3 for disturbing glare,
and 32.4~38.8 for intolerable glare (Figure 5-11). These calculated glare score ranges
are very close to the VCP’s existing glare category ranges, which are 80~100 for
imperceptible glare, 60~80 for perceptible glare, 40~60 for disturbing glare, and
0~40 for intolerable glare. The calculated ranges better match the VCP’s existing
ranges when using the human eye FOV rather than the full fisheye FOV. With the
human eye FOV, the calculated disturbing and intolerable score ranges are pretty
much the same as VCP’s existing glare ranges, but the calculated imperceptible score
range is within the perceptible glare range. Furthermore, the calculated perceptible
score range is in between the VCP’s perceptible and disturbing glare ranges. This
shows that VCP’s existing glare categories are more accurately defined than CGI,
UGR, and DGI’s, but still need some improvement for imperceptible and perceptible
glare categories.
189
Imperceptible Per ceptible Disturbing Intolerable
90
80
70
60
50
40
30
20
Glare category
VCP- FE
40
60
80
25.0
30.5 36.1
46.9
47.8
55.7
65.9
71.4

Intolerable Disturbing Pe r ceptible Imperceptible
90
80
70
60
50
40
30
20
Glare category
VCP- HE
40
60
80
32.4
38.8
44.7
55.3
56.7
64.2
74.5
79.3

Figure 5-11. Interval plot of VCP score ranges with full fisheye FOV (top) and
human eye FOV (bottom).

Intolerable
Disturbing
Perceptible
Imperceptible
Intolerable
Disturbing
Perceptible
Imperceptible
190
Finally, a one-way ANOVA test was performed for DGP to compare the calculated
glare scores to the subjective glare evaluations. Glare score ranges were calculated in
Evalglare in either full fisheye FOV or the human eye FOV. The ANOVA test for the
full fisheye FOV shows a p-value of 0.000, R-sq of 61.98%, and a standard deviation
of 0.1593, while the human eye FOV shows a p-value of 0.000, R-sq of 61.98%, and
a standard deviation of 0.1588. Both R-sq values are much higher than those of the
other glare indices, and there is no difference between the two FOVs. The DGP
scores were calculated using externally measured illuminance values at the range of
the human eye; thus, the calculated glare score ranges are almost identical for both
FOVs. As DGP shows the highest accuracy among the five glare indices, the
calculated glare score ranges also match to DGP’s existing glare ranges, as shown in
Figure 5-12. In fact, the calculated perceptible glare score range exactly matches to
DGP’s existing perceptible glare range of 0.35 to 0.40. The calculated imperceptible
glare score range is not identical to DGP’s existing imperceptible glare range, but it
is still within the range. The calculated glare score ranges for both disturbing and
intolerable glare are within the DGP’s existing intolerable glare range, which is
above 0.45. It is possible to adjust these score ranges for disturbing and intolerable
glare categories, so that the two different glare levels can be accurately evaluated.
191
Intolerable Disturbing Pe r ce ptible Imperceptible
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
Glare category
DGP- FE
0.3 5
0.4
0.4 5
0.809
0.933
0.558
0.693
0.344
0.394
0.232
0.252

Intolerable Disturbing Pe r ce ptible Imperceptible
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
Glare category
DGP- HE
0.3 5
0.4
0.4 5
0.808
0.932
0.557
0.692
0.343
0.392
0.230
0.250

Figure 5-12. Interval plot of DGP score ranges with full fisheye FOV (top) and
human eye FOV (bottom).

Imperceptible
Perceptible
Disturbing
Intolerable
Imperceptible
Perceptible
Disturbing
Intolerable
192
The interior glare study data has been evaluated using the existing glare indices. The
analysis results indicate that DGP shows the best evaluation accuracy among the five
indices when the subjective evaluations are used as the baseline for determining
accuracy. Unfortunately, the evaluation accuracy of all the existing glare indices is
too low to be trusted. The inconsistent evaluation issue arose again throughout this
analysis. Furthermore, each glare index shows higher or lower evaluation accuracy
depending on the glare levels of a scene, indicating inter-index inconsistency. It was
determined that the human eye FOV improves the glare evaluation performance of
the existing glare indices. These findings can help users of the existing indices to
find a better understanding of the indices. The various research groups should
consider and discuss adjusting the existing glare score ranges.

193
Chapter 6 Results and Analysis
Development of the AGF and RGF Method Based on Human Subject Study
Results and Analysis

This chapter analyzes the collected human subject study data along with the captured
High Dynamic Range (HDR) images and luminance measurements. A statistical
analysis approach was used to find correlations between variables such as luminance,
vertical illuminance, and contrast and subjective evaluations such as visual comfort,
visual satisfaction, and glare categories. Interior and exterior glare data were
analyzed separately, to see whether a new glare analysis method should be developed
for each glare issue, or if one could be developed for both glare issues.

6.1 Interior Glare Study Results
More than 450 HDR images were captured from interior glare human subject studies.
Each captured scene is also described by subjective evaluation data from the human
subjects. Figure 6-1 shows nine different interior scenes captured from a single test
for one subject. As described in the methodology chapter (Chapter 4), combinations
of three different tasks and three different blind conditions were given to each subject
for a single test. The order of the combinations was shuffled randomly, so that the
subjects would experience inconsistent lighting environment transitions.

194
No-task Typing task Writing task
Fully open blind condition
No-task Typing task Writing task
Roller blind condition
No-task Typing task Writing task
Venetian blind condition
Figure 6-1. Nine different HDR images captured from a single test.

These captured HDR images were then analyzed in the developed MATLAB code to
calculate crucial values for analysis, such as minimum luminance, maximum
luminance, mean background luminance, glare source luminance range, mean glare
195
source luminance, glare ratio, and glare size. The calculated values from the HDR
images were then compared to the visual maps and subjective evaluations that
subjects provided during the tests. Figure 6-2 shows the captured scenes that were
processed in the developed MATLAB code. For each scene, the MATLAB code
calculates the crucial lighting parameters. Then it indicates potential glare sources in
the field of view. First, the calculated values from the captured HDR images were
compared to human subject evaluations. After a thorough analysis of the values and
subjective evaluations, the indicated potential glare sources from the MATLAB code
were compared to the visual maps marked up by human subjects. This analysis
approach enabled the researcher to fully understand how a specific lighting condition
caused visual discomfort to a subject and what exactly caused the discomfort to the
subject. The entire data set was analyzed together first. Then the entire data set was
separated into small data groups based on task type, blind condition, and glare
category. This approach helped determine whether human subjects experienced
different glare sensations by performing different tasks or experiencing different
blind conditions.

196
No-task Typing task Writing task
Fully open blind condition
No-task Typing task Writing task
Roller blind condition
No-task Typing task Writing task
Venetian blind condition
Figure 6-2. Nine different HDR images processed in MATLAB, wherein red fields
indicate potential glare sources as calculated in MATLAB.

6.1.1 Percentages of Subject Processed in MATLAB
After the interior glare study was completed, the researcher checked to see if the
197
human subjects experienced the entire range of visual comfort and discomfort, from
imperceptible glare to intolerable glare. Figure 6-3 shows the percentage of each
glare category that was identified by subjects. Of all the subjective data, 45% was
evaluated as imperceptible glare and 55% was evaluated among the various other
glare categories. A breakdown of the 55% shows that 27% was evaluated as
perceptible glare, 17% was evaluated as disturbing glare, and 9% was evaluated as
intolerable glare. The worse glare category shows a lower percentage. This was
exactly as intended, since human subjects were able to control the roller blinds or
venetian blinds to avoid any discomfort glare issues experienced by the fully open
blind condition. The percentage of intolerable glare is the lowest among the glare
categories, since intolerable glare was for the most part experienced only during the
fully open blind condition.

198
Intolerable Disturbing Pe r ceptible Imperceptible
50
40
30
20
10
0
Glare category
Percent
9.55556
17.3333
27.7778
45.3333
Per cent within all data.
Figure 6-3. Percentages of glare sensation experienced by subjects.

The entire data set was plotted in Figure 6-4 to compare the percentages of subjects’
various visual satisfaction levels. Since a subject’s visual satisfaction levels were
measured with a seven-point Likert scale, Figure 6-4 provides percentage
information for more detailed categories than those of the previous graph, which
shows only four categories. The subjects’ visual satisfaction levels are also evenly
distributed. Of all the subjects, 36.8% indicated that they were visually dissatisfied,
while 50.6% indicated that they were visually satisfied. Furthermore, 12.4%
indicated that they were neither satisfied nor unsatisfied. The percentage of positive
and neutral satisfaction combined is 63.0 percent, which is much higher than the
percentage of imperceptible glare in Figure 6-3. This means that 18% of the scenes
199
were satisfactory even when the subjects noticed a certain level of glare in their
fields of view.

3 2 1 0 - 1 - 2 - 3
30
25
20
15
10
5
0
How satistied?
Percent
11.3333
21.3333
18
12.4444
18.4444
11.3333
7.11111
Per cent within all data.
Figure 6-4. Percentages of subjects’ visual satisfaction levels.

Figure 6-5 was created to show the percentages of each visual comfort level
experienced by subjects. The graph shows that subjects experienced the entire range
of visual comfort levels. The bar chart was expected to be almost identical to the
previous graph, since the researcher assumed that the subjects would equate visual
satisfaction with visual comfort. However, there are small discrepancies between the
percentages of satisfaction and comfort levels in Figures 6-4 and 6-5. The subjects
translated these two subjective terms slightly differently in the same visual scene,
200
and found some level of glare to be “satisfactory.” 35.7% were evaluated as being
visually uncomfortable in scenes, while 52.7% were evaluated as being visually
comfortable scenes. 11.9% were neutral to visual comfort. Again, the total
percentage of visual comfort and neutral comfort combined is 64.6 percent, which is
higher than the imperceptible glare percentage. This confirms that subjects were not
dissatisfied or uncomfortable even when some levels of glare existed in their fields
of view.

3 2 1 0 - 1 - 2 - 3
30
25
20
15
10
5
0
How comfortable?
Percent
12.6667
22.8889
17.3333
11.3333
17.1111
11.3333
7.33333
Per cent within all data.
Figure 6-5. Percentages of subjects’ visual comfort levels.

Subjects were asked to assume that this research setting was their daily office
environment and to evaluate how visually comfortable the lighting environment of
201
the research setting was. The collected responses were quite similar to the data
presented in the previous two figures. However, the percentage of strongly
uncomfortable subjects (-3) increased from 7% to 9.3% in the chart below. It is
possible that the condition of assuming the setting to be a daily office environment
might have affected the subjects’ evaluation of visual comfort. However, the
discrepancy might be too minimal to make a strong argument; thus, it was not further
analyzed.

3 2 1 0 - 1 - 2 - 3
30
25
20
15
10
5
0
Ho w co mfortable if d aily w o rk?
Percent
12
21.7778
17.5556
11.1111
16.8889
11.3333
9.33333
Per cent within all data.
Figure 6-6. Percentages of subjects’ visual comfort levels when the daily office
condition was assumed.

As explained above, the subjects were asked various subjective questions about
general light level, visual comfort, visual satisfaction, and glare sensation to evaluate
202
the same visual condition. Each subject’s responses to different questions were
compared to check the consistency of subjective evaluations on a same visual scene,
and to see if there is any correlation between different subjective questions.

6.1.2 Analysis of Variance (ANOVA) Test: Subjective Response Consistency
Glare sensation was evaluated in a bar scale from 0.0 to 3.0 according to the four
glare categories (imperceptible, perceptible, disturbing, and intolerable glare), and so
the researcher needed to translate the bar scale into four different glare sensation
categories. One-way analysis of variance (ANOVA) tests were performed to define
glare level ranges that are strongly correlated to glare categories from imperceptible
to intolerable glare. After glare levels were defined for the glare categories, the one-
way ANOVA test was performed again to compare the defined glare levels and glare
categories. The following glare level ranges were determined for each category:
imperceptible glare ranges from 0 to 0.3, perceptible glare ranges from 0.4 to 1.4,
disturbing glare ranges from 1.5 to 2.5, and intolerable glare ranges from 2.6 to 3.0.
Mean values for imperceptible, perceptible, disturbing, and intolerable glare are 0.05,
0.79 1.97, and 2.84, as shown in Figure 6-7. This ANOVA test shows a p-value of
0.000 with 95% accuracy confidence. The coefficient of determination value is
93.81%, which supports the idea that glare levels were successfully defined for glare
categories. Thus, the defined glare level ranges will work accurately with the
majority of future data.
203
Intolerable Disturbing Pe r ceptible Imperceptible
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Glare category
Glare lev el
2.84186
1.97179
0.7928
0.05
Figure 6-7. One-way ANOV A: Glare level vs. glare sensation category (CI = 95%, F
= 2251.23, P = 0.00, R-sq = 93.81%, R-sq(adj) = 93.76%, StDev = 0.2433).

Prior to the analysis of captured HDR images, it was important to check the
consistency of participant responses among the different subjective questions, such
as visual satisfaction, comfort, and glare sensation. Again, a one-way ANOVA test
was performed to check whether or not the subjects’ visual satisfaction or visual
comfort evaluations match their glare category evaluations. This procedure can
assure that subject evaluations were consistent across all the questions. First, the
researcher compared the subjects’ answers concerning visual satisfaction and glare
sensation. As the ANOVA test result shows, glare categories are well associated with
visual satisfaction levels. The coefficient of determination value for this comparison
is 45.98%, while the p-value is 0.000. The visual satisfaction range from -2 to 3 falls
204
into the imperceptible glare category, while the middle 50% ranges from 0 to 2. The
bottom 25% is within -2 to 0. The subjects felt dissatisfied even when there was no
visible glare source in the scene. Based on the collected data, subjects were not
satisfied when the research setting was dark, even without any glare sources.
Subjects evaluated perceptible glare to be present when visual satisfaction levels
were between -2 and 3. Again, the middle 50% ranges from -1 to 1.

Before performing this analysis, it was not clear whether or not the existence of
perceptible glare bothered building occupants. As shown in Figure 6-8, the red
dashed line shows a neutral satisfaction level (0), and the perceptible glare range is in
the exact center of the positive and negative satisfaction levels. This shows that
perceptible glare can be visually satisfying or dissatisfying. This finding needs
further investigation in order to avoid potential confusion in the future use of the
existing glare categories.

The disturbing glare range is from satisfaction level -1 to 3. The middle 50% sits
between -2 and 0. 75% of disturbing glare falls onto the negative side, but 25% still
sits on the positive side of visual satisfaction. This can be interpreted to indicate that
the existence of disturbing glare does not always make people visually dissatisfied.

Unlike the other three glare sensations, intolerable glare falls onto the negative side
only, with satisfaction levels of -1 to -3. The middle 50% sits between -2 and -3. This
205
shows that intolerable glare absolutely makes people dissatisfied.

Intolerable Disturbing Pe r ceptible Imperceptible
3
2
1
0
- 1
- 2
- 3
Glare category
How satistied?
- 2.32558
- 1.01282
0.344
1.37745
Figure 6-8. One-way ANOV A: Visual satisfaction vs. glare sensation (CI = 95%,
F = 126.52, P = 0.00, R-sq = 45.98%, R-sq(adj) = 45.61%, StDev = 1.326).

The following ANOVA test compares subjective evaluations of visual comfort level
to those of glare category. The results are depicted in Figure 6-9. The coefficient of
determination value for this comparison is 49.19%, which is higher than that of the
previous comparison. The visual comfort range for imperceptible glare is between 0
and 3, while the middle 50% sits between 1 and 2. Disturbing glare sits between
visual comfort level 0 and -3, and intolerable glare sits between -1 and -3. This
comparison clearly shows that visual comfort has a direct relationship to glare
category, unlike the relationship between visual satisfaction and glare category.
206
However, perceptible glare is still between visual comfort level -2 and 3, with the
middle 50% between -1 and 2. It is thus possible to claim that perceptible glare is not
always visually uncomfortable. It can be either visually comfortable or
uncomfortable.

Intolerable Disturbing Pe r ceptible Imperceptible
3
2
1
0
- 1
- 2
- 3
Glare category
Ho w co mfo rtable?
- 2.39535
- 1.07692
0.504
1.47549
Figure 6-9. One-way ANOV A: Visual comfort level vs. glare category (CI = 95%,
F = 143.92, P = 0.00, R-sq = 49.19%, R-sq(adj) = 48.85%, StDev = 1.312).

The next ANOVA test compares subjects’ glare category evaluations to their visual
comfort levels, with the assumption that the research setting is their daily workplace.
The results are depicted in Figure 6-10. The coefficient of determination value for
this comparison is 55.66% and the p-value is 0.000. The visual comfort range for
imperceptible glare is between 0 and 3, while the middle 50% sits between 1 and 2.
207
Disturbing glare sits between visual comfort level -3 and 0, while the middle 50%
sits between -1 and -2. Intolerable glare sits between -2.5 and -3. Again, perceptible
glare is between visual comfort level -2 and 3, with middle 50% between -1 and 1.
This indicates that perceptible glare cannot be assumed to create visual discomfort.

Intolerable Disturbing Pe r ceptible Imperceptible
3
2
1
0
- 1
- 2
- 3
Glare category
Ho w co mfo rtable if d aily w o rk?
- 2.67442
- 1.41667
0.37594
1.48515
Figure 6-10. One-way ANOV A: Visual comfort level with daily workplace
assumption vs. glare category (CI = 95%, F = 186.65, P = 0.00, R-sq = 55.66%,
R-sq(adj) = 55.37%, StDev = 1.252).

The ANOVA tests deliver the following three findings: a) the subjective evaluations
of visual satisfaction, visual comfort, and glare category are consistent; b) the
perceptible glare category cannot be clearly defined as creating visual dissatisfaction
or visual discomfort; and c) the glare category shows a stronger correlation to visual
208
comfort than to visual satisfaction.

6.1.3 Field of View: Full Fisheye vs. Human Eye
After the subjects’ responses were thoroughly analyzed, the captured HDR images
were also analyzed in the MATLAB code. Before calculating luminance values and
contrast ratios from the captured HDR images, it was important to check which field
of view would provide more accurate analysis: the 180-degree full fisheye view or
the human eye field of view (FOV). As mentioned in Chapter 3, a human eye FOV is
smaller than a full 180 vertical and horizontal degree angle (Figure 3-18). However,
most previous research has used a full fisheye view for glare analysis, due to the ease
of doing so. A few studies tried to compare different fields of view, but the results
were not strong enough to claim which FOV is more accurate.

