University of Southern California Dissertations and Theses

Beyond greenhouse gases and towards urban-scale climate mitigation: understanding the roles of black carbon aerosols and the urban heat island effect as local to regional radiative forcing agents

(USC Thesis Other)

Beyond greenhouse gases and towards urban-scale climate mitigation: understanding the roles of black carbon aerosols and the urban heat island effect as local to regional radiative forcing agents

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Beyond greenhouse gases and towards urban-scale climate mitigation:
understanding the roles of black carbon aerosols and the urban heat island effect
as local to regional radiative forcing agents
by
Joseph Ko
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ENVIRONMENTAL ENGINEERING)
December 2022
ii
Acknowledgements
First and foremost, I want to acknowledge my parents, Yong and Joohee Ko, who worked tirelessly
and sacrificed so much to give me a chance at a better life. Without them, I wouldn’t be who I am today. I
also want to acknowledge my sister Joyce, all my cousins, and all my relatives who have supported me
unconditionally, not only through the PhD, but throughout my life.
I’m also grateful for all the wonderful colleagues, labmates, and friends I’ve met at USC while
working on my PhD. A special shout out to all my labmates from the Ban-Weiss and Sanders groups:
Trevor Krasowsky, Jiachen Zhang, Arash Mohegh, Mo Chen, Yun Li, Hannah Schlaerth, McKenna
Peplinski, Kayley Butler, Diego Aguilera, Andrew Jin, Stepp Mayes, and Zoia Comarova.
A huge thank you to my advisor, Dr. Kelly Sanders, who graciously adopted me into her group in
my final year. Kelly is a super-human, and it would take more than a dissertation to figure out how she does
it all. Her passion and enthusiasm for science and effective policy is infectious, and beyond that, it’s clear
that she cares deeply about each of us in her group. She’s the best!
I’d also like to thank my qualifying and defense committee members: Dr. Amy Childress, Dr. Erika
Garcia, Dr. Felipe de Barros, and Dr. Sam Silva. They gave me helpful feedback on my dissertation and
contributed to rich discussions about my research.
A special thanks to my partner, Maho, for being my rock during my PhD years. It’s never a dull
moment with you, and I’m lucky to have someone who gets me like you do. I love you!
Finally, I want to dedicate this PhD to my advisor in Memorium, Dr. George Ban-Weiss (1981-
2021). I’ve never met anyone like him, and I likely never will. His ability to ask the important questions,
both in research and in life, have made a profound impact on how I view the world. His kindness, empathy,
and uncanny ability to make anyone feel seen, were his super-powers. He showed me a life well lived, and he
left us way too soon. We’ll forever miss you, George.

iii
Table of Contents
Acknowledgements ................................................................................................................................................... ii
List of Tables ............................................................................................................................................................ vi
List of Figures ......................................................................................................................................................... vii
Abstract.................................................................................................................................................................. xvii
Chapter 1 – Introduction ......................................................................................................................................... 1
1.1 Background ..................................................................................................................................................... 1
1.2 Significance of research .................................................................................................................................. 2
1.2.1 Black carbon aerosols .............................................................................................................................. 2
1.2.2 The urban heat island effect ................................................................................................................... 3
Chapter 2 - Measurements to determine the mixing state of black carbon emitted from the 2017–2018
California wildfires and urban Los Angeles ............................................................................................................ 6
2.1 Introduction .................................................................................................................................................... 6
2.2 Methods ......................................................................................................................................................... 10
2.2.1 Measurement location and time periods .............................................................................................. 10
2.2.2 Instrumentation ..................................................................................................................................... 11
2.2.3 Auxiliary data ......................................................................................................................................... 12
2.2.4 Estimation of source-to-receptor timescale ........................................................................................ 15
2.2.5 Time series filtering ............................................................................................................................... 15
2.2.6 Lag-time method ................................................................................................................................... 16
2.2.7 Leading-edge-only (LEO) method ....................................................................................................... 17
2.3 Results and discussion .................................................................................................................................. 18
2.3.1 Source identification and meteorology ................................................................................................ 18
2.3.2 rBC mass and number concentration .................................................................................................. 22
iv
2.3.3 Lag-time analysis: number fraction of thickly-coated rBC-containing particles ............................... 25
2.3.4 Negative lag-times and rBC morphology ............................................................................................ 27
2.3.5 Leading-edge-only (LEO) fit analysis: rBC coating thickness ........................................................... 31
2.3.6 rBC core size .......................................................................................................................................... 36
2.3.7 Impact of emissions source and aging on rBC mixing state .............................................................. 39
2.3.8 Comparison to past studies quantifying CT
BC
using the SP2 ............................................................. 50
2.4 Conclusion ..................................................................................................................................................... 53
2.5 Video supplement ......................................................................................................................................... 56
2.6 Funding and support .................................................................................................................................... 56
Chapter 3 - Measuring the impacts of a real-world neighborhood-scale cool pavement deployment on
albedo and temperatures in Los Angeles .............................................................................................................. 57
3.1 Introduction .................................................................................................................................................. 57
3.2 Methods ......................................................................................................................................................... 61
3.2.1 Site details ............................................................................................................................................... 61
3.2.2 Measurement details .............................................................................................................................. 64
3.2.3 Quantifying the direct impact of cool pavement on temperatures: difference-in-difference (DID)
method ............................................................................................................................................................. 65
3.3 Results and discussion .................................................................................................................................. 67
3.3.1 Albedo .................................................................................................................................................... 67
3.3.2 Surface temperature ............................................................................................................................... 71
3.3.3 Air temperature ...................................................................................................................................... 73
3.3.4 Implications on pedestrian thermal comfort ....................................................................................... 76
3.4 Conclusion ..................................................................................................................................................... 80
3.5 Funding and support .................................................................................................................................... 81
v
Chapter 4 - Modeling the impacts of anthropogenic heat on the regional climate of Los Angeles ................ 82
4.1 Introduction .................................................................................................................................................. 82
4.2 Methods ......................................................................................................................................................... 85
4.2.1 Buildings ................................................................................................................................................. 87
4.2.2 Traffic ..................................................................................................................................................... 90
4.2.3 Human metabolism ............................................................................................................................... 93
4.2.4 Regional climate modeling .................................................................................................................... 95
4.3 Results and discussion .................................................................................................................................. 95
4.3.1 AHF in Los Angeles .............................................................................................................................. 95
4.3.2 Impacts on meteorology ..................................................................................................................... 104
4.4 Conclusion ................................................................................................................................................... 113
4.5 Funding and support .................................................................................................................................. 115
Conclusion ............................................................................................................................................................. 116
References ............................................................................................................................................................. 118
Appendices ............................................................................................................................................................ 136
Appendix A: Supplement to Chapter 2 ........................................................................................................... 136
Appendix B: Supplement to Chapter 3 ........................................................................................................... 164

vi
List of Tables
Table 2-1. Major wildfires that were active during the three campaigns. Only the two largest fires
from each campaign (in terms of burn area) are listed in the table below. Note that there
were numerous other smaller fires that were active during the three campaigns, but not listed
in this table. ................................................................................................................................................ 10
Table 2-2. Details of the ten different LEO time periods. Further details about the source-to-
receptor characteristic timescales can be found in Appendix A, section S1. ........................................ 32
Table 2-3. Summary table of rBC coating thickness values reported in previous studies using the
SP2. ............................................................................................................................................................. 52
Table 3-1. Albedo spot measurement details and results. .................................................................................. 69
Table 3-2. Comparison of reported T
air
reductions per 0.1 increase in albedo from previous
modeling studies. ....................................................................................................................................... 76
Table 4-1. List of modeling studies that quantify the impact of anthropogenic heat ...................................... 83
Table 4-2. List of the main datasets used to construct the AHF dataset. Meaning of the acronyms
used here can be found in the acronym dictionary provided in the Appendix. ................................... 85
Table 4-3. Mean AHF for various geospatial aggregations and by sector categories. The percent
contribution of each sector is also shown for each spatial aggregation. The Near Highways
aggregation includes all areas in LA County within 1 km of a major highway (i.e., freeways). ........... 96

vii
List of Figures
Figure 2-1. Overview map showing the location of the sampling site with respect to the Greater Los
Angeles (LA) area. Map data © 2015 Google. ........................................................................................ 11
Figure 2-2. Wind roses for the September 2017 (first row), December 2017 (second row), and
November 2018 (third row) sampling periods. Wind roses are based on five-minute ASOS
airport data from LAX (first column), LGB (second column), and AVX (third column),
provided by NOAA. .................................................................................................................................. 20
Figure 2-3. HYSPLIT back-trajectories for all three campaigns. The star denotes the start location
of each back-trajectory, i.e., the sampling location. The trajectories for the first period
(September 2017) (i.e., panel a) represent week-long back-trajectories for each day of the
campaign. The trajectories for the (b) second (December 2017) and (c) third (November
2018) periods represent 72-hour back-trajectories for each hour of the campaign. Panels (d)
and (e) show more zoomed-in maps of the second and third campaign back-trajectories
along with active Southern California fires. Map data: Google DigitalGlobe. ..................................... 21
Figure 2-4. Time series of (a) BC absolute coating thickness, (b) number fraction of thickly-coated
rBC particles, (c) rBC count median diameter, and (d) rBC concentrations, for all three
measurements campaigns. The boxed annotations (i.e., L1 to L10) refer to specific LEO
periods, which are further described in Section 2.3.5. In panel (a), each blue dot represents
an individual particle. The hourly median is shown in the dotted pink line, and the
corresponding 10
th
and 90
th
percentiles are shown in purple. In panel (b), green dots
represent one-minute means while the black curve shows hourly means. Panel (c) shows the
one-minute mean for the count mean diameter. Panel (d) shows the one-minute means for
rBC concentration...................................................................................................................................... 24
viii
Figure 2-5. Meteorological variables and rBC concentrations during the second campaign
(December 2017). Panel (a) shows wind speed and (b) shows wind direction measured by a
NOAA weather station located at Los Angeles International Airport (LAX). Panel (c) shows
rBC mass and number concentrations and identifies three peaks of interest. The two dashed
ovals in panel (b) highlight periods of northerly and easterly winds, which occur ~0.5-1 days
before each of the three peaks, suggesting that the elevated rBC concentrations included
important contributions from the local Thomas Fire (and other smaller fires) and urban
emissions from the Los Angeles basin. .................................................................................................... 25
Figure 2-6. Panel (a) shows the 10-minute mean time series for number fraction of rBC particles
with negative lag-times (f
neg, lag
). The threshold for negative lag-times was set to -1.25 s to
account for uncertainties in the lag-time determination (Sedlacek et al., 2012). Panel (b)
shows the time series of lag-time values for each individual particle, corresponding to
individual dots on the graph. Panel (c) shows the one-minute mean rBC number
concentration for reference. ...................................................................................................................... 28
Figure 2-7. Scatter plot of 10-minute mean negative lag-times versus 10-minute mean rBC core
diameters. .................................................................................................................................................... 30
Figure 2-8. Distributions of BC coating thickness (CT
BC
) aggregated by campaign are shown in red
(1
st
campaign), green (2
nd
campaign), and purple (3
rd
campaign). The combined distributions
for all campaigns are shown in black. Panels (a) and (b) show the normalized frequency
distributions, while panels (c) and (d) show the absolute frequency distributions. The
distributions are also distinguished by the rBC core diameter ranges included in the LEO
analysis. The top panels (a) and (c) show distributions for particles with rBC core diameters
between 180 and 220 nm. The bottom panels (b) and (d) show distributions for particles
with rBC core diameters between 240 and 280 nm. ............................................................................... 32
ix
Figure 2-9. Violin plots that show the distribution of rBC coating thickness values calculated for
each LEO time period, L1 through L10. Each circle marker in the plot represents a particle
analyzed by the LEO analysis and the curves for each “violin” shape represents the
normalized probability density function of the coating thickness for each LEO period. The
violin shape results from mirroring each probability density distribution along a vertical axis.
Box-and-whiskers plots are also overlaid to show the quartiles (25
th
, 50
th
, and 75
th

percentiles) of the coating thickness distributions. The 95% confidence intervals (CI) based
on Student’s t-distribution are shown above each violin plot to demonstrate when the mean
coating thickness values are statistically distinguishable from one another. The mean (unfilled
diamond) and median (solid diamond) coating thicknesses are also indicated above each
violin plot, and a brief description of sources for each LEO period is annotated below each
distribution. ................................................................................................................................................ 34
Figure 2-10. Measured rBC core size distributions and corresponding log-normal fits to the
measurements for LEO periods L1, L5, and L10. .................................................................................. 37
Figure 2-11. Median rBC core diameter for both mass and number size distribution lognormal fits. ........... 38
Figure 2-12. rBC coating thickness versus rBC core diameter. Each point on the plot represents a 1-
minute mean. Data from all three campaigns are shown. CT
BC
values are calculated for
particles with rBC core diameters between 200–250 nm. The line represents the least-squares
linear regression to the one-minute mean data points. There is a statistically significant
positive correlation shown between CT
BC
and rBC core diameter, as shown in the summary
box in the top left corner. ......................................................................................................................... 40
Figure 2-13. Distributions of BC coating thickness (CT
BC
) aggregated by campaign and varying rBC
core diameter ranges used in the LEO analysis. Panels (a) through (d) in the left column
show the normalized frequency distributions, while panels (e) through (h) in the right
x
column show the absolute frequency distributions. Within each panel, each line represents a
distribution for a particular rBC core diameter range, with darker lines representing larger
diameter ranges and vice versa. ................................................................................................................ 41
Figure 2-14. Matrix of scatter plots showing the time evolution of CT
BC
(nm) and rBC count mean
diameter (nm) for the second campaign (December 2017). Axes labels are shown in the
upper left. A scatter plot is shown for each six-hour time interval of the day, starting at 00:00
Pacific Time, and for each day of the campaign. The columns of the matrix denote the time
interval of the day, and the rows of the matrix denote the days of the campaign. Each point
within a plot represents a one-minute mean value for both CT
BC
and count mean diameter. ............ 43
Figure 2-15. Matrix of scatter plots showing the time evolution of CT
BC
(nm) and rBC count mean
diameter (nm) for the third campaign (November 2018). Axes labels are shown in the upper
left. A scatter plot is shown for each six-hour time interval of the day, starting on 00:00
Pacific Time, and for each day of the campaign. The columns denote the time interval of the
day, and the rows denote the day of the campaign. Each point within a plot represents a one-
minute mean value within that six-hour interval for both CT
BC
and count mean diameter. ............... 44
Figure 2-16. Contour plots of count as a function of one-minute mean BC coating thickness (CT
BC
)
and one-minute mean rBC core diameter. This figure can be interpreted as a 2-d joint
histogram, converted to a contour plot. Each count represents a single one-minute mean
data point. The contours are created based on the 2-d joint histogram that is calculated using
a 50x50 grid within the range of all one-minute mean data. Panels (a), (b), and (c) in the first
row show mass mean diameter on the horizontal axes, while panels (d), (e), and (f) in the
second row show count mean diameter. ................................................................................................. 46
Figure 3-1. Map of the measurement site in Covina, CA. The impact area (with cool pavements) is
shown in blue, while the control area (without cool pavements) is shown in red. The inset
xi
map in the upper-left corner shows the Covina site with respect to downtown Los Angeles
and the greater Los Angeles metropolitan area. Map data ©2020 Google. .......................................... 62
Figure 3-2. Map of cool pavement installation by product. Modified figure based on original map
supplied by Los Angeles County Department of Public Works. Below the map is a table that
describes the corresponding installation schedule. The red dashed box on the map shows the
portion of Queenside Dr that was installed after the first phase of installation. .................................. 63
Figure 3-3. Illustration of the difference-in-difference (DID) method. Δ
post
is the post-event
difference between the impact and control measurements (e.g., impact – control). Δ
pre
is the
pre-event difference between the impact and control measurements. The DID is defined as
the difference between Δ
post
and Δ
pre
. This DID value is therefore the attributable impact of
the event on the variable of interest, Y. “Dummy” values are shown in the figure as a simple
illustration. In this example, the event reduced Y by 1 unit (i.e, DID = -1)......................................... 66
Figure 3-4. Stationary albedo measurements made at seven distinct locations (S1 - S7) throughout
the impact area. The markers represent discrete measurements. The markers and lines are
color-coded by the different cool pavement products that were installed in the impact area.
Pre-installation albedo represents the mean albedo of the pavement in the impact area
before installation. The pre-installation albedo is based on mobile albedo measurements
taken on 2019 Sep 17-18. A heavy rain event occurred on 2019 Nov 11 with a recorded 24-
hour precipitation of 8.9 mm (NCEI 2021). ........................................................................................... 68
Figure 3-5. Panels (a) – (c) show a “heat-map” of albedo ~1 week, ~1 month, and ~1 year after
cool pavement installation, respectively. Panels (d) – (f) show satellite imagery of
representative areas of cool pavement with marked albedo decrease (highlighting with red
circles) in the impact area. Map data ©2020 Google. ............................................................................. 70
xii
Figure 3-6. Matrix of barplots showing the surface temperature DID for unique pairs of dates. The
columns identify the first date used in the DID calculation and the rows identify the second
date. The barplots outlined in red show the DID values that represent the impact of cool
pavement installations on surface temperature. The barplots outlined in blue highlight the
null DID impacts that occur when the DID calculations were made with two pre-installation
dates or two post-installation dates. A negative DID value indicates a surface temperature
reduction. The error bars represent the 95% confidence intervals for the mean. ................................ 72
Figure 3-7. A scatter plot for hourly mean Δ
post
versus Δ
pre
, where Δ is the difference between 3 m
air temperature in the impact area versus control area (i.e., impact minus control). “Post” or
“pre” in the subscript signifies whether the measurements were made post-installation or
pre-installation, respectively. The values inside each circle represents the hour of day and the
error bars represent the 95% confidence interval for the mean. Values above the 1:1 line
imply a warming effect due to cool pavement, while values below imply a cooling effect. The
air temperature DID can be calculated as (Δ
post
- Δ
pre
). Only non-daylight hours were included
in this analysis (see section 3.2.2). The first non-daylight hour was 18:00 LST. Data from
hours 00:00 LST to 06:00 LST showed a null effect (i.e., neither warming nor cooling effect)
and are thus not included in the figure. ................................................................................................... 74
Figure 4-1. Flowchart describing the different datasets used and the overall process of constructing
the AHF data products. ............................................................................................................................. 86
Figure 4-2. This is an example of dasymetric mapping from a USGS report. This figure illustrates
how population distribution is mapped at finer resolution using auxiliary land use data. Panel
(a) on the left shows a choropleth plot using data from the US Census Bureau at the census
block-group level. Panel (b) shows an improved representation of population distribution
using land cover data and dasymetric mapping. A similar approach is taken for the AHF
xiii
datasets in our study by using auxiliary data on road and building geometries. (Source: USGS
2008, last access 7/1/2022) ...................................................................................................................... 89
Figure 4-3. AHF
building
from (a) residential, (b) commercial, (c) industrial, and (d) all buildings for a
subset of Central LA. This includes the neighborhoods of Downtown, Koreatown,
Westlake, Pico-Union, and Harvard Heights. The AHF shown here is representative of an
average hour in the year of 2016. The final dataset will incorporate temporal variability, but
here we show a representative average to illustrate the spatial variability in AHF
building
. Note,
the range of the colorbar in panel (d) is different than the colorbars in panels (a) – (c). .................... 90
Figure 4-4. This figure shows the SCAG VMT dataset for a subset of Central LA. Each road
segment has an assigned VMT value that is representative of the year 2016. This snapshot
shows the total VMT during morning hours (6 am to 9 pm local time). Diagonal lines
represent the aggregated VMT of minor roads within their respective TAZ. ...................................... 91
Figure 4-5. Panel (a) shows AHF
traffic
during morning rush hour (8 am) and panel (b) shows
AHF
traffic
during evening non-rush hour (8 pm). This figure shows AHF for a subset of
Central LA as described in section 4.2.1. ................................................................................................. 93
Figure 4-6. AHF
metabolism
during the (a) daytime and (b) nighttime. This figure shows AHF for a
subset of Central LA as described in section 4.2.1. ................................................................................ 95
Figure 4-7. Heat maps of AHF within LA County from (a) buildings, (b) metabolism, (c) traffic,
and (d) total. ............................................................................................................................................... 97
Figure 4-8. Zoomed in heat map showing annual mean AHF. There are noticeable clusters of
hotspots around dense industrial and commercial areas, as well as elevated AHF over
highways. .................................................................................................................................................... 98
Figure 4-9. Unitless diurnal profiles for sub-components of AHF. Panels (a) – (g) illustrate the
mean normalized diurnal load profiles for a typical weekday in August. Panels (h) and (i)
xiv
shows the fraction of nighttime and daytime metabolism contributing to total AHF
metabolism
on
a weekday. I.e., AHF
metabolim
is a linear sum of nighttime and daytime metabolism to capture
the diurnal variability in commute patterns for each neighborhood. Institutional buildings
were assumed to follow the same profile as commercial buildings. ...................................................... 99
Figure 4-10. Unitless seasonal profiles for sub-components of AHF, normalized to the annual
mean. Industrial (Ind) and Institutional (Inst) categories are not shown here because the
Commercial (Com) seasonal profiles were substituted as an approximation. .................................... 100
Figure 4-11. Diurnal profiles of mean total AHF, averaged over LA County (gray), urban areas
(yellow), City of LA (purple), and within 1 km of a major highway (green). ...................................... 101
Figure 4-12. Annual mean AHF for the top 20 neighborhoods in Los Angeles. .......................................... 102
Figure 4-13. Grouped barplot showing the relative contributions of traffic, buildings, and
metabolism to the total mean AHF, for the top 10 neighborhoods. .................................................. 103
Figure 4-14. Panel (a) shows the diurnal mean AHF profiles for all 270 neighborhoods in LA
County. Each profile represents an average weekday in August. Red indicates higher AHF
and blue indicates lower AHF. Panels (b) and (c) show the diurnal profile of mean AHF for
the top and bottom 5 neighborhoods, respectively. Red ..................................................................... 104
Figure 4-15. Nested model domains for all WRF simulations in this study. The inner-most domain
is labeled “d03” and has a grid cell resolution of 2 km. ....................................................................... 105
Figure 4-16. WRF simulation scenarios conducted for this study. AHF-on means that the gridded
AHF dataset was used as an input into the WRF model, and vice versa. Each simulation
period was 24 hours. ................................................................................................................................ 105
Figure 4-17. Diurnal plot showing hourly, mean changes in 2 m air temperatures (ΔT
air
) for both
summer (red) and winter (blue) days. The mean is spatially representative of urban areas
xv
within Los Angeles County. The error shading for both lines indicates the 95% confidence
interval for the mean. .............................................................................................................................. 106
Figure 4-18. Histograms showing the distributions of ΔT
air
for both summer (red) and winter (blue).
Each histogram is normalized such that the area under the curve is equal to one............................. 107
Figure 4-19. Heatmaps of ΔT
air
from both the summer and winter simulations. Red indicates an
increase, while blue indicates a decrease. Each plot shows the spatial distribution of ΔT
air
for
the hours of maximum mean ΔT
air
, which were 08:00 and 10:00 LST for summer and winter,
respectively. .............................................................................................................................................. 107
Figure 4-20. Diurnal plot showing hourly, mean changes in planetary boundary layer height
(ΔPBLH) for both summer (red) and winter (blue) days. The mean ΔPBLH is spatially
representative of urban areas within Los Angeles County. The error shading for both lines
indicates the 95% confidence interval for the mean. ............................................................................ 109
Figure 4-21. Histograms showing the distributions of ΔPBLH for both summer (red) and winter
(blue). Each histogram is normalized such that the area under the curve is equal to one. ................ 110
Figure 4-22. Heatmaps of ΔPBLH

from both the summer and winter simulations. Red indicates an
increase, while blue indicates a decrease. Each plot shows the spatial distribution of ΔPBLH
for the hours of maximum mean ΔPBLH, which were 08:00 and 16:00 LST for summer and
winter, respectively. ................................................................................................................................. 111
Figure 4-23. Diurnal plot showing hourly, mean changes in wind speed (Δ(Wind Speed)) for both
summer (red) and winter (blue) days. The mean is spatially representative of urban areas
within Los Angeles County. The error shading for both lines indicates the 95% confidence
interval for the mean. .............................................................................................................................. 112
xvi
Figure 4-24. Histograms showing the distributions of Δ(Wind Speed) for both summer (red) and
winter (blue). Each histogram is normalized such that the area under the curve is equal to
one. ............................................................................................................................................................ 113
Figure 4-25. Heatmaps of Δ(Wind Speed)

from both the summer and winter simulations. Red
indicates an increase, while blue indicates a decrease. Each plot shows the spatial distribution
of Δ(Wind Speed) for the hours of maximum mean Δ(Wind Speed), which were 00:00 and
10:00 LST for summer and winter, respectively. ................................................................................... 113

xvii
Abstract
There is broad agreement that emissions of greenhouse gases (GHG) have led to human-induced
climate change, increasing global average temperatures as well as the frequency, duration, and intensity of
extreme weather events. However, in addition to GHGs, there are also non-GHG agents that contribute to
atmospheric warming at local-to-regional spatial scales. Two prominent non-GHG warming agents are (1)
black carbon (BC) aerosols, and (2) the urban heat island (UHI) effect. BC aerosols are small (~0.1 - 1 μm)
carbonaceous particles emitted from incomplete combustion of fossil fuels and biomass burning. BC warms
the atmosphere by absorbing shortwave radiation, and it is often cited as the second strongest contributor
to atmospheric warming after carbon dioxide. The UHI effect describes the phenomena of warmer urban
areas compared to surrounding rural areas, caused by heat trapping properties of the urban built
environment, as well as heat generated from anthropogenic activities. Using field measurements and
modeling techniques, this dissertation seeks to constrain important properties of non-GHG agents, assess
their impacts on local-to-regional climate, and quantify the efficacy of proposed urban heat mitigation
strategies. First, I will describe how in-situ, single-particle field measurements were conducted on Catalina
Island, CA to constrain BC microphysical properties, which modulate its atmospheric warming efficiency.
Second, I will share how micrometeorological measurements, in tandem with causal inference methods,
were used to quantify the first observable air temperature reductions from real-world installations of solar
reflective cool pavement. Lastly, I will present how a high-resolution (100 m, hourly) anthropogenic heat
flux (AHF) dataset was developed and subsequently used to simulate the impacts of AHF on the urban
climate of Los Angeles. This dissertation highlights the intricate connections between air quality, the urban
environment, and regional climate; broadens our understanding of non-GHG warming agents; and
investigates mitigation pathways towards sustainable cities in the context of our warming planet.

1
Chapter 1 – Introduction
1.1 Background
There is unequivocal evidence that humans have altered the Earth’s climate system through
anthropogenic activities (IPCC 2021). According to the latest IPCC assessment report (2021),
anthropogenic climate change is already influencing weather and climate extremes all over the globe, and the
frequency and intensity of such extreme events are projected to increase along with global warming. The
main driver of anthropogenic climate change is the emission of greenhouse gases (GHGs), such as carbon
dioxide and methane, from human activities such as the combustion of fossil fuels, industrial activities,
agricultural activities, and oil/gas production. GHGs absorb upwelling longwave radiation from the Earth’s
surface and historical emissions of GHGs due to human activities have already increased global average
temperatures by ~1.1 °C compared to the pre-industrial (1850 – 1900) average (IPCC 2021).
While it is abundantly clear that urgent actions must be taken worldwide to reduce GHGs to prevent
even more severe impacts of global climate change in the future, there are also various non-GHG factors
that will affect the degree of atmospheric warming felt at smaller spatial and temporal scales. Furthermore,
there is a greater degree of uncertainty about the extent and impacts of non-GHG warming agents
compared to GHGs, which have relatively direct impacts on global warming. In addition to these
uncertainties, there is also a need to quickly identify how to best mitigate and adapt to the realities of climate
change at smaller local and regional scales, especially in urban environments, where the impacts of
increasing extreme heat are particularly pronounced (Oke 1973; Stone, Vargo, and Habeeb 2012). Through
the studies presented here, we seek to characterize and constrain factors that influence non-GHG warming,
assess the impacts of non-GHG warming at local to regional scales, and quantify the effectiveness of
proposed mitigation strategies. In particular, the two non-GHG warming agents that will be discussed in
this work are (1) black carbon (BC) aerosols and (2) the urban heat island (UHI).
2

1.2 Significance of research
1.2.1 Black carbon aerosols
BC is emitted from incomplete fossil-fuel combustion, and it is often cited as the second strongest
warming agent in the atmosphere after carbon dioxide (Bond et al. 2013; Fierce et al. 2020). Unlike GHGs,
BC aerosols have relatively short lifetimes (~days to weeks) and heterogeneous physical and chemical
characteristics. The variable characteristics of BC make it difficult to constrain its climate impacts because
these characteristics strongly influence their atmospheric distribution and radiative absorption efficiency
(Bond and Bergstrom 2006; Cappa et al. 2012; Bond et al. 2013; Fierce et al. 2020). The complexity of BC is
largely due to the fact that BC aerosols evolve as they age in the atmosphere and get internally mixed with
other aerosols and condensing gaseous species. Ultimately, the physical and chemical composition of BC
aerosols influence their deposition rates and radiative forcing, making it crucial to understand these
underlying characteristics of BC.
One particular characteristic that we seek to constrain is the BC mixing state. In other words, the
extent of physical coatings on BC aerosols directly influences its deposition propensity and its radiative
absorption. Generally, BC coating tends to increase hygroscopicity and enhance radiative absorption
efficiencies. The increased hygroscopicity of BC due to thicker coatings will directly impact its propensity
for wet deposition, altering its atmospheric lifetime and therefore the general regional distribution of BC
(McMeeking, Good, et al. 2011; J. Zhang et al. 2015). Previous literature shows that BC coating will induce a
“lensing effect” that enhances its ability to absorb shortwave radiation, ultimately increasing its radiative
forcing impact (Moteki et al. 2007; Q. Wang et al. 2014).
While there have been several laboratory and field experiments conducted to study BC mixing state,
there have been considerable variability in results (Metcalf et al. 2012; Krasowsky et al. 2018; Joshua P.
Schwarz, Gao, et al. 2008; Q. Wang et al. 2018). Furthermore, understanding regional variability of BC
3
mixing state is a crucial component in properly characterizing BC radiative impacts at sub-global scales,
since BC emission sources and atmospheric aging processes vary widely from region to region (J. Zhang et
al. 2015). In our work here, we seek to quantify the characteristic coating thickness of various BC
populations in Southern California as a function of its emission source and atmospheric aging, using single
particle, in-situ field measurements. The results of our study, as presented in chapter 2, help to confirm, and
constrain previous estimates of BC mixing state, adding to a limited body of BC mixing state field
measurements that exist to date.
1.2.2 The urban heat island effect
Another non-GHG warming agent is the UHI effect, which can simply be described as higher urban
temperatures relative to surrounding rural or suburban temperatures (Oke 1973). The UHI is primarily
caused by (1) urban materials (e.g., asphalt concrete) that absorb and retain heat due to their optical and
thermal properties, (2) urban geometries that trap radiation in the urban canopy, (3) anthropogenic waste
heat from human activities (e.g., air conditioning), and (4) the replacement of natural vegetation with
impervious surfaces (Howard 1833; Oke 1973). Previous studies have discussed the impact of elevated
urban temperatures on increased energy demand for air conditioning (M. Chen, Ban-Weiss, and Sanders
2018; Hassid et al. 2000), increased risk of critical infrastructure failure (e.g., electrical grid failures) (Stone et
al. 2021), potential deterioration of air quality (Sarrat et al. 2006), and increased risk of heat-related illness
and mortality (Patz et al. 2005; D. Li and Bou-Zeid 2013). With the compounded effects of global climate
change and increasing urban populations around the globe (United Nations 2019), it is important that we
have a firm understanding of the underlying factors contributing to the UHI, as well as the various
mitigation strategies that we can employ to potentially reduce the negative impacts of the UHI (Vahmani
and Ban-Weiss 2016; Zhao, Lee, and Schultz 2017).
4
1.2.2.1 Anthropogenic heat
As mentioned above, anthropogenic waste heat can be a significant contributor to the overall UHI
effect (Ichinose, Shimodozono, and Hanaki 1999; Sailor and Lu 2004). For example, Ichinose et al. (1999)
estimated that certain regions of Tokyo could emit as much as 1590 W m
-2
of anthropogenic heat during
early morning hours, with city-scale temperature impacts of ~1.0 °C. Subsequently, several modeling studies
have characterized the anthropogenic heat flux (AHF) in various cities and regions around the world (Chow
et al. 2014; Sailor and Lu 2004; Y. Zheng and Weng 2018). However, the anthropogenic heat contribution
to the UHI is uncertain in the Los Angeles region due to the limited number of studies, as well the lack of
accessible data, for the region. For example, Sailor and Lu (2004) estimated that the City Los Angeles has a
maximum average AHF of ~30 W m
-2
, but the results were aggregated to the city-level and the spatial
variability of anthropogenic heat at more granular resolutions was not presented. On the other hand, a more
recent estimate by Zheng et al. (2018) estimated that the maximum average AHF of Los Angeles County
was less than 8 W m
-2
and provided an example of an AHF heatmaps highlighting the spatial variability of
AHF for a select hour of the day. However, results were limited to the spatial aggregated selected by the
authors of the study (e.g., average AHF not presented for City of LA alone), and the underlying dataset was
not provided in a publicly accessible manner. Furthermore, there is no existing study that quantifies the
direct impact of anthropogenic heat on the regional climate of the Los Angeles using mesoscale
meteorological modeling. In our work, we propose to first constrain and improve the regional estimates of
anthropogenic heat in Los Angeles using methods described in chapter 4, then proceed to use a mesoscale
climate model to directly quantify the regional climate impact of anthropogenic heat in Los Angeles.
1.2.2.2 Urban heat mitigation strategies
In the past few decades, there has been an increasing interest in heat mitigation and adaptation
strategies, particularly by local and regional government. A host of urban heat mitigation strategies have
been proposed, but the most prominent proposals include solar reflective cool surfaces (e.g., roofs,
5
pavements, and walls), urban greening to increase shading and evapotranspiration, “green” vegetative roofs
and walls, and non-vegetative shading structures (Jiachen Zhang et al. 2018; Rosenfeld et al. 1998; Taleghani,
Sailor, and Ban-Weiss 2016; Taleghani et al. 2019; Hashem Akbari et al. 2016; Georgescu et al. 2014). In
short, these mitigation strategies seek to alter land cover and/or land use in such a way that the surface
energy balance is altered and ultimately traps less heat in the urban boundary layer. One such strategy that
has gained traction in some policy circles is the installation of solar reflective cool pavements. Although a
host of modeling studies and some smaller scale measurement studies have confirmed that cool pavements
lower pavement surface temperatures and likely air temperatures at large spatial scales, there is a lack of real-
world field measurements to confirm these estimates (Rosenfeld et al. 1998; Santamouris, Gaitani, et al.
2012; Taleghani, Sailor, and Ban-Weiss 2016; H. Li, Harvey, and Kendall 2013). Only a single field
measurement-based study exists to date (Middel et al. 2020), and they were unable to confirm if air
temperatures were reduced with statistical robustness, due to the lack of pre-installation field measurements.
In our study (see chapter 3), we used real-world field measurements to assess the impacts of neighborhood-
scale cool pavements on surface and air temperatures in the Los Angeles region, as well as monitor the
performance of the cool pavement over the span of one year.

