University of Southern California Dissertations and Theses

The impact of definition criteria on mapped wildland-urban interface: a case study for ten counties along the Oregon-California border

(USC Thesis Other)

The impact of definition criteria on mapped wildland-urban interface: a case study for ten counties along the Oregon-California border

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright © 2022 Dakota Alexander Slaton

The Impact of Definition Criteria on Mapped Wildland-Urban Interface:

A Case Study for Ten Counties along the Oregon-California Border

by

Dakota Alexander Slaton

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

August 2022

ii
Acknowledgements
I am grateful for the guidance and feedback from my advisor Dr. Sedano and Dr. Ruddell and
my 594 classmates. I would like to thank my family and friends for their support and interest in
GIS. I am also thankful for the suite of data compiled and made available by various government
agencies and the people who support that mission.

iii
Table of Contents
Acknowledgements ......................................................................................................................... ii
List of Tables .................................................................................................................................. v
List of Figures ................................................................................................................................ vi
Abbreviations ............................................................................................................................... viii
Abstract .......................................................................................................................................... ix
Chapter 1 Introduction .................................................................................................................. 10
1.1. Defining the WUI .............................................................................................................12
1.2. Motivation .........................................................................................................................14
1.3. Study Area ........................................................................................................................15
1.4. Project Overview ..............................................................................................................24
Chapter 2 Related Work................................................................................................................ 26
2.1. Vegetation and Population-Based WUI Delineation ........................................................26
2.2. Identifying the WUI with Indirect Evidence of Human Presence ....................................30
2.3. Sociology of WUI Communities ......................................................................................32
Chapter 3 Methods ........................................................................................................................ 36
3.1. Methods Overview ............................................................................................................37
3.2. Selection of Variable Thresholds ......................................................................................39
3.3. Data ...................................................................................................................................40
3.3.1. NLCD Pre-processing ..............................................................................................45
3.3.2. Census Data Pre-Processing ....................................................................................46
3.4. Identifying the WUI ..........................................................................................................47
3.4.1. Calculation of Wildland Vegetation Cover ..............................................................47
3.4.2. Dasymetric Mapping ................................................................................................52
3.4.3. WUI Outputs ............................................................................................................54
iv
Chapter 4 Results .......................................................................................................................... 57
4.1. Dasymetric Mapping Results ............................................................................................57
4.2. Comparing Input Parameters ............................................................................................61
4.3. Evaluating WUI Outputs...................................................................................................64
Chapter 5 Conclusion .................................................................................................................... 75
5.1. Identification of Vulnerable Communities .......................................................................75
5.2. Long-Term Functionality ..................................................................................................82
5.3. Limitations ........................................................................................................................83
5.4. Next Steps and Future Research .......................................................................................84
References ..................................................................................................................................... 86
Appendix ....................................................................................................................................... 90

v
List of Tables
Table 1. Federal Register WUI types ............................................................................................ 13

Table 2. Population totals for study area by county and 20-year population change ................... 19

Table 3. WUI identification criteria and thresholds from previous studies .................................. 40

Table 4. Project data ..................................................................................................................... 41

Table 5. Land cover type and reclassification values ................................................................... 49

Table 6. Complete WUI parameter matrix ................................................................................... 61

Table 7. Federal Register WUI communities identified at each variable threshold ..................... 65

Table 8. Federal Register WUI community population change 2000-2020,
based on available US Census Bureau Data ................................................................................. 66

vi
List of Figures
Figure 1. Study area and major population centers ....................................................................... 18

Figure 2. Federal Register listed WUI communities .................................................................... 20

Figure 3. National Interagency Fire Center (NIFC) Fire Perimeters 2000-2018 .......................... 23

Figure 4. Generalized WUI identification workflow .................................................................... 39

Figure 5. Hierarchy of census geographic entities (via census.gov) ............................................. 43

Figure 6. NLCD pre-processing workflow ................................................................................... 46

Figure 7. Census block pre-processing workflow......................................................................... 47

Figure 8. Calculation of wildland vegetation cover workflow ..................................................... 50

Figure 9. Wildland vegetation landcover percentage by block group .......................................... 51

Figure 10. Dasymetric mapping workflow ................................................................................... 54

Figure 11. Appending population density attributes ..................................................................... 55

Figure 12. Populated census blocks used as input for dasymetric mapping ................................. 58

Figure 13. Federal and state lands used as input for dasymetric mapping ................................... 59

Figure 14. Final occupied census block groups ............................................................................ 60

Figure 15. Comparison of WUI outputs at ≥10 % wildland vegetation
compared with ≥75 % wildland vegetation ................................................................................. 63

Figure 16. WUI at ≥1 household/60 acres and ≥10% wildland vegetation cover,
with Federal Register WUI communities ..................................................................................... 68

Figure 17. WUI at ≥1 household/20 acres and ≥75 % wildland vegetation cover,
with Federal Register WUI communities ..................................................................................... 69

Figure 18. WUI at ≥1 household/40 acres and ≥50 % wildland vegetation cover,
with Federal Register WUI communities ..................................................................................... 70

Figure 19. WUI at ≥1 household/40 acres and ≥10 % wildland vegetation cover,
with Federal Register WUI communities ..................................................................................... 71

vii
Figure 20. Optimal WUI at ≥1 household/400 acres and
≥25 % wildland vegetation cover, with Federal Register WUI communities .............................. 73

Figure 21. Optimized WUI with proportion of population below poverty income ...................... 77

Figure 22. Wildfire hazard potential by block group .................................................................... 79

Figure 23. Most vulnerable WUI areas based on fire risk and poverty status .............................. 81

Figure 24. WUI output with all households and ≥10 % wildland vegetation cover ..................... 90

Figure 25. WUI output with all households and ≥25 % wildland vegetation cover ..................... 91

Figure 26. WUI output with all households and ≥50 % wildland vegetation cover ..................... 92

Figure 27. WUI output with all households and ≥75% wildland vegetation cover ...................... 93

Figure 28. WUI at ≥1 household/60 acres and no wildland vegetation criteria ........................... 94

Figure 29. WUI at ≥1 household/60 acres and ≥10% wildland vegetation cover ........................ 95

Figure 30. WUI at ≥1 household/60 acres and ≥25% wildland vegetation cover ........................ 96

Figure 31. WUI at ≥1 household/60 acres and ≥50% wildland vegetation cover ........................ 97

Figure 32. WUI at ≥1 household/60 acres and ≥75% wildland vegetation cover ........................ 98

Figure 33. WUI at ≥1 household/40 acres and no minimum wildland vegetation criteria ........... 99

Figure 34. WUI at ≥1 household/40 acres and ≥10% wildland vegetation cover ...................... 100

Figure 35. WUI at ≥1 household/40 acres and ≥25% wildland vegetation cover ...................... 101

Figure 36. WUI at ≥1 household/40 acres and ≥50% wildland vegetation cover ...................... 102

Figure 37. WUI at ≥1 household/40 acres and ≥75% wildland vegetation cover ...................... 103

Figure 38. WUI at ≥1 household/20 acres and no minimum wildland vegetation criteria ......... 104

Figure 39. WUI at ≥1 household/20 acres and ≥10% wildland vegetation cover ...................... 105

Figure 40. WUI at ≥1 household/20 acres and ≥25% wildland vegetation cover ...................... 106

Figure 41. WUI at ≥1 household/20 acres and ≥50% wildland vegetation cover ...................... 107

Figure 42. WUI at ≥1 household/20 acres and ≥75% wildland vegetation cover ...................... 108
viii

Abbreviations
ACS American Community Survey
BLM Bureau of Land Management
FEMA Federal Emergency Management Agency
GIS Geographic information system
MAUP Modifiable areal unit problem
MRLC Multi-Resolution Land Characteristics
NIFC National Interagency Fire Center
NLCD National Land Cover Database
WUI Wildland-Urban Interface
USDA US Department of Agriculture
USDI US Department of Interior
USFS US Forest Service
USGS US Geological Survey
ix
Abstract
Large and severe wildfires have become the norm in many parts of the western US, including the
region along the Oregon-California border. As populations in this area continue to grow, they
encroach on undeveloped land with abundant wildland fuels and high fire risk. Communities that
inhabit this wildland-urban interface (WUI) are increasingly imperiled as climate change
exacerbates catastrophic fire activity. While previous, national-level studies have established a
methodological baseline for WUI identification using vegetation and population density data, the
impacts of variable criteria on small-scale study areas remains under investigated. This is a key
area of concern because the adequate identification of WUI communities is a vital first step for
effective public policy decision making, emergency planning, and resource allocation. This
project attempts to bridge the current research gap by analyzing the impact of vegetation and
population variable parameters on the size and character of identified WUI areas for ten counties
along the Oregon-California border. This analysis is used to generate an optimal WUI definition
for the project area, which defines the WUI as census block groups with ≥1 household/400 acres
and ≥25 % wildland vegetation cover. This project finds that, compared with previous national-
level studies, a much lower population density threshold is necessary to adequately identify
plausible WUI communities. This study also supports previous findings, which indicate that
vegetation density thresholds are of secondary importance when compared to population density.
These findings are of interest to land managers who are tasked with resource allocation and
wildland firefighting in the WUI, along with residents who inhabit these communities.

10

Chapter 1 Introduction
In recent years residents of western states have come to begrudgingly expect a fifth season in late
summer. Driven by catastrophic, landscape-level wildfires, the “smoky” season typically
includes weeks of poor air quality and the evacuation of rural communities, followed by multi-
year rebuilding efforts. Though different regions are spared in a given year, this new paradigm,
endemic to the Anthropocene in the era of climate change, is annually consistent. One only needs
to skim the headlines in late summer to read an updated report of acres burned and homes lost in
a given year’s mega-fire. Fire is, of course, ecologically indigenous to the western landscape.
Natural and anthropogenic fire has long been a fact in many environments across the globe,
leading to a proliferation of fire-adapted species and ecological communities reliant on routine,
low intensity fires (Pyne 1997). What is unique about the current fire paradigm is the abundance
of people and infrastructure vulnerable to fire and the increased frequency of severe, large-scale
fire events resulting from a hotter, drier climate and alteration to historic fire regimes.
Wildland-urban interface (WUI) communities are located where “humans and their
development meet or intermix with wildland fuels” (USDA and USDI 2001, 752-753). WUI
communities across the US have seen sustained growth in recent decades (Hammer, Stewart, and
Radeloff 2009) and have emerged as the fastest growing land use type in the country (Radeloff et
al. 2018). With their close proximity to abundant, fire-prone fuels, WUI communities and homes
frequently receive the brunt of wildfire impacts. While drier and hotter conditions continue to
increase the frequency and severity of wildfires across the west, wildfire impacts to homes and
communities have become a standard feature of the annual fire season. As communities grapple
with frequent and severe fire events, the onus for prevention and risk reduction on private
property is placed on individual home and landowners (Downing et al. 2022). The presence of
11

homes and related infrastructure also complicates firefighting efforts (Radeloff et al. 2018) and
increases the probability of catastrophic outcomes for individuals and communities. Between
1990 and 2014, the number of homes lost in wildfires in the US increased by 300% (Downing et
al. 2022). This issue has also led to a marked increase in fire suppression costs by government
entities that are tasked with protecting private property adjacent to public lands (Mell et al.
2010). Likewise, wildfire risk reduction and fire prevention has seen increased political
prominence in recent years, as worsening fire conditions near populous western communities
have garnered national attention.
Understanding the location and geographic extent of the WUI is a fundamental issue for
wildfire planning, hazard mitigation, and related resource allocation. For instance, fuels
reduction programs are a common method of wildfire hazard mitigation by land management
agencies. Logistically, it is often not feasible to uniformly conduct these treatments across an
entire landscape, so identifying high priority areas, such as the WUI or especially vulnerable
portions of the WUI, is imperative (Wimberly, Zhang, and Stanturf 2006). Though the general
definition of the WUI is easily understood, the question of how to actually demarcate WUI from
non-WUI locations is much more complicated. Methodological approaches to WUI identification
vary, but they typically combine population and landcover data or, for smaller scale studies,
remote sensing and object detection. WUI identification is often the preliminary step for a wider
analysis of emergency preparedness or risk assessment.
Extant research has provided a solid methodological framework that combines vegetation
and population data, but the impact of variable thresholds on mapped WUI locations remains
under-investigated. For instance, a given proportion of wildland vegetation cover is a common
WUI identifier, but the selected proportion (e.g. 20% versus 40% of the landcover in a given
12

area) might significantly impact the geographic extent of the area defined as WUI. This is
especially true for smaller, regional study areas where variable characteristics may differ from
the national norm established in large scale studies. This project attempts to bridge this research
gap by investigating the impact of land cover and population density variable thresholds on WUI
identification for ten counties along the Oregon-California border. By testing variable threshold
combinations, this study attempts to identify the relative impacts of vegetation and population
density and determine the optimal combination of variables for the project area.
1.1. Defining the WUI
Across the western US, much of the land that abuts private property in the WUI is
administered by various federal, state, and local government agencies. Federal land management
agencies administer the largest swathes of public wildland. The five largest federal land
managers (Department of Defense, Bureau of Land Management (BLM), US Forest Service
(USFS), National Park Service, and Fish and Wildlife Service) collectively administer over 600-
milion acres of the country. These agencies administer 52.3% of the state of Oregon and 45.4%
of California (Congressional Research Service 2020). Often, federal agencies and their state
counterparts are tasked with wildfire risk mitigation, prevention, and emergency response. In the
1980s, the USFS, along with the National Fire Protection Association and the Federal
Emergency Management Agency (FEMA), formed a partnership to address fire and the WUI.
This partnership, which later drew collaboration with the BLM and the National Association of
State Foresters, held the first conference dealing with fire and the WUI in 1986 (Cortner and
Gale 1990). National, state, and local governments continue to prioritize WUI community
planning and resilience. As recently as 2021, the Bipartisan Infrastructure Bill allocated funding
13