While the test for this study was performed, human subjects were asked not to
change their fields of view, especially when they performed typing and writing tasks,
and they were asked to mark up visual maps with exactly what they saw in their
FOVs. It was expected that glare source locations and sizes marked up on visual
maps would help to define a boundary of the field of view that was actually seen by
the subjects.

The number of scenes with glare sources and without glare sources were separately
counted for each task type. Then, glare source locations and sizes as marked up by
209
subjects were carefully compared to the human eye FOV and full fisheye FOV, to
check which FOV matches the glare source locations and sizes that the subjects
marked up on visual maps. The number of scenes accurately defined by the human
eye FOV or full fisheye FOV were separately counted, and then the total frequencies
were plotted onto three graphs. The scenes in which both FOVs accurately detected
glare sources were also counted and plotted. As shown in Figure 6-11, a total of 94
out of 152 scenes have glare sources detected by human subjects when no task was
performed by subjects. Out of these ninety-four glare scenes, eighty show no
difference in glare source detections between the two FOVs. Of the remaining
fourteen glare scenes, nine show accurate glare source detections only with the
human eye FOV, while five do so only with the full fisheye FOV. Based on the
results shown here, the FOV does not matter for the majority of the glare scenes
when people do not perform a task and simply consider their general impression of a
visual condition. The FOV does matter when one is performing different tasks,
which may help to explain why some researchers did not find a difference between
fisheye and human eye FOV .
210
Glare_Sour ces_Detected No_Glare_Sources_Detected
10 0
80
60
40
20
0
Fre q uency
HE
FE
Both
Vie w
Fie ld of
94
85
80
58 58 58
Figure 6-11. Total frequency of glare source detection outside human eye FOV for
the no-task condition.

For the typing task, a total of 153 scenes were analyzed. Of those, 108 scenes have
glare sources detected by human subjects. This is depicted in Figure 6-12. Out of the
108 glare scenes, 57 have glare sources detected within both human eye and full
fisheye FOVs. For these fifty-seven glare scenes, either FOV can be used to
accurately detect glare sources. The full fisheye FOV cannot accurately detect glare
sources in 48 out of 108 glare scenes, while human eye FOV cannot detect glare
sources only in 3 glare scenes. This might have occurred because the human eye
FOV is more accurate than a full fisheye FOV for typing task scenes, since there are
many more cases that full fisheye FOV cannot accurately detect glare sources than
human eye FOV. It appears that there is a significant difference, and the human eye
211
FOV is more accurate.

Glare_Sour ces_Detected No_Glare_Sources_Detected
10 0
80
60
40
20
0
Fre q uency
HE
FE
Both
Vie w
Fie ld of
108
60
57
45 45 45
Figure 6-12. Total frequency of glare source detection outside human eye FOV in the
typing task condition.

The writing task scenes show a much clearer difference between the two different
FOVs than the other two task conditions. Figure 6-13 shows the frequency of glare
source detection outside the human eye FOV . Of the 145 writing task scenes, 72 have
no glare sources detected by subjects. A total of seventy-three scenes have glare
sources detected by subjects. Among these glare scenes, the human eye FOV
accurately detects glare sources in fifty-three scenes, while full fisheye FOV
accurately detects glare sources in only seven scenes. The remaining thirteen glare
scenes were accurate with both FOVs.
212
Glare_Sour ces_Detected No_Glare_Sources_Detected
10 0
80
60
40
20
0
Fre q uency
HE
FE
Both
Vie w
Fie ld of
73
20
13
72 72 72
Figure 6-13. Total frequency of glare source detection outside the human eye FOV
during the writing task.

Based on this analysis, it seems more appropriate to use the human eye FOV than the
full fisheye FOV, especially when people are reading and writing on paper in an
office environment. Even though the data above shows that the human eye FOV
shows a higher frequency of accuracy to a human subject’s glare detection than does
the full fisheye FOV, it is still difficult to argue that the human eye FOV should be
used for the daylight glare analysis method introduced in this study. After all, the
research plan was not developed to investigate the field of view issue. However,
these findings may be useful for future research about FOV in glare analysis.

The following section describes how to detect glare sources in a field of view using
213
the MATLAB code.

6.1.4 Glare Source Detection in Visual Map and HDR Image
While the researcher compared the calculated luminance values in HDR images to
human subject glare evaluation data, the subjects’ markups on visual maps were also
compared to the glare sources that were automatically detected by the MATLAB
code. As explained in Chapter 4, potential glare sources were detected using a
relative luminance threshold that is five times the background mean luminance in the
field of view. Thus, any pixel with higher than five times the mean background
luminance is identified as a potential glare source. This luminance threshold value
for glare source detection can thus change when background mean luminance
changes under different lighting conditions. More than 450 HDR images and visual
maps were compared to check whether or not the identified glare sources matched.

The following three figures (Figures 6-14, 6-15, and 6-16) show no-task, typing task,
and writing task scenes to compare glare source detection between the visual maps
and the HDR images assessed using the MATLAB code. Figure 6-14 shows visual
maps and HDR images for no-task scenes for three different blind settings. This
subject experienced disturbing glare under the fully open blind condition and roller
bind condition, and experienced imperceptible glare under the venetian blind
condition. As visual map and HDR images from the MATLAB code show, glare
sources were detected pretty accurately by using the multiplier 5 with the first two
214
blind conditions. The first two comparisons show that the glare source locations of
the visual map and the HDR image are a close match, even though the glare source
sizes are somewhat different. Under the venetian blind condition, the HDR image has
detected a totally different glare source than the visual map. The HDR image shows a
huge potential glare source on the left windows, while the human subjects did not
consider there to have been a perceptible glare at all. The existing formula to detect
potential glare sources by using a multiplier thus does not seem to be accurate,
especially when ambient lighting is low. With this existing glare detection approach,
glare sources will always be detected in HDR images, regardless of whether the glare
sources are in the subject’s field of view. Because these automatically detected glare
sources are considered by the imager to be potential glare sources, end users can be
misled to believe that there will be glare issues from potential glare sources that will
not actually have a perceivable effect.

215
Human subject HDR image detection
Fully open condition: Disturbing glare
Human subject HDR image detection
Roller blind condition: Disturbing glare
Human subject HDR image detection
Venetian blind condition: Imperceptible glare
Figure 6-14. Glare source detection example under the no-task condition.

216
Figure 6-15 compares the three different blind conditions in the context of the user’s
engagement in a typing task. Visual maps include red ‘X’ marks to indicate glare
sources in each FOV . Each subject clicked ‘X’ marks on a visual map that was shown
on a computer monitor right after they completed a typing task. In this context, the
subject evaluated disturbing glare and indicated glare sources on the corner of the
front and side windows for the fully open blind condition. For the roller blind
condition, glare sources were indicated at similar locations.

The venetian blind condition also has glare sources on the corner, but its size is
smaller than those of the other two blind conditions. The subject evaluated both
roller and venetian blind conditions as perceptible glare. Again, glare sources
detected on HDR images pretty well match to the glare sources marked up in visual
maps, even though detected glare source sizes in HDR images are larger than glare
sources in visual maps. Thus, it may be possible to use a different multiplier than 5
for glare source detections, since the multiplier 5 might not always work in different
lighting conditions. A more accurate and consistent method to detect glare source
locations and sizes in FOV was found, which can use a multiplier factor, an absolute
luminance value, or a combination of the two.

217
Human subject HDR image detection
Fully open condition: Disturbing glare
Human subject HDR image detection
Roller blind condition: Perceptible glare
Human subject HDR image detection
Venetian blind condition: Perceptible glare
Figure 6-15. Glare source detection example under the typing task condition.

218
Figure 6-16 compares visual maps and HDR images for a writing task under three
different blind conditions. Much as with visual maps for no-task, subjects were asked
to draw red circles or lines on hard copies of visual maps. A visual map was
presented to subjects right after they had completed the writing task. The subjects did
not need to change their FOV to fill in the survey and mark up the visual maps. As
shown in Figure 6-16, the FOV for the writing task is quite different from the FOVs
for no-task and the typing task, since subjects were asked to look down while they
were reading and writing on a paper that was supported by the desk. They were also
asked to maintain their FOVs while performing the writing task and completing the
survey.

219
Human subject HDR image detection
Fully open condition: Disturbing glare
Human subject HDR image detection
Roller blind condition: Perceptible glare
Human subject HDR image detection
Venetian blind condition: Perceptible glare
Figure 6-16. Glare source detection example under the writing task condition.

220
The fully open blind condition was evaluated as disturbing glare. The visual map and
HDR image show similar glare locations, but glare sizes are quite different. The
roller blind condition was subjectively evaluated as having perceptible glare. Glare
sources detected in the HDR image do not match with those on the visual map. The
venetian blind condition was also subjectively evaluated as having perceptible glare.
While both the visual map and the HDR image show identical glare sources at the
end of the desk, the HDR image also indicates additional potential glare sources
through the venetian blinds. The examples shown in this section show how the
multiplier 5 conveniently visualizes potential glare sources in computer renderings or
HDR images, but might not be enough to correctly indicate glare sources for various
lighting conditions. However, this visualization might be adjusted by the absolute
range within which the values fall.

The glare source detection comparisons on the entire 450 scenes show that glare
source locations in the visual map match well with the automatically detected glare
source locations in the HDR images. However, glare source sizes are quite different
between visual maps and HDR images. Also, the HDR image tends to indicate more
glare sources than do the subjects.

The following section investigates captured luminance information in HDR images
and seeks its correlation with human subject visual discomfort data. It also describes
how to find absolute luminance values to detect glare sources in place of using the
221
multiplier 5.
6.1.5 ANOVA Test: Glare Source Minimum Luminance
The researcher first checked whether or not there is a strong correlation between the
analyzed luminance values from the HDR images and subjective glare evaluations in
order to define glare source luminance ranges for different glare sensations.

The ANOVA test was performed to compare subjects’ glare sensation levels to the
calculated glare source minimum luminance values (Figure 6-17). Glare level data
was collected from the surveys as subjects marked their glare sensation levels in a
bar scale. The data was transferred to numeric values ranging from 0.0 to 3.0. As
shown in the following graph, the calculated glare source minimum luminance
ranges vary widely for different glare levels, but mean values gradually increase as
glare levels go up. The P-value of this ANOVA test is 0.000, with a 95% confidence
level. The coefficient of determination value for this comparison is 54.41%.

222
3.0
2.9
2.8
2 .7
2 .6
2 .5
2 .4
2.3
2.2
2.1
2.0
1 .8
1 .7
1 .6
1 .5
1.4
1.3
1.2
1.0
0.9
0 .8
0 .7
0 .6
0 .5
0.4
0.3
0.2
0.1
0.0
25 00 0
20 00 0
15 00 0
10 00 0
50 00
0
Glare lev el
Glare source minimum luminance (cd/m2)
Figure 6-17. One-way ANOV A: Glare source minimum luminance vs. glare level
(CI = 95%, F = 17.95, P = 0.00, R-sq = 54.41%, R-sq(adj) = 51.38%, StDev = 3366).

After the glare sensation levels were accurately categorized into four different glare
categories, another ANOVA test was performed to compare glare categories and
calculated glare source minimum luminance values from the captured HDR images.
These results are shown in Figure 6-18. The calculated glare source minimum
luminance ranges are well defined for each of four glare categories. P-value is 0.000,
with a 95% confidence level. The coefficient of determination value for this
comparison is slightly higher than the last—from 54.41% with glare level to 56.67%
with glare category. The increase could be caused by the decrease in the number of
bins, from 29 to 4, for glare sensation. However, both ANOVA tests show significant
differences between the defined luminance ranges. Figure 6-18 also shows that mean
223
luminance values linearly increase as glare sensation grows from imperceptible to
intolerable glare.

Intolerable Disturbing Pe r ceptible Imperceptible
25 00 0
20 00 0
15 00 0
10 00 0
50 00
0
Glare category
Glare source minimum luminance (cd/m2)
Figure 6-18. One-way ANOV A: Glare source minimum luminance vs. glare
sensation category (CI = 95%, F = 194.45, P = 0.00, R-sq = 56.67%,
R-sq(adj) = 56.38%, StDev = 3188).

A general linear model analysis was then performed to double check whether or not
the other subjective responses, such as visual comfort and visual satisfaction,
correlates to glare source minimum luminance values. As previously shown, the
glare category has a p-value of 0.000, while visual satisfaction has a p-value of 0.300,
which is not significant. Visual comfort, meanwhile, shows a p-value of 0.006, which
is certainly better than that of visual satisfaction but is still considered not significant
224
enough for further analysis. Visual comfort with the assumption that the location is a
daily workplace shows a p-value of 0.835, which is the worst among these variables.
Since the p-value for glare category is the most significant to glare source luminance
among the surveyed subjective evaluation data, only the glare category data was
utilized for further analysis.

6.1.6 Absolute Luminance Ranges for Glare Category
Captured HDR images were analyzed in the MATLAB code to calculate values for
the following within each visual scene: glare source minimum luminance, glare
source average luminance, background minimum luminance, background average
luminance, background maximum luminance (glare source maximum luminance),
glare source size, and glare ratio. The researcher analyzed the calculated values and
the surveyed subjective evaluation data. Furthermore, they recorded two vertical and
two horizontal illuminance values during the tests and compared these to the
subjective data.

Since one of the objectives is to define AGF, glare source luminance values from the
captured HDR images were first compared to human subjects’ glare evaluations.
Then a series of graphs were plotted to understand what absolute luminance values
could cause different levels of discomfort glare (Figures 6-19, 6-20, 6-21, and 6-22).
These graphs are interval plots with a 95% confidence level. They show four
different absolute luminance ranges for different glare categories, from imperceptible
225
to intolerable glare.

Intolerable Disturbing Pe r ceptible Imperceptible
18 00 0
16 00 0
14 00 0
12 00 0
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
12143.9
14520
6852.86
9551.42
3777.16
4854.53
1617.05
1977.51
13332
8202.14
4315.84
1797.28
Figure 6-19. Glare source minimum luminance vs. glare category for all task
conditions.

As expected, absolute luminance ranges gradually increase as the glare sensation
becomes worse. Among the many values plotted on the graphs, there are three highly
critical values: the highest value of the imperceptible glare luminance range, the
lowest value of the disturbing glare luminance range, and the lowest value of the
intolerable glare luminance range. These three values define a range that can be
visually comfortable or uncomfortable. An analysis of the entire data set found that
absolute luminance up to 1,977 cd/m
2
is dominant with imperceptible glare.
Disturbing glare is dominant from 6,852 cd/m
2
to 9,551 cd/m
2
, and intolerable glare
Imper c eptible g l a r e thr e shold
Int o ler a ble g l a r e t h r e s h old
Distu r bing glar e t hr es ho ld
226
occurs from 12,143 cd/m
2
and up. Based on the imperceptible and disturbing glare
thresholds, perceptible glare can be defined as ranging from 1,977 cd/m
2
to

6,852
cd/m
2
.

After defining the absolute luminance ranges from the entire data set, three interval
plots were created to see whether a different office task activity would change the
defined absolute luminance ranges (Figures 6-20, 6-21, and 6-22). Figure 6-20 was
plotted with no-task scenes only, and its absolute luminance ranges are quite
different from those of the previous plot. The absolute luminance range of
imperceptible glare goes up to 2,752 cd/m
2
under the no-task condition.

227
Intolerable Disturbing Pe r ceptible Imperceptible
18 00 0
16 00 0
14 00 0
12 00 0
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
12522.7
16180.9
7000.91
11694
3294.49
5778.75
1891.17
2752.26
14351.8
9347.47
4536.62
2321.71
Figure 6-20. Glare source minimum luminance vs. glare category under the no-task
condition.

Disturbing glare is dominant from 7,000 cd/m
2
to 11,694 cd/m
2
, and intolerable glare
occurs from 12,522 cd/m
2
and up. The perceptible glare luminance range can be
defined from 2,752 cd/m
2
to

7,000 cd/m
2
, as it is between imperceptible and
disturbing glare. Based on these ranges, it is possible to say that an absolute
luminance value of 7,000 cd/m
2
will cause a visual discomfort issue to occupants
when no office task is involved.

Figure 6-21 was plotted using data from the typing task condition only. Again, the
absolute luminance ranges for the typing task scenes are quite different from the
ranges for no-task scenes. Imperceptible glare luminance range is up to 1,920 cd/m
2
.
Imper c eptible g l a r e thr e shold
Int o ler a ble g l a r e t h r e s h old
Distu r bing glar e t hr es ho ld
228
The disturbing glare range is from 5,014 cd/m
2
to 8,875 cd/m
2
, while the intolerable
glare range is from 11,718 cd/m
2
and up. Perceptible glare can be defined between
the imperceptible and disturbing glare thresholds, and thus ranges from 1,920 cd/m
2
to

5,014 cd/m
2
. Typing task luminance ranges are certainly lower than the luminance
ranges found from no-task scenes. Based on these ranges, an absolute luminance
value of 5,014 cd/m
2
will cause visual discomfort to occupants when a computer-
based typing task is involved.

Intolerable Disturbing Pe r ceptible Imperceptible
18 00 0
16 00 0
14 00 0
12 00 0
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
11718.5
15113.3
5014.52
8875.32
3420.75
4986.55
1449.65
1920.04
13415.9
6944.92
4203.65
1684.84
Figure 6-21. Glare source minimum luminance vs. glare category under the typing
task condition.

The difference between no-task luminance ranges and typing task luminance ranges
suggests that the subjects were more sensitive to glare source luminance levels when
Imper c eptible g l a r e thr e shold
Int o ler a ble g l a r e t h r e s h old
Distu r bing glar e t hr es ho ld
229
they performed a typing task than when they did not perform any task.

Figure 6-22 was created to describe the results from the writing task scenes. This plot
shows a somewhat different result from the previous plots. Luminance ranges for
disturbing and intolerable glare have a big overlap. For the writing task, subjects
experienced intolerable glare at much lower luminance values than for the other
tasks. After thoroughly checking the captured writing task images, the researcher
found that the direct sunlight through the windows illuminated the papers on the desk,
which made the task area a potential glare source. Compared to the sun and windows,
reflected light on the paper has relatively lower luminance levels, but it still causes
intolerable glare. This explains why intolerable glare shows a big range in Figure 6-
22. For writing tasks, the imperceptible glare luminance range caps at 1,696 cd/m
2
.
Disturbing glare is dominant from 5,263 cd/m
2
to 11,354 cd/m
2
, and intolerable glare
occurs beyond 6,940 cd/m
2
. Since the ranges for disturbing and intolerable glare
have a big overlap, it is better to say that an absolute luminance threshold for both
disturbing and intolerable glare is 5,263 cd/m
2
. Therefore, the perceptible glare
luminance range can be defined between 1,696 cd/m
2
and 5,263 cd/m
2
. The writing
task luminance ranges are very close to the typing task luminance ranges, except for
the overlap between disturbing and intolerable glare. Based on the
disturbing/intolerable glare threshold, an absolute luminance value of 5,263 cd/m
2

can be expected to cause visual discomfort issue to occupants performing a paper-
based writing task.
230

These results show that an office activity such as computer-based typing or paper-
based writing makes occupants visually more sensitive to glare sources. They can
experience worse discomfort glare with the same glare source luminance ranges than
when no office task is involved.