6
Chapter 2 - Measurements to determine the mixing state of black
carbon emitted from the 2017 –2018 California wildfires and urban
Los Angeles
Published in Atmospheric Chemistry and Physics (Ko, Krasowsky, and Ban-Weiss 2020)
2.1 Introduction
Atmospheric black carbon (BC) is a carbonaceous aerosol that can result from the incomplete
combustion of carbon-containing fuels. Major energy-related sources of BC include vehicular combustion,
power plants, residential fuel-use, and industrial processes. Biomass burning, which can be either
anthropogenic or natural, is another significant BC source. BC is a pollutant of particular interest for two
main reasons: (1) it absorbs solar radiation, which results in atmospheric warming (Ramanathan and
Carmichael 2008), and (2) it is associated with increased risk of cardiopulmonary morbidity and mortality
(World Health Organization 2012). Regarding its effect on climate, BC is widely considered to be the second
strongest contributor to climate warming, after carbon dioxide (Bond et al. 2013). Although it has been
established that BC is a strong radiative forcing agent in Earth’s atmosphere, there remains considerable
uncertainty about the extent to which BC affects Earth’s radiative budget, from regional to global scale
(IPCC 2021; Bond et al. 2013).
Since the lifetime of BC is relatively short (~days to weeks), the spatiotemporal distribution of BC is
highly heterogeneous and thus difficult to quantify (Krasowsky et al. 2018). The quantification of where and
when BC is emitted around the world is also a challenging task that causes significant uncertainty (Bond et
al. 2013). In addition to the difficulties that come with tracking the emissions and distribution of BC, there
are complex physical and chemical processes that govern the transformation of BC in the atmosphere,
which ultimately impact its climate and health effects. These atmospheric processes, in addition to the
emissions source type, influence the BC mixing state in a highly dynamic manner. A BC particle that is
physically separate from other non-BC aerosol species is considered externally mixed. On the other hand, BC
7
is considered internally mixed if it is physically combined with another non-BC aerosol species (Bond and
Bergstrom 2006; Joshua P. Schwarz, Gao, et al. 2008). As freshly emitted BC particles are transported in the
atmosphere, they can obtain inorganic and organic coatings from either gaseous pollutants that condense
onto the BC, oxidation reactions on the BC surface, or the coalescence of other aerosol species onto the
BC, making them more internally mixed (He et al. 2015). In general, the mixing state of BC describes the
degree to which BC is internally mixed (Bond et al. 2013). The BC mixing state near the point of emission as
well as the evolution during aging in the atmosphere of the mixing state can vary widely, depending on the
source of emissions and atmospheric context.
The evolution of rBC mixing state as BC ages in the atmosphere is crucial to understand for two
reasons. First, it has been shown that non-refractory coatings on BC can enhance its absorption efficacy,
implying that internally mixed BC with thick coatings can have stronger warming potential in the
atmosphere compared to uncoated or thinly-coated BC (Moteki et al. 2007; Q. Wang et al. 2014). Second,
coatings on BC can alter the aerosol’s hygroscopicity and effectively shorten its lifetime by increasing the
probability of wet deposition (McMeeking, Good, et al. 2011; J. Zhang et al. 2015). In short, freshly emitted
BC particles are generally hydrophobic, but coatings acquired during the aging process can make BC-
containing particles hydrophilic, and therefore, more susceptible to wet deposition. Thus, uncertainties in
the evolution of rBC mixing state directly translate to uncertainties regarding BC’s impact on Earth’s climate
due to both the radiative impact per particle mass and spatiotemporal variation of atmospheric BC loading.
Although there have been several laboratory experiments (Q. Wang et al. 2018; He et al. 2015;
Slowik et al. 2007; Knox et al. 2009) and field campaigns (Krasowsky et al. 2018; Metcalf et al. 2012; Cappa
et al. 2012; Joshua P. Schwarz, Gao, et al. 2008) studying rBC mixing state, there is considerable variability
in results. For example, field studies in China suggest that the mass absorption cross-section (MAC) of BC
that has aged for more than a few hours should be enhanced by a factor of ~2 (Q. Wang et al. 2014), while
other studies in California reported an absorption enhancement factor of ~1.06 (Cappa et al. 2012) and
8
~1.03 (Krasowsky et al. 2016). Preceding these studies, Bond et al. (2006) suggested an enhancement factor
of ~1.5 based on a review of laboratory and field studies. The wide range of reported values is not
surprising given that rBC mixing state is expected to be influenced by a variety of spatiotemporal factors
such as source type, season, and regional atmospheric composition (Krasowsky et al. 2018). In other words,
BC aged in different places and at different times may have significantly varying mixing states, resulting in a
wide range of absorption and hygroscopicity enhancements in the real world.
Quantifying BC mixing state is challenging because it requires single-particle analysis (Bond and
Bergstrom 2006). There are two main methods to measure rBC mixing state: (1) microscopy (Johnson et al.
2005; Adachi, Chung, and Buseck 2010; Adachi et al. 2016), and (2) real-time, in-situ measurements (Hughes
et al. 2000). In our study, we quantify rBC mixing state by taking real-time, in-situ measurements with a
single particle soot photometer (SP2). The SP2 uses laser-induced incandescence to measure refractory black
carbon (rBC) mass per particle, which can be used to directly compute the mass concentration, number
concentration, and mass size distribution, and indirectly compute the number size distribution (Stephens,
Turner, and Sandberg 2003). The SP2 can also measure the mixing state of rBC using one of two different
methods. In the lag-time method, each sensed rBC-containing particle is deemed as either thinly-coated or
thickly-coated using the measured time difference between the peak of the incandescence and scattering
signals induced by the particle (Moteki and Kondo 2007, 2008). In the leading-edge-only (LEO) method, the
actual coating thickness for rBC-containing particles can be explicitly quantified (Gao et al. 2007). Further
detail regarding these two methods can be found in sections 2.2.6 and 2.2.7. In this study, we used both
methods to quantify the rBC mixing state.
In this study, we measured rBC with an SP2 on Catalina Island, California (~70 km southwest of
Los Angeles) during three different time periods, with the goal of observing how rBC loading and mixing
state varied as a function of source type and source-to-receptor timescale. During the first campaign
(September 2017), westerly winds dominated and thus the sampling location was upwind of the dominant
9
regional sources of rBC (i.e., urban emissions from the Los Angeles basin). We suspect measurements
during this period to represent well-aged particles; evidence suggests that some of the measured particles
originated from wildfires in Oregon and Northern California. In contrast, the second and third campaigns
(December 2017, November 2018) were dominated by northerly-to-easterly “Santa Ana conditions”, which
advected fresh and aged rBC-containing particles from both biomass burning emissions and urban
emissions. Several significant wildfires were active in the Southern and Northern California regions
throughout the second and third campaigns. In particular, the Thomas Fire, which was active in Southern
California during the second campaign, was the second largest wildfire in modern California history. The
Camp Fire, which was active in Northern California during the third campaign, was the 16
th
largest fire in
terms of burn area size, and was also considered the deadliest and most destructive wildfire in modern
California history. Table 2-1 lists the two most significant wildfires for each campaign period that impacted
our rBC measurements, along with the total burn area and time period of non-containment for each fire.
Mass and number concentrations of rBC-containing particles, rBC size distributions, the number fraction of
thickly-coated rBC-containing particles (i.e., using the lag-time method), and absolute coating thickness
values (i.e., using the LEO method) are reported. We then evaluate how the rBC loading, size distribution,
and mixing state relate to the meteorology and major sources at the time of measurements in order to
further understand the microphysical transformation of BC as it ages in the atmosphere. While a few past
studies have investigated the mixing state of rBC in the Los Angeles region using the SP2 (Metcalf et al.
2012; Cappa et al. 2012; Krasowsky et al. 2018), this study is the first to use fixed ground-based
measurements off the coast of Los Angeles to focus on how (a) wildfire source-to-receptor travel time, and
(b) wildfire versus urban emissions, influence rBC mixing state.

10
Table 2-1. Major wildfires that were active during the three campaigns. Only the two largest fires from each campaign (in terms
of burn area) are listed in the table below. Note that there were numerous other smaller fires that were active during the three
campaigns, but not listed in this table.

2.2 Methods
2.2.1 Measurement location and time periods
All measurements reported in this study were conducted at the USC Wrigley Institute for
Environmental Studies on Catalina Island (~33°26'41.68"N, 118°28'55.98"W). Catalina Island is located
approximately 70 km (43.5 miles) southwest of downtown Los Angeles. Figure 2-1 shows the location of
the sampling site relative to the Los Angeles metropolitan area. The three campaigns were conducted from 7
to 14 September 2017, 20 to 22 December 2017, and 12 to 18 November 2018, Pacific Time (local time).
Campaign Wildfire name Location Area (km
2
) Start date Containment date
First
(September
2017)
Chetco Bar Fire
Rogue River–
Siskiyou National
Forest, Oregon
773 12 July, 2017 2 November, 2017
Eclipse Complex Siskiyou, California 318 15 August, 2017 29 November, 2017
Second
(December
2017)
Thomas Fire
Ventura and Santa
Barbara, California
1,140 4 December, 2017 12 January, 2018
Creek Fire
Los Angeles,
California
63 5 December, 2017 9 January, 2018
Third
(November
2018)
Camp Fire Butte, California 620 8 November, 2018 25 November, 2018
Woolsey Fire
Ventura and Los
Angeles, California
392 8 November, 2018 22 November, 2018
11

Figure 2-1. Overview map showing the location of the sampling site with respect to the Greater Los Angeles (LA) area. Map data
© 2015 Google.
2.2.2 Instrumentation
An SP2 (Droplet Measurement Technologies, Boulder, CO) was used to quantify the physical
characteristics of rBC-containing particles for all three campaigns. In short, the SP2 uses laser-induced
incandescence to quantify rBC mass on a particle-by-particle basis. The SP2 uses a continuous Nd:YAG
laser (λ = 1064 nm) that is oriented perpendicular to the flow of air containing rBC-containing particles. As
each particle passes through the intra-cavity laser, any coating on the rBC particle vaporizes while the core
incandesces and emits thermal radiation. The scattered and thermally emitted radiation is measured by
optical sensors and converted to signals that can then be used to obtain information about the mass and
mixing state of the sampled rBC-containing particles. In this study, an assumed rBC density of 1.8 g cm
-3

was used. The SP2 has detection limits from ~0.5 to 50 fg rBC per particle. Further details regarding the
12
governing principles and operation of the SP2 can be found in numerous publications (Stephens, Turner,
and Sandberg 2003; Joshua P. Schwarz et al. 2006; Gao et al. 2007; Moteki and Kondo 2007; Laborde,
Schnaiter, et al. 2012; Dahlkötter et al. 2014; Krasowsky et al. 2016; Joshua P. Schwarz, Gao, et al. 2008;
Joshua P. Schwarz, Spackman, et al. 2008).
The inlet of the SP2 was positioned on the roof of a three-story research building at the Wrigley
Institute as shown in Fig. S1 (Appendix A). The height of the inlet was approximately 15 meters above
ground level. A fine mesh was secured to the tip of the inlet to prevent clogging by small insects, and a small
plastic cone was also attached to block any potential precipitation from entering the inlet. The inlet tube was
fed in through a window of a secure laboratory room on the top floor of the building where the SP2 was
housed for the duration of sampling. The SP2 ran continuously for the duration of the three measurement
periods. Desiccant used to remove moisture from the sample air was replaced daily, and the data during
these replacement periods were subsequently removed during the data analysis.
2.2.3 Auxiliary data
Model simulations and publicly available auxiliary datasets were used to supplement our SP2
measurements. The National Oceanic and Atmospheric Administration’s (NOAA) Hybrid Single-Particle
Lagrangian Integrated Trajectory (HYSPLIT) model (Stein et al. 2015) was the primary tool used to identify
dominant emissions sources. The HYSPLIT back-trajectories were also used to estimate the age range of
measured rBC-containing particles and the path of the air masses carrying these particles. The HYSPLIT
trajectory model requires the user to specify the following input parameters: meteorological database,
starting point of the back-trajectory, height of source location, run time, and the vertical motion method. A
height of 15 meters above ground level was chosen to approximately represent the height of the SP2 inlet
positioned on the roof of the laboratory facility. For the first campaign (September 2017), the Global Data
Assimilation System (GDAS) meteorology database with 1-degree resolution (~110 km for 1-degree latitude
and ~85 km for 1-degree longitude) was selected, and one-week back-trajectories were simulated for every
13
day of the first campaign. For the second and third campaigns (December 2017, November 2018), the
High-Resolution Rapid Refresh (HRRR) meteorology database with a 3-km resolution was selected, and 72-
hour back-trajectories were simulated starting on every hour. The GDAS database was selected for the first
campaign simulations because a 1-degree resolution was sufficient to show that the measured air masses
were generally coming from the west. In contrast, the HRRR database was used for the second and third
campaigns because a finer resolution helped determine the sources that contributed to measured rBC. The
default vertical motion method was selected for all back-trajectory simulations.
Data from local weather stations were used to identify the meteorological regimes during all three
campaigns, and to supplement the HYSPLIT back-trajectories used for source characterization. Hourly
weather data from Los Angeles International Airport (LAX), Long Beach Airport, Avalon (Catalina Island),
Santa Barbara, and Oxnard, during September 2017, December 2017, and November 2018, were obtained
using the NOAA National Center for Environmental Information online data tool
(https://www.ncdc.noaa.gov/cdo-web/datatools/lcd, last access: 26 August 2019). Five-minute weather
data at the same weather stations and time periods were obtained from the Iowa Environmental Mesonet
website (Todey, Herzmann, and Takle 2002). Wind data from the USC Wrigley Institute on Catalina Island
were also examined when available (7 to 13 September 2017) on the Weather Underground website
(https://www.wunderground.com/weather/us/ca/catalina, last access: 26 August 2019), though these data
are not validated by NOAA. Data from Santa Barbara, Oxnard, and USC Wrigley Institute were assessed to
support conclusions made in this study but are not directly presented in any of the analyses here.
In addition to meteorological data, weather information from local news reports, NASA satellite
imagery, and global aerosol model data were used in conjunction to explain the variability in rBC
concentrations and mixing state during the sampling campaigns. Local weather news reports between 20
December and 22 December 2017 were used to obtain information about the active fires in Southern
California and the dominant wind conditions for each day in the second campaign (December 2017) (CBS
14
Los Angeles 2017a, 2017c, 2017b, 2017e, 2017d, 2017f). There were generally two local weather reports
retrievable per day: one in the early morning and one later in the evening. The information from these
reports was used to get a holistic picture of the local fire and weather conditions at the time of sampling.
Data from the California Department of Forestry and Fire Protection (https://www.fire.ca.gov/incidents/,
last access: 26 August 2019) was also used to verify basic spatial and temporal information about significant
fires occurring during sampling periods. The local weather reports were used to cross-validate wildfire
timelines, but they are not directly presented here.
NASA satellite imagery and data were accessed through NASA’s Worldview online application
(https://worldview.earthdata.nasa.gov/, last access: 26 August 2019), which provides public access to
NASA’s Earth Observing System Data and Information System (EOSDIS). Moderate Resolution Imaging
Spectroradiometer (MODIS) images taken from two satellites (Aqua and Terra) were examined for all
sampling days. MODIS images were used to identify visible plumes of aerosols, particularly those from large
wildfires. The general movement of air masses was also assessed from the visible movement of large-scale
clouds from these satellite images. In addition to the MODIS images, aerosol index, aerosol optical depth
(AOD), and fires and thermal anomalies data products were examined to supplement the source
identification process. For aerosol index, the OMAERUV (Torres 2006) and OMPS_NPP_NMTO3_L2
(Jaross 2017) products were used. For AOD, the MYD04_3K MODIS/Aqua and MYD04_3K
MODIS/Terra products were used (Levy et al. 2013). For fires and thermal anomalies, the
VNP14IMGTDL_NRT (Schroeder et al. 2014) and MCD14DL (Justice et al. 2002) products were used.
Examples of NASA data products used for source identification analysis can be found in Appendix A.
An open-source online visualization tool (earth.nullschool.net, last access: 26 August 2019) was used
to visually assess the European Centre for Medium-Range Weather Forecasts (ECMWF) Copernicus
Atmosphere Monitoring Service (CAMS) model output data (Beccario n.d.). The CAMS model provides
“near-real-time” forecasts of global atmospheric composition on a daily basis. Specifically, the PM
2.5

15
concentration output data from CAMS were examined using earth.nullschool.net. The CAMS output
visualizations were particularly helpful for understanding where certain sources were located and when they
were likely affecting our measurements. The concentration gradients of PM
2.5
were examined on the
visualization tool on an hourly interval for every day of active sampling in order to supplement the
HYSPLIT analysis and confirm the contribution of certain emission sources. Access to the CAMS
visualizations for the three campaigns can be found in Appendix A and Video Supplement.
2.2.4 Estimation of source-to-receptor timescale
Characteristic timescales of transport between the sampling site and nearest source(s) were estimated
based on the HYSPLIT trajectories simulated for source identification. The approximate source-to-receptor
timescale characterizations by HYSPLIT were cross-validated with approximate calculations of transport
time performed with representative length scales between sources and the sampling site, and the average
wind speeds during the time periods of interest. Further details regarding the calculations of the timescale
characterizations are in section S1 of Appendix A. Although we cannot fully capture the intricacies of
particle aging timescales with our estimates, they are meant to be conservative approximations based on
available meteorological data. These estimated source-to-receptor timescales were used to help categorize
different LEO periods by source(s) (see Table 2-2 and Fig. 2-9), and also used in our discussion of how rBC
mixing state evolves with particle aging (see Section 2.3.7).
2.2.5 Time series filtering
rBC mass and number concentrations during the first campaign (September 2017) showed
anomalous spikes likely due to unexpected local sources. In an effort to obtain representative background
concentrations, we filtered these spikes by removing values above a threshold of 0.08 μg m
-3
and 40 cm
-3
for
mass and number concentrations, respectively. Figure S2 in Appendix A shows the time series for the first
campaign before and after removal of spikes. Figure S3 in Appendix A shows the median rBC concentration
16
for the first campaign as a function of the cut-off threshold value. Median rBC mass and number
concentrations appeared to asymptote at cut-off values of approximately 0.08 μg m
-3
and 40 cm
-3
, suggesting
that the median rBC concentration values become insensitive to the choice of cut-off threshold above these
values.
2.2.6 Lag-time method
The mixing state of rBC was examined using two different methods. The first method used to
characterize mixing state is called the lag-time method. This method categorizes each rBC particle as either
“thickly-coated” or “thinly-coated” based on a measured time delay (i.e., “lag-time”) between the scattering
and incandescence signal peaks. This method has been previously described and used in various studies
(Moteki and Kondo 2007; McMeeking, Good, et al. 2011; Metcalf et al. 2012; Sedlacek et al. 2012; Q. Wang
et al. 2014; Krasowsky et al. 2016, 2018). In short, as a coated rBC-containing particle passes through the
SP2 laser, the sensors will detect a scattering signal as the coating vaporizes. Shortly after, there will be a
peak in the incandescence signal as the rBC core heats up and emits thermal radiation. A probability density
function of the lag-time values often results in a bimodal distribution. Based on the data for a particular
campaign, a lag-time cut-off is chosen between the two peaks of the bimodal distribution to bin each rBC
particle as either thinly or thickly-coated. The fraction of rBC particles that are thickly-coated (f
BC
) is then
determined based on this categorization. For our study, a lag-time cut-off of 1.8 μs was chosen to quantify
whether an rBC-containing particle was thickly-coated. Only particles with an rBC core diameter greater
than 170 nm were included in the calculation of f
BC
to account for the scattering detection limit of the
instrument. As discussed previously by Krasowsky et al. (2018), the lag-time method is inherently
susceptible to biases since f
BC
can depend on the selection of the lag-time cut-off value. For example,
Krasowsky et al. selected a cut-off value of 1 μs for a near-highway SP2 campaign in the Los Angeles Basin,
which is significantly different than the value of 1.8 μs used in this study and others. There remains an
unresolved issue of maintaining consistency between different studies utilizing the lag-time method, while
17
simultaneously representing the unique mixing state characterization of each measured rBC population; the
definition of “thickly-coated” likely varies by the aerosol population sampled and thus is not necessarily
comparable from one study to the next.
2.2.7 Leading-edge-only (LEO) method
BC mixing state was also characterized using the LEO method. In brief, this method reconstructs a
Gaussian scattering function from the leading edge of the scattering signal for each rBC-containing particle.
The width and location of the reconstructed Gaussian scattering function is determined by a two-element
avalanche photodiode. Assuming a core-shell morphology, the rBC coating thickness is subsequently
calculated from the reconstructed scattering signal and the incandescence signal (Gao et al. 2007; Moteki
and Kondo 2008). The Paul Scherrer Institute’s single-particle soot photometer toolkit version 4.100b
(developed by Martin Gysel et al.) was used to perform the LEO method in Igor Pro version 7.09.
In this study, the LEO “fast-fit” method was used with three points, and particles analyzed were
restricted to those with rBC core diameters between 180 and 300 nm. Although the SP2 has been reported
to accurately measure the volume equivalent diameter (VED) of scattering particles down to ~170 nm, a
more conservative lower threshold of 180 nm was used for our study to reduce instrument noise at smaller
VED values near the detection limit (Krasowsky et al. 2018). Specific rBC core diameter ranges were used
for different analyses in this study and these ranges are explicitly defined within each respective discussion.
One exception was made to the 180–300 nm rBC core diameter restriction in section 2.3.7. For the analyses
and discussion presented in section 2.3.7, the LEO coating thickness was calculated for all detectable rBC
particles with non-saturated scattering signals. The rBC core size was not restricted in this section because
the relative comparisons between characteristic coating thickness values were more important for the
analysis, rather than the absolute value (which would likely be biased, as discussed further in section 2.3.8).
In other words, the LEO-derived coating thickness values in section 2.3.7 were not used to report
18
representative averages for selective time periods, but rather were used for comparative and/or qualitative
purposes.
2.3 Results and discussion
This section starts by discussing the major identifiable sources and meteorological patterns in each
of the three campaigns (section 2.3.1). Then, the overall mass and number loading of rBC is discussed and
compared to past literature values (section 2.3.2). Following that, the rBC mixing state results from the lag-
time and LEO analyses are discussed (sections 2.3.3–2.3.5). The impacts of emissions source type and
atmospheric aging on rBC mixing state and core size are subsequently discussed (sections 2.3.6, 2.3.7).
Section 2.3 then ends by comparing rBC coating thickness values calculated in this study to reported values
from past studies
2.3.1 Source identification and meteorology
In this section, we summarize the dominant pollutant sources and wind patterns for each of the
three campaigns. For all three campaigns, we used HYSPLIT back-trajectories, HYSPLIT dispersion model,
CAMS model data, and NASA data products (i.e., satellite imagery, aerosol index products, and AOD
products) in conjunction to identify the most likely sources of measured rBC-containing particles. For the
first campaign (September 2017), the Oregon wildfires were identified as probable sources of measured rBC.
Furthermore, we also identified long-range transport from East Asia and ship/aviation emissions as
potential sources contributing to measured rBC. Overall, we expect measured rBC during the first campaign
to be aged. For the second campaign (December 2017), fresh urban emissions from the Los Angeles basin
and biomass burning emissions from the Thomas Fire in Santa Barbara and Ventura County (along with
other smaller Southern California fires) were the main sources identified by our analysis. For the third
campaign (November 2018), fresh urban emissions from the Los Angeles basin and fresh biomass burning
emission from the Woolsey Fire in Ventura (along with other smaller Southern California fires) were the
19
main sources identified for approximately the first four days of the campaign. For the last two days of the
third campaign, the Camp Fire in Northern California (along with other smaller fires in Northern and
Central California) contributed significantly to measured rBC. Figure 2-2 displays wind roses for each
campaign at three different weather station locations (public data provided by NOAA, see section 2.2.3).
Furthermore, Fig. 2-3 shows HYSPLIT back-trajectories simulated for each of the three campaigns and
further highlights the differences in wind conditions between the three campaigns. These figures show the
distinct meteorological regimes of each campaign. A more detailed description of the source identification
process can be found in section S2 of Appendix A.
For the remainder of the manuscript, we refer to rBC measured when the dominant source was
biomass burning emissions as BC
bb
, and rBC measured when the dominant source was fossil fuel (i.e.,
urban) emissions will be referred to as BC
ff
. rBC measured in the first campaign (September 2017), when the
measured air masses were representative of well-aged background over the Pacific Ocean, will be referred to
as BC
aged,bg
.

20

Figure 2-2. Wind roses for the September 2017 (first row), December 2017 (second row), and November 2018 (third row)
sampling periods. Wind roses are based on five-minute ASOS airport data from LAX (first column), LGB (second column), and
AVX (third column), provided by NOAA.

21

Figure 2-3. HYSPLIT back-trajectories for all three campaigns. The star denotes the start location of each back-trajectory, i.e.,
the sampling location. The trajectories for the first period (September 2017) (i.e., panel a) represent week-long back-trajectories
for each day of the campaign. The trajectories for the (b) second (December 2017) and (c) third (November 2018) periods
represent 72-hour back-trajectories for each hour of the campaign. Panels (d) and (e) show more zoomed-in maps of the second
and third campaign back-trajectories along with active Southern California fires. Map data: Google DigitalGlobe.
22
2.3.2 rBC mass and number concentration
Figure 2-4 shows time series for rBC mass and number concentrations, rBC coating thickness
(CT
BC
), number fraction of thickly-coated particles (f
BC
), and rBC count mean diameter (CMD) for all three
measurement campaigns. The mixing state (CT
BC
and f
BC
) and rBC size are discussed in following sections.
The mean mass and number concentration (±standard deviation) for the first campaign (September 2017)
was 0.04 (±0.01) μg m
-3
and 20 (±7) cm
-3
, respectively. For the second campaign (December 2017), the
corresponding mean concentrations were 0.1 (±0.1) μg m
-3
and 63 (±74) cm
-3
, with concentrations reaching
as high as 0.6 μg m
-3
and 381 cm
-3
. Likewise, for the third campaign (November 2018), the corresponding
mean concentrations were 0.15 (±0.1) μg m
-3
and 80.2 (±54.5) cm
-3
. The range of observed rBC
concentrations is larger for the second and third campaigns compared to the first campaign, and there are
distinct prolonged peaks in concentrations that can be observed during these times. In comparison, the first
campaign shows relatively stable concentrations.
Given the remote location of the sampling site and the consistent westerly winds during the first
campaign (September 2017), the observed rBC concentrations establish an appropriate baseline for ambient
conditions away from the broader urban plume in the Los Angeles basin. On the other hand, the
concentrations during the second and third campaigns (December 2017, November 2018) were more
variable, with mean concentrations that were higher than the first campaign due to periods of northerly-to-
easterly winds driven by Santa Ana wind conditions as described in section 2.3.1. Figure 2-5 shows rBC
mass and number concentrations along with wind speed and direction during the second campaign. Wind
direction was directly related to elevated concentrations for all three peaks shown. Peak P1 is clearly
preceded by a prolonged period of northerly winds. Similarly, Peaks P2 and P3 are preceded by periods of
easterly winds. An analogous plot for the third campaign is shown in Fig. S8 (Appendix A), but the
relationship between wind direction measured at LAX and the rBC concentration is not clearly discernible
since long distance biomass emissions were impacting the measurements in addition to local sources near
23
the LA basin. The impacts of different sources on measurements during the third campaign are described in
detail in section S2 of Appendix A.
The mean concentration for the first campaign (September 2017) was approximately an order of
magnitude lower than the mean concentration of ~0.14 μg m
-3
observed by Krasowsky et al. (2018) near the
downwind edge (assuming dominant westerly wind flows) of the LA Basin (i.e., Redlands, CA).
Concentrations during the most polluted time periods in our measurements were comparable to recently
measured concentrations in the Los Angeles basin (Krasowsky et al., 2018) but at least one to two orders of
magnitude lower than average concentrations found in other heavily polluted cities around the world. Mass
concentration values of ~0.9 μg m
-3
, ~0.5 to 2.5 μg m
-3
, ~0.9 to 1.74 μg m
-3
, and ~0.6 μg m
-3
were measured
with an SP2 in Paris, Mexico City, London, and Houston, respectively (Laborde et al. 2013; Baumgardner,
Kok, and Raga 2007; D. Liu et al. 2014; Joshua P. Schwarz, Gao, et al. 2008). In urban areas of China, an
average mass concentration of ~9.9 μg m
-3
was reported for a polluted period in Xi’an (Q. Wang et al.
2014).

24

Figure 2-4. Time series of (a) BC absolute coating thickness, (b) number fraction of thickly-coated rBC particles, (c) rBC count
median diameter, and (d) rBC concentrations, for all three measurements campaigns. The boxed annotations (i.e., L1 to L10) refer
to specific LEO periods, which are further described in Section 2.3.5. In panel (a), each blue dot represents an individual particle.
The hourly median is shown in the dotted pink line, and the corresponding 10
th
and 90
th
percentiles are shown in purple. In panel
(b), green dots represent one-minute means while the black curve shows hourly means. Panel (c) shows the one-minute mean for
the count mean diameter. Panel (d) shows the one-minute means for rBC concentration.
25

Figure 2-5. Meteorological variables and rBC concentrations during the second campaign (December 2017). Panel (a) shows
wind speed and (b) shows wind direction measured by a NOAA weather station located at Los Angeles International Airport
(LAX). Panel (c) shows rBC mass and number concentrations and identifies three peaks of interest. The two dashed ovals in
panel (b) highlight periods of northerly and easterly winds, which occur ~0.5-1 days before each of the three peaks, suggesting
that the elevated rBC concentrations included important contributions from the local Thomas Fire (and other smaller fires) and
urban emissions from the Los Angeles basin.
2.3.3 Lag-time analysis: number fraction of thickly-coated rBC-containing particles
Figure 2-4, panel (b) shows both one-minute and one-hour means for f
BC
over the course of all three
campaigns. On average, f
BC
was larger during the first campaign (September 2017) than during the second
and third campaigns (December 2017, November 2018). The mean values (±standard deviation) of f
BC
were
0.27 (±0.19), 0.03 (±0.09), and 0.14 (±0.15) for the first, second, and third campaigns, respectively. This
implies that about one-quarter of the rBC-containing particles that were measured in the first campaign
either had sufficient time in the atmosphere to become aged with thick coatings or originated from biomass
burning emission sources, which have been shown to emit more thickly-coated particles compared to fossil
fuel emissions (Dahlkötter et al. 2014; Laborde et al. 2013; Joshua P. Schwarz, Gao, et al. 2008). Most of the
rBC particles measured in the second campaign were thinly-coated, implying BC
ff
dominated measurements.
26
The rBC from the third campaign exhibited mostly thinly-coated rBC for the first ~four days of the
campaign and an increased f
BC
for the last ~two days of the campaign.
Compared to past studies in the Los Angeles region, the mean f
BC
for the first campaign (September
2017) (f
BC
= 0.27) is close to the lower end of values from aircraft measurements (f
BC
= 0.29) (Metcalf et al.
2012) and the upper end of previous ground-based measurements (f
BC
= 0.21) (Krasowsky et al. 2016). In
contrast, the mean value of f
BC
for the second campaign (December 2017) is almost an order of magnitude
lower than for the first campaign. There are some periods with slightly elevated f
BC
during the second
campaign, but the overall trend suggests that most of the rBC-containing particles in this period are thinly-
coated or essentially uncoated. The Santa Ana wind conditions during the second campaign advected fresh
(a) urban emissions from the Los Angeles basin, and (b) biomass burning emissions from active fires in
Southern California, as discussed in section 2.3.1.
The third campaign (November 2018) is unique in that both “fresh” and “aged” BC
bb
, in addition to
fresh BC
ff
were measured. As shown in Fig. 2-4, there is a distinct period of relatively higher f
BC
and rBC
concentrations starting at ~noon on 16 November 2018 and lasting through the end of the campaign on 18
November 2018. This is the only period from all three measurement campaigns where we observed both
high rBC mass/number loadings and high f
BC
values. In section 2.3.1, we identified the Camp Fire to be the
dominant source during this time period within the third campaign. Thus, the biomass burning rBC particles
measured in this portion of the third campaign are more thickly-coated than our measured urban rBC.
Previous field studies have reported that BC
ff
generally have a lower f
BC
relative to BC
bb
(Joshua P. Schwarz,
Gao, et al. 2008; Sahu et al. 2012; Laborde et al. 2013; McMeeking, Morgan, et al. 2011; Akagi et al. 2012).
For example, Schwarz, Gao, et al. (2008) reported that f
BC
~ 10% for urban emissions and f
BC
~ 70% for
biomass burning emissions. The impact of source type on rBC mixing state will be further discussed in
section 2.3.7.
27
2.3.4 Negative lag-times and rBC morphology
A number of previous studies (Moteki and Kondo 2007; Sedlacek et al. 2012; Moteki, Kondo, and
Adachi 2014; Dahlkötter et al. 2014; Sedlacek et al. 2015) reported negative lag-times from both laboratory
and field measurements of rBC. It has been hypothesized that a negative lag-time is observed when rBC
fragments from its coating material, resulting in a scattering signal that follows an incandescent signal.
Dahlkötter et al. (2014) summarized that negative lag times can occur when either: (i) rBC is very thickly-
coated in a core-shell configuration, (ii) rBC is thickly-coated and the core is offset from the center in an
eccentric arrangement, or (iii) rBC is located on or near the surface of an rBC-free particle. The morphology
of rBC-containing particles is of importance because the enhancement of BC light absorption can vary
widely depending on whether the morphology more closely resembles a core-shell configuration or near-
surface attachment (Moteki, Kondo, and Adachi 2014). Although the fraction of negative lag-times (f
lag,neg
)
cannot definitively identify the morphology of individual rBC-containing particles (Sedlacek et al. 2015) or
accurately quantify the actual percentage of all fragmenting rBC-containing particles (Dahlkötter et al. 2014),
it can offer some general insights about rBC morphology, especially when it is paired with other information
like the emissions source type and rBC coating thickness. f
lag,neg
is a conservative lower-bound estimate for the
fragmentation rate since there may be rBC particles with positive lag-times that still fragment in the SP2
(Dahlkötter et al. 2014). Dahlkötter et al. (2014) used a method examining the tail end of the time-
dependent scattering cross-section in order to determine if a rBC-containing particle was fragmenting,
thereby calculating a higher fragmentation rate relative to f
lag,neg
. Details of the time-dependent scattering
cross-section method can be found in Laborde et al. (2012) and Dahlkötter et al. (2014). This method to
calculate a refined fragmentation rate was not used in Sedlacek et al. (2012), nor in this study.
Furthermore, Sedlacek et al. (2012, 2015) suggest that f
lag,neg
and the lag-time distributions may assist
in source attribution. More specifically, Sedlacek et al. (2012) measured a confirmed biomass burning plume
28
in August 2011 and found high positive correlation between biomass burning tracers and f
lag,neg
during the
period of impact, suggesting that f
lag,neg
may be a useful indicator of biomass burning influence.
In this study, we observed negative lag-times, although at a relatively low rate, with f
lag,neg
calculated to
be much less than 0.1 throughout most of the measurement periods (see Fig. 2-6). We defined f
lag,neg
to be
identical to the "fraction of near surface rBC particles" metric used by Sedlacek et al. (2012), using a lag-time
threshold of -1.25 s to account for uncertainties associated with the lag-time determination. The campaign-
wide f
lag,neg
was 0.017 for the first campaign (September 2017), 0.018 for the second campaign (December
2017), and 0.026 for the third campaign (November 2018). Comparatively, Dahlkötter et al. (2014) observed
f
lag,neg
of ~0.046 during an airborne field campaign measuring an aged biomass burning plume, and
additionally calculated a higher fragmentation rate of ~0.4 to 0.5, based on their aforementioned alternative
method (Laborde, Schnaiter, et al. 2012). Sedlacek et al. (2012) reported f
lag,neg
> 0.6 for ground-based
measurements of a biomass burning plume in Long Island, New York, originating from Lake Winnipeg,
Canada.