to improve the fire regime condition and restore 10,000,000 acres of land in the WUI by 2027
(Public Law 117–58, Sec. 40803. Wildfire Risk Reduction).
Defining and identifying WUI communities is a fundamental task for emergency
planning and resource allocation. Definitions vary slightly, but ultimately describe the same core
phenomenon (Stewart et al. 2007). An official definition was established in a 2001 Federal
Register article: “the urban wildland interface community exists where humans and their
development meet or intermix with wildland fuel” (USDA and US Department of Interior
(USDI) 2001, 752-753). The Federal Register publication identifies three categories of WUI:
interface, intermix, and occluded (Table 1). Interface communities are those that abut wildland
fuels, with a “clear line of demarcation” between the two (2001, 753). These areas are defined as
having a density of three or more structures per acre or a population density of 250 or more per
square mile. Intermix WUI lacks a “clear line of demarcation” and has more intermixing of
wildland fuels and development (2001, 753). Structure density can be as low as one structure per
40 acres or 28-250 people per square mile. Occluded WUI occurs where structures abut an island
of wildland fuel in otherwise developed areas, such as an urban park. The Federal Register
definition has become the baseline for WUI research because it provides a standardized,
quantifiable measure.
Table 1. Federal Register WUI types
WUI Type Definition Criteria
Interface Settlement borders wildland fuels ≥3 structures/acre or ≥250 people/sq mi
Intermix
Settlement mixed with wildland fuels, no
separation
≥1 structure/40 acres or 28-250 people/sq
mi
Occluded Island of fuels surrounded by settlement No specific criteria

14

In the western US, the geographic extent of the WUI has increased dramatically over
time. Between 1990 and 2000, the WUI in western states grew 25%; collectively, this comprised
45% of housing units in the region (Hammer, Stewart, and Radeloff, 2009). This trend has
continued into the 21st century. The WUI collectively increased by 25-million residents between
1990 and 2010. In that same span of time, the number of houses within fire perimeters grew
62%, to 286,000 (Radeloff et al. 2018). Though the long-term impacts of the 2020 COVID-19
pandemic are still playing out, WUI growth has almost certainly escalated in the first years of the
decade. Outmigration, prompted by remote work opportunities and rising housing costs in urban
areas, has led to an influx of new residents into smaller communities (Frey 2022).
1.2. Motivation
Previous GIS-based WUI identification projects often focus on determining and
operationalizing the criteria for identifying the WUI at a national or regional scale. Though a
useful exercise for broad decision-making and emergency planning, national level WUI mapping
does not account for nuanced local differences. Factors like local economic history, geography,
and land management practices have varying degrees of impact on local settlement patterns and
community makeup, and therefore, the contemporary WUI. The social and demographic makeup
of WUI communities can also vary greatly, which has real-world ramifications for fire resiliency
and economic capacity. Because of the nuanced nature of WUI communities and fire behavior
across the country, localized, smaller-scale mapping projects have the potential to yield greater
insight into a specific area, making them a valuable tool to improve wildfire planning. Small,
geographic regions provide an optimal area to test and improve up the federal definition and
criteria established in national-level studies.
15

Though federal and state agencies are directly involved in fire suppression and fuels
reduction, the responsibility for risk mitigation in the WUI is shared with homeowners and local
communities. Recent research demonstrates that the majority of wildfires in the US that burn
structures begin on private property before spreading across ownership lines onto USFS land
(Downing et al. 2022). The authors suggest that, while fire is inevitable, private landowners and
homeowners are “the actors best positioned to reduce fire risk to homes and other high-value
assets” (2625). It is also the case that most wildfires are caused by humans. Between 1992 and
2012, the majority (84%) of wildfires in the US were anthropogenic (Balch et al. 2017).
Common sources of anthropogenic fire include arson, debris burning, campfires, and equipment
or vehicle use (NIFC n.d.). Though human-caused wildfires do start in more remote areas, it is
inevitable that the majority occur along the WUI where people and their activities are more
concentrated. Fire in the WUI has been a topic of increased importance in recent years,
especially as fire seasons have become more severe and long-lasting across much of the western
US. This issue is not unique to the American West. Recent summers have seen above-average
fire severity in Siberia (Patel 2021), the Mediterranean (Abnett 2021), and Australia (Gramling
2021), among other parts of the world. Continued population growth and sprawl into WUI areas,
along with the deleterious impacts of climate change, indicate that proactive WUI identification
and resource allocation will be essential for public planning and protecting communities and
their residents.
1.3. Study Area
The ten counties selected for this project encompass the entirety of the Klamath-Siskiyou
region and incorporate portions of the southern Willamette Valley and the very northern portion
of California’s Central Valley. The project area is bounded by the Cascade Mountains in the east
16

and the Coast Range in the west. These north/south trending ranges are connected by the
Umpqua and Klamath Mountains, which include many subranges such as the Calapooya,
Siskiyou, Marble, Scotts, Salmon, Russian, Trinity Alps, and Yolla-Bolly mountains. Much of
the project area is characterized by rugged topography with most larger settlements occupying
the Umpqua, Rogue, Klamath, and Upper Sacramento River valleys. Climatically, the area
occurs at the junction of numerous ecological zones, including the Great Basin, Sierra Nevada,
Cascade Mountains, Central Valley, and Coastal temperate zone. The region experiences a
generalized west-east rain gradient, with higher precipitation along the Pacific coast, generally
tapering off as weather systems move inland and over mountainous terrain. The rugged
topography and high relief create complex, localized climatic zones, as well as numerous
vegetation communities and microclimates, all of which impact localized fire regimes (Morton
2017). Portions of the project that lie east of the Cascade mountains, such as eastern Siskiyou,
are given a Köppen Climate Classification of “Humid Continental Climate - Dry Cool Summer.”
Most of the project area that lies between the Cascades and Coast Ranges is categorized as
“Warm-Summer Mediterranean Climate” or “Hot-Summer Mediterranean Climate” (Kottek et
al. 2006). These areas experience comparatively hot, dry summers, and tend to have the highest
burn risk and most frequent fire intervals.
This study area straddles the Oregon-California border (Figure 1). Together, the ten
counties studied encompass 30,746 square miles and approximately 800,000 residents. In
addition to many smaller communities, the main urban spheres are Eureka-Arcata and Redding
on the California side and Medford-Ashland-Grants Pass and Roseburg on the Oregon side. This
area has mostly seen sustained population growth through the first decades of the twenty-first
century (Table 2). Population growth is most pronounced in counties with larger urban centers,
17

while more rural counties such as Siskiyou and Coos counties have seen a slower increase. Only
one county, Del Norte, has experienced a slight decrease in residents over the last twenty years.
Curry county has the lowest population, at just over 23,000. Jackson county is the most
populous, with over 223,000 residents.
18

Figure 1. Study area and major population centers
19

Table 2. Population totals for study area by county and 20-year population change
County State
2000
Population
2010
Population
2020
Population
Population Change
Del Norte CA 27509 28565 27221* -1.04 %
Humboldt CA 126476 134353 136002* 7.50 %
Shasta CA 162889 177248 178271* 9.40 %
Siskiyou CA 44281 44962 44612* 0.74 %
Trinity CA 13031 13811 13635* 4.63 %
Coos OR 62779 63043 63315 0.85 %
Curry OR 21137 22364 23005 8.83 %
Douglas OR 100399 107607 112530 12.08 %
Jackson OR 181269 203206 223240 23.15 %
Josephine OR 75726 82713 86560 12.51 %
* 2018 population
Source: State of Oregon (2021) and California State Association of Counties (2019)
For context, a 2001 Federal Register listing identified 23 WUI communities in the project
area. Though the identification methods are not clarified, and the list is clearly incomplete for
anyone familiar with the area, it nonetheless serves as a convenient starting point for
contextualizing the region (Figure 2). In Oregon, the listed WUI communities include Applegate,
Ashland, Merlin, Murphy, Ruch, Sam’s Valley, Shady Cove, and Williams. On the California
side, Burney, Dorris, Dunsmuir, Etna, Fort Jones, Happy Camp, Hayfork, Hoopa, Klamath,
McCloud, Mount Shasta, Redding, Weed, Willow Creek, and Yreka were identified (USDA and
USDI 2001).
20

Figure 2. Federal Register listed WUI communities

21

According to the US Census Bureau’s 2021 population estimates, the mean poverty rate
for the project area is 15.1%. Del Norte County has the highest poverty rate of 18.5%, while the
lowest is 11.9% in Jackson County. Median annual household income ranges from $57,139 in
Shasta County to $41,780 in neighboring Trinity County. The largest proportion of the
population, 78.6%, is White alone, not Latino or Hispanic. Hispanic or Latino residents are the
next largest demographic group within the project area, accounting for 11.2% of the population.
In descending order, the next largest demographic groups are people of two or more races,
Native Americans, Asian Americans, and Black or African Americans.
Historically, the region was typified by frequent, naturally occurring, low-intensity fires.
Beginning in the late 19th century, this regime has been interrupted by fire suppression policies
(Frost and Sweeney 2000). Routine landscape-level burning by Native American groups prior to
colonization was also ubiquitous across Oregon (Boyd 2019) and California (Marks-Block et al.
2021). Seasonal low-severity burning of the landscape helped to improve habitat for game
animals and plant foods while maintaining a more fire-resilient landscape with less fuel buildup.
Following the establishment of land management agencies like the USFS and BLM in the early
twentieth century, wildfire suppression became the common approach to land management.
Suppression of low-severity burns, in addition to myriad other factors, laid the groundwork for
comparatively infrequent, but much more severe fire seasons.
The project area has seen a number of large fires in recent decades including many high
profile, destructive fires that have ravaged homes and communities. According to the National
Interagency Fire Center’s (NIFC) “Wildland Fire Perimeters” dataset, over 5.2 million acres
burned in the project area between 2000 and 2018 (Figure 3). Recent fires of note include the
2018 Carr Fire, which burned into the outskirts of Redding, California. According to CalFire, the
22

Carr Fire ultimately killed three people, torched 229,651 acres, and destroyed over 1,600
structures. On the Oregon side of the project area, similarly large and catastrophic fires have
occurred in recent years. Early in the century, the 2002 Biscuit Fire burned over 500,000 acres,
setting records as the largest fire in Oregon history (LaLande 2022). In early September 2020, a
number of fires occurred. These included the Obenchain, Almeda, Slater, Archie Creek, and
Thielson. Together, the Labor Day fires burned over 300,000 acres and destroyed more than
3,000 structures.
23

Figure 3. National Interagency Fire Center (NIFC) Fire Perimeters 2000-2018

24

1.4. Project Overview
The overarching hypothesis of this project is that the choice of thresholds set for the
levels of vegetation coverage and population density, which serve to demarcate areas as “WUI”,
has an impact on the extent and location of areas identified as WUI, and further that, because
both vegetative and population landscapes vary from place to place, different thresholds might be
appropriate in different regions for designating the WUI. Though the 2001 Federal Register
listing provided a consistent definition, the effect of vegetation and population criteria remains
under-investigated for small study areas. The effects of variable thresholds are potentially
significant and may be the source of under- or over-representation of the WUI in a given area.
This could have substantial consequences if a given WUI definition is the basis for public
planning and resource allocation. This impact inevitably differs based on local variables, and is
therefore best assessed in relatively small, homogenous study areas. Because of the importance
and sensitivity of WUI areas, a more nuanced understanding of how selection criteria can impact
WUI designation is necessary for better GIS-based identification. This is especially true given
the nuanced, mosaic quality of WUI, demonstrated by the suite of sociological research
described in the next chapter. To better understand the relationship between thresholds for WUI
criteria and the specific portions of a landscape which are designated as WUI according to a
chosen set of thresholds, this project tests different WUI variable thresholds, adjusting and
combining wildland vegetation cover and population density proportions. The goal of this
research is to identify the optimal criteria and thresholds for WUI identification in the project
area.
More specifically, this study attempts to answer the following questions: How does the
areal extent of WUI differ when the proportion of wildland vegetation landcover is adjusted?
25

What impact does the population density threshold have on the WUI extent? Which combination
of landcover and population density parameters is most suitable for the project area? A number
of smaller hypotheses can also be tested. For example, it is reasonable to suspect that a lower
population density threshold may better represent (or over-represent) higher income WUI areas
by including households on larger plots of land. Conversely, a higher population density
threshold may better account for lower income WUI areas such as rural trailer parks or
residences associated with smaller plots of land. This could be especially significant as a way to
better incorporate interface homes on the outskirts of larger communities, which may not be
associated with large land holdings despite occupying a wildland interface. Similarly,
adjustments to the wildland vegetation cover part of the equation could reasonably be expected
to have a diminishing level of return. Wildland vegetation covers as low as 10% have proven
efficacy in other studies (Hanberry 2020), but those effects inevitably vary based on local
environmental factors. Portions of the project area, especially nearer the Pacific Ocean, have
comparatively dense vegetation due to abundant moisture and a relatively moderate climate. For
these areas, then, a higher vegetation threshold may be ideal to definitively rule out non-wildland
areas. Conversely, interior valleys dominated by sparse chaparral or juniper scrub may require a
lower vegetation threshold to adequately identify wildland areas. The findings of this project,
regardless of which pattern they support, would be valuable to land managers and other
researchers who are interested in the relationship described above. The findings could help
inform public policy and resource allocation, along with community planning efforts and
emergency preparedness.