Intolerable Disturbing Pe r ceptible Imperceptible
18 00 0
16 00 0
14 00 0
12 00 0
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
6940.11
15011.8
5263.48
11354
3374.03
5101.49
1237.76
1696.66
10976
8308.73
4237.76
1467.21
Figure 6-22. Glare source min luminance vs. glare category for the writing task
condition.

Similar interval plots were plotted with the data on background mean luminance,
background minimum luminance, glare source average luminance, and glare source
maximum luminance. These plots show quite similar luminance range patterns to the
glare source minimum luminance plots.
Imper c eptible g l a r e thr e shold
Distu r bing/ I nt ole r able
glar e t h r es h o ld
231

An interval plot for glare ratio shows the opposite pattern, indicating higher ratio
value ranges for imperceptible glare and gradually decreasing ratio value ranges for
perceptible, disturbing, and intolerable glare. This unexpected result occurred
because the research setting was designed to focus on discomfort glare caused by
absolute glare factor (AGF), which can cause glare with excessive brightness no
matter how high or low the glare ratios are. The research setting avoided glare
conditions from low glare source luminance and high glare ratio. Therefore, it had
wide ranges of glare source luminance ranges with quite even ratio ranges between
backgrounds and glare sources. This issue will be discussed further in the section
about relative glare factor.

The following items summarize the findings concerning absolute luminance ranges:
1. When a task is not performed, the
a. imperceptible glare luminance range caps at 2,752 cd/m
2
.

b. perceptible glare luminance range is from 2,752 cd/m
2
to

7,000 cd/m
2
.
c. disturbing glare luminance range is from 7,000 cd/m
2
to 12,522 cd/m
2
.
d. intolerable glare luminance range extends beyond 12,522 cd/m
2
.
2. When a computer-based typing task is performed, the
a. imperceptible glare luminance range caps at 1,920 cd/m
2
.

b. perceptible glare luminance range is from 1,920 cd/m
2
to

5,014 cd/m
2
.
c. disturbing glare luminance range is from 5,014 cd/m
2
to

11,718 cd/m
2
.
232
d. intolerable glare luminance range extends beyond 11,718 cd/m
2
.
3. When a paper-based writing task is performed, the
a. imperceptible glare luminance range caps at 1,696 cd/m
2
.
b. perceptible glare luminance range is from 1,696 cd/m
2
to

5,263 cd/m
2
.
c. disturbing/intolerable glare luminance range extends beyond 5,263
cd/m
2
.

These absolute luminance ranges have been utilized to define AGF. The following
sections describe the procedure that was used to generate this definition.

After the absolute luminance ranges for different office activities were successfully
defined, a series of scatter plots were then created to show how well the defined
absolute luminance ranges work with subjective evaluation on each scene. With the
glare source luminance on the x-axis and glare ratio on the y-axis, each plotted dot
represents a glare category as evaluated by subjects.

The first graph was plotted with no task scenes only (Figure 6-23). Three vertical
guidelines are drawn at 2,752 cd/m
2
, 7,000 cd/m
2
, and 12,522 cd/m
2
, to indicate the
boundary of each glare category range. Each range is color coded as follows:
imperceptible glare range in white, perceptible glare range in blue, disturbing glare
range in orange, and intolerable glare range in red. Imperceptible glare scenes are
plotted within the imperceptible glare range for the most part,

but a few scenes are
233
also found between 2,752 cd/m
2
and 7,000 cd/m
2
. Disturbing glare scenes are plotted
in all ranges. Some disturbing glare scenes are even located within the imperceptible
glare range. Intolerable glare scenes are mostly located beyond 7,000 cd/m
2
.
Perceptible glare scenes are widely scattered throughout the imperceptible to
intolerable glare ranges, thus indicating once again that perceptible glare can be
either visually comfortable or uncomfortable.

250 00 20 00 0 15 00 0 10 00 0 50 00 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Glare source minimum luminance (cd/m2)
Glare ratio
27 52 70 00 12 52 2
Impe rce ptible
Pe rce ptible
Disturbing
Intole rable
Glare ca te gory
Figure 6-23. Scatter plot of glare ratio and glare source minimum luminance for no-
task scenes.

The second scatter plot was created with data only from the typing task scenes
(Figure 6-24). In this graph, imperceptible glare scenes are most frequently located
within imperceptible glare range,

but a few scenes are also found between 1,920
cd/m
2
and 5,000 cd/m
2
. Perceptible glare scenes are found within the imperceptible
234
and disturbing glare ranges, as perceptible glare can be either comfortable or
uncomfortable. Disturbing glare scenes are mostly located in the disturbing glare
range,

but are also found in the other ranges. There are five disturbing glare scenes
plotted inside the imperceptible glare range; these scenes require further
investigation. It is possible that higher glare ratios caused more serious glare
sensations even when glare source luminance was low, or that the dark ambient
lighting environment with low glare source luminance values might have caused
visual discomfort to subjects.

25 00 0 20 00 0 15 00 0 10 00 0 50 00 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Glare source minimum luminance (cd/m2)
Glare ratio
19 20 5 0 1 4 11 71 8
Impe rce ptible
Pe rce ptible
Disturbing
Intole rable
Glare ca te gory
Figure 6-24. Scatter plot of glare ratio and glare source minimum luminance for
typing task.

A total of 153 writing task scenes are plotted in Figure 6-25 and coded by glare
category. Imperceptible glare scenes are dominantly located within the imperceptible
235
glare range,

but a few scenes are also found from 1,696 cd/m
2
to 5,263 cd/m
2
.
Disturbing glare scenes are located only within the disturbing glare range. Intolerable
glare scenes are located mostly within the disturbing/intolerable glare ranges.
Compared to the graphs depicted in Figures 6-23 and 6-24, the total number of
disturbing and intolerable glare scenes is much less with the paper-based writing task,
while the number of imperceptible glare scenes is much higher than the other two
task scenes. No disturbing or intolerable glare scene is found within the
imperceptible glare range, and only one intolerable glare scene is found outside the
disturbing/intolerable glare ranges. No imperceptible glare scene is located within
the disturbing or intolerable glare range. The absolute luminance ranges fit the
writing task scene data very well. Much as in the previous plots, perceptible glare
scenes are found from imperceptible to disturbing glare ranges.

236
250 00 20 00 0 15 00 0 10 00 0 50 00 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Glare source minimum luminance (cd/m2)
Glare ratio
16 9 6 52 63
Impe rce ptible
Pe rce ptible
Disturbing
Intole rable
Glare cate gor y
Figure 6-25. Scatter plot of glare ratio and glare source minimum luminance for
writing task.

Absolute luminance ranges for glare categories thus are found to provide very
accurate and consistent evaluations for both typing and writing task scenes.

6.1.7 Glare Source Detection with Absolute Luminance Thresholds
After the first HDR image analysis was performed in the MATLAB code by using
the multiplier 5 to detect potential glare sources in FOV, the absolute luminance
thresholds that were found from the interval plots were used to detect potential glare
sources in the captured HDR images. Potential glare sources that were detected using
either the multiplier 5 or the luminance thresholds were compared to the visual maps
that were marked up by human subjects. This procedure examined what method
237
provides more accurate glare source detection in HDR images. The entire data set
was re-analyzed with the new luminance thresholds found in the previous section.
For no-task scenes, luminance above 7,000 cd/m
2
was considered as potential glare
sources, since the luminance certainly causes disturbing or intolerable glare issues.
The luminance value of 5,000 cd/m
2
was used to detect potential glare sources in
typing and writing task scenes.

Figure 6-26 shows discrepancies in glare detection when the multiplier 5 was used
and when the new luminance threshold of 5,000 cd/m
2
was used to detect glare
sources in the same HDR images. All images shown in Figure 6-26 are typing task
scenes. The left images are visual maps that were marked up by a subject for three
different blind conditions. The red “X” marks represent the glare source locations
and sizes. The middle images are HDR images with the glare sources detected by
multiplying 5 to the background mean luminance. These glare sources are shown as a
red field. The right images are HDR images with the glare source detected by the
luminance threshold of 5,000 cd/m
2
. Both the middle and right images were created
from the MATLAB code.

238
Visual map Multiplier 5 5,000 cd/m
2

Fully open blind condition: Perceptible glare
Visual map Multiplier 5 5,000 cd/m
2

Roller blind condition: Perceptible glare
Visual map Multiplier 5 5,000 cd/m
2

Venetian blind condition: Imperceptible glare
Figure 6-26. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the typing task condition.

Glare sources detected by the new luminance thresholds match the visual maps better
than does the multiplier 5. In this example, the multiplier 5 tends to detect bigger
239
glare source sizes than what subjects experienced. The Venetian blind scene was
subjectively evaluated as having imperceptible glare, without any glare sources
marked up, but the multiplier 5 still detects large potential glare sources. The new
threshold, like the subjective evaluation, does not detect any glare sources in the
FOV.

Some might claim that potential glare detection is not a big issue as long as this
scene is evaluated to have imperceptible glare. Of course, accurate glare evaluation is
more important than getting accurate glare sizes and locations, but it is still important
to note that incorrectly detected glare sources can confuse end-users, who might
think that detected glare sources will cause glare problems even though there is
actually no glare issue. Finding a correct method to detect glare sources is thus
absolutely crucial in daylight glare analysis.

The example shown in Figure 6.26 shows that the new luminance threshold more
accurately detects glare sources than the multiplier 5. However, this does not mean
that the new luminance threshold always performs accurately for various lighting
conditions. Figure 6-27 shows a different result from the previous example, in which
neither the multiplier nor the luminance threshold accurately detects glare source
locations and sizes.

240
Visual map Multiplier 5 5,000 cd/m
2

Fully open blind condition: Intolerable glare
Visual map Multiplier 5 5,000 cd/m
2

Roller blind condition: Disturbing glare
Visual map Multiplier 5 5,000 cd/m
2

Venetian blind condition: Perceptible glare
Figure 6-27. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the typing task condition.

For fully open blind and roller blind scenes, the new luminance threshold detects
much larger areas of glare source than the subjects indicated, while the multiplier
241
detects similar or larger glare sources. The glare sources detected by the multiplier
are smaller than the luminance threshold in the fully open and roller blind scenes.
The Venetian blind scene shows the opposite condition, as the luminance threshold
detects smaller glare source sizes than the multiplier. This comparison indicates that
the new threshold exaggerates glare source sizes when the ambient light level is high.

Based on this finding, it can be assumed that the new glare threshold works well only
when ambient light levels are relatively low and the multiplier works better when
ambient light levels are relatively high. The glare source in HDR images should be
detected using a combination of both the multiplier and the luminance thresholds.
One potential advantage of using the luminance threshold would be that its
exaggeration of glare source sizes in high ambient light level conditions could
actually help to represent the seriousness of glare levels.
Similar results were found from the glare source detection comparisons for no-task
and writing task scenes. Figure 6-28 shows three no-task scenes under different blind
conditions. The fully open and venetian blind conditions scenes were subjectively
evaluated as having disturbing glare, while the roller blind condition was evaluated
as having imperceptible glare. The new luminance threshold of 7,000 cd/m
2
for no-
task scenes was used to detect glare sources in the FOV; the detected glare sources
were compared to the ones with the multiplier 5. The fully open blind condition has
disturbing glare source in the upper left corner and the multiplier 5 detected the glare
source at the same location, even though the size is slightly bigger. The luminance
242
threshold of 7,000 cd/m
2
detected glare sources in much larger areas; thus, most of
the glazing and desk areas are shown in red.

Visual map Multiplier 5 7,000 cd/m
2

Fully open blind condition: Disturbing glare
Visual map Multiplier 5 7,000 cd/m
2

Roller blind condition: Imperceptible glare
Visual map Multiplier 5 7,000 cd/m
2

Venetian blind condition: Disturbing glare
Figure 6-28. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the no-task condition.
243

Even though the glare source sizes and locations do not match the visual map
information, these large red colored areas explain how bright the interior surfaces
and the glazing were at that moment. Thus, it is possible to utilize the luminance
threshold to visualize the illumination condition of the space rather than to analyze
glare. Further research could work to develop an illumination condition analysis
method using the luminance threshold.

The roller blind scene shows a quite different result. This scene was subjectively
evaluated as having imperceptible glare, and the luminance threshold actually works
better by detecting no glare source in the FOV, except for in the small gaps between
the roller blinds. The multiplier, on the other hand, detected glare where there was
none. Thus, it is again shown that the luminance threshold works better with low
ambient lighting environments, while the multiplier works better with high ambient
lighting environments. The venetian blind condition was subjectively evaluated as
having disturbing glare. Both the multiplier and the threshold show almost the same
glare source detection, but neither of them quite matches the glare location and size
marked up on the visual map.

Figure 6-29 shows the subject responses for the writing task scene under three
different blind conditions. As subjects looked down on the paper sitting on the desk,
most of the FOV is occupied by the desk and interior surfaces, rather than by the
glazing areas. Again, the luminance threshold of 5,000 cd/m
2
was compared to the
244
analysis made by multiplier 5 in terms of glare source detection. Fully open and
roller blind scenes were subjectively evaluated as having disturbing glare and the
venetian blind scene was evaluated as having imperceptible glare. In the fully open
and roller blind scenes, the luminance threshold detects larger areas than those
marked by the subjects on the visual maps. Meanwhile, the multiplier detects smaller
areas than what subjects indicated as glare sources in the visual maps. Almost no
glare sources were detected for the venetian blind condition when using the new
threshold, but the multiplier still detected glare sources through the front and side
venetian blinds. For this writing task example, the luminance threshold performs
better than the multiplier.

245
Visual map Multiplier 5 5,000 cd/m
2

Fully open blind condition: Disturbing glare
Visual map Multiplier 5 5,000 cd/m
2

Roller blind condition: Disturbing glare
Visual map Multiplier 5 5,000 cd/m
2

Venetian blind condition: Imperceptible glare
Figure 6-29. Comparison of visual map and HDR image glare detections using
different luminance thresholds under the writing task.

Note that the human subject did not identify the visible sun in the fully open blind
condition as having glare. The human eye FOV helps to avoid detecting glare
246
sources that are on the periphery or outside the boundaries of the human eye FOV. In
the roller blind condition, this subject did not detect the severe glare sources located
at the upper right portion of full fisheye FOV while looking down on the paper.
Therefore, we may conclude that the human eye FOV should not be underestimated
or ignored in discomfort glare analysis.

The findings in this section show that glare source detection requires using both the
multiplier and the luminance thresholds for task scenes and no-task scenes. The
multiplier should be calculated first, then the calculated luminance value should be
compared to the new luminance thresholds. Between these two values, a higher
luminance value should be used to detect glare sources in HDR images.

6.1.8 Absolute Illuminance Ranges for Glare Category
Two vertical and two horizontal illuminance values were recorded at four different
locations every thirty seconds to document the interior lighting environment while
subjects were participating in tests. Horizontal illuminance values were recorded for
the light levels on the desk. One of the two vertical illuminance values was recorded
for the amount of incident light shed into the subject’s eyes; the other one was
mounted on the backside of the computer monitor, and recorded for the amount of
daylight passing through the glazing. Even though the researcher intended to utilize
absolute luminance values to define AGF, these recordings also provide valuable
information concerning whether or not vertical illuminance in the human eye can be
used to define AGF.
247

Three analyses were performed to discover the different absolute illuminance ranges
for each glare category experienced by subjects. Each graph was plotted for a
different task type. The first interval plot shows the absolute illuminance ranges for
no-task scenes only (Figure 6-30). In this condition, the illuminance range for
imperceptible glare is up to 2,086 lux. The disturbing glare illuminance range is from
12,357 lux to 28,606 lux, and intolerable glare occurs from 31,553 lux and up. Based
on the imperceptible and disturbing glare illuminance thresholds, perceptible glare
can be defined as existing between 2,086 lux and 12,357 lux. As an office
environment is normally designed with horizontal light levels between 300 to 1,000
lux at task height, the large amount of natural light coming through the glazing in
this research setting indicates a great opportunity to save lighting energy
consumption. Of course, the harvesting of natural light also increases the chance that
there will be discomfort glare issues for occupants.

248
Intolerable Disturbing Pe r ceptible Imperceptible
50 00 0
40 00 0
30 00 0
20 00 0
10 00 0
0
Glare category
Vertical illuminance (lux )
31553.8
49510.2
12357.7
28606.4
2841.96
5335.88
1366.38
2086.83
40532
20482.1
4088.92
1726.6
Figure 6-30. Vertical illuminance vs. glare category under the no-task condition.

The typing task scenes were analyzed to discover the absolute illuminance ranges for
each glare category (Figure 6-31). The imperceptible glare range caps at 1,479 lux.
The disturbing glare range is from 8,624 lux to 21,137 lux, and intolerable glare
occurs from 26,573 lux and up. Perceptible glare can be defined within the
imperceptible and disturbing illuminance thresholds from 1,479 lux and 8,624 lux.
Vertical illuminance ranges for typing task scenes are lower than the ranges for no-
task scenes. This trend was also found from the absolute luminance range
comparisons between task scenes and no-task scenes. Subjects became more
sensitive to the amount of light when performing office tasks compared to when
performing no task. This held true for the interval plots with absolute luminance
Imper c eptible g l a r e thr e shold
Int o ler a ble g l a r e t h r e s h old
Distu r bing glar e t hr es ho ld
249
ranges, as well.

Intolerable Disturbing Pe r ceptible Imperceptible
50 00 0
40 00 0
30 00 0
20 00 0
10 00 0
0
Glare category
Vertical illuminance (lux )
26573.4
40204.8
8624.75
21137.5
2810.03
4006.87
1023.24
1479.04
33389.1
14881.1
3408.45
1251.14
Figure 6-31. Vertical illuminance vs. glare category within the typing task condition.