Figure 2-6. Panel (a) shows the 10-minute mean time series for number fraction of rBC particles with negative lag-times (f neg, lag).
The threshold for negative lag-times was set to -1.25 s to account for uncertainties in the lag-time determination (Sedlacek et al.,
2012). Panel (b) shows the time series of lag-time values for each individual particle, corresponding to individual dots on the
graph. Panel (c) shows the one-minute mean rBC number concentration for reference.
29
The widely varying f
lag,neg
between these different studies (including our study) suggests that f
lag,neg
may
not be a useful metric when comparing between studies. One of the key findings from Sedlacek et al. (2015)
shows that SP2 operating conditions strongly affects the frequency of negative lag-times, and suggests that
inter-study comparisons of f
lag,neg
could be meaningless, or at worst misleading, if the laser power and sample
flow rate are not reported. See the section S3 in Appendix A for more details.
The higher mean value of f
lag,neg
(0.026) during the third campaign (November 2018), relative to the
first (0.017) and second (0.018) campaigns, shows that f
lag,neg
could potentially be a useful as a supplemental
metric when identifying impacts from biomass burning sources, as mentioned by Sedlacek et al. (2012,
2015). Figure 2-7 also shows that the 10-minute mean negative lag-times increase in magnitude with
increasing rBC core diameter between the range of ~100 to 115 nm (i.e., higher rates of fragmentation with
increasing core size). This follows a similar trend observed by Sedlacek et al. (2012, 2015), who attributed
this trend to increased heat dissipation to surrounding gases for smaller rBC cores, which in turn decreases
the particle heating rate and consequently decreases the fragmentation rate. Our observations add to the
limited past observations that show that the fragmentation rate of rBC particles in the SP2 depend on
physical factors like the core size. This further complicates the practical use of f
lag,neg
as a biomass burning
indicator.
30

Figure 2-7. Scatter plot of 10-minute mean negative lag-times versus 10-minute mean rBC core diameters.
Certain trends in f
lag,neg
for this study further indicate that it should not be used in isolation to verify
the relative abundance of biomass burning aerosol versus non-biomass burning aerosols. There are peaks in
the f
lag,neg
time series (Fig. 2-6) that do not follow the expected trends based on identifiable source impact
time periods. For example, the two peaks on 22 December 2017 (BC
ff
periods) correspond to f
lag,neg
values
exceeding 0.02, but f
lag,neg
hovers around 0.02 on 17 November 2018, when we had expected direct impact
from the Camp Fire. As evidenced from the meteorology (Section 2.3.1), mixing state (Section 2.3.3), rBC
concentrations (Section 2.3.2), and rBC core size (to be discussed in Section 2.3.6), measurements on 17
November 2018 were dominated by biomass burning emissions, but f
lag,neg
fails to show that independently.
These anomalous observations show that f
lag,neg
needs to be used with caution, and that future studies are
necessary to extensively quantify the relationship between f
lag,neg
and source type.
The observations of negative lag-times in this study confirm that ambient rBC likely do not adhere
strictly to core-shell morphology. The exact morphology of measured rBC cannot be quantified based on
our measurements, but the presence of negative lag-times in this study highlights the need to further
understand rBC morphology and its effect on absorption enhancement in future studies, as well as the
potential for f
lag,neg
to be used as a supplemental source identification tool.
31
2.3.5 Leading-edge-only (LEO) fit analysis: rBC coating thickness
To further examine the mixing state of rBC-containing particles, the leading-edge-only (LEO) fit
method was used to quantify rBC coating thickness (CT
BC
) on a particle-by-particle basis. Figure 2-4 shows
the time series of CT
BC
throughout all three campaigns. The time series of CT
BC
shows that each campaign
was characterized by different mixing states, and that there are distinct trends within each campaign.
The inter-campaign differences are highlighted in Fig. 2-8, which shows the CT
BC
distribution for
each campaign, as well as the distribution including rBC from all campaigns. For both rBC core diameter
ranges (180–220 nm and 240–280 nm), we observe that the first campaign has the largest mean CT
BC
,
followed by the third and second campaign, respectively. The mean CT
BC
(  standard deviation) for the first,
second, and third campaign was 52.5 (  45.5) nm, 22.3 (  25.0) nm, and 40.3 (  41.5) nm, respectively, for
particles with a rBC core diameter between 180 and 220 nm.
Comparing the time series of CT
BC
to the time series of f
BC
(Fig. 2-4), we observe similar trends over
time, which is expected and also reported in past studies that have employed both the lag-time and LEO
methods (Metcalf et al. 2012; Laborde, Mertes, et al. 2012; McMeeking, Good, et al. 2011). Figure S9 shows
that there is a statistically significant correlation between 10-minute mean CT
BC
and f
BC
throughout all three
campaigns. The Pearson correlation test (i.e., linear correlation test) was conducted to quantify the linear
correlation between the two variables and to infer the statistical significance of the potential correlation. For
the test, α = 0.05 (i.e., 95% confidence level). The test resulted in p<0.001, suggesting a statistically
significant correlation. The sample Pearson correlation (r) was 0.82. The statistically significant, positive
correlation confirms that these two methods are in general agreement, and that they can be used together to
robustly describe the rBC mixing state.
32

Figure 2-8. Distributions of BC coating thickness (CT BC) aggregated by campaign are shown in red (1
st
campaign), green (2
nd

campaign), and purple (3
rd
campaign). The combined distributions for all campaigns are shown in black. Panels (a) and (b) show
the normalized frequency distributions, while panels (c) and (d) show the absolute frequency distributions. The distributions are
also distinguished by the rBC core diameter ranges included in the LEO analysis. The top panels (a) and (c) show distributions for
particles with rBC core diameters between 180 and 220 nm. The bottom panels (b) and (d) show distributions for particles with
rBC core diameters between 240 and 280 nm.
In addition to aggregating CT
BC
by campaign, we also examined ten discrete time periods of interest
to get a more detailed understanding of the mixing state variability. Two time periods from the September
campaign, three time periods from the December campaign, and five time periods from the November
campaign were selected to represent a diverse range of meteorological conditions, emission sources, and age
of aerosols. Table 2-2 lists the ten LEO-fit periods, and their median and mean CT
BC
. The LEO-fit periods
are also annotated on the rBC concentration time series (see Fig. 2-4) to show when they occurred in the
context of all three campaigns. The median CT
BC
for the LEO periods ranged from –0.4 to 54.0 nm. L6 had
the lowest median CT
BC
(–0.4 nm), while L9 had the highest median CT
BC
(54.0 nm).
Table 2-2. Details of the ten different LEO time periods. Further details about the source-to-receptor characteristic timescales
can be found in Appendix A, section S1.
LEO
Time
Period
Date/Time (Pacific Time)
Period
Length
(mins)
Total number
of rBC particles
analyzed
a
Mean
coating
thickness
(nm)
Median
coating
thickness
(nm)
Characteristic
timescale
33
L1 9 Sep. 2017, 12:00-1:00am 60 397 62.2 53.5 ~days to week
L2 13 Sep. 2017, 11:59-12:58pm 59 467 28.1 23.6
~minutes to
hours
L3 20 Dec. 2017, 12:59-2:00pm 61 79 49.3 47.7 ~days to week
L4 21 Dec. 2017, 12:29-1:00pm 31 318 14.3 12.0 ~3 hours
L5 22 Dec. 2017, 9:59-10:15am 16 1,176 14.6 12.2 ~12 hours
L6 12 Nov. 2018, 12:00-1:00pm 60 1,752 5.6 -0.4 ~8 hours
L7 14 Nov. 2018, 5:00-6:00am 60 2,879 10.7 8.2 ~17 hours
L8 17 Nov. 2018, 5:00-6:00am 60 2,712 57.2 48.4 ~days to week
L9 17 Nov. 2018, 7:00-8:00pm 60 1,254 67.2 54.0 ~days to week
L10 18 Nov. 2018, 10:00-11:00am 60 4,778 40.6 31.2 ~days to week

a
LEO coating thickness calculations shown in the table only include rBC-containing particles with core sizes between 200 and
250 nm.

Figure 2-9 illustrates the CT
BC
distributions and statistics of each LEO period. L1 and L2 were from
the first campaign (September 2017). L1 is representative of ambient background rBC-containing particles
from the first campaign. A period that did not exhibit any anomalously large rBC mass concentration values
was chosen so that contributions from possible nearby sources would not skew the mean CT
BC
. On the
other hand, L2 intentionally spans a period with many anomalously high rBC mass concentration values.
Although these anomalous values were removed from the concentration time series discussed previously in
section 2.2.4, the values were not removed for the LEO analysis of L2 to examine the relationship between
CT
BC
and possible nearby emissions. As hypothesized, the rBC-containing particles from L2 were generally
more thinly-coated than those from L1. The median CT
BC
from L2 was ~30 nm lower than that for L1,
which corroborates our hypothesis that the anomalously high mass concentration values in the first
campaign included contributions from nearby, unidentified fossil fuel sources.
34

Figure 2-9. Violin plots that show the distribution of rBC coating thickness values calculated for each LEO time period, L1
through L10. Each circle marker in the plot represents a particle analyzed by the LEO analysis and the curves for each “violin”
shape represents the normalized probability density function of the coating thickness for each LEO period. The violin shape
results from mirroring each probability density distribution along a vertical axis. Box-and-whiskers plots are also overlaid to show
the quartiles (25
th
, 50
th
, and 75
th
percentiles) of the coating thickness distributions. The 95% confidence intervals (CI) based on
Student’s t-distribution are shown above each violin plot to demonstrate when the mean coating thickness values are statistically
distinguishable from one another. The mean (unfilled diamond) and median (solid diamond) coating thicknesses are also indicated
above each violin plot, and a brief description of sources for each LEO period is annotated below each distribution.
L3 through L5 are time periods from the second campaign (December 2017). L3 represents a period
near the start of the second campaign (December 2017). The predominant wind direction during L3 was
westerly, with a mean wind speed of ~4.5 m s
-1
. HYSPLIT back-trajectories and CAMS data show that L3
likely included important contributions from the Thomas Fire in Santa Barbara and Ventura County. The
PM
2.5
concentration gradient from CAMS was examined over time to track the movement of plumes that
influenced the measurements during this time period. A few days prior to the start of the second campaign,
the Thomas Fire resulted in a large aerosol plume westward over the Pacific Ocean. From visually tracking
PM
2.5
concentration gradients, it appears that a large-scale, clockwise, atmospheric circulation brought
aerosols from the Thomas Fire to Catalina Island around the time of L3 (see video 2 of Video Supplement).
35
The average concentration during L3 was about an order of magnitude lower than the average concentration
for the September campaign. This could be partially attributed to the fact that L3 was around 1 to 2 pm,
when the planetary boundary layer would be expected to increase in height, causing pollutant concentrations
to decrease due to dilution. The median CT
BC
for L3 was 47.7 nm, which is slightly lower than for L1, which
is representative of the ambient background conditions. The slightly larger CT
BC
for L3 likely reflects the
fact that mixing state is sensitive to the source of emissions. In this time period, urban emissions were likely
mixed into the regional air mass, slightly lowering the median CT
BC
. A number of previous studies have
suggested that rBC from biomass burning emissions are generally more thickly-coated (Sahu et al. 2012;
Joshua P. Schwarz, Gao, et al. 2008; Dahlkötter et al. 2014). In this case, we have evidence to support that a
larger fraction of measured rBC during L3 came from the local Thomas Fire, while L1 represents a mix of
influences, including, but not limited to, aged biomass burning aerosols. The effect of emissions sources on
rBC mixing state is discussed in section 2.3.7.
L4 through L7 represent periods when the Los Angeles basin, Santa Barbara/Ventura counties, and
San Diego county (to a lesser degree) were identified as major sources. Air masses measured during these
periods likely contained a mixture of both urban emissions and biomass burning emissions (see Appendix A
section S2 and accompanying figures), although urban emissions were likely dominant. Overall, these LEO
periods exhibit the lowest median CT
BC
, ranging from -0.4 to 12.2 nm. The potential relationship between
aging time and CT
BC
is discussed further in section 2.3.7.
L8, L9, and L10 are the unique LEO periods from the third campaign (November 2018) with
concurrently increased rBC concentrations and f
BC
(discussed in the previous section). We also observed
significantly higher CT
BC
values during these periods compared to L4–L7, with median CT
BC
ranging from
31.2 to 54.0 nm. We have strong evidence to support that the sampled particles include important
contributions from aged rBC from the Northern California fires, particularly the Camp Fire (see section S2
in Appendix A). The relatively high CT
BC
values in L8 and L9 (compared to other LEO periods) further
36
support our claim that rBC-containing particles from Northern California fires were dominating our
measurements during this time. L10 has a median CT
BC
of 31.2 nm, which is ~23 nm lower than the median
value for L9. This reduction in the median CT
BC
is also reflected in the decrease of the f
BC
values near the
end of the campaign. Meteorological data, MODIS satellite images, and CAMS data during this time period
suggest that sources from the Southern California (and possibly Central Valley) region contributed more to
measurements during L10 than they did during L8 and L9, explaining the lower CT
BC
and higher overall
concentrations. Wind speeds were lower on average for L10 compared to L8 and L9. The mean wind speed
for L10 at LAX, based on 5-minute NOAA data, was ~1.3 m s
-1
, while the mean wind speeds for L8 and L9
were ~2.1 m s
-1
and 1.6 m s
-1
, respectively. There was also a general shift of wind direction from westerly to
north-easterly, approximately a half day before L10 (see Fig. S8 in Appendix A). MODIS satellite imagery
and CAMS data also confirm that local to regional sources were likely impacting the measurements more
during this period (see video 3 and 4 of Video Supplement), compared to L8 and L9. The meteorology, in
addition to local to regional sources of emissions from the Los Angeles basin and Southern California, likely
explain the reduction in CT
BC
and the near doubling of the rBC concentration level.
2.3.6 rBC core size
The number- and mass-based size distributions for rBC cores were assessed for periods L1 to L10.
Similar to past studies, rBC core mass equivalent diameters between 70 and 450 nm are reported (Gao et al.
2007; Moteki and Kondo 2007; Dahlkötter et al. 2014; Krasowsky et al. 2018). Figure 2-10 shows both log-
normal fits of the rBC core size distributions and measured rBC core diameters for three LEO periods (L1,
L5, and L10); we investigated these three LEO periods to assess whether log-normal fits adequately
represent the actual rBC size distributions before presenting log-normal fits for all LEO periods. Previous
studies have shown that rBC core size distributions are generally log-normal in the accumulation mode
(Metcalf et al. 2012). Figure 2-10 shows that log-normal fits adequately capture the measured size
distributions, though we cannot rule out the possibility of another rBC mode outside the detection limits of
37
the SP2. Although the peak of the observed size distribution is not always discernible (e.g., number size
distribution for L5 in Fig. 2-10), it is reasonable to fit these points assuming that a log-normal distribution is
a realistic representation of ambient rBC number size distributions in the Aitken mode.

Figure 2-10. Measured rBC core size distributions and corresponding log-normal fits to the measurements for LEO periods L1,
L5, and L10.
A survey of past studies that have reported log-normal fit rBC mass median diameter (MMD
fit
) and
count median diameter (CMD
fit
) shows that the source of emissions has a strong influence on rBC core
diameter (Cheng et al. 2018). The MMD
fit
[CMD
fit
] for BC
bb
, which has been reported to range from ~130
nm to 210 nm [100 to 140 nm], is generally larger than the MMD
fit
[CMD
fit
] for BC
ff
, which has been
reported to range from ~100 nm to 178 nm [38 to 80 nm] (Joshua P. Schwarz, Gao, et al. 2008; Metcalf et
al. 2012; McMeeking et al. 2010; Kondo et al. 2011; Sahu et al. 2012; Shiraiwa et al. 2007; Cappa et al. 2012;
Laborde et al. 2013; D. Liu et al. 2014; Taylor et al. 2014; Krasowsky et al. 2018). The MMD
fit
[CMD
fit
] for
well-aged background BC were reported to range from ~180 nm to 225 nm [90 nm to 120 nm] (Shiraiwa et
al. 2008; D. Liu et al. 2010; McMeeking et al. 2010; J. P. Schwarz et al. 2010).
38
Figure 2-11 shows the rBC MMD
fit
and CMD
fit
for each LEO period in this study. Based on the
source identification discussed in section 2.3.1 and section S2 in Appendix A, the MMD
fit
and CMD
fit
values
in this study are generally consistent with the ranges reported in past studies. For BC
bb
(L3, L8, L9, L10),
MMD ranged from 149 nm to 171 nm, which is within the range of ~130 nm to 210 nm reported in past
studies. For BC
ff
(L2, L4, L7), the MMD
fit
dropped, ranging from 112 nm to 129 nm. This falls within the
range of ~100 nm to 178 nm previously reported for measurements of urban emissions.

Figure 2-11. Median rBC core diameter for both mass and number size distribution lognormal fits.
There were some periods in which we observed a relative increase in the MMD
fit
, but concurrent
decrease in the CMD
fit
. For example, L5 exhibits a relatively high MMD
fit
(~171 nm), which suggests that
this was a biomass burning dominated time period, but the CMD
fit
is the second lowest of all the LEO
periods (~53 nm). L10 exhibits a similar pattern. In such seemingly contradictory situations, it is likely that
biomass burning aerosols are entrained into a broader urban plume (e.g., from Los Angeles basin). An urban
plume with no biomass burning influence is expected to exhibit a very low CMD
fit
. As biomass burning
aerosol gets entrained, MMD
fit
is expected to change more than CMD
fit
due to its larger rBC core size
39
relative to urban rBC cores. (For a unit increase in diameter, the mass weighting will increase proportionally
to the third power, while the size weighting will increase proportionally on a first order basis.) This may
explain why in some cases we observe relatively high MMD
fit
values along with relatively low CMD
fit
values,
highlighting the need to examine both the number and mass size distributions for rBC core size analysis.
Another explanation for varying rBC core size is coagulation (Bond et al. 2013). Shiraiwa et al.
(2008) observed an increase in rBC core diameters in aged plumes compared to fresher urban plumes,
suggesting that coagulation can alter the rBC size distribution during atmospheric transport (i.e., aging).
Although the emissions source type appears to be the dominant influence on rBC core sizes in our study,
there is evidence to suggest that coagulation also played a role during transport from the Los Angeles basin
to Catalina Island (~70 km away). For example, we observed a MMD
fit
[CMD
fit
] of 112 nm [53 nm] during
L4, when BC
ff
was measured. This is noticeably larger than values of 93 nm [42 nm] reported in Krasowsky
et al. (2018) for measurements conducted 114 meters downwind of a major highway in Los Angeles.
Furthermore, Laborde et al. (2013) observed an MMD
fit
of ~100 nm for BC
ff
in Paris, which is again lower
than the value of 112 nm calculated for L4. Even though it was determined that L4 was characterized by
BC
ff
, we cannot rule out the effects of local wildfires influencing the size distribution as well (as explained in
Appendix A section S2). While the rBC size distribution from L4 suggests that coagulation plays at least a
minor role, both factors (source type and coagulation) likely influence rBC size distributions to varying
degrees in areas with varying emissions source types and relatively elevated rBC concentrations (e.g.,
polluted urban areas).
2.3.7 Impact of emissions source and aging on rBC mixing state
The dominant factors that influence rBC core size (i.e., emission source type and aging) also
influence rBC mixing state. Figure 2-12 shows a scatter plot of one-minute mean CT
BC
versus one-minute
mean rBC core diameter. A statistically significant, positive correlation (p < 0.001) was found, with r = 0.55.
Further details regarding the statistical tests used to calculate the correlation coefficients and to conduct the
40
hypothesis testing can be found in section 2.3.5. The significant correlation confirms that larger
contributions from biomass burning (as opposed to fossil fuel) and longer aging timescales are associated
with increases in both the rBC core size and the BC coating thickness. Figure 2-13 shows the CT
BC

distributions for different rBC core size ranges, and a similar relationship between the two variables can be
observed. As the core size increases (lighter to darker curves), a broader right-hand side tail is observed in
the CT
BC
normalized distributions for each campaign, implying higher mean CT
BC
for particles with larger
rBC cores.

Figure 2-12. rBC coating thickness versus rBC core diameter. Each point on the plot represents a 1-minute mean. Data from all
three campaigns are shown. CT
BC
values are calculated for particles with rBC core diameters between 200–250 nm. The line
represents the least-squares linear regression to the one-minute mean data points. There is a statistically significant positive
correlation shown between CT
BC and rBC core diameter, as shown in the summary box in the top left corner.
41

Figure 2-13. Distributions of BC coating thickness (CT
BC
) aggregated by campaign and varying rBC core diameter ranges used in
the LEO analysis. Panels (a) through (d) in the left column show the normalized frequency distributions, while panels (e) through
(h) in the right column show the absolute frequency distributions. Within each panel, each line represents a distribution for a
particular rBC core diameter range, with darker lines representing larger diameter ranges and vice versa.
The time evolution of both CT
BC
and rBC core size is represented in a series of scatter plots in Fig.
2-14 and 2-15. In each of the figures, the scatter between one-minute mean CT
BC
and rBC CMD are
grouped into six-hour time intervals for both the second (December 2017) and third (November 2018)
periods, respectively. In these figures, the time evolution of the rBC physical properties can be examined in
detail and compared to periods of known emissions source impacts. There are a few significant patterns
42
worth mentioning here. First, the influence of BC
bb
can be observed between 16 and 18 November 2018 in
Fig. 2-15. Both CT
BC
and CMD drastically increase for a prolonged period of time, implying an impact from
the Camp Fire plume from Northern California. Second, the scatter plots for 20 to 22 December 2017 and
12 to 15 November 2018 show that there is some variability over time in the cluster shapes, which can be
explained by the local wildfires that were confirmed to influence the broader LA basin plume (see section S2
for details regarding source attribution). Although the scatter plots during these time periods support that
BC
ff
was largely dominant, there are some periods where the CMD spread deviates quite noticeably (e.g.,
Fig. 2-14, 06:00 21 Dec. 2017), or even periods that show two distinct clusters (e.g., Fig. 2-15, 12:00 12 Nov.
2018), supporting our claim that local wildfires were indeed influencing our measurements. A similar figure
for the first campaign (September 2017) is included in Appendix A as Fig. S23, but not shown here because
of the relatively stable mixing state and size of BC
aged,bg
. These figures confirm the general patterns noted in
previous sections regarding the effects of different sources on rBC mixing state.

43

Figure 2-14. Matrix of scatter plots showing the time evolution of CT
BC
(nm) and rBC count mean diameter (nm) for the second
campaign (December 2017). Axes labels are shown in the upper left. A scatter plot is shown for each six-hour time interval of the
day, starting at 00:00 Pacific Time, and for each day of the campaign. The columns of the matrix denote the time interval of the
day, and the rows of the matrix denote the days of the campaign. Each point within a plot represents a one-minute mean value for
both CT
BC
and count mean diameter.
44

Figure 2-15. Matrix of scatter plots showing the time evolution of CT BC (nm) and rBC count mean diameter (nm) for the third
campaign (November 2018). Axes labels are shown in the upper left. A scatter plot is shown for each six-hour time interval of the
day, starting on 00:00 Pacific Time, and for each day of the campaign. The columns denote the time interval of the day, and the
45
rows denote the day of the campaign. Each point within a plot represents a one-minute mean value within that six-hour interval
for both CT BC and count mean diameter.
When the scatter plots of one-minute mean CT
BC
and rBC mean diameter are aggregated by
campaign, distinct patterns emerge. Contour plots representing the 2-d joint histograms of these two
variables are shown in Fig. 2-16. Each campaign exhibits a distinct pattern that is representative of the
emissions sources and relative age of the measured air masses. Figures 2-16b and 2-16e show a single cluster
for the second campaign (September 2017) characterized by relatively thin coatings and smaller rBC core
diameters, compared to the other campaigns. Figures 2-16c and 2-16f on the other hand show two distinct
clusters for the third campaign (November 2018). One cluster represents thickly-coated particles with larger
rBC core diameters, and the other represents more thinly-coated particles with smaller rBC core diameters.
The thinly-coated/smaller rBC core cluster for the third campaign exhibits some similarities to the single
cluster for the second campaign. Figures 2-16a and 2-16d show two overlapping clusters for the first
campaign (September 2017), which fall loosely in between the thickly-coated and thinly-coated clusters from
the third campaign. For easy reference, a cluster characterized by thin coatings and smaller rBC cores will be
referred to as a “BC
ff
cluster,” a cluster with thick coatings and larger rBC cores will be referred to as a
“BC
bb
cluster,” and the bimodal, mixed cluster will be referred to as the “BC
aged,bg
cluster.”
46

Figure 2-16. Contour plots of count as a function of one-minute mean BC coating thickness (CT
BC
) and one-minute mean rBC
core diameter. This figure can be interpreted as a 2-d joint histogram, converted to a contour plot. Each count represents a single
one-minute mean data point. The contours are created based on the 2-d joint histogram that is calculated using a 50x50 grid
within the range of all one-minute mean data. Panels (a), (b), and (c) in the first row show mass mean diameter on the horizontal
axes, while panels (d), (e), and (f) in the second row show count mean diameter.
Within the context of the identifiable sources discussed in previous sections (section 2.3.1 and S2 in
Appendix A), it is evident that these distinct clusters in Fig. 2-16 are strongly influenced by emissions source
type. A BC
bb
cluster is present in the third campaign (November 2018) when impacts from long-range
transported biomass burning emissions were identified, but not in the second campaign (December 2018).
Furthermore, a BC
ff
cluster is present in both the second (December 2017) and third campaign, but not in
the first campaign (September 2017). This shows that fresh (age < 1 d), urban emissions from the LA basin
and the surrounding southern California region are characterized by thin coatings and smaller core size,
confirming what has also been observed in other past field studies (Laborde, Mertes, et al. 2012; D. Liu et al.
2014; Krasowsky et al. 2018).
The BC
aged,bg
cluster (Fig. 2-16a, 2-16d) exhibits two distinct modes within the same cluster. One
mode is characterized by a peak CT
BC
[CMD] that is ~20 nm [~10 nm] larger than the other mode. Within
the context of BC
aged,bg
, this mode is referred to as the larger mode, while the other mode with smaller CT
BC

47
and CMD is referred to as the smaller mode. This suggests that BC
aged,bg
advecting over the Pacific Ocean
during typical meteorological conditions contain rBC from both biomass burning and fossil fuel (i.e., urban)
emissions sources. Aged background air masses are likely to contain aerosol from a mix of sources.
In addition to emissions source type, atmospheric aging also appears to have an observable effect on
the mixing state. Table 2-3 lists the range of estimated “source-to-receptor” timescales for rBC-containing
particles measured during LEO time periods L1 to L10. In short, the first campaign (September 2017) is
broadly characterized by source-to-receptor timescales on the order of days to a week. The second campaign
(December 2017) is characterized by timescales of less than one day. And the third campaign (November
2018) is characterized by timescales of less than one day for the first four days of the campaign, and
timescales of approximately days to a week for the last two days of the campaign.
With regards to aging, we first observe that BC
ff
particles do not develop thick coatings within the
timescales observed in this study. This suggests that a timescale of less than one day is not sufficient to
thickly coat urban rBC-containing particles in the lower boundary layer, in the Los Angeles region. Although
a modestly higher CT
BC
is observed during urban-dominated time periods, relative to CT
BC
~ 0 nm observed
by Krasowsky et al. (2018) inside the LA basin, this is likely due to the effects of local biomass burning
emissions mixing into the broader urban plume in both December 2017 and November 2018, as discussed
above (also see section S2 in Appendix A). While we observed mostly thinly-coated BC from these urban-
dominated time periods, we acknowledge that the timescale required to acquire coatings on BC will likely
differ by location because of variations in local meteorology, pollution concentrations, and emission source
profiles.
On the other hand, BC
bb
were generally more thickly-coated, although the time evolution of the
mixing state could not be quantified directly in this study. Fresh BC
bb
had slightly lower CT
BC
compared to
that of aged BC
bb
(e.g., L3 vs. L9), but higher CT
BC
compared to that of fresh BC
ff
(e.g., L3 vs. L4). The
overall higher CT
BC
for aged BC
bb
relative to fresh BC
bb
indicates that significant coating formation can
48
occur within timescales of ~1 day to ~1 week for BC
bb
, even after rapid coating formation that occurs soon
after emission. An important caveat is that CT
BC
of BC
bb
may not be monotonically increasing over time.
Past studies have observed rapid coating of BC
bb
within one day to more than 100 nm (Perring et al. 2017;
Morgan et al. 2020), but we observed a median CT
BC
of 47.7 nm for L3, which suggests that CT
BC
for BC
bb

might decrease during atmospheric transport under certain conditions and could again increase later at
longer timescales (e.g., median CT
BC
of 54.0 nm for L9), although we would need simultaneous
measurements near the point of biomass burning emissions in order to confirm this for a specific plume.
Previous studies have noted that the competing processes of dilution-driven evaporation and oxidation-
driven condensation determine the abundance of organic aerosol relative to carbon monoxide (ΔOA/ΔCO)
in biomass burning plumes (Garofalo et al. 2019). Although the time evolution of ΔOA/ΔCO is not
necessarily indicative of rBC mixing state evolution, the same physical processes (i.e., evaporation and
condensation) must apply to OA rBC coating formation and loss. The conflicting observations from various
studies showing ΔOA/ΔCO either increasing, decreasing, or staying relatively stable in the near-field
(timescale of ~hours) suggest that rBC coating in dense fresh biomass burning plumes may undergo similar
competing physical mechanisms. Preliminary results from developing research showed that well-aged BC
bb

(>7 days) had thinner coatings than fresh BC
bb
(< 5 h), providing emerging evidence that rBC coating may
not always monotonically increase (Sedlacek et al. 2019). Further research is necessary to confirm this
process in more field measurements, and to determine the various mechanisms that may be driving the loss
of rBC coating in biomass burning plumes. We make no definitive claims about the rate of change of CT
BC

for BC
bb
throughout atmospheric transport since we measured CT
BC
at one location. Nonetheless, our
measurements suggest that CT
BC
for fresh Southern California BC
bb
were generally lower than CT
BC
for aged
Northern California BC
bb
.
The contour plots for the first campaign (September 2017), shown in Fig. 2-16a and 2-16d, offer
additional perspective on how aging can affect rBC mixing state within well-aged background air masses
49
over longer aging timescales (~days to week). The first notable feature of the BC
aged,bg
cluster is that the
smaller mode is significantly more coated than the BC
ff
clusters found in Fig. 2-16 for the second
(December 2017) and third (November 2018) campaigns. The peak of the smaller mode of the BC
aged,bg

cluster is at least 35 nm higher than the peak of the respective BC
ff
clusters in Fig. 2-16b, 2-16c, 2-16e, and
2-16f. Assuming that this smaller mode represents fossil fuel influenced BC (i.e., urban BC), this confirms
that while urban BC may not become thickly-coated within a day, they seem to acquire coatings over longer
timescales.
Another prominent feature of the BC
aged,bg
cluster is the shift in mass mean diameter (MMD) (Fig. 2-
16a) and count mean diameter (CMD) (Fig. 2-16d) peaks, relative to the BC
bb
and BC
ff
clusters in the
December 2017 (Fig. 2-16b, 2-16e) and November 2018 (Fig. 2-16c, 2-16f) plots. Comparing peaks of the
smaller mode of the BC
aged,bg
cluster in both Fig. 2-16a and 2-16d to the respective peaks of the BC
ff
clusters
in Fig. 2-16b, 2-16c, 2-16e, and 2-16f; we observe that the MMD is generally lower for the smaller mode of
BC
aged,bg
while the CMD is generally higher. The lower MMD in BC
aged,bg
compared to BC
ff
can be explained
by the impact of local biomass burning sources in both the second (September 2017) and third (November
2018) campaigns, as previously mentioned in section 2.3.6. On the other hand, the overall higher peak CMD
for the lower mode of the BC
aged,bg
cluster implies that either (i) urban BC is coagulating over long aging
timescales (days to week), (ii) source-specific variables like fuel type and combustion conditions are
influencing the initial core size distribution, or (iii) the urban area in which the rBC is emitted contains a
much higher concentration of rBC, leading to more coagulation in the near-field before continental-scale
transport. Any combination of these three explanations could contribute to the overall increase in the peak
rBC diameter of the urban mode. We do not attempt to quantify the extent to which each factor contributes
to core size increases in this study, though (i) is unlikely to be important given the dependence of
coagulation rate on number concentrations. Further research needs to be conducted to accurately
characterize the relative importance of each factor.
50
Shifting focus to the larger mode in the BC
aged,bg
cluster in Fig. 2-16a and 2-16d, which we attribute
to biomass burning BC, we notice lower CMD and MMD, relative to the BC
bb
clusters in Fig. 2-16c and 2-
16f. Based on the assumption that the initial rBC size distribution of the biomass burning rBC from the first
campaign and third campaign are similar, selective wet deposition and/or increased hygroscopicity of
thickly-coated rBC could explain this apparent decrease in the overall rBC core size. Moteki et al. (2012)
found that larger rBC particles were more effectively removed through wet deposition. McMeeking, Good,
et al. (2011) reported that more thickly-coated particles were more hygroscopic, which would in turn lead to
a higher likelihood of wet scavenging. Either, or both, of these mechanisms could explain why the peak rBC
diameter of the larger mode in the BC
aged,bg
cluster is lower than the peak rBC diameter of the BC
bb
cluster.
The evolution of rBC mixing state and rBC size distribution has important implications on
accurately assessing the regional climate benefits of black carbon reductions, particularly in California, and
also reducing uncertainty in global radiative forcing of BC. Understanding the impact of varying emissions
source types and atmospheric aging in different regional contexts is crucial for accurately quantifying the
enhancement of BC light absorption, and also for determining BC lifetime in the atmosphere since
hygroscopic coating material can enhance the particle’s susceptibility to wet deposition (J. Zhang et al.
2015). The rBC mixing state results from this study add to a growing body of evidence that suggests that
biomass burning emissions and longer aging timescales generally lead to more thickly-coated rBC particles.
These results also emphasize the need for more field measurements of rBC mixing state in various regions
around the world to further understand how different emissions source profiles and atmospheric aging
ultimately effect rBC physical properties in various, real-world atmospheric contexts.
2.3.8 Comparison to past studies quantifying CTBC using the SP2
Overall, the range of CT
BC
calculated in this study agree with reported values from past studies.
Table 2-3 presents a comprehensive list of CT
BC
values from various studies, categorized by dominant
emissions source type and sorted alphabetically by first author name.
51
For BC
bb
, the mean CT
BC
ranged between ~40–70 nm in this study. This range overlaps with the
range of values reported by Morgan et al. (2020), Pan et al. (2017), Sahu et al. (2012), Joshua P. Schwarz,
Gao, et al. (2008), and Sedlacek et al. (2012). For BC
ff
, the mean CT
BC
ranged between ~5–15 nm in this
study. This range overlaps with the range of values reported by Krasowsky et al. (2018), Laborde et al.
(2012), Liu et al. (2014), Sahu et al. (2012), and Joshua P. Schwarz, Gao, et al. (2008). For BC
aged,bg
, the mean
CT
BC
was ~60 nm in this study. This value falls within the range of values reported by Laborde et al. (2013),
Joshua P. Schwarz, Gao, et al. (2008), and Shiraiwa et al. (2008).
An important caveat to note when making inter-study comparisons is that the studies that reported
higher CT
BC
ranges (relative to this study) tended to have a lower value for the lower rBC core diameter
threshold. For example, Gong et al. (2016) reports a CT
BC
range of 110–300 nm for biomass burning
emissions using an rBC core diameter range of 80–180 nm. Since the scattering detection limit is accurate
down to ~170 nm for the SP2, this implies that the inclusion of particles with rBC core sizes smaller than
170 nm will bias the average CT
BC
values higher because smaller rBC particles with optical diameters below
the scattering detection will not be included in the LEO analysis. Dahlkötter et al. (2014), Gong et al. (2013),
Perring et al. (2017), Taylor et al. (2014), Cheng et al. (2018), Metcalf et al. (2012), Raatikainen et al. (2015),
and Sharma et al. (2017) all reported CT
BC
for rBC-containing particles in a size range that includes rBC
cores smaller than 170 nm. There is value in reporting CT
BC
for rBC particles with core sizes smaller than
170 nm because it will show the relative abundance of coated rBC-containing particles exceeding the lower
scattering detection limit, but care must be taken when comparing CT
BC
values calculated with varying rBC
core size restrictions.
For future studies using the SP2, we suggest that at a minimum, the rBC core size range be explicitly
stated if CT
BC
is being quantified and reported. Furthermore, it would be useful to establish some
standardized guidelines for reporting CT
BC
so that future inter-study comparisons can serve as reliable
benchmarks. As shown in Figure 2-13 and discussed earlier, the range of rBC core diameters used for the
52
calculation of CT
BC
has a significant effect on the CT
BC
statistics. These ranges must be considered to
accurately represent the physical parameterization of BC mixing state and size distributions in models.
Table 2-3. Summary table of rBC coating thickness values reported in previous studies using the SP2.
Dominant
source
Coating
thickness
(nm)
rBC core
diameter
(nm)
rBC age Description Time period Reference
Biomass
burning
emissions
~40–70
a
200–250
~days–
wk
Ground-based measurements on Catalina
Island (~70 km SW of Downtown LA)
17–18 Nov 2018 This study
105–136 140–220 ~3–4 d
Airborne measurements of the Pagami Creek
Fire plume (Minnesota, US) conducted over
Germany
16 Sep 2011 Dahlkötter, 2014
110–300 80–130 –
Ground-based measurements in Shanghai,
China
5–10 Dec 2013 Gong, 2016
11–15 200–260 – Ground-based measurements in Paris, France 15 Jan–15 Feb 2010 Laborde, 2013
100–300 – –
Ground-based measurements in London,
during periods significantly influenced by solid
fuel burning
22–24 Jan 2012 Liu, 2014
< 30 – –
Ground-based measurements near central
Manchester, UK
3–16 Aug 2010 McMeeking, 2011
40–120 – < 3 h
Airborne measurements across the Amazon
and Cerrado
Sep and Oct 2012 Morgan, 2020
11–54 190–210 < 10 s
Burning experiments in laboratory combustion
chamber
– Pan, 2017
90–110 160–185 < 2 d
Airborne measurements of the Yosemite Rim
Fire, CA
Aug 2013 Perring, 2017
20–80 200 –
Airborne measurements over California during
ARCTAS-CARB campaign
15–30 Jun 2008 Sahu, 2012
65 12 190–210 0.5–1.5 h
Airborne measurements over Houston and
Dallas, TX
20–26 Sep 2006
Joshua P. Schwarz,
Gao, et al. 2008
40–70 – ~days
Ground-based measurements of a wildfire
plume from the Lake Winnipeg area in Canada,
conducted in Long Island, NY
2 Aug 2011 Sedlacek, 2012
79–110 130–230 1–2 d
Airborne measurements over wildfires in
eastern Canada and North Atlantic
Jul–Aug 2011 Taylor, 2014
Fossil fuel
emissions
~5–15
a
200–250 < 1 d
Ground-based measurements on Catalina
Island (~70 km SW of Downtown LA)
17–18 Nov 2018 This study
22–40 130–160
< 3 h
Airborne measurements over the Athabasca oil
sands in Canada
13 Aug–7 Sep 2013 Cheng, 2018
17–39 160–190
50–130 60-80 –
Ground-based measurements in Shanghai,
China
5–10 Dec 2013 Gong, 2016
~0–24 240–280 < 7 h
Ground-based measurements in the Los
Angeles basin
Aug–Oct 2016 Krasowsky, 2018
2 10 200–260 – Ground-based measurements in Paris, France 15 Jan–15 Feb 2010 Laborde, 2013
0–50 – –
Ground-based measurements in London,
during periods dominated by traffic sources
31 Jan–1 Feb 2012 Liu, 2014
99 20 90-260 –
Airborne measurements in the Los Angeles
Basin and surrounding outflows
May 2010 Metcalf, 2012
88 4
b
180 ~hours
Ground-based measurements in Gual Pahari,
India
3 Apr–14 May 2014 Raatikainen, 2015
0–40 200 –
Airborne measurements over California during
ARCTAS-CARB campaign
15–30 Jun 2008 Sahu, 2012
20 10 190–210 2–3.5 d
Airborne measurements over Houston and
Dallas, TX
20–26 Sep 2006
Joshua P. Schwarz,
Gao, et al. 2008
30–40 200 ~ 6 h
Ground-based measurements of fresh
emissions from Japan, conducted on Fukue
Island, Japan
Mar–Apr 207 Shiraiwa, 2008
Remote /
background
/
continental
/ highly-
aged
~60
a
200–250
~days–
wk
Ground-based measurements on Catalina
Island (~70 km SW of Downtown LA)
7–14 Sep 2017 This study
130–300 60–80 –
Ground-based measurements in Shanghai,
China
5–10 Dec 2013 Gong, 2016
37–93 200–260 – Ground-based measurements in Paris, France 15 Jan–15 Feb 2010 Laborde, 2013
188 31 90-260 –
Airborne measurements in the free
troposphere
May 2010 Metcalf, 2012
53
75–100 150–200 –
Ground-based measurements at the Pallas
GAW (Finnish Arctic)
Dec 2011–Jan 2012 Raatikainen, 2015
90 5
b
180

~hours
Ground-based measurements in Mukteshwar,
India
9 Feb–31 Mar 2014 Raatikainen, 2015
48 14 190–210 –
Airborne measurements over Houston and
Dallas, TX
20–26 Sep 2006
Joshua P. Schwarz,
Gao, et al. 2008
< 30 nm 190–210 –
Airborne measurements over Costa Rica, 1-5
km
6–9 Feb 2006
Joshua P. Schwarz,
Spackman, et al.
2008
20–36 160–180 –
Ground-based measurements in Alert,
Nunavut, Canada (within Arctic Circle)
Mar 2011–Dec 2013 Sharma, 2017
~60 200 ~days
Ground-based measurements of Asian
continental air masses, conducted on Fukue
Island, Japan
Mar–Apr 207 Shiraiwa, 2008
a
The range of values shown represent the approximate range of the mean CT BC.
b
The absolute coating thickness was calculated from the ratio of rBC core diameter to particle mobility diameter as presented in
the study.
Note: A dash (“-“) indicates that the value was not reported, or it could not be identified.
2.4 Conclusion
This study investigates the concentration, size distribution, and mixing state of rBC on Catalina
Island (~70 km southwest of Los Angeles) using a single-particle soot photometer (SP2). Measurements
were taken during three separate campaigns with varying meteorological conditions and emission sources, in
September 2017, December 2017, and November 2018. During the first campaign (7 to 14 September
2017), westerly winds dominated and thus the sampling location was upwind of the dominant regional
sources of BC (i.e., urban emissions from the Los Angeles basin). The measurements from the first
campaign were largely characteristic of ambient background levels of rBC over the Pacific Ocean, away
from the broader urban Los Angeles plume. During the second and third campaigns (20 to 22 December
2017, 12 to 18 November 2018), atypical Santa Ana wind conditions caused measured rBC to include
important contributions from large wildfires in California and urban emission from the Los Angeles basin.
Furthermore, during the third campaign, rBC from the Camp Fire in Northern California was measured,
allowing us to compare the mixing state of aged biomass burning particles (from Camp Fire) to fresher
particles (from Southern California fires and urban Los Angeles emissions). The measurements from these
three campaigns showed that rBC physical properties (rBC core size and mixing state) were influenced by (i)
emissions source type, and (ii) atmospheric aging.
54
BC
bb
generally had larger core diameters than BC
ff
. The MMD [CMD] of BC
bb
was observed to be
~180 nm [120 nm], while MMD [CMD] of BC
ff
was observed to be ~160 nm [100 nm]. BC
aged,bg
showed a
bimodal rBC core size distribution, with MMD [CMD] peaks at ~170 nm [115 nm] for the larger mode, and
~153 nm [109 nm] for the smaller mode. The bimodal rBC core size distribution in the aged background air
mass suggests that background rBC above the Pacific Ocean during typical meteorological conditions are
likely a mix of both urban (i.e., smaller rBC cores) and biomass burning (i.e., larger rBC cores) emissions.
The larger CMD of the smaller mode for BC
aged,bg
compared to the CMD of BC
ff
suggests that either (i)
coagulation increased the size of BC
aged,bg
somewhere between the source and receptor, and/or (ii) the initial
source and combustion conditions for BC
aged,bg
were different than for BC
ff
in this study. More accurate
methods of source apportionment would be needed to quantify the relative contribution of each factor, but
both factors likely effect aged urban rBC particles. The CMD [MMD] of the larger mode for BC
aged,bg
was
smaller than that of BC
bb
, which suggests that (i) selective wet deposition of larger particles, and/or (ii)
increased wet scavenging of thickly-coated particles due to increased hygroscopicity, contributes to the shift
in the rBC size distribution for biomass burning rBC-containing particles over long-range atmospheric
transport.
Similar trends are observed for the impact of emissions source type on rBC mixing state. On
average, BC
ff
was either uncoated or very thinly-coated, with mean coating thickness (CT
BC
) ranging from
~5 to 15 nm and mean fraction of thickly coated particles (f
BC
) of less than 0.15. In contrast, BC
bb
was more
thickly-coated, with mean CT
BC
ranging from ~40 to 70 nm and f
BC
ranging from ~0.23 to 0.47. BC
aged,bg
was
characterized by a mean CT
BC
of ~60 nm and f
BC
of ~0.27, confirming that a mix of biomass burning and
urban emissions sources were likely entrained into these aged background air masses.
By estimating approximate source-to-receptor timescales (i.e., age) and also comparing the physical
properties of fresh rBC to that of BC
aged,bg
, we assessed the effect of aging on both BC
bb
and BC
ff
. For BC
ff
,
we observed that timescales of less than one day were not sufficient for urban rBC particles to become
55
thickly coated. This is in direct contrast to biomass burning rBC, which has been shown in previous studies
to acquire thick coatings within hours or even minutes, near the source of emissions. For BC
bb
, we observed
higher values of f
BC
and CT
BC
during periods that included contributions from the Camp Fire in Northern
California, compared to periods of fresh biomass burning impacts from local Southern California fires (e.g.,
L3). The average CT
BC
during the Camp Fire impacted period was ~18 nm higher than the average CT
BC

during L3, when we identified Southern California fires as the main emission source. Likewise, we also
observed an increase in the urban rBC-containing particles by comparing the aged urban mode of the
BC
aged,bg
distribution to fresh BC
ff
during periods when emissions from the LA basin dominated. We found
that coatings on aged urban particles within BC
aged,bg
were ~35 nm thicker than BC
ff
from fresh LA basin
emissions. Overall, our measurements suggest that aging increases the coating thickness on both BC
ff
and
BC
bb
, which is consistent with previous research. We did not quantify the rate of change of coating
thickness since we were unable track the evolution of the mixing state during source-to-receptor transport.
The measurements reported in this study agree with past research that investigates impacts of source
type and aging on rBC physical properties. This study further highlights the complexity of rBC mixing state
and demonstrates how meteorology, emissions source type, and atmospheric aging can drastically affect the
size distribution and mixing state of BC, even within the same region. Further measurements of rBC
physical properties, along with pollutant measurements that allow for robust source apportionment, would
improve our understanding of BC mixing state in various regions with different atmospheric contexts.
Given that there are less than 20 studies that quantify CT
BC
using the LEO method, this study confirms that
further measurements are necessary to narrow the quantitative bounds of rBC mixing state in our climate
system, which has important implications on BC absorption enhancement and atmospheric lifetime. We also
suggest that future studies further examine the BC mixing state as a function of altitude, as well as the role
of combustion conditions on mixing state (e.g., flaming versus smoldering), especially in real-world field
measurements.
56
2.5 Video supplement
CAMS model output showing the Camp Fire and Southern California plumes during the November 2017
campaign:
https://doi.org/10.5446/42893

NASA MODIS images showing the Camp Fire plume during the November 2017 campaign:
https://doi.org/10.5446/42892

CAMS model output showing the Camp Fire plume reaching Southern California during the December
2018 campaign:
https://doi.org/10.5446/42943

Large-scale circulation of aerosols off the California coast during the December 2018 Campaign:
https://doi.org/10.5446/42942

2.6 Funding and support
This research was supported by the National Science Foundation under CAREER grant CBET-
1752522. This research was also funded in part by the Indo-US Science and Technology Forum.
We acknowledge the use of data from the European Centre for Medium-Range Weather Forecasts
(ECMWF). Neither the European Commission nor ECMWF is responsible for any use that may be made of
the information it contains.
We acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT
transport and dispersion model and/or READY website (http://www.ready.noaa.gov) used in this
publication.
We acknowledge the use of imagery from the NASA Worldview application
(https://worldview.earthdata.nasa.gov/), part of the NASA Earth Observing System Data and Information
System (EOSDIS).

57
Chapter 3 - Measuring the impacts of a real-world neighborhood-
scale cool pavement deployment on albedo and temperatures in
Los Angeles
Published in Environmental Research Letters (Ko et al. 2022)
3.1 Introduction
The urban heat island (UHI) broadly refers to the phenomenon of elevated surface temperatures
(T
surface
) and air temperatures (T
air
) in urban areas relative to surrounding rural areas (Oke 1973, 1982). Luke
Howard (1833) described this phenomenon in scientific literature more than 200 years ago in The Climate of
London, and the UHI has been subsequently confirmed by many observational studies (Howard 1833;
Stewart 2011; Oke 1973, 1995; Fortuniak, Kłysik, and Wibig 2006; Fast, Torcolini, and Redman 2005). The
UHI is primarily caused by (1) urban materials (e.g., asphalt concrete) that absorb and retain heat due to
their optical and thermal properties, (2) urban geometries that trap radiation in the urban canopy, (3) waste
heat from human activities (e.g., air conditioning), and (4) the replacement of natural vegetation with
impervious surfaces (Howard 1833; Oke 1973, 1982). The UHI has important implications on energy
consumption, air quality, critical infrastructure, and human health and comfort in urban environments.
Previous studies have discussed the impact of elevated urban temperatures on increased energy demand for
air conditioning (M. Chen, Ban-Weiss, and Sanders 2018; M. Chen, Sanders, and Ban-Weiss 2019; M. Chen,
Ban-Weiss, and Sanders 2020; Hassid et al. 2000), increased risk of critical infrastructure failure (e.g.,
electrical grid failures) (Stone et al. 2021), potential deterioration of air quality (Sarrat et al. 2006), and
increased risk of heat-related illness and mortality (Patz et al. 2005; D. Li and Bou-Zeid 2013). In the United
States, extreme heat is the leading weather-related cause of mortality (Vaidyanathan et al. 2020).
Compounded with the rise of global average temperature and elevated risks of extreme weather events due
to climate change, along with an increasing proportion of the global population living in urban areas (United
Nations 2019), the negative consequences of the UHI are expected to get worse, barring measures to
58
mitigate the effects of both global climate change and the UHI (Vahmani and Ban-Weiss 2016; Zhao, Lee,
and Schultz 2017; McCarthy, Best, and Betts 2010; Jiachen Zhang et al. 2018).
With most American cities warming at twice the rate of globally averaged warming between 1961
and 2010 (Stone, Vargo, and Habeeb 2012), a host of urban heat mitigation strategies have been proposed in
the last few decades to counteract the UHI (Santamouris, Synnefa, and Karlessi 2011; Stella Tsoka et al.
2020; H. Akbari, Pomerantz, and Taha 2001; Hashem Akbari and Matthews 2012; Taha 2015; Taleghani,
Sailor, and Ban-Weiss 2016; Taleghani et al. 2019; Rosenfeld et al. 1995; Middel et al. 2020; Croce et al.
2021; Vahmani and Ban-Weiss 2016; Mohegh et al. 2017; Jiachen Zhang et al. 2016, 2018, 2019). These
urban heat mitigation strategies (with the exception of anthropogenic heat release control) involve the
modification of urban land cover and/or land use. The main heat mitigation strategies that have been
studied and discussed in the recent literature include solar reflective cool surfaces (e.g., roofs, pavements,
and walls), urban greening to increase shading and evapotranspiration, “green” vegetative roofs and walls,
and non-vegetative shading structures (Jiachen Zhang et al. 2018; Rosenfeld et al. 1998; Taleghani, Sailor,
and Ban-Weiss 2016; Taleghani et al. 2019; Hashem Akbari et al. 2016; Georgescu et al. 2014). All of these
strategies alter the surface energy balance with the goal of reducing the heat that is ultimately trapped in the
urban environment.
The use of solar-reflective cool surfaces has garnered attention in the past few decades as a relatively
straightforward strategy that can partially counteract the UHI. In short, by increasing the albedo (i.e., solar
reflectivity) of urban surfaces, an increased fraction of incident solar radiation at the surface is reflected back
to space. The reduction in absorbed sunlight leads to lower T
surface
, which results in lower sensible heat and
longwave radiation fluxes from the cool surfaces. Given a sufficiently large spatial extent of these cool
surfaces, the reduction in T
surface
should also translate to reductions in near-surface T
air
(Rosenfeld et al. 1995;
H. Akbari, Pomerantz, and Taha 2001; Hashem Akbari and Matthews 2012; Taha 2015; Taleghani, Sailor,
59
and Ban-Weiss 2016; Taleghani et al. 2019; Santamouris, Synnefa, and Karlessi 2011; Middel et al. 2020;
Croce et al. 2021).
The impacts of cool pavements have not been studied as extensively as the impacts of cool roofs,
which have been investigated by numerous studies at different spatial scales (Rosenfeld et al. 1998; Taha,
Douglas, and Haney 1997; Taha, Konopacki, and Gabersek 1999; Jacobson and Ten Hoeve 2012; Jiachen
Zhang et al. 2016; Oleson, Bonan, and Feddema 2010; Synnefa et al. 2008; Millstein and Menon 2011;
Hashem Akbari, Menon, and Rosenfeld 2009; Sharma et al. 2016). These studies have led to real-world
implementation of policy for cool roofs in major cities in the United States, with cities like New York, Los
Angeles, and Chicago passing ordinances that include cool roof requirements (EPA 2020). Some cities are
likewise pushing to implement cool pavements as part of their plans to mitigate the UHI (Garcetti 2019),
but the potential impacts (i.e., both benefits and penalties) of cool pavements, especially at the
neighborhood scale, are not well understood.
To the best of our knowledge, less than 20 peer-reviewed studies to date have examined the isolated
impacts of cool pavements on the urban environment. A majority of these studies are based on numerical
modeling, using either mesoscale climate models (Rosenfeld et al. 1998; Taha 2015; Mohegh et al. 2017) or
micrometeorological computational fluid dynamics models (Santamouris, Xirafi, et al. 2012; Taleghani,
Sailor, and Ban-Weiss 2016; Taleghani et al. 2019; S. Tsoka et al. 2018; Sen, Fernandèz, and Roesler 2020;
Bartesaghi-Koc et al. 2021; Battista, Vollaro, and Vollaro 2021; Croce et al. 2021). One additional modeling
study (Yaghoobian and Kleissl 2012)( used an indoor-outdoor coupled building energy model to study the
effects of cool pavements on building energy use. Regarding observational studies, Rossi et al (2018) used
field measurements of cool pavement test sections to quantify the potential energy savings for artificial
lighting (Rossi, Iacomussi, and Zinzi 2018). Only a single study (Middel et al. 2020) presents data from field
measurements of cool pavements in real-world neighborhoods. A number of studies have also studied the
thermal performance of small test segments of cool pavements in both laboratory and controlled field
60
measurement settings, but these studies were not included in the count mentioned above because they did
not investigate the direct impacts on the urban environment (H. Li, Harvey, and Kendall 2013; M. Zheng et
al. 2015; Anting et al. 2018; Sha et al. 2017; You et al. 2019; Jiang et al. 2020; Y. Zhang et al. 2021).
Although cool pavements are undoubtedly effective at reducing T
surface
, there remains a great deal of
uncertainty regarding their real-world impact on urban climate, especially T
air
and human thermal comfort.
Even the most basic questions concerning the efficacy of cool pavements in reducing T
air
in the real-world
have proven to be difficult to address, with previous estimates of T
air
reductions ranging widely, from
anywhere between 0.1 C to > 2 C. Another complication with cool pavements is that they are applied at
ground level (e.g., as opposed to cool roofs), which means that the reflected shortwave radiation from
pavement surfaces interact with surrounding surfaces in the urban canopy, including buildings and
pedestrians. This complication leads to a tradeoff between the benefits of lower T
air
and the penalties of
increased reflected shortwave radiation at ground level. These penalties have been noted by a number of
previous papers but remain essentially unstudied and highly dependent on the context of the surrounding
physical environment (Santamouris 2013; Taleghani, Sailor, and Ban-Weiss 2016; Middel et al. 2020).
Furthermore, cool pavement albedo may decrease rapidly due to tire wear from vehicular traffic, in addition
to natural weathering. The rate at which cool pavement albedo decreases in the real world and how this
decrease in albedo affects its mitigation performance has not been comprehensively reported in previous
studies.
In this study, we report on field measurements of the largest known residential deployment of cool
pavements to date. The field site was located in the Los Angeles area, and measurements were made before
and after cool pavements were installed at the site during the summer of 2019. Measurements were
subsequently used to determine the neighborhood-scale impacts of cool pavements under real-world
conditions. In particular, the following results are presented in this paper: (1) the spatial and temporal
variability of pavement albedo, (2) the effect of cool pavement on T
surface
, and (3) the effect of cool pavement
61
on T
air
. A brief caveat on the implication of our results on thermal comfort, as well as suggestions for future
research suggestions, are also included in the discussion that follows.
3.2 Methods
3.2.1 Site details
Between 2019 Oct 07 and 2019 Nov 15, ~59,020 m
2
of cool pavements were deployed in a
residential neighborhood in the city of Covina, California, which is located in Los Angeles County ~30 km
east of downtown Los Angeles (see figure 3-1). The cool pavement installation area is shown in figure 3-1
and is hereafter referred to as the “impact area.” The impact area measures approximately 0.8 km x 0.8 km
(0.64 km
2
) and is centered at approximately (34.09, -117.92). In the first phase of installation (2019 Oct 07 to
2019 Oct 16), ~51,670 m
2
of cool pavements were installed. This phase included products 1 to 3, with the
exception of a small segment on Queenside Dr. (see figure 3-2). In the second phase of installation, an
additional ~7,350 m
2
of cool pavements were installed, approximately a month after the first phase. This
additional installation included the rest of product 1 on Queenside Dr and product 4. Further details
regarding the installation schedule can be found in figure 3-2, since measurements were taken before,
during, and after all installations were completed.
62

Figure 3-1. Map of the measurement site in Covina, CA. The impact area (with cool pavements) is shown in blue, while the
control area (without cool pavements) is shown in red. The inset map in the upper-left corner shows the Covina site with respect
to downtown Los Angeles and the greater Los Angeles metropolitan area. Map data ©2020 Google.
63

Start date End Date Description
10/7/2019 10/9/2019 Product 1 installation
10/10/2019 10/14/2019 Product 2 installation
10/15/2019 10/16/2019
Product 3 installation (post-installation mobile T
air
measurements taken after
product 3 installation)
11/13/2019 11/14/2019 Product 4 installation
11/15/2019 11/15/2019
Product 1 extended on Queenside Dr between Coney and Vincent (red dashed
box above)
2/2/2020 2/5/2020 Product 2 re-installed due to skid resistance issues
Figure 3-2. Map of cool pavement installation by product. Modified figure based on original map supplied by Los Angeles
County Department of Public Works. Below the map is a table that describes the corresponding installation schedule. The red
dashed box on the map shows the portion of Queenside Dr that was installed after the first phase of installation.
Directly west of the impact area is a region that we deem the “control area,” which is of comparable
size and land cover features, with no cool pavement installations (see figure 3-1). Measurements in an
64
adjacent and upwind control area were necessary to properly control for factors other than the cool
pavement installation that may affect the temperature comparisons (see details in section 3.2.3).
This pilot installation was coordinated and supervised by the Los Angeles County Department of
Public Works (LADPW) in order to test the real-world performance and impacts of cool pavement. The
companies that produced each cool pavement product were responsible for the installation of the cool
pavement, under the direction of LADPW. The site was chosen by LADPW primarily for logistical reasons,
because the area was under their jurisdiction; but the site also exhibits ideal traits for a cool pavement pilot
study. The site is located in census tract 4057.01, which has an average max temperature of 37.3 °C. This is
2.8 °C higher than the average max temperature of downtown Los Angeles. According to the California
Heat Assessment Tool (www.cal-heat.org, last access 2021 Nov 18), census tract 2057.01 has a tree canopy
cover of only ~3.3% and an estimated average urban heat island delta of ~2.7 °C. These census tract
characteristics make this site a good candidate for a pilot cool pavement installation.
3.2.2 Measurement details
Albedo, T
surface
, and T
air
were measured at varying sampling frequencies, time intervals, and spatial
coverage. Pavement albedo was measured using two Kip & Zonen SMP6 Smart Pyranometers. Stationary
spot measurements of albedo were taken at seven distinct locations, both before and after cool pavement
installation. These seven locations were chosen to represent variability due to the different cool pavement
products and the spatial variability of road wear from traffic (e.g., more tire wear expected near
intersections). Albedo spot measurements were sampled at 1 Hz and average albedo was computed using
representative 1-minute mean values. These spot sampling locations are shown in figure s1 (Appendix B) for
reference. Mobile albedo measurements were also taken at 1 Hz with the same instrument mounted on a
cart, and the data were paired with simultaneous GPS measurements in order to assess the spatial
distribution of albedo over the impact area.
65
Mobile transect measurements of 1.6 m T
air
and pavement T
surface
were taken simultaneously at a
sampling rate of 1 Hz with an Apogee ST-110 thermistor and Apogee SI-111SS infrared radiometer,
respectively. The thermistor was shielded with a white PVC pipe segment (3.81 cm nominal diameter, 15 cm
length) to minimize the effects of incident and reflected shortwave radiation. These instruments were
mounted on a car and measurements were taken first in the impact area and then the control area for each
unique transect (see figure s2, Appendix B). For each day of measurements, full transects were performed
every three hours at the top of the hour, with the first transect starting at 09:00 local standard time (LST)
and the last transect starting at 21:00 LST. Transect measurements were taken on three days pre-installation,
and three days post-installation.
Stationary T
air
measurements were continuously sampled every five minutes, between 2019 Aug 27
and 2019 Nov 06, at four locations in the impact area and five locations in the control area (see figure s3,
Appendix B). HOBO U23-004 External Temperature Data Loggers were used with HOBO RS3-B solar
radiation shields for the stationary T
air
measurements. These T
air
sensors were placed at approximately 3 m
above ground level on county-owned street sign poles (see figure s2, Appendix B). Note that only data from
non-daylight hours were used for the stationary T
air
analysis because of increased reflection of shortwave
radiation at the surface potentially heating the sensor, causing an unintended warming signal.
All GPS measurements were taken with an iPhone 6s at a sampling frequency of 1 Hz, using an iOS
application called GPS Tracker. Representative photos of the instrumentation set-ups can also be found in
Appendix B in figure s2.
3.2.3 Quantifying the direct impact of cool pavement on temperatures: difference-in-
difference (DID) method
To quantify the direct impact of cool pavements on T
surface
and T
air
, we used a technique called the
difference-in-difference (DID) method. The DID method has its origins in econometrics but has since been
used in a wide array of quantitative fields (Gertler et al. 2016; Qiu and He 2017; Wing, Simon, and Bello-
66
Gomez 2018; Henneman, Choirat, and Zigler 2019; Casey et al. 2018; Dempsey and Plantinga 2013; Branas
et al. 2011; Dong et al. 2019; Buerger and Bifulco 2019; Card and Krueger 1994; Poortinga et al. 2018). In
short, the DID method calculates the direct impact of a treatment/event on some variable of interest by
using measurements of the variable of interest in both a control group and an impact group, both before
and after the event of interest. This results in four distinct subsets of measurements: (1) pre-event impact,
(2) pre-event control, (3), post-event impact, and (4) post-event control. These measurements allow us to
calculate how the difference between the impact and the control changes over time. The post-event versus
pre-event difference in the impact versus control difference (i.e., the DID) can then be attributed as the
direct impact of the event on the variable of interest. A simplified illustration of how the DID method
works for any arbitrary variable is shown in figure 3-3.

Figure 3-3. Illustration of the difference-in-difference (DID) method. Δ post is the post-event difference between the impact and
control measurements (e.g., impact – control). Δ pre is the pre-event difference between the impact and control measurements. The
DID is defined as the difference between Δ post and Δ pre. This DID value is therefore the attributable impact of the event on the
variable of interest, Y. “Dummy” values are shown in the figure as a simple illustration. In this example, the event reduced Y by 1
unit (i.e, DID = -1).
67
The advantage of the DID method is that it is able to control for both (1) time-varying confounding
factors, and (2) inherent differences between the control and impact group. In a typical before versus after
comparison, there is no control for time-varying confounding factors. One example of a time-varying
confounding factor in the context of our study is meteorology, which may be different in pre-event and
post-even periods. Likewise, in a typical control versus impact comparison without any before versus after
comparisons, there is no control for inherent differences between the two groups of measurements. One
example in the context of our study would be the difference in land cover features in the control versus
impact area. By using the DID method, the pitfalls of the two single-differencing methods are avoided by
applying an additional round of differencing.
Like all quantitative methods, the DID method has its set of drawbacks and does not necessarily
generate perfect results, but it does provide a more robust estimate of the direct impact of a
treatment/event compared to single differencing methods. There are two main assumptions that are built in
to the DID method. First, it assumes that the characteristics and compositions of the control and impact
groups remain constant, aside from the treatment event. Second, the temporal trends of the variable of
interest in the control and impact groups are assumed to be parallel in the absence of any treatment (i.e., the
rate of change of the variable of interest is assumed to be the same for the control and impact groups).
These assumptions were examined in the context of our study and more details about the statistical
robustness of our estimates can be found in Appendix B.
3.3 Results and discussion
3.3.1 Albedo
Based on mobile albedo measurements taken on 2019 Sep 17-18, the mean (  standard deviation)
pre-installation pavement albedo of the impact area was 0.08  0.02 (see figure 3-4). After cool pavement
installation, the pavement albedo increased significantly, as expected. Seven stationary measurement
68
locations (S1 through S7) were distributed throughout the area that was completed in the first phase of cool
pavement installation. As shown in figure 3-4, the albedo at the seven stationary measurement locations
ranged from 0.22 to 0.38 on 2019 Oct 21, five days after cool pavement installation. This corresponds to an
albedo increase of 175% to 363% relative to the original asphalt pavement (i.e., 2.75x to 4.63x original
asphalt pavement). Subsequently, we observed a steady decline in albedo at all stationary measurement
locations. On 2019 Nov 18, the stationary albedo measurements ranged from 0.20 to 0.29. The albedo
degradation from the first set of measurements on 2019 Oct 21 to the third set of measurements on 2019
Nov 18 ranged from 8% to 30%.

Figure 3-4. Stationary albedo measurements made at seven distinct locations (S1 - S7) throughout the impact area. The markers
represent discrete measurements. The markers and lines are color-coded by the different cool pavement products that were
installed in the impact area. Pre-installation albedo represents the mean albedo of the pavement in the impact area before
installation. The pre-installation albedo is based on mobile albedo measurements taken on 2019 Sep 17-18. A heavy rain event
occurred on 2019 Nov 11 with a recorded 24-hour precipitation of 8.9 mm (NCEI 2021).
A heavy rain event occurred on 2019 Nov 20, and the albedo increased at all stationary
measurement locations. Albedo measurements on 2019 Nov 22, two days after the rain event, ranged from
69
0.23 to 0.31. The relative albedo increases due to the rainstorm (compared to measurements two days prior
to the rainstorm) ranged from 7% to 17%, corresponding to an absolute albedo increase of 0.02 to 0.04.
Albedo measurements were also taken approximately 10 months following the first phase of cool
pavement installation (figure 3-4). Albedo at all seven measurement locations decreased markedly, with
values ranging from 0.16 to 0.24. Pavement albedo degradation over the 10-month timeframe ranged from
12% to 49%. Results for each spot measurement location are summarized in Table 3-1. Of note, we
observed a wide variability in albedo degradation, which we attribute to the performance variability between
different cool pavement products and the spatially heterogenous road wear caused by vehicular traffic.
Table 3-1. Albedo spot measurement details and results.
Location
Product
number
Located
near
intersection,
corner, or
cul-de-sac?
Percent change
(%) between
mean pre-
installation
albedo (0.08) and
initial albedo
(2019 Oct 21)
Percent change
(%) between
2019 Oct 21 and
2019 Nov 18 (~1
month)
Percent change
(%) between
2019 Oct 21 and
2020 Aug 18
(~10 months)
Percent change
(%) between
mean pre-
installation
albedo (0.08) and
10-month albedo
(2020 Aug 18)
S1 1 No +213 -8 -12 +175
S2 1 No +200 -8 -13 +163
S3 3 No +363 -22 -49 +138
S4 2 Yes +313 -30 -33* +175*
S5 2 Yes +263 -14 -17* +200*
S6 1 No +225 -15 -15 +175
S7 1 Yes +175 -9 -27 +100
* Note: Product 2 was reapplied in February 2020 (i.e., four months after initial application) due to issues with skid resistance (see
table s1, Appendix B). Therefore, these values are not necessarily directly comparable to areas that were aged for ~10 months
continuously without reapplication (i.e., albedo degradation for Product 2 at 10 months is likely underestimated slightly).
Nevertheless, these values are presented to show the range of albedo values at different locations throughout the impact area.

The spatial heterogeneity of albedo can be further observed in figure 3-5. Panels a, b, and c show
pavement albedo approximately one week, one month, and one year after cool pavement installation,
respectively. One week after installation we observed substantial spatial variability in albedo, which was
mainly driven by different initial albedos of the different products. The southern half of the impact area was
coated with product 3, while the northern half was coated with products 1 and 2. The corresponding
70
measured albedo is generally higher in the northern area relative to the southern area. This spatial pattern is
consistent with the stationary albedo measurements (see figure 3-4).

Figure 3-5. Panels (a) – (c) show a “heat-map” of albedo ~1 week, ~1 month, and ~1 year after cool pavement installation,
respectively. Panels (d) – (f) show satellite imagery of representative areas of cool pavement with marked albedo decrease
(highlighting with red circles) in the impact area. Map data ©2020 Google.
An additional source of spatial variation in albedo was caused by different patterns of tire wear.
Corners, intersection, cul-de-sacs, and bends were more likely to exhibit greater tire wear and therefore
lower albedo. This effect was observable in the mobile albedo measurements (see figures 3-5a – 3-5c), and
from visual inspection of satellite imagery (see figures 3-5d – 3-5f). Over the timespan of a year, the average
albedo of the three cool pavement products decreased from 0.26 to 0.18. This corresponds to a spatially
averaged albedo degradation of 31% over the year.
While others have measured the effects of aging on cool pavement albedo in either controlled
laboratory settings or using small-scale field measurements (X. Cao et al. 2016; Shirakawa et al. 2020;
Alchapar, Correa, and Cantón 2013), we present here the most extensive results to date from long-term
albedo monitoring of neighborhood-scale cool pavements subjected to real-world exposure, which we
71
define here as the combination of vehicular traffic and natural weathering. One other study (Lontorfos,
Efthymiou, and Santamouris 2018) reported a one-year albedo reduction from 0.26 to 0.15 for a 4,600 m
2

cool pavement installation on a single street in Athens, Greece. Their reported values are consistent with the
range of albedo degradation we observed in this study. One important distinction between Lontorfos et al
(2018) and our study is that they reported a majority of the albedo degradation occurring within a month of
installation, while we saw that albedo values did not asymptote to a steady state value at the one-month
timeframe (see figure 3-4). This implies that cool pavement albedo performance over time is not universally
consistent and can be dependent on the context of the physical environment and the cool pavement
product.
Akbari and Matthews (2012) reported that aged asphalt pavements have albedos that range from
~0.10 to 0.18 (Hashem Akbari and Matthews 2012). We found that the mean cool pavement albedo in this
study decreased to 0.18 within one year. This implies that aged cool pavements (at least for the products
measured in our study) subjected to real-world exposure may have albedo values similar to that of aged
asphalt pavement. An oft-repeated concern with cool pavements is that vehicular traffic could result in
albedo reductions that reduce their efficacy for heat mitigation. This study highlights the critical need to plan
for degradation in albedo if cool pavements are deployed in the real world, either through improved
maintenance, higher performance materials, and/or regular reapplication.
3.3.2 Surface temperature
Pavement T
surface
was measured as part of the mobile transects described earlier in section 3.2.2.
Figure 3-6 shows a matrix of barplots, where each barplot presents the DID at different hours throughout
the day at 3-hour intervals. These DID values were calculated using the mobile T
surface
measurements from a
pair of distinct measurement days, which are labelled at the top of each column and end of each row. For
example, the barplot in the top-left corner shows DID values for T
surface
calculated using measurements from
2019-09-23 and 2019-10-05. In figure 3-6, the DID values shown in the red rectangular outline refer to the
72
impact of cool pavement installation on T
surface
. Our measurements show that T
surface
reductions due to cool
pavement installation follow a consistent diurnal cycle, with a peak reduction of 5 ºC observed at 15:00 LST.
The smallest reduction of 0.9 ºC was observed at 09:00 LST.

Figure 3-6. Matrix of barplots showing the surface temperature DID for unique pairs of dates. The columns identify the first
date used in the DID calculation and the rows identify the second date. The barplots outlined in red show the DID values that
represent the impact of cool pavement installations on surface temperature. The barplots outlined in blue highlight the null DID
impacts that occur when the DID calculations were made with two pre-installation dates or two post-installation dates. A negative
DID value indicates a surface temperature reduction. The error bars represent the 95% confidence intervals for the mean.
The barplots outside of the red outline in figure 3-6 show the DID values calculated for either pairs
of pre-installation dates or pairs of post-installation dates; the barplot in the top-left corner presents DID
values for pairs of pre-installations dates, and the three barplots in the lower-right corner present DID
values for pairs of post-installation dates. The pre-installation and post-installation DID values in the blue
boxes in figure 3-6 are nearly all centered about zero, with the 95% confidence intervals implying a null
effect. These results were expected since there was no intervention that would have changed the DID for
73
these date pairs. These DID results for pre-installation and post-installation date pairs confirm that the DID
method is correctly controlling for external variabilities and can robustly identify the isolated effect of cool
pavement installation.
In the two post-installation pair DID plots in the last row of figure 3-6, the slightly positive DID
values (with the exception of 09:00 LST) imply that there was a “treatment effect” that occurred between
2019-10-25 and 2020-08-14 that increased the T
surface
of the cool pavement. The treatment effect in this
context was likely the degradation of the cool pavement albedo due to real-world weathering as explained in
the previous section. In other words, the albedo degradation within a 10-month time period reduced the
ability of the cool pavement to reduce T
surface
by ~1 ºC during the daylight hours.
Overall, an increase of the average pavement albedo by 0.18 led to a maximum T
surface
reduction of 5
ºC. This corresponds to a T
surface
reduction rate of 2.7 ºC per 0.1 increase in pavement albedo. These T
surface

reductions are consistent with the range of what previous studies have reported. Most recently, Middel et al
(2020) reported that cool pavements were up to 6 ºC cooler than nearby non-cool pavements. Furthermore,
we found that an albedo degradation of 0.08 after approximately one year of real-world exposure reduced
the cooling efficacy for T
surface
by ~1 ºC (Middel et al. 2020). By linear extrapolation, we estimate that an
albedo degradation of 0.1 will lower the cooling efficacy for T
surface
by ~1.25 ºC. To our knowledge, this is the
first time that the loss of T
surface
reduction potential was directly measured and related to albedo changes over
time using neighborhood-scale field measurements.
3.3.3 Air temperature
As described in section 3.2.2, we measured both stationary and mobile T
air
. Figure 3-7 shows a
scatterplot of Δ
post
versus Δ
pre
using data from the stationary T
air
measurements, where Δ signifies the
difference in T
air
between the impact area and the control area, and “post” or “pre” signifies whether the
measurements were taken post-installation or pre-installation, respectively. Note, the post period for air
temperature measurements described here refers to the period between the completion of the first
74
installation phase and the start of the second installation phase. This means that the impact on air
temperature described here is attributable to the cool pavements installed in the first phase (see figure 3-2).
Our measurements indicate that cool pavement installation induced a reduction in 3 m T
air
between 19:00
and 23:59 LST, though only reductions for the hours of 20:00, 21:00, and 22:00 were statistically
distinguishable from zero at a 95% confidence level. A maximum evening 3 m T
air
reduction of 0.19 ºC
occurred at 21:00 LST.