26

Chapter 2 Related Work
This project integrates methodology from a selection of previous WUI identification projects.
Previous scholarship can be broken down into a few broad, overlapping categories. First are the
studies that have set the groundwork for the standard approach to WUI designation, which
integrate vegetation and population density data to identify areas that meet the Federal Register
definition. The second body of research incorporates various technical and methodological
improvements for identifying aspects of a landscape that support a WUI designation, such as
dasymetric mapping and object detection to more accurately locate settlements and apportion
populations across a landscape. A final section provides greater context for the project by
describing the sociological nature of WUI communities.
2.1. Vegetation and Population-Based WUI Delineation
The most common approach to WUI identification combines population and vegetation
data to identify communities that fit the minimum Federal Register thresholds for settlements
and wildland vegetation. Radeloff et al. (2000) provide an early example of GIS-based
integration of vegetation and population data, setting the groundwork for ensuing WUI-oriented
research. Recognizing the potential of GIS for data integration and interdisciplinary research, the
authors combine census block data with Landsat imagery to assess the relationship between
housing density and vegetation for a region in northern Wisconsin. Methodologically, the study
uses total housing and land area at the census block level to calculate housing density. The
housing data is overlaid with a classified Landsat raster. Spatial patterning of population density
is first visually assessed before the population density data is converted to a raster at the same
resolution as the Landsat data. The two raster layers are then overlaid and the housing density
and land cover classification for each cell is determined. Though not specifically WUI-related,
27

this project provides a methodological groundwork that the authors expand upon in later WUI
identification efforts.
Developing on this previous study, Radeloff et al. (2005) undertake a national-level
assessment of the WUI using a similar methodology. The study also incorporate a sensitivity
analysis of WUI mapping criteria using California, New Hampshire, and North Carolina as test
samples. The authors overlay census block-level housing data with land cover data from the
National Land Cover Database (NLCD), calculating the housing density and percentage of
wildland vegetation cover for each block. The result is then considered against the 2001 Federal
Register definitions for interface and intermix WUI. Next, a sensitivity analysis of housing
density and vegetation cover is performed for the three states. The test varies the vegetation and
housing density criteria to assess changes in the WUI output. The results indicate that the choice
of housing density threshold tends to have a more significant impact on amount of land
designated as WUI than the threshold of vegetation cover. Of the three sample states, this finding
is especially true for California. The authors have two additional findings that are significant for
the methodology developed in the project herein. First, they find that, nationally, intermix WUI
accounts for over 80% of the WUI. The second finding is that the WUI proportion by state
follows an east-west gradient, where WUI accounts for a higher proportion of the eastern area of
each state.
The approach developed in Radeloff et al. (2005) has been improved upon and validated
in follow up studies. For example, Stewart et al. (2007) runs through essentially the same
methodology as the Radeloff et al. (2005) study, using updated datasets. A similar sensitivity
analysis is conducted, this time for seven sample states: California, Colorado, Florida, Michigan,
North Carolina, New Hampshire, and Washington. As with the previous study, vegetation and
28

housing density parameters are adjusted above and below the Federal Register thresholds. The
findings indicate that adjusting housing density thresholds has the most statistically significant
impact on the WUI output, with Florida and California especially sensitive to these adjustments.
Overall, the authors find that operationalizing the Federal Register WUI definition leads to
plausible outcomes and that altering individual parameters did not significantly alter the
“prevalence or pattern” of WUI at a national level (Stewart et al. 2007, 206).
The Modifiable Areal Unit Problem (MAUP) refers to an issue inherent in overlaying
spatial data. Depending on the size and location of a polygon, the individual data that it
encompasses will differ. If census blocks are used to determine population density, as is the case
with WUI identification, the size and extent of the census polygon is somewhat arbitrary. If
shifted or expanded, the population density could differ significantly (Bolstad 2017, 392). A few
generalized approaches to addressing the MAUP have been suggested. First, if possible, data
should be addressed individually instead of at the aggregate level. A second approach is to
optimize zoning by minimizing differences within individual aggregations. Finally, a third
recommendation is to conduct a sensitivity analysis by comparing aggregations at different levels
(Bolstad 2017, 393).
Zhang and Wimberly (2007), address the MAUP issue inherent in using census data for
identifying WUI areas, basically selecting the latter two options. As the authors describe, “the
hierarchal way census data are organized provides researchers with various potential scales of
analysis. However, it is not immediately clear which scale is the best for specific research
questions” (Zhang and Wimberly 2007,139). Regarding methodology, the authors aggregate
census data at the county, tract, census block group, and census block. These census aggregations
are then used to compare WUI growth in southern states between 1990 and 2000. Significantly,
29

the study uses housing density instead of population density to account for secondary homes,
which may be underrepresented when using population. The study demonstrates a key point:
county-level census data provides a comparatively coarse WUI output and is most suitable for
looking at broader regional trends. The census block level was found to be most suited to more
nuanced for landscape-level analysis that requires the inclusion of demographic data (145-146).
The smallest possible level of aggregation, then, provides the most detailed and accurate WUI.
Dasymetric mapping is “a geospatial technique that uses information such as land cover
types to more accurately distribute data that have been assigned to selected boundaries like
census blocks” (US Environmental Protection Agency 2022). Phrased differently, dasymetric
mapping is an additional set of steps that categorically excludes geographic portions of an area
based on a specified criterion. Many of the above studies implicitly integrate dasymetric
mapping by counting only wildland fuel portions of a given study area, but more overt and
intentional dasymetric mapping steps can be added to the WUI identification workflow.
Wilmer and Aplet (2005, as cited in Greetan 2016) integrate a dasymetric mapping step
that is especially apt for the study area in this project. The authors, as part of their workflow for
identifying the WUI for three case study areas, omit public land from the census blocks before
calculating population density. This is a key step because, if much of the land area within a
census block is public land, the population density will skew lower than it actually is. If a
minimum housing density of ≥1 household/40 acres is used, census blocks with already low
population densities may not be included at all if they have enough public land within their
boundaries. By only calculating population density for residential portions of the census block,
the authors are able to get a much more accurate measure.
30

2.2. Identifying the WUI with Indirect Evidence of Human Presence
The process of WUI mapping is iterative, with successive studies providing small
improvements and alterations, adding to the established methodology. This section describes
recent studies that rely upon data that suggests the presence of humanity rather than or in
addition to direct population data via the census.
Greetan (2016), in developing a WUI identification methodology for Lassen County,
California, incorporates more extensive dasymetric mapping steps than previous studies. In lieu
of census data, the author uses county cadastral data to identify residential plots. The use of
residential parcels allows for the inclusion of only demonstrably occupied areas. This is an
improvement on previous studies which ascribe a uniform population density across an entire
census block. The integration of cadastral data is not possible for all studies, due to data
accessibility and privacy issues, but the higher resolution is well-suited to very fine scale studies.
Compared with census block data, residential parcel-level data allows for a more accurate WUI
output, especially when paired with quality land cover data. Greetan’s (2016) workflow also
integrates vegetation data by including a .5-mile buffer around residential areas in the final WUI
output (to account for the community’s footprint) and omitting non-wildland NLCD data.
Though census-based population data combined with landcover is the most common
approach to WUI identification, a number of other methods have been successfully employed.
For example, Johnston and Flannigan (2018) take a novel approach to mapping the Canadian
WUI based on the presence of infrastructure rather than population data. The authors add greater
nuance to the standard WUI by distinguishing between three types of interfaces. The “wildland-
industrial interface (or WUI-Int)” occurs where industrial “values”, like oil and gas facilities,
electricity stations, and industrial areas, intermingle with wildland fuels. The “wildland-
31

infrastructure interface (or WUI-Inf)” similarly incorporates infrastructural values like
powerlines, roads, and transmission lines. Finally, “wildland-human interface” is selected based
on the presence of public or private structures like homes, hospitals, and railway stations.
Identification of these areas was accomplished via CanVec+, a government produced public data
source that includes structure information. Methodologically, these areas are rasterized, buffered,
and overlain with landcover data to identify each WUI type. This project is unique because it
distinguishes different WUI types based on remote sensing data. By integrating building
presence data, rather than assuming a census block is uniformly populated, the study may be
better able to account for actual residential habitation for human interface areas. For example,
small rural communities may cluster within a fraction of the area of a census block, but if
population data alone is used without further dasymetric steps, the entire block is assumed to be
inhabited. The effects would be less pronounced in core urban or suburban locales, but may be
significant in rural settings. Further, distinguishing industrial and infrastructural interfaces can
generate a more useful end product, which is ultimately the goal of WUI mapping. The
identification of key economic resources allows for better proactive planning and risk reduction
and, as the authors demonstrate, is well-suited for national-level mapping.
Caggiano et al. (2016) offer an alternate approach to WUI mapping. Using National
Agriculture Imagery Program (NAIP) imagery, the authors utilized semi-autonomous object-
based image extraction to identify structures within ten small (5.28 km x 6.94 km) sample blocks
in northern Colorado. In comparing their results with county structure data, the study concludes
that this approach is potentially more accurate than census data based approaches at similarly
small scales. The authors argue in favor of this approach by detailing some of the shortcomings
of census based WUI mapping. The central concern is inaccuracy due to census data being based
32

on the aggregation of individual population data (e.g. households or persons) into a census block.
Though dasymetric mapping steps can help remedy this issue, the authors argue that, at small
scales, object-based image extraction is more accurate.
In a recent study, Hanberry (2020) advocates a more deductive approach by using
previous fire locational data to identify and classify WUI areas. Methodologically, the project
utilizes a fire occurrence database, in conjunction with vegetation and decadal census data, for
classifier-based statistical modeling. The author demonstrates that fire is statistically likely to
occur in areas with as little as 10% vegetation cover. Given the Federal Register WUI threshold
of at least 50% vegetation cover, this is a key finding, because depending on nearby vegetation
type, land cover data may not adequately capture fire risk. For example, lower elevation
chapparal may appear sparsely vegetated when using NLCD data but is comparatively high fire
risk relative to higher elevation conifer forests which appear more densely vegetated. The study
also reaffirms previous findings that indicate housing density is a more sensitive determinant of a
WUI designation than vegetation density in ecologically homogenous areas.
2.3. Sociology of WUI Communities
To better understand what a map of the WUI represents, it is important to have a nuanced
understanding of the sociological characteristics of those communities. The WUI is not a
uniform entity, but instead a complex patchwork with varying demographic, economic, and
cultural contexts, which act together to influence response and adaptive capacity to wildfire risk.
Paveglio et al. (2009) provide a useful starting point for contextualizing these differences. The
authors demonstrate that WUI communities are better understood as a “mosaic,” which are
distinguished by their demographic and social context. They determine that a community’s
adaptive capacity is the result of its demography, access to information, place-based knowledge,
33

and informal relationships. Differences in these criteria between communities results in varying
capacities and approaches to firewise planning, even in geographically similar contexts.
Winkler et al. (2007) add greater nuance to this discussion, addressing the shifting “social
landscapes” of the intermountain west. The authors distinguish between “old west” and “new
west” communities. Old west communities are characterized by a continued reliance on
extractive industries like logging, mining, and ranching. New west communities, in contrast, are
typically associated with in-migration of more affluent and educated populations and a shift
towards natural resource tourism, conservation, and development. Discussing trends more
broadly in the west, Paveglio et al. (2015) add an important caveat to the new west demographic
shift. These changes are not uniform across the landscape. The influx of the new west and
associated demographic and economic trends is concentrated in areas that are “particularly scenic
or resource rich…with cultural and recreational resources” (300). The expansion and sprawl of
WUI areas away from city centers was also found to exacerbate urban poverty and reify greater
sprawl in the southeast US. Cho et al. (2012) examined this relationship, showing the impacts of
white outmigration. Outmigration to WUI communities and homes has the dual effect of drawing
money out of cities and shifting businesses into exurban areas, both of which have the potential
to exacerbate urban poverty.
Collectively, these sources suggest broad demographic shifts in WUI areas, in which
higher income residents move away from city centers, increasing urban poverty, and supplant or
change existing rural communities. Per Paveglio (2015), this relationship is not equal across
space but is concentrated in high amenity areas. The relationship between this demographic force
and fire preparedness is not overtly clarified but is alluded to in various studies. For example,
Bright and Burtz (2006b) look at differences in perceptions and behaviors about fire-wise
34

activities for year-round versus seasonal residents in northern Minnesota. The authors find that
seasonal residents are less likely to engage in and have favorable opinions of firewise activities.
This would suggest that permanent residents, with deeper local knowledge, may have greater
personal investment in wildfire preparedness. Though an overt connection between year-round
residency and economics was not made in this study, seasonal residents were largely comprised
of people who own summer vacation homes in the community, suggesting a degree of economic
privilege.
This body of research suggests that higher income, new west communities may be less
invested in and knowledgeable of local fire issues. Interestingly, though, the opposite
relationship is suggested in other research. For example, Paveglio et al. (2015) describe
community archetypes and their WUI adaptive capacity. This study finds that “working
landscape/ resource dependent WUI communities,” which closely align with the old west
communities described by Winkler et al., have the highest “local ecological knowledge”, but
tend to be “highly independent and distrustful of the government” (306). Individual and
community political ideology undoubtedly plays a role in firewise planning. This is shown in
Bright and Burtz (2006a), which notes the influence of individualistic values, personal property
rights, and personal freedom on defensible space. This would, presumably, suggest greater
reluctance to follow government advice about fuels reduction and firewise planning. Typical
advice, such as removing vegetation, replacing roof material, and landscaping with native plants
(Radeloff et al. 2018) would likely be met with relatively greater resistance, and perhaps, less
economic capacity.
By focusing on an explicitly spatial workflow, this project is unable to integrate a full
sociological synthesis of WUI communities in the project area. The above body of research,
35

though, suggests some of the broader demographic trends that occur within various WUI
communities. These trends undoubtably occur, to some degree, within the current project area
and could be investigated further in future research.

36

Chapter 3 Methods
As demonstrated in the previous chapter, previous WUI assessments have generated a series of
overlapping methodological approaches and refinements. The scope and methods employed in
this project attempt to combine relevant methods developed in previous studies, while avoiding
potential downfalls. WUI identification typically combines two key data types: vegetation and
population. As Zhang and Wimberly (2007) demonstrate, this can be done at different
hierarchical scales, depending on the requirements and logistics of the study. The geographic
scope of this project is local, focusing on a discrete area with common ecological and
demographic characteristics. This scale allows for a more nuanced level of data relative to
national or state-level WUI mapping, which is in direct response to the MAUP issue inherent in
large scale WUI mapping. The overall goal of the project is to evaluate different land cover and
population density thresholds for identification of the WUI in the study area.
It is impossible to fully integrate every refinement and method described in the previous
chapter. Many key findings, though useful, may apply only to a specific study area and remain
untested for other geographic locations. For example, Radeloff et al. (2005) find that California
is most sensitive to changes in housing density thresholds, compared with New Hampshire and
North Carolina. This finding is echoed in Stewart et al. (2007), which finds Florida and
California more sensitive to housing density adjustments compared to other sample states. This is
an interesting finding, but its significance is unclear. Would this result hold true across other
larger states when compared with smaller states in the northeast? Because California and Florida
are large, diverse, densely populated states, is this finding the result of an inherent MAUP issue
when using state-level aggregations? If, for example, the same relationship was examined at a
county level across California, would this finding be true across the board or is it skewed by
37

more populous urban nodes? Radeloff et al. (2005) also find that the proportion of WUI land
relative to the state was higher on the east coast compared to the west coast. Is this due to the fact
that states on the east coast are smaller and have less public land? Would the relationship hold
consistent if dasymetric mapping was added to the workflow, omitting public land in western
states?
3.1. Methods Overview
This project seeks to better understand how WUI mapping parameters impact the extent
and character of a mapped WUI output. As described earlier, the way a study or project defines
and identifies the WUI can have significant downstream impacts, with real-world consequences
if that data is used to support public planning and resource allocation. Ten counties on either side
of the Oregon-California border were selected as a case study for this project. These counties
encompass the entirety of the Klamath-Siskiyou region and surrounding areas, which, due to
their geographic proximity, share many characteristics such as climate, fuel type, and settlement
patterns. This relative homogeneity across the project area allows for a sort of control variable,
where regional differences are less likely to skew results compared with national or statewide
WUI mapping.
Methodologically, this project overlays data layers showing vegetation cover and total
household density to test a series of WUI variable thresholds. Household density was used as the
population input and was derived from the total number of households per block group and the
total geographic area of residential land per block group. As argued in Stewart et al. (2007),
housing is a better suited metric for human habitation when compared with other potential
measures like population density. This is due, in part, to the importance of structure protection in
WUI fire suppression and the growth of housing density compared to population. Depending on
38

the family makeup of an area, total households may represent different population totals.
Households, though, measure the number of housing units, which is a better metric when looking
at the WUI. Because most construction in rural areas is single family housing, a household can
be reasonably expected to be a single structure, with associated outbuildings like sheds or
garages.
Based on previous research, several additional steps were added to this basic workflow.
This project incorporates robust dasymetric mapping steps to refine the extent of areas that meet
the vegetation and population density criteria. This additional step is not included in Stewart et
al. (2007) or Radeloff et al. (2005) but is particularly well-suited to the project area. As described
in Section 3.4.2, this step allowed for the exclusion of large swathes of federal and state land, in
addition to unoccupied census blocks. This step is especially appropriate for western states where
a large proportion of the total and area is administered by federal and state agencies and is, by
definition, non-residential. Omitting these plots of land from the workflow before calculating
housing density generated a more accurate household density measure, in turn enabling a more
accurate assessment of WUI criteria.
The entirety of this workflow was completed via ArcGIS Pro using various cartographic
and geoprocessing tools (Figure 4). Data for this project, described more extensively later in this
chapter, was projected using the North American Albers Equal Area Conic coordinate system.
This is the default projection for the land cover and census data used in the analysis. State
ownership data was reprojected into this coordinate prior to analysis. Generally, the methods
employed in this project condense and aggregate WUI identification data to the census block
group level. Land cover data was clipped and reclassified into wildland and non-wildland
vegetation types, before being summarized based on census block groups. With the requisite land
39

cover and population data appended, census block groups were next whittled down to only
populated areas in a dasymetric mapping step. Finally, WUI selection criteria were used to select
and export block groups that met the appropriate criteria. WUI selection criteria tested for
wildland vegetation cover ranged from 0-75%, while household density thresholds of ≥1
household/20 acres, /40 acres, and /60 acres were tested. The output WUI from each criteria
combination was then assessed based on its total population, area, relative poverty status, and
identification of Federal Register WUI communities.

Figure 4. Generalized WUI identification workflow
3.2. Selection of Variable Thresholds
Variable thresholds for the workflow described above were determined based on previous
sensitivity analyses detailed in Stewart et al. (2007) and Radeloff et al. (2005) (Table 3). The
standard wildland vegetation cover of ≥50% was used as the control. The value is increased and
decreased by half for comparison. A low range of ≥10% was also used, based on the findings by
Hanberry (2020). Previous studies found that household density was the most significant variable
40

for WUI identification, so a wildland vegetation threshold of ≥0% was also tested to address the
impact of household density alone. For household density, the same thresholds used in previous
studies were selected, with ≥40 acres/household as the control and ≥60 acres and ≥20 acres, a
50% increase and decrease, as test parameters. As with vegetation criteria, a test null household
density was also included to show the effects of wildland vegetation cover alone.
Table 3. WUI identification criteria and thresholds from previous studies
Author Land Cover
Criteria
Population/Housing
Density
Buffer Distance Additional Steps
Radeloff et
al. (2005)
NLCD >50%
wildland veg
cover
Census block. Minimum
density of 1 house/40
acres
<2.4 (≈1.5 mi) from
heavily vegetated
area = interface
Sensitivity analysis
of vegetation and
population density.
Stewart et al.
(2007)
NLCD >50%
wildland veg
cover
Census block. Minimum
density of 1 house/40
acres
<2.4 (≈1.5 mi) from
heavily vegetated
area = interface
Sensitivity analysis
of vegetation and
population density.
Zhang and
Wimberly
(2007)
Does not
include
vegetation
criteria
Housing density between
1/40 acres to 1/1.67 acres
No buffer distance
used, only housing
density
Test WUI output
using county, tract,
census block group,
and census block
Greetan
(2016)
Reclassified
NLCD
Buffered parcel ownership .5 mi around houses Dasymetric mapping
to omit uninhabited
areas
Hanberry
(2020)
As low as 10%
valid
Census block. Determines
high-med-low density
classes for WUI type. Low
end is <6.17 houses/km
2
None Uses fire locations
to determine WUI
location

3.3. Data
This project made use of six datasets for the identification of WUI areas (Table 4). Each
is described in more detail below.
41

Table 4. Project data
Dataset Description Format Source and Date
National Land
Cover Database
Land cover classified into 16
categories
30-meter
raster
MRLC Consortium 2019
TIGER/line Shapfile County, census block group, and
census block
Polygon US Census Bureau 2019
American
Community Survey
Demographic characteristics at
census block group level
Tabular US Census Bureau 2015-
2019
Federal Lands Land administered by federal
agencies
Polygon Esri Living Atlas of the
World 2022
California State
Lands
Land administered by the state Polygon California State Geoportal
2022
Oregon State Lands Land administered by the state Polygon Oregon Dept of Forestry
2022

Landcover data, used to identify wildland fuels, was derived from the 2019 National
Landcover Database (NLCD). This raster dataset is freely available via the Multi-Resolution
Land Characteristics (MRLC) Consortium, “a group of federal agencies who coordinate and
generate consistent and relevant land cover information at the national scale for a wide variety of
environmental, land management, and modeling applications” (MRLC, n.d.). NLCD data is
generated from Landsat imagery at a 30-meter resolution. After testing with finer resolution data,
it was determined that the 30-meter resolution of the NLCD raster was most suitable for the
current mapping scale. For example, Esri hosts a 10-meter resolution land cover dataset derived
from the European Space Agency Sentinel-2 Imagery. Though providing greater detail, the
dataset has less land cover classifications (ten instead of sixteen) and proved to be unwieldy
during geoprocessing tasks due to its comparatively large size. NLCD data is also advantageous
because of its geographic and temporal extent. The same high-resolution dataset covers the entire
continental US with the same vegetation classifications. At present, NLCD data is available in
two to three-year increments going back to 2001. The ubiquity and availability of this dataset
allows for scalability and consistency for WUI mapping across geographic and temporal extents.
42

The second key data source is the US Census Bureau, a federal agency under the
Department of Commerce. The mission of the US Census Bureau is to collect and provide
demographic and economic data about the US. This project used two Census Bureau datasets,
both made available to the public through the US Census Bureau data interface. First are
TIGER/Line shapefiles. TIGER (short for Topologically Integrated Geographic Encoding and
Referencing)/line shapefiles are “extracts of selected geographic and cartographic information
from the Census Bureau's Master Address File” (US Census Bureau 2019, 1-1). In short, they are
polygon shapefiles representing, among other things, various levels of census statistical units.
The US Census Bureau collects several types of data at different hierarchal levels (Figure 5).
Below the state and county level are census tracts. According to the US Census Bureau, these are
“relatively permanent statistical subdivisions of a county or equivalent entity,” comprised of
“1,200 to 8,000 people with an optimum size of 4,000” (2021, 4-22). The census tract is
subdivided into one or more census block groups, which contain 600 to 3,000 people. The
smallest unit of aggregation is the census block, which are subdivisions of the block group.
Census blocks are bounded on all sides by roads, city boundaries, or other geographic features.
In urban areas, they often consist of a single city block, though they can be geographically larger
in more rural locales to encompass a similar amount of people (US Census Bureau 2019). This
project used census block groups as the primary unit of aggregation, but also incorporated census
blocks as part of the workflow. The county boundaries used for the project area were also
sourced from the same dataset. For consistency with the NLCD data, this project used 2019
TIGER/Line shapefiles.
43

Figure 5. Hierarchy of census geographic entities (via census.gov)
The US Census Bureau was also the source for the demographic data, which is appended
to the block group polygons based on a shared identification code. This project specifically
incorporated American Community Survey (ACS) population and poverty data published in
2019 based on 5-year estimates from 2015-2019. The ACS is an annual survey used to
supplement the decadal census. The survey consists of various “social, economic, housing, and
demographic” questions, which are more detailed than the shorter questions asked on the
decennial census (US Census Bureau 2017, 1). The annual ACS is sent to a sample of the
population and is aggregated at various scales. For privacy reasons, the smallest level at which
ACS data is available is the census block group level. ACS data is available at 1-year and 5-year
estimates. Because ACS data is based on sampling of the general population, 5-year estimates
44

contain a larger breadth of data and are more statistically reliable, especially for less populated
areas (US Census Bureau 2022).
One key question when mapping WUI areas is the appropriate scale of the data. This
determination is inevitably made based on both logistical constraints (e.g. data size and
geoprocessing capability) and on the required level of detail for the output. For the given project
area, comprised of ten counties, there are a total of 651 census block groups, a manageable
number for use with geoprocessing tools. In comparison, there are over 44,000 census blocks for
the same area. Though a robust and detailed data source, the 44,000 plus census blocks proved to
be inefficient during geoprocessing tasks. This project, then, used census block groups as the
main areal population unit. Though geographically larger, this dataset was much more suitable
for efficient geoprocessing and provided a sufficient level of detail for WUI identification,
especially after the dasymetric mapping steps described later in the chapter. Because the census
block group is also the smallest scale for which ACS data is available, using block groups
allowed for the integration of detailed demographic data. This became important later in the
workflow to determine household density and poverty status.
The final category of data includes the miscellaneous sources for land ownership, which
were combined in the WUI mapping workflow for dasymetric mapping. Three different sources
of information were incorporated. Federal lands in the project area, administered by the BLM,
USFS, National Park Service, Bureau of Reclamation, Department of Defense, and Fish and
Wildlife Service, were sourced from Esri’s Living Atlas of the World. The Living Atlas is a
curated collection of ready-to-use spatial data provided by Esri for visualization and analysis
(ESRI n.d.). According to the source’s metadata, it was compiled using data from each of the
federal agencies it represents and was last updated earlier in 2022. Though federal land data can
45

be acquired from other sources, the Esri provided data was preferable because the various
datasets were already aggregated into one comprehensive feature class. State land ownership was
sourced separately for Oregon and California. The Oregon ownership data was provided by the
Oregon Department of Forestry. This shapefile was last updated earlier in 2022 and depicts state
owned forest lands. For California, state ownership data was sourced from the California State
Geoportal, which provides centralized, authoritative spatial data for the public. This dataset was
also last updated in 2022. Notably, it also includes county and city ownership categories. Only
state ownership was included in this project because a similarly authoritative data source for
Oregon was not available.
3.3.1. NLCD Pre-processing
A number of pre-processing steps were taken to prepare each dataset for further analysis.
The workflow to process the land cover data had two main steps (Figure 6). The first step was to
limit the data to the study area boundary. First, the study area counties were selected from the
county TIGER/line shapefile and exported as a separate layer. The raster data was then clipped to
the project area using the Clip Raster tool. NLCD data is only available at a much larger extent
than is needed for this study, so this initial step was necessary to whittle down the raster to a
more manageable size. The Clip Raster tool maintains individual raster cells, but consequently
does not match the raster to the exact clipping input, so additional steps are necessary. The
clipped raster was run through the Extract by Mask tool. This tool is similar to the Clip Raster
tool, but the output raster is defined exactly by the mask, in this case the project area counties.
The Extract by Mask tool also differs from the Clip Raster tool because it resamples cells along
the edge of the mask. According to the Esri technical guidance for the tool: “when the input
mask is feature data, cells in the input raster whose center falls within the perimeter of the feature
46

will be included in the output, while cells whose center falls outside it will receive No Data”
(ESRI n.d.b). For certain types of raster analysis, this may represent a potential issue. For this
project, a sufficient number of cells with vegetation values remained in each catchment (census
block group) that the effect of resampled null values was negligible when the mean vegetation
value was calculated in later steps.

Figure 6. NLCD pre-processing workflow
3.3.2. Census Data Pre-Processing
As with the NLCD raster, the census block group data required a number of pre-
processing steps to generate a dataset with the appropriate information and geographic extent
(Figure 7). To start, TIGER/Line shapefiles for California and Oregon Census block groups were
added. The appropriate ACS tables were selected and joined with each state’s block groups
based on their matching GEOIDs. For this project, the “Poverty” and “Household, Family, and
Subfamily” ACS tables were used. Specifically, this project utilized the total household and ratio
of income to poverty level data. The latter is a measure of the total number of people per census
block whose income was below the poverty income level. Other similar poverty measures could
be used in lieu of poverty income level. This variable was selected because it provided a general
baseline to measure poverty relative to the total population. Similar ACS poverty measures
breakdown populations based on race/ethnicity, education, et cetera. While these variables could
47

be integrated into a deeper analysis of WUI demography, they were too nuanced to provide the
generalized information needed in this project. With poverty and population data appended, the
next step was to export the California and Oregon block groups, which makes the joins
permanent and allows the features to be more easily edited later. Similarly, the two block group
datasets (for Oregon and California) were clipped to the project area and merged together into a
single feature class.

Figure 7. Census block pre-processing workflow
3.4. Identifying the WUI
With data pre-processing complete, this section describes the overarching workflow
developed to identify WUI areas. Methodology pulled from Radeloff et al. (2005), Stewart et al.
(2007), Zhang and Wimberly (2007), Greetan (2016), and Hanberry (2020).
3.4.1. Calculation of Wildland Vegetation Cover
With the block group data and the NLCD raster prepared, the next step in the process was
to use both datasets as inputs for the Zonal Statistics as Table tool to calculate wildland
vegetation cover proportions by block group. First, the NLCD raster needed to be reclassified
based on cover type. As produced by the MLRC, NLCD raster cells are assigned one of sixteen
landcover classifications. For this workflow, the raster was simplified into Boolean categories of
non-wildland and wildland fuel. The categorization of land cover type borrows from Radeloff et
48

al. (2005) and Stewart et al. (2007). Both studies are fairly intuitive in their classifications,
defining any developed land, including agricultural land, as non-wildland. Open water, ice, and
bare rock/sand/clay are also considered non-wildland vegetation land cover. The remaining land
cover classes are defined as wildland. Using the 2019 NLCD data, the following categories were
defined as non-wildland fuels: open water, perennial ice/snow, developed open space, developed
low intensity, developed medium intensity, developed high intensity, pasture/hay, cultivated
crops, barren land (rock/sand/clay), and no data. All of these categories are present in the study
area. Wildland vegetation includes all remaining categories, which are deciduous forest,
evergreen forest, mixed forest, dwarf scrub, shrub/scrub, grassland/herbaceous,
sedge/herbaceous, lichens, moss, woody wetlands, and emergent herbaceous wetlands. Of these
landcover types, dwarf scrub, sedge/herbaceous, lichens, and moss do not occur in the study
area. Logistically, this recategorization was accomplished using the Reclassify Raster tool. The
numerical code for each landcover type is changed to either 0, for non-wildland vegetation, or 1,
for wildland vegetation (Table 5).
49

Table 5. Land cover type and reclassification values
Original Classification Land Cover Type Reclassification Value
11 Open Water 0
12 Perennial Ice/Snow 0
21 Developed, Open Space 0
22 Developed, Low Intensity 0
23 Developed, Medium Intensity 0
24 Developed, High Intensity 0
31 Barren Land 0
41 Deciduous Forest 1
42 Evergreen Forest 1
43 Mixed Forest 1
51 Dwarf Scrub Not Present
52 Shrub/Scrub 1
71 Grassland/Herbaceous 1
72 Sedge/Herbaceous Not Present
73 Lichens Not Present
74 Moss Not Present
81 Pasture/Hay 0
82 Cultivated Crops 0
90 Woody Wetlands 1
95 Emergent Herbaceous Wetlands 1

The next step used the reclassified raster and block groups to calculate zonal statistics. As
the name suggests, this tool calculates various statistical outputs for a dataset based on specified
zones, with the end result being a separate table that contains the specified values. In this case,
the zones were the census block groups, with the GEOID specified as the zone field for each
block group. The GEOID was used so the table output could be joined to the block group later.
The target zonal statistic was the mean of the reclassified landcover value representing wildland
vegetation in each zone. The output table, then, indicated the mean wildland vegetation
percentage for each census block group. Because non-wildland vegetation was given a
reclassified value of 0 and wildland vegetation is given a value of 1, a zonal mean of .4 would
indicate that, in the given block group, 40% of the landcover is wildland vegetation. As a final
50

step, this value was multiplied by 100 to give a whole number percentage. Finally, the new zonal
statistics table was joined to the block group dataset based on matching GEOIDs, completing this
portion of the workflow. The end result of this section of the workflow (Figure 8) was a block
group feature class with the percentage of wildland vegetation (Figure 9) and selected ACS
demographic data appended as attributes of each polygon.

Figure 8. Calculation of wildland vegetation cover workflow
51

Figure 9. Wildland vegetation landcover percentage by block group

52

3.4.2. Dasymetric Mapping
The next step of the process involved refining the block groups, with appended data, to a
more accurate geographic extent. Put simply, this was accomplished by omitting portions of the
block group that were unoccupied. Compared with other population aggregations, such as
counties and census blocks, census block groups vary greatly in size. In rural portions of the
project area (i.e., where WUI areas are most likely to occur) individual block groups are much
larger than in more urban areas. Amongst the larger, rural census block groups, much of the land
is not actually inhabited or zoned for residential use. There is the potential for overestimating the
extent of the WUI in these areas if it is assumed that the entire landscape is uniformly sparsely
populated, instead of populated with small clusters of settlement amidst large swathes of
uninhabited land. This issue underscores the importance of additional dasymetric mapping steps
that provide more nuance when added to population density and vegetation cover.
To counteract the large scale of census block group data, this project omitted portions of
the census block group based on two criteria. The first was census blocks with a total population
of zero. The second criterion was land that is, by definition, not occupied. This includes federal
and state lands administered by agencies such as the US Forest Service, the BLM, and the
Oregon Department of Forestry.
Methodologically, the dasymetric mapping was straightforward. Government-
administered state and federal land polygons were merged and given a negative 1-mile buffer.
The negative buffer excludes a 1-mile internal band within the boundary of a given polygon. The
negative buffer is used to account for wildland adjacent to private property. Due to the proximity
to residential land, adjacent public land acts as a WUI extension. Statistically, fire incidents that
occur on public land often occur near private property (and are therefore likely the result of
actions by nearby residents) (Downing et al. 2022). Previous projects typically employ a buffer
53

distance of 1 to 1.5-miles to account for the distance a firebrand can travel and ignite (e.g.,
Greetan 2016 and Stewart et al. 2007). The distance is typically based on local slope and fuel
conditions. Because this project area encompasses a large region with mixed fuel types and
topography, a 1-mile distance for the buffer was selected as a conservative estimate. Instead of
buffering the entire WUI output, as is typically done, this project applied only a negative buffer
to public land. This accomplished basically the same thing, but only incorporated areas that are
actually adjacent to wildland by targeting only public land. In comparison, buffering the census
block group would have also included more densely populated urban areas that border WUI
block groups in exurban areas. Public land data, after being merged, was laid over the census
block group dataset prepared in the previous steps. Areas of the block group that were
overlapped by the public land layer were omitted using the Erase tool.
Next, unpopulated census blocks were reincorporated using a similar workflow (Figure
10). As described earlier, census blocks are a smaller level of aggregation than the block groups.
Within each block group are numerous census blocks with different population totals. Of the
over 44,000 census blocks within the study area, approximately half are unoccupied. These
census blocks were selected and exported as their own feature class, which was then overlaid on
the block group data. As with federal and state land, the areas of the block group layer that
overlapped unoccupied blocks were erased. The rationale for this step was straightforward.
Block groups in rural areas tend to be very large, so omitting demonstrably unoccupied census
blocks allowed the block group to be whittled down to only its occupied portions.
54

Figure 10. Dasymetric mapping workflow
3.4.3. WUI Outputs
The final product of the above workflow was the dasymetric block group with appended
data. In summary, this feature class shows the geographic extent of the inhabited portions of each
census block group. With a more accurate geographic extent of the occupied portions of the
block group complete, household density was then calculated and added as an attribute to the
dasymetric block groups. Based on Stewart et al. (2007) and Zhang and Wimberly (2007),
household density was used in lieu of other population figures. Four household density attributes
were added to the census block groups (Figure 11). First, the acreage of each block group, minus
the portions erased during previous dasymetric mapping, was calculated. Next, the total number
of households per 20, 40, and 60 acres was calculated. Total household density attributes were
important when selecting blocks that meet WUI criteria, which was the next step in the
workflow.
55

Figure 11. Appending population density attributes
At this point in the workflow, each block group had all of the key attributes to identify
WUI status. This included the percentage of wildland vegetation landcover and the household
density per 20, 40, and 60 acres. The total population and total population below poverty income
were also appended to the census blocks. With this main dataset ready, WUI outputs were
generated by selecting the features with attributes that fit the WUI criteria for landcover and
population density. Because all essential data is joined to the same feature class, generating WUI
outputs only required the input of relevant selection criteria using the Select by Attributes query
builder. For example, the following input was used to select block groups with ≥10% wildland
vegetation and ≥1 household per 20 acres:
“SELECT WHERE Percent Land Cover >= 10 And Household per 20 acres >= 1”
The selected blocks that met those criteria could then be exported as a unique feature class.
To track the effect of WUI definition criteria, different combinations of landcover
percentage and population density were tested. A table was generated so that combinations of
each variable could be tracked. For each combination of vegetation cover and household density,
four attributes were recorded: the total land area, the population percentage relative to the entire
population, the proportion of the WUI population below poverty income, and the number of
56

federal register communities identified. To measure the total land area of the given WUI output,
a new attribute called “Sq Mi” was added to each WUI attribute table. The square mileage was
then calculated using the Calculate Geometry function. The population percentage relative to the
entire population was calculated by adding up the total population of each selected block group
in each WUI output. The total population of the given WUI was then compared with the total
population of the entire project area and the percentage of the population present in the selected
WUI was calculated. The same basic process was followed to determine the proportion of
residents with a monthly income below the poverty rate. The total number of residents with
below poverty income was calculated using the Statistics attribute tool and the sum calculated
against the total population of the WUI to determine the relative proportion. To identify Federal
Register WUI communities identified by each WUI output, the Select by Location tool was used.
The selection criteria identified Federal Register WUI communities that intersected the WUI
output. Because points were used for each WUI community, this tool was supplemented by
visual inspection. In some cases, only some of the block groups around a community were
identified as WUI without clearly overlapping the community point. When this occurred, visual
inspection was used to determine if the community was included in the output. Pending data
availability, future studies may opt to use community polygons for larger datasets where visual
inspection would be impractical.

57

Chapter 4 Results
This chapter describes the results of the WUI identification process. It first discusses the
dasymetric mapping results, demonstrating the effectiveness of omitting the specified areas. The
next section compares the results from combining different variable parameters, drawing
conclusions based on the performance of each variable. The third section evaluates the WUI
outputs based on their ability to adequately identify known WUI communities. Finally, the
information gained from the analysis is used to determine optimal WUI parameters for the
project area.
4.1. Dasymetric Mapping Results
Following the extensive data processing steps described in the previous chapter,
dasymetric mapping proved effective for excluding unoccupied areas. As previously described,
this is an important step prior to identifying WUI areas because it provides a better indicator of
actual residential areas and, therefore, likely household density. The first step in the dasymetric
mapping workflow allows for the exclusion of unoccupied census blocks. Of the 44,000+ census
blocks in the project area, approximately half are uninhabited (Figure 12). The next step omits
public (state and federal administered) lands, with a negative 1-mile buffer (Figure 13). In total,
these unoccupied lands accounted for approximately 18,000 square miles, out of a total project
area of approximately 31,000 square miles (Figure 14).

58

Figure 12. Populated census blocks used as input for dasymetric mapping

59

Figure 13. Federal and state lands used as input for dasymetric mapping
60

Figure 14. Final occupied census block groups

61

4.2. Comparing Input Parameters
The above methods were employed to successfully generate nineteen unique WUI extents
and populate a WUI parameter matrix (Table 6).
Table 6. Complete WUI parameter matrix
Household (hh) Density
Wildland Vegetation Cover
Most Restrictive Least Restrictive
Least Restrictive Most Restrictive

≥0% (test) ≥10% ≥25% ≥50% ≥75%
All hh
(test)
Area: 12,700mi
2
Pop: 897,090
Below Pov
Income: 16.5%
a)12,576 mi
2

b) 65.3 %
c) 15.1 %
a) 12, 482 mi
2

b) 55.4 %
c) 15.2 %
a) 12,125 mi
2

b) 43.4 %
c) 14.8 %
a) 11,040 mi
2

b) 26.9%
c) 14.4 %
1hh/60
acres
a) 3,069 mi
2

b) 87.9 %
c) 16.6 %
a) 2,945 mi
2

b) 53.3 %
c) 14.9 %
a) 2,851 mi
2

b) 43.4 %
c) 15.0 %
a) 2,494.1mi
2

b) 31.4 %
c) 14.4 %
a) 1,956 mi
2

b) 15.6 %
c) 13.2 %
1hh/40
acres
a) 2,265 mi
2

b) 84.9 %
c) 16.7 %
a) 2,141 mi
2

b) 50.3 %
c) 15.0 %
a) 2,047 mi
2

b) 40.4 %
c) 15.1 %
a) 1,713 mi
2

b) 28.5 %
c) 14.5%
a) 1,268 mi
2

b) 13.1 %
c) 13.0 %
1hh/20
acres
a) 1,205 mi
2

b) 78.7 %
c) 17.1 %
a) 1,083 mi
2

b) 44.0 %
c) 15.4 %
a) 989 mi
2

b) 34.2 %
c) 15.6 %
a) 808 mi
2

b) 23.0 %
c) 14.8 %
a) 499 mi
2

b) 8.7 %
c) 12.6 %
a) Land Area
b) Percentage of Total Population
c) Percentage of Population Below Poverty Income

Five wildland vegetation cover thresholds were tested between ≥0% and ≥75%. The
correlation is straightforward, but ≥0% was the most permissive vegetation criterion because
does not exclude any area based on vegetation cover. Though the WUI extents generated are
insufficient because they only incorporate the household density variable, the inclusion of this
null criterion yielded interesting results. Adjusting the wildland vegetation cover from ≥0% to
≥10% led to the exclusion of approximately 30% of the total population. For instance, at the ≥1
household/60 acres, adjusting the wildland vegetation criteria from 100% (≥0%) of block groups
to 90% (≥10%) led to a 34% decrease in total population. This relationship holds true across
other household density levels. At the ≥0% wildland vegetation level, the effects of household
62

density alone are most pronounced. Predictably, the land area decreases substantially as the
household density input is made more restrictive (from ≥60 acres to ≥40 to ≥20). Interestingly
though, the proportion of the total population remains relatively homogenous (87% to 78%).
This indicates that most of the block groups that have a household density of ≥1 household/60
acres also have three times that quantity i.e. ≥1 household/20 acres. The proportion of the block
group population who makes less that the poverty rate also stayed fairly static. This is likely the
result of the dasymetric mapping step, which, by omitting large geographic areas, more
accurately represents a higher household density in rural areas.
As with vegetation parameters, a null population density input was used, allowing for the
inclusion of all block groups based on the number of households per acre. Similar to the null
vegetation input, the WUI generated without a household density limit is insufficient as an
accurate rendition of the WUI but is useful for looking at the role vegetation plays in the
equation. As wildland vegetation parameters increase from ≥10% up to ≥75% a unique trend
plays out. The land area remained fairly static, decreasing by only 1,536 square miles or 8% as
the vegetation criterion became more restrictive. Across that same span, the percentage of the
WUI population decreased dramatically from 53% to approximately 16% of the total population.
This indicates that, as the vegetation criterion is increased, it effectively omits smaller, sparsely
vegetated, populous areas (Figure 15). This suggests that the ≥10% wildland vegetation
parameter might be overly permissive. Conversely, it suggests that the ≥75% wildland vegetation
parameter might be overly restrictive, potentially excluding interface WUI.
63

Figure 15. Comparison of WUI outputs at ≥10 % wildland vegetation compared with ≥75 %
wildland vegetation

64

Outside of the test null parameters, the analysis produced a total of twelve potential WUI
outputs. These test vegetation criteria from 10-75% and household density criteria between 20
acres per household and 60 acres per household. The overall, obvious trend is, as each variable
gets more restrictive, the output decreases in land area, percentage of the population, and
percentage of population below poverty income. Land area decreased most notably as household
density parameters increased. This is reasonable as, all block groups with ≥1 household/60 acres
necessarily also have at least 1 household/20 acres. The same relationship was not true when
looking at increases in wildland vegetation cover. Land area decreases were small or negligible
as vegetation cover requirements increased. As suggested above, the decrease in area was largely
in urban or suburban area, indicated by a comparatively high decrease in population as
vegetation requirements increase. This finding is supported by the previous vegetation density
map (Figure 10), which shows relative uniformity in wildland vegetation cover outside of urban
areas around cities like Medford and Redding.
4.3. Evaluating WUI Outputs
The above findings are interesting but do not readily offer a way to compare mapped
WUI results. One reasonable way to evaluate the various outputs is to compare mapped WUI
coverage to known WUI communities. As discussed in the introductory chapter, 23 WUI
communities were identified in the project area for a 2001 Federal Register listing (see Figure 2).
The original rationale for selecting these communities was not clarified and the list is
incomplete. Nonetheless, an adequate GIS-based WUI assessment ought to be able to reidentify
these known communities.

65

As with the above table detailing geographic extent and demographic character, a table
was generated to track the total number of Federal Register WUI communities identified at each
pair of variable parameters (Table 7). Generally speaking, the number of Federal Register WUI
communities identified decreases when vegetation density thresholds are increased. This effect is
more pronounced with household density, but a slight decrease from 23 to 22 occurs when
wildland vegetation cover is increased from 10% to 75% when household density is not
accounted for. Similarly weak decreases occur at all household density levels between 25 and
50% wildland vegetation cover. Household density had a strong impact on the number of Federal
Register WUI communities identified. From ≥1 household/60 acres to ≥1 household/20 acres, the
total number of WUI communities decreased by almost half across all vegetation densities, with
the strongest drop at 75% wildland vegetation cover.
Table 7. Federal Register WUI communities identified at each variable threshold
Household (hh) Density
Wildland Vegetation Cover

≥0% (test) ≥10% ≥25% ≥50% ≥75%
All hh (test)

23/23 23/23 23/23 22/23
1hh/60 acres 20/23 20/23 20/23 19/23 15/23
1hh/40 acres 17/23 17/23 17/23 16/23 13/23
1hh/20 acres 13/23 13/23 13/23 12/23 9/23

Population totals for the Federal Register WUI communities are not uniformly available
because many are technically unincorporated and often lumped into larger neighboring
communities. Available population data, sourced from the US Census Bureau, shows that some
degree of population change has occurred since 2001 (Table 8). For example, Shady Cove and
Redding both grew considerably (33% and 15%) over the twenty-year period. Conversely,
Dunsmuir and Etna both decreased by over 10% in the same period. These shifts in population
66

do not necessarily prohibit comparison with WUI identified using more recent data. Instead,
these changes indicate that the population density criteria used to identify the same WUI
communities today will be slightly different than they would have been when they were first
listed in 2001. For instance, communities that grew over time would qualify as WUI using more
densely populated household criteria (e.g. 1 household/20 acres versus 1 household/40 acres).
The opposite would be true for communities with shrinking populations.
Table 8. Federal Register WUI community population change 2000-2020, based on available US
Census Bureau Data
WUI Community 2000 Population 2020 Population Population Change
Ashland 19,522 21,413 + 9.6 %
Shady Cove 2,307 3,085 + 33.7 %
Dorris 886 863 - 2.6 %
Dunsmuir 1,923 1,705 - 12.6 %
Etna 781 677 - 11.3 %
Fort Jones 660 696 + 5.5%
Redding 80,865 93,559 + 15.7 %
Weed 2,978 2,864 + 3.8 %
Yreka 7,290 7,809 + 7.1 %

To simplify the visualization of results, this study will look at four of the twelve WUI
outputs. First, the WUI output using ≥10% wildland vegetation cover and ≥1 household/60 acres
will be assessed. This output is the most permissive in the sense that it incorporates the largest
area and population. On the opposite end of the spectrum is the WUI output generated with
≥75% wildland vegetation cover and ≥1 household/20 acres. This pair of criteria is the most
restrictive and generated the smallest WUI output, both in geographic extent and population.
Two additional WUI outputs will be tested alongside this pair. The WUI output from ≥50%
vegetation and ≥1 household/40 acres, which is the basic WUI mapping criteria, will be tested.
67

Finally, the WUI output using ≥10% wildland vegetation and ≥1 household/40 acres will be
used, as these values are around the median for area and population percentage.
The WUI output at ≥1 household/60 acres and ≥10% wildland vegetation cover
performed best when compared with the Federal Register WUI communities (Figure 16). Of the
23 listed, all but three were within the mapped WUI. The communities of Dorris, Fort Jones, and
McCloud were missed using this output. Interestingly, further experimentation indicates that this
is due to a lower population density than ≥1 household/60 acres. Of the four test outputs, the
WUI at ≥1 household/20 acres and ≥75 % wildland vegetation cover performed the worst (Figure
17). This output only overlapped nine of the 23 listed WUI communities. Experimentation and
comparison with other WUI outputs indicates that the higher vegetation threshold has little
impact on the overlap with WUI communities, instead it is due, almost entirely, to population
density. This finding is reaffirmed when comparing the results of ≥50% (Figure 18) and ≥10%
(Figure 19) vegetation, both at ≥1 household/40 acres. Both outputs are visually similar, with the
lower vegetation threshold incorporating more small block groups near cities. At ≥50%
vegetation cover, seven Federal Register WUI communities are missed (Ruch, Applegate, Fort
Jones, Dorris, McCloud, Klamath, and Etna. Adjusting the from ≥50% to ≥10% only allows for
the identification of Etna, still missing the other six communities.

68

Figure 16. WUI at ≥1 household/60 acres and ≥10% wildland vegetation cover, with Federal
Register WUI communities
69

Figure 17. WUI at ≥1 household/20 acres and ≥75 % wildland vegetation cover, with Federal
Register WUI communities
70

Figure 18. WUI at ≥1 household/40 acres and ≥50 % wildland vegetation cover, with Federal
Register WUI communities
71

Figure 19. WUI at ≥1 household/40 acres and ≥10 % wildland vegetation cover, with Federal
Register WUI communities
72

Based on these findings, the optimal WUI criteria for the project area needs a lower
population density than ≥1 household/60 acres to account for the most sparsely populated areas,
such as Dorris. The effects of the vegetation criteria have been shown to be most pronounced in
including or omitting areas near urban cores. At ≥10%, this variable may overcount non-WUI
urban areas. Conversely, a high threshold of ≥75% visibly decreases the amount of land on the
outskirts of cities, where interface WUI is also likely to occur. An optimized ≥25% wildland
vegetation would incorporate some areas near urban cores, without overestimating the extend of
WUI in those areas. For population density, a much lower criterion is necessary to incorporate
very sparsely populated areas. The community of Dorris, for example, has a population density
of ≥1 household/400 acres. This low density threshold was determined by slowly decreasing the
household density from 80, testing progressively smaller densities until the block groups around
Dorris qualified for the selection criteria. This optimized WUI (Figure 20) covers 11,860 mi
2
,
much more than any of the other outputs. This total accounts for 93% of the total land area. Even
though this proportion of the land area skews much higher than any other WUI test, it makes
intuitive sense. The study area only has a handful of population centers, with most of the area
being sparsely populated. In contrast, the total population for the optimized WUI is only
495,794, 55% of the total population. The proportion of people below poverty income in 15.1%,
similar to most of the other outputs and 1% below the proportion of the entire project area
population.

73

Figure 20. Optimal WUI at ≥1 household/400 acres and ≥25 % wildland vegetation cover, with
Federal Register WUI communities
74

Interpretation of the results can be based both on technical measures of the output, but
also on an understanding of how the WUI should reasonably be distributed in the study area.
This is one of the advantages to limiting the geographic scope to a specific area. The WUI
ultimately describes locations where human development meets wildland fuels. This area is at a
comparatively high risk of being impacted by catastrophic wildfire due to proximity to wildland
vegetation. In this specific project area, dense urban settlement occurs only in the small handful
of urban cores (see Figure 1). Outside of these communities, most residential areas sprawl
outwards from cities, along highways, and around small towns. Generally speaking, much of the
non-public land is sparsely occupied outside of cities and towns. The WUI output with the most
verisimilitude, then, would error on the side of expansive rather than restricted.

75

Chapter 5 Conclusion
This project demonstrates how a streamlined and location specific WUI mapping methodology
can be established based on existing research. The impacts of different combinations of
vegetation cover and population density were tested and an optimal WUI definition was
established for the project area. This definition employs ≥1 household/400 acres and ≥25%
wildland vegetation cover as WUI identification criteria for the given study area. Despite the
expedience of national WUI mapping efforts, this study demonstrates that standardized variable
definitions may not be uniformly optimal across the country. Before moving forward with WUI
locational data, then, it is necessary to assess whether the selected criteria are well-suited for the
project area or selected arbitrarily based on previous projects.
The heighted vulnerability of WUI communities in the age of climate change necessitates
an honest and good faith effort to adequately identify vulnerable communities, especially if their
identification is tied to resource allocation and public policy decision making. This chapter
demonstrates how the main methodology for identifying WUI can be augmented by teasing out
the differing vulnerabilities of communities within the WUI. The chapter then discusses the long-
term utility of the project’s results before considering the project’s limitations and proposing
future work that could continue this research.
5.1. Identification of Vulnerable Communities
With an optimized WUI established, the identification of especially vulnerable interface
communities is an important next step. To demonstrate this workflow, the process was
completed using the optimized WUI established in Chapter 4. First, the proportion of people with
below poverty income relative to the total population is calculated for each WUI block group.
The feature class is then symbolized using graduated colors and natural breaks, which divides the
76

feature into five categories (Figure 21). Based on the sociological context described in Chapter 1,
there is reason to believe that both higher and lower income communities may be vulnerable to
wildfire based on demographic characteristics. For demonstration purposes, vulnerable
communities will be defined as those with the highest proportion of poverty income relative to
the total population (33%+).
77

Figure 21. Optimized WUI with proportion of population below poverty income

78

Identification of the poorest communities may be sufficient to identify vulnerability, but
an additional step is to overlay fire risk data. This allows for the identification of the poorest
communities with the highest burn risk. Wildfire hazard potential data is sourced from the Esri
Living Atlas. The dataset depicts relative burn risk on a ranking of one to five, with five being
the highest (Figure 22). Data is aggregated at various levels from state and county to block group
and hex bin. For consistency across datasets, the median burn risk by block group was used. The
median burn probability data is ultimately derived from the USDA’s Fire Modeling Institute,
which represents fire risk in a 270-meter resolution raster (Holtzclaw 2020). Sourcing the data
from Esri provides a convenient shortcut, skipping extensive pre-processing steps necessary to
prepare the dataset as it comes from the USDA.
79

Figure 22. Wildfire hazard potential by block group

80

This dataset was clipped to the optimized WUI, and the union tool was used to combine
the two datasets. From here, the select by attributes tool was used to identify most vulnerable,
second most vulnerable, and third most vulnerable communities (Figure 23). The selection
criteria are as follows:
• Most vulnerable = % of population below poverty income>= 33 AND Median Wildfire
Potential Score = 5
• Second most vulnerable= % of population below poverty income >= 21 AND Median
Wildfire Potential Score = 4
• Third most vulnerable= % of population below poverty income >=13 AND Median
Wildfire Potential Score = 3
The results of this analysis indicate that the most extensive area of high burn risk and
high poverty status occurs near the community of Hoopa, California. This area includes the
reservation of the Hoopa Valley Tribe and other communities adjacent to the Six Rivers National
Forest. Additional small pockets are located around Happy Camp, California and west of
Williams, Oregon near the communities of Cave Junction and Kerby.
81

Figure 23. Most vulnerable WUI areas based on fire risk and poverty status

82

5.2. Long-Term Functionality
For long-term WUI data functionality, variable parameters should be retested using new
census and landcover data as it is released and updated WUI extents should be generated. In any
region with sustained and consistent population growth (or decline), the WUI is constantly
expanding and shifting over time. This means that any estimation of WUI extent is necessarily
approximate and, depending on the rate and nature of local growth, may be quickly outdated.
The nature of this shift may differ across an area too. For example, WUI areas that abut federal
and state land, which are common across the west, may be limited in their geographic spread. By
definition, public land is not open for residential development and represents a growth barrier.
As populations in these types of communities grow, population density rather than physical
extent is more likely to increase. The opposite may be true along the periphery of urban centers,
where new construction tends to sprawl outwards across previously undeveloped, private lands.
ACS data is available annually and updated population and demographic data can be reintegrated
into the workflow. As decadal census data is released, it can also be integrated.
In the longer term, land cover can also be expected to shift and change. Urban sprawl is
inevitably the likeliest source of land cover change, from wildland vegetation to developed, non-
wildland categories. But landcover change may also occur over time due to climate change and
wildfire events. In some areas, warmer and drier conditions may shift vegetation communities
e.g. from mixed forest to scrub. While these would both register as wildland vegetation, they
may represent different levels of fire risk. It is also feasible that severe enough burn scars may be
reclassified from forest to bare land cover types in some areas. Finally, areas previously
classified as open water or perennial snow/ice may transition to bare ground or vegetation as a
result of drier, hotter conditions. NLCD data is typically available in 2 to 3-year increments. At
83

this time, the 2019 NLCD data used in this study is the most updated version. As with population
data, newer versions should be integrated as they are released. More detailed landcover data can
also be substituted in future iterations of this workflow.
5.3. Limitations
Any GIS-based analysis is limited by data quality. This project relies extensively on
NLCD and US Census Bureau data. Both of these datasets are considered authoritative resources
and are sufficiently accurate for the project at hand. Both have some amount of inherent
inaccuracy that is worth mentioning though. The metadata associated with the 2019 NLCD
dataset clarifies that a formal accuracy assessment has not been made for that year. The previous
edition, which dates to 2016, includes an overall accuracy assessment of 91% for landcover data
for the conterminous US. Similarly, because ACS data is based on a sample of the population,
there is an intrinsic amount of statistical error when extrapolating population level conclusions.
ACS datasets also have “non-sampling error,” which are non-statistical errors that effect data
accuracy (US Census Bureau 2019b, 20). The US Census Bureau gives data entry errors and
systematic undercounting of “groups who are difficult to enumerate” (US Census Bureau 2019b,
20). The latter category is especially important when considering the role of socioeconomic
status in WUI communities. For example, undocumented farm workers and rural homeless
residents may be underrepresented in rural census data.
An additional source of potential inaccuracy can arise from mismatched census and
landcover data. This project matches 2019 census data with 2019 NLCD data. Both datasets
include data from previous years, due to the lapse of time between observation and data
publication. Direct temporal correlation between datasets is not possible but using roughly
contemporaneous data is preferable. For example, pairing 2010 census data with 2016 NLCD
84

data would likely yield decent WUI results, but more inaccuracy would be introduced because of
the temporal difference between the datasets.
Beyond data accuracy and compatibility, the nature of this project represents a potential
limitation. GIS-based research can be an invaluable tool. The project described here enables the
remote and fairly accurate identification of large expanses of WUI areas. This identification is an
important first step for a number of potential actions, but GIS-based research needs to be
integrated with other research methods to fully understand, not just where the WUI is, but what
that finding implies. For example, after identifying potential WUI locations, a reasonable next
step would be applied sociological or ethnographic research to better understand the issues and
challenges residents face and how they might interact with wildfire risk. If, for example, cost is a
major limiting factor in firesafe planning for local residents, a grant system could be established
by policy makers to better meet community needs. Alternatively, if negative perception of
government and authority is a factor (as suggested in Bright and Burtz 2006a), then rapport
building with the local community may be imperative to build trust and collaboration.
5.4. Next Steps and Future Research
With an optimized WUI definition established for the study area, a number of next steps
are possible for policy makers and future researchers. In their analysis of WUI areas across the
US, Radeloff et al. (2005) suggest that WUI locational data needs to be integrated with other
detailed spatial data, such as a building materials database, detailed topography, and
accessibility, to fully ascertain fire risk across the interface. This sentiment is echoed in Mell et
al. (2010) who argue that, to assess the effectiveness of WUI wildfire risk mitigation programs,
community characteristics and structure exposure conditions need to be integrated into a wider
WUI model. Because the WUI extent is necessarily generated at a specific time series based on
85

the census and vegetation data used in the workflow, the mapping process is well-suited to
longitudinal assessments of WUI change through time. Currently, the temporal scope is limited
by the range of NLCD data, which is available as far back as 2001. As new NLCD datasets are
released in the future, the temporal scope of WUI data will grow and be a potentially fruitful
avenue for analysis.
As is, the dataset generated here could be a valuable decision-making tool. For instance,
if any of the neighboring land management agencies were tasked with WUI adjacent wildland
fuels reduction, they could prioritize land near lower income, higher fire risk communities.
Similarly, if state or county governments received funding to construct new rural fire
infrastructure, they might identify high priority areas based on the same criteria. Individuals
within communities might also use the dataset to understand their relative risk and, potentially,
use it as an impetus for community-level planning and organization.
The full realization of WUI mapping, then, is not in the minutiae of variable selection,
but in the big picture functionality and integration of a well-designed dataset for long-term
planning and analysis. Effective research approaches can integrate quality WUI locational
information, using it as a building block to address key public planning and emergency
preparedness decisions. Because of the foundational importance of WUI locational information,
it is imperative to have a fully realized mapping methodology, such as the one developed in this
project.

86

References
Abnett, K. 2021. “Mediterranean Has Become a 'Wildfire Hotspot', Eu Scientists Say.” Reuters.
https://www.reuters.com/business/environment/mediterranean-has-become-wildfire-
hotspot-eu-scientists-say-2021-08-04/.
Balch, J., B. Bradley, J. Abatzoglou, R. Nagy, E. Fusco, and A. Mahood. 2017. “Human-started
Wildfires Expand the Fire Niche Across the United States” Proceedings of the National
Academy of Sciences 114(11): 2946-2951.
Boyd, R. 2019. “Indian Use of Fire in Early Oregon.” Oregon Encyclopedia, Oregon Historical
Society.
Bright, A. and R. Burtz. 2006a. “Creating Defensible Space in the Wildland-Urban Interface:
The Influence of Values on Perceptions and Behavior.” Environmental Management
37(2): 170-185.
Bright, A. and R. Burtz. 2006b. “Firewise Activities of Full-Time Versus Seasonal Residents in
the Wildland-Urban Interface.” Journal of Forestry 104 (6): 307-315.
Bolstad, P. 2017. GIS Fundamentals: A First text on Geographic Information Systems. 5
th
ed.
White Bear Lake, MN: Eider Press.
Caggiano, M. D., W. Tinkham, C. Hoffman, A. Cheng, and T, Hawbaker. 2016. “High
Resolution Mapping of Development in the Wildland-Urban Interface Using Object
Based Image Extraction.” Heliyon 2(e00174): 1-19.
California State Association of Counties. 2019. “Population by Jurisdiction and by County 1970
to 2018.” California State Association of Counties. Database accessed February 14, 2022.
https://www.counties.org/data-and-research.
Cho, S., S. Jung, R. Roberts, and S. Kim. 2012. “Interrelationship Between Poverty and the
Wildland-Urban Interface in Metropolitan Areas of the Southern US.” Applied
Economics 2144 (11): 1405-1416.
Congressional Research Service. 2020. “Federal Land Ownership: Overview and Data.”
Accessed July 11, 2022. https://sgp.fas.org/crs/misc/R42346.pdf
Cortner, H., and R. Gale. 1990. “People, Fire, and Wildland Environments.” Population and
Environment 11(4): 245-257.
Downing, W., C. Dunn, M. Thompson, M. Caggiano, and K. Short. 2022. “Human Ignitions on
Private Lands Drive USFS Cross-boundary Wildfire Transmission and Community
ESRI. n.d. “ArcGIS Living Atlas of the World.” Accessed September 3, 2022.
https://www.esri.com/en-us/arcgis/products/living-atlas
87

ESRI. n.d.b. “Extract by Mask (Spatial Analyst).” Access September 4, 2022.
https://pro.arcgis.com/en/pro-app/2.8/tool-reference/spatial-analyst/extract-by-mask.htm
Frey, W.H. 2022. “New Census Data Shows a Huge Spike in Movement Out of Big Metro Areas
During the Pandemic.” The Avenue, Brookings Institute, April 14, 2022.
https://www.brookings.edu/blog/the-avenue/2022/04/14/new-census-data-shows-a-huge-
spike-in-movement-out-of-big-metro-areas-during-the-pandemic/
Frost, E. and R. Sweeney. 2000. “Fire Regimes, Fire History and Forest Conditions in the
Klamath-Siskiyou Region: An Overview and Synthesis of Knowledge” Prepared for the
World Wildlife Fund, Klamath-Siskiyou Ecoregion Program, Ashland, OR.
Gramling, C. 2021. “Australian Fires in 2019–2020 Had Even More Global Reach than
Previously Thought.” Science News, September 15, 2021.
https://www.sciencenews.org/article/australia-wildfires-climate-change-carbon-dioxide-
ocean-algae.
Greetan, J. 2016. “The Wildland-Urban Interface in Lassen County, California: A Change
Analysis 2000-2015.” Master’s Thesis, University of Southern California.
Hammer, R., S. Stewart, and V. Radeloff. 2009. “Demographic Trends, the Wildland-Urban
Interface, and Wildfire Management.” Society & Natural Resources 22(8): 777-782.
Hanberry, B. 2020. “Reclassifying the Wildland-Urban Interface Using Fire Occurrences for the
United States.” Land (Basel) 9(7): 225-.
Holtzclaw, J. 2020. “Wildfire Hazard Potential Enriched with Demographics now in ArcGIS
Living Atlas.” ESRI ArcGIS Blog. https://www.esri.com/arcgis-blog/products/arcgis-
online/analytics/wildfire-hazard-potential-demographics/
Johnston L. and M. Flannigan. 2018. “Mapping Canadian Wildland Fire Interface Areas.”
International Journal of Wildland Fire 27: 1-14.
Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel. 2006. World Map of the Köppen-
Geiger climate classification updated. Meteorol. DOI: 10.1127/0941-2948/2006/0130.
LaLande, J. 2022. “Biscuit Fire of 2002.” Oregon Encyclopedia, Oregon Historical Society.
Accessed July 7, 2022.
https://www.oregonencyclopedia.org/articles/biscuit_fire_of_2002/#:~:text=With%20a%
20burned%20area%20totaling,summer%20and%20fall%20of%202002.
Marks-Block, T., F. Lake, R Bird, and L. Curran. 2021. “Revitalized Karuk and Yurok Cultural
Burning to Enhance California Hazelnut for Basketweaving in Northwestern California,
USA.” Fire Ecology 17(6): 1-20. https://doi.org/10.1186/s42408-021-00092-6
Mell, W., S. Manzello, A. Maranghides, D. Butry, and R. Rehm. “The Wildland-Urban Interface
Fire Problem – Current Approaches and Research Needs.” International Journal of
Wildland Fire 19: 238-251.
88

Morton, S. 2017. “Spatial and Temporal Variability of Fire Regimes Across the Biodiverse
Klamath-Siskiyou Ecoregion, Northern California and Southern Oregon, USA.” Master’s
Thesis, University of Colorado.
MRLC. n.d. “Multi-Resolution Land Characteristics (MRLC) Consortium.” Accessed June 28,
2022. https://www.mrlc.gov/
NIFC. n.d. “Wildland Fire Investigation: Common Wildfire Causes.” Accessed August 1, 2022.
https://www.nifc.gov/fire-information/fire-prevention-education-mitigation/wildfire-
investigation
Patel, K. 2021. “Wildfires and Smoke in Siberia.” NASA Earth Observatory. NASA, April 22,
2021. https://earthobservatory.nasa.gov/images/148238/wildfires-and-smoke-in-siberia.
Paveglio, T., C. Moseley, M. Carroll, D. Williams, E. Davis, and A. Fischer. 2015. “Categorizing
the Social Context of the Wildland Urban Interface: Adaptive Capacity for Wildfire and
Community ‘Archetypes.’” Forest Science 61(2): 298–310.
Paveglio, T., P. Jakes, M. Carroll, and D. Williams. 2009. “Understanding Social Complexity
Within the Wildland-Urban Interface: A New Species of Human Habitation?”
Environmental Management 43: 1085-1095.
Pyne, S. J. 1997. Fire in America: A Cultural History of Wildland and Rural Fire. University of
Washington Press: Seattle.
Radeloff, V., D. Helmers, H. Kramer, M. Mockrin, P. Alexandre, A. Bar-Massada, V. Butsic, T.
Hawbakere, S. Martinuzzia, A. Syphardf, and S. Stewart. 2018. “Rapid Growth of the US
Wildland-Urban Interface Raises Wildfire Risk.” Proceedings of the National Academy
of Sciences 115(13): 3314-3319. https://doi.org/10.1073/pnas.1718850115.
Radeloff, V., R. Hammer, S. Stewart, J. Fried, S. Holcomb, and J. McKeefry. 2005.“The
Wildland-Urban Interface in the United States.” Ecological Applications 15(3): 799-805.
Radeloff, V., A. Hagen, P. Voss, D. Field, and D. Mladenoff. 2000. “Exploring the Spatial
Relationship Between Census and Land-Cover Data.” Society & Natural Resources
13(6): 599-609.
State of Oregon. 2021. “County Populations.” In Oregon Blue Book. Accessed February 13,
2022. https://sos.oregon.gov/blue-book/Pages/local/county-
population.aspx#InplviewHash04115801-432c-47bf-aef3-
abafd774261d=SortField%3DCounty-SortDir%3DAsc.
Stewart, S., V. Radeloff, R. Hammer, and T. Hawbaker. 2007. “Defining the Wildland-Urban
Interface.” Journal of Forestry 105(4): 201-207.
US Census Bureau. 2022. “American Community Survey 5-Year Data (2009-2020)” Accessed
August 5, 2022. https://www.census.gov/data/developers/data-sets/acs-
89

5year.html#:~:text=The%205%2Dyear%20estimates%20from,areas%20and%20small%2
0population%20subgroups.
US Census Bureau. 2019. “TIGER/Line Shapefiles Technical Documentation” Accessed July
27, 2022. https://www2.census.gov/geo/pdfs/maps-
data/data/tiger/tgrshp2019/TGRSHP2019_TechDoc.pdf
US Census Bureau. 2019b. “American Community Survey Accuracy of the Data (2019)”
Accessed August 3, 2022. https://www2.census.gov/programs-
surveys/acs/tech_docs/accuracy/ACS_Accuracy_of_Data_2019.pdf
US Census Bureau. 2017. “American Community Survey Information Guide” Accessed July 27,
2022. https://www.census.gov/content/dam/Census/programs-
surveys/acs/about/ACS_Information_Guide.pdf
US Environmental Protection Agency. n.d. “Dasymetric Toolbox.” Last modified July 26, 2022.
Access August 3, 2022. https://www.epa.gov/enviroatlas/dasymetric-
toolbox#:~:text=Dasymetric%20mapping%20is%20a%20geospatial,solely%20on%20the
%20census%20data.
USDA and USDI. 2001. “Urban wildland interface communities within vicinity of Federal lands
that are at high risk from wildfire.” Federal Register 66: 751–777.
Wilmer B. and G. Aplet. 2005. “Targeting the Community Fire Planning Zone: Mapping
Matters.” The Wilderness Society, Washington, D.C.
Wimberly, M., Y. Zhang, and J. Stanturf. 2006. “Digital Forestry in the Wildland-Urban
Interface.” In Computer Applications in Sustainable Forest Management, 201–222.
Dordrecht: Springer Netherlands.
Winkler, R., D. Field, A. Luloff, R. Krannich, and T. Williams. 2007. “Social Landscapes of the
Inter-Mountain West: A Comparison of ‘Old West’ and ‘New West’ Communities.”
Rural Sociology 72(3): 478–501.
Wong D. 2009. “The Modifiable Areal Unit Problem (MAUP).” In The SAGE Handbook of
Spatial Analysis, edited by A. Fotheringham and P. Rogerson, 105-122. Thousand Oaks,
CA: SAGE Publications Inc.
Zhang, Y., and M. Wimberly. 2007. “The Importance of Scale in Using Hierarchical Census
Data to Identify the Wildland-Urban Interface.” Southern Journal of Applied Forestry
31(3): 138–147.

90

Appendix

Figure 24. WUI output with all households and ≥10 % wildland vegetation cover
91

Figure 25. WUI output with all households and ≥25 % wildland vegetation cover
92

Figure 26. WUI output with all households and ≥50 % wildland vegetation cover
93

Figure 27. WUI output with all households and ≥75% wildland vegetation cover

94

Figure 28. WUI at ≥1 household/60 acres and no wildland vegetation criteria
95

Figure 29. WUI at ≥1 household/60 acres and ≥10% wildland vegetation cover
96

Figure 30. WUI at ≥1 household/60 acres and ≥25% wildland vegetation cover
97

Figure 31. WUI at ≥1 household/60 acres and ≥50% wildland vegetation cover
98

Figure 32. WUI at ≥1 household/60 acres and ≥75% wildland vegetation cover

99

Figure 33. WUI at ≥1 household/40 acres and no minimum wildland vegetation criteria
100

Figure 34. WUI at ≥1 household/40 acres and ≥10% wildland vegetation cover
101

Figure 35. WUI at ≥1 household/40 acres and ≥25% wildland vegetation cover
102

Figure 36. WUI at ≥1 household/40 acres and ≥50% wildland vegetation cover
103

Figure 37. WUI at ≥1 household/40 acres and ≥75% wildland vegetation cover

104

Figure 38. WUI at ≥1 household/20 acres and no minimum wildland vegetation criteria

105

Figure 39. WUI at ≥1 household/20 acres and ≥10% wildland vegetation cover
106

Figure 40. WUI at ≥1 household/20 acres and ≥25% wildland vegetation cover
107

Figure 41. WUI at ≥1 household/20 acres and ≥50% wildland vegetation cover
108

Figure 42. WUI at ≥1 household/20 acres and ≥75% wildland vegetation cover

Abstract (if available)

Abstract Large and severe wildfires have become the norm in many parts of the western US, including the region along the Oregon-California border. As populations in this area continue to grow, they encroach on undeveloped land with abundant wildland fuels and high fire risk. Communities that inhabit this wildland-urban interface (WUI) are increasingly imperiled as climate change exacerbates catastrophic fire activity. While previous, national-level studies have established a methodological baseline for WUI identification using vegetation and population density data, the impacts of variable criteria on small-scale study areas remains under investigated. This is a key area of concern because the adequate identification of WUI communities is a vital first step for effective public policy decision making, emergency planning, and resource allocation. This project attempts to bridge the current research gap by analyzing the impact of vegetation and population variable parameters on the size and character of identified WUI areas for ten counties along the Oregon-California border. This analysis is used to generate an optimal WUI definition for the project area, which defines the WUI as census block groups with ≥1 household/400 acres and ≥25 % wildland vegetation cover. This project finds that, compared with previous national level studies, a much lower population density threshold is necessary to adequately identify plausible WUI communities. This study also supports previous findings, which indicate that vegetation density thresholds are of secondary importance when compared to population density. These findings are of interest to land managers who are tasked with resource allocation and wildland firefighting in the WUI, along with residents who inhabit these communities.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The wildland-urban interface in Lassen County, California: a change analysis 2000–2015

PDF

Pre-incident plan mapping in Kern County's wildland urban interface

PDF

Impacts of vegetation management on wildfire severity: a study of the 2021 Caldor fire

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

Spatiotemporal hotspots of 2018-2020 crime in Houston, Texas

PDF

Exploring the pernicious effects of redlining and discriminatory policies on an American city: a spatio-temporal case study of New York City

PDF

Spatiotemporal analysis of the SLOSH and ADCIRC storm surge models: a case study of hurricane Ida

PDF

The impact of severe coastal flooding on economic recovery disparities: a study of New Jersey communities following Hurricane Sandy

PDF

Investigating bus route walkability: comparative case study in Orange County, California

PDF

A comparison of pixel-based, object-based, deep learning, and data fusion image classification approaches in delineating urban tree canopy in City Terrace

PDF

The community wildfire protection plan repository: using VGI to create a national collection of wildfire management information

PDF

A spatial and temporal exploration of how satellite communication devices impact mountain search and rescue missions in California’s Sierra Nevada mountain range

PDF

Pharmacy shortage areas across the United States: a visual representation by a web mapping application

PDF

A threat-based least-cost path decision support model for national security resource allocation along the US-Mexico border

PDF

Operational optimization model for Hungry Marketplace using geographic information systems

PDF

Building a spatial database for agricultural record keeping and management on a regenerative farm

PDF

San Diego Wildfire Hazards Information Center mashup

PDF

A comparison of urban land cover change: a study of Pasadena and Inglewood, California, 1992‐2011

PDF

Geographic information systems and marketing: a transdisciplinary approach to curriculum development

PDF

Preparing for immigration reform: a spatial analysis of unauthorized immigrants

Asset Metadata

Creator Slaton, Dakota (author)

Core Title The impact of definition criteria on mapped wildland-urban interface: a case study for ten counties along the Oregon-California border

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Degree Conferral Date 2022-08

Publication Date 09/17/2022

Defense Date 08/23/2022

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

Tag emergency preparedness,mapping criteria,OAI-PMH Harvest,wildfire,wildfire GIS,wildland urban interface,WUI

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elizabeth (committee chair), Ruddell, Darren (committee member), Wu, An-Win (committee member)

Creator Email daslaton27@gmail.com,dslaton@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC111992650

Unique identifier UC111992650

Legacy Identifier etd-SlatonDako-11206

Document Type Thesis

Format application/pdf (imt)

Rights Slaton, Dakota

Type texts

Source 20220917-usctheses-batch-980 (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