Figure 6-32 shows the vertical illuminance ranges for the writing task scenes.
Illuminance ranges for imperceptible glare are similar between typing and writing
task scenes. However, illuminance ranges for disturbing and intolerable glare are
much lower in the writing task scenes than in the typing task scenes. The highest
illuminance range is only around 10,000 lux, which is much lower than that of the
other tasks. This is because the illuminance sensor was tilted down to mimic the
subject’s head angle and eye direction when it was recording vertical illuminance
values. Since most of the light came through the windows in front of and beside
Imper c eptible g l a r e thr e shold
Int o ler a ble g l a r e t h r e s h old
Distu r bing glar e t hr es ho ld
250
subjects, the tilted angle of the illuminance sensor drastically reduced the incident
illuminance level.

Intolerable Disturbing Pe r ceptible Imperceptible
50 00 0
40 00 0
30 00 0
20 00 0
10 00 0
0
Glare category
Vertical illuminance (lux )
4805.76
8309.5
5081.23
10401.5
2508.32
3887.06
888.178
1342.05
6557.63
7741.38
3197.69
1115.12
Figure 6-32. Vertical illuminance vs. glare category under the writing task condition.

Imperceptible glare is dominant up to 1,342 lux, while disturbing glare is dominant
from 5,081 lux to 10,401 lux. This range is much lower than what is found from the
typing task scenes. The intolerable glare range is lower than the disturbing glare
range, as it ranges from 4,805 lux to 8,309 lux. Thus, it is best to say that both
disturbing and intolerable glare are dominant beyond 4,805 lux. Based on the
thresholds of imperceptible and disturbing/intolerable glare, perceptible glare can be
defined between 1,342 lux and 4,805 lux.
Distu r bing/ I nt ole r able
glar e t h r es h o ld
Imper c eptible g l a r e thr e shold
251
No-task scenes were plotted onto the graph, which shows vertical illuminance values
on the x-axis and glare ratio values on the y-axis (Figure 6-33). Many scenes have
illuminance values above 30,000 lux. It was surprising to see the brightness of
vertical illuminance levels from natural lighting in an office context.

70 00 0 60 00 0 50 000 40 00 0 30 00 0 20 00 0 10 00 0 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Vertical illuminance (lux )
Glare ratio
12 35 7 20 86 31 55 3
Impe rce ptible
Pe r ce ptible
Disturbing
Intole rable
Glar e ca te gor y
Figure 6-33. Scatter plot of glare ratio and vertical illuminance at human eyes for no-
task condition.

The imperceptible glare range contains imperceptible and perceptible glare scenes
only. The disturbing glare range has perceptible, disturbing, and intolerable glare
scenes. The intolerable glare range has disturbing and intolerable glare scenes only.
Perceptible glare scenes are mostly seen in imperceptible glare and perceptible glare
ranges, except the one scene found in the disturbing glare range.

252
Figure 6-34 was enlarged from Figure 6-33 in order to better distinguish the
imperceptible and perceptible glare ranges. Two disturbing glare scenes are near the
threshold of the imperceptible glare range. These two scenes have almost the same
illuminance value, but have two extremely different glare ratios. It is possible that
glare ratio affected these scenes to have more serious glare issue. Further
investigation is required to find the cause of the discrepancy.

20 00 0 15 00 0 10 00 0 50 00 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Vertical illuminance (lux )
Glare ratio
12 35 7 20 86
Impe rce ptible
Pe r ce ptible
Disturbing
Intole rable
Glar e ca te gor y
Figure 6-34. Enlarged scatter plot from Figure 6-33 shows glare ratio and vertical
illuminance at human eyes for no-task condition.

Imperceptible glare scenes are mostly located at the bottom left of the chart, while
disturbing and intolerable glare scenes are located at the top right of the perceptible
glare range. This explains why a combination of high glare ratio and vertical
253
illuminance can increase the possibility that an individual will experience more
serious glare problems than one in an environment with a low glare ratio and vertical
illuminance.

Typing task scenes were also plotted with the absolute illuminance ranges for glare
categories (Figure 6-35). Much as with no-task scenes, many disturbing and
intolerable glare scenes are located within the intolerable glare range for a
moderately high amount of incident light for indoor conditions. Imperceptible glare
evaluations were not found within the disturbing or intolerable glare range.

50 00 0 40 00 0 30 00 0 20 00 0 10 00 0 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Vertical illuminance (lux )
Glare ratio
14 79 8 6 2 4 26 57 3
Impe rce ptible
Pe r ce ptible
Disturbing
Intole rable
Gla r e c a tego r y
Figure 6-35. Scatter plot of glare ratio and vertical illuminance at human eyes for the
typing task condition.

As with the previous example, the scenes inside the perceptible glare range need
254
further investigation for more accurate evaluation. As it is difficult to see the scenes
located within the imperceptible glare range, an enlarged graph was created. Figure
6-36 shows imperceptible, perceptible, and disturbing glare ranges. Perceptible glare
scenes are dominantly located in the perceptible glare range, and imperceptible glare
scenes are mostly located within the imperceptible glare range. There are three
disturbing glare scenes located in the imperceptible glare range. Further investigation
is required to understand why these scenes were subjectively evaluated as having
disturbing glare.

20 00 0 15 00 0 10 00 0 50 00 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Vertical illuminance (lux )
Glare ratio
14 79 8 6 2 4
Impe rce ptible
Pe r ce ptible
Disturbing
Intole rable
Gla r e c a tego r y
Figure 6-36. Enlarged scatter plot from Figure 6-35 shows glare ratio and vertical
illuminance at human eyes for the typing task condition.

Paper-based writing task scenes were plotted in Figure 6-37. The imperceptible glare
illuminance thresholds are similar to those of the typing and writing tasks, but
255
disturbing/intolerable glare illuminance thresholds of the writing task are much
lower than those of the typing task. There are only a few disturbing and intolerable
glare scenes evaluated with the writing task. Most disturbing and intolerable glare
scenes are located within the disturbing/intolerable glare range. The imperceptible
glare range contains imperceptible and perceptible glare scenes only. As before,
perceptible glare scenes can be found in all three glare ranges.

20 00 0 15 00 0 10 00 0 50 00 0
20 .0
17 .5
15 .0
12 .5
10 .0
7.5
5.0
Vertical illuminance (lux )
Glare ratio
48 05 1 342
Impe rce ptible
Pe rce ptible
Disturbing
Intole rable
Glare ca te gory
Figure 6-37. Scatter plot of glare ratio and vertical illuminance at human eyes for the
writing task condition.

Even though absolute illuminance thresholds cannot be used for glare source
detection, absolute illuminance thresholds work pretty well to define AGF. Thus, it
will be included for further analysis.
256

6.1.9 Absolute Glare Factor (AGF) and Relative Glare Factor (RGF)
Before the statistical analysis, it was expected that glare ratio would be one of the
major causes determining the existence of discomfort glare. It is widely accepted that
higher glare ratios can cause discomfort glare, especially in a low level lighting
environment, as explained in Chapter 2. Therefore, glare ratios have been thoroughly
investigated to define RGF.

In the first analysis of the captured HDR images, the glare ratio was calculated from
the background mean luminance and glare source average luminance, but there were
no strong correlations found between the glare ratios and different glare categories.
Therefore, the researcher calculated glare ratios between task area luminance and
glare source luminance to find a strong correlation. This approach of calculating
glare ratios can be only applied to the typing and writing task scenes, which include
task areas in the field of view. No-task scenes do not have a task area for subjects to
perform, and so did not qualify.

The second run of one-way ANOVA tests was performed using the glare ratios of
task area mean luminance and glare source average luminance. A third run of one-
way ANOVA tests was performed using the glare ratio of task area mean luminance
and glare source minimum luminance. Table 6-1 shows the statistical analysis results
from the three runs of one-way ANOV A tests.
257

Table 6-1. ANOV A test results of glare ratio and glare category.
Glare ratio Task DF P-value F value R-sq R-sq (adj) StDev
Background mean
lum to glare source
avg. lum
No 151 0.043 2.79 5.35% 3.43% 2.102
Typing 152 0.018 3.46 6.51% 4.63% 2.013
Writing 144 0.001 5.41 10.32% 8.41% 3.003
Task area mean lum
to glare source
avg. lum
Typing 152 0.000 37.09 42.75% 41.60% 18.00
Writing 144 0.019 3.43 6.80% 4.82% 7.895
Task area mean lum
to glare source min.
lum
Typing 152 0.000 37.58 43.07% 41.93% 11.18
Writing 144 0.096 2.16 4.39% 2.35% 2.461

The glare ratio between background and glare sources is not strongly correlated to
glare category for all three task scenes. In summary, the glare ratio has no influence
on perceived discomfort glare levels in the research setting of a high ambient lighting
environment with natural lighting.

However, typing task scenes show a very strong correlation between glare ratios and
glare category when the ratio is calculated either between task area mean luminance
and glare source minimum luminance or between task area mean luminance and
glare source average luminance.

Unlike typing scenes, writing task scenes still do not show a significant correlation,
even with the glare ratio calculated between task areas and glare sources. It was
258
discovered that the task areas in the writing task scenes often became a part of glare
sources, as the direct sunlight through the glazing hit the task area. When the writing
task area became part of the glare sources, the calculated glare ratio between task
area and glare source drastically decreases, which causes an insignificant correlation.
Thus, no-task and writing task scenes were not included for further analysis to define
RGF.

Among the two glare ratios with the most significant correlations, the glare ratio
between task mean luminance and glare source average luminance was excluded
when it showed a poor p-value from calculating a regression equation. The glare
ratio between task mean luminance and glare source minimum luminance was used
to define RGF. Figure 6-38 was created to describe the glare ratio ranges for typing
task scenes per glare category.

259
Intolerable Disturbing Pe r ceptible Imperceptible
50
40
30
20
10
Glare category
Glare ratio: task lum to glare source min. lum
32.0
46.0
22.3
34.5
19.4
26.2
9.1
11.9
Figure 6-38. Glare ratio of task luminance to glare source minimum luminance vs.
glare category, within the typing task condition.

Much as with the process to find out the absolute luminance range, glare ratio ranges
for different glare categories can be found from this interval plot. Three absolute
thresholds were found as shown in Figure 6-38. These glare ratio thresholds may be
used to define the intolerable glare category as being dominant from 32.0 and the
disturbing glare category as being dominant from 22.3 to 32.0. Imperceptible glare
range caps at 11.9, while perceptible glare can be defined between 11.9 and 22.3.
With the new glare ratio definition between task and glare sources, a revised scatter
plot was created with typing task scenes only (Figure 6-39). Once AGF is based on
luminance thresholds, RGF can be applied to the graph with the calculated glare
ratios.

Imper c eptible g l a r e thr e shold
Int o ler a ble g l a r e t h r e s h old
Distu r bing glar e t hr es ho ld
260
25 00 0 2 000 0 15 00 0 10 000 50 00 0
80
70
60
50
40
30
20
10
0
Glare source minimum luminance (cd/m2)
Glare ratio: task lum to glare source min. lum
Impe rce ptible
Pe rce ptible
Disturbing
Intole rable
Glare cate gor y
Figure 6-39. Scatter plot of glare ratio and glare source minimum luminance for the
typing task condition.

Figure 6-40 illustrates the four different zones of glare categories based on absolute
luminance and glare ratios. The imperceptible glare zone is defined by a luminance
range from 0 to 1,920 cd/m
2
and a ratio from 0 to 11.9. The perceptible glare zone is
defined by a luminance range from 1,920 to 5,014 cd/m
2
and a ratio from 11.9 to
22.3. The disturbing glare zone is defined by a luminance range from 5,014 to 11,718
cd/m
2
and a ratio from 22.3 to 32. The intolerable glare zone is beyond these zones.
Thus, perceptible glare can be either visually comfortable or uncomfortable, and
perceptible glare scenes are widely scattered in all four glare category zones.

261
25 00 0 2 000 0 15 00 0 10 000 50 00 0
80
70
60
50
40
30
20
10
0
Glare source minimum luminance (cd/m2)
Glare ratio: task lum to glare source min. lum
19 20
12
22 .3
32
50 14 11 71 8
Impe rce ptible
Pe rce ptible
Disturbing
Intole rable
Glare cate gor y
Figure 6-40. Imperceptible (white), perceptible (blue), disturbing (orange), and
intolerable (red) glare zones on the scatter plot of glare ratio and glare source
luminance for the typing task condition.

There are a few outliers (disturbing glare scenes) in the imperceptible glare zone and
a few imperceptible glare scenes within the disturbing glare zone. These specific
scenes require further investigation.

Figure 6-41 illustrates AGF and RGF in the graph. The red area at the bottom
represents AGF, and the yellow at the top represents RGF. AGF and RGF are the
factors that cause glare issues, and so neither is present in the imperceptible glare
zone. Although these two factors are always involved together in glare scenes, one of
them can be a more dominant factor than the other. The AGF area gradually
increases and RGF area gradually decreases as the luminance value increases. AGF
Int o ler a ble g l a r e zone Distu r bing glar e Z o n e
262
becomes a more dominant factor when the absolute luminance value increases, and .
RGF becomes a more dominant factor when the absolute luminance value decreases.
The plotted location of each glare scene indicates which glare factor is more
dominant in that instance, which allows one to determine the glare category of the
scene. When a scene falls within AGF, the scene’s glare issue is caused by an
absolute luminance value rather than a glare ratio. It is important to note that these
AGF and RGF predictions were developed only from the computer-based typing task
scenes. The no-task and writing task scenes do not have glare ratio thresholds—thus,
they only have AGF information. It is expected that similar AGF and RGF
predictions can be developed for those scenes, however. A specific RGF-focused
research setting is required to examine the relationship of RGF and AGF under the
no-task and writing task conditions.

263
Figure 6-41. AGF and RGF dominant zones for typing task.

After AGF and RGF were defined with absolute luminance and glare ratio thresholds,
regression analysis was performed to create the following regression equation.

Glare Level
for typing task
= 0.420 + 0.000232 *L
s
- 0.0253* R
t

Where L
s
is glare source luminance and R
t
is the ratio between task mean luminance
and glare source luminance. The equation has a p-value of 0.000 for both glare
source minimum luminance and glare ratio and a R-sq value of 56.9%.

The coefficient of determination became more significant when the analysis was
264
performed by weighting the factor of task area mean luminance. The weighted
analysis has a p-value of 0.000 and R-sq value of 65.0%. The final equation is as
follows:

Glare Level
for typing task
= 0.496 + 0.000244 *L
s
- 0.0310* R
t

Based on the glare level value calculated from the formula, a glare scene can be
evaluated as one of the following four categories.
 Imperceptible glare: below 0.59
 Perceptible glare: 0.59 ~ 1.03
 Disturbing glare: 1.03 ~ 2.36
 Intolerable glare: beyond 2.36

The scenes falling between imperceptible glare and disturbing glare might be
visually comfortable or uncomfortable.

Figure 6-42 was plotted to indicate the vertical illuminance and glare ratio between
task luminance and glare source luminance. The X-axis is vertical illuminance value,
while the y-axis is glare ratio.

265
50 00 0 40 00 0 30 00 0 20 00 0 10 00 0 0
80
70
60
50
40
30
20
10
0
Vertical illuminance (lux)
Glare ratio: task lum to glare source min. lum
Impe rce ptible
Pe rc e ptible
Disturbing
Intole rable
Glar e ca te gor y
Figure 6-42. Scatter plot of glare ratio and vertical illuminance for typing task.

Four different glare categories are zoned based on the absolute illuminance and glare
ratios (Figure 6-43). The imperceptible glare zone is defined as having luminance
from 0 to 1,479 lux and a ratio from 0 to 11.9. The perceptible glare zone is defined
as having luminance from 1,479 to 8,624 lux and a ratio from 11.9 to 22.3. the
disturbing glare zone is defined as having luminance from 8,624 cd/m
2
to 26,573
cd/m
2
and a ratio from 22.3 to 32. The intolerable glare zone is beyond these zones.
Unlike Figure 6-40, the glare scenes are mostly plotted on the left end of the graph.
Furthermore, there is a group of disturbing and intolerable glare scenes on the right
side of the graph.

266
50 00 0 40 00 0 30 00 0 20 00 0 10 00 0 0
80
70
60
50
40
30
20
10
0
Vertical illuminance (lux)
Glare ratio: task lum to glare source min. lum
14 79
12
86 24
22
26 57 3
32
Impe rce ptible
Pe rc e ptible
Disturbing
Intole rable
Glar e ca te gor y
Figure 6-43. Imperceptible (white), perceptible (blue), disturbing (orange), and
intolerable (red) glare zones on the scatter plot of glare ratio and vertical illuminance
for the typing task condition.

Much as in Figure 6-41, AGF and RGF are illustrated in Figure 6-44. The red area at
the bottom represents AGF, and the yellow area at top represents RGF. AGF and
RGF are the factors that cause glare issues and thus do not occur in the imperceptible
glare zone. These two factors are always involved together in glare scenes, although
one of them can be a more dominant factor than the other. The AGF area gradually
increases and the RFG area gradually decreases as the vertical illuminance value
increases. AGF becomes the more dominant factor when the absolute luminance
value increases, while RFG becomes the more dominant factor when the absolute
luminance value decreases. Based on the plotted location of each glare scene, it is
possible to know which glare factor is more dominant, while helps to determine a
Int o ler a ble g l a r e zone Distur bing glar e z one
267
glare category of the scene. When a scene falls within AGF, the scene’s glare issue is
due to a vertical illuminance value rather than a glare ratio. Again, these AGF and
RGF predictions were developed only from the computer-based typing task scenes.
No-task and writing task scenes do not have glare ratio thresholds—thus, they only
have AGF information. It is expected that similar AGF and RGF predictions can be
developed for those scenes. An RGF-focused research setting is required to
determine the relationship between RGF and AGF under the no-task and writing task
conditions.

Figure 6-44. AGF and RGF dominant zones for typing task.

Regression analysis was performed to find the regression equation from vertical
illuminance and glare ratio between task luminance and glare source luminance. The
268
equation is as follows:
Glare Level
for typing task
= 0.170 + 0.000041 *E
v
+ 0.0236* R
t

Where, E
v
is vertical illuminance at human eyes and R
t
is the ratio between task
mean luminance and glare source luminance. The equation was created with a p-
value of 0.000 and an R-sq value of 60.2%.

The coefficient of determination becomes more significant when weighted analysis is
performed with task area mean luminance. The weighted analysis has a p-value of
0.000 and an R-sq value that has increased to 68.7%.

After AGF and RGF were defined by vertical illuminance and glare ratio thresholds,
regression analysis was performed to create the following regression equation.