Figure 3-7. A scatter plot for hourly mean Δ post versus Δ pre, where Δ is the difference between 3 m air temperature in the impact
area versus control area (i.e., impact minus control). “Post” or “pre” in the subscript signifies whether the measurements were
made post-installation or pre-installation, respectively. The values inside each circle represents the hour of day and the error bars
represent the 95% confidence interval for the mean. Values above the 1:1 line imply a warming effect due to cool pavement, while
values below imply a cooling effect. The air temperature DID can be calculated as (Δ post - Δ pre). Only non-daylight hours were
included in this analysis (see section 3.2.2). The first non-daylight hour was 18:00 LST. Data from hours 00:00 LST to 06:00 LST
showed a null effect (i.e., neither warming nor cooling effect) and are thus not included in the figure.
From our mobile transect measurements, we found that cool pavements reduced 1.6 m T
air
by a
mean  95% confidence interval of 0.20  0.06 ºC at 12:00 LST. A statistically distinguishable DID signal
was only obtainable for 12:00 LST due to high micrometeorological variability, and the small T
air
reduction
75
signal expected from the scale of the cool pavement installation in this study. Further details regarding these
complications and the statistical procedures implemented can be found in Appendix B. Note that the
absence of statistically significant results for other hours of the day does not preclude cool pavements from
having caused changes in 1.6 m T
air
at those other hours. For these null hours, the noise due to
micrometeorological variability within the timeframe of the mobile transect made it difficult to distinguish
the magnitude of the T
air
signal (see Appendix B).
Our stationary and mobile T
air
measurements imply that cool pavements reduce near-surface T
air

both during the daytime, as well as during the evening. We were unable to identify a definitive diurnal
pattern of near-surface T
air
reduction. However, we can conclude that the neighborhood-scale cool
pavement installation investigated here can reduce near-surface T
air
by 0.20 °C around noon, and at least
0.19 ºC in the late evening, within the context of the local meteorology and land surface features specific to
our field site (Covina, CA).
To date, this is the first study to report statistically significant T
air
reductions caused by the isolated
impact of cool pavement installations, using a controlled experimental technique to account for
confounding variables. Middel et al (2020) reported maximum T
air
reductions of ~0.4 to 0.5 ºC, but pre-
installation measurements were not taken as part of that study, which means that confounding variables
could not be accounted for. We also find that our results are relatively consistent (i.e., on the same order of
magnitude) with past modeling studies examining the effects of cool pavements in the Los Angeles area.
Taleghani et al (2019) used micrometeorological CFD modeling to investigate impacts of cool pavements
(and other heat mitigation strategies) at neighborhood scale and found that increasing pavement albedo by
0.3 could lead to a neighborhood-averaged reduction in T
air
of ~0.26 ºC. Taha (2015a, 2015b) used
mesoscale modeling to estimate T
air
reductions of approximately 0.25 to 0.5 ºC in Los Angeles for moderate
(0.15) to high (0.25) increases in pavement albedo for a hypothetical city-scale installation. Mohegh et al
(2017) also used mesoscale modeling to investigate cool pavement impacts and found that increasing
76
pavement albedo by 0.4 in California cities could reduce T
air
by ~0.18 to 0.86 ºC, depending on the city.
Millstein and Levinson (2018) used an idealized heat transfer model of air flowing over a 200 m hot plate
and estimated a ~0.15 ºC reduction in 1.5 m T
air
when albedo was increased by 0.3. This is remarkably close
to the maximum reduction of ~0.20 ºC we observed, considering that the maximum east-west street length
in our impact area was ~650 m. Table 3-2 places our results in the context of these previous studies by
summarizing the abovementioned T
air
reductions per 0.1 increase in albedo. Though the results reported
here may be specific to neighborhood characteristics and local meteorology during the measurement
campaign, this study can serve as a real-world reference for hot and dry regions like Los Angeles.
Table 3-2. Comparison of reported T
air
reductions per 0.1 increase in albedo from previous modeling studies.
Study Spatial scale
T
air
reduction per 0.1
increase in albedo (°C)
This study Neighborhood-scale 0.11
Taleghani et al (2019) Neighborhood-scale 0.09
Taha (2012, 2013) City-scale 0.17 – 0.2
Mohegh et al (2017) City-scale 0.05 – 0.22
Millstein and Levinson (2018) Hypothetical hotplate (~200 m) 0.05
3.3.4 Implications on pedestrian thermal comfort
Although we do not report on measurements of pedestrian thermal comfort here, we find it
important to discuss the implications that cool pavements can have on thermal comfort. The cool pavement
penalty of increased upwelling shortwave radiation at the ground level has been noted in previous studies
(Santamouris 2013; Taleghani, Sailor, and Ban-Weiss 2016; Middel et al. 2020), yet the nuances of cool
pavement tradeoffs are often overlooked by the public and policy makers, potentially leading to the
misconception that cool pavements can serve as a “policy panacea” for urban heat mitigation (Middel et al.
2020). Here we briefly review the main tradeoffs related to cool pavements’ impact on thermal comfort,
discuss relevant questions that have not been clearly addressed to date, and discuss our results in the context
of these tradeoffs.
77
The increased albedo of cool pavements relative to asphalt pavements will by definition lead to an
increase in reflected shortwave radiation at the ground level. This results in an increased shortwave radiation
load on all surrounding surfaces in the vicinity of the cool pavements, including pedestrians. One main
concern is that the increased shortwave radiation can potentially decrease pedestrian thermal comfort
significantly during daylight hours. Most notably, Middel et al (2020) reported that cool pavement increased
the shortwave radiation load over the pavement by up to 168 Wm
-2

during the daytime. This corresponded
to an observed mean radiant temperature (T
MRT
) increase of 4 ºC.
Although this observed increase in mid-day T
MRT
is certainly a cause for increased scrutiny of cool
pavements, we suggest that an increased radiant load directly over the pavement does not fully account for
the complexities that impact outdoor pedestrian thermal comfort. Controlling for inter-subject variability of
biometeorological factors, such as evaporative cooling rates from sweating, effects of clothing, and
metabolic rates; there are only two ways that cool pavement can directly impact thermal comfort: (1)
radiative heat transfer, or (2) convective heat transfer. With respect to radiative heat transfer, cool
pavements will increase shortwave radiative load on nearby pedestrians, but longwave radiative loads will be
decreased due to lower pavement T
surface
. With respect to convective heat transfer, cool pavements can lower
near-surface T
air
, which can in turn decrease the convective heat load on pedestrians during days for which
the T
air
is higher than the pedestrian skin temperature, or increase convective cooling during days for which
the T
air
is lower than the pedestrian skin temperature. In either case, the reduction in T
air
would tend to make
the pedestrian more comfortable, all else being equal.
Based on previous approximations (ASHRAE 1993; De Dear et al. 1997), the radiative heat transfer
coefficient (h
r
) and the convective heat transfer coefficients (h
c
) for the human body are approximately the
same magnitude at low wind speeds (< 1 m/s). This implies that an increase in radiative heat transfer due to
a unit increase in T
MRT
is similar in magnitude to an increase in convective heat transfer due to a unit
increase in T
air
. As a hypothetical example, if T
MRT
increased by 4 ºC (as observed by Middel et al (2020)
78
directly over cool pavement), but T
air
was also reduced by 4 ºC, the thermal comfort change should be
negligible under low wind speeds.
The influence of wind speed further complicates this radiation-convection tradeoff. For illustration,
the average wind speed at 12:00 LST recorded at Los Angeles International Airport (LAX) during the peak
of summer is ~3 m s
-1
(NCEI 2021). According to measurements conducted by de Dear et al (1997), h
c
/h
r
is
~4 when the wind speed is 3 m s
-1
. In this case, if T
MRT
increased by 4 ºC, a simultaneous reduction in T
air
of
~1 ºC would compensate for the increased radiative load. The reduction in T
air
of 0.20 ºC at 12:00 LST
observed in our study is not sufficiently large to compensate for a T
MRT
increase of 4 ºC. However, this
balance would change as T
air
reductions exceed 1 ºC. Based on a number of modeling studies (Santamouris,
Gaitani, et al. 2012; Taleghani, Sailor, and Ban-Weiss 2016; Mohegh et al. 2017), it is generally expected that
larger spatial extents of cool pavement deployment would lead to larger reductions in T
air
. It is interesting to
note that while T
air
reductions are expected to be strongly dependent on the spatial extent, the radiant load
effect from increased shortwave radiation is not. Currently, there are no studies that adequately address this
radiation-convection tradeoff, which will be largely dependent on local meteorology and the spatial extent of
cool pavements.
In addition, to the radiation-convection tradeoff, there is also considerable uncertainty regarding the
effect of reflected shortwave radiation on pedestrians on the sidewalks, where most foot traffic is expected. It
is well-known that the albedo of concrete sidewalks is already substantially higher than that of black asphalt
pavement, with concrete albedo values ranging but generally exceeding 0.2 (Hashem Akbari and Matthews
2012; Sen, Roesler, and King 2019; H. Li, Harvey, and Kendall 2013; Xu et al. 2019). To add to the
complexity, the geometry and composition of the “buffer zones” between the sidewalk and the road vary
considerably. For example, some sidewalks may have no buffer at all, while other streets may have segments
of vegetation distancing the sidewalks substantially from the road, or even parked cars or shrubs that could
block a substantial proportion of any reflected shortwave radiation from the pavement. Although it is clear
79
that the radiative load would increase substantially for a pedestrian standing directly over the cool pavement
during the day, it is less clear how thermal comfort would be impacted for pedestrians on sidewalks. Middel
et al (2020) reported T
MRT
on sidewalks versus on cool pavement, and also reported an additional 20 to 30 W
m
-2
of reflected shortwave radiation on the sidewalks during the early evening. But there still exists vast
uncertainty in how cool pavements impact thermal comfort for pedestrians on sidewalks in a range of real-
world configurations. It would be helpful for future research to quantify how sidewalk-specific factors (e.g.,
distance to roadway, buffer characteristics, surrounding vegetation) can attenuate the impact of cool
pavement on thermal comfort of pedestrians on sidewalks.
Lastly, the potential tradeoffs between nighttime decreases in T
air
and daytime increases in T
MRT
need
to be critically examined. There is growing evidence that nighttime temperatures are increasing faster than
daytime temperatures, and there is increasing concern about resulting public health impacts (USGCRP 2017;
Murage, Hajat, and Kovats 2017; Kovats and Hajat 2008; Laaidi et al. 2012). Although cool pavements may
decrease nighttime temperatures, it remains uncertain whether the potential benefits of nighttime
temperature reductions outweigh the consequences of potentially decreasing daytime outdoor thermal
comfort.
We assert that the various tradeoffs (i.e., benefits and penalties) of cool pavements with respect to
thermal comfort need to be quantified more explicitly in future research. A large portion of the uncertainty
can be attributed to the wide variety of hyper-local features (e.g., sidewalk buffer characteristics,
meteorology, vegetation) that may affect the impacts of cool pavements, and the general lack of controlled
real-world field measurements. This study confirms that cool pavement can generally reduce T
air
, but it
remains an open question whether cool pavements can play a viable role in increasing net human health and
wellbeing, and just as importantly, in what context.
80
3.4 Conclusion
This study reports on field measurements of the largest neighborhood-scale installation of cool
pavements to date. Measurements were used to investigate the direct impacts of cool pavement installation
on T
surface
and T
air
. Albedo was also monitored to evaluate changes over the time span of a year.
The mean pavement albedo of the impact site (i.e., the neighborhood where cool pavements were
installed) increased significantly after installation of cool pavement, from 0.08 to 0.26. This initial increase in
albedo caused average T
surface
reductions ranging from 0.9 to 5 ºC, with maximum reductions observed at
15:00 LST. T
air
reductions were also observed using both mobile and stationary measurements. Mobile
measurements showed that cool pavements reduced 1.6 m T
air
by 0.20 °C at 12:00 LST. Stationary
measurements showed that cool pavement installation induced a reduction in 3 m T
air
between 19:00 and
23:59 LST, though only reductions for the hours of 20:00, 21:00, and 22:00 were statistically distinguishable
from zero at a 95% confidence level. A maximum evening 3 m T
air
reduction of 0.19 ºC occurred at 21:00
LST. Our measurements suggest that cool pavement installations at neighborhood-scale (0.8 km x 0.8 km)
can lead to a statistically significant reduction in T
air
, both during daytime and in the evening.
After a year of real-world exposure (i.e., natural weathering plus tire wear from vehicles), cool
pavement albedo decreased to 0.18, which corresponded to a degradation of 30%. We found that albedo
degradation was dependent on the type of cool pavement product used, as well as tire wear, which led to
more rapid degradation at curved road sections, cul-de-sacs, and intersections where vehicles would be
turning. Albedo degradation was also found to have a significant effect on T
surface
, as expected. An albedo
degradation of 0.08 after approximately one year of real-world weathering corresponded to a lower cooling
efficacy for T
surface
by ~1 ºC.
Future research should aim to reduce uncertainties associated with the impact of cool pavements on
pedestrian thermal comfort. We suggest that future research focus on reducing uncertainties in (1) the
radiation-convection tradeoff, (2) the pedestrian thermal comfort on sidewalks, and (3) the tradeoff between
81
nighttime cooling versus increased daytime shortwave radiation. Reducing such uncertainties will help policy
makers and urban planners make decisions about when and where to install cool pavements based on a
nuanced understanding of their benefits and penalties.
3.5 Funding and support
The authors would like to thank Christopher Sheppard, Clarence Su, Van Truong, Yonah Halpern,
and all other associated staff of Los Angeles County Public Works for coordinating and executing the cool
pavement pilot installation, and for their technical and administrative support throughout this study.
82
Chapter 4 - Modeling the impacts of anthropogenic heat on the
regional climate of Los Angeles
4.1 Introduction
As noted in the chapter 3, the UHI effect has substantial implications on human health and comfort.
The impacts of the UHI are projected to get worse in the future with the compounded effects of global
climate change and increasing urban populations. Of the various factors that contribute to the UHI,
anthropogenic heat is arguably the most neglected (Sailor and Lu 2004). For urban dwellers, it is easy to
observe the heat trapping and absorbing materials that are used widely in our built environments, but the
waste heat that is generated from anthropogenic activities goes largely unnoticed by the public.
In the context of the UHI effect and this study, anthropogenic heat is defined as any waste heat that
is generated by any kind of human activity and is ultimately released into the atmosphere. The main sources
of anthropogenic heat include vehicular traffic, buildings, industrial activities, and to a lesser degree, human
metabolism. For example, fossil-fuel powered vehicles emit waste heat from their combustion processes.
Waste heat is produced from HVAC systems in buildings, primarily through heating and cooling activities.
Waste heat is also emitted from a variety of industrial activities including, but not limited to, the extraction,
transformation, production, and manufacturing of materials and goods. Lastly, humans themselves emit
waste heat through metabolism, and when gathered in dense populations, this waste heat has a non-
negligible contribution to the overall anthropogenic heat budget (Sailor and Lu 2004).
Efforts to quantify the extent and impact of anthropogenic heat can be traced back decades, to the
1970’s and 1980’s (Torrance and Shun 1976; Clark, Bornstein, and Tam 1985; Harrison, McGoldrick, and
Williams 1984), but uncertainties remain, especially with regards to its spatiotemporal distribution in
complex urban environments. In this study, we will use a hybrid top-down and bottom-up methodology to
develop a high spatial resolution anthropogenic heat emissions dataset for the Los Angeles region. Although
previous efforts have been made to constrain and characterize anthropogenic heat in the Los Angeles region
83
(Sailor and Lu 2004; Y. Zheng and Weng 2018), our dataset will be constructed using novel datasets and
methodologies that will provide more accurate representations of real-world processes, as further described
in section 4.1.3. Furthermore, we will use this newly generated dataset to conduct regional climate modeling
using the Weather and Research Forecasting (WRF) model to assess the impact of anthropogenic heat on
the UHI in the Los Angeles region. Additional energy use scenarios (e.g., 100% renewable energy) will also
be modeled using WRF to assess the potential mitigating impact of renewable energy systems on the UHI
effect.
Sailor and Lu (2004) developed a top-down methodology that incorporated publicly accessible data
to estimate spatially averaged anthropogenic heating profiles for six large US cities, including Los Angeles.
Based on their estimates, they found that the City of Los Angeles had maximum average AHF values
exceeding 30 W m
-2
. More recently, Zheng et al. (2018) developed a gridded 120 m resolution AHF dataset
for Los Angeles County, using a hybrid approach incorporating both inventory and GIS approaches. Using
their constructed dataset, they estimated that the maximum average AHF value for Los Angeles County was
7.76 W m
-2
.
After conducting a comprehensive literature review, we found that there were at least 27 existing
studies that assessed the regional climate impacts of anthropogenic heat on various cities, using mesoscale
modeling. 18 of these modeling case studies were based in East Asia, four in Southeast Asia, four in Europe,
and only one in the US. None of the 27 studies modeled anthropogenic heat impacts in the Los Angeles
region. Further details summarizing these studies can be found in Table 4-1.
Table 4-1. List of modeling studies that quantify the impact of anthropogenic heat
Study Model details Location Region
(Z. Cao et al. 2019) WRF/UCM Guangzhou, China East Asia
(Feng Chen, Yang, and Zhu 2014) WRF/UCM Hangzhou City, China East Asia
(L. Chen and Frauenfeld 2016) WRF/UCM Hangzhou City, China East Asia
(Feng et al. 2012) WRF/UCM China East Asia
84
(Feng, Wang, and Ma 2015) WRF/UCM China East Asia
(Ichinose, Shimodozono, and Hanaki
1999) CSU-MM Tokyo, Japan East Asia
(Kikegawa et al. 2014) WRF/BEP Tokyo and Osaka, Japan East Asia
(Lin et al. 2008) WRF/UCM Taipei, Taiwan East Asia
(R. Liu and Han 2016) WRF/UCM Beijing-Tianjin-Hebei, China East Asia
(R. Liu et al. 2017) WRF/UCM Beijing, China East Asia
(B. Liu et al. 2021) WRF/UCM Beijing, China East Asia
(Ryu, Baik, and Lee 2013) WRF/CMAQ Seoul, South Korea East Asia
(Takane et al. 2019) WRF/BEP Osaka City, Japan East Asia
(X. Wang et al. 2015) WRF/UCM Yangtze River Delta, China East Asia
(Y. Wang et al. 2017) WRF/BEP Hong Kong East Asia
(Y. Wang et al. 2018) WRF/BEP Hong Kong East Asia
(Wen et al. 2020) WRF/UCM Pearl River Delta, China East Asia
(Xie et al. 2016) WRF-Chem/UCM South China East Asia
(Yu et al. 2014) WRF-Chem/UCM Beijing, China East Asia
(Giovannini et al. 2014) WRF/BEP Trento, Italy Europe
(Molnár, Kovács, and Gál 2020) WRF/UCM Szeged, Hungary Europe
(Salamanca, Martilli, and Yagüe 2012) WRF/BEP Madrid, Spain Europe
(Vitanova and Kusaka 2018) WRF/UCM Sofia, Bulgaria Europe
(Doan, Kusaka, and Nguyen 2019) WRF+modified UCM Hanoi, Vietnam Southeast Asia
(X. X. Li et al. 2013) WRF/UCM Singapore Southeast Asia
(X. X. Li et al. 2016) WRF/UCM Singapore Southeast Asia
(Mughal et al. 2019) WRF/BEP Singapore Southeast Asia
(Salamanca et al. 2011) WRF (variations) Houston, Texas US

In the single US study mentioned above, Salamanca et al. (2011) found that anthropogenic heat can
increase nighttime temperatures in Houston, Texas by up to 2 °C. Houston is the fourth largest city in the
US, with a total 2020 population of 2,304,580 and a population density of 3,502 mi
-2
. In comparison, Los
Angeles is the second largest city in the US, with a much larger total 2020 population of 3,792,621 and a
population density of 8,092 mi
-2
(US Census Bureau 2020). Since anthropogenic heat is expected to scale
directly as a function of population density and human activity, Los Angeles may have an even greater
regional climate impact due to anthropogenic heat. With growing evidence of nighttime temperatures
increasing more rapidly than daytime temperatures, the potential for large impacts of anthropogenic heat on
85
nighttime temperatures is particularly important to quantify for the sake of public health (Murage, Hajat, and
Kovats 2017; Kovats and Hajat 2008; Laaidi et al. 2012).
The need for more anthropogenic heat studies, particularly for the Los Angeles region, is evident
from our literature review. Although several modeling studies have been conducted as mentioned above,
none of them are representative of the regional climate, population, and human activity characteristics that
are unique to the Los Angeles region. Our study will be the first to explicitly characterize the impacts of
anthropogenic heat on the Los Angeles region using a customized, high resolution AHF dataset and
mesoscale climate modeling. This will provide timely information to policy makers and the general public as
we make urgent efforts as a society to mitigate against and adapt for the UHI and climate change.
4.2 Methods
For our dataset, we used a hybrid approach by integrating GIS-based methods with the inventory
method. A variety of datasets were used to construct each component of the dataset. Table 4-2 lists the data
sources used to produce the final dataset with brief descriptions of each. Figure 4-1 provides an overview of
the whole dataset construction process in a diagram format.
Table 4-2. List of the main datasets used to construct the AHF dataset. Meaning of the acronyms used here can be found in the
acronym dictionary provided in the Appendix.

Category Data Open-access? Source Spatial aggregation Temporal aggregation Year
Building energy consumption yes UCLA Energy Atlas census tract annual 2016
parcel information yes LA County parcel continuously updated last updated 2022
building geometry yes LARIAC building annual 2017
energy consumption yes CEC county annual 2016
smart meter electricity consumption no SCE climate zone hourly 2016
Comstock natural gas consumption yes DOE building hourly 2016
smart meter natural gas consumption no SocalGas city hourly 2019
Traffic simulated VMT and VHT no SCAG road segment annual 2016
arterial traffic count no LA Metro road segment hourly 2016
highway traffic count yes Caltrans road segment hourly 2016
Metabolism CTPP commute patterns data yes U.S. Census Bureau census tract 5-year average 2006-2010
ACS population data yes U.S. Census Bureau census tract annual 2010, 2016
86

Figure 4-1. Flowchart describing the different datasets used and the overall process of constructing the AHF data products.
In general, the inventory method assumes that all energy consumed will be converted to sensible or
latent heat flux that is directly released into the atmosphere. Numerous studies on anthropogenic heat have
used some form of the inventory method to estimate AHF (Taha 1997; Torrance and Shun 1976; Kimura
and Takahashi 1991; Lee et al. 2009; Ichinose, Shimodozono, and Hanaki 1999; Fan and Sailor 2005;
Hamilton et al. 2009; Block, Keuler, and Schaller 2004; Fortuniak, Kłysik, and Wibig 2006). Although the
inventory method is useful because of its practicality and relative ease of implementation, it has some key
limitations. First, the inventory method assumes that all energy consumed is instantaneously converted to
sensible heat, with no time lag. Although attempts can be made to correct this with some form of a time lag,
this is not commonly done in other studies (Torrance and Shun 1976; Kimura and Takahashi 1991; Sailor
2011). Second, it is often hard to obtain accurate energy consumption data at fine temporal and spatial
scales, which means that further assumptions or data are needed to downscale to finer spatiotemporal
resolutions.
Although the time lag issue is not accounted for in our dataset, previous studies show that estimates
using physics-based models and inventory methods without time lag adjustments are comparable (Chow et
87
al. 2014; S. Wang et al. 2020). To address the issue regarding the limited resolution of energy consumption
data, we used auxiliary geospatial datasets to supplement the consumption data and increase both spatial and
temporal resolution. The following sections below describe in further detail how AHF was estimated for
each category.
4.2.1 Buildings
Building sources included in our dataset were grouped into three categories: residential, commercial,
and industrial buildings. The UCLA Energy Atlas provides annual electricity and natural gas consumption
data for each category, aggregated to census tracts. The 2016 Energy Atlas dataset serves as the backbone
“ground truth” of our dataset, as it has been thoroughly validated and used in several previous studies
(Pincetl et al. 2019; Pincetl and Newell 2017; Burillo et al. 2019).
To resolve the seasonal and diurnal temporal variability of energy consumption, we relied on three
additional datasets. For residential electricity consumption, we used smart meter electricity consumption
data provided by Southern California Edison (SCE). This smart meter dataset consists of hourly electricity
consumption of a representative sample of individual households in Southern California. We assumed that
households in similar climates would have similar diurnal patterns in electricity consumption throughout the
year, so we created an average annual hourly diurnal profile of electricity consumption for each building
climate zone (designated by the California Energy Commission (CEC)) in each month of the year based on
2016 electricity consumption data.
For commercial electricity consumption, NREL’s ComStock dataset was used. The ComStock
dataset provides sub-hourly load profiles aggregated to Public Use Microdata Areas (PUMAs), which are
“non-overlapping, statistical geographic areas that partition each state … into geographic areas containing
no fewer than 100,000 people each (U.S. Census Bureau, last access 7/1/2022).” ComStock was created
using physics-based, building energy models, which were also calibrated with smart meter data. Since smart
meter data for commercial buildings was not available to us, the modeled ComStock dataset serves as a
88
reasonable alternative to direct observations. Monthly-averaged, diurnal profiles of energy consumption
with 15-minute time resolutions are provided by the Comstock data. The Comstock dataset we used is
representative of the year 2018. Hourly-averaged load profiles for each month were created using the
ComStock 15-minute load profile dataset.
For natural gas consumption, we use aggregated smart meter data representative of the year 2019,
provided by Southern California Gas Company (SoCalGas). The SoCalGas data consists of average, hourly
natural gas consumption aggregated to the city level, and split by building category. This allowed us to
characterize the temporal variability of natural gas consumption for the three distinct building categories,
based on directly observed data. Smart meters were not fully implemented by SoCalGas yet in 2016, so 2019
data was used as a proxy.
Finally, for industrial electricity consumption, we assumed an idealized load profile based on
estimated load profiles provided in a 2013 report released by Oak Ridge National Laboratory (ORNL 2013).
See section 4.3.1 and Figure 4-9 for further details regarding the industrial load profiles.
To spatially resolve the building AHF dataset further from the census tract level down to the
gridded resolution of 100 m, we used two geospatial datasets that describe the building characteristics in LA
County. The first dataset is the Los Angeles Region Imagery Acquisition Consortium (LARIAC) building
geometry dataset, which is publicly accessible through the LA City Geohub website (City of Los Angeles
Geohub, last access 7/1/2022). The LARIAC building geometry dataset is based on digital aerial imagery
data acquired through LiDAR measurements. The LARIAC dataset provides accurate building geometries
for over 3 million buildings in LA County. The second dataset is the LA County Parcels dataset, which
describes the building vintage (i.e., year built), square footage of the building, and most importantly, the
building use type (i.e., category). The parcels dataset is also available on the LA City Geohub website (City
of Los Angeles Geohub, last access 7/1/2022). By joining these two geospatial datasets, we generate both a
detailed understanding of the geospatial distribution of buildings, as well as a building level understanding of
89
their individual characteristics. Since we know the building geometry for each building, as well as its use
type, we then used geospatial methods to “weight” each grid cell by a building size metric of choice. For
example, one could weight each grid cell by its fraction of building volume, building footprint area, or
building floor area, relative to the total respective quantity in the whole census tract. For our initial dataset
prototype, we used building volume to geospatially distribute the AHF within each census tract. This
method is a variation on a geospatial re-distribution method called dasymetric interpolation (Petrov 2012).
Figure 4-2 illustrates how dasymetric mapping works with an example of mapping population distribution
from the US Geological Survey (USGS).
To further calibrate our estimates, we applied an additional top-down constraint with the total
county-wide energy consumption reported by the CEC. Figure 4-3 shows a heatmap of anthropogenic heat
from buildings (AHF
building
), with each individual category represented in panels (a) – (c), and the total shown
in panel (d). Note, that these figures only show AHF for a segment of Central LA. The neighborhoods
included in this subset of Central LA include Downtown LA, Koreatown, Pico-Union, Westlake, and
Harvard Heights. The full-scale dataset is shown and described further in the Results section below.

Figure 4-2. This is an example of dasymetric mapping from a USGS report. This figure illustrates how population distribution is
mapped at finer resolution using auxiliary land use data. Panel (a) on the left shows a choropleth plot using data from the US
Census Bureau at the census block-group level. Panel (b) shows an improved representation of population distribution using land
cover data and dasymetric mapping. A similar approach is taken for the AHF datasets in our study by using auxiliary data on road
and building geometries. (Source: USGS 2008, last access 7/1/2022)
90

Figure 4-3. AHF building from (a) residential, (b) commercial, (c) industrial, and (d) all buildings for a subset of Central LA. This
includes the neighborhoods of Downtown, Koreatown, Westlake, Pico-Union, and Harvard Heights. The AHF shown here is
representative of an average hour in the year of 2016. The final dataset will incorporate temporal variability, but here we show a
representative average to illustrate the spatial variability in AHF building. Note, the range of the colorbar in panel (d) is different than
the colorbars in panels (a) – (c).
4.2.2 Traffic
The AHF dataset for traffic (AHF
traffic
) was constructed in a similar manner to AHF
buildings
. Modeled
traffic flow dataset provided by the Southern California Association of Governments (SCAG) served as the
backbone of the AHF
traffic
dataset. The SCAG data provides estimates of annual VMT for every major
arterial road and highway segment in LA County, and it also provides an estimate of VMT for minor roads
aggregated to traffic analysis zones (TAZs). The SCAG dataset provides a comprehensive spatial
distribution of annual VMT on all road segments in LA County for the year 2016. Figure 4-4 shows a subset
of the SCAG data, where each road is shaded according to its annually averaged VMT value.
91

Figure 4-4. This figure shows the SCAG VMT dataset for a subset of Central LA. Each road segment has an assigned VMT
value that is representative of the year 2016. This snapshot shows the total VMT during morning hours (6 am to 9 pm local time).
Diagonal lines represent the aggregated VMT of minor roads within their respective TAZ.
Since the SCAG dataset does not resolve the traffic flow to the hourly resolution necessary for the
final AHF dataset, two additional datasets are used for temporal downscaling. The first is highway traffic
count data collected by Caltrans. The second is arterial traffic count data collected by LA Metro. These
traffic count data were measured using inductive loop detectors that are installed underneath roadways,
usually at major intersections or near highway ramps. Both the highway and arterial traffic count data are
consolidated in a unified database called the Archived Database Management System (ADMS), which is
managed and maintained by the METRANS Transportation Center, which is a collaboration between USC
and CSULB.
In 2016, there were 12,568 active sensors for arterial traffic counts in the City of LA and 4,647 active
sensors for highway traffic counts in LA County. This sums to 17,215 active traffic count sensors scattered
throughout LA County. Each sensor records traffic count approximately every 30 to 60 seconds, providing
92
a fine time resolution of traffic flow. The data for each sensor was aggregated to an hourly resolution,
aggregated by weekend vs. weekday, and then finally averaged over each month. This resulted in a monthly-
averaged diurnal profile of traffic count for an average weekday and weekend, for each individual sensor.
Then, the VMT of each road segment in the SCAG dataset was scaled by the mean-normalized diurnal
profile of the nearest traffic count sensor. By doing this, we downscaled the annual SCAG VMT dataset to
an hourly resolution, for each month of the year.
GIS methods were then used to spatially distribute the VMT from road segments to the appropriate
grid cells of the final gridded dataset. Major road and highway segments were represented as line geometries.
If the road segment crossed multiple grid cells, then the line segments were cut at the intersections of the
line and the grid cells, and the VMT value for the whole segment was distributed proportionally to the size
of each road sub-segment in each cell. For example, if a road segment had a value of 10 VMT and it was
equally split between two different grid cells, each of the two grid cells would be assigned 5 VMT. Since
minor roads were represented as aggregated TAZs, the TAZ polygons were simply spatially resampled to
the final AHF dataset grid. What results after these series of geospatial operations is a gridded dataset of
VMT that varies hourly, for each month of the year.
Once we quantified VMT for each grid cell of AHF
traffic
, we used estimates of fuel efficiency for each
type of vehicle. As a conservative estimate for the prototype AHF
traffic
dataset, we assumed that light- and
medium-duty vehicles (LMDVs) had a fuel efficiency of 25.2 miles per gallon (MTI report, last access
11/30/2022). For heavy-duty vehicles (HDVs), we assumed a fuel efficiency of 5.85 miles per gallon
(NACFE, last access 11/30/2022). Once we calculated the amount of fuel used per each grid cell, energy
conversion values of 120,286 and 137,381 Btu per gallon of fuel were used for gasoline and diesel,
respectively, based on an estimate from the EIA (US Energy Information Administration, last access
7/1/2022). As we did for buildings, we assumed that all energy consumed is converted instantaneously to
sensible heat. Once we assigned hourly energy consumption to each grid cell, we finally converted to AHF
93
with the standard units of W/m
2
. Figures 4-5 shows how daytime versus nighttime AHF
traffic
for the same
test segment of Central LA shown earlier in the AHF
building
section.

Figure 4-5. Panel (a) shows AHF traffic during morning rush hour (8 am) and panel (b) shows AHF traffic during evening non-rush
hour (8 pm). This figure shows AHF for a subset of Central LA as described in section 4.2.1.
4.2.3 Human metabolism
AHF from human metabolism (AHF
metabolism
) is proportional to population density, assuming an
average diurnal profile of human metabolic rates. Although AHF
metabolism
is the smallest contributor to overall
AHF, past studies have shown that it is non-negligible, especially in densely populated areas (Sailor and Lu
2004; Y. Zheng and Weng 2018).
Since commute patterns largely dictate the change in population density throughout the day,
commute-adjusted population density was estimated using data from the US Census Bureau's American
Community Survey (ACS) data, and well as the Census Bureau’s Census Transportation Planning Product
(CTPP) data. The commute-adjusted population was calculated as (Total resident population + Total
workers working in area - Total workers living in area), following the methods suggested by the US Census
Bureau (US Census Bureau, last access 7/1/2022). “Total resident population” (i.e., nighttime population)
and “Total workers living in area” were obtained from the ACS dataset. “Total workers working in area”
was obtained through the CTPP dataset. Each of the datasets were spatially aggregated to the census tract
level. The ACS data for the year 2016 was used, while the CTPP data representative of the year 2006-2010
94
was used. Although a more recent version of the CTPP data for the year 2010-2016 is available, this dataset
had some “missing workers” in the dataset due to geocoding uncertainties. The 2006-2010 CTPP dataset
used data imputation to account for these missing workers, but the 2010-2016 CTPP dataset did not. To
account for the change in overall population between 2010 and 2016, we scaled the number of “workers
working in area” by the percent increase in total population of each census tract between 2010 and 2016.
For the prototype dataset, we assumed that the population remained static from 9 am – 3 pm local
time (i.e., assuming people at work), and from 7 pm – 6 am (i.e., assuming people at their residence).
Between 9 am – 3 pm, it was assumed that the population was equal to the commute-adjusted population
calculated per the method outlined above. Between 7 pm – 6 am, it was assumed that the population was
equal to the resident population. For all the other hours, the population in each census tract was linearly
interpolated between the commute-adjusted population and the resident population, to reflect the changing
population during commute hours.
After calculating the hourly population per census tract, the population was then re-sampled to the
100 m AHF grid. To convert to AHF
metabolism
[W/m2], estimated average metabolic rates from Sailor and Lu
(2004) were used (Sailor and Lu 2004). The metabolic rate profile represents the average human rate of
energy consumption [W] for each hour of the day. Finally, to calculate the final AHFmetabolism, we
multiply the population density [m-2] of each grid cell to the corresponding metabolic rate [J/s] for the hour
of interest. Figure 4-6 shows AHF
metabolism
in the daytime (commute-adjusted population) versus the
nighttime (resident population).
95

Figure 4-6. AHF metabolism during the (a) daytime and (b) nighttime. This figure shows AHF for a subset of Central LA as
described in section 4.2.1.
4.2.4 Regional climate modeling
To assess the regional climate impacts of anthropogenic heat, we used the Weather and Research
Forecasting model (WRF) v4.3 to simulate meteorological fields at a 2 km resolution. WRF is a state-of-the-
science nonhydrostatic mesoscale numerical meteorological model that facilitates physics-based simulations
of physical atmospheric processes (Grell et al. 2005). We also used the urban canopy model (UCM) within
WRF to resolve land-atmosphere exchange of water, momentum, and energy for impervious surfaces in
urban areas (Kusaka et al. 2001; Fei Chen et al. 2011). The UCM parameterizes the effects of urban
geometry on energy fluxes from urban surfaces (i.e., roofs, walls, and roads) and wind profiles within
canyons (Kusaka et al. 2001). The WRF code was also modified to handle the gridded AHF dataset that we
describe above.
4.3 Results and discussion
4.3.1 AHF in Los Angeles
The annual mean AHF for LA County based on our dataset is 2.32 W m
-2
. The annual mean AHF
changes significantly depending on the extent of spatial aggregation before averaging. For example, the
annual mean AHF for urban areas and the City of LA are 6.29 W m
-2
and 7.28 W m
-2
, respectively. This
96
difference is caused by the large range of pixels values and the heterogenous nature of the relevant
geographic features, like roads and buildings. For example, values of individual 100 x 100 m pixels could
exceed 1000 W m
-2
, but large rural swaths of LA County (e.g., the high desert regions that are sparsely
population) have values of ~0 W m
-2
. The spatially averaged mean values are further broken down into the
three major sectors (buildings, traffic, metabolism) below in Table 4-3. For LA County, Urban Areas, and
the City of LA, traffic is the contributing category, making up more than 50% of the total AHF in all three
cases. The Near Highways aggregation includes pixels that were within 1 km of a major highway, in LA
County. For these near-highway pixels, we observe a significantly higher total AHF, as well as a higher
contribution from traffic, as expected. Traffic contributes nearly 70% of the total AHF for near-highway
areas.
Table 4-3. Mean AHF for various geospatial aggregations and by sector categories. The percent contribution of each sector is
also shown for each spatial aggregation. The Near Highways aggregation includes all areas in LA County within 1 km of a major
highway (i.e., freeways).
LA County Urban Areas City of LA Near Highways
Sector AHF mean % of total AHF mean % of total AHF mean % of total AHF mean % of total
Buildings 0.92 40% 2.61 41% 2.95 41% 2.63 28%
Traffic 1.29 56% 3.40 54% 3.98 55% 6.56 69%
Metabolism 0.10 4% 0.28 5% 0.35 5% 0.27 3%
Total 2.32 6.29 7.28 9.47

Figure 4-7 shows heat maps of total AHF for LA County, as well as the AHF from traffic, buildings,
and metabolism, separately. These maps capture important geographical details, such as dense clusters of
large buildings, as well as large roads and highways with significant traffic flow. The distribution of average
population is also represented by AHF
metabolism
. The high resolution of the dataset allows for identification of
hotspots, which would not be possible with a dataset created with more coarse spatial aggregations. Figure
4-8 shows a zoomed in version of the AHF heat map. The extent covers approximately Central LA, West
LA, South Central LA, and parts of the San Fernando Valley in the northern part of the city. At this scale,
97
important geographic clusters emerge, such as hotspots associated with dense buildings and high traffic
roads.

Figure 4-7. Heat maps of AHF within LA County from (a) buildings, (b) metabolism, (c) traffic, and (d) total.
98

Figure 4-8. Zoomed in heat map showing annual mean AHF. There are noticeable clusters of hotspots around dense industrial
and commercial areas, as well as elevated AHF over highways.
As described in section 4.2, temporal variability was infused into the dataset for all three sectors
(buildings, traffic, metabolism) by using an assortment of available energy consumption data. Figures 4-9
and 4-10 show the normalized diurnal and seasonal load profiles (respectively) for the different AHF
categories. Figure 4-9 shows the mean normalized, unitless load profile for all panels, except for the
metabolism profiles (panels (h) and (i)). The metabolism at any given hour was represented as a linear
combination of nighttime metabolism and daytime metabolism. The profiles in Figure 4-9 (h) and (i)
represent the fraction of nighttime and daytime metabolism contributing to the total metabolism. The load
99
profiles for institutional buildings are not shown in Figure 4-9 because we assumed that they would have a
similar load profile to commercial buildings.

Figure 4-9. Unitless diurnal profiles for sub-components of AHF. Panels (a) – (g) illustrate the mean normalized diurnal load
profiles for a typical weekday in August. Panels (h) and (i) shows the fraction of nighttime and daytime metabolism contributing
to total AHF metabolism on a weekday. I.e., AHF metabolim is a linear sum of nighttime and daytime metabolism to capture the diurnal
variability in commute patterns for each neighborhood. Institutional buildings were assumed to follow the same profile as
commercial buildings.
Figure 4-10 shows the mean normalized seasonal load profiles for the various AHF categories. One
notable trend is that natural gas and electricity consumption has opposite trends, for both commercial and
residential sectors. Electricity use peaks during mid-summer, likely due to increased use of air conditioning,
while natural gas use increase in the winter, likely due to increased heating. Even though LA is a moderate
climate, the clearly discernable trend shows that the seasonal extremes even in LA are large enough to
induce consumption behavior changes. Traffic stays relatively flat throughout the year, although summer
traffic is slightly higher, at about 4% above the mean.
100

Figure 4-10. Unitless seasonal profiles for sub-components of AHF, normalized to the annual mean. Industrial (Ind) and
Institutional (Inst) categories are not shown here because the Commercial (Com) seasonal profiles were substituted as an
approximation.
Figure 4-11 shows the diurnal variability in mean AHF, aggregated over various spatial extents.
These profiles represent AHF for an average weekday in August, which is historically the warmest month in
Los Angeles. The temporal variability is important to consider because the daily maximum AHF can be 30-
40% higher than the daily mean AHF. As shown in Table 4-3, we see here once again that areas near
highways have significantly higher AHF, throughout all hours of the day. Here, the daily maximum AHF for
near-highway areas is ~4x the daily maximum for LA County.
101

Figure 4-11. Diurnal profiles of mean total AHF, averaged over LA County (gray), urban areas (yellow), City of LA (purple), and
within 1 km of a major highway (green).
To examine how AHF varies from neighborhood to neighborhood, the annual mean AHF dataset
was spatially joined to the geographical boundaries of LA neighborhoods, and then averaged over each
neighborhood. Figure 4-12 shows the top 20 neighborhoods with the highest average annual mean AHF.
Downtown LA unsurprisingly takes the number one spot with the highest annual mean AHF of 31.30 W m
-
2
. A common pattern among the top 20 neighborhoods is relatively high residential or commercial density,
or proximity to a major highway. For example, Elysian Valley is a small neighborhood in northeast LA, with
modest residential and commercial density, but it directly borders the 101 freeway, which has high traffic
volume throughout the day. One anomalous neighborhood in the top 20 ranking is Universal City, which is
mainly comprised of the theme park, Universal Studios, and its accompanying commercial and recreational
area, Universal Citywalk.
102

Figure 4-12. Annual mean AHF for the top 20 neighborhoods in Los Angeles.
Furthermore, we disaggregated the total AHF to its three main components, for the top 10
neighborhoods. These results are shown in Figure 4-13 in a grouped barplot that shows the relative
proportion of each category to the total AHF in each neighborhood. By breaking down the AHF to its
individual components, we observe a wide variability is relative contributions from traffic, buildings, and
metabolism. For example, Downtown has a relatively even split between AHF from traffic and buildings,
but Elysian Valley has 90% of the total AHF coming from traffic. Overall, seven of the top 10
neighborhoods have traffic as the higher contributor to overall mean AHF. In contrast, Century City, which
is comprised of large commercial buildings and high-rises, have 74% of its AHF coming from buildings.
103

Figure 4-13. Grouped barplot showing the relative contributions of traffic, buildings, and metabolism to the total mean AHF, for
the top 10 neighborhoods.
Figure 4-14 (a) shows the diurnal profiles of mean AHF for all 270 neighborhoods in LA County.
These profiles are representative of an average weekday in August. It is apparent from visual inspection that
there is a large spread between the different neighborhoods in LA county. This can be seen more clearly in
Figures 4-14 (b) and (c), which shows the diurnal mean AHF profiles for the top 5 neighborhoods and
bottom 5 neighborhoods. Note the difference in the y-axis magnitudes between Figures 4-14 (b) and (c).
The AHF in the highest emitting and least emitting neighborhoods vary by approximately two orders of
magnitudes. Figure 4-14 (d) shows the location of the top 5 and bottom 5 neighborhoods on a map of LA
County. The shapes of the markers on the map correspond to the marker shapes in panels (b) and (c). With
geospatial context, it becomes apparent that the top 5 neighborhoods are concentrated in the high-density
104
regions near the city center, while the bottom 5 neighborhoods are scattered throughout the sparsely
populated, rural regions of the high desert.

Figure 4-14. Panel (a) shows the diurnal mean AHF profiles for all 270 neighborhoods in LA County. Each profile represents an
average weekday in August. Red indicates higher AHF and blue indicates lower AHF. Panels (b) and (c) show the diurnal profile
of mean AHF for the top and bottom 5 neighborhoods, respectively. Red
4.3.2 Impacts on meteorology
4.2.2.1 Model Description
As described in section 4.2.4, Weather and Research Forecasting model (WRF) v4.3 was used to
simulate meteorological fields at a 2 km resolution. Figure 4-15 below shows the spatial extents of the three,
two-way nested domains of our WRF simulations. Domains d01, d02, and d03 have grid cell resolutions of
18, 6, and 2 km, respectively. For this memo, four 24-hour simulations were conducted as shown in Figure
4-16. Both a summer weekday (2016-08-01) and a winter weekday (2016-12-19) were run, with AHF turned
on and turned off. The differences (i.e., Δ’s) between the “AHF-on” and “Control” scenarios were
interpreted as the impacts of AHF on the relevant meteorological variables of interest. We analyzed the
105
impacts on 2 m air temperature, planetary boundary layer height, and wind speed, as described further in the
following sections.

Figure 4-15. Nested model domains for all WRF simulations in this study. The inner-most domain is labeled “d03” and has a grid
cell resolution of 2 km.

Figure 4-16. WRF simulation scenarios conducted for this study. AHF-on means that the gridded AHF dataset was used as an
input into the WRF model, and vice versa. Each simulation period was 24 hours.

d01
Summer Weekday
(2016-08-01)
Winter Weekday
(2016-12-19)
AHF-on 1) AH_SW 2) AH_WW
Control (AHF-off) 3) CONTROL_SW 4) CONTROL_WW
106
4.2.2.2 Air Temperature
We observed 2 m air temperature increases at all hours of the day, for both the summer and winter
simulations. Figure 4-17 shows the diurnal profiles of hourly mean 2 m air temperature changes (ΔT
air
) for
both the summer and winter days. The error shading indicates the 95% confidence interval for the mean.
Air temperature values were spatially averaged over the urban grid cells within Los Angeles County. Urban
designations were adopted from the US Census Bureau, based on population density (US Census Bureau,
last access 11/30/22). The same spatial averaging procedure, using the US Census Bureau’s urban
designation, was used for the planetary boundary layer and wind speed analyses as well (see following
sections). Simulations show mean maximum air temperature increases of 0.20 and 0.37 °C for summer and
winter, respectively. The summer and winter distributions of ΔT
air
are shown in Figure 4-18. We observe a
clear right skewness for both winter and summer distributions, confirming that ΔT
air
increased on average,
over urban areas. Figure 4-18 also shows that some pixels within the WRF domain can exceed 1 °C,
particularly for the winter distribution.

Figure 4-17. Diurnal plot showing hourly, mean changes in 2 m air temperatures (ΔT air) for both summer (red) and winter (blue)
days. The mean is spatially representative of urban areas within Los Angeles County. The error shading for both lines indicates
the 95% confidence interval for the mean.
107

Figure 4-18. Histograms showing the distributions of ΔT air for both summer (red) and winter (blue). Each histogram is
normalized such that the area under the curve is equal to one.

Figure 4-19. Heatmaps of ΔT air from both the summer and winter simulations. Red indicates an increase, while blue indicates a
decrease. Each plot shows the spatial distribution of ΔT air for the hours of maximum mean ΔT air, which were 08:00 and 10:00
LST for summer and winter, respectively.
Figure 4-19 shows the spatial distribution of ΔT
air
at the hours of maximum mean ΔT
air
, for both the
summer and winter days. Red shading indicates an increase in air temperature due to AHF, while blue
shading indicates a cooling. The higher maximum mean ΔT
air
for winter is visually represented here in
Figure 4-19 by the darker shade of red in most of the visible spatial extent. Another notable feature apparent
from Figure 4-19 is the heterogenous spatial distribution of ΔT
air
, which does not always correlate with the
108
actual distribution of AHF. This is because the atmospheric conditions strongly dictate how sensible heat
will be distributed throughout the LA Basin. For example, atmospheric convection spreads the sensible heat
in areas away from the actual sources of heat. An examination of ΔT
air
throughout a full 24-hour diurnal
cycle shows that regional-scale convection indeed carries sensible heat into neighboring areas outside of LA
County. In essence, AHF from LA County does not stay neatly confined within the boundaries of AHF
emissions. In fact, temperatures may increase downwind of the AHF sources. Another notable observation
is the qualitative difference in the spatial distribution of ΔT
air
between the summer and winter days shown in
Figure 4-19. In addition to the time-of-day difference between the two sub-panels in Figure 4-19, other
atmospheric conditions, such as the dominant wind direction and planetary boundary layer height may also
play a key role in determining the distribution and magnitude of air temperature impacts.
4.2.2.3 Planetary boundary layer height (PBLH)
The planetary boundary layer height (PBLH) is the height of the lowest vertical segment of the
atmosphere. The PBLH is dictated by the vertical temperature profile of the atmosphere, and it changes
throughout the day, largely influenced by the warming of the land surface due to solar radiation. In short, a
higher PBLH results in a larger volume of the PBL, and vice versa. A practical implication of the PBLH is
its impact on concentrations of relevant meteorological scalar values, such as heat and pollution. For
example, given a constant rate of pollutant emissions from the land surface, a lower PBLH will result in
higher concentrations, assuming no chemistry impacts. Similarly, a lower PBLH can also result in larger air
temperature increases, given an equal rate of sensible heat flux emissions at the surface.
Figure 4-20 shows the diurnal profile of the mean changes in PBLH due to AHF, for both a
summer and winter day. Over urban areas, we observed maximum mean ΔPBLH of ~10 m and ~36 m for
summer and winter, respectively. There is also a clear seasonal influence on ΔPBLH, evident by the
different peak locations for summer versus winter. The summer mean ΔPBLH peaks in the morning at ~8
am LST, while the winter mean ΔPBLH peaks in the late afternoon, at ~4:00 pm LST.
109

Figure 4-20. Diurnal plot showing hourly, mean changes in planetary boundary layer height (ΔPBLH) for both summer (red) and
winter (blue) days. The mean ΔPBLH is spatially representative of urban areas within Los Angeles County. The error shading for
both lines indicates the 95% confidence interval for the mean.
The distributions of summer and winter ΔPBLH are shown in Figure 4-21. There is a modest, but
visibly noticeable shift in both summer and winter distributions to the right of zero, indicating an average
increase over the whole day. The larger increase in the winter can be explained by the temperature
dependence of the atmosphere’s heat capacity, as well as the differences in baseline temperatures. Colder air
has a lower heat capacity than warmer air, which means that for a given unit of heat transferred to the air,
the colder air’s temperature will increase more than the corresponding warmer air. This explains the larger
values of ΔT
air
we observed in Figure 4-17, as well as the larger peak increases in PBLH shown in Figure 4-
20.
Figure 4-22 shows the corresponding spatial distributions of ΔPBLH at the hours of maximum
mean ΔPBLH. The summer panel of Figure 4-22 shows lower magnitudes of ΔPBLH, as expected from
Figure 4-20. The spatial distributions of ΔPBLH hotspots are qualitatively different, between the summer
and winter day. As mentioned above in the previous section, various atmospheric conditions likely drive this
difference. Aside from the difference in the hour of the day, the overall synoptic winds were different
between the summer and winter day. On an average summer day, LA experiences a land-sea breeze, where a
110
high-pressure system off the coast drives winds from over the Pacific, towards the low-pressure system
within the LA Basin. Conversely, when high-pressure systems develop in the high desert regions of Nevada
and Utah, Santa Ana winds blow in the opposite direction, from the north and/or east towards the LA
Basin. According to historical weather data, the winter day we simulated had Santa Ana winds, with winds
blowing from the north and east. This insight highlights that regional scale wind patterns can play an
important role in the ultimate distribution of sensible heat and its corresponding impacts on other
meteorological variables like PBLH.

Figure 4-21. Histograms showing the distributions of ΔPBLH for both summer (red) and winter (blue). Each histogram is
normalized such that the area under the curve is equal to one.
111

Figure 4-22. Heatmaps of ΔPBLH from both the summer and winter simulations. Red indicates an increase, while blue indicates
a decrease. Each plot shows the spatial distribution of ΔPBLH for the hours of maximum mean ΔPBLH, which were 08:00 and
16:00 LST for summer and winter, respectively.
4.2.2.4 Wind speed
We also examined the impacts of AHF on wind speed. Figure 4-23 shows the diurnal profiles of
mean Δ(wind speed) for urban areas of LA County, for both summer and winter. On average, the summer
day had larger increases in wind speed, while the winter day showed more modest changes throughout the
day. We saw peaks of ~0.13 and ~0.07 m s
-1
for mean Δ(wind speed), for summer and winter, respectively.
The summer day also saw increases in wind speed throughout all hours of the day, while the winter day
showed increases in the mean at some hours and decreases at other hours.
112

Figure 4-23. Diurnal plot showing hourly, mean changes in wind speed (Δ(Wind Speed)) for both summer (red) and winter (blue)
days. The mean is spatially representative of urban areas within Los Angeles County. The error shading for both lines indicates
the 95% confidence interval for the mean.
The histograms of wind speed changes (Figure 4-24) accentuate the differences between the summer
and winter day. The winter distribution is nearly centered at zero, with the area under the curve
approximately equal on both sides of zero. In contrast, the summer distribution displays a clear shift right,
indicating a systematic increase in average daily wind speed. Figure 4-25 shows the spatial distribution of
wind speed changes for both the summer day and winter day. The summer day in Figure 4-25 shows a more
consistent increase throughout the spatial extent, while the winter day exhibits pockets of red and some
pockets of blue, resulting in a lower spatially averaged Δ(wind speed).
As noted in the previous section, the differences in synoptic wind conditions (i.e., land-sea breeze
vs. Santa Ana winds) between the summer and winter day likely play a large role in overall impacts on wind
speed. In future work, winter days without Santa Ana conditions should be simulated to control for the
differences in synoptic wind patterns. Nevertheless, the comparison shown in figure 4-25 highlights the
importance of considering all potential impacts of atmospheric conditions that may play a role in driving
seasonal differences.
113

Figure 4-24. Histograms showing the distributions of Δ(Wind Speed) for both summer (red) and winter (blue). Each histogram is
normalized such that the area under the curve is equal to one.

Figure 4-25. Heatmaps of Δ(Wind Speed) from both the summer and winter simulations. Red indicates an increase, while blue
indicates a decrease. Each plot shows the spatial distribution of Δ(Wind Speed) for the hours of maximum mean Δ(Wind Speed),
which were 00:00 and 10:00 LST for summer and winter, respectively.
4.4 Conclusion
In this final report, we outline the framework and methods used to create a high-resolution, gridded
AHF dataset for LA County. By incorporating GIS methods and auxiliary datasets into our inventory
114
approach, we improve the spatial and temporal representation of AHF in LA. We also show WRF modeling
results, utilizing the gridded AHF dataset that we created.
We found that the spatial distribution of AHF in LA County is highly heterogenous, and the
magnitude of AHF varies widely between different neighborhoods. The mean AHF for LA County peaks at
~3 W m
-2
for a summer weekday, but we found that the choice of spatial aggregation can dramatically
influence the mean AHF values. For example, the City of LA has a much higher maximum mean AHF of
~10 W m
-2
, which is more than 3x greater than the maximum mean AHF of the whole county.
Furthermore, we found that the maximum mean AHF for pixels within 1 km of a major highway exceeded
13 W m
-2
, which is more than 4x greater than the county-wide average. When aggregated to neighborhoods,
we found that mean AHF varied more than two orders of magnitude, with the top five neighborhoods
exceeding 40 W m
-2
and the bottom five neighborhoods staying under 0.15 W m
-2
. Individual pixels were
found to exceed 2,000 W m
-2
in the most extreme cases.
WRF simulations confirmed non-negligible impacts on the urban climate of Los Angeles. The
impacts of AHF on 2 m air temperature, planetary boundary layer height, and wind speed were quantified.
We found that AHF generally increased air temperature, planetary boundary layer height, and wind speeds,
though the magnitudes of change varied throughout the day and by season. Mean T
air
increased up to 0.2
and 0.4 °C for summer and winter, respectively. Mean PBL increased up to 10 and 40 m for summer and
winter, respectively. Finally, mean wind speed increased up for 0.14 m s
-1
for summer, while the winter
impacts on wind speed were more ambiguous.

The impacts of AHF also have secondary implications on air quality, which were not fully
considered within the scope of this project. Since meteorological variables such as air temperature, PBLH,
and wind conditions play such a crucial role in the dynamics of air pollution formation, transport, and
115
exposure, we recommend that this secondary impact of AHF on air quality should be studied in more detail
in future research.
Moving forward, we are currently working to do final quality assurance tests on the AHF dataset.
Upon full validation, we plan to upload the full dataset to a publicly accessible data repository, which will
also be shared with LADWP upon completion. Furthermore, we plan to evaluate the impact of future
potential energy scenarios (e.g., LA100 scenarios) on AHF, and the subsequent impacts on the climate of
LA. Since the electrification of buildings and vehicles will inevitably impact energy consumption at the local
level, this will in turn impact AHF from these various sources. We believe that quantifying the impacts of
future electrification and efficiency on AHF would be useful for policymakers and public agencies. This
continuing part of the study will help confirm whether future energy use scenarios will also have non-
negligible co-benefits in reducing negative impacts of the urban heat island, in addition to air pollution
emission reductions that have already been quantified (NREL 2021, last access 7/1/2022).
4.5 Funding and support
This research was supported by the Los Angeles Department of Water and Power.

116
Conclusion
Climate change is one of the most pressing issues of the anthropogenic age. There is no question
that greenhouse gases have already caused irreversible warming around the world. In the context of a
warming planet, I show in this dissertation that in addition to greenhouse gases, it is also crucial to examine
other significant sources of local and regional radiative forcing, whose effects may be compounded in the
context of global climate change and increasing urbanization around the world. This dissertation focuses on
two significant, non-GHG drivers of local and regional radiative forcing: black carbon (BC) aerosols and the
urban heat island (UHI) effect.
Chapter 2 describes how in-situ, field measurements campaigns were utilized to constrain the
microphysical properties of BC aerosols from both wildfires and urban emissions. BC is often cited as the
second strongest radiative forcing pollutant in the atmosphere, after CO
2
. However, due to the
heterogenous nature of BC spatial distribution, as well as its varying microphysical and chemical properties,
large uncertainties remain about its radiative forcing. This leads to crude approximations and
parameterizations of BC in climate models. Our study presented in Chapter 2 contributes to constrain the
uncertainties associated with BC by providing limited real-world measurements of BC microphysical
properties, such as its internal mixing state and core size distributions. We also describe how BC mixing
state varies as a function of aging timescales and emissions sources, which is important to consider when
improving aerosol parameterizations in climate models.
Chapter 3 takes a shift to the neighborhood scale, where we measured the real-world impacts of
solar reflective cool pavements using a host of micrometeorological field measurements in a Los Angeles
neighborhood. Here we present the first statistically significant air temperature reductions due to cool
pavements, observed using controlled field measurements. We also highlight that there are potential pitfalls
associated with cool pavements that need to be considered more carefully before widespread adoption in
urban areas. For example, the albedo of cool pavements we measured were found to degrade by more than
117
30% in the time span of less than one year. The thermal comfort tradeoffs between increased shortwave
radiation above the pavement versus decreased longwave radiation and air temperatures, must also be
carefully considered on a neighborhood-by-neighborhood basis.
Finally, in Chapter 4, we present a high spatiotemporal resolution dataset of AHF in LA County that
was developed using a wide variety of energy consumption data and geospatial information related to the
built environment. Our estimates constrain the estimates that were done in the past with slightly different
methods and provide highly resolved spatial information such that AHF hotspots can be identified and
potentially diagnosed. Although AHF averaged over LA County was found to be less than 3 W m
-2
, we
found huge variability amongst the 100 x 100 m pixels, as well as between different neighborhoods in Los
Angeles. For example, AHF above highways well exceed 100 W m
-2
on average, and some pixels were
found to exceed 1000 W m
-2
. We were also able to rank the highest AHF emitting neighborhoods, and
attribute the relative contributions of traffic, buildings, and metabolism to the total AHF in the various
neighborhoods of LA. Furthermore, preliminary WRF modeling results with parameterized AHF values
show that 2 m air temperatures can increase up to 1 °C for large portions of LA County. This serves as only
the second study that modeled the impacts of total AHF on air temperatures in the US, and the first for Los
Angeles specifically.
In this dissertation, we constrained the properties of BC and the UHI effect, explored the impacts of
local radiative forcing, and investigated the efficacy of proposed mitigation strategies. Along with global
mitigation strategies to limit the impacts of greenhouse gases on the global climate, it is also important to
consider steps we can take to minimize the negative impacts of local and regional radiative forcing agents,
such as black carbon and the urban heat island. As this dissertation shows, there is a complex coupling
between air pollution, the built urban environment, and regional to local climate. Understanding each
component and their inter-connections are more important than now than ever before, as we seek to adapt
to a warming and urbanizing future.
118
References
Adachi, Kouji, Serena H. Chung, and Peter R. Buseck. 2010. “Shapes of Soot Aerosol Particles and
Implications for Their Effects on Climate.” Journal of Geophysical Research Atmospheres 115 (15): 1–9.
https://doi.org/10.1029/2009JD012868.
Adachi, Kouji, Nobuhiro Moteki, Yutaka Kondo, and Yasuhito Igarashi. 2016. “Mixing States of Light-
Absorbing Particles Measured Using a Transmission Electron Microscope and a Single-Particle Soot
Photometer in Tokyo, Japan.” Journal of Geophysical Research 121 (15): 9153–64.
https://doi.org/10.1002/2016JD025153.
Akagi, S. K., J. S. Craven, J. W. Taylor, G. R. McMeeking, R. J. Yokelson, I. R. Burling, S. P. Urbanski, et al.
2012. “Evolution of Trace Gases and Particles Emitted by a Chaparral Fire in California.” Atmospheric
Chemistry and Physics 12 (3): 1397–1421. https://doi.org/10.5194/acp-12-1397-2012.
Akbari, H., M. Pomerantz, and H. Taha. 2001. “Cool Surfaces and Shade Trees to Reduce Energy Use and
Improve Air Quality in Urban Areas.” Solar Energy 70 (3): 295–310. https://doi.org/10.1016/S0038-
092X(00)00089-X.
Akbari, Hashem, Constantinos Cartalis, Denia Kolokotsa, Alberto Muscio, Anna Laura Pisello, Federico
Rossi, Matheos Santamouris, Afroditi Synnefa, Nyuk Hien Wong, and Michele Zinzi. 2016. “Local
Climate Change and Urban Heat Island Mitigation Techniques - The State of the Art.” Journal of Civil
Engineering and Management 22 (1): 1–16. https://doi.org/10.3846/13923730.2015.1111934.
Akbari, Hashem, and H. Damon Matthews. 2012. “Global Cooling Updates: Reflective Roofs and
Pavements.” Energy and Buildings 55: 2–6. https://doi.org/10.1016/j.enbuild.2012.02.055.
Akbari, Hashem, Surabi Menon, and Arthur Rosenfeld. 2009. “Global Cooling: Increasing World-Wide
Urban Albedos to Offset CO 2.” Climatic Change 94 (3–4): 275–86. https://doi.org/10.1007/s10584-
008-9515-9.
Alchapar, N. L., E. N. Correa, and M. A. Cantón. 2013. “Solar Reflectance Index of Pedestrian Pavements
and Their Response to Aging.” Journal of Clean Energy Technologies 1 (4): 281–85.
https://doi.org/10.7763/jocet.2013.v1.64.
Anting, Nickholas, Mohd Fadhil Md Din, Kenzo Iwao, Mohanadoss Ponraj, Alvin John Lim Meng Siang,
Lee Yee Yong, and Joewono Prasetijo. 2018. “Optimizing of near Infrared Region Reflectance of Mix-
Waste Tile Aggregate as Coating Material for Cool Pavement with Surface Temperature
Measurement.” Energy and Buildings 158: 172–80. https://doi.org/10.1016/j.enbuild.2017.10.001.
ASHRAE. 1993. “Physiological Principles and Thermal Comfort.” In ASHRAE Handbook of Fundamentals,
8.1-8.29. Atlanta.
Bartesaghi-Koc, Carlos, Shamila Haddad, Gloria Pignatta, Riccardo Paolini, Deo Prasad, and Mattheos
Santamouris. 2021. “Can Urban Heat Be Mitigated in a Single Urban Street? Monitoring, Strategies,
and Performance Results from a Real Scale Redevelopment Project.” Solar Energy 216 (September
2020): 564–88. https://doi.org/10.1016/j.solener.2020.12.043.
119
Battista, Gabriele, Emanuele de Lieto Vollaro, and Roberto de Lieto Vollaro. 2021. “How Cool Pavements
and Green Roof Affect Building Energy Performances.” Heat Transfer Engineering 0 (0): 1–15.
https://doi.org/10.1080/01457632.2021.1874667.
Baumgardner, D., G. L. Kok, and G. B. Raga. 2007. “On the Diurnal Variability of Particle Properties
Related to Light Absorbing Carbon in Mexico City.” Atmospheric Chemistry and Physics 7 (10): 2517–26.
https://doi.org/10.5194/acp-7-2517-2007.
Beccario, Cameron. n.d. “Earth Nullschool.” Accessed December 13, 2019. earth.nullschool.net.
Block, Alexander, Klaus Keuler, and Eberhard Schaller. 2004. “Impacts of Anthropogenic Heat on Regional
Climate Patterns.” Geophysical Research Letters 31 (12): 2–5. https://doi.org/10.1029/2004GL019852.
Bond, Tami C., and Robert W. Bergstrom. 2006. “Light Absorption by Carbonaceous Particles: An
Investigative Review.” Aerosol Science and Technology. https://doi.org/10.1080/02786820500421521.
Bond, Tami C., S. J. Doherty, D. W. Fahey, P. M. Forster, T. Berntsen, B. J. Deangelo, M. G. Flanner, et al.
2013. “Bounding the Role of Black Carbon in the Climate System: A Scientific Assessment.” Journal of
Geophysical Research Atmospheres 118 (11). https://doi.org/10.1002/jgrd.50171.
Branas, Charles C., Rose A. Cheney, John M. MacDonald, Vicky W. Tam, Tara D. Jackson, and Thomas R.
Ten Havey. 2011. “A Difference-in-Differences Analysis of Health, Safety, and Greening Vacant
Urban Space.” American Journal of Epidemiology 174 (11): 1296–1306.
https://doi.org/10.1093/aje/kwr273.
Buerger, Christian, and Robert Bifulco. 2019. “The Effect of Charter Schools on Districts’ Student
Composition, Costs, and Efficiency: The Case of New York State.” Economics of Education Review 69
(July 2016): 61–72. https://doi.org/10.1016/j.econedurev.2019.01.003.
Burillo, Daniel, Mikhail V. Chester, Stephanie Pincetl, Eric D. Fournier, and Janet Reyna. 2019. “Forecasting
Peak Electricity Demand for Los Angeles Considering Higher Air Temperatures Due to Climate
Change.” Applied Energy 236 (November 2018): 1–9. https://doi.org/10.1016/j.apenergy.2018.11.039.
Cao, Xuejuan, Boming Tang, Ying Yuan, and Xiaoling Zou. 2016. “Indoor and Outdoor Aging Behaviors
of a Heat-Reflective Coating for Pavement in the Chongqing Area.” Journal of Materials in Civil
Engineering 28 (1): 04015079. https://doi.org/10.1061/(asce)mt.1943-5533.0001347.
Cao, Zheng, Zhifeng Wu, Lin Liu, Yinbiao Chen, and Yuxuan Zou. 2019. “Assessing the Relationship
between Anthropogenic Heat Release Warming and Building Characteristics in Guangzhou: A
Sustainable Development Perspective.” Science of the Total Environment 695: 133759.
https://doi.org/10.1016/j.scitotenv.2019.133759.
Cappa, C. D., T. B. Onasch, P. Massoli, D. R. Worsnop, T. S. Bates, E. S. Cross, P. Davidovits, et al. 2012.
“Radiative Absorption Enhancements Due to the Mixing State of Atmospheric Black Carbon.” Science
337 (6098): 1078–81. https://doi.org/10.1126/science.1223447.
Card, David, and Alan B. Krueger. 1994. “Minimum Wages and Employment: A Case Study of the Fast-
Food Industry in New Jersey and Pennsylvania.” American Economic Review 84 (4): 772–93.
120
Casey, Joan A., Deborah Karasek, Elizabeth L. Ogburn, Dana E. Goin, Kristina Dang, Paula A. Braveman,
and Rachel Morello-Frosch. 2018. “Retirements of Coal and Oil Power Plants in California:
Association with Reduced Preterm Birth among Populations Nearby.” American Journal of Epidemiology
187 (8): 1586–94. https://doi.org/10.1093/aje/kwy110.
CBS Los Angeles. 2017a. “Danielle Gersh’s Weather Forecast (Dec. 20, 2017).” 2017.
https://youtu.be/p8dEJ-qoPwA.
———. 2017b. “Danielle Gersh’s Weather Forecast (Dec. 21, 2017).” 2017.
https://youtu.be/j_0P09YZuKc.
———. 2017c. “Evelyn Taft’s Weather Forecast (Dec. 20, 2017).” 2017.
https://youtu.be/gDH73gLbGDk.
———. 2017d. “Evelyn Taft’s Weather Forecast (Dec. 22, 2017).” 2017. https://youtu.be/I7KkZiduQq4.
———. 2017e. “Garth Kemp’s Weather Forecast (Dec. 21, 2017).” 2017.
https://youtu.be/vpv65O1MP0A.
———. 2017f. “Markina Brown’s Weather Forecast (Dec. 22, 2017).” 2017.
https://youtu.be/N7htXoFLVcw.
Chen, Fei, Hiroyuki Kusaka, Robert Bornstein, Jason Ching, C. S.B. Grimmond, Susanne Grossman-Clarke,
Thomas Loridan, et al. 2011. “The Integrated WRF/Urban Modelling System: Development,
Evaluation, and Applications to Urban Environmental Problems.” International Journal of Climatology 31
(2): 273–88. https://doi.org/10.1002/joc.2158.
Chen, Feng, Xuchao Yang, and Weiping Zhu. 2014. “WRF Simulations of Urban Heat Island under Hot-
Weather Synoptic Conditions: The Case Study of Hangzhou City, China.” Atmospheric Research 138:
364–77. https://doi.org/10.1016/j.atmosres.2013.12.005.
Chen, Liang, and Oliver W. Frauenfeld. 2016. “Impacts of Urbanization on Future Climate in China.”
Climate Dynamics 47 (1–2): 345–57. https://doi.org/10.1007/s00382-015-2840-6.
Chen, Mo, George A Ban-Weiss, and Kelly T Sanders. 2018. “The Role of Household Level Electricity Data
in Improving Estimates of the Impacts of Climate on Building Electricity Use.” Energy and Buildings
180: 146–58.
———. 2020. “Utilizing Smart-Meter Data to Project Impacts of Urban Warming on Residential Electricity
Use for Vulnerable Populations in Southern California.” Environmental Research Letters 15 (6): 64001.
Chen, Mo, Kelly T Sanders, and George A Ban-Weiss. 2019. “A New Method Utilizing Smart Meter Data
for Identifying the Existence of Air Conditioning in Residential Homes.” Environmental Research Letters
14 (9): 94004.
Cheng, Yuan, Shao Meng Li, Mark Gordon, and Peter Liu. 2018. “Size Distribution and Coating Thickness
of Black Carbon from the Canadian Oil Sands Operations.” Atmospheric Chemistry and Physics 18 (4):
2653–67. https://doi.org/10.5194/acp-18-2653-2018.
121
Chow, Winston T.L., Francisco Salamanca, Matei Georgescu, Alex Mahalov, Jeffrey M. Milne, and
Benjamin L. Ruddell. 2014. “A Multi-Method and Multi-Scale Approach for Estimating City-Wide
Anthropogenic Heat Fluxes.” Atmospheric Environment 99: 64–76.
https://doi.org/10.1016/j.atmosenv.2014.09.053.
Clark, E. C., R. D. Bornstein, and Y. T. Tam. 1985. “Current and Potential Anthropogenic Moisture Effects
on the New York City Planetary Boundary Layer.” Journal of the Air Pollution Control Association 35 (8).
https://doi.org/10.1080/00022470.1985.10465963.
Croce, Silvia, Elisa D’Agnolo, Mauro Caini, and Rossana Paparella. 2021. “The Use of Cool Pavements for
the Regeneration of Industrial Districts.” Sustainability (Switzerland) 13 (11).
https://doi.org/10.3390/su13116322.
Dahlkötter, F., M. Gysel, D. Sauer, A. Minikin, R. Baumann, P. Seifert, A. Ansmann, M. Fromm, C. Voigt,
and B. Weinzierl. 2014. “The Pagami Creek Smoke Plume after Long-Range Transport to the Upper
Troposphere over Europe – Aerosol Properties and Black Carbon Mixing State.” Atmospheric
Chemistry and Physics 14 (12): 6111–37. https://doi.org/10.5194/acp-14-6111-2014.
Dear, Richard J. De, Edward Arens, Zhang Hui, and Masayuki Oguro. 1997. “Convective and Radiative
Heat Transfer Coefficients for Individual Human Body Segments.” International Journal of Biometeorology
40 (3): 141–56. https://doi.org/10.1007/s004840050035.
Dempsey, Judith A., and Andrew J. Plantinga. 2013. “How Well Do Urban Growth Boundaries Contain
Development? Results for Oregon Using a Difference-in-Difference Estimator.” Regional Science and
Urban Economics 43 (6): 996–1007. https://doi.org/10.1016/j.regsciurbeco.2013.10.002.
Doan, Van Q., Hiroyuki Kusaka, and Truong M. Nguyen. 2019. “Roles of Past, Present, and Future Land
Use and Anthropogenic Heat Release Changes on Urban Heat Island Effects in Hanoi, Vietnam:
Numerical Experiments with a Regional Climate Model.” Sustainable Cities and Society 47 (January):
101479. https://doi.org/10.1016/j.scs.2019.101479.
Dong, Feng, Yuanju Dai, Shengnan Zhang, Xiaoyun Zhang, and Ruyin Long. 2019. “Can a Carbon
Emission Trading Scheme Generate the Porter Effect? Evidence from Pilot Areas in China.” Science of
the Total Environment 653: 565–77. https://doi.org/10.1016/j.scitotenv.2018.10.395.
EPA. 2020. “Heat Island Community Actions Database.” 2020. https://www.epa.gov/heatislands/heat-
island-community-actions-database.
Fan, Hongli, and David J. Sailor. 2005. “Modeling the Impacts of Anthropogenic Heating on the Urban
Climate of Philadelphia: A Comparison of Implementations in Two PBL Schemes.” Atmospheric
Environment 39 (1): 73–84. https://doi.org/10.1016/j.atmosenv.2004.09.031.
Fast, Jerome D., Joel C. Torcolini, and Randy Redman. 2005. “Pseudovertical Temperature Profiles and the
Urban Heat Island Measured by a Temperature Datalogger Network in Phoenix, Arizona.” Journal of
Applied Meteorology 44 (1): 3–13. https://doi.org/10.1175/JAM-2176.1.
Feng, Jin Ming, Yong Li Wang, and Zhu Guo Ma. 2015. “Long-Term Simulation of Large-Scale
Urbanization Effect on the East Asian Monsoon.” Climatic Change 129 (3–4): 511–23.
https://doi.org/10.1007/s10584-013-0885-2.
122
Feng, Jin Ming, Yong Li Wang, Zhu Guo Ma, and Yong He Liu. 2012. “Simulating the Regional Impacts of
Urbanization and Anthropogenic Heat Release on Climate across China.” Journal of Climate 25 (20):
7187–7203. https://doi.org/10.1175/JCLI-D-11-00333.1.
Fierce, Laura, Timothy B. Onasch, Christopher D. Cappa, Claudio Mazzoleni, Swarup China, Janarjan
Bhandari, Paul Davidovits, et al. 2020. “Radiative Absorption Enhancements by Black Carbon
Controlled by Particle-to-Particle Heterogeneity in Composition.” Proceedings of the National Academy of
Sciences of the United States of America 117 (10): 5196–5203. https://doi.org/10.1073/pnas.1919723117.
Fortuniak, K., K. Kłysik, and J. Wibig. 2006. “Urban - Rural Contrasts of Meteorological Parameters in
Łódź.” Theoretical and Applied Climatology 84 (1–3): 91–101. https://doi.org/10.1007/s00704-005-0147-y.
Gao, R. S., J. P. Schwarz, K. K. Kelly, D. W. Fahey, L. A. Watts, T. L. Thompson, J. R. Spackman, et al.
2007. “A Novel Method for Estimating Light-Scattering Properties of Soot Aerosols Using a Modified
Single-Particle Soot Photometer.” Aerosol Science and Technology 41 (2): 125–35.
https://doi.org/10.1080/02786820601118398.
Garcetti, Eric. 2019. “Los Angeles Mayor Garcetti Kicks off ‘cool Streets LA’ (2019, Oct 22).” Targeted
News Service. http://libproxy.usc.edu/login?url=https://www.proquest.com/wire-feeds/los-angeles-
mayor-garcetti-kicks-off-cool-streets/docview/2307707015/se-2?accountid=14749.
Garofalo, Lauren A., Matson A. Pothier, Ezra J.T. Levin, Teresa Campos, Sonia M. Kreidenweis, and
Delphine K. Farmer. 2019. “Emission and Evolution of Submicron Organic Aerosol in Smoke from
Wildfires in the Western United States.” Research-article. ACS Earth and Space Chemistry 3 (7): 1237–47.
https://doi.org/10.1021/acsearthspacechem.9b00125.
Georgescu, Matei, Philip E. Morefield, Britta G. Bierwagen, and Christopher P. Weaver. 2014. “Urban
Adaptation Can Roll Back Warming of Emerging Megapolitan Regions.” Proceedings of the National
Academy of Sciences of the United States of America 111 (8): 2909–14.
https://doi.org/10.1073/pnas.1322280111.
Gertler, Paul J., Sebastian Martinez, Patrick Premand, Laura B. Rawlings, and Christel M. J. Vermeersch.
2016. “Impact Evaluation in Practice, Second Edition. 2nd Ed.” Washington D.C.: World Bank
Publications.
Giovannini, Lorenzo, Dino Zardi, Massimiliano De Franceschi, and Fei Chen. 2014. “Numerical
Simulations of Boundary-Layer Processes and Urban-Induced Alterations in an Alpine Valley.”
International Journal of Climatology 34 (4): 1111–31. https://doi.org/10.1002/joc.3750.
Grell, Georg A., Steven E. Peckham, Rainer Schmitz, Stuart A. McKeen, Gregory Frost, William C.
Skamarock, and Brian Eder. 2005. “Fully Coupled ‘Online’ Chemistry within the WRF Model.”
Atmospheric Environment 39 (37). https://doi.org/10.1016/j.atmosenv.2005.04.027.
Hamilton, Ian G., Michael Davies, Philip Steadman, Andrew Stone, Ian Ridley, and Stephen Evans. 2009.
“The Significance of the Anthropogenic Heat Emissions of London’s Buildings: A Comparison against
Captured Shortwave Solar Radiation.” Building and Environment 44 (4): 807–17.
https://doi.org/10.1016/j.buildenv.2008.05.024.
Harrison, R., B. McGoldrick, and C. G.B. Williams. 1984. “Artificial Heat Release from Greater London,
123
1971-1976.” Atmospheric Environment (1967) 18 (11). https://doi.org/10.1016/0004-6981(84)90001-5.
Hassid, S., M. Santamouris, N. Papanikolaou, A. Linardi, N. Klitsikas, C. Georgakis, and D. N.
Assimakopoulos. 2000. “Effect of the Athens Heat Island on Air Conditioning Load.” Energy and
Buildings 32 (2): 131–41. https://doi.org/10.1016/S0378-7788(99)00045-6.
He, C., K. N. Liou, Y. Takano, R. Zhang, M. Levy Zamora, P. Yang, Q. Li, and L. R. Leung. 2015.
“Variation of the Radiative Properties during Black Carbon Aging: Theoretical and Experimental
Intercomparison.” Atmospheric Chemistry and Physics 15 (20). https://doi.org/10.5194/acp-15-11967-
2015.
Henneman, Lucas R.F., Christine Choirat, and Corwin M. Zigler. 2019. “Accountability Assessment of
Health Improvements in the United States Associated with Reduced Coal Emissions between 2005 and
2012.” Epidemiology 30 (4): 477–85. https://doi.org/10.1097/EDE.0000000000001024.
Howard, Luke. 1833. The Climate of London: Deduced from Meteorological Observations Made in the Metropolis and at
Various Places around It. Vol. 3. Harvey and Darton, J. and A. Arch, Longman, Hatchard, S. Highley
[and] R. Hunter.
Hughes, L. S., J. O. Allen, P. Bhave, M. J. Kleeman, G. R. Cass, D. Y. Liu, D. P. Fergenson, B. D. Morrical,
and K. A. Prather. 2000. “Evolution of Atmospheric Particles along Trajectories Crossing the Los
Angeles Basin.” Environmental Science and Technology 34 (15): 3058–68.
https://doi.org/10.1021/es9908671.
Ichinose, Toshiaki, Kazuhiro Shimodozono, and Keisuke Hanaki. 1999. “Impact of Anthropogenic Heat on
Urban Climate in Tokyo.” Atmospheric Environment 33 (V): 23. papers3://publication/uuid/D904B94A-
7096-4D0A-8327-C8CC022D10D0.
IPCC. 2021. “Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the
Sixth Assessment Report of the Intergovernmental Panel on Climate Change.”
Jacobson, Mark Z., and John E. Ten Hoeve. 2012. “Effects of Urban Surfaces and White Roofs on Global
and Regional Climate.” Journal of Climate 25 (3): 1028–44. https://doi.org/10.1175/JCLI-D-11-00032.1.
Jaross, G. 2017. “OMPS-NPP L2 NM Ozone (O3) Total Column Swath Orbital V2.” 2017.
doi:10.5067/0WF4HAAZ0VHK.
Jiang, Yingjun, Changqing Deng, Zhejiang Chen, and Yuhang Tian. 2020. “Evaluation of the Cooling Effect
and Anti-Rutting Performance of Thermally Resistant and Heat-Reflective Pavement.” International
Journal of Pavement Engineering 21 (4): 447–56. https://doi.org/10.1080/10298436.2018.1483506.
Johnson, G. R., Z. D. Ristovski, B. D’Anna, and L. Morawska. 2005. “Hygroscopic Behavior of Partially
Volatilized Coastal Marine Aerosols Using the Volatilization and Humidification Tandem Differential
Mobility Analyzer Technique.” Journal of Geophysical Research D: Atmospheres 110 (20): 1–14.
https://doi.org/10.1029/2004JD005657.
Justice, C. O., L. Giglio, S. Korontzi, J. Owens, J. T. Morisette, D. Roy, J. Descloitres, S. Alleaume, F.
Petitcolin, and Y. Kaufman. 2002. “The MODIS Fire Products.” Remote Sensing of Environment 83 (1–2):
244–62. https://doi.org/10.1016/S0034-4257(02)00076-7.
124
Kikegawa, Yukihiro, Ai Tanaka, Yukitaka Ohashi, Tomohiko Ihara, and Yoshinori Shigeta. 2014. “Observed
and Simulated Sensitivities of Summertime Urban Surface Air Temperatures to Anthropogenic Heat in
Downtown Areas of Two Japanese Major Cities, Tokyo and Osaka.” Theoretical and Applied Climatology
117 (1): 175–93. https://doi.org/10.1007/s00704-013-0996-8.
Kimura, Fujio, and Shunji Takahashi. 1991. “The Effects of Land-Use and Anthropogenic Heating on the
Surface Temperature in the Tokyo Metropolitan Area: A Numerical Experiment.” Atmospheric
Environment. Part B, Urban Atmosphere 25 (2): 155–64. https://doi.org/10.1016/0957-1272(91)90050-O.
Knox, A., G. J. Evans, J. R. Brook, X. Yao, C. H. Jeong, K. J. Godri, K. Sabaliauskas, and J. G. Slowik.
2009. “Mass Absorption Cross-Section of Ambient Black Carbon Aerosol in Relation to Chemical
Age.” Aerosol Science and Technology 43 (6): 522–32. https://doi.org/10.1080/02786820902777207.
Ko, Joseph, Trevor Krasowsky, and George Ban-Weiss. 2020. “Measurements to Determine the Mixing
State of Black Carbon Emitted from the 2017-2018 California Wildfires and Urban Los Angeles.”
Atmospheric Chemistry and Physics 20 (24): 15635–64. https://doi.org/10.5194/acp-20-15635-2020.
Ko, Joseph, Hannah Schlaerth, Alexandra Bruce, Kelly Sanders, and George Ban-Weiss. 2022. “Measuring
the Impacts of a Real-World Neighborhood-Scale Cool Pavement Deployment on Albedo and
Temperatures in Los Angeles.” Environmental Research Letters 17 (4). https://doi.org/10.1088/1748-
9326/ac58a8.
Kondo, Y., H. Matsui, N. Moteki, L. Sahu, N. Takegawa, M. Kajino, Y. Zhao, et al. 2011. “Emissions of
Black Carbon, Organic, and Inorganic Aerosols from Biomass Burning in North America and Asia in
2008.” Journal of Geophysical Research Atmospheres 116 (8): 1–25. https://doi.org/10.1029/2010JD015152.
Kovats, R. Sari, and Shakoor Hajat. 2008. “Heat Stress and Public Health: A Critical Review.” Annual Review
of Public Health 29: 41–55. https://doi.org/10.1146/annurev.publhealth.29.020907.090843.
Krasowsky, Trevor S., Gavin R. Mcmeeking, Constantinos Sioutas, and George Ban-Weiss. 2018.
“Characterizing the Evolution of Physical Properties and Mixing State of Black Carbon Particles: From
near a Major Highway to the Broader Urban Plume in Los Angeles.” Atmospheric Chemistry and Physics 18
(16): 11991–10. https://doi.org/10.5194/acp-18-11991-2018.
Krasowsky, Trevor S., Gavin R. McMeeking, Dongbin Wang, Constantinos Sioutas, and George A. Ban-
Weiss. 2016. “Measurements of the Impact of Atmospheric Aging on Physical and Optical Properties
of Ambient Black Carbon Particles in Los Angeles.” Atmospheric Environment 142: 496–504.
https://doi.org/10.1016/j.atmosenv.2016.08.010.
Kusaka, Hiroyuki, Hiroaki Kondo, Yokihiro Kikegawa, and Fujio Kimura. 2001. “A Simple Single ‐layer
Urban Canopy Model for Atmospheric Models: Comparison with Multi ‐layer and Slab Models.”
Boundary-Layer Meteorology 101 (3): 329–58.
Laaidi, Karine, Abdelkrim Zeghnoun, Bénédicte Dousset, Philippe Bretin, Stéphanie Vandentorren,
Emmanuel Giraudet, and Pascal Beaudeau. 2012. “The Impact of Heat Islands on Mortality in Paris
during the August 2003 Heat Wave.” Environmental Health Perspectives 120 (2): 254–59.
https://doi.org/10.1289/ehp.1103532.
Laborde, M., M. Crippa, T. Tritscher, Z. Jurányi, P. F. Decarlo, B. Temime-Roussel, N. Marchand, et al.
125
2013. “Black Carbon Physical Properties and Mixing State in the European Megacity Paris.” Atmospheric
Chemistry and Physics 13 (11): 5831–56. https://doi.org/10.5194/acp-13-5831-2013.
Laborde, M., P. Mertes, P. Zieger, J. Dommen, U. Baltensperger, and M. Gysel. 2012. “Sensitivity of the
Single Particle Soot Photometer to Different Black Carbon Types.” Atmospheric Measurement Techniques 5
(5): 1031–43. https://doi.org/10.5194/amt-5-1031-2012.
Laborde, M., M. Schnaiter, C. Linke, H. Saathoff, K.-H. Naumann, O. Möhler, S. Berlenz, et al. 2012.
“Single Particle Soot Photometer Intercomparison at the AIDA Chamber.” Atmospheric Measurement
Techniques 5 (12): 3077–97. https://doi.org/10.5194/amt-5-3077-2012.
Lee, S. H., C. K. Song, J. J. Baik, and S. U. Park. 2009. “Estimation of Anthropogenic Heat Emission in the
Gyeong-In Region of Korea.” Theoretical and Applied Climatology 96 (3–4): 291–303.
https://doi.org/10.1007/s00704-008-0040-6.
Levy, R. C., S. Mattoo, L. A. Munchak, L. A. Remer, A. M. Sayer, F. Patadia, and N. C. Hsu. 2013. “The
Collection 6 MODIS Aerosol Products over Land and Ocean.” Atmospheric Measurement Techniques 6
(11): 2989–3034. https://doi.org/10.5194/amt-6-2989-2013.
Li, Dan, and Elie Bou-Zeid. 2013. “Synergistic Interactions between Urban Heat Islands and Heat Waves:
The Impact in Cities Is Larger than the Sum of Its Parts.” Journal of Applied Meteorology and Climatology 52
(9): 2051–64. https://doi.org/10.1175/JAMC-D-13-02.1.
Li, H., J. Harvey, and A. Kendall. 2013. “Field Measurement of Albedo for Different Land Cover Materials
and Effects on Thermal Performance.” Building and Environment 59: 536–46.
https://doi.org/10.1016/j.buildenv.2012.10.014.
Li, Xian Xiang, Tieh Yong Koh, Dara Entekhabi, Matthias Roth, Jagabandhu Panda, and Leslie K. Norford.
2013. “A Multi-Resolution Ensemble Study of a Tropical Urban Environment and Its Interactions with
the Background Regional Atmosphere.” Journal of Geophysical Research Atmospheres 118 (17): 9804–18.
https://doi.org/10.1002/jgrd.50795.
Li, Xian Xiang, Tieh Yong Koh, Jagabandhu Panda, and Leslie K. Norford. 2016. “Impact of Urbanization
Patterns on the Local Climate of a Tropical City, Singapore: An Ensemble Study.” Journal of Geophysical
Research 121 (9): 4386–4403. https://doi.org/10.1002/2015JD024452.
Lin, Chuan Yao, Fei Chen, J. C. Huang, W. C. Chen, Y. A. Liou, W. N. Chen, and Shaw C. Liu. 2008.
“Urban Heat Island Effect and Its Impact on Boundary Layer Development and Land-Sea Circulation
over Northern Taiwan.” Atmospheric Environment 42 (22): 5635–49.
https://doi.org/10.1016/j.atmosenv.2008.03.015.
Liu, Bin, Zhenghui Xie, Peihua Qin, Shuang Liu, Ruichao Li, Longhuan Wang, Yan Wang, et al. 2021.
“Increases in Anthropogenic Heat Release from Energy Consumption Lead to More Frequent
Extreme Heat Events in Urban Cities.” Advances in Atmospheric Sciences 38 (3): 430–45.
https://doi.org/10.1007/s00376-020-0139-y.
Liu, D., J. D. Allan, D. E. Young, H. Coe, D. Beddows, Z. L. Fleming, M. J. Flynn, et al. 2014. “Size
Distribution, Mixing State and Source Apportionment of Black Carbon Aerosol in London during
Winter Time.” Atmospheric Chemistry and Physics 14 (18): 10061–84. https://doi.org/10.5194/acp-14-
126
10061-2014.
Liu, D., M. Flynn, M. Gysel, A. Targino, I. Crawford, K. Bower, T. Choularton, et al. 2010. “Single Particle
Characterization of Black Carbon Aerosols at a Tropospheric Alpine Site in Switzerland.” Atmospheric
Chemistry and Physics 10 (15): 7389–7407. https://doi.org/10.5194/acp-10-7389-2010.
Liu, Ruiting, and Zhiwei Han. 2016. “The Effects of Anthropogenic Heat Release on Urban Meteorology
and Implication for Haze Pollution in the Beijing-Tianjin-Hebei Region.” Advances in Meteorology 2016.
https://doi.org/10.1155/2016/6178308.
Liu, Ruiting, Zhiwei Han, Jian Wu, Yonghong Hu, and Jiawei Li. 2017. “The Impacts of Urban Surface
Characteristics on Radiation Balance and Meteorological Variables in the Boundary Layer around
Beijing in Summertime.” Atmospheric Research 197 (February): 167–76.
https://doi.org/10.1016/j.atmosres.2017.07.006.
Lontorfos, V., C. Efthymiou, and M. Santamouris. 2018. “On the Time Varying Mitigation Performance of
Reflective Geoengineering Technologies in Cities.” Renewable Energy 115: 926–30.
https://doi.org/10.1016/j.renene.2017.09.033.
McCarthy, Mark P, Martin J Best, and Richard A Betts. 2010. “Climate Change in Cities Due to Global
Warming and Urban Effects.” Geophysical Research Letters 37 (9).
McMeeking, G. R., N. Good, M. D. Petters, G. McFiggans, and H. Coe. 2011. “Influences on the Fraction
of Hydrophobic and Hydrophilic Black Carbon in the Atmosphere.” Atmospheric Chemistry and Physics 11
(10): 5099–5112. https://doi.org/10.5194/acp-11-5099-2011.
McMeeking, G. R., T. Hamburger, D. Liu, M. Flynn, W. T. Morgan, M. Northway, E. J. Highwood, et al.
2010. “Black Carbon Measurements in the Boundary Layer over Western and Northern Europe.”
Atmospheric Chemistry and Physics 10 (19): 9393–9414. https://doi.org/10.5194/acp-10-9393-2010.
McMeeking, G. R., W. T. Morgan, M. Flynn, E. J. Highwood, K. Turnbull, J. Haywood, and H. Coe. 2011.
“Black Carbon Aerosol Mixing State, Organic Aerosols and Aerosol Optical Properties over the United
Kingdom.” Atmospheric Chemistry and Physics 11 (17): 9037–52. https://doi.org/10.5194/acp-11-9037-
2011.
Metcalf, A. R., J. S. Craven, J. J. Ensberg, J. Brioude, W. Angevine, A. Sorooshian, H. T. Duong, H. H.
Jonsson, R. C. Flagan, and J. H. Seinfeld. 2012. “Black Carbon Aerosol over the Los Angeles Basin
during CalNex.” Journal of Geophysical Research Atmospheres 117 (8): 1–24.
https://doi.org/10.1029/2011JD017255.
Middel, Ariane, V. Kelly Turner, Florian A. Schneider, Yujia Zhang, and Matthew Stiller. 2020. “Solar
Reflective Pavements-A Policy Panacea to Heat Mitigation?” Environmental Research Letters 15 (6).
https://doi.org/10.1088/1748-9326/ab87d4.
Millstein, Dev, and Surabi Menon. 2011. “Regional Climate Consequences of Large-Scale Cool Roof and
Photovoltaic Array Deployment.” Environmental Research Letters 6 (3). https://doi.org/10.1088/1748-
9326/6/3/034001.
Mohegh, Arash, Pablo Rosado, Ling Jin, Dev Millstein, Ronnen Levinson, and George Ban-Weiss. 2017.
127
“Modeling the Climate Impacts of Deploying Solar Reflective Cool Pavements in California Cities.”
Journal of Geophysical Research 122 (13): 6798–6817. https://doi.org/10.1002/2017JD026845.
Molnár, Gergely, Attila Kovács, and Tamás Gál. 2020. “How Does Anthropogenic Heating Affect the
Thermal Environment in a Medium-Sized Central European City? A Case Study in Szeged, Hungary.”
Urban Climate 34 (April): 100673. https://doi.org/10.1016/j.uclim.2020.100673.
Morgan, William T., James D. Allan, Stéphane Bauguitte, Eoghan Darbyshire, Michael J. Flynn, James Lee,
Dantong Liu, et al. 2020. “Transformation and Ageing of Biomass Burning Carbonaceous Aerosol
over Tropical South America from Aircraft in Situ Measurements during SAMBBA.” Atmospheric
Chemistry and Physics 20 (9): 5309–26. https://doi.org/10.5194/acp-20-5309-2020.
Moteki, Nobuhiro, Y. Kondo, Y. Miyazaki, N. Takegawa, Y. Komazaki, G. Kurata, T. Shirai, D. R. Blake, T.
Miyakawa, and M. Koike. 2007. “Evolution of Mixing State of Black Carbon Particles: Aircraft
Measurements over the Western Pacific in March 2004.” Geophysical Research Letters 34 (11).
https://doi.org/10.1029/2006GL028943.
Moteki, Nobuhiro, and Yutaka Kondo. 2007. “Effects of Mixing State on Black Carbon Measurements by
Laser-Induced Incandescence.” Aerosol Science and Technology 41 (4): 398–417.
https://doi.org/10.1080/02786820701199728.
———. 2008. “Method to Measure Time-Dependent Scattering Cross Sections of Particles Evaporating in
a Laser Beam.” Journal of Aerosol Science 39 (4): 348–64. https://doi.org/10.1016/j.jaerosci.2007.12.002.
Moteki, Nobuhiro, Yutaka Kondo, and Kouji Adachi. 2014. “Identification by Single-Particle Soot
Photometer of Black Carbon Particles Attached to Other Particles: Laboratory Experiments and
Ground Observations in Tokyo.” Journal of Geophysical Research 119 (2).
https://doi.org/10.1002/2013JD020655.
Mughal, M. O., Xian Xiang Li, Tiangang Yin, Alberto Martilli, Oscar Brousse, Maria Angela Dissegna, and
Leslie K. Norford. 2019. “High-Resolution, Multilayer Modeling of Singapore’s Urban Climate
Incorporating Local Climate Zones.” Journal of Geophysical Research: Atmospheres 124 (14): 7764–85.
https://doi.org/10.1029/2018JD029796.
Murage, Peninah, Shakoor Hajat, and R. Sari Kovats. 2017. “Effect of Night-Time Temperatures on Cause
and Age-Specific Mortality in London.” Environmental Epidemiology, 1.
https://doi.org/10.1097/ee9.0000000000000005.
NCEI. 2021. “2021 Integrated Surface Dataset (Global) Electronic Dataset.” 2021.
https://doi.org/https://doi.org/10.7910/DVN/2PIUHE.
Oke, Timothy R. 1973. “City Size and the Urban Heat Island.” Atmospheric Environment (1967) 7 (8): 769–79.
———. 1982. “The Energetic Basis of the Urban Heat Island.” Quarterly Journal of the Royal Meteorological
Society 108 (455): 1–24.
———. 1995. “The Heat Island of the Urban Boundary Layer: Characteristics, Causes and Effects.” In
Wind Climate in Cities, 81–107. Springer.
128
Oleson, K. W., G. B. Bonan, and J. Feddema. 2010. “Effects of White Roofs on Urban Temperature in a
Global Climate Model.” Geophysical Research Letters 37 (3). https://doi.org/10.1029/2009GL042194.
ORNL. 2013. “Assessment of Industrial Load for Demand Response across U.S. Regions of the Western
Interconnect.” https://doi.org/https://doi.org/10.2172/1111445.
Patz, Jonathan A., Diarmid Campbell-Lendrum, Tracey Holloway, and Jonathan A. Foley. 2005. “Impact of
Regional Climate Change on Human Health.” Nature 438 (7066): 310–17.
https://doi.org/10.1038/nature04188.
Perring, Anne E., Joshua P. Schwarz, Milos Z. Markovic, David W. Fahey, Jose L. Jimenez, Pedro
Campuzano-Jost, Brett D. Palm, et al. 2017. “In Situ Measurements of Water Uptake by Black Carbon-
Containing Aerosol in Wildfire Plumes.” Journal of Geophysical Research 122 (2): 1086–97.
https://doi.org/10.1002/2016JD025688.
Petrov, Andrey. 2012. “One Hundred Years of Dasymetric Mapping: Back to the Origin.” Cartographic
Journal 49 (3): 256–64. https://doi.org/10.1179/1743277412Y.0000000001.
Pincetl, Stephanie, H. Gustafson, L. Strug, and D. Cheng. 2019. “UCLA Energy Atlas (Version 2.0).” Los
Angeles: California Center for Sustainable Communities., UCLA. http://energyatlas.ucla.edu.
Pincetl, Stephanie, and Joshua P. Newell. 2017. “Why Data for a Political-Industrial Ecology of Cities?”
Geoforum 85: 381–91. https://doi.org/10.1016/j.geoforum.2017.03.002.
Poortinga, Ate, Nicholas Clinton, David Saah, Peter Cutter, Farrukh Chishtie, Kel N. Markert, Eric R.
Anderson, et al. 2018. “An Operational Before-after-Control-Impact (BACI) Designed Platform for
Vegetation Monitoring at Planetary Scale.” Remote Sensing 10 (5). https://doi.org/10.3390/rs10050760.
Qiu, Lu Yi, and Ling Yun He. 2017. “Can Green Traffic Policies Affect Air Quality? Evidence from a
Difference-in-Difference Estimation in China.” Sustainability (Switzerland) 9 (6): 1–10.
https://doi.org/10.3390/su9061067.
Ramanathan, V., and G. Carmichael. 2008. “Global and Regional Climate Changes Due to Black Carbon.”
Nature Geoscience. Nature Publishing Group. https://doi.org/10.1038/ngeo156.
Rosenfeld, Arthur H., Hashem Akbari, Joseph J. Romm, and Melvin Pomerantz. 1998. “Cool Communities:
Strategies for Heat Island Mitigation and Smog Reduction.” Energy and Buildings 28 (1): 51–62.
https://doi.org/10.1016/S0378-7788(97)00063-7.
Rosenfeld, Arthur H, Hashem Akbari, Sarah Bretz, Beth L Fishman, Dan M Kurn, David Sailor, and Haider
Taha. 1995. “Mitigation of Urban Heat Islands: Materials, Utility Programs, Updates.” Energy and
Buildings 22 (3): 255–65.
Rossi, Giuseppe, Paola Iacomussi, and Michele Zinzi. 2018. “Lighting Implications of Urban Mitigation
Strategies through Cool Pavements: Energy Savings and Visual Comfort.” Climate 6 (2).
https://doi.org/10.3390/cli6020026.
Ryu, Young Hee, Jong Jin Baik, and Sang Hyun Lee. 2013. “Effects of Anthropogenic Heat on Ozone Air
Quality in a Megacity.” Atmospheric Environment 80: 20–30.
129
https://doi.org/10.1016/j.atmosenv.2013.07.053.
Sahu, L. K., Y. Kondo, N. Moteki, N. Takegawa, Y. Zhao, M. J. Cubison, J. L. Jimenez, et al. 2012.
“Emission Characteristics of Black Carbon in Anthropogenic and Biomass Burning Plumes over
California during ARCTAS-CARB 2008.” Journal of Geophysical Research Atmospheres 117 (16): 1–20.
https://doi.org/10.1029/2011JD017401.
Sailor, David J. 2011. “A Review of Methods for Estimating Anthropogenic Heat and Moisture Emissions
in the Urban Environment.” International Journal of Climatology 31 (2): 189–99.
https://doi.org/10.1002/joc.2106.
Sailor, David J., and Lu Lu. 2004. “A Top-down Methodology for Developing Diurnal and Seasonal
Anthropogenic Heating Profiles for Urban Areas.” Atmospheric Environment 38 (17): 2737–48.
https://doi.org/10.1016/j.atmosenv.2004.01.034.
Salamanca, Francisco, Alberto Martilli, Mukul Tewari, and Fei Chen. 2011. “A Study of the Urban Boundary
Layer Using Different Urban Parameterizations and High-Resolution Urban Canopy Parameters with
WRF.” Journal of Applied Meteorology and Climatology 50 (5): 1107–28.
https://doi.org/10.1175/2010JAMC2538.1.
Salamanca, Francisco, Alberto Martilli, and Carlos Yagüe. 2012. “A Numerical Study of the Urban Heat
Island over Madrid during the DESIREX (2008) Campaign with WRF and an Evaluation of Simple
Mitigation Strategies.” International Journal of Climatology 32 (15): 2372–86.
https://doi.org/10.1002/joc.3398.
Santamouris, M. 2013. “Using Cool Pavements as a Mitigation Strategy to Fight Urban Heat Island - A
Review of the Actual Developments.” Renewable and Sustainable Energy Reviews 26: 224–40.
https://doi.org/10.1016/j.rser.2013.05.047.
Santamouris, M., N. Gaitani, A. Spanou, M. Saliari, K. Giannopoulou, K. Vasilakopoulou, and T.
Kardomateas. 2012. “Using Cool Paving Materials to Improve Microclimate of Urban Areas - Design
Realization and Results of the Flisvos Project.” Building and Environment 53: 128–36.
https://doi.org/10.1016/j.buildenv.2012.01.022.
Santamouris, M., A. Synnefa, and T. Karlessi. 2011. “Using Advanced Cool Materials in the Urban Built
Environment to Mitigate Heat Islands and Improve Thermal Comfort Conditions.” Solar Energy 85
(12): 3085–3102. https://doi.org/10.1016/j.solener.2010.12.023.
Santamouris, M., F. Xirafi, N. Gaitani, A. Spanou, M. Saliari, and K. Vassilakopoulou. 2012. “Improving the
Microclimate in a Dense Urban Area Using Experimental and Theoretical Techniques - The Case of
Marousi, Athens.” International Journal of Ventilation 11 (1): 1–16.
https://doi.org/10.1080/14733315.2012.11683966.
Sarrat, C., A. Lemonsu, V. Masson, and D. Guedalia. 2006. “Impact of Urban Heat Island on Regional
Atmospheric Pollution.” Atmospheric Environment 40 (10): 1743–58.
https://doi.org/10.1016/j.atmosenv.2005.11.037.
Schroeder, Wilfrid, Patricia Oliva, Louis Giglio, and Ivan A. Csiszar. 2014. “The New VIIRS 375m Active
Fire Detection Data Product: Algorithm Description and Initial Assessment.” Remote Sensing of
130
Environment 143. https://doi.org/10.1016/j.rse.2013.12.008.
Schwarz, J. P., J. R. Spackman, R. S. Gao, A. E. Perring, E. Cross, T. B. Onasch, A. Ahern, et al. 2010. “The
Detection Efficiency of the Single Particle Soot Photometer.” Aerosol Science and Technology 44 (8): 612–
28. https://doi.org/10.1080/02786826.2010.481298.
Schwarz, Joshua P., R. S. Gao, D. W. Fahey, D. S. Thomson, L. A. Watts, J. C. Wilson, J. M. Reeves, et al.
2006. “Single-Particle Measurements of Midlatitude Black Carbon and Light-Scattering Aerosols from
the Boundary Layer to the Lower Stratosphere.” Journal of Geophysical Research Atmospheres 111 (16).
https://doi.org/10.1029/2006JD007076.
Schwarz, Joshua P., R. S. Gao, J. R. Spackman, L. A. Watts, D. S. Thomson, D. W. Fahey, T. B. Ryerson, et
al. 2008. “Measurement of the Mixing State, Mass, and Optical Size of Individual Black Carbon
Particles in Urban and Biomass Burning Emissions.” Geophysical Research Letters 35 (13): 1–5.
https://doi.org/10.1029/2008GL033968.
Schwarz, Joshua P., J. R. Spackman, D. W. Fahey, R. S. Gao, U. Lohmann, P. Stier, L. A. Watts, et al. 2008.
“Coatings and Their Enhancement of Black Carbon Light Absorption in the Tropical Atmosphere.”
Journal of Geophysical Research Atmospheres 113 (3): 1–10. https://doi.org/10.1029/2007JD009042.
Sedlacek, Arthur J., Ernie R. Lewis, Lawrence Kleinman, Jianzhong Xu, and Qi Zhang. 2012.
“Determination of and Evidence for Non-Core-Shell Structure of Particles Containing Black Carbon
Using the Single-Particle Soot Photometer (SP2).” Geophysical Research Letters 39 (6): 2–7.
https://doi.org/10.1029/2012GL050905.
Sedlacek, Arthur J., Ernie R. Lewis, T. B. Onasch, and Lawrence Kleinman. 2019. “Improving Our
Understanding of Light Absorbing Aerosols from Biomass Burning.” In American Geophysical Union Fall
Meeting. San Francisco, CA.
Sedlacek, Arthur J., Ernie R. Lewis, Timothy B. Onasch, Andrew T. Lambe, and Paul Davidovits. 2015.
“Investigation of Refractory Black Carbon-Containing Particle Morphologies Using the Single-Particle
Soot Photometer (SP2).” Aerosol Science and Technology 49 (10).
https://doi.org/10.1080/02786826.2015.1074978.
Sen, Sushobhan, Juan Pablo Ricardo Mendèz Ruiz Fernandèz, and Jeffery Roesler. 2020. “Reflective Parking
Lots for Microscale Urban Heat Island Mitigation.” Transportation Research Record 2674 (8): 663–71.
https://doi.org/10.1177/0361198120919401.
Sen, Sushobhan, Jeffery Roesler, and Daniel King. 2019. “Albedo Estimation of Finite-Sized Concrete
Specimens.” Journal of Testing and Evaluation 47 (2): 20170059. https://doi.org/10.1520/jte20170059.
Sha, Aimin, Zhuangzhuang Liu, Kun Tang, and Pinyi Li. 2017. “Solar Heating Reflective Coating Layer
(SHRCL) to Cool the Asphalt Pavement Surface.” Construction and Building Materials 139: 355–64.
https://doi.org/10.1016/j.conbuildmat.2017.02.087.
Sharma, A., P. Conry, H. J.S. Fernando, Alan F. Hamlet, J. J. Hellmann, and F. Chen. 2016. “Green and
Cool Roofs to Mitigate Urban Heat Island Effects in the Chicago Metropolitan Area: Evaluation with a
Regional Climate Model.” Environmental Research Letters 11 (6). https://doi.org/10.1088/1748-
9326/11/6/064004.
131
Shiraiwa, M., Y. Kondo, N. Moteki, N. Takegawa, Y. Miyazaki, and D. R. Blake. 2007. “Evolution of Mixing
State of Black Carbon in Polluted Air from Tokyo.” Geophysical Research Letters 34 (16): 2–6.
https://doi.org/10.1029/2007GL029819.
Shiraiwa, M., Y. Kondo, N. Moteki, N. Takegawa, L. K. Sahu, A. Takami, S. Hatakeyama, S. Yonemura, and
D. R. Blake. 2008. “Radiative Impact of Mixing State of Black Carbon Aerosol in Asian Outflow.”
Journal of Geophysical Research Atmospheres 113 (24): 1–13. https://doi.org/10.1029/2008JD010546.
Shirakawa, M. A., L. N. de Lima, C. C. Gaylarde, J. A. Silva Junior, P. H.F. Loz, and V. M. John. 2020.
“Effects of Natural Aging on the Properties of a Cool Surface Exposed in Different Brazilian
Environments.” Energy and Buildings 221. https://doi.org/10.1016/j.enbuild.2020.110031.
Slowik, Jay G., Eben S. Cross, Jeong Ho Han, Paul Davidovits, Timothy B. Onasch, John T. Jayne, Leah R.
Williams, et al. 2007. “An Inter-Comparison of Instruments Measuring Black Carbon Content of Soot
Particles.” Aerosol Science and Technology 41 (3): 295–314. https://doi.org/10.1080/02786820701197078.
Stein, A. F., R. R. Draxler, G. D. Rolph, B. J.B. Stunder, M. D. Cohen, and F. Ngan. 2015. “Noaa’s Hysplit
Atmospheric Transport and Dispersion Modeling System.” Bulletin of the American Meteorological Society 96
(12): 2059–77. https://doi.org/10.1175/BAMS-D-14-00110.1.
Stephens, Michelle, Nelson Turner, and Jon Sandberg. 2003. “Particle Identification by Laser-Induced
Incandescence in a Solid-State Laser Cavity.” Applied Optics 42 (19): 3726.
https://doi.org/10.1364/ao.42.003726.
Stewart, I. D. 2011. “A Systematic Review and Scientific Critique of Methodology in Modern Urban Heat
Island Literature.” International Journal of Climatology 31 (2): 200–217. https://doi.org/10.1002/joc.2141.
Stone, Brian, Evan Mallen, Mayuri Rajput, Carina J. Gronlund, Ashley M. Broadbent, E. Scott Krayenhoff,
Godfried Augenbroe, Marie S. O’Neill, and Matei Georgescu. 2021. “Compound Climate and
Infrastructure Events: How Electrical Grid Failure Alters Heat Wave Risk.” Environmental Science and
Technology 55 (10): 6957–64. https://doi.org/10.1021/acs.est.1c00024.
Stone, Brian, Jason Vargo, and Dana Habeeb. 2012. “Managing Climate Change in Cities: Will Climate
Action Plans Work?” Landscape and Urban Planning 107 (3): 263–71.
https://doi.org/10.1016/j.landurbplan.2012.05.014.
Synnefa, A., A. Dandou, M. Santamouris, M. Tombrou, and N. Soulakellis. 2008. “On the Use of Cool
Materials as a Heat Island Mitigation Strategy.” Journal of Applied Meteorology and Climatology 47 (11):
2846–56. https://doi.org/10.1175/2008JAMC1830.1.
Taha, Haider. 1997. “Urban Climates and Heat Islands: Albedo, Evapotranspiration, and Anthropogenic
Heat.” Energy and Buildings 25 (2): 99–103. https://doi.org/10.1016/s0378-7788(96)00999-1.
———. 2015. “Meteorological, Emissions and Air-Quality Modeling of Heat-Island Mitigation: Recent
Findings for California, USA.” International Journal of Low-Carbon Technologies 10 (1): 3–14.
https://doi.org/10.1093/ijlct/ctt010.
Taha, Haider, Sharon Douglas, and Jay Haney. 1997. “Mesoscale Meteorological and Air Quality Impacts of
Increased Urban Albedo and Vegetation.” Energy and Buildings 25 (2): 169–77.
132
https://doi.org/10.1016/s0378-7788(96)01006-7.
Taha, Haider, Steven Konopacki, and Sasa Gabersek. 1999. “Impacts of Large-Scale Surface Modifications
on Meteorological Conditions and Energy Use: A 10-Region Modeling Study.” Theoretical and Applied
Climatology 62 (3–4): 175–85. https://doi.org/10.1007/s007040050082.
Takane, Yuya, Yukihiro Kikegawa, Masayuki Hara, and C. Sue B. Grimmond. 2019. “Urban Warming and
Future Air-Conditioning Use in an Asian Megacity: Importance of Positive Feedback.” Npj Climate and
Atmospheric Science 2 (1): 1–11. https://doi.org/10.1038/s41612-019-0096-2.
Taleghani, Mohammad, Peter J. Crank, Arash Mohegh, David J. Sailor, and George A. Ban-Weiss. 2019.
“The Impact of Heat Mitigation Strategies on the Energy Balance of a Neighborhood in Los Angeles.”
Solar Energy 177 (September 2018): 604–11. https://doi.org/10.1016/j.solener.2018.11.041.
Taleghani, Mohammad, David Sailor, and George A. Ban-Weiss. 2016. “Micrometeorological Simulations to
Predict the Impacts of Heat Mitigation Strategies on Pedestrian Thermal Comfort in a Los Angeles
Neighborhood.” Environmental Research Letters 11 (2): 24003. https://doi.org/10.1088/1748-
9326/11/2/024003.
Taylor, J. W., J. D. Allan, G. Allen, H. Coe, P. I. Williams, M. J. Flynn, M. Le Breton, et al. 2014. “Size-
Dependent Wet Removal of Black Carbon in Canadian Biomass Burning Plumes.” Atmospheric
Chemistry and Physics 14 (24): 13755–71. https://doi.org/10.5194/acp-14-13755-2014.
Todey, D.P., D. E. Herzmann, and G. S. Takle. 2002. “The Iowa Environmental Mesonet – Combining
Networks into a Single Network.” In 6th Symposium on Integrated Observing Systems, Annual Meeting of the
American Meteorological Society. Orlando, FL.
Torrance, K. E., and J. S.W. Shun. 1976. “Time-Varying Energy Consumption as a Factor in Urban
Climate.” Atmospheric Environment (1967) 10 (4): 329–37. https://doi.org/10.1016/0004-6981(76)90174-
8.
Torres, O. 2006. “OMI/Aura Near UV Aerosol Optical Depth and Single Scattering Albedo 1-Orbit L2
Swath 13x24 Km V003.” Goddard Earth Sciences Data and Information Services Center (GES DISC).
2006. https://doi.org/10.5067/Aura/OMI/DATA2004.
Tsoka, S., T. Theodosiou, K. Tsikaloudaki, and F. Flourentzou. 2018. “Modeling the Performance of Cool
Pavements and the Effect of Their Aging on Outdoor Surface and Air Temperatures.” Sustainable Cities
and Society 42 (March): 276–88. https://doi.org/10.1016/j.scs.2018.07.016.
Tsoka, Stella, Katerina Tsikaloudaki, Theodoros Theodosiou, and Dimitrios Bikas. 2020. “Urban Warming
and Cities’ Microclimates: Investigation Methods and Mitigation Strategies—A Review.” Energies 13 (6).
https://doi.org/10.3390/en13061414.
United Nations. 2019. World Urbanization Prospects: The 2018 Revision. Sustainable Development.
https://doi.org/https://doi.org/10.18356/b9e995fe-en.
USGCRP. 2017. “Climate Science Special Report: Fourth National Climate Assessment, Volume I.”
Washington, DC. https://doi.org/10.7930/J0J964J6.
133
Vahmani, P, and G A Ban-Weiss. 2016. “Impact of Remotely Sensed Albedo and Vegetation Fraction on
Simulation of Urban Climate in WRF-Urban Canopy Model: A Case Study of the Urban Heat Island in
Los Angeles.” Journal of Geophysical Research: Atmospheres 121 (4): 1511–31.
Vaidyanathan, Ambarish, Josephine Malilay, Paul Schramm, and Shubhayu Saha. 2020. “Heat-Related
Deaths — United States , 2004 – 2018” 69 (24).
Vitanova, Lidia Lazarova, and Hiroyuki Kusaka. 2018. “Study on the Urban Heat Island in Sofia City:
Numerical Simulations with Potential Natural Vegetation and Present Land Use Data.” Sustainable Cities
and Society 40 (March): 110–25. https://doi.org/10.1016/j.scs.2018.03.012.
Wang, Qiyuan, Junji Cao, Yongming Han, Jie Tian, Yue Zhang, Siwatt Pongpiachan, Yonggang Zhang, et al.
2018. “Enhanced Light Absorption Due to the Mixing State of Black Carbon in Fresh Biomass
Burning Emissions.” Atmospheric Environment 180 (December 2017): 184–91.
https://doi.org/10.1016/j.atmosenv.2018.02.049.
Wang, Qiyuan, R. J. Huang, Junji Cao, Yongming Han, Gehui Wang, Guohui Li, Yichen Wang, Wenting
Dai, Renjian Zhang, and Yaqing Zhou. 2014. “Mixing State of Black Carbon Aerosol in a Heavily
Polluted Urban Area of China: Implications for Light Absorption Enhancement.” Aerosol Science and
Technology 48 (7): 689–97. https://doi.org/10.1080/02786826.2014.917758.
Wang, Shasha, Deyong Hu, Chen Yu, Shanshan Chen, and Yufei Di. 2020. “Mapping China’s Time-Series
Anthropogenic Heat Flux with Inventory Method and Multi-Source Remotely Sensed Data.” Science of
the Total Environment 734: 139457. https://doi.org/10.1016/j.scitotenv.2020.139457.
Wang, Xiaomei, Xuguang Sun, Jianping Tang, and Xiuqun Yang. 2015. “Urbanization-Induced Regional
Warming in Yangtze River Delta: Potential Role of Anthropogenic Heat Release.” International Journal of
Climatology 35 (15): 4417–30. https://doi.org/10.1002/joc.4296.
Wang, Y., Y. Li, S. Di Sabatino, A. Martilli, and P. W. Chan. 2018. “Effects of Anthropogenic Heat Due to
Air-Conditioning Systems on an Extreme High Temperature Event in Hong Kong.” Environmental
Research Letters 13 (3). https://doi.org/10.1088/1748-9326/aaa848.
Wang, Y., S. Di Sabatino, A. Martilli, Y. Li, M. S. Wong, E. Gutiérrez, and P. W. Chan. 2017. “Impact of
Land Surface Heterogeneity on Urban Heat Island Circulation and Sea-Land Breeze Circulation in
Hong Kong.” Journal of Geophysical Research 122 (8): 4332–52. https://doi.org/10.1002/2017JD026702.
Wen, Junpeng, J. I. Chen, Wenshi Lin, Baolin Jiang, Suishan Xu, and Jing Lan. 2020. “Impacts of
Anthropogenic Heat Flux and Urban Land-Use Change on Frontal Rainfall near Coastal Regions: A
Case Study of a Rainstorm over the Pearl River Delta, South China.” Journal of Applied Meteorology and
Climatology 59 (3): 363–79. https://doi.org/10.1175/JAMC-D-18-0296.1.
Wing, Coady, Kosali Simon, and Ricardo A. Bello-Gomez. 2018. “Designing Difference in Difference
Studies: Best Practices for Public Health Policy Research.” Annual Review of Public Health 39: 453–69.
https://doi.org/10.1146/annurev-publhealth-040617-013507.
World Health Organization. 2012. “Health Effects of Black Carbon” 41 (28): 5908–20.
https://doi.org/10.1016/j.atmosenv.2007.03.042.
134
Xie, Min, Kuanguang Zhu, Tijian Wang, Wen Feng, Da Gao, Mengmeng Li, Shu Li, et al. 2016. “Changes in
Regional Meteorology Induced by Anthropogenic Heat and Their Impacts on Air Quality in South
China.” Atmospheric Chemistry and Physics 16 (23): 15011–31. https://doi.org/10.5194/acp-16-15011-
2016.
Xu, Kuo, Enshen Long, Jin Li, and Luting Xu. 2019. “Field Measurement and Influence Mechanism
Analysis of the Albedo for a Typical Urban Concrete Surface.” Indoor and Built Environment 28 (6): 837–
47. https://doi.org/10.1177/1420326X18799398.
Yaghoobian, Neda, and Jan Kleissl. 2012. “Effect of Reflective Pavements on Building Energy Use.” Urban
Climate 2: 25–42. https://doi.org/10.1016/j.uclim.2012.09.002.
You, Zhilang, Mingyi Zhang, Jiwei Wang, and Wansheng Pei. 2019. “A Black Near-Infrared Reflective
Coating Based on Nano-Technology.” Energy and Buildings 205: 109523.
https://doi.org/10.1016/j.enbuild.2019.109523.
Yu, Man, Gregory R. Carmichael, Tong Zhu, and Yafang Cheng. 2014. “Sensitivity of Predicted Pollutant
Levels to Anthropogenic Heat Emissions in Beijing.” Atmospheric Environment 89: 169–78.
https://doi.org/10.1016/j.atmosenv.2014.01.034.
Zhang, J., J. Liu, S. Tao, and G. A. Ban-Weiss. 2015. “Long-Range Transport of Black Carbon to the Pacific
Ocean and Its Dependence on Aging Timescale.” Atmospheric Chemistry and Physics.
https://doi.org/10.5194/acp-15-11521-2015.
Zhang, Jiachen, Yun Li, Wei Tao, Junfeng Liu, Ronnen Levinson, Arash Mohegh, and George Ban-Weiss.
2019. “Investigating the Urban Air Quality Effects of Cool Walls and Cool Roofs in Southern
California.” Environmental Science & Technology 53 (13): 7532–42.
Zhang, Jiachen, Arash Mohegh, Yun Li, Ronnen Levinson, and George Ban-Weiss. 2018. “Systematic
Comparison of the Influence of Cool Wall versus Cool Roof Adoption on Urban Climate in the Los
Angeles Basin.” Environmental Science & Technology 52 (19): 11188–97.
Zhang, Jiachen, Kai Zhang, Junfeng Liu, and George Ban-Weiss. 2016. “Revisiting the Climate Impacts of
Cool Roofs around the Globe Using an Earth System Model.” Environmental Research Letters 11 (8): 1–
11. https://doi.org/10.1088/1748-9326/11/8/084014.
Zhang, Yi, Peiyuan Wei, Lei Wang, and Yinghong Qin. 2021. “Temperature of Paved Streets in Urban
Mockups and Its Implication of Reflective Cool Pavements.” Atmosphere 12 (5): 1–12.
https://doi.org/10.3390/atmos12050560.
Zhao, Lei, Xuhui Lee, and Natalie M. Schultz. 2017. “A Wedge Strategy for Mitigation of Urban Warming
in Future Climate Scenarios.” Atmospheric Chemistry and Physics 17 (14): 9067–80.
https://doi.org/10.5194/acp-17-9067-2017.
Zheng, Mulian, Lili Han, Fei Wang, Haichen Mi, Yifeng Li, and Litao He. 2015. “Comparison and Analysis
on Heat Reflective Coating for Asphalt Pavement Based on Cooling Effect and Anti-Skid
Performance.” Construction and Building Materials 93: 1197–1205.
https://doi.org/10.1016/j.conbuildmat.2015.04.043.
135
Zheng, Yuanfan, and Qihao Weng. 2018. “High Spatial- and Temporal-Resolution Anthropogenic Heat
Discharge Estimation in Los Angeles County, California.” Journal of Environmental Management 206:
1274–86. https://doi.org/10.1016/j.jenvman.2017.07.047.

136
Appendices
Appendix A: Supplement to Chapter 2
S1 Source-to-receptor timescale estimations
This section describes the methods used to arrive at the approximate source-to-receptor timescales
shown in Table 2-2. For L1, a semi-quantitative range was estimated by examining CAMS model output and
MODIS imagery/data. The CAMS model and MODIS data both show that long-range transport from East
Asia and Oregon wildfires during the first campaign (September 2017) took at a minimum several days to
arrive at the sampling site from the approximate location of the respective sources. The upper end of the
range for L1 is determined by the approximate maximum lifetime of BC as mentioned in previous literature
(Bond et al., 2006; Lund et al., 2018). The upper end for L3, L8, L9, and L10 were also approximated by the
same logic.
The timescale for L2 was also determined semi-quantitatively. Since it was determined that local
sources were heavily impacting the measurements during this period, we estimated that a timescale on the
order of minutes to hours was appropriate, given that the closest source of substantial emissions would be
right off the dock at the USC Wrigley Institute, which was ~100 m away from the inlet.
The lower end for L3 was determined by observing CAMS model output and determining how long
it took for the PM
2.5
emissions plume from the Thomas Fire (and urban emissions) to recirculate back to
Catalina Island. See video 2 of the Video Supplement for visual support.
The approximations for L4 to L7 were determined using the following steps. First, four HYSPLIT
back-trajectories were chosen for examination: (1) a back-trajectory starting an hour prior to the beginning
of the LEO period, (2) a back-trajectory starting at the beginning of the LEO period, (3) a back-trajectory
starting at the end of the LEO period, and finally (4) a back-trajectory starting an hour after the end of the
LEO period. For each of these four back-trajectories, the first hour at which the trajectory crossed over the
California coast was determined. From this, we approximated how many hours it took for a particle to travel
137
between the inlet and the coastal edge of the Los Angeles basin. The mean “inlet-to-coast” timescale was
then calculated for the four back-trajectories. The floor of this value is what is shown in Table 2-3 as the
approximate characteristic timescale for the time periods, L4 to L7. The floor of the mean was used because
this would conservatively round to the nearest integer value of the timescale by representing the shortest
length of time it would take for a particle to travel from an emission source on land to the sampling site on
Catalina Island.

For L8 to L10, CAMS model and MODIS imagery/data were used to establish an upper and lower limit of
approximate timescales. For example, video 3 and 4 in the Video Supplement show that the plume of
aerosols from the Camp Fire takes at least a few days to reach the Southern California region. This
established the lower end limit of ~days. As mentioned previously, the upper end of ~weeks is from
established knowledge about the approximate lifetime of aerosols in the atmosphere.
S2 Details regarding section 3.1: source identification and meteorology
The red trajectories in Fig. 3 represent seven-day back-trajectories starting at 00:00 Pacific Daylight
Time for every day of the first campaign (September 2017). These back-trajectories show that the dominant
wind-flows during the first campaign were westerly, which is consistent with the typical synoptic winds that
blow toward the Los Angeles coast from the Pacific Ocean, as well as the westerly mesoscale flows driven
by the sea breeze. Wind data from Los Angeles International Airport, Long Beach Airport, and Avalon (Fig.
2) further confirm that winds were generally blowing from the west during the first campaign. The back-
trajectories and wind data suggest that measured particles during the first campaign are “aged” since there
are limited major nearby sources that are upwind of the sampling site. Although the exact sources of BC
during this period cannot be ascertained, the plausible sources of rBC-containing particles would be from
(1) nearby ships and aviation, (2) aged urban and/or biomass burning emissions, and/or (3) inter-
continental transport from East Asia. Using CAMS model data, we further identified two long-range sources
138
that likely contributed to measured rBC-containing particles; these model data identified large biomass
burning events during the first campaign in Oregon and Northern California. Although these fires were
much further away than the Southern California fires that were active during the second and third
campaigns (December 2017, November 2018), close visual tracking of plumes from these fires with the
CAMS visualization tool and NASA aerosol index product shows that PM
2.5
from these fires reached the
coast of California, even as far south as Catalina Island (i.e., our sampling site) (see Fig. S4, Appendix A).
We also identified an example of inter-continental transport of PM
2.5
from East Asia around the time of the
first campaign using CAMS data (see Fig. S6, Appendix A). We did not attempt to determine which
potential sources were dominant for the first campaign, but regardless of the source it is fair to say that
measured rBC-containing particles were aged.
In strong contrast to the first campaign (September 2017), the second and third campaigns
(December 2017, November 2018) included periods in which the sampling location was downwind of
biomass burning and urban emissions. The yellow and green trajectories in Fig. 3b and 3c represent 72-hour
back-trajectories for each hour of the second and third campaigns. These back-trajectories, along with
supporting airport wind data (Fig. 2), confirm that winds were easterly-to-northerly for a significant fraction
of these campaigns. These “Santa Ana” conditions, in which winds originate from dry desert regions north
and east of Los Angeles and advect through the mountain ranges of Southern California to the Los Angeles
basin (Small, 1995), are infamous for exacerbating wildfires. Figs. 3d and 3e show several HYSPLIT
trajectories either going through or coming within close proximity to active wildfires in the Southern
California region (see Table 2-1 for information on wildfires). This provides evidence that the measured
rBC-containing particles in the second and third campaign included important contributions from both
fresh biomass burning emissions, and fresh urban emissions from the Los Angeles basin, which are largely
from motor vehicles.
139
Since meteorology varied a lot more during the second and third campaigns compared to the first
campaign, different periods of interest within these campaigns were examined more carefully in an attempt
to assess the relative impact of different known sources. For the second campaign (December 2017), we
examined the time periods defined by unique peaks in rBC loading (discussed in section 3.2 and labeled in
Fig. 5). After investigating HYSPLIT back-trajectories, CAMS model data, and airport wind data, we
concluded that contributions from the Thomas Fire (Table 2-1) and urban emissions from the Los Angeles
basin were too difficult to accurately distinguish from one another. The Thomas Fire and Los Angeles urban
plumes were within ~110 km (~70 miles) of each other, and even though the CAMS model output (Fig. S5,
Appendix A) shows two distinct plumes around the time of Peak P1 from the second campaign (Fig. 5 and
discussed in Section 3.2), both plumes likely interacted and impacted measurements throughout the second
campaign. Thus, we conclude that both the Thomas Fire and Los Angeles urban emissions were the main
sources of measured rBC during the second campaign, when Santa Ana wind conditions were blowing
emissions towards the sampling site.
Although the relatively low fraction of thickly-coated rBC particles and low average coating
thickness (discussed in section 3.3 and 3.5) might seem to suggest that local biomass burning emissions were
not impacting the second and third campaigns, various data show otherwise. First, Soleimanian et al. (2020)
reported elevated concentrations of levoglucosan (a biomass burning tracer) during the second and third
weeks of November, which overlaps with the third campaign (12-18 November 2018). The weekly mean
concentration of levoglucosan was reported to be 187.5 ng m
-3
for the week of 7 to 14 November 2018, and
83.89 ng m
3
for the week of 15 to 22 November 2018. For reference, levoglucosan concentrations during
July 2018 (non-wildfire season) ranged between ~4 and 17 ng m
-3
. This removes any lingering doubt
whether or not biomass burning emissions contributed to the broader LA basin plume during this period.
There were no levoglucosan measurements available for the second campaign (December 2017), but given a
similar geographic situation (large wildfire close to the LA basin) and similar Santa Ana wind conditions, it
140
would be hard to imagine a fire like the Thomas Fire not affecting measurements on Catalina Island during
the second campaign as well.
To provide an additional line of evidence to fill in for the missing levoglucosan measurements
during the second campaign (December 2017), the HYSPLIT dispersion model was run for both 21 and 22
December 2017, simulating a controlled burn with an identical area as the burn area of the Thomas Fire and
the Creek Fire. The results are shown in Fig. S20 and S21 (Appendix A), and they show beyond reasonable
doubt that the plume fingerprint reaches the sampling location on Catalina Island during the second
campaign. CALIPSO lidar images for the second and third (November 2018) campaigns were also
qualitatively analyzed and reported here in Fig. S9 to S19 (Appendix A). According to the CALIPSO aerosol
classification algorithm, there were many instances during the second and third campaign in which biomass
burning aerosols were present off the coast of Southern California. This supplemental data further bolsters
our initial hypothesis that local Southern California wildfires were contributing to the broader LA plume.
Nonetheless, we cannot quantitatively apportion the relative effect of the urban emissions versus local
biomass burning emissions during these time periods. Based on the mixing state measurements and rBC
core size distributions, we conservatively conclude that although biomass burning emissions were affecting
measurements during this time of relatively low coating thickness, urban emissions seem to be the dominant
source type. Although we cannot eliminate the possibility of relatively thinly-coated biomass burning rBC
particles during the second and third campaigns, there is also no way for us to definitively prove or disprove
this with the available data.
For the third campaign (November 2018), the CAMS model and MODIS satellite imagery show a
large plume of aerosols from Northern and Central California fires (particularly the Camp Fire, Table 2-1)
reaching Southern California (see video 3 of Video Supplement). Figure S7 confirms that back-trajectories
from this period originate from the area of the plume that is visible in the eight-day mean AOD overlay (9
to 16 November 2018). The AOD overlay represents the general extent of the Camp Fire plume ~1-4 days
141
before the start of the period with increased concentrations. The back-trajectories overlaid on the AOD
layer suggest that the measured rBC-containing aerosols during the last two days of the third campaign did
indeed originate from within the Camp Fire plume. This particular period can be further examined in the
supplemental video of animated PM
2.5
concentration isopleths output by the CAMS model for the third
campaign (see video 1 of Video Supplement). In the first part of the third campaign prior to Camp Fire
contributions, measured rBC likely included contributions from a mixture of fresh biomass burning and
urban emissions because the Woolsey fire was located within ~45 km (~30 miles) of the broader Los
Angeles urban plume. We could not isolate contributions from the Woolsey Fire (Ventura) and Los Angeles
urban emissions during this period.

S3 Details regarding section 3.4: negative lag-times and rBC morphology
According to Sedlacek et al. (2015) higher laser power and higher sample flow rates result in more
rapid rBC heating, and therefore higher rates of fragmentation (i.e., larger f
lag,neg
). In this study, a laser current
of 1600 mA was used for the first and second campaign, and a laser current of 1850 mA was used for the
third campaign. The sample flow rate was set to 120 cm
3
min
-1
. Comparatively, Sedlacek et al. (2012)
reported a laser current of 3000 mA and sample flow rate of 120 cm
3
min
-1
. Given that Sedlacek et al. (2015)
found a six-fold increase in f
lag,neg
by increasing the laser current from 2000 mA to 3000 mA (with the same
sample flow rate), it was expected that f
lag,neg
in this study would be much lower than f
lag,neg
reported by
Sedlacek et al. (2012). Taking the operating conditions into account and multiplying the f
lag,neg
in this study by
an adjustment factor of ~7 (1850 mA versus 3000 mA, see Fig. 8 in Sedlacek et al., 2015 ), the November
2018 peaks observed in the f
lag,neg
time series in Fig. 6 would increase from ~0.06 to ~0.41, which is much
closer to the 0.6 value reported by Sedlacek et al. (2012).
142
S3 Supplemental figures

Figure S1. Photo showing the sample inlet, which was positioned on the roof of USC’s Wrigley Institute on
Catalina Island. Mapdata: Google DigitalGlobe.

Figure S2. rBC mass and number concentration for the September campaign. The circle markers indicate
all points, including the spikes. The solid lines represent the concentration time series after filtering out the
spikes.

143

Figure S3. Median number concentrations (a) and mass concentration (b) for the September campaign,
shown as a function of the spike threshold cut-off values (see Section 2.5). The median concentrations
display an asymptotic behaviour.
144

Figure S4. Aerosol optical depth (AOD) and aerosol index layers (NASA MODIS) imposed on top of each
other, from 1 to 14 September 2017 (during the first campaign). Darker red indicates higher concentrations
of aerosols. The large aerosol plume is largely from the Oregon fires that were active during this period. The
blue star indicates the approximate location of the sampling site, and the progression of images show that
the plume from the Oregon fires likely contributed to measured rBC during the first campaign (September
2017). Aerosol index was layered below the AOD layer in order to visually “fill” some gaps in the qualitative
representation of the aerosol plume. Gaps in the AOD and aerosol index layer are due to cloud cover, and
spatial variation in the gaps are due to NASA satellite routes.

145

Figure S5. CAMS PM
2.5
model visualizations from 06 to 11 Pacific Time, on 21 December 2017, ordered
chronologically for every hour, starting from the top-left and ending at the bottom-right sub-figure. This
time period corresponds to Peak P1 of the December campaign (see Figure 5 in the main body). The blue
circle shows the distinct plume from the Thomas Fire in Santa Barbara/Ventura County. The red circle
shows a less pronounced, but distinct, plume from emissions in the Los Angeles urban basin, presumably
mostly from motor vehicle emissions. The green circle denotes the sampling location on Catalina Island.
Concentration colorbar gradient ranges from red to blue, where red represents high concentrations and vice
versa.

146

Figure S6. CAMS PM
2.5
model visualizations showing a plume originating from the East Asia region and
advecting across the Pacific Ocean towards California during the time of the first campaign (September
2017). The first sub-figure is a snapshot at 11 Pacific Time on 2 September 2017. The last sub-figure is a
snapshot at 03 Pacific Time on 10 September 2017. The red arrows track the movement of the PM
2.5
plume
originating from East Asia. The time elapsed between the first and the last frame is approximately eight
days. The movement of the plume illustrates that inter-continental transport from East Asia could
contribute to heavily coated rBC-containing particles measured during the first campaign. Concentration
colorbar gradient ranges from red to blue, where red represents high concentrations and vice versa.
147

Figure S7. Week-long back-trajectories starting every six hours for the period of elevated f
BC
between
November 16-18, 2018 shown on top of an eight-day averaged AOD layer (9 to 16 November 2018) from
MODIS. The AOD scale is shown in the colorbar at the bottom-right corner. The fire symbols represent
active fires in Northern and Central California during the time of measurements. The northern-most fire on
the map represents the Camp Fire, which encompassed a burn area of more than 600 km
2
.

148

Figure S8. Meteorological variables and rBC concentrations during the third campaign (November 2018).
Panel (a) shows wind speed and (b) shows wind direction measured by a NOAA weather station located at
Los Angeles International Airport (LAX). Panel (c) shows rBC mass and number concentrations.

149

Figure S9. Scatter plot as a function of lag-time (i.e., delay time) and BC coating thickness (CT
BC
). Each
point on the plot represents a 10-minute mean value. Data shown includes average values from all three
campaigns. A significant correlation is confirmed using a linear correlation test. The actual numerical p-value
is not shown because it is so small that the test returned it as a zero

150

Figure S10. Images collected by NASA’s CALIPSO lidar on 20 December 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
151

Figure S11. Images collected by NASA’s CALIPSO lidar on 21 December 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
152

Figure S12. Images collected by NASA’s CALIPSO lidar on 22 December 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
153

Figure S13. Images collected by NASA’s CALIPSO lidar on 12 November 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
154

Figure S14. Images collected by NASA’s CALIPSO lidar on 13 November 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
155

Figure S15. Images collected by NASA’s CALIPSO lidar on 14 November 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
156

Figure S16. Images collected by NASA’s CALIPSO lidar on 15 November 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
157

Figure S17. Images collected by NASA’s CALIPSO lidar on 16 November 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
158

Figure S18. Images collected by NASA’s CALIPSO lidar on 17 December 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The image for the daytime
transect was not available on this day.

159

Figure S19. Images collected by NASA’s CALIPSO lidar on 18 November 2017. Panel (a) shows the lidar
image for the nighttime path, shown in the map on the left. The path is segmented into four parts, separated by
different colors. Panel (b) shows the lidar image for the segment of the path that is closest to the Southern
California region. Panel (b) shows the aerosol subtypes identified by the CALIPSO automated classification
algorithm. Orange, brown, and black areas signify biomass burning aerosols. The same information is shown in
panels (d), (e), and (f) for the daytime path, for the same day.
160

Figure S20. HYSPLIT dispersion model simulating a prescribed burn that encompasses the burn
area of the Thomas Fire and Creek Fire. Panel (a) shows the simulation starting on 00:00 Pacific
Time on December 21, 2017. Panel (b), (c), (d), and (e) shows concentration isopleths at 06:00,
12:00, 18:00, and 24:00 Pacific Time, respectively.

161

Figure S21. HYSPLIT dispersion model simulating a prescribed burn that encompasses the burn
area of the Thomas Fire and Creek Fire. Panel (a) shows the simulation starting on 00:00 Pacific
Time on December 22, 2017. Panel (b), (c), (d), and (e) shows concentration isopleths at 06:00,
12:00, 18:00, and 24:00 Pacific Time, respectively.

162

Figure S22. HYSPLIT back-trajectories for LEO periods L1 through L10. See section 2.3 of
manuscript for details regarding HYSPLIT. Mapdata: Google DigitalGlobe.

163

Figure S23. Scatter plots for each day of the first campaign (September 2017), as a function of BC
coating thickness (CT
BC
) and rBC count mean diameter. Each point represents a one-minute mean
value.

164
Appendix B: Supplement to Chapter 3
DID method
The advantage of the DID method is that it can control for both (1) time-varying
confounding factors, and (2) inherent differences between the control and impact group. In a typical
before versus after comparison, there is no control for time-varying confounding factors. One
example of a time-varying confounding factor in the context of our study is meteorology, which may
be different in pre-event and post-even periods. Likewise, in a typical control versus impact
comparison without any before versus after comparisons, there is no control for inherent
differences between the two groups of measurements. One example in the context of our study
would be the difference in land cover features in the control versus impact area. By using the DID
method, the pitfalls of the two single-differencing methods are avoided by applying an additional
round of differencing.
Like all quantitative methods, the DID method has its set of drawbacks and does not
necessarily generate perfect results, but it does provide a more robust estimate of the direct impact
of a treatment/event compared to single differencing methods. There are two main assumptions
that are built in to the DID method.
First, the characteristics of the control and the impact groups must remain constant
throughout the duration of the study. In the context of this study, this means that the control and
impact areas could not be modified in any significant way, aside from the cool pavement installation
itself. There were no major land use changes or modifications to any residential structures that we
observed during the duration of our measurements.
Second, the parallel trends assumption must be met. This assumption states that the pre-
treatment trends for both the control and impact groups must be the same. In the context of this
study, this means that the pre-treatment changes in temperatures in both the control and impact area
165
over a given time period are approximately equal. As an illustration, the null effects found in figure 5
for the pre-installation date pair shows that the parallel trend assumption was satisfied. The null
effects show that the change in temperature for different time periods during the pre-installation
phase were approximately equal for both control and impact areas.

Details regarding DID calculations
Here we summarize the procedures used to determine the DID’s for both mobile and
stationary T
air
measurements. For the mobile T
air
measurements, the first step of the data processing
involved detrending the background diurnal temperature trend from the raw data. In order to do
this, we first performed a simple linear regression using stationary T
air
data from the control area, for
the time period covering each discrete mobile transect. A single mobile transect is defined here as
one round of mobile measurements, first through the impact area and then through the control area.
Each mobile transect was generally ~20-30 minutes long. We assumed that the averages between all
the stationary T
air
measurements in the control area would be representative of the general
background temperature for the local area. After the slope for the linear regression line was found
for each mobile transect, the raw mobile T
air
data were detrended using this slope. This detrending
process was repeated for all mobile T
air
measurements. Figure s5 illustrates this detrending process
for a mobile transect performed on 26 Oct 2019, starting at 12:00 LST.
After the linear detrending was complete, a bootstrapping approach was used to calculate the
estimated mean DID and 95% confidence interval for each date pair and each representative time
period (i.e., 06:00, 09:00, …, 18:00, 21:00). In short, bootstrapping is a procedure that involves the
estimation of a sampling distribution, using random sampling of the raw data with replacement. In
our case, we estimated the DID sampling distribution using 1000 iterations (N = 1000). The
mean DID results for 1.6 m mobile T
air
are shown in figure s6. Upon close examination, we found
166
that only results for 12:00 LST were statistically reliable. The three pre-installation date pairs (top left
of figure s6) show that only 12:00 LST exhibited a null effect in the pre-installation period. This
means that 12:00 LST was the only hour for which micrometeorological variability within the
timescale of one mobile transect was not confounding the DID estimate. Since the expected air
temperature reduction signal was so small (relative to the spatiotemporal variability of temperature
within one mobile transect), we could not distinguish from these measurements whether the cool
pavement reduced temperatures for the other hours. The 1.6 m mobile T
air
result presented in the
manuscript is from the 23 Sep 2019 and 25 Oct 2019 date pair (see row 3 and column 2 from figure
s6). A similar procedure was conducted for the T
surface
measurements, except detrending was not
conducted before bootstrapping the DID estimates. This is because the variability in surface
temperature within a single mobile transect was very small compared to the change in surface
temperature caused by the cool pavement.
For the stationary T
air
measurements, a slightly different procedure was used to estimate the
DID and associated 95% confidence intervals (see figure 6). The first step was to calculate hourly
averages from the 5-minute raw data. This resulted in an hourly time series of T
air
for each stationary
air temperature sensor (see figure s4 for reference). This means that for each discrete datetime of the
time series, there would be nine hourly averages (one for each sensor). Then, Δ (i.e., impact minus
control) was calculated for each distinct control/impact sensor pair. There were five control area
sensors and four impact area sensors, so this resulted in 20 distinct control/impact sensor pairs and
therefore, 20 distinct Δ values for each datetime in the hourly time series. The Δ values were then
further subset into either Δ
pre
or Δ
post
, where the subscript denotes whether the measurements were
taken pre-installation or post-installation. Finally, the mean Δ
pre
and Δ
post
values were calculated for
each hour of day. The errors bars in the vertical and horizontal directions (see figure 6) represent the
95% confidence intervals of Δ
pre
and Δ
post
, respectively, for each hour of day. No detrending was
167
necessary for stationary T
air
because all sensors took measurements on a 5-minute sampling interval
at the exact same time.

168
Supplementary figures and tables

Figure s1. Albedo spot measurement locations. Seven locations shown in total. Map data ©2020
Google.

169

Figure s2. Photos of instrumentation set-up. Panel (a) shows the set-up for stationary albedo
measurements. Panel (b) shows the set-up for mobile albedo measurements. Panel (c) shows the set-
up for mobile transect measurements of 1.6 m air temperature and surface temperature. Panel (d)
shows the set-up for stationary 3 m air temperature. The sensor is highlighted in the red circle.

170

Figure s3. Map showing location of stationary air temperature measurements. The red-tinted area
(right) indicates the impact area, and the blue-tinted area (left) indicates the control area. The
markers show the locations of the stationary temperature monitors. Note: T3 is not present on the
map because it was stolen in the middle of active monitoring. Map data ©2020 Google.

171

Figure s4. Visual illustration of how raw T
air
data from mobile transects are detrended. Detrending
removes the background diurnal temperature change signal.

Figure s5. Matrix of barplots showing the mean 1.6 m air temperature DID for unique pairs of
dates. The columns identify the first date used in the DID calculation and the rows identify the
second date. The barplots outlined in red show the DID values that represent the impact of cool
pavement installations on surface temperature. A negative DID value indicates a surface temperature
reduction.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

From tugboats to trees: investigating the coupled systems of urban air pollution and meteorology

PDF

Impacts of heat mitigation strategies and pollutant transport on climate and air quality from urban to global scales

PDF

Identifying and mitigating the effects of urban heat islands in California

PDF

Evaluating the role of energy system decarbonization and land cover properties on urban air quality in southern California

PDF

Characterization of black carbon: from source to evolution of physical and optical properties in the atmosphere

PDF

Residential electricity demand in the context of urban warming: leveraging high resolution smart meter data to quantify spatial and temporal patterns…

PDF

Impact of urban source emissions on ambient particulate matter (PM) composition and associated toxicity in local and regional scales using source appointment models

PDF

Investigating the role of climate in affecting residential electricity consumption through high spatiotemporal resolution observations

PDF

Developing a replicable approach for the creation of urban climatic maps for urban heat island analysis: a case study for the city of Los Angeles, California

PDF

Analysis of sources and profiles of organic carbon in ambient particulate matter across fine and coarse sizes and introducing an optical technique for real-time urban dust measurement

PDF

Developing frameworks to quantify the operational and environmental performance of energy systems within the context of climate change

PDF

Chemical and toxicological characteristics of particulate matter in urban environments with a focus on its sources, associated health impacts and mitigation policies

PDF

Mitigating the urban heat island effect: thermal performance of shade-tree planting in downtown Los Angeles

PDF

Integrating systems of desalination and potable reuse: reduced energy consumption for increased water supply

PDF

Investigating the role of urban emission sources on redox-active PM compounds and the chemical analysis of the standardized diesel exhaust particles

$Hydraulic fracturing and the environment: risk assessment for groundwater contamination from well casing failure$

PDF

Hydraulic fracturing and the environment: risk assessment for groundwater contamination from well casing failure

PDF

Toxicity of urban particulate matter: long-term health risks, influences of surrounding geography, and diurnal variation in chemical composition and the cellular oxidative stress response

Asset Metadata

Creator Ko, Joseph (author)

Core Title Beyond greenhouse gases and towards urban-scale climate mitigation: understanding the roles of black carbon aerosols and the urban heat island effect as local to regional radiative forcing agents

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Engineering (Environmental Engineering)

Degree Conferral Date 2022-12

Publication Date 01/05/2023

Defense Date 12/01/2022

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag aerosols,air quality,anthropogenic heat,black carbon,climate change,OAI-PMH Harvest,urban climate,urban heat island

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sanders, Kelly (committee chair), Childress, Amy (committee member), Silva, Sam (committee member)

Creator Email joseph.ko.91@gmail.com,kojoseph@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC112710939

Unique identifier UC112710939

Identifier etd-KoJoseph-11400.pdf (filename)

Legacy Identifier etd-KoJoseph-11400

Document Type Dissertation

Format theses (aat)

Rights Ko, Joseph

Internet Media Type application/pdf

Type texts

Source 20230111-usctheses-batch-1000 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

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

Repository Email cisadmin@lib.usc.edu