Glare Level
for typing task
= 0.143 + 0.000040 *E
v
+ 0.0261* R
t

Based on the glare level value calculated from the formula, a glare scene can fall into
one of the following four categories:
 Imperceptible glare: below 0.52
 Perceptible glare: 0.52 ~ 1.06
 Disturbing glare: 1.06 ~ 2.04
 Intolerable glare: beyond 2.04
269

The scenes that fall between imperceptible glare and disturbing glare can be assumed
to be either visually comfortable or uncomfortable.

These two equations can help to solve the inaccuracy and inconsistency issues of the
existing glare indices while utilizing the existing glare categories. These equations
also help to simplify the complexity hidden behind the existing glare metrics by
providing a clear understanding of the dominant factors in different glare scenes.
Table 6-2 summarizes absolute luminance and illuminance thresholds for the
different glare categories. Unlike the no-task and writing task conditions, the typing
task condition includes the defined glare ratio thresholds for different glare
categories.

270
Table 6-2. Absolute luminance and illuminance thresholds.
Luminance &
Glare Ratio
No-task Typing task Writing task
Imperceptible Below 2,752 cd/m
2

below 1,920 cd/m
2
&
below 12 glare ratio
below 1,696 cd/m
2

Perceptible 2,752 ~ 7,000 cd/m
2

1,920 ~ 5,014 cd/m
2
&
12 ~ 22 glare ratio

1696 ~ 5263 cd/m
2

Disturbing 7,000 ~ 12,522 cd/m
2

5,014 ~ 11,718 cd/m
2
&
22 ~ 32 glare ratio
beyond 5,263 cd/m
2

Intolerable beyond 12,522 cd/m
2

beyond 11,718 cd/m
2
&
beyond 32 glare ratio
Illuminance &
Glare Ratio
No-task Typing task Writing task
Imperceptible below 2,086 lux
below 1,479 lux &
below 12 glare ratio
below 1,342 lux
Perceptible 2,086 ~ 12,357 lux
1,479 ~ 8,624 lux &
12 ~ 22 glare ratio
1,342 ~ 4,805 lux
Disturbing 12,357 ~ 31,553 lux
8,624 ~ 26,573 lux &
22 ~ 32 glare ratio
beyond 4,805 lux
Intolerable beyond 31,553 lux
beyond 26,573 lux &
beyond 32 glare ratio

In summary, the goal was to develop a simplified daylight glare analysis method
without compromising analysis accuracy and consistency. The new method utilizes
only two values to help determine the existence of discomfort glare while identifying
whether AGF or RGF is the dominant glare factor. AGF is clearly defined by
absolute luminance thresholds or vertical illuminance thresholds in the no-task,
typing task, and writing task conditions. RGF is also clearly defined in typing task
scenes. AGF and RGF are always involved in any glare scene as long as a certain
271
level of discomfort glare exists. The important question for designers trying to
analyze potential glare issue is which glare factor is more dominant to a glare scene
and to what extent? Even though the method was not able to complete RGF analysis
for all task types, it is recommended that future research continue to use a method
including the two glare factors to analyze glare, as this method would greatly
improve the practicality, accuracy, and consistency of daylight glare evaluation
procedures. The daylight glare evaluation procedure would thereby be more widely
adopted in practice, and designs could more easily accommodate potential glare
issues while reducing energy consumption and improving human health through the
harvesting of natural light.

6.2 Exterior Glare Study Results
The exterior glare study was performed simultaneously with the interior glare study,
and used the same expert group. More than 84 HDR images were captured from the
exterior glare human subject study. Figure 6-45 shows six different scenes captured
by the HDR technique for a single test set. Each test included two different lighting
conditions with three different tasks per each lighting condition. The two lighting
conditions were: a) with visible glare sources in the field of view and b) without
visible glare sources in the field of view. Under these two conditions, subjects were
asked to perform three tasks: 1) do nothing; 2) read text on an iPad; and 3) read a
paper book. Immediately after each task, subjects were asked to fill in a survey form
and mark up a visual map.
272

No-task Reading iPad Reading Paper
Reflected sunlight in FOV
No-task Reading iPad Reading Paper
No reflected sunlight in FOV
Figure 6-45. Six different scenes captured from a single test.

As with the interior glare data analysis procedure, these captured HDR images were
analyzed in the MATLAB code to calculate important values such as minimum
luminance, maximum luminance, mean background luminance, glare source
luminance range, mean glare source luminance, glare ratio, and glare size. The
calculated values from the captured HDR images were then compared to the
subjective evaluations from the tests. Figure 6-46 shows those graphical outputs
processed in the MATLAB code. Each subject marked a visual map for each of the
scenes. The entire group of scenes, including all three tasks, was analyzed. Then
273
further analysis was performed for each task.

No-task Reading iPad Reading Paper
Reflected sunlight in FOV
No-task Reading iPad Reading Paper
No reflected sunlight in FOV
Figure 6-46. Six different HDR images processed in MATLAB.

6.2.1 Percentages of Subject Glare Sensations
More than 84 data sets were retrieved from the exterior glare human subject study.
The same number of HDR images was captured—one for every scene that a human
subject experienced. Using the same methodology as the interior study analysis
procedure, exterior glare data was first checked to see if human subjects experienced
various visual conditions that cover all four glare categories.
274

The percentages describing the frequency of each glare category identified by
subjects are plotted in Figure 6-47. The bar chart shows that 50% of the gathered
subjective data was evaluated as imperceptible glare and the remaining 50% was
evaluated as glare in various levels. Of all the scenes, 36% were evaluated as having
perceptible glare and 13% were evaluated as having disturbing glare. No intolerable
glare was identified by the subjects. Possible reasons for the lack of intolerable glare
are as follows: a) subjects are generous when rating exterior glare issues compared to
interior glare issues; b) reflected sunlight in an exterior glare setting may not be as
bright as the glare sources in the interior glare research setting; c) there was no
intolerable glare for subjects to experience; and d) the background light levels were
much higher in the exterior setting. Even though there were no intolerable glare
evaluations in the exterior glare study, it is important to note that subjects evaluated
disturbing glare for many scenes. Thus, exterior glare issues from specular building
envelopes do exist, even though they are not evaluated as severely as interior glare
issues.

275
Disturbing Pe r ceptible Imperceptible
50
40
30
20
10
0
Glare category
Percent
13.4146
36.5854
50
Percent within all data.
Figure 6-47. Percentages of exterior glare sensation by subjects.

The same data set was plotted in Figure 6-48 to show the percentages of subjective
evaluation on visual satisfaction levels. A seven-point Likert scale was used to assess
visual satisfaction levels. Visual satisfaction levels are well distributed to cover from
negative 3 to positive 3. Most responses are shown at negative 1, positive 1, or
positive 2. Negative 3 received 3.6% of total responses, but subjects still did not
consider these scenes to have intolerable glare.

276
3 2 1 0 - 1 - 2 - 3
50
40
30
20
10
0
How satistied?
Percent
9.7561
20.7317
19.5122
8.53659
24.3902
13.4146
3.65854
Percent within all data.
Figure 6-48. Percentages of subjects’ visual satisfaction levels.

The percentages of different visual comfort levels experienced by subjects are shown
in Figure 6-49. This chart was expected to be similar to the previous graph, since
visual satisfaction was expected to be similar to visual comfort. As shown in Figures
6-48 and 6-49, there are small discrepancies between the percentages of satisfaction
and comfort levels. The percentage of negative 2 scores is lower in visual comfort
than in visual satisfaction, while the percentage of neutral scores is higher. Both
graphs show that subjects experienced various visual satisfaction and comfort levels.

277
3 2 1 0 - 1 - 2 - 3
50
40
30
20
10
0
How comfortable?
Percent
10.9756
20.7317
19.5122
12.1951
25.6098
8.53659
2.43902
Percent within all data.
Figure 6-49. Percentages of subjects’ visual comfort levels.

During the tests, subjects were also asked to assume that this exterior research setting
was their daily rest place, and they were asked to evaluate their visual comfort levels
according to this assumption. The collected responses are quite different from the
previous two figures. The strongly uncomfortable category (-3) disappeared. The
negative 1 and 2 scores increased, while the percentage of neutral scores decreased.
The score percentages are shown in Figure 6-50.

278
3 2 1 0 - 1 - 2
50
40
30
20
10
0
Ho w co mfo rtable if d aily re st?
Percent
9.7561
19.5122 19.5122
4.87805
31.7073
14.6341
Percent within all data.
Figure 6-50. Percentages of subjects’ visual comfort levels when they assumed the
area was a daily resting place.

Various questions were asked to subjects to evaluate a same visual condition.
Subjects were able to experience various visual comfort and satisfaction levels from
the study. The following section compares the subjective responses to these various
questions.

6.2.2 ANOVA Test: Subjective Response Consistency
Glare sensation was evaluated on a bar scale by subjects, rather than on the seven-
point Likert scale used for visual satisfaction and comfort questions. Therefore, it
was necessary to convert the subjective bar scale evaluations into the existing four
279
different glare categories. Since this procedure was already performed for the interior
glare data analysis, the glare level ranges found from the interior glare study were
used for the exterior glare categories. Imperceptible glare ranged from 0 to 0.3,
perceptible glare ranged from 0.4 to 1.4, and disturbing glare ranged from 1.5 to 2.5.
After categorizing the glare levels, the researcher performed a one-way ANOVA test
to compare glare levels and glare categories (Figure 6-51). Mean values for
imperceptible, perceptible, and disturbing glare are 0.05, 0.82, and 1.89. P-value is
0.000 with 95% confidence, and the coefficient of determination value is 90.83%,
which supports the idea that this glare level categorization will be effective for most
future data. This analysis confirms that glare levels were successfully categorized,
thus allowing for further analysis using the four glare sensation categories.

280
Disturbing Pe r ce ptible Imperceptible
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Glare category
Glare lev el
1.89091
0.826667
0.0560976
Figure 6-51. One-way ANOV A comparing glare level and glare sensation category
(CI = 95%, F = 391.25, P = 0.000, R-sq = 90.83%, R-sq(adj) = 90.60%, StDev =
0.2022)

The consistency of subjective responses to various questions was first checked prior
to the analysis of the captured HDR images from the exterior glare study. A one-way
ANOVA test was performed to check whether or not subjects’ answers concerning
visual satisfaction and comfort are consistent with their discomfort glare rating for
the same exterior scene. If ANOVA test results prove the consistency, subjective
glare ratings can be analyzed further using the captured HDR images. This procedure
can assure that each subject is using a consistent set of standards to evaluate visual
satisfaction, comfort, and discomfort glare issue. Subjects’ answers on visual
satisfaction and glare sensation were compared. Figure 6-52 shows that subjective
281
responses on glare sensation categories are not consistent with subjects’ visual
satisfaction levels. P-value is 0.091 with 95% confidence and the coefficient of
determination value for this comparison is 5.89%. Visual satisfaction for
imperceptible glare ranges from -2 to 3. The middle 50% of imperceptible glare
scenes ranges from -1 to 2. This shows that subjects were not always visually
satisfied, even the when discomfort glare level was imperceptible. Subjects evaluated
perceptible glare with visual satisfaction levels between -3 and 3. Again, the middle
50% of perceptible glare scenes ranges from -2 and 1. This was expected from the
interior glare study analysis, which showed both negative and positive visual
satisfaction in cases of perceptible glare. Disturbing glare ranges from -3 to 3, while
the middle 50% ranges from -2 to 0. This shows that 75% of disturbing glare scenes
were visually dissatisfying to subjects. Subjects did not experience any intolerable
glare during the exterior glare tests. It certainly indicates that the exterior glare data
does not have a similar pattern to the interior glare data. Furthermore, the exterior
glare evaluations were not consistent, which also differed from the interior glare area.

282
Disturbing Pe r ceptible Imperceptible
3
2
1
0
- 1
- 2
- 3
Glare category
How satistied?
- 0.363636
- 0.0333333
0.682927
Figure 6-52. One-way ANOV A: Visual satisfaction vs. glare sensation (CI = 95%,
F = 2.47, P = 0.091, R-sq = 5.89%, R-sq(adj) = 3.50%, StDev = 1.693).

Visual comfort levels were also compared to glare categories to verify whether or not
surveyed glare level evaluations can be trusted for further analysis. Figure 6-53
compares subjects’ answers to visual comfort level and glare sensation category
questions for a same scene. P-value is 0.004 with 95% confidence, and the
coefficient of determination value for this comparison is 13.14%. This value
indicates that populations in this dataset are not significantly different for statistical
analysis.

283
Disturbing Pe r ceptible Imperceptible
3
2
1
0
- 1
- 2
- 3
Glare category
Ho w co mfo rtable?
- 0.818182
0.233333
0.926829
Figure 6-53. One-way ANOV A: Visual comfort level vs. glare category (CI = 95%,
F = 5.98, P = 0.004, R-sq = 13.14%, R-sq(adj) = 10.94%, StDev = 1.542).

The visual comfort range for imperceptible glare is between -2 and 3 while the
middle 50% sits between -1 and 2. It is unusual that almost 50% of imperceptible
glare scenes were still visually uncomfortable to subjects. Perceptible glare was
indicated between visual comfort levels -3 and 3 with the middle 50% between -1
and 1. Disturbing glare sits between visual comfort level -3 and 1, with the middle 50%
between -2 and 0. These outcomes are very similar to the comparison of visual
satisfaction and glare categories.

Figure 6-54 compares subjects’ glare sensation categorization to their visual comfort
levels when they assume that the exterior research setting is their daily resting place.
284
P-value is 0.000 with 95% confidence and the coefficient of determination value for
this comparison is 20.50%. The visual comfort range for imperceptible glare is
between -2 and 3, while the middle 50% sits between -1 and 2. Perceptible glare is
between visual comfort levels -2 and 2 with the middle 50% between -1 and 1.
Disturbing glare sits between visual comfort levels -2 and 1 while the middle 50%
sits between -2 and 0. It is difficult to make consistent definitions of each glare
category based on the visual comfort assessments under the assumption that the
scene is a daily resting place.

Disturbing Pe r ceptible Imperceptible
3
2
1
0
- 1
- 2
- 3
Glare category
Ho w co mfo rtable if d aily rest?
- 1.09091
- 0.166667
0.95122
Figure 6-54. One-way ANOV A: Visual comfort level with daily rest place
assumption vs. glare category (CI = 95%, F = 10.19, P = 0.000, R-sq = 20.50%,
R-sq(adj) = 18.49%, StDev = 1.488).

285
These ANOVA tests deliver the following two findings: a) subjective responses for
imperceptible and perceptible glare are inconsistent and b) subjects did not
experience intolerable glare issues from the exterior glare study. These findings are
unexpected and currently unexplained. It is possible that subjects expect and tolerate
much wider ranges and higher absolute values in exterior environments.

6.2.3 Field of View: Full Fisheye vs. Human Eye
Much as with the interior study analysis, the researcher checked the exterior glare
scenes to see whether full fisheye or the human eye FOV would provide more
accurate glare analysis. The interior study indicated that the full fisheye FOV works
better for no-task scenes, while the human eye FOV works better with typing and
writing task scenes. Figure 6-55 shows the total frequency of glare source detection.
Out of thirty-six scenes, twenty-three were detected to have glare sources by subjects
under the no-task condition. Out of these twenty-three glare scenes, eighteen show
identical glare source detections between full fisheye and human eye FOVs. For the
remaining five glare scenes, the full fisheye FOV provides more accurate glare
source detections than the human eye FOV .

286
Glare_Sour ces_Detected No_Glare_Sources_Detected
25
20
15
10
5
0
Fre q ue ncy
FE
Both
Vie w
Fie ld of
23
18
13 13
Figure 6-55. Total frequency of field of view matching glare source detection during
the no-task condition.

For the iPad reading task scenes, a total of twenty-three scenes were analyzed, and
twelve scenes have glare sources detected by human subjects (Figure 6-56). Out of
those twelve glare scenes, four have accurate glare source detections in both human
eye FOV and full fisheye FOV. The other eight were not accurately analyzed with
full fisheye FOV. This data can be interpreted to indicate that the human eye FOV is
more accurate than a full fisheye FOV for iPad reading task scenes in an external
environment.

287
Glare_Sour ces_Detected No_Glare_Sources_Detected
25
20
15
10
5
0
Fre q ue ncy
HE
Both
Vie w
Fie ld of
12
4
11 11
Figure 6-56. Total frequency of glare source detection outside human eye FOV under
the iPad reading task condition.

For the paper reading task scenes, eleven out of twenty-three scenes have no glare
sources detected by subjects (Figure 6-57). A total of twelve scenes have glare
sources detected by subjects. Among the twelve, seven scenes were accurately
detected with the human eye FOV. Thus, it seems more appropriate to use the human
eye FOV rather than full fisheye FOV for paper reading task scenes.

288
Glare_Sour ces_Detected No_Glare_Sources_Detected
25
20
15
10
5
0
Fre q ue ncy
HE
Both
Vie w
Fie ld of
12
5
11 11
Figure 6-57. Total frequency of glare source detection outside human eye FOV under
the paper reading task condition.

Both interior and exterior data show that human eye FOV is more appropriate for
task scenes than full fisheye FOV. Although the research plan was not designed to
investigate the field of view issue for daylight glare analysis, fairly concrete results
were found to support the idea that the human eye FOV is more accurate. Further
study is required to analyze the field of view issue more thoroughly.

This exterior glare study utilized an FOV in the MATLAB code to exactly match
what subjects marked up on the visual maps. The following section compares the
visual maps to the processed HDR images in the MATLAB code to find out the best
way to detect glare sources in a scene.
289

6.2.4 Glare Source Detection in Visual Map and HDR Image
The researcher compared the captured HDR images to the surveyed visual maps to
check whether or not the glare detections in the HDR images matched the subjects’
glare detections. Potential glare sources can be detected by assessing the captured
HDR images or simulated PIC files using a relative luminance threshold that is five
times the background mean luminance value. It is also possible to use an absolute
luminance value to detect glare sources, instead of using a multiplier.

The researcher compared the glare detection accuracy of no-task, iPad reading task,
and paper reading task scenes. Figure 6-58 compares the visual maps and the
processed HDR images for the scenes without any task. The top images show an
FOV of a subject sitting in a chair facing the south windows, which reflected
sunlight toward the subject. The subject detected two glare sources: one on the
windows and the other on the top of the table. This particular scene was evaluated as
having perceptible glare.

290
Human subject HDR image detection
Reflected sunlight in FOV: Perceptible glare
Human subject HDR image detection
No reflected sunlight in FOV: Imperceptible glare
Figure 6-58. Glare source detection example in the no-task condition.

The bottom images show an FOV that is 90-degree oriented toward the right from
the first field of view. This FOV was intended not to include any directly reflected
sunlight from the windows. Subjects evaluated this scene as having imperceptible
glare. For the FOV with reflected sunlight, the multiplier 5 accurately detected glare
sources that match the glare sources marked up on the visual map, even though glare
source sizes are somewhat different. The FOV without reflected sunlight shows that
the processed HDR image detects completely different glare sources than did the
291
subjects. The processed image shows large potential glare sources on the exterior
walls, whereas nothing was marked on the visual map. This discrepancy in glare
source detection was also found throughout the interior glare study. Glare sources are
nearly always detected in the HDR images when using the multiplier, regardless of
the existence of discomfort glare. These automatically detected glare sources do not
necessarily indicate that there is a discomfort issue, but they might confuse users into
thinking that there is. As the interior glare study found an effective combined method
using both the multiplier and the absolute luminance thresholds to assess glare, this
method can be also applied to the exterior glare scenes.
Figure 6-59 compares two lighting conditions for the iPad reading task. This task has
a slightly different FOV from the no-task scene, as subjects were looking down on an
iPad sitting on the table. Right after performing the reading task, subjects were asked
to indicate glare sources directly on the iPad, and were thus able to maintain the
same field of view for the iPad reading task. The FOV with reflected sunlight was
evaluated as perceptible glare, and glare sources were detected in the middle of the
table and on the iPad screen. The processed HDR image shows a glare source on the
table, but it also shows a big glare source on the exterior wall of the building. The
FOV without reflected sunlight was also evaluated as having perceptible glare, even
though no glare source was marked up in the visual map. After the test, the
participant described that it was uncomfortable to read due to a veiling reflection on
the iPad. It is possible that this veiling reflection issue might have caused
inconsistent subjective responses in the exterior glare study.

292
Human subject HDR image detection
Reflected sunlight in FOV: Perceptible glare
Human subject HDR image detection
No reflected sunlight in FOV: Perceptible glare
Figure 6-59. Glare source detection example under the iPad reading task condition.

Figure 6-60 compares the visual maps and the captured HDR images for the paper
reading task under two different lighting conditions. Subjects were asked to draw red
circles or lines on the printed visual maps. The FOV for the paper reading task is
same as the FOV for the iPad reading task, as both FOVs look down on the table.
Both scenes were evaluated as having perceptible glare. The visual map and the
processed HDR image show similar glare locations for the reflected sunlight. For the
FOV without the reflected sunlight, glare detection in the HDR image does not
293
match that of the visual map. This scene was also evaluated as having perceptible
glare, but the causes of the glare are unknown.

The exterior glare study has many unknown causes and results that cannot be solved
using the collected data. For example, the multiplier 5 does not work in all lighting
conditions.

Human subject HDR image detection
Reflected sunlight in FOV: Perceptible glare
Human subject HDR image detection
No reflected sunlight in FOV: Perceptible glare
Figure 6-60. Glare source detection example under the paper reading task condition.

294
To summarize, the potential glare source locations in the visual map match those of
the automatically detected glare source locations in the HDR images when there is a
strong glare source in the field of view. However, glare source sizes do not always
match between the visual maps and the HDR images. As found in the interior study,
both the multiplier and luminance threshold should be utilized to give users more
accurate graphical representations of glare sources.

6.2.5 ANOVA Test: Glare Source Minimum Luminance
An ANOVA test was performed to check whether or not calculated luminance values
from the captured HDR images are correlated with subjects’ glare evaluations, and
also to check whether absolute luminance ranges can be defined for different glare
categories. Figure 6-61 shows the results of a one-way ANOVA test performed to
compare subjects’ glare sensation levels and glare source minimum luminance values.
Glare sensation level was collected from the surveys and the data was converted to
numeric values ranging from 0.0 to 3.0. The glare source minimum luminance ranges
vary widely for different glare levels, and even mean values show a random pattern,
even with gradually increasing glare sensation levels (Figure 6-61). P-value is 0.311
with a 95% confidence level. This means that there will be 31% random data outside
the pattern shown on the graph. The coefficient of determination value for this
comparison is 25.09%. This indicates that the populations in this ANOV A test are not
significantly different for statistical analysis.

295
2.5 2.1 2.0 1.9 1.7 1.6 1.5 1.2 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
10 00 0
80 00
60 00
40 00
20 00
0
Glare lev el
Glare source minimum luminance (cd/m2)
Figure 6-61. One-way ANOV A: Glare source minimum luminance vs. glare level
(CI = 95%, F = 1.17, P = 0.311, R-sq = 25.09%, R-sq(adj) = 3.69%, StDev = 2648).

Once glare sensation levels were defined into the three different glare categories,
another one-way ANOVA test was performed to compare glare categories and glare
source minimum luminance values calculated from the captured HDR images. Figure
6-62 shows three different luminance ranges for three different glare categories. The
mean values of the three luminance ranges linearly increase as glare becomes more
intense. However, p-value is still 0.218 with a 95% confidence level, and the
coefficient of determination value is only 3.78%, which is considered to be very low.
Based on the two ANOV A tests, the exterior glare data appears to show no consistent
pattern for statistical analysis.

296
Disturbing Pe r ceptible Imperceptible
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
Figure 6-62. One-way ANOV A: Glare source minimum luminance vs. glare
sensation category (CI = 95%, F = 1.55, P = 0.218, R-sq = 3.78%, R-sq(adj) = 1.34%,
StDev = 2680).

6.2.6 Absolute Luminance Ranges for Glare Category
Even though the exterior glare study data does not show a significant difference for
the different glare categories, the subjective responses were compared to the captured
HDR images. The captured HDR images were analyzed in the MATLAB code, and
the following calculated values were compared to subjective data: glare source
minimum luminance, glare source average luminance, glare ratio, and background
minimum, average, and maximum luminance values. In addition, vertical and
horizontal illuminance values recorded during the tests were compared to subjective
evaluation data.
297
Among the calculated values from the captured HDR images, glare source minimum
luminance value was compared to the glare categories, and a series of graphs were
plotted to show the luminance ranges that cause different discomfort glare issues.
The resulting interval plots show three different glare source minimum luminance
ranges for each glare category. Again, no intolerable glare was subjectively evaluated
from the exterior glare study. As expected from the ANOVA test, the glare source
minimum luminance ranges are not significantly different from each other. Thus, it is
difficult to define the absolute luminance range for each glare category. Based on the
entire data set, including no-task, iPad reading task, and paper reading task
conditions, imperceptible glare is dominant up to 5,480 cd/m
2
and perceptible glare
ranges from 4,466 cd/m
2
to 6,387 cd/m
2
. Disturbing glare is dominant from 4,524
cd/m
2
and up. As shown in Figure 6-63, the luminance ranges for different glare
sensations have large overlaps that make the difference less significant.

298
Disturbing Pe r ceptible Imperceptible
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
4524.11
7356.31
4466.7
6387.64
3664.9
5480.78
Figure 6-63. Glare source minimum luminance vs. glare category under all task
conditions.

After checking the entire exterior glare study dataset, three interval plots were
created—one for each of the three different task conditions. Figure 6-64 was plotted
only with no-task scenes. Imperceptible glare is dominant up to 8,487 cd/m
2
, while
perceptible glare is dominant from 7,189 cd/m
2
to 8,682 cd/m
2
. Disturbing glare is
dominant from 6,282 cd/m
2
to 8503 cd/m
2
. It is not possible to define absolute
luminance ranges for glare categories, since the three ranges all overlap.

299
Disturbing Pe r ceptible Imperceptible
10 00 0
80 00
60 00
40 00
20 00
0
Glare category
Glare source minimum luminance (cd/m2)
6282.59
8503.51
7189.28
8682.26
5385.18
8487.55
Figure 6-64. Glare source minimum luminance vs. glare category under the no-task
condition.

Figure 6-65 shows the results from the iPad reading task scenes only. The
imperceptible glare is dominant up to 2,836 cd/m
2
, the perceptible glare range is
from 1,987 cd/m
2
to 3,240 cd/m
2
, and the disturbing glare range is from -7,882 cd/m
2
to 14,458 cd/m
2
. The perceptible glare luminance range is almost the same as the
imperceptible glare luminance range. These luminance ranges are not significantly
different for different glare categories; thus, it is not possible to define absolute
luminance thresholds based on the results of the exterior study.

300
Disturbing Pe r ceptible Imperceptible
15 00 0
10 00 0
50 00
0
- 50 00
- 10 00 0
Glare category
Glare source minimum luminance (cd/m2)
- 7882.98
14458.6
1987.5
3240.56
1990.15
2836.33
Figure 6-65. Glare source minimum luminance vs. glare category under the iPad
reading task condition.

Finally, Figure 6-66 was created only for the paper reading task scenes. This plot is
similar to the previous plot of iPad reading task scenes. The imperceptible glare
range is up to 4,063 cd/m
2
, the perceptible glare luminance ranges from 3,591 cd/m
2

to 5,439 cd/m
2
, and the disturbing glare range is dominant from 265 cd/m
2
to 9,339
cd/m
2
. Luminance ranges for imperceptible, perceptible, and disturbing glare all
overlap; thus, the absolute luminance thresholds cannot be defined.

301
Disturbing Pe r ceptible Imperceptible
15 00 0
10 00 0
50 00
0
- 50 00
- 10 00 0
Glare category
Glare source minimum luminance (cd/m2)
265.618
9339.96
3591.07
5439.87
2704.95
4063.29
Figure 6-66. Glare source minimum luminance vs. glare category under the paper
reading task condition.

Similar graphs were plotted to define the three glare categories in terms of calculated
background mean luminance, background minimum luminance, glare source average
luminance, glare source maximum luminance, and glare ratio. The vertical
illuminance ranges are not significant enough to define glare categories. Since none
of these values shows significant differences for the glare categories, they are not
discussed in this chapter.

6.3 Analysis Summary
Subjective data for interior and exterior glare studies were surveyed and analyzed for
302
different glare categories for the interior glare study only, not for the exterior glare
study. For interior glare conditions, imperceptible glare thresholds are 2,752 cd/m
2

without a task and 1,920 cd/m
2
with a task. Disturbing glare thresholds are 7,000
cd/m
2
without a task and 5,000 cd/m
2
with a task. Perceptible glare is defined as
existing between the imperceptible and disturbing glare thresholds. The intolerable
glare threshold is above 12,522 cd/m
2
without a task and 11,718 cd/m
2
with a task.
The glare ratio between computer typing task luminance and glare source minimum
luminance showed significant difference for the glare categories, as imperceptible
glare is up to 11.9, perceptible glare is from 11.9 to 22, disturbing glare is from 22 to
32, and intolerable glare is beyond 32.

AGF and RGF dominant zones were illustrated based on absolute luminance and
illuminance thresholds and the glare ratios between task luminance and glare source
minimum luminance. Two glare equations were developed using AGF and RGF to
evaluate discomfort glare issues by using two different variables, such as absolute
luminance and glare ratios or absolute illuminance and glare ratios.

The exterior glare study did not provide statistically significant data to define
absolute luminance and illuminance thresholds or glare ratios. Therefore, the new
daylight glare evaluation method using AGF and RGF is currently applicable to
interior glare issues only, and not to exterior glare issues. Even though the exterior
glare study did not provide strong data, it still explains what caused inconsistent and
303
random data from the human subject study and suggests the issues that could be
fixed to create a more accurate research methodology for future exterior glare study.

304
Chapter 7 Conclusions and Future Work
AGF- and RGF-Based Daylight Glare Analysis Method

7.1 Conclusions
Indoor and outdoor daylit spaces were investigated to verify the process and results
of discomfort glare issues inside and outside building envelopes. Two separate
human subject studies were performed—one in interior and one in exterior daylit
spaces. Many interesting findings were deduced from the interior glare study, while
the exterior study did not provide any concrete findings.

Three primary objectives were pursued:
1. Examine existing discomfort glare metrics and tools in terms of consistency,
accuracy, and practicality to understand why they are not often used in
daylighting practice.
2. Define absolute glare factor (AGF) and relative glare factor (RGF) based on
human subject study results.
3. Develop a consistent and accurate daylight glare analysis method using AGF
and RGF that can overcome the consistency, accuracy, and practicality issues
of existing glare indices.

The existing glare indices were thoroughly investigated using computer simulations,
HDR image technology, and human subject surveys. The studies carefully compared
calculated glare evaluations to human subject data. The comparisons clearly showed
305
that the existing glare indices are inconsistent and inaccurate. The researcher also
found that the existing metrics and tools do not determine the dominant factor (AGF
or RGF) causing discomfort glare, when this knowledge is critical if one wishes to
avoid or fix the glare issue. The findings of the human subject studies provided a
better understanding of each existing glare index. Thus, that portion of the study
indicates the areas for improvement of each glare index for more accurate evaluation
performance.

The second objective was achieved by successfully defining two new glare
definitions: absolute glare factor and relative glare factor. The development of AGF
and RGF was the main objective, and the human subject study data successfully
helped to define these two glare factors. The various absolute luminance thresholds
that define different glare categories were found for various office tasks. When no
task was being performed inside the building, subjects reported no glare up to 3,000
cd/m
2
and disturbing glare above 7,000 cd/m
2
. When subjects performed a computer
typing task inside the building, they reported no glare up to 1,920 cd/m
2
and
disturbing glare above 5,014 cd/m
2
. When subjects performed a paper writing task
inside the building, they reported no glare up to 1,696 cd/m
2
and disturbing glare
above 5,263 cd/m
2
. Furthermore, various absolute illuminance thresholds at the
human eye were found to define different glare categories for various office tasks.
When no task was being performed inside the building, subjects reported no glare up
to 2,086 lux and disturbing glare above 12,357 lux. When subjects performed a
306
computer typing task inside the building, they reported no glare up to 1,479 lux and
disturbing glare above 8,624 lux. When subjects performed a paper writing task
inside the building, they reported no glare up to 1,342 lux and disturbing glare above
4,805 lux.

Glare ratio thresholds were also defined for the computer based typing task.
Imperceptible glare is dominant up to 12:1, while disturbing glare is dominant
beyond 22:1. Using these thresholds, the researcher illustrated imperceptible,
disturbing, and intolerable glare zones on a graph of luminance. Any glare scene can
be plotted onto the graph to visually represent what glare category zone the scene
falls into.

Furthermore, AGF and RGF were graphically represented to show which glare factor
is more dominant, which helps to determine what glare category is present.
Ascertaining a dominant glare factor will assist our understanding of the causes
hidden behind the complicated formulas of the existing glare metrics and help to
avoid the issue of discomfort glare inside buildings.

307
Figure 7-1. AGF and RGF dominant zones.

The third objective was achieved by developing the AGF/RGF-based glare analysis
method, which is simple enough to be practical but still provides high evaluation
accuracy and consistency. Two equations were successfully developed: one using
luminance and glare ratio and the other using vertical illuminance and glare ratio. It
is believed that this new glare analysis method can simplify daylight glare evaluation
procedures by skipping the complicated calculation process, which is often quite
difficult to understand. This new method can be incorporated into existing building
standards and design guidelines; thus, it can make a big impact on building energy
and occupant comfort. In fact, several glare scenes otherwise falling onto the
imperceptible glare zone are evaluated as disturbing glare, even with low luminance
values and low contrast ratios. There must be other unknown factors or variables that
308
were not considered, especially for the daylit environments with low luminance and
a low contrast ratio. It is believed that these other factors can be revealed through
future studies. The defined thresholds can be compared to the previously developed
thresholds in order to improve the accuracy and consistency of the existing indices.

The equation for glare that combines the two factors is as follows:

Glare Level
for typing task
= 0.496 + 0.000244 *L
s
- 0.0310* R
t

Where L
s
is glare source luminance and R
t
is the ratio between task mean luminance
and glare source luminance.

Based on the glare level value calculated from the formula, a glare scene can be
evaluated as belonging to one of the following four categories.
 Imperceptible glare: below 0.59
 Perceptible glare: 0.59 ~ 1.03
 Disturbing glare: 1.03 ~ 2.36
 Intolerable glare: beyond 2.36

In addition to the achievements explained above, the analysis procedure provided
other contributions, including the following:
 It indicated that occupants have different visual sensitivity when performing
309
different office activities.
 It supported the need to use the human eye FOV in glare analysis.
 It defined the meaning of perceptible glare based on statistical analysis.
 It investigated the existing glare indices using human subject study data and
suggested how to improve the existing metrics.
 It initiated an exterior glare human subject study and proved that specular
building envelopes can cause exterior visual discomfort issues.
 It suggested a new human subject research method using an expert group and
visual maps.

7.1.1 Interior Glare: Visual Sensitivity for Different Office Activities
Another crucial contribution from the interior glare study is its finding that occupants
experienced different visual sensitivity for different office activities. The same
luminance value can cause different glare sensations depending on the tasks
performed inside an office space. When subjects did not perform any task, luminance
thresholds for different glare sensations are as follows: 2,752 cd/m
2
for imperceptible
glare, 7,000 cd/m
2
for disturbing glare, and 12,522 cd/m
2
for intolerable glare. When
subjects performed a typing task viewed on computer monitor, luminance thresholds
are as follows: 1,920 cd/m
2
for imperceptible glare, 5,000 cd/m
2
for disturbing glare,
and 11,718 cd/m
2
for intolerable glare. Much as with the typing task, luminance
thresholds for the writing task are as follows: 1,696 cd/m
2
for imperceptible glare
and 5,263 cd/m
2
for disturbing/intolerable glare. Comparisons among these
310
luminance thresholds clearly support that subjects became more visually sensitive to
light when they were performing an office task. For example, a glare source
luminance of 6,000 cd/m
2
causes perceptible glare when performing no task, while
the same luminance value causes disturbing glare when performing a typing or
writing task.

The vertical illuminance thresholds also support the idea that occupants are more
visually sensitive when office activity is performed in a daylit space. It certainly
shows that daylight analysis procedures should consider various indoor activities
separately, in order to avoid underestimating or overestimating daylight glare issues
inside a building.

7.1.2 Interior Glare: Total FOV vs. Human Eye FOV
The field of view issue has been frequently discussed in regards to discomfort glare
analysis. However, the importance of FOV has been somewhat underestimated due
to the focus on simplifying and increasing the speed of use for glare analysis
procedures. Furthermore, it is not easy to discover the extent to which full fisheye
FOV can reduce the accuracy of glare analysis performance compared to human eye
FOV. It was not the researcher’s original intent to investigate this issue, but it was
possible to utilize the captured HDR images and the surveyed visual maps to look
into the FOV issue. It was found that both full fisheye FOV and the human eye FOV
perform well to detect glare sources from no-task scenes, while the human eye FOV
311
better detects glare sources from typing and writing task scenes. This finding shows
that the human eye FOV might be better to use for glare analysis if only one FOV
can be assessed. The typing and writing task scenes certainly seem to require the
human eye FOV to correctly detect glare sources. Considering a glare source that
cannot be seen or felt by occupants would create excess labor for designers to create
solutions to a problem that does not exist. Some people might claim that using the
human eye FOV can make glare analysis more complicated, but everyone should
keep in mind that there is a certain impact from using the full fisheye FOV in some
contexts.

7.1.3 Interior Glare: The Meaning of Perceptible Glare Sensation
Statistical analysis was performed to determine the meaning of different glare
categories. Based on the comparison between visual comfort and glare levels, it was
found that imperceptible glare is always visually comfortable to occupants. Also, it
was found that 100% of disturbing and intolerable glare scenes were visually
uncomfortable to occupants. Perceptible glare, on the other hand, showed visual
comfort levels evenly distributed between visual comfort and visual discomfort.
Similar results were found from the comparison between visual satisfaction and glare
levels. The comparison between glare levels and visual comfort with the assumption
that the location was a daily workplace showed similar results. It might be possible
to assume that the meaning of perceptible glare is visual discomfort in a noticeable
but tolerable level. However, this assumption is still too vague to make an important
312
decision in daylighting practice. When a daylit space is evaluated as having
perceptible glare, it becomes tricky for designers to determine whether the
perceptible glare should be avoided or whether it can remain as is without disturbing
potential occupants. Therefore, it is very important to define the real meanings of
different glare categories to provide scientific explanations. This was achieved by
statistical analysis of the human subject study. The findings can help designers to
make good design decisions in terms of discomfort glare.

7.1.4 Interior Glare: Validations of Existing Glare Indices
Another big contribution of the study is its extensive examination of the existing
glare metrics and the glare analysis tool Evalglare. Since the existing glare metrics
and standards are not consistent, they are of little help to daylighting design decision
makers. Developing a new glare metric might simply add another inconsistent metric,
and could worsen the inconsistency issue among different metrics. Therefore, the
researcher used the collected human subject data to carefully examine the existing
metrics prior to developing a new method. It was hoped that this validation
procedure would help improve the existing metrics to provide more consistent
evaluations, or at least provide some clues to solve the consistency issue. The human
subject study data was large enough and valuable statistically to attempt to validate
the existing glare indices.

After the validation procedure, it was found that DGP shows the most accurate
313
evaluation for interior daylit space with the various blind conditions only when
vertical illuminance values measured by an illuminance meter are used. Without the
measured illuminance, DGP shows very low accuracy. DGI and VCP perform the
second best among the five glare indices, and the study shows that they work better
for a specific lighting condition only. DGI showed good accuracy for imperceptible
and perceptible glare scenes. This means that DGI works better with a low ambient
lighting environment that lacks visible glare sources. VCP shows good accuracy for
imperceptible, disturbing, and intolerable glare scenes. Also, VCP shows even higher
accuracy when using the human eye FOV. CGI and UGR show the lowest accuracy
among the five indices, even though they have relatively higher accuracy for
disturbing or intolerable glare scenes. The results indicate what index performs better
than the others on any given lighting condition. It is critical that users of the glare
indices correctly understand what index they should utilize for different daylit spaces.

Glare scores were calculated for each glare index, and the scores were
statistically analyzed to attain glare score ranges for each glare category. The
calculated glare scores were then compared to the existing glare score ranges.
The results show that the existing glare score ranges cannot accurately
differentiate perceived glare levels. In CGI, UGR, and DGI, the existing glare
score ranges are completely different from what was calculated from the human
subject data. On the other hand, the existing glare score ranges of DGP and VCP
match well to the calculated glare score ranges. VCP shows closely matching
314
score ranges for disturbing and intolerable glare, while DGP shows a very good
match for imperceptible, perceptible, and disturbing glare score ranges.

Based on the findings, it is possible to make suggestions to each glare index to
adjust the defined glare score ranges in order to improve glare evaluation
accuracy and consistency. However, it is important to note that the calculated
glare scores are only applicable to a closed office space with natural lighting.
When artificial light sources or a nighttime environment is included, the
findings from this study cannot guarantee any improvement of accuracy. It is
possible that the calculated glare score ranges are worth consideration by other
researcher groups that can develop further investigation on discomfort glare
issues in buildings. Future studies are encouraged to validate the new findings
and to find errors that were not thoroughly considered.

7.1.5 Exterior Glare: Investigation of the Exterior Glare Issue from Specular
Building Envelopes
The study also contributed the discovery that exterior glare has its own unique and
complex situations that each require different research methodologies. After
completing both interior and exterior studies, the researcher concluded that an
exterior glare study should be designed differently from an interior glare study.
Unlike the interior glare study, an exterior glare study has many unknown factors that
can affect human subject comfort. Ambient temperature, humidity, and wind could
315
not be controlled outside. During the exterior glare study, some subjects appreciated
reflected sunlight when it was cold outside. On the other hand, subjects complained
about reflected sunlight when it was hot outside. It certainly showed that the thermal
aspects of reflected sunlight play a crucial role in visual comfort evaluation. Strong
wind and pollution also affected the subjects’ comfort levels. It was difficult for
subjects to differentiate the causes and results of discomfort when multiple factors
mingled to cause a certain discomfort level.

Privacy issues could also have caused inconsistent and inaccurate subjective
evaluations. Unlike the closed office space, the outdoor patio is open to the public,
and sitting on a chair in that particular research setting made the subjects feel
abnormal and uncomfortable. Their social discomfort could cause inaccurate and
inconsistent evaluations, as well. Thus, developing a study method to keep the
subjects’ privacy would be critical for exterior glare human subject studies.

The study methodology did not sufficiently consider that those factors could greatly
affect a human subject’s comfort and satisfaction levels. The results from the exterior
glare study clearly showed that the research setting was not well designed or
controlled to collect accurate human subject evaluation data. However, it did identify
opportunities for designing a better research setting for exterior glare human subject
study.

316
The data collected from the exterior glare study was not very useful in terms of data
analysis and methodology development. However, the collected data clearly showed
that reflected sunlight from specular building envelopes could cause visual
discomfort issues such as visual discomfort, visual dissatisfaction, and disturbing
glare. This clearly indicates that exterior glare exists and can be an issue for those
outside a building.

An accurate evaluation method can be developed specifically to analyze exterior
glare. The development of an exterior glare analysis method would help new and
existing specular building façades to avoid causing serious visual discomfort to
neighbors, pedestrians, and drivers.

7.1.6 Research Method: Expert Group and Visual Map
Two unique research method ideas were utilized in this study: an expert group and
visual maps. Other human subject research in architecture normally tries to test a
large number of random subjects to get unbiased subjective responses from non-
experts. This approach is a well-validated method to collect unbiased answers and
avoids results influenced by the subjects’ preference or knowledge.

However, the following additional objectives were set for using an expert group: a)
to get feedback from future architectural professionals, b) to educate subjects about
discomfort glare issues through personal experience, and c) to help subjects become
317
daylight glare evaluation experts. In addition, the study intended to see whether an
expert group could help to ensure consistent subjective data. These objectives were
successfully achieved. The expert group did not have knowledge on daylight glare
issues prior to the study, but they started to understand the issue as they participated
in multiple tests. The use of the visual map also helped to make direct comparisons
between subjective data and captured glare scenes, since subjects were able to
visually locate problematic areas on the map. It certainly helped to determine
whether glare sources were visible in subjects’ FOVs. Therefore, the use of the visual
map allowed the study to discover that the human eye FOV is more accurate for
discomfort glare analysis in many cases. These two methods successfully provided
valuable data to develop a new glare analysis method.

7.2 Limitations
Even though the interior glare study was successful, it still had many limitations. The
following limitations were found in the interior and exterior studies. These
limitations were well considered and the consequences from these limitations were
also well understood.
 Exterior glare involves a lot more variables than interior glare. For example,
reflected sunlight moves quickly as the sun moves. It was difficult to plan a
test for the condition with reflected sunlight. Sometimes, there was only time
to perform one or two tests in the proper conditions. A controlled research
setting is required to minimize the variables.
318
 The new methodology only works with daylight glare. Future study is
required to include glare from artificial light sources.
 Human subject tests were performed only during the day, and thus the study
does not include any applicability to a nighttime environment.
 As with many existing glare methods and thresholds, it is difficult to claim
that this new method is better than the rest. Validation of the method will
require independent review.
 Thermal issues always come with discomfort glare issues, especially for
daylight glare. Thermal data was collected along with visual data, but it was
not included in the analysis.
 An expert group was used instead of random subjects. They provided
consistent subjective responses that were valuable for statistical analysis, but
their experience might have been somewhat different from that of a group of
random subjects.
 An HDR imaging technique is a very powerful tool to capture various
luminous environments, but does not guarantee 100% accuracy for captured
luminance information. It was especially difficult to capture luminous daylit
space with lots of clouds and wind, as lighting conditions continuously
changed.
 Various sensors, data loggers, a luminance meter, and an illuminance meter
with high accuracy were carefully chosen and calibrated prior to the main
study, but it is still important to note that the equipment does not have 100%
319
accuracy.

7.3 Future Work
Based on the limitations explained above, several future research topics are
suggested, including exterior glare research in controlled settings, a relative glare
focused study for high contrast conditions, a perceptible glare focused study, and a
validation study of the AGF and RGF method.

7.3.1 Exterior Glare Research
Continuation of exterior glare study is absolutely encouraged. The findings from the
exterior glare study show that more thorough research plans need to be developed to
assess the issue of exterior glare. When performing human subject study outside
buildings, the most essential consideration would be to control the unknown factors,
such as ambient temperature, noise, wind, and privacy issues. Even though weather
conditions cannot be completely controlled, there are still several ways to control
some unknown factors. The researcher first recommends that the exterior glare
research setting be performed in private rather than in public. In addition, using a
specular façade unit would be more beneficial than using a real building’s façade,
since it was found that reflected sunlight moves quickly as the sun moves, which
makes it difficult to plan and perform the test within the small time slot provided.
Being able to modify a specular façade unit would help to overcome this issue. It
would also allow for the researcher to reflect sunlight into subjects’ lines of sight.
320
One of the problems was that reflected sunlight occurred at windows on the third and
fourth floors of the specular building façade, and so the glare was too high to be seen
by subjects. Testing different types of specular materials, such as glass and stainless
steel, would be also interesting. This new research setting would provide subjects
with more consistency in evaluation.

Future research could also examine whether or not other environmental elements
such as heat, wind, and noise can affect people’s visual comfort levels. One could
perform a study comparing the different effects on visual comfort levels under
various ambient temperature conditions.

7.3.2 Relative Glare Factor Focused Study
As noted in Chapter 6, the collected data from both interior and exterior glare studies
do not include a large enough number of glare conditions in which RGF is dominant.
Future study is encouraged with a research setting that has lower light levels. For
example, smaller windows or lower transmissivities would help to further test RGF.
These new settings can increase contrast while decreasing ambient luminance values.
The findings from such a future study can utilize the conclusions of this study; thus,
an improved methodology can be developed to cover various daylit spaces.

7.3.3 Perceptible Glare Focused Study
The analysis results indicate that perceptible glare can be either visually comfortable
321
or uncomfortable. When quantifying perceived glare levels, the perceptible glare
category might be too broad to represent the actual levels of discomfort glare.
Therefore, future study is required to understand what really causes perceptible glare
to be either visually comfortable or uncomfortable. It was expected that glare ratio
would cause the difference within perceptible glare scenes, but concrete evidence
was not found to support the assumption. To make a more thorough investigation,
future study is suggested to test only perceptible glare scenes. Based on the absolute
luminance ranges and glare ratios discussed earlier, it would be possible to exclude
imperceptible, disturbing, and intolerable glare scenes and to design a research
setting with perceptible glare. It would then be possible to investigate how different
factors affect visual comfort and discomfort in perceptible glare scenes.

7.3.4 Validation Study on AGF and RGF Method
A validation study on the new glare analysis method developed using AGF and RGF
would be valuable, especially to determine the accuracy of the method with
additional data. The current method does not consider the effects of electrical
lighting and is not yet applicable to exterior glare. It is important to note that, at this
stage, the new method has only been shown to be applicable to daylight glare scenes
with a computer-based typing task. However, it can be expanded to include other
office tasks, such as no-task and the paper-based writing task, with the future
research plan of an RGF-focused research setting.
322
8 References

Adelson, Edward H. “Lightness Perception and Lightness Illusions.” In The New
Cognitive Neurosciences, 2nd ed., M. Gazzaniga, ed. Cambridge, MA:
MIT Press (2000): 339-351.

Baker, Nick and Steemers, Koen. Daylight Design of Buildings. James & James Ltd.
ISBN 1-873936-88-5, 2002.

Bellia, Laura, Arcangelo Cesarano, Giuseppe F. Iuliano, and Gennaro Spada.
“DAYLIGHT GLARE: A REVIEW OF DISCOMFORT INDEXES.” Paper
presented at visual quality and energy efficiency in indoor lighting: Today
for tomorrow, International Workshop, Rome, Italy, 31 March 31, 2008.

Boubekri, Mohamed. Daylighting, Architecture and Health: Building Design
Strategies. Burlington, MA: Architectural Press, 2008.

Bulow-Hube, Helena. “Daylight in Glazed Office Buildings: A Comparative Study of
Daylight Availability, Luminance and Illuminance Distribution for an
Office Room with Three Different Glass Areas.” Lund University, Sweden,
2008.

Cai, Hongyi. “Luminance Gradient for Lighting Evaluation.” Lighting Research and
Technology (2013): 0-21. Accessed December 23, 2013. doi:
10.1177/1477153513512501.

Carmody, John, Stephen Selkowitz, Eleanor S. Lee, and Dariush Arasteh. Window
Systems for High-Performance Buildings. W. W. Norton & Company Inc.,
2004.

Chauvel, P., J. B. Collins, R. Dogniaux, J. Longmore. Glare from windows: current
views of problem, Lighting Research and Technology 14-1 (1982): 31–46.
323
Debevec, Paul E. and Malik, Jitendra. “Recovering high dynamic range radiance
maps from photographs.” Computer Science Division, University of
California at Berkely, 1997.

Drago, Frederic, William Martens, Karol Myszkowski, and Hans-Peter Seidel.
“Perceptual Evaluation of Tone Mapping Operators with Regard to
Similarity and Preference.” Forschungsbericht Research Report MPI-I-
2002-4-002, Saarbrücken, Germany, 2002.

Dubois, Marie-Claude. “Impact of Shading Devices on Daylight Quality in Offices:
Simulations with Radiance.” Lund University, Sweden, 2001.

Duell, Mark and Webb, Sam. “Now the Walkie Talkie building is melting
BICYCLES: Dazzling light reflected from giant London skyscraper
scorches bike seats (and you can even fry an egg).” Dailymail, September 3,
2013. Accessed December 4, 2014.
http://www.dailymail.co.uk/news/article-2409710/Walkie-Talkie-building-
melting-bicycles-Light-reflected-construction-City-skyscraper-scorches-
seat.html

Egan, David M. (1983). Concepts in Architectural Lighting. McGraw-Hill Book
Company, 1983

Egan, David M. and Olgyay, Victor W. Architectural Lighting. McGraw-Hill Book
Company, 2002.

Energy Information Administration. “2003 Commercial Buildings Energy
Consumption Survey.” Released April 2009. Accessed December 4, 2013.
http://www.eia.gov/totalenergy/data/annual/pdf/sec2_30.pdf

324
Goodyear, Bradley G. and Menon, Ravi S. “Effect of Luminance Contrast on BOLD
fMRI Response in Human Primary Visual Areas.” Journal of
Neurophysiology 79 (1998): 2204-2207.

Hardin, Brad, Brad Nies, and Dimitri Contoyannis. “BIM and sustainable design:
Current capabilities and future possibilities.” Paper presented at Living
Future 08, Vancouver, BC. April 16-18, 2008.

Hopkinson, R. G. “Evaluation of glare.” Illuminating Engineering, (1957): 305-316.

Hwang, Sung C. and Kim, Kang S. “Control of Daylight Environment in Interior
Spaces: Prediction of Daylight Glare from Windows by Using a Computer
Program.” Review of Architecture and Building Science 18-5 (2002): 117-
124.

Inanici, Mehlika. “Transformations in Architectural Lighting Analysis: Virtual
Lighting Laboratory.” PdD diss., University of Michigan, 2004.

Inanici, Mehlika and Galvin, Jim. “ Evaluation of High Dynamic Range Photography
as a Luminance Mapping Technique.” Lawrence Berkeley National
Laboratory, December 2004.

Inanici, Mehlika. “Per-Pixel Lighting Data Analysis.” Lawrence Berkeley National
Laboratory, August 2005. LBNL-58659.

Inanici, Mehlika. “Per-Pixel Lighting Data Acquisition and Analysis with High
Dynamic Range Photography.” Paper presented at the CIE Midterm
Meeting, Leon, Spain, May 21, 2005.

Inanici, Mehlika. “Evaluation of High Dynamic Range Photography as a Luminance
Data Acquisition System.” Lighting Research and Technologies 38- 2
(2006): 123-136.

325
Inanici, Mehlika. “Application of image based rendering in lighting simulation:
Development and evaluation of image based sky models.” Paper presented
at the 11
th
International IBPSA Conference. Glasgow, Scotland, July 27-30,
2009.

Isoard, Gillian, Veronica Garcia-Hansen, and Michael Hirning. “Evaluation of the
Luminous Environment in Open-plan Offices with Skylights.” Paper
presented at the World Renewable Energy Forum, SOLAR 2012, Denver,
Colorado, May 13-17, 2012.

Jakubiec, J. Alstan and Reinhart, Christoph. “The use of glare metrics in the design
of daylit spaces: Recommendations for practice.” Paper presented at the 9
th

International Radiance Workshop, Freiburg, Germany, September 20-21,
2010.

Jakubiec, J. Alstan and Reinhart, Christoph. “The ‘adaptive zone’ – A concept for
assessing discomfort glare throughout daylit spaces.” Lighting Research
and Technology 44 (2012): 149-170.

Japee, Shweta and Schiler, Marc. “A Method of Post Occupancy Glare Analysis for
Building Energy Performance Analysis.” Paper presented at the American
Solar Energy Conference, Minneapolis, MN, 1995.

Kim, Wonwoo and Kim, Jeong T. “The Variation of the Glare Source Luminance
according to the Average Luminance of Visual Field.” Indoor and Built
Environment 21-1 (2012): 98-108.

Kleindienst, Sian A. and Andersen, Marilyne. “The adaptation of daylight glare
probability to dynamic metrics in a computational setting.” Paper presented
at the Lux Europa- 11
th
European Lighting Conference, Istanbul, Turkey,
September 9-11, 2009.

326
Köster, Helmut. Dynamic Daylighting Architecture: Basics, Systems, Projects.
Birkhäuser-Publishers for Architecture, 2004.

Küller, Rikard and Lindsten, Carin. “Health and behavior of children in classrooms
with and without windows.” Journal of Environmental Psychology 12-4
(1992): 305–317.

Kumaragurubara, Viswanathan. “High Dynamic Range Image Processing Toolkit for
Lighting Simulations and Analysis.” PhD diss., University of Washington,
2012.

Linney, Andrew S. “Maximum Luminances and Luminance Ratios and their Impact
on Users’ Discomfort Glare Perception and Productivity in Daylit Offices.”
Master thesis, Victoria University of Wellington. 2008.

Lovell, Jenny. Building envelopes: An Integrated Approach. New York: Princeton
Architectural Press, 2010.

Luckiesh, Matthew and Guth, S.K. “Brightness in Visual Field at Borderline
Between Comfort and Discomfort (BCD).” Journal of Illuminating
Engineering Society November (1949): 650–670.

Mayerowitz, Scott. “Vegas Hotel Pool ‘Death Ray’ Burns Tourists.” ABC News,
September 28, 2010. Accessed January, 15 2014.
http://abcnews.go.com/Travel/las-vegas-hotel-pool-sunlight-swimming-
tourists/story?id=11739234

Ngai, P. and Boyce, P. “The Effect of Overhead Glare on Visual Discomfort.” Journal
of Illuminating Engineering Society November (2000): 29-38.

327
Neal, Leon. “London’s Walkie Talkie tower.” ABC News, September 3, 2013.
Accessed December 4, 2013. http://www.abc.net.au/news/2013-09-
03/london27s-walkie-talkie-tower/4932362

Osterhaus, Werner. “Recommended Luminance Ratios and Their Applications in the
Design of Daylighting Systems for Offices.” Paper presented at the 35th
Annual Conference of the Australia and New Zealand Architectural
Science Association (ANZAScA), Geelong, Australia, November 1-4, 2002.

Osterhaus, Werner. “Discomfort glare assessment and prevention for daylight
applications in office environments.” Solar Energy 79-2 (2005): 140-158.

Osterhaus, Werner. “Design guidelines for glare-free daylit work environments.”
Paper presented at 11
th
European Lighting Conference, Lux Europa,
Istanbul, Turkey, September 9-11, 2009.

Parker, James F. and West, Vita R. Bioastronautics Data Book: Second Edition.
National Aeronautics & Space Administration (NASA) SP-3006, 1973.

Rangi, K. and Osterhaus, Werner. “Windowless environments: are they affecting our
health?” Paper presented at LIGHTING ‘99––Then, Now and Beyond,
44th Annual Convention of the Illuminating Engineering Society of
Australia and New Zealand (IESANZ), Australia, October 18-19, 1999.

Rea, Mark S. and Ouelletue, Michael J. “Relative Visual Performance: A Basis for
Application.” Lighting Research and Technology 23- 3 (1991): 135-144

Reinhard, Erik, Greg Ward, Sumanta Pattanaik, Paul Debevec, Wolfgang Heidrich,
and Karol Myszkowski. High dynamic range imaging: Acquisition, display,
and image-based lighting. (2
nd
ed.). Burlington, MA: Morgan Kaufmann
Publishers, 2010.

328
Saxena, Mudit, Lisa Heschong, Kevin V .D. Wymelenberg, and Seth Wayland. “61
Flavors of Daylight.” Paper presented at 2010 ACEEE Summer Study on
Energy Efficiency in Buildings, Pacific Grove, CA, August 15-20, 2010.

Schiler and Associates. “Disney Hall Glare Study-Final Report.” 2004.

Schiler, Marc. “Toward a Definition of Glare: Can Qualitative Issues Be Quantified?”
Paper presented at the ARCC Conference on Architectural Research, Paris,
France, July 4-8, 2000.

Schiler, Marc and Valmont, Elizabeth. “Microclimatic Impact: Glare Around the Walt
Disney Concert Hall.” Paper presented at the Solar World Congress 2005
Joint American Solar Energy Society / International Solar Energy Society
Conference, Orlando, August 6-12, 2005.

Schiler, Marc. “Examples of glare remediation techniques: four buildings.” Paper
presented at the 26
th
Conference on Passive and Low Energy Architecture,
Quebec City, Canada, June 22-24, 2009.

Shin, Ju Y ., Geun Y . Yun, and Jeong T. Kim. “View types and luminance effects on
discomfort glare assessment from windows.” Energy and Buildings 46
(2012): 139–145.

Spencer, Greg, Peter Shirley, Kurt Zimmerman, and Donald P. Greenberg.
“Physically-based glare effects for digital images”, SIGGRAPH 95- 22
nd

Annual Conference on Computer Graphics and Interactive Techniques
(1995): 325-334.

Steffy, Gary. Architectural lighting design. (3
rd
ed.). Hoboken, New Jersey: John
Wiley & Sons, 2008.

329
Stone, Peter. “The effects of environmental illumination on melatonin, bodily
rhythms and mood states: a review.” Lighting Research and Technology
31-3 (1999): 71–79.

Suk, Jae Y . “POST-TREATMENT ANALYSIS OF THE GLARE REMEDIATION
OF THE WALT DISNEY CONCERT HALL.” Master thesis, University of
Southern California, 2007.

Suk, Jae Y ., Marc Schiler, and Kensek Kensek. “POST-TREATMENT ANALYSIS
OF THE GLARE REMEDIATION OF THE WALT DISNEY CONCERT
HALL.” Paper presented at the American Solar Energy Society Conference,
Cleveland, OH, 2007.

Suk, Jae Y . and Schiler, Marc. “ Interior glare problems of Watt Hall 3rd floor in
University of Southern California.” Paper presented at the American Solar
Energy Society Conference, Cleveland, OH, 2007.

Suk, Jae Y . and Schiler, Marc. “ Investigation of Evalglare software, daylight glare
probability and high dynamic range imaging for daylight glare analysis.”
Lighting Research and Technology 45 (2013): 450-463.

Suk, Jae Y ., Marc Schiler, and Kensek Kensek. “Development of new daylight glare
analysis methodology using absolute glare factor and relative glare factor.”
Energy and Buildings 64 (2013): 113-122.

Tedjakusuma, Jonathan. “A Visual and Digital Method for predicting Discomfort
Glare.” Master thesis, University of Southern California, 2003.

Veitch, Jennifer A. and Newsham, Guy R. “Preferred luminous conditions in open-
plan offices: research and practice recommendations.” Lighting Research
and Technology 32-4 (2000): 199–212.
330
Veitch, Jennifer A. and Newsham, Guy R. “Determinants of Lighting Quality I: State
of the Science.” Journal of the Illuminating Engineering Society 27, 1998.
Accessed May 1, 2010. doi: 10.1080/00994480.

Velds, Martine. “User Acceptance Studies to Evaluate Discomfort glare in Daylit
Room.” Solar Energy 73, No. 2 (2002): 95–103.

Ward, Gregory. “RADIANCE Visual Comfort Calculation.” LESO-EPFL, 1992.

Ward, Gregory. “Universal version of Photosphere.” Anywhere Software. Accessed
March 2009. http://www.anyhere.com/

Wienold, Jan and Christoffersen, Jens. “ Towards a New Daylight Glare Rating.”
Paper presented at Lux Europa, Berlin, September 19-21, 2005.

Wienold, Jan and Christoffersen, Jens. “Evaluation methods and development of a
new glare prediction model for daylight environments with the use of CCD
cameras.” Energy and Buildings 38, Issue 7 (2006).

Wienold, Jan. “Dynamic daylight glare evaluation.” Paper presented at the 11th
International Building Performance Simulation Association (IBPSA)
Building Simulation, Glasgow, Scotland, 2009.

Wienold, Jan. “Evalglare Version 0.9f. Release notes (Important changes from 0.9).”
Fraunhofer ISE, October 2009. Accessed December 12, 2009.
http://www.ise.fraunhofer.de/search?SearchableText=evalglare

Wienold, Jan. “Evalglare Version 1.0. Release notes (Important changes from 0.9).”
Fraunhofer ISE, February, 2012. Accessed July 9, 2012.
http://www.ise.fraunhofer.de/search?SearchableText=evalglare

Wymelenberg, Kevin V .D., Mehlika Inanici, and Peter Johnson. “The effect of
luminance distribution patterns on occupant preference in a daylit office
environment.” LEUKOS VII (II) 2010.

331
Wymelenberg, Kevin V .D. “valuating Human Visual Preference and Performance in
an Office Environment Using Luminance-based Metrics.” PhD diss.,
University of Washington, 2012.

Yin, Hanshu and Schiler, Marc. “Study on HDR image as a Luminance Mapping
Tool.” Paper presented at the IESNA Annual Meeting, Austin, October 31-
November 1, 2011.

Yin, Hanshu and Schiler, Marc. “Study on Glareindices- A User-Friendly Glare
Evaluating Program Based on HDR Images.” Paper presented at the
IESNA Annual Meeting, Austin, October 31- November 1, 2011.

332
Appendix A. USC Institutional Review Boards Approvals

UNIVERSITY OF SOUTHERN CALIFORNIA
UNIVERSITY PARK INSTITUTIONAL REVIEW BOARD
FWA 00007099

Date: Jan 24, 2013, 08:39am
Principal Investigator: Jae Yong Suk
Faculty Advisor: Douglas Noble
Co-Investigators:
Project Title: Interior and Exterior Daylight Glare Analysis Study
USC UPIRB # UP-13-00023

The iStar application and attachments were reviewed by UPIRB staff on 1/24/2013.

The project was APPROVED.

The study has been reviewed and determined to qualify for exemption under the USC
Human Research Protection Program Flexibility Policy. The study is not subject to 45 CFR
46 including informed consent requirements and further IRB review, unless there is
modifications to the study that increase risks to subjects or the funding status changes.

If there are modifications that increase risk to subjects or the funding status of this
research is to change, you are required to submit an amendment to the IRB for review
and approval.

The following documents were reviewed and approved:
Certified Information Sheet, dated 01-23-2013
Certified Recruitment Script, dated 01-23-2013

Minor revisions were made to the recruitment and consent documents by the IRB
Administrator (IRBA). The IRBA revised documents have been uploaded into the
333
relevant iStar sections. Please use the IRBA revised documents if an amendment is
submitted and future revisions are required.

To access IRB-approved documents, click on the “Approved Documents” link in the
study workspace. These are also available under the “Documents” tab.

Sincerely,

RoseAnn Fleming, CIP
Funding Source(s): N/A - no funding source listed

334
Appendix B. Participant Recruitment Email

335
Appendix C. Information Sheet for Non-Medical Research

336

337
338
339
Appendix D. Human Subjects Certificate: Human Subjects Education
Program (CITI)
340

341
Appendix E. UTA/HOBO connector

342

Abstract (if available)

Abstract Building facades comprise the first layer between the occupant and the outside world. They are largely responsible for conserving building energy, encouraging occupant comfort, providing views, and allowing for the expression of design aesthetics. The trend toward sustainable architecture has inspired an increasing number of transparent building facades. Transparent facades allow more natural light into the interior spaces and are more thermally reflective, which lessens the solar heat gain of the building. The use of transparent materials to harvest daylight can result in occupant discomfort from glare or veiling reflections, while the use of reflective and specular materials can cause momentary blindness and produce excessive thermal discomfort for those outside the building. Complex building geometry, including curved or faceted facades, have made it even more difficult to predict when or where sunlight will cause discomfort glare. Even though many research groups have developed their own analysis metrics and tools to address interior glare issues, existing glare metrics remain too inconsistent, inaccurate, and complicated to be incorporated into daylighting practice. Hence, the existing metrics are used by only a few researchers. ❧ Many absolute luminance thresholds and specific contrast ratio values have also been developed in various lighting codes and standards, such as the Swedish energy authority NUTEK, ISO Standard 9241-6, and ANSI/IESNA RP-1 VDT Lighting Standard. These thresholds are inconsistent with each other. Due to these problems and limitations, this study aimed to develop a new daylight glare analysis method that is practical enough to be utilized in practice without compromising evaluation accuracy. Human subject studies have been performed inside and outside buildings with the help of high dynamic range (HDR) imaging techniques and light sensors. An HDR image analysis tool was developed in MATLAB to analyze digitally captured glare scenes, to calculate luminance values, and to visualize glare sources in the field of view. Along with the MATLAB tool, a glare analysis program called Evalglare was used to calculate the five existing glare index scores of the HDR images. Collected subjective evaluation data and the visual scenes have been analyzed statistically to examine the advantages and disadvantages of the existing glare indices and then to develop a new method based upon these revelations. ❧ Absolute glare factor (AGF) and relative glare factor (RGF) were used to explain the different causes of daylight glare problems inside and outside the buildings. Every glare scene has both AGF and RGF, although one of them can be more dominant in causing discomfort glare than the other. AGF causes discomfort glare with excessive glare source brightness, while RGF causes glare with high contrast ratios between task luminance and glare source luminance. Based on the human subject study results, the categories were successfully defined within absolute luminance ranges and glare ratios for different office task activities. These categories were: imperceptible, perceptible, disturbing, and intolerable. When no task is being performed inside the building by the occupant, an imperceptible glare has up to 2,752 cd/m², a perceptible glare occurs between 2,752 cd/m² and 7,000 cd/m², a disturbing glare is dominant from 7,000 cd/m² to 12,522 cd/m², and an intolerable glare occurs from 12,522 cd/m² and up. When a computer‐based typing task is being performed, an imperceptible glare occurs up to 1,920 cd/m², a perceptible glare occurs between 1,920 cd/m² and 5,000 cd/m², a disturbing glare is dominant from 5,000 cd/m² to 11,718 cd/m², and an intolerable glare occurs from 11,718 cd/m² and up. When a paper‐based writing task is being performed, there is an imperceptible glare up to 1,696 cd/m², a perceptible glare between 1,696 cd/m² and 5,263 cd/m², and a disturbing or intolerable glare occurs from 5,263 cd/m² and up. These defined glare categories with and without task activity clearly show that occupants have different visual sensitivities depending on whether and what kind of office task activity is involved. Vertical illuminance and glare ratio ranges were also successfully defined for different glare categories. Unlike the luminance and vertical illuminance ranges, glare ratios were defined only for typing task scenes, since the no‐task and writing task scenes were not RGF dominant in the interior glare research setting. When a computer‐based typing task is being performed, imperceptible glare is dominant up to 12, the glare ratio between task luminance and glare source luminance while disturbing glare is dominant from 22 to 32, and intolerable glare occurs beyond the glare ratio of 32. Perceptible glare occurs between 12 and 22 glare ratios. ❧ Finally, AGF‐ and RGF‐based glare equations were developed to evaluate perceived glare categories. AGF and RGF zones were also displayed on a scatter plot graph to visually explain which glare factor is more dominant to cause discomfort glare. After validating the existing glare indices using human subject study data, the study presents several ideas to future users to assist their understanding and application of existing glare metrics. Even though an exterior glare study was not conclusive enough to develop absolute and relative glare factors, we hope that the study will initiate new exterior glare human subject studies to prove the existence of exterior visual discomfort issues caused by strong sunlight reflections on specular building envelopes. The findings suggest that such an exterior glare study needs a different research method from that of the interior glare research method used in this study. The glare factor analysis method can provide the causes and results of discomfort glare

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Post-treatment analysis of the glare remediation of the Walt Disney Concert Hall

PDF

Glare studies: Comparison of three glare indices, HDR imaging and measured values

PDF

Multi-domain assessment of a kinetic facade: determining the control strategy of a kinetic façade using BIM based on energy performance, daylighting, and occupants’ preferences; Multi-domain asse...

PDF

Daylight prediction: an evaluation of daylighting simulation software for four cases

PDF

Evaluation and development of solar control screens: using daylight simulation to improve the performance of facade solar control screens

PDF

Double skin façades performance: effects on daylight and visual comfort in office spaces

PDF

Daylighting in Riyadh and Los Angeles: comparison of cultural factors in potential market penetration

PDF

Daylight and health: exploring the relationship between established daylighting metrics for green building compliance and new metrics for human health

PDF

Simulation-based electric lighting control algorithm: integrating daylight simulation with daylight-linked lighting control

PDF

Blind to light loss: evaluating light loss through commercial building facades as a contribution to urban light pollution

PDF

Effective light shelf and form finding: development of a light shelf design assistant tool using parametric methods

PDF

Guidelines to airport design: accounting for glare from buildings during takeoff and landing – an LAX case study

PDF

Adaptive façade controls: A methodology based on occupant visual comfort preferences and cluster analysis

PDF

Tilted glazing: angle-dependence of direct solar heat gain and form-refining of complex facades

PDF

Human-building integration: Investigation of human eye pupil sizes as a measure of visual sensation in the workstation environment

PDF

Skin fit and retrofit: challenging the sustainability of curtainwall practice in tall buildings

PDF

Building bridges: filling gaps between BIM and other tools in mechanical design

PDF

Exploring the influence of scheduling on daylighting performance evaluation in assisted living: testing two new metrics based on sDA and ASE

PDF

Detecting joint interactions between sets of variables in the context of studies with a dichotomous phenotype, with applications to asthma susceptibility involving epigenetics and epistasis

Asset Metadata

Creator Suk, Jae Yong (author)

Core Title Absolute glare and relative glare factors: predicting and quantifying levels of interior glare and exterior glare caused by sunlight and daylight

School School of Architecture

Degree Doctor of Philosophy

Degree Program Architecture

Degree Conferral Date 2014-06

Publication Date 07/01/2014

Defense Date 04/03/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag absolute glare factor,daylighting,discomfort glare,exterior glare,interior glare,OAI-PMH Harvest,relative glare factor

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Schiler, Marc (committee chair), Kensek, Karen M. (committee member), Tjan, Bosco S. (committee member)

Creator Email jsuk@usc.edu,jysuk77@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-428264

Unique identifier UC11286496

Identifier etd-SukJaeYong-2593.pdf (filename),usctheses-c3-428264 (legacy record id)

Legacy Identifier etd-SukJaeYong-2593.pdf

Dmrecord 428264

Document Type Dissertation

Format application/pdf (imt)

Rights Suk, Jae Yong

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA