University of Southern California Dissertations and Theses

Maui’s wildland-urban interface: enhancements for the unique vegetative and agricultural landscape

(USC Thesis Other)

Maui’s wildland-urban interface: enhancements for the unique vegetative and agricultural landscape

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Maui’s Wildland-Urban Interface:
Enhancements for the Unique Vegetative and Agricultural Landscape
by
Amber Jean Birdwell
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)
May 2025
Copyright 2025 Amber Birdwell

ii
Dedication
To my parents, brother, and grandparents

iii
Acknowledgements
I am grateful to my thesis advisor, Dr. Elisabeth Sedano, and my faculty advisors, Dr. Guoping
Huang and Dr. Yi Qi, for inspiring me and assisting me writing this thesis. I would like to thank
my family and friends, who have shown me endless love and support as well as my employer,
Esri, who allowed me to take a semester off to finish this degree. I am grateful for the data
available through Maui County and the Hawaii Statewide GIS Program.

iv
Table of Contents
Dedication....................................................................................................................................... ii
Acknowledgements........................................................................................................................iii
List of Tables ................................................................................................................................ vii
List of Figures.............................................................................................................................. viii
Abbreviations................................................................................................................................. xi
Abstract......................................................................................................................................... xii
Chapter 1 Introduction .................................................................................................................... 1
1.1 Study Area .......................................................................................................................... 1
1.2 Background......................................................................................................................... 6
1.2.1 Wildfire Conditions in Maui, Hawaii ........................................................................ 7
1.2.2 The August 2023 Maui Wildfires............................................................................ 10
1.3 The Wildland-Urban Interface.......................................................................................... 13
1.4 Research Goals and Objectives......................................................................................... 15
1.5 Thesis Organization .......................................................................................................... 16
Chapter 2 Related Literature......................................................................................................... 17
2.1 Wildfire-Related Studies on Maui .................................................................................... 17
2.1.1 Maui Drought and Wildfire Studies......................................................................... 17
2.1.2 August 2023 Wildfire Studies.................................................................................. 18
2.2 WUI and Mapping Practices............................................................................................. 20
2.2.1 Building Density vs Housing Unit Density ............................................................. 20
2.2.2 WUI Mapping Frameworks..................................................................................... 21
2.2.2.1 Point-based...................................................................................................... 22
2.2.2.2 Zonal-based..................................................................................................... 23
2.2.3 WUI Definition Thresholds ..................................................................................... 24
2.2.4 Agricultural Mapping in WUI ................................................................................. 25
2.3 WUI on Maui .................................................................................................................... 25
2.3.1 Maui WUI Estimations............................................................................................ 26
2.3.2 Maui in Global WUI................................................................................................ 26
2.3.3 Maui in Hawaii WUI ............................................................................................... 28
2.3.4 Comparison to WUI in Fire-Prone Western US ...................................................... 29
2.3.5 Maui WUI Guidelines and Community Protection Plans........................................ 30
Chapter 3 Methods........................................................................................................................ 32
3.1 Workflow Overview ......................................................................................................... 32
3.2 Data Description ............................................................................................................... 33
3.2.1 Development-Oriented Datasets.............................................................................. 34
3.2.2 Landcover Datasets.................................................................................................. 35

v
3.3 WUI Mapping ................................................................................................................... 43
3.3.1 Building Density ...................................................................................................... 44
3.3.2 Vegetation Cover ..................................................................................................... 47
3.3.3 Wildland Areas ........................................................................................................ 53
3.3.4 Standard WUI Map .................................................................................................. 56
3.3.5 Forest and Grassland Cover..................................................................................... 59
3.3.6 WUI Map Enhancements with Forest Cover........................................................... 62
3.4 Refining Landcover Mapping........................................................................................... 63
Chapter 4 Results.......................................................................................................................... 67
4.1 Final WUI Map................................................................................................................. 67
4.1.1 WUI Map with Standard WUI Classifications ........................................................ 67
4.1.2 Final WUI Map with Enhanced WUI Classifications.............................................. 70
4.2 Landcover Dataset Refinement WUI Maps...................................................................... 75
4.2.1 LANDFIRE EVT Dataset WUI Map....................................................................... 76
4.2.2 LANDFIRE FVC Dataset WUI Map....................................................................... 78
4.2.3 ESA WorldCover Dataset WUI Map....................................................................... 80
4.2.4 Landcover WUI Comparisons ................................................................................. 82
4.2.5 ESA WorldCover and All Hawaii Statewide GIS Program Agricultural Data WUI
Map ................................................................................................................. 84
Chapter 5 Dashboard Creation...................................................................................................... 87
5.1 Web-Sharing Related Literature ....................................................................................... 87
5.2 Dashboard Methods.......................................................................................................... 88
5.2.1 Dashboard Workflow Overview.............................................................................. 88
5.2.2 Dashboard Data Description.................................................................................... 89
5.2.3 Adding Attributes to Census Blocks........................................................................ 92
5.2.4 Preparing Web Map ................................................................................................. 93
5.2.5 Dashboard Structure Formatting.............................................................................. 94
5.2.6 Static and Dynamic Widgets.................................................................................... 96
5.3 Dashboard Results ............................................................................................................ 98
Chapter 6 Conclusions................................................................................................................ 102
6.1 Contributions................................................................................................................... 102
6.1.1 Theoretical Contributions ...................................................................................... 102
6.1.2 Contribution to Methodology ................................................................................ 103
6.1.3 Contribution to Technology................................................................................... 105
6.2 Findings........................................................................................................................... 106
6.3 Comparison to Existing Maui WUI Maps...................................................................... 107
6.3.1 Comparison to SILVIS Lab’s Global WUI ........................................................... 107
6.3.2 Comparison to State of Hawaii’s WUI Definition................................................. 109
6.4 WUI Significance............................................................................................................ 110
6.4.1 Awareness.............................................................................................................. 110
6.4.2 Supporting Maui and Maui County Mapping Efforts............................................ 110
6.5 Dashboard Audience and Applications........................................................................... 111
6.6 Limitations...................................................................................................................... 112
6.6.1 Multiple WUI-Calculation Factors........................................................................ 113

vi
6.6.2 Zonal-Based Census Block Approach ................................................................... 113
6.6.3 Dataset Limitations................................................................................................ 114
6.7 Future Research .............................................................................................................. 115
6.7.1 Future WUI Research ............................................................................................ 115
6.7.2 Enhancing Existing Dashboard.............................................................................. 117
6.7.3 Maui County Feedback.......................................................................................... 119
References................................................................................................................................... 120

vii
List of Tables
Table 1. WUI definitions .............................................................................................................. 14
Table 2. Data descriptions............................................................................................................. 34
Table 3. Maui’s agricultural land area by agriculture type........................................................... 43
Table 4. Standard WUI classification outcomes........................................................................... 69
Table 5. ESA WorldCover WUI classification outcomes ............................................................ 72
Table 6. Land, buildings, and people in WUI............................................................................... 75
Table 7. EVT WUI classification outcomes................................................................................. 78
Table 8. FVC WUI classification outcomes................................................................................. 80
Table 9. ESA WUI classification outcomes ................................................................................. 82
Table 10. ESA and all agricultural data WUI classification outcomes......................................... 86
Table 11. Dashboard data descriptions......................................................................................... 90

viii
List of Figures
Figure 1. Maui, Hawaii study area.................................................................................................. 2
Figure 2. Rainfall in Maui............................................................................................................... 6
Figure 3. Hawaii drought conditions on August 8, 2023................................................................ 9
Figure 4. Historical wildfires on Maui from 2005-2020............................................................... 10
Figure 5. Historical sites in Lahaina before and after the wildfires.............................................. 12
Figure 6. Maui August 2023 wildfire burn extents....................................................................... 13
Figure 7. Diagram of two WUI mapping approaches. Figure by Bar-Massada 2021. ................. 22
Figure 8. SILVIS Lab – Maui WUI map ...................................................................................... 27
Figure 9. Hawaii state-defined CARs and WUI. Figure by Hawaii DLNR 2010. ....................... 29
Figure 10. Workflow diagram....................................................................................................... 33
Figure 11. Census blocks and distribution of buildings................................................................ 35
Figure 12. LANDFIRE EVT dataset ............................................................................................ 37
Figure 13. LANDFIRE FVC dataset ............................................................................................ 39
Figure 14. ESA WorldCover dataset ............................................................................................ 41
Figure 15. Agricultural land use dataset ....................................................................................... 42
Figure 16. Workflow diagram for calculating building density ................................................... 45
Figure 17. Building density........................................................................................................... 46
Figure 18. Building density distribution ....................................................................................... 47
Figure 19. Workflow diagram for calculating vegetation cover................................................... 48
Figure 20.Vegetation cover before agricultural enhancements .................................................... 49
Figure 21. Included agricultural plots........................................................................................... 50
Figure 22. Vegetation cover with agriculture considered............................................................. 51

ix
Figure 23. Vegetation cover per census block .............................................................................. 53
Figure 24. Workflow diagram for identifying wildlands and creating a wildland buffer............. 54
Figure 25. Wildland areas............................................................................................................. 55
Figure 26. Wildland buffer ........................................................................................................... 56
Figure 27. Workflow diagram creating a standard WUI map ...................................................... 58
Figure 28. Workflow diagram for forest and grassland cover...................................................... 60
Figure 29. Forest and grassland cover .......................................................................................... 61
Figure 30. Workflow diagram for enhancing the WUI map......................................................... 63
Figure 31. Taro, macadamia, and pasture agricultural land.......................................................... 65
Figure 32. Standard WUI map with agricultural enhancements................................................... 68
Figure 33. Standard WUI classification makeup .......................................................................... 70
Figure 34. Final WUI map with Maui-specific subclassifications ............................................... 71
Figure 35. Final WUI classification makeup ................................................................................ 73
Figure 36. Buildings and population by WUI classification......................................................... 74
Figure 37. LANDFIRE EVT WUI map........................................................................................ 77
Figure 38. LANDFIRE FVC WUI map........................................................................................ 79
Figure 39. ESA WUI map............................................................................................................. 81
Figure 40. WUI classification makeup per landcover dataset ...................................................... 83
Figure 41. ESA and all agriculture WUI map............................................................................... 85
Figure 42. ESA and all agricultural data WUI classification makeup.......................................... 86
Figure 43. Workflow diagram creating ArcGIS Dashboard......................................................... 89
Figure 44. Maui fire stations......................................................................................................... 91
Figure 45. CARs on Maui............................................................................................................. 92

x
Figure 46. ArcGIS Dashboard elements....................................................................................... 95
Figure 47. Maui WUI ArcGIS Dashboard.................................................................................... 98
Figure 48. Dynamic metrics on Dashboard .................................................................................. 99
Figure 49. Static portion of Dashboard....................................................................................... 100
Figure 50. Data layers for additional context.............................................................................. 101

xi
Abbreviations
CARs Communities at Risk
DLNR Department of Land and Natural Resources
ESA European Space Agency
HWMO Hawaii Wildfire Management Organization
LANDFIRE Landscape Fire and Resource Management Planning Tools
NIST National Institute of Standards and Technology
NLCD National Land Cover Dataset
WUI Wildland-Urban Interface
WUIF Wildland-Urban Interface (subcategory)
WUIX Wildland-Urban Intermix

xii
Abstract
In August 2023, Maui experienced a series of massive wildfires, including one that destroyed
most of the historic town of Lahaina. Climate change has increased the frequency and intensity
of wildfires, and the Maui wildfires exhibit the dangers of living next to wildlands. Identifying
an area’s wildland-urban interface (WUI) is a crucial part of wildfire management. To provide
insight into the relationship between development and the surrounding wildlands in Maui, this
thesis studies the spatial distribution of the WUI in Maui before the 2023 wildfires. This thesis
creates the first census block-based WUI map of Maui. Standard WUI map creation entails
determining census blocks’ structural density, vegetation cover percentage, and distance from
wildland areas. In census block-based WUI maps, each census block is assigned a WUI
classification. This thesis experiments with multiple landcover datasets and a local agricultural
dataset to assess their effects on WUI classification and determine the most appropriate datasets
for mapping WUI on Maui. For the final WUI analysis, this thesis utilizes 2020 US Census
Bureau census block data, European Space Agency 2021 WorldCover landcover data, and
Hawaii Statewide GIS Program agricultural land use data to map WUI. The final WUI map
product shows the arrangement of Maui in relation to the WUI before the wildfires. The findings
of the WUI analysis show that 27.46% of land, 96.82% of buildings, and 99.03% of the
population on Maui are in WUI. The final WUI map is developed into an ArcGIS Dashboard that
allows users to explore the WUI in relation to other wildfire-related data, while providing
transparency into WUI calculation. The findings of this thesis are useful for wildfire
management, urban planning, the private sector, and the general public by providing insight into
the spatial arrangement of the WUI in Maui, where to target priority areas for wildfire prevention
interventions, and the relative safety of buildings and homes in the WUI.

1
Chapter 1 Introduction
On August 8, 2023, four wildfires began on Hawaii’s Maui Island, which together caused terrible
loss of life and widespread destruction. The Lahaina wildfire was one of the deadliest wildfires
in United States (US) history. The Olinda, Kula, and Pūlehu wildfires did not cause fatalities, but
caused building damage and thousands of acres of burned land. One year after these wildfires, at
the time of writing this thesis, the Maui landscape and its communities are still recovering.
Leading up to the wildfires, many parts of Maui faced wildfire risk factors, including the
proximity of homes to wildland areas. Maui is known to have much of its population living in the
wildland-urban interface (WUI), a concept created in the 1980s by the department of Forest Fire
and Atmospheric Sciences Research to improve mass fire predictions and to improve wildfire
protection, wildfire management, natural resource management, and land use planning (Sommers
2008). This thesis creates the first census-block based, Maui-specific WUI map and shares the
map publicly in an ArcGIS Dashboard that interactively presents WUI contributing factors and
other wildfire management data. This chapter provides background into wildfires in Maui, the
Maui 2023 wildfires, and the research goals and objectives of this thesis.
1.1 Study Area
The second largest of the Hawaiian Islands, Maui spans approximately 1,886 square
kilometers (or 728 square miles) (Encyclopedia Britannica 2024b). Maui is divided into six
geographic regions (Figure 1). Maui is called the “Valley Isle” because of the large valley
between the two volcanoes that created it, which lie to the east and west sides of the island
(Encyclopedia Britannica 2024b). The volcano on the western side is no longer active and the

2
remnants of the eroded volcano are the West Maui Mountains. On the east, Haleakala is an active
volcano over 3,000 meters tall.
Figure 1. Maui, Hawaii study area
Maui is part of Maui County, along with the islands of Lanai, Molokai, and Kahoolawe.
As of 2022, Maui County has a population of about 164,000 and a population density of 366
people per square kilometer (US Census Bureau 2022). The communities on Maui developed
densely. Maui is significantly denser than the overall US, with a population density of 34 people
per square kilometer in 2022 (Macrotrends 2024) The largest demographic group on Maui is
White, making up about 29% of Maui County’s population, followed closely by the Asian

3
population at about 28% (Data USA 2022). About 10% of the population is Native Hawaiian or
Pacific Islander.
Maui, named after a demigod from Polynesian mythology, has a rich history and cultural
significance regarding the Hawaiian Kingdom. King Kamehameha, regarded as the greatest King
in Hawaii, united the Hawaiian Islands for the first time in 1810 and he chose the capital of the
Hawaiian Kingdom to be in Lahaina (Lahaina Town 2024b). King Kamehameha chose Lahaina
because it is said to be the home of the goddess Kihawahine in Native Hawaiian religion (LaPier
2023). Many Hawaiian monarchs are buried in the Waiola Church Cemetery, which is also
regarded as sacred (Encyclopedia Britannica 2024a). The capital remained on Maui until the
switch was made to Honolulu in 1845.
Before westerners colonized Hawaii, native Hawaiians had a sophisticated land
management system that was not based on private land ownership (County of Maui 2008).
Islands were divided into sections called ahupua’a, which ran from the mountains to the ocean
(County of Maui 2008). The native Hawaiians stewarded the land and carefully managed
resources, understanding the interconnectivity of the various ecosystems across an ahupua’a
(County of Maui 2010). Complex agriculture and aquaculture systems supported the population.
In the 1850s, Maui had 141 ahupua’a, over 300 villages, and about 35,000 inhabitants (County of
Maui).
In the 1800s, Maui underwent many changes as western colonizers introduced a new
intensive agricultural economy based on private land ownership and greatly changed the
vegetative landscape. The native Hawaiians’ natural resource management was destroyed as
newly introduced crops and livestock depleted native vegetation and westerners cut down forests
(County of Maui 2010). Maui continued to change in the 1900s, with a rise in infrastructure and

4
housing after World War II. Hawaii became a state in 1959. Planned communities grew in the
latter half of the 20th century, along with large resorts (County of Maui 2008). As Maui
developed over time, the diverse native landscape degraded.
Maui’s agricultural landcover decreased considerably from 1976-2000s while
development increased (Brewington 2020). Agricultural landcover loss was between 46-63%.
Contrarily, developed land increased significantly, between 105-273%. Furthermore, grassland
on Maui increased between 113-196%, and replaced almost half of agricultural land (Brewington
2020). When sugarcane plantations closed by the late 1990s, the abandoned agricultural plots
became unruly and filled with invasive grasses (Harrison 2020; Parsons and Martin 2023). Urban
development in Maui has altered the natural landscape, allowing highly flammable invasive plant
species to intrude into native vegetation. Homes in the WUI in Maui have an increased wildfire
risk due to the prevalence of highly flammable invasive species in the surrounding wildlands.
While agriculture is still prominent today, the largest economy on Maui is the tourism
industry (Encyclopedia Britannica 2024b). For attractions, Maui boasts beaches, rainforests, a
volcano, waterfalls, and more. Lahaina is the biggest tourist destination on Maui, receiving 80%
of the island’s tourism and about two million visitors annually (Lahaina Town 2024a). Lahaina
has suffered greatly economically after the 2023 wildfires, losing over $13 million in tourismrelated revenue per day (Bond-Smith et al. 2023). Tourism to Maui is encouraged to support the
island’s economy in the aftermath of the 2023 wildfires (Hawai’i Tourism Authority 2024a;
Hawai’i Tourism Authority 2024b).
The climate and wildfire susceptibility vary greatly across Maui, and a map of annual
rainfall in Maui is shown in Figure 2 (Hawaii Statewide GIS Program 2022b). The northern and
eastern parts of the island consist of the windward ‘wet’ side of the island with rainforests and

5
heavy rainfall (Blum 2024). The southern and western portions of the island are the leeward
‘dry’ side that experience arid and desert-like conditions. West Maui is dry and receives little
rain because the prevailing winds from the northeast are blocked by the mountains, producing
heavy rainfalls on the eastern side and creating a rain shadow on the western side (University of
Hawaii at Manoa n.d.). South Maui and parts of West and Upcountry Maui receive under 20
inches of rain per year. This sharply contrasts other parts of Maui that are rainforests, such as the
Big Bog in East Maui, one of the wettest areas on the earth, which receives up to 400 inches of
rain per year (Dejournett 2023). The Lahaina and Pūlehu wildfires occurred in dry parts of the
island, in areas that receive about 15 inches of rain per year. The area around the Kula wildfire
receives more rain, between 25 and 30 inches annually. Of the August 2023 wildfires, the Olinda
wildfire occurred in the wettest area, receiving about 65 inches of rain annually.

6
Figure 2. Rainfall in Maui
1.2 Background
As climate change intensifies, wildfires have become stronger and more destructive
(Jones et al. 2022). The top eight largest wildfires in the history of the US occurred in the last 20
years (Earth.org 2024). The state of Hawaii has been ravaged by many wildfires in recent years,
and the Hawaiian Islands have experienced some of the highest increases in wildfires in the
country (Trauernicht 2019). In the past century, the burn areas of Hawaii have increased
drastically, over 400% (Parsons and Martin 2023). Wildfires affect thousands of people each
year, and wildfire awareness is a crucial part of the lives of Maui residents.

7
1.2.1 Wildfire Conditions in Maui, Hawaii
Factors that contribute to wildfire risk in Maui include the rise of invasive species
populations and dry weather. Due to the isolation of Hawaii from the rest of the world, native
plants on Hawaii lack the ability to defend themselves from invasive species, as they evolved
without competition (Hawaii Invasive Species Council 2024b). This background makes Hawaii
especially susceptible to the detriments of non-native plant invasions. Native plants in Hawaii
did not evolve to adapt to fire because lightning does not normally occur in Hawaii (Parsons and
Martin 2023). Only a few native species, if any, can regenerate after a wildfire; only volcanic
areas experience wildfires naturally and have vegetation adapted to wildfires (Hawaii Dept. of
Land and Natural Resources 2010).
Wildfire risk is exacerbated because of the large number of invasive species that have
been recorded invading the Hawaiian Islands for decades (Tunison, D’Antonio, and Loh 2001;
Parsons and Martin 2023). About a quarter of Hawaii is overrun by invasive, drought-resistant,
highly flammable grasses and shrubs (Parsons and Martin 2023). According to a 2018 USGS
report, Hawaii’s density of invasive species is 200% when compared to the mainland (Parsons
and Martin 2023). When native plant species are destroyed in fires, nonnative dry grasslands
populate the islands; the nonnative grasslands in turn are more susceptible to wildfires, therefore
furthering the continual destruction of the native biodiversity (Trauernicht 2019; Tunison,
D’Antonio, and Loh 2001). Guinea grass and buffelgrass are native to Africa and the Middle
East and are some of the most common grasses seen across Maui. Buffelgrass was prominent
before the 2023 wildfires occurred; this invasive grass has one of the highest wildfire risks of any
in Hawaii (Parsons and Martin 2023). The invasive grasses around Lahaina are known to quickly
spread in areas that are disturbed after fires (Romero and Kovaleski 2023). Large open fields of
grasslands are dangerous because of their fuel content, especially during drier seasons.

8
Since the late 1990s, Hawaii has experienced two of its most severe droughts in the past
century, which has decreased the moisture in the soil and in plants (Parsons and Martin 2023).
The wildfire risk from the invasive grasses that dominate Maui’s landscape increases as they
become drier and more flammable. Drought does not deter the growth of the invasive grasses;
during droughts, native plants suffer but invasive species are less affected and often replace the
native plants (Kunz 2021). Drought conditions are detrimental even after Maui has wet winter
conditions; the invasive grasses grow swiftly with rainfall, so there are larger masses of fuel
vegetation that dry out when drought returns (County of Maui 2021). Due to Maui’s geography,
the southern and western sides of the island are predisposed to have drier conditions, and
therefore heightened wildfire risk. When the August 8, 2023, wildfires occurred, Western Maui
was split between the drought classifications of Abnormally Dry and Moderate Drought, as seen
in Figure 3 (US Drought Monitor 2023). South Maui and parts of Upcountry Maui were in
Severe Drought, and East Maui was Abnormally Dry. Maui was under worse drought conditions
than all the other Hawaiian Islands, none of which had Severe Drought conditions.

9
Figure 3. Hawaii drought conditions on August 8, 2023
Wildfires occur across most of Maui. A map with point locations of incidence locations
for historical wildfires on Maui between 2005 and 2020 is shown in Figure 4, with less
transparency for more recent wildfires (HWMO and Trauernicht 2023). Relatively few fires
occur on the wetter parts of the island, including East Maui and the West Maui Mountains. No
wildfires have begun on the eastern side of Haleakala, and wildfires in East Maui have only
ignited along the coast. Many wildfire occurrences have been in Central Maui, South Maui, and
the coast of West Maui. The more recent fires have also been in these densely populated areas,
especially around Lahaina and the western side of the North Shore. Wildfire incidents are less
dense in Upcountry Maui than in Central Maui and the West and South Maui coasts, but are still

10
numerous, and some recent fires stand prominent in the open land between Upcountry Maui and
Central Maui.
Figure 4. Historical wildfires on Maui from 2005-2020
1.2.2 The August 2023 Maui Wildfires
Beginning on August 8, 2023, Maui experienced a disastrous series of wildfires. All the
Hawaiian Islands were experiencing an extremely high-pressure system, leading to high winds
(Partyka and Erdman 2023). In the early morning hours of August 8, Lahaina experienced winds
reaching 60-80mph, and police closed multiple roads due to many electric poles fracturing and
falling (Maui Police Dept. 2024). At 6:35am, a fire started near Lahaina Intermediate School, but

11
it was largely contained within two hours and fully extinguished by 2:17pm (Maui Police Dept.
2024). Soon after, at 2:55pm, another fire started at the same location and spread quickly,
ultimately destroying most of the town (Kerber and Alkonis 2023). The Lahaina wildfires swept
through old, abandoned plantation fields, the slopes surrounding the town, and up to homes, all
connected by invasive grasses (Romero and Kovaleski 2023). The fires did not conclude until
8:30am the morning of August 9 (Fire Safety Research Institute 2024). In total, 2,170 acres
burned in the wildfire (Maui Police Dept. 2024). Over 100 people were killed, over 2,000
buildings were destroyed, and damages were over $5 billion (US Fire Administration 2024). The
Lahaina wildfires were designated a WUI (specifically, Interface) event by the National Institute
of Standards and Technology (NIST) (Link 2024).
Many historical sites in Lahaina with cultural significance were destroyed (Hawai’i
Tourism Authority 2024a). Lahaina’s historic district is compared before the fire, on March 24,
2023, to soon after the fire on August 11, 2023, in Figure 1Figure 5 (Google Earth 2023a;
Google Earth 2023b). These maps show some of the most popular landmarks. In Lahaina, some
historic landmarks trace back to the 1800s when the first missionaries and whalers arrived, and
others trace back to the Hawaiian Kingdom. The wildfires swept through the historic Front
Street, around which many historic sites are located. The famous Banyan tree, planted in 1873,
was damaged but not destroyed. However, the Lahaina Public Library, Masters Reading Room,
Baldwin Home, Pioneer Inn, and Old Lahaina Courthouse were destroyed. Many artifacts and
archives were lost from the destruction of the Lahaina Public Library and the Old Lahaina
Courthouse. Slightly north, the Wo Hing Temple Museum was destroyed. Slightly to the south,
the Waiola Church was destroyed, and its cemetery was damaged. As of writing this paper, most
of these landmarks are still closed.

12
Figure 5. Historical sites in Lahaina before and after the wildfires
Three wildfires occurred in other regions of Maui, with the Olinda and Kula wildfires in
Upcountry Maui, and the Pūlehu wildfire in South Maui (Maui Police Dept. 2024). These
wildfires were not reviewed by the NIST or assigned WUI event classifications. The burn extents
of the wildfires are shown in Figure 6, covering thousands of acres (County of Maui GIS 2024).
The Olinda wildfire started on a residential street shortly after midnight, at 12:22am on August 8,
spreading quickly and leading police to evacuate residents in the middle of the night (Maui
Police Dept. 2024). The Olinda wildfire burned 1,081 acres. The Kula wildfire started at
11:27am on a small street off Route 377 (Kerber and Alkonis 2024). The Kula wildfire burned
202 acres and damaged 25 homes (Maui Police Dept. 2024; Yamane 2024). At 5:59pm, the
Pūlehu fire started on a small residential road and burned mostly ranch land (Maui Police Dept.
2024; Kerber and Alkonis 2024; Hawaii Dept. of Land and Natural Resources 2023). This fire
burned the largest area, covering 3,240 acres. These three wildfires affected mostly ranch land
and had no casualties. The wildfires were not completely contained until September 30, well
after a month after they began (Federal Emergency Management Agency 2023).

13
Figure 6. Maui August 2023 wildfire burn extents
The Maui wildfires made news headlines around the country and brought widespread
attention to Lahaina, as well as globally reviving conversations on the danger of wildfires and
importance of emergency management (Synolakis and Karagiannis 2024). Resources poured in
from all around the world. As of the writing of this paper, the parts of Lahaina that were
impacted most from the wildfires are still closed to the public, and tourists are asked not to visit
those areas and not to take photos.
1.3 The Wildland-Urban Interface
WUI is defined as the area where developed areas meet or mix with wildland areas
(Radeloff et al. 2005). While WUI has a broad, qualitative definition, thresholds are standardized

14
to quantitatively classify WUI. WUI definitions mostly stem from the definitions set by the
Federal Register in 2001 (Federal Register 2001). There are two subtypes of WUI, wildlandurban interface (from here on, WUIF) and wildland-urban intermix (from here on, WUIX), and
when referred to collectively, they are called WUI. WUI has two main components, structures
and vegetation. Commonly implemented thresholds regarding structural density and vegetation
cover are in Table 1 (NIST 2023; Federal Register 2001). An additional threshold regarding
distance from wildlands is included, which is derived from vegetation cover data. Both WUIF
and WUIX have a structural density threshold of over 6.18 structures per square mile (NIST
2023). WUIX has a vegetation cover threshold of over 50%, while WUIF is less than 50%
vegetation cover, so WUIF areas tend to be more developed than WUIX areas. WUIX does not
have a defined relation with wildlands, but WUIF areas are under 2.4 km from wildlands (NIST
2023). In WUI literature, wildlands are areas of 5 square kilometers or more with greater than
75% vegetation cover (Bar-Massada 2021).
Table 1. WUI definitions
Wildland-Urban Interface Wildland-Urban Intermix
Structural density threshold ≥ 6.18 structures per sq km ≥ 6.18 structures per sq km
Vegetation cover threshold < 50% vegetation ≥ 50% vegetation
Relation with wildland
threshold
< 2.4 km from land (≥ 5 sq km)
with ≥ 75% vegetative cover
N/A
Source: NIST 2023

15
WUI was first proposed as a US Forest Service research initiative in 1987 due to wildfire
management challenges where urban areas and wildlands met (Sommers 2008). Concerns at the
time included water resource conflict and post-war concerns about wildfires caused by bombings
(Sommers 2008). Additional intent in conceptualizing WUI was to quantify wildfire-related
issues along with demographic information, and to combine information on both structural and
wildland fuels (Sommers 2008).
WUI became crucial for wildfire management and policy and is widely used in wildfire
management efforts today. WUI maps can be used to identify areas of high wildfire risk and
target educational efforts. The maps assist wildfire management in prioritizing where to evaluate
and thin fuels in and around urban areas, and where to conduct controlled burns. Targeting where
to implement defensible space policies is informed by WUI maps. WUI maps assist urban
planners in choosing where to enforce fire-resistant building codes, as well as evaluate proposed
developments. WUI maps inform the allocation of firefighting resources and support the design
of evacuation routes.
1.4 Research Goals and Objectives
The goal of this thesis is to use cartography and spatial analysis to support wildfire
management and wildfire-related research in Maui, creating a WUI map made specifically for
Maui and sharing the map publicly in an ArcGIS Dashboard. The overarching research questions
for this thesis are, where does WUI exist on Maui and what changes should be made to standard
WUI mapping practices to best answer this question given the climate and landscape of Maui?
To answer these questions, this thesis first finds the most appropriate datasets to represent the
vegetative and agricultural landscape of Maui by assessing how different landcover datasets
affect WUI classification. WUI maps of Maui with standard WUI classifications and enhanced

16
microclimate-related subclassifications are created. The information is presented in an
interactive, dynamic ArcGIS Dashboard that showcases the factors affecting WUI classification
alongside other wildfire-related data for additional context. The analyses conducted and datasets
created provide insight for planners in Maui concerning the community’s relationship with the
WUI to inform wildfire management planning decisions, as well as provide insight for other
stakeholders invested in wildfire issues such as Maui County, Maui fire departments, insurance
agencies, homebuyers, homeowners, real estate developers, and the public.
1.5 Thesis Organization
The remainder of this thesis consists of five chapters. Chapter 2 provides an overview of
related works regarding studies on wildfires in Maui, WUI mapping practices, and existing WUI
maps that include Maui. Chapter 3 covers the methodology of this thesis by detailing the data
used, the steps for mapping the WUI in Maui, and the experimentation with different landcover
datasets that led to the final methodology. Chapter 4 contains the final WUI map products for
Maui and the WUI maps from the landcover dataset experimentations. Chapter 5 focuses on the
creation of the ArcGIS Dashboard, which showcases the WUI mapping results alongside other
wildfire-related data. Chapter 6 is for discussion of the final products and findings of this thesis,
as well as limitations and additional research going forward.

17
Chapter 2 Related Literature
WUI mapping efforts have increased in the last decade and scholars have created WUI maps for
many areas that have frequent wildfires. WUI studies in the US are mostly concentrated in the 48
conterminous states, thereby excluding Hawaii. This chapter covers existing wildfire studies on
Maui, the many WUI mapping methodologies present in the literature, and existing WUI studies
that include Maui.
2.1 Wildfire-Related Studies on Maui
Before the August 2023 wildfires, wildfire-related studies covered a variety of topics
including drought and wildfire mitigation, but few were specific to wildfire events. Trauernicht
and Lucas (2016) record historical wildfire ignition points for all of Hawaii, but this is the
exception in the literature. Some research is on Maui specifically, but research on Maui is more
frequently part of studies on the state of Hawaii. After August 2023, research poured into Maui.
The 2023 wildfires and their causes and effects were studied extensively.
2.1.1 Maui Drought and Wildfire Studies
Drought was frequently studied on Maui before August 2023. Frazier et al. (2019) find
that the impact of drought on agricultural land is dire, specifically in non-irrigated pastures. Over
half of Maui’s pastureland receives little rainfall, and the agricultural industry suffers heavy
financial losses through cattle (Frazier et al. 2019). Heavy financial losses are also felt in
Upcountry Maui during drought because many profitable vegetables are grown in that region
(Frazier et al. 2019). Dolling, Chu, and Fujioka (2005) find correlation between a drought index
and wildfires on Maui. Studying the various effects of drought on Hawaii, Kunz (2021) gathers
knowledge from wildfire experts across the Hawaiian Islands on the different tactics for wildfire

18
mitigation. Before drought, emphasis is put on fuels management, road management, training,
and taking steps to preserve biodiversity after fires; during drought, emphasis is much higher on
active wildfire prevention and preparation (Kunz 2021). Hawaii is in fire season throughout the
entire year, and this is exacerbated in the drier southern and western regions (Parsons and Martin
2023).
Since wildfires are expected, many mitigation efforts are taken and researched on Maui
and Hawaii as a whole. Maui County has an extensive hazard mitigation plan that covers wildfire
risk, hazard mitigation, implementation plans, and more (Maui Emergency Management Agency
2020). Ritchey (2022) explores the potential benefits of controlled burns in wildfire-prone
regions of the Hawaiian Islands. The study determines that for Maui, the most appropriate
locations for controlled burns would be rural, flat, and dry (Ritchey 2022). Harrison (2020)
explores the potential of mitigating wildfire risk through the untapped strategy of cattle grazing
targeted at invasive grasses. Although the case study is on the island of Hawaii, the applications
extend to all the Hawaiian Islands, especially Maui due to its extensive pastureland and ranching.
De Roode and Martinac (2020) research and select three potential sites for resilience hubs on
Maui that would support communities during emergencies. Corlew (2015) creates a handbook
for community preparedness in Maui, for wildfires, drought, and disasters and hazards.
2.1.2 August 2023 Wildfire Studies
Wildfire susceptibility analysis and mapping has been conducted for the island of Maui
after the 2023 wildfires. Using remote sensing data, Ramayanti et al. (2024) include a multitude
of wildfire factors and use machine-learning techniques to develop a categorical map of wildfire
susceptibility levels across the island. The study also creates an inventory map of Maui wildfires
from 2017-2023, including the August 2023 wildfires.

19
Several studies model the path of the August 2023 wildfires and create maps of the
wildfires’ spread. Roy et al. (2024) use MODIS, VIIRS, and PlanetScope satellite data to track
incidences of the Maui wildfires and characterize their intensities over time. Several map
products are created, including maps of active fires and their extents, symbolized by day and/or
time. This is significant for showing exactly where the fire reached and at what times during the
days that the fire was active. With the Lahaina 2023 wildfire as a case study, Zhou (2024) creates
a fire spread model for WUI wildfires and estimates monetary damage to wildlands and
structures based on their material type. Juliano et al. (2024) use the Weather Research and
Forecasting model and the Streamlined Wildland-Urban Interface Fire Tracing model to model
the predicted path of the Lahaina wildfire at timed intervals and compare the predictions to the
actual extent.
Many studies on the Maui wildfires focus on topics other than the fire itself. The August
2023 wildfires occurred under extreme wind conditions, and the meteorological aspect of the
event is studied (Mass and Ovens 2024). Several studies assess the response to the fire and
provide critiques for emergency planning and how the local government handled the situation
(Byren and TESA Tech Team 2023; Voda 2023). One study uses vector data to create map
products for network analysis from fire stations and cost analysis of damage from parcel data in
Lahaina (Mengote 2024). Balmes and Holm (2023) research how wildfire smoke affects human
health and mentions that WUI fires, such as the one in Lahaina, are especially dangerous because
of the synthetic materials that burn and add toxic substances into the smoke. Averett (2024)
studies contaminants and pollutants around the burned areas of the Lahaina wildfires and notes
that Lahaina stands out amongst WUI fires because of the high density of homes adjacent to the
burned areas.

20
2.2 WUI and Mapping Practices
Many WUI maps have been created on local, national, and global scales, but choices in
WUI mapping methodology vary. The definition of structural density can be altered, either
building density or housing unit density. There are two main methodology frameworks for
mapping WUI: zonal-based and point-based (Bar-Massada 2021). Many studies focus on editing
the standard density and vegetation cover thresholds from the Federal Register WUI definitions
to be more appropriate for local study areas. The type of agricultural land included in WUI
analysis can also be changed.
2.2.1 Building Density vs Housing Unit Density
Whether a study uses buildings or housing units for its structural density metric is an
important distinction. Building footprints include non-residential structures, while housing units
are based on census data and may count multiple structures within the same building, such as in
an apartment complex. Many studies perform multiple iterations, testing the effect of either
choice. Montana has a large history of wildfires, and Ketchpaw et al. (2022) compare three
methods of calculating structural density: local Montana address data, the Microsoft building
footprint dataset, and census housing units. The study found that the census housing unit method
outputs the least WUI territory, and that using buildings results in more WUI coverage. The
study recommends using the Microsoft building footprint dataset for assessing defensible space
around structures (Ketchpaw et al. 2022). Carlson et al. (2021) compare using the Microsoft
building dataset for structural density to using census housing units for structural density.
Carlson et al. (2021) also find that the choice of building data results in more WUI area than
census housing units. The study finds that using building data provides insight into the effect of

21
non-residential structures on the WUI and recommends using building data in study areas
without reliable census data.
2.2.2 WUI Mapping Frameworks
Bar-Massada (2021) compares point-based and zonal-based WUI mapping methods
across California (Figure 7). The point-based approach is based on pixels. Structural density is
calculated as the number of points within a specified buffer distance around the centroid of each
pixel. The percentage of vegetation cover is calculated as the percentage of pixels with
vegetation cover within a radius around each pixel centroid as well. Each pixel is determined to
be inside or outside of the buffer around wildland areas. Alternatively, the zonal-based approach
uses census block polygons to determine structural density and the percentage of vegetation
cover. For distance from wildlands, the census block polygon is determined to be inside or
outside of a buffer around wildland areas. The census block is used in zonal-based WUI mapping
because it is the finest resolution of census data available, smaller than both census tracts and
census block groups. A benefit of the point-based approach for calculating density is that the
varying size and shape of census block polygons are not an issue, as in the zonal-based approach.
However, there are no standards for how far of a radius should be used for calculating density in
the point-based approach.

22
Figure 7. Diagram of two WUI mapping approaches. Figure by Bar-Massada 2021.
2.2.2.1 Point-based
Some WUI maps cover a continental or global extent, and they use point-based WUI
mapping practices. Mapping across national boundaries and beyond the extent of consistent
census data makes the point-based scale appropriate. Schug et al. (2023) use the point-based
approach to map WUI globally with a 500m search radius. Johnston and Flannigan (2018) create
the first WUI map for Canada and use a pixel-based approach. With a building location dataset
and a 30m resolution LANDSAT landcover dataset, the study calculates density in hexagonal
grid cells of 3,400 square kilometers. Johnston and Flannigan (2018) show that almost 4% of
Canada is WUI and the WUI maps pave the way for future wildfire-related research. Carlson et

23
al. (2021) create a WUI map of the conterminous US; this study uses 30m 2016 NLCD data and
Microsoft building footprints. Carlson et al. (2021) calculate building density and vegetation
cover per pixel using a moving window method, with thresholds determined by the Federal
Register. The project tests radii of 100, 250, 500, 750, 1000, and 1500m. Density is more
controlled in the pixel-based method than when using census blocks of various sizes in the zonalbased method, but results change significantly based on the radius used; the area counted as WUI
increases when the radius increases. After testing the various radii, Carlson et al. (2021) choose
and recommend the 500m radius as the most appropriate for WUI mapping across the
conterminous US because this radius captures WUI around development next to wildlands and
excludes WUI around isolated buildings.
2.2.2.2 Zonal-based
Using the zonal-based approach, Radeloff et al. (2005) mapped WUI for the 48
conterminous states in one of the earliest WUI mapping efforts. This study is one of the most
significant and commonly cited papers in WUI mapping literature. The zonal approach is popular
due to the availability of census block data across the country. The study uses decennial census
housing unit counts and NLCD landcover to calculate housing unit density and vegetation cover
per census block. While using standard WUI threshold definitions, Randeloff et al. (2005)
perform a sensitivity analysis to determine how much the threshold adjustments affect results.
The study tests housing density thresholds of 3.09, 6.17, and 12.34 housing units per square
kilometer. Vegetation cover thresholds are tested at 25, 50, and 75%. The wildland buffer is
tested at 1.2, 2.4, and 4.8 kilometers. Furthermore, wildland vegetation cover classifications are
part of the sensitivity analysis. Standard wildland classifications, classifications excluding woody
and emergent wetlands, and forest classifications alone are tested. The results of the sensitivity

24
analysis show that the housing density threshold affects WUI area the most, WUIF is especially
sensitive to the buffer distance, WUIX is especially sensitive to the vegetation cover threshold,
and shrublands should not be excluded from vegetation cover because fires are frequent in
chaparral ecosystems. Overall, the ranking of WUI area remained consistent across states
regardless of the thresholds used, showing the reliability of WUI analysis (Radeloff et al. 2005).
Radeloff et al. (2018) use the same zonal-based methodology to compare WUI in 1990 and 2010.
2.2.3 WUI Definition Thresholds
While many studies use the Federal Register definitions for density and vegetation cover
thresholds, some studies alter these thresholds to be more suitable for a study area. Li et al.
(2022) map the WUI in California, emphasizing the need for WUI maps to be continually
updated in such a fire-prone state, and the study tests the definition thresholds of vegetation
cover percentages and the distance between structures and wildland. The study tests vegetation
cover percentages in 10% increments, and tests 1.2-, 2.4-, and 4.8-kilometer buffers. Ultimately,
a standard buffer of 2.4 km is chosen because the other tested buffers do not significantly change
results. A vegetation cover threshold of 40%, slightly lower than the standard 50%, is chosen
because the study finds that change in WUI area is stable under this threshold.
Slaton (2022) explores the most appropriate WUI definition for the study area of the
Oregon-California border, emphasizing the importance of custom thresholds for small-scale
studies. The study tests vegetation cover thresholds between 0 and 75%, and tests housing
density thresholds of 1 home per 20, 40 (standard), 60, and 400 acres. A 25% vegetation
threshold is chosen for the study area to represent WUI near urban cores. Slaton (2022) chooses
a 1 home per 400 acres (approximately 0.6 homes per square kilometer) density threshold,

25
finding that a lower density threshold than national definitions is critical to properly identify
WUI in the study area because there are many sparsely populated areas that must be represented.
2.2.4 Agricultural Mapping in WUI
In WUI mapping practice, agricultural land is not considered to be vegetation cover
(Radeloff et al. 2005; Schug et al. 2023; Slaton 2023). The most important characteristic of
agricultural land that removes it from vegetation cover classification in WUI mapping is its
lower level of flammability; agricultural land is typically maintained and irrigated. Agricultural
plant species vary in their level of flammability, and some pasture grasses have low wildfire risk
(Pagadala et al. 2024). Pastureland is considered agricultural land and removed from vegetation
cover calculations in Radeloff et al. (2005). Fu et al. (2023) find that irrigated, low-flammability
crop species serve well as fire buffers. Specifically, the study highlights bananas. Fu et al. (2023)
recommend banana buffers as a profitable method of fire mitigation, especially in WUI.
2.3 WUI on Maui
Mapping WUI is a popular subject in literature, however not for the study area of Hawaii.
Broadly, WUI estimates about Maui have been made, but without detailed WUI mapping. No
WUI map has been made specifically for the study area of Maui, but insight into Maui’s WUI
can be gleaned from WUI maps of larger study areas, of the world and of Hawaii, that include
Maui. Furthermore, WUI studies of other fire prone areas provide insight into what may be
expected for WUI outcomes. Centered around WUI, many guidelines and community plans have
been created for Maui.

26
2.3.1 Maui WUI Estimations
Nominal WUI assignments have been attributed to Maui without the support of WUI
analysis. Per the HWMO and Maui County in a 2014 report, all Western Maui is broadly
considered to be WUI and share a wildfire protection boundary (Pickett, Grossman, and HWMO
2014). However, this map is not detailed and does not give insight into the spatial distribution of
the island in relation to WUI. Based on field reconnaissance after the 2023 wildfire, the NIST
declared the Lahaina wildfire to be an interface event, in a WUI subclassification of intermix or
interface (Link 2024). However, a full case study was not conducted by the NIST for Lahaina, so
there is no map or further description of this classification. The other 2023 Maui wildfires in
Olinda, Kula, and Pūlehu were not reviewed by the NIST or assigned WUI event classifications.
2.3.2 Maui in Global WUI
The SILVIS Lab at the University of Wisconsin-Madison released a global WUI map
based on the year 2020 at 10m resolution (Schug et al. 2023). The global WUI map provides a
detailed WUI map of Maui, shown in Figure 8. The map does not follow the standard building
density threshold of 6.18 buildings per square kilometer because it uses a global building density
dataset, rather than building footprints. The study uses the Global Human Settlement worldwide
built-up surface estimate 10m raster dataset to calculate building density for pixels with a 500m
radius. Unable to follow the Federal Register building density threshold, the study considers
pixels as potentially qualifying as WUI if they have at least 0.5% aggregated building density of
pixels in the 500m radius around them. The paper claims that because this threshold is higher
than the Federal Register definition, the WUI estimates are conservative, so there are likely more
WUI areas on Maui than designated in this map.

27
Figure 8. SILVIS Lab – Maui WUI map
For landcover and vegetation data, the study uses the European Space Agency
WorldCover 2020 dataset, reclassifying the data into wildland and non-wildland vegetation.
Following standard WUI definitions, the study finds wildland areas that are greater than 5 square
kilometers of ≥ 75% vegetation cover and considers pixels within 2.4 square kilometers of these
wildlands as potentially WUIF. Whether grassland cover, or forest, shrubland, and wetlands
cover is greater determines if a WUI pixel receives a grassland or forest subclassification.
The target audience for the global WUI map is other researchers, and map interpretation
is done for large areas such as countries and biomes. In continental summaries, Schug et al.
(2023) disclose that under 5% of land area and over 60% of people in Oceania are in WUI, and
that forest-dominated WUI is more prominent than grassland-dominated WUI in Oceania. While
viewable on a web application and accessible through ArcGIS Online, this dataset is exceedingly
large and downloadable at continental levels.

28
2.3.3 Maui in Hawaii WUI
The Hawaii Dept. of Land and Natural Resources (DLNR) has created its own definition
of WUI for Hawaii based on Communities at Risk (CARs), which are communities of varying
wildfire risk identified by the Hawaii Department of Forestry and Wildlife (DOFAW). The
CARs are designated based on vegetation type, climate, and wildfire history, and organized into
High, Medium, or Low Risk classifications (Hawaii DLNR 2010). The state of Hawaii defines
WUI for Hawaii as areas within 1-mile of CARs. The CARs and WUI defined by Hawaii are
shown in Figure 9Figure 6 (Hawaii DLNR 2010). These assessments were begun in 2005 and
were published in 2010. The WUI definition does not conform to the 2001 Federal Register
definitions of WUI, which are based on density and vegetation cover. The polygon outlines do
not align with census-defined community boundaries. The intended audience of the Hawaiispecific CAR and WUI designations is Hawaii-based wildfire management and planning
professionals.

29
Figure 9. Hawaii state-defined CARs and WUI. Figure by Hawaii DLNR 2010.
2.3.4 Comparison to WUI in Fire-Prone Western US
WUI studies are common in the conterminous US. Radeloff et al. (2023a) provide WUI
statistics for the conterminous US for the year 2020, and WUI information on the fire-prone
western states can be gathered. While percentages of land area in WUI are quite low for all the
fire-prone western states, ranging from 0.9% in Nevada to 8.2% in Washington, the percentage
of housing within the WUI is significantly higher, ranging from 32.5% in Washington to 80.1%
in Wyoming. The national average for housing percentage in WUI is 31.6%, and many of the
western states do not stray far from this average. Notably, while only 1% of Wyoming’s land
area is within WUI, 80.1% of housing is within the WUI, the highest of the western states. These

30
statistics highlight the disproportionate amount of development within the WUI in fire-prone
states.
2.3.5 Maui WUI Guidelines and Community Protection Plans
Designing communities with WUI consideration is addressed in literature to protect
communities from fires in nearby wildlands. Calkin et al. (2023) address WUI fires, such as the
Lahaina fire, and place the responsibility on individual communities to prevent major wildfire
spread in densely developed areas. Specifically, Calkin et al. (2023) take the position that
communities must have proper landscaping, construction site, and material guidelines that are
designed to prevent the spread of wildfire. Since the greatest risk of wildfire spreading from
structure to structure is within 100 ft, the immediate surroundings of buildings should not
connect flammable materials together, whether by fences or otherwise (Calkin et al. 2023).
Necessary precautions must be taken by homeowners and local governments to prevent wildfires
from spreading between structures at such a rapid rate (Calkin et al. 2023). The extension of
utility networks in WUI areas increases ignition risks, as in the case of the Lahaina wildfire
(Mahmoud 2024), and utility planning must consider WUI as well. All these steps are necessary
for WUI areas to coexist with wildland fires and prevent the fires from spreading amongst
dwellings.
Maui has many wildfire protection plans such as the West Maui Community Wildfire
Protection Plan and the Maui County Multi-Hazard Mitigation Plan Update. Fire management
requirements include that projects must have defensible space around them and keep up with
maintenance per Maui Fire Dept. guidelines (Maui County Department of Planning 2022). Plans
to develop firebreaks with multipurpose recreational functions around and between communities
are encouraged. The plan mentions that the transportation system needs to be improved to better

31
address wildfire hazards and that a wildfire information campaign should be held to encourage
native plant landscapes and firebreaks.
In a 2021 report, Maui County identified alien grasses as a dangerous wildfire fuel source
on the island that must be removed (County of Maui 2021). A strategy recommended for wildfire
prevention is replacing invasive grasses with native plants, especially in abandoned sugarcane
plantations. The report also recommends an assessment program that identifies properties with
hazardous overgrown vegetation and supports their development of proper firebreaks.
Furthermore, the report suggests that increasing the required width of firebreaks countywide can
reduce the spread of wildfires (County of Maui 2021).
The Maui Fire Department has released a Community Risk Reduction Program for WUI
on Maui County’s website. The plan has general guidelines for reducing hazards but is not
spatialized (Purdy 2018). The plan recommends self-assessment for fire code violations, fire
hazard inspections, removing vegetation deemed to be a hazard by the fire department, keeping
highly flammable vegetation fuels more than 30 ft away from buildings (or 100 ft, depending on
circumstance), and incorporating fuel breaks.

32
Chapter 3 Methods
This chapter details the data as well as the zonal-based methodology used in this thesis for
mapping Maui’s WUI. A WUI map is created with standard WUI classifications, and then this
map is refined to create a WUI map with Maui-specific WUI subclassifications. To arrive at this
methodology, several landcover datasets were compared before choosing the most appropriate
dataset to use for vegetation consideration. Furthermore, agricultural data was also incorporated
before finalizing the WUI methodology, and these enhancements are discussed.
3.1 Workflow Overview
A workflow diagram of all steps in this thesis’s final analysis is shown in Figure 10. The
analysis in this thesis is entirely replicable. The only data needed to complete the WUI map are a
building footprint dataset, census block data, landcover data, and agricultural data. This thesis
project falls into six steps: Step 1 is calculating building density, Step 2 is calculating vegetation
cover with agricultural enhancements, Step 3 is identifying wildlands, Step 4 is creating a
standard WUI map, Step 5 is calculating grassland and forest cover, and Step 6 is incorporating
the forest cover to enhance the WUI map by distinguishing WUI and wildlands that are forestdominated.

33
Figure 10. Workflow diagram
3.2 Data Description
The multiple facets of WUI analysis require data products from the Census Bureau, the
Hawaiian government, Maui County, and more. The data and relevant characteristics are
portrayed in Table 2. Data for WUI mapping includes building footprints, census block
boundaries, landcover, and agricultural landcover. More landcover datasets are included in the

34
data table than are used for creating the final WUI map; these extra datasets are used in WUI
mapping experiments to compare how vegetation and agricultural classifications affect WUI
classification outcomes. The data sources are all free and publicly accessible. All data is
projected to Hawaii State Plane 2 (m), per Hawaii government standards (Hawaii DLNR 2013).
Raster data is cropped to a bounding box surrounding Maui and vector data is cropped to Maui
before analysis begins. All analysis is completed in ArcGIS Pro.
Table 2. Data descriptions
Data Source Year Resolution Use Data Description
Building
Footprints
Maui County
GIS Dept.
2020 N/A,
vector
Calculate building
density
Building footprint
polygons for Maui
Census Blocks Esri Federal
Data, sourced
from US
Census Bureau
2020 N/A,
vector
Boundaries for WUI
mapping and zonal
analysis
Census block polygons
Hawaii
Existing
Vegetation
Type
LANDFIRE 2022 30m Vegetation cover
dataset
experimentation
Per pixel, the type of
vegetation cover
Hawaii Fuels
Cover
LANDFIRE 2022 30m Vegetation cover
dataset
experimentation
Per pixel, the % coverage
of flammable fuels
WorldCover
Landcover
European
Space Agency
Sentinel-2
2021 10m Designate
developed/nonwildland areas vs
vegetation/ wildland
Raster of landcover
classification
Agricultural
Land Use
Hawaii
Statewide GIS
Program
2020 N/A,
vector
Eliminate appropriate
vegetation types from
vegetation cover
Polygons of existing
agricultural lands
categorized by crop/use
3.2.1 Development-Oriented Datasets
Census blocks are nationally available with information from the 2020 decennial census
(Esri US Federal Data 2024) and building footprints for the year 2020 are provided by Maui
County (County of Maui GIS 2020). There are over 1,400 census blocks and over 70,000

35
buildings on Maui, and they are shown in Figure 11. Census blocks vary in size across the island,
small in densely developed areas and large in unpopulated wildlands. Buildings are distinctly
concentrated in West Maui, Central Maui, South Maui, and parts of Upcountry Maui. Clusters of
development are present but sparser in the North Shore and East Maui regions.
Figure 11. Census blocks and distribution of buildings
3.2.2 Landcover Datasets
The included landcover datasets for mapping vegetation cover are LANDFIRE Existing
Vegetation Type (EVT), LANDFIRE Fuel Vegetation Cover (FVC), and European Space

36
Agency (ESA) WorldCover. The National Land Cover Dataset (NLCD) is the standard dataset
for WUI vegetation cover in the US because of its wide coverage, relatively fine resolution
(30m), and its timely updates (the most recent being 2021). However, the NLCD updates do not
regularly include Hawaii, and the last Hawaii Land Cover dataset is from 2011, so NLCD is not
considered in this analysis.
LANDFIRE is a national government program from a collaboration between the US
Department of Agriculture Forest Service and the US Department of the Interior, and the datasets
are specifically intended to support wildfire management. LANDFIRE datasets are included
because of their wildfire-oriented nature, and they have local datasets that are Hawaii-specific.
Hawaii has unique vegetation, so Hawaii-specific datasets are advantageous. They are also 30m
resolution, like NLCD, and are based on the year 2022, one year more recent. The data is created
from field-referenced data, biophysical layers, and machine learning. LANDFIRE datasets have
been used in WUI literature (Li et al. 2021).
The EVT dataset is considered for the WUI mapping vegetation cover dataset because it
has detailed, vegetation-specific vegetation classifications. The EVT dataset is shown in Figure
12. This dataset has many distinctions, with over 50 existing vegetation type classifications.
Whether the vegetation is Hawaiian or Polynesian is distinguished, and the Polynesian
classifications are mostly dominated by invasive species. The EVT dataset has multiple
agriculture classifications, including cultivated crops, pasture/hay, and orchards. There are also
multiple subclassifications for grassland, forests, shrubland, and development. The dataset has
significant agricultural misclassifications, especially around Central Maui and West Maui.

37
Figure 12. LANDFIRE EVT dataset

38
LANDFIRE also has a wildfire fuels landcover dataset. The FVC dataset is considered
for mapping WUI in Maui because it bases its landcover classification on flammability rather
than vegetation type, and vegetation cover classification based on flammability may be
beneficial in a WUI context. The FVC dataset is shown in Figure 13 and contains over 30
classifications. This dataset does not detail the specific vegetation type, as in the EVT dataset,
but generalizes flammable fuel cover into three categories: tree cover, shrub cover, and herb
cover. Within these three categories, the data distinguishes the vegetation cover in increments of
10%. There is one agricultural classification, and it covers significantly less land than the
agricultural classifications in the EVT dataset. Like the EVT dataset, this dataset has five
development distinctions, but the FVC dataset differs by having only one agricultural
classification, cultivated crops. Most of the land area that is classified as agricultural land in the
EVT dataset is classified as shrub- or herb-covered land in the FVT dataset. However, large
areas of West Maui are still classified as agricultural.

39
Figure 13. LANDFIRE FVC dataset
The ESA Sentinel-2 global landcover dataset has a higher resolution than the other
datasets, with a fine, 10m resolution. This ESA dataset is based on 2021, the same year as the
most recent NLCD release that did not include Hawaii. It is also used by the SILVIS Lab in the
WUI map of the world, although the SILVIS Lab used 2020 WorldCover data. This dataset is

40
shown in Figure 14. With nine classifications, the ESA dataset has significantly fewer vegetation
distinctions than either LANDFIRE dataset. All tree cover, shrubland, grassland, development,
and agriculture are represented by one classification each. Compared to the LANDFIRE datasets,
almost no land on Maui is classified as agricultural. While the LANDFIRE datasets identified
croplands in West Maui and around Central Maui, the ESA dataset classifies these areas as
grassland. Only small patches of land around Central Maui are classified as agricultural by the
ESA dataset. The ESA dataset recognizes significantly less developed land, most notably in
Upcountry Maui. The resolution of the ESA dataset is much finer than the LANDFIRE datasets,
which brings higher accuracy; however, some parts of the communities in Upcountry Maui are
underrepresented in their development coverage, and more landcover in Upcountry Maui is
classified as vegetation cover.

41
Figure 14. ESA WorldCover dataset
The three datasets that can be used to calculate vegetation cover vary greatly and show
that agricultural land is difficult to classify on Maui, whether from remote sensing or predictive
modeling. Notably, the EVT dataset presents a great swath of agricultural land in Central Maui,
whereas this is mostly grassland in the FVC and ESA datasets. Historically, this area was sugar
cane cropland, however it is no longer agricultural. The ESA dataset classifies little area of Maui
as agricultural.
The Hawaii Statewide GIS Program is the source for an authoritative agricultural
landcover dataset. The agricultural land use data for Maui for the year 2020 is shown in Figure
15 (Hawaii Statewide GIS Program 2022a). Maui has many different agricultural land uses: seed

42
production, pasture, commercial forestry, banana, tropical fruits, pineapple, flowers, taro,
diversified crop, macadamia nuts, and coffee.
Figure 15. Agricultural land use dataset
By area, pastureland is the most prevalent agricultural land use, covering about 25% of
the land area of Maui. As seen in Table 3, pastureland consists of about 466 square kilometers
out of Maui’s almost 500 square kilometers of agricultural land. The second most prominent
agricultural type is Diversified Crop, covering about 14 square kilometers. The remaining nine
agricultural types all cover less than four square kilometers.

43
Table 3. Maui’s agricultural land area by agriculture type
Agriculture Type Area (square km)
Banana 0.26
Coffee 3.82
Commercial Forestry 0.15
Diversified Crop 13.77
Flowers / Foliage /
Landscape
0.62
Macadamia Nuts 3.20
Pasture 465.83
Pineapple 3.88
Seed Production 2.62
Taro 0.46
Tropical Fruits 2.44
Source: Hawaii Statewide GIS Program 2022a
3.3 WUI Mapping
This thesis sources its definition of the WUI from the Federal Register and its
methodology from WUI literature. This thesis follows zonal-based WUI mapping techniques, as
outlined in Bar-Massada (2021). Structural density, sourced from building footprints as in
Carlson et al. (2022), and vegetation percentages, sourced from ESA as in Schug et al. (2023),
are calculated within census blocks to map WUI classifications.
According to the Federal Register, WUIX areas are areas with over 50% vegetation and
structural density above 6.18 structures per square kilometer. WUIF areas have under 50%
vegetation, structural density above 6.18 structures per square kilometer, and are located less
than 2.4 km away from wildland areas. Building density, vegetation cover, and wildland areas

44
are calculated to map WUI and classify WUIX, WUIF, Non-WUI, and Non-WUI Wildlands.
After these standard classifications are mapped, vegetation cover metrics are enhanced by
calculating grassland and forest cover. The new forest cover percentages are used to further
classify Non-WUI Wildlands and WUIX into Non-WUI Wildlands Forest and WUIX Forest.
3.3.1 Building Density
Building density is chosen as the structural density metric for WUI calculation in this
thesis because this metric is suitable for Maui. Commercial areas were destroyed in Lahaina and
received extensive media attention. Housing and commercial areas are in close proximity in
Maui and fire spreads between buildings, regardless of how many residential units are within
them. Excluding non-residential buildings in WUI analysis would neglect a major aspect of the
wildfires’ impact in Maui. After investigation, although the density metrics changed significantly
(because housing unit density is significantly higher than building density in Maui), most census
blocks are far above the density threshold whether the metric is housing unit or building density.
So, the different density types will not affect the WUI classification, and building density
remains the chosen metric.
Census block polygon data and building footprints are needed to calculate building
density for Maui as part of WUI classification because WUI areas have a structural density of ≥
6.18 structures per square kilometer. WUI mapping with census blocks is precedented by the
foundational study, Radeloff et al. (2005). This thesis uses the zonal-based WUI mapping
approach, and structural density calculations per census block are the foundation of zonal-based
WUI mapping (Bar-Massada 2021).
Calculating building density follows Step 1 of the workflow diagram, and a detailed
workflow diagram of this step is shown in Figure 16. To calculate building density in ArcGIS

45
Pro, the Summarize Within tool is used to create an attribute field of the number of buildings in
each census block. The Maui census blocks are the Input Polygons, and the building footprints
are the Input Summary Features; the output feature class with the new attribute field for the
number of buildings is used for further analysis. The Calculate Geometry tool is used to gather
the area of each census block in square kilometers in a new attribute field. The Calculate Field
tool is used to divide the number of buildings in each census block by the area in square
kilometers of each census block to gather each census block’s building density per square
kilometer.
Figure 16. Workflow diagram for calculating building density
A map of the building density on Maui is shown in Figure 17. The threshold value of
6.18 buildings per square kilometer is used as one of the symbology cut-offs. The vast areas in
white are census blocks that do not qualify for WUI designation because they do not meet the
building density threshold. Most areas that do meet the building density threshold of 6.18

46
buildings per square kilometer have a much higher density than this minimum. The densest
development on the island is in West Maui, Central Maui, and South Maui.
Figure 17. Building density
The building density distribution on Maui is heavily skewed to the right, as shown in the
histogram in Figure 18. A large proportion of Maui’s landcover has very low building density,
but the fewer highly populated areas are incredibly dense. The minimum building density is 0,
the maximum is over 5,700 buildings per square kilometer, and the average building density is
about 574 buildings per square kilometer. Out of 1,447 total census blocks on Maui, 1,131 have a

47
building density above the threshold of 6.18 buildings per square kilometer. These census blocks
reach the building density threshold to qualify as WUI census blocks.
Figure 18. Building density distribution
3.3.2 Vegetation Cover
The percent vegetation cover is also calculated per census track, consistent with standard
zonal-based WUI mapping (Bar-Massada 2021). This thesis uses ESA WorldCover 2021 as the
source of vegetation cover. Calculating the percentage vegetation cover per census block follows
Step 2 of the workflow diagram, and a detailed workflow diagram of this step is shown in Figure
19. The vegetation cover calculation methodology stems from Slaton (2023). The ESA dataset
comes in nine classes originally: tree cover, shrubland, grassland, cropland, built-up, bare/spare
vegetation, permanent water bodies, herbaceous wetland, and mangroves. These are reclassified
with the Reclassify tool to represent either vegetation cover (1) or non-vegetation cover (0). The
tree cover, shrubland, grassland, herbaceous wetland, cropland, and mangroves are assigned to 1.
Built-up, bare/sparse vegetation, and permanent water bodies are assigned to 0. Cropland is
typically assigned as non-vegetation in WUI literature; however, it is initially assigned as
vegetation in this step because authoritative, accurate agricultural plots will be removed from
vegetation cover in a later step and be assigned as non-vegetation landcover.

48
Figure 19. Workflow diagram for calculating vegetation cover
The results of the initial reclassification are shown in Figure 20. By area, most of the
island is covered by vegetation. The developed communities of West Maui, Central Maui, and
South Maui are distinguishable as areas with high concentrations of non-vegetation cover.
Furthermore, the area around the volcano is notable as non-vegetation, as well as the tip of the
southern coast.

49
Figure 20.Vegetation cover before agricultural enhancements
Agricultural land other than pastures is removed from the vegetation cover before further
calculations are run. Using the Select by Attributes tool, the included agricultural land without
pastures is gathered by selecting the inverse of agricultural plots that are equal to the pastureland
classification. After they are selected, these less-flammable agricultural plots are exported to a
new feature layer, as is shown in Figure 21. Without pastures, the total area of agricultural land
decreases significantly, and the remaining agricultural plots are small and scattered. While they
span across the whole island, the largest concentrations of large agricultural plots are in Central
Maui and Upcountry Maui.

50
Figure 21. Included agricultural plots
The remaining agricultural land (without pastures) is removed from vegetation cover
classification. The Extract by Mask tool is used on the reclassified vegetation raster to extract the
areas outside of these agricultural plots. The Extraction Area must be set to Outside, and the
Feature Mask Data Input is the feature layer of agricultural plots without pastures. The Analysis
Extent must be set to the extent of the island. The output from this tool shows gaps where the
plots were. The Reclassify tool is run again so that the no data values are also assigned to 0 (nonvegetation cover). The output can optionally be cropped to the extent of Maui by using the
Extract by Mask tool again. The final vegetation raster with appropriate agricultural
considerations is shown in Figure 22. Compared to the previous vegetation raster, the amount of

51
non-vegetation cover increased. The shapes of the agricultural plots removed from vegetation
cover calculations are seen across the island but are especially noticeable in Central Maui and
Upcountry Maui. The agricultural adjustment shifts the vegetation cover classifications in census
blocks that have agricultural plots in them. These are the final vegetation cover classifications.
Figure 22. Vegetation cover with agriculture considered
The final vegetation cover classifications are used to calculate the percent vegetation
cover per census block. The Zonal Statistics as Table tool is used to get the mean of the
vegetation values (0s and 1s) within each census block, as in Slaton (2023), representing the
percentage of vegetation cover. The Input Feature Zone Data is the census block layer, the Zone

52
Field is the Geographic Identifier, and the Input Value Raster is the vegetation raster. The
Statistics Type is set to Mean. The values in the output table are multiplied by 100 to get the true
percentage value. Then, this attribute field is joined back to the census block feature layer with
the Join Field tool. The Input Table is the census block feature layer, the Join Table is the table,
the Input Field is the Geographic Identifier, the Join Field is GEOID, and the attribute field
holding the percentage vegetation cover is the Transfer Field. The census block layer then
contains the vegetation cover attribute field.
The percent vegetation cover for census blocks in Maui is symbolized and shown in
Figure 23. The vegetation cover ranges from 0 to 100%, and most of the island is highly
vegetated. The map includes a symbology break at 50% vegetation cover, so the distinction is
clear between census blocks that may qualify for WUIF (< 50%) or WUIX (≥ 50%) per the
vegetation cover aspect of their definitions. Most census blocks on Maui have greater than 50%
vegetation cover, with the exceptions being in the most developed areas. Another symbology
distinction on the map is the 75% vegetation cover threshold, which is used to determine areas
that qualify as wildlands. Most census blocks that have ≥ 50% vegetation cover also have ≥ 75%
vegetation cover. The ≥ 75% range covers most of the island.

53
Figure 23. Vegetation cover per census block
3.3.3 Wildland Areas
Wildlands are mapped after vegetation cover is calculated. Part of WUIF classification
requires the census block to be within 2.4 kilometers of wildlands, which are greater than 5
square kilometers (Bar-Massada 2021). In this thesis, the vegetation cover data is also used to
determine the wildland areas. Per Federal Registrar definition, the wildland areas must be ≥ 75%
vegetation cover. Identifying wildlands census blocks and creating a buffer around them follows
Step 3 in the workflow diagram, and a detailed workflow diagram of this step is shown in Figure
24. The Select by Attributes tool is used to select census blocks that have ≥ 75% vegetation

54
cover and the Export Features tool is used to export these census blocks to a new feature layer.
The Dissolve tool is used to dissolve the newly exported census blocks with ≥ 75% vegetation
cover into continuous polygons. The Create Multipart Features option is unchecked. In the
dissolved output feature layer, the Calculate Geometry tool is used to calculate the area in square
kilometers of each continuous, ≥ 75% vegetation cover polygon. The Select by Attributes tool is
used to select polygons that are < 5 square kilometers. These selected polygons are deleted
because they do not qualify as wildland areas.
Figure 24. Workflow diagram for identifying wildlands and creating a wildland buffer
The remaining census blocks areas that are ≥ 5 square kilometers of area with ≥ 75%
vegetation cover qualify as wildlands, and they are shown in Figure 25. Maui has 1,649 square

55
kilometers of wildlands; wildlands cover most of the island. Few census blocks were removed
for not meeting the ≥ 5 square kilometer threshold. The areas that do not qualify as wildlands are
developed areas, some areas that contained agricultural plots, and the southern coast and
volcano, which did not have high vegetation cover classification. The wildlands layer is used
later for Non-WUI Wildlands classification in WUI mapping.
Figure 25. Wildland areas
The wildland areas layer is used to create the 2.4-kilometer buffer that is a part of the
WUIF definition. The final wildlands layer is exported with the Export Features tool to make a
copy, and the Buffer tool is used on this layer to create a 2.4-kilometer buffer around the

56
wildlands. This buffer is shown in Figure 26. The 2.4-kilometer buffer covers almost the entire
island, except for two barren areas (the southern coast and volcano), and extends into the ocean.
This buffer is used later in WUI mapping to impact WUIF classifications.
Figure 26. Wildland buffer
3.3.4 Standard WUI Map
Standard WUI maps have four classifications: WUIF, WUIX, Non-WUI, and Non-WUI
Wildlands. Initial WUI mapping is Step 4 of the workflow diagram, and a detailed workflow
diagram of this step is shown in Figure 27. A new empty text attribute field is added to the

57
census blocks layer to hold WUI classifications. WUIF and WUIX census blocks are identified
first. The Select by Attributes tool is used to select census blocks that fit the WUIF definition,
with ≥ 6.18 buildings per square kilometer and < 50% vegetation cover. The selected census
blocks are assigned “Interface” in the new attribute field with the Calculate Field tool. The
WUIF census blocks are exported with the Export Features tool to a new layer to be compared
with the 2.4-kilometer wildland buffer. None of the WUIF census blocks fall outside of the
buffer, so the wildland buffer aspect of the WUIF definition does not affect WUIF classification
on Maui. All census blocks labeled “Interface” maintain their classification. If there were WUIF
census blocks outside the buffer, they would be assigned “Non-WUI.” The WUIF selection is
cleared, and the Select by Attribute tool is used to select census blocks that fit the WUIX
definition, with ≥ 6.18 buildings per square kilometer and ≥ 50% vegetation cover. The selected
census blocks are assigned “Intermix” with the Calculate Field tool, and the selection is cleared
again.

58
Figure 27. Workflow diagram creating a standard WUI map
The non-WUI classifications are assigned after the WUI classifications. Identifying
wildland census blocks requires multiple steps and the use of the wildlands layer created
previously. The Select by Attributes tool is used to select census blocks with ≥ 75% vegetation
cover and where the WUI classification is Null. Then, to ensure these selected areas qualify as
continuous wildlands that are ≥ 5 square kilometers, the Select by Location tool is used with the
census blocks layer as the Input Feature, the Relationship as Within, and the Selecting Feature as
the dissolved wildlands layer with unqualified areas removed. The Selection Type is set to Select
Subset from the Current Selection. The updated selection (although nothing changes in this

59
scenario) is assigned to “Non-WUI Wildlands” using the Calculate Field tool. After the selection
is cleared, the Select by Attributes tool is used to select census blocks that have a Null WUI
Classification attribute field, and these selected census blocks are assigned to “Non-WUI” with
the Calculate Field tool. The initial WUI map product consists of these four classifications that
strictly adhere to the Federal Register WUI definitions. Cartographically, WUIF and WUIX are
symbolized in WUI literature with warm tones that are associated with fire. Non-WUI Wildlands
are shown with green tones and Non-WUI is shown with white tones.
3.3.5 Forest and Grassland Cover
Distinctions between the dry, grassland ecosystems and the wet, rainforest ecosystems on
Maui are crucial because rainforests have lowered wildfire risk due to high moisture content and
higher rainfall. Rainforest census blocks have less fire risk than other wildland census blocks;
buildings in WUIX census blocks that are dominated by rainforest are at less wildfire risk than
their counterparts that are amidst highly flammable, dry vegetation. Representing all WUIX
census blocks on Maui as being equal with each other, through one classification, is misleading
and disregards the varying climate and regional biomes of Maui, and the same applies to NonWUI Wildlands census blocks. Critical distinctions are necessary to communicate the varying
fire risk within the original standard WUI classifications. To show the critical distinctions
between areas of Maui that have vegetation of varying levels of flammability, the percentage of
forest cover per census block is calculated and used to create a more detailed WUI map with
vegetation subclassifications. The forest cover is representative of Maui’s rainforests. The
percentage of grassland coverage is also calculated for additional context regarding Maui’s
flammable landscape.

60
Calculating the percent forest and percent grassland cover per census block is Step 5 in
the workflow diagram, and a more detailed workflow diagram of this step is shown in Figure 28.
To calculate forest and grassland cover, several of the steps for vegetation cover percentage
calculation are repeated two more times, for both forest and grassland cover, based on the
original ESA landcover dataset. The ESA dataset is reclassified so that only tree cover receives a
value of 1 and all other classifications are assigned a 0. This is repeated for grassland cover as
well. The agricultural plots are removed from the results of both. Then, the Zonal Statistics as
Table tool is used twice, to gather both the mean forest and grassland cover values per census
block. The values of each are multiplied by 100 to calculate percentages. The target attribute
fields from the grassland table and the forest table are joined to the census block feature layer
with the Join Field tool.
Figure 28. Workflow diagram for forest and grassland cover
The forest and grassland cover are shown in Figure 29. The island’s shrubland cover is
not represented on this map because of its relatively proportion, but it fills in many vegetation
cover gaps. Together, forest and grassland cover almost the entire island. For census blocks that
qualify as Non-WUI Wildlands, vegetation cover per census block on Maui is dominated either
by grassland or forest cover. East Maui is dominated by forest cover, as well as the mountainous

61
parts of West Maui. These two areas are Maui’s large rainforests. The rest of the island is
dominated by grassland.
Figure 29. Forest and grassland cover
The newly calculated forest cover is used to alter WUI classifications in the enhanced
WUI map product. Census blocks that are dominated by forest cover are renamed to show that
their qualities stray from the assumption of flammability. WUIX is divided into WUIX and
WUIX Forest, and Non-WUI Wildlands is divided into Non-WUI Wildlands and Non-WUI
Wildlands Forest; these classifications are divided because they already have ≥ 50% vegetation
cover, and the other classifications are unaffected. Census blocks that are ≥ 50% forest cover

62
receive the additional “Forest” descriptor, indicating that they are majority forest-cover and that
they have less wildfire risk compared to their counterparts.
3.3.6 WUI Map Enhancements with Forest Cover
The process of incorporating the forest cover data to enhance the standard WUI map with
vegetation subclassifications is Step 6 of the workflow diagram, and a detailed workflow
diagram of this step is shown in Figure 30. The Select by Attributes tool is used to select census
blocks where the WUI classification is “Intermix” and the forest cover is ≥ 50%. With the
Calculate Field tool, this selection is assigned to “Intermix Forest.” The selection is cleared, and
the Select by Attributes tool is used to select census blocks where the WUI classification is
“Non-WUI Wildlands” and the forest cover is ≥ 50%. This selection is assigned to “Non-WUI
Wildlands Forest” with the Calculate Field tool. The grassland cover percentages are not used to
affect classification because grassland on Maui matches the standard assumption that WUI areas
are highly-flammable. The Non-WUI Wildlands and WUIX census blocks that are not
dominated by forest are dominated by grassland, and they keep their “Non-WUI Wildlands” and
“Intermix” classifications. The WUIX Forest classification is symbolized with a browner tone of
orange than its WUIX counterpart. The Non-WUI Wildlands classification is symbolized with a
paler green to represent grassland, and the Non-WUI Wildlands Forest classification is
symbolized with a darker green to represent forest.

63
Figure 30. Workflow diagram for enhancing the WUI map
3.4 Refining Landcover Mapping
Refining the WUI methodology to be suitable for Maui required landcover dataset
experimentation. Standard WUI mapping practice uses one landcover dataset to gather both
agriculture and vegetation landcover information and removes all agricultural classifications
from vegetation cover calculations due to their relatively lower flammability. However, these
standard practices are inappropriate for the study area of Maui due to its unique agricultural
landscape. Choosing the most accurate landcover dataset and appropriately incorporating
agriculture is critical. In this thesis, different landcover datasets are tested to assess see how their
use affects WUI classification outcomes. Three landcover datasets are tested to assess their WUI
outputs, then the chosen landcover dataset is tested with different agricultural landcover
classification refinements.
A WUI map is created from each of the EVT, FVC, and ESA datasets with their original
agricultural classifications to compare the effect of their use on WUI mapping outcomes. The
three datasets show large discrepancies in agricultural landcover classification. Due to the

64
disagreements among landcover datasets in agricultural classification and the potential of the
agriculture classification to affect WUI outcomes, authoritative agricultural data from the Hawaii
Statewide GIS Program is incorporated into the analysis to increase accuracy in the vegetation
cover aspect of WUI calculation. The agricultural plot boundaries from the authoritative dataset
do not align with the agricultural classifications in the EVT, FVC, or ESA datasets. Because of
the strong agricultural past in Maui and the overrunning of many old croplands with invasive
grasses, the difference between historic agricultural lands and current agricultural lands can be
difficult to distinguish in remote sensing.
The ESA dataset is chosen as the landcover dataset for further analysis and enhancement
over the LANDFIRE datasets due to its finer 10m resolution, its worldwide coverage, and remote
sensing source. However, comparing the agricultural classification in the ESA dataset to the
authoritative agricultural dataset shows that the ESA classification is inaccurate. Where the ESAclassified agriculture does not overlap with an existing agricultural plot, the misclassified land is
mostly likely grassland, due to Maui’s history of abandoned agricultural plots being overrun with
grasses. For this reason, the land that is classified as agriculture by the ESA dataset is initially set
to vegetation in the final methodology, so it is effectively disregarded. Then, the official
agricultural land (some of which does overlap with the ESA classification) is removed from
vegetation and wildland calculations.
Removing pastureland from vegetation cover calculations along with other agricultural
land uses as in standard practice is inappropriate for Maui, and pastureland should be separated
from other agricultural land use types in the Hawaii Statewide GIS Program dataset. Pastureland
stands out among the other agricultural land uses in the dataset because it is not highly
maintained or irrigated. For example, in Figure 31, land growing taro (left) and macadamia nuts

65
(center) is lush, upkept, and green. In sharp contrast, pastureland (right) is dry, untamed,
unirrigated, flammable grassland. Although different agricultural species have varying levels of
flammability, and some grasses used for pastureland have low flammability (Pagadala et al.
2024), this is not the case on Maui because of the nature of the flammable invasive grasses
present in the pastureland. Kikuyu grass is the most important pasture grass on Hawaii
(Fukumoto and Lee 2003), but kikuyu grass is one of the highly flammable invasive grasses that
threaten the Hawaiian Islands, along with buffelgrass, fountain grass, and others that dominate
the landscape in Maui (Hawaii Invasive Species Council 2024a).
Figure 31. Taro, macadamia, and pasture agricultural land
The ESA dataset is used to make two more WUI maps. The second ESA WUI map is
made using all agricultural land use classifications from the Hawaii Statewide GIS Program
agricultural dataset, including pastures. Removing all agricultural plots from vegetation cover
calculations alters the WUI classification outcomes negatively and is inappropriate for Maui, so
the agricultural considerations are refined. The third ESA WUI map excludes all agriculture plots
except for pastureland from vegetation cover calculations, and this map is used as the final WUI
map with standard WUI classifications.

66
This thesis’s methodology and final WUI map products, using the ESA landcover dataset,
the Hawaii Statewide GIS Program agricultural dataset, and including pastures in vegetation
cover, stem from these experiments. The agricultural landcover classification from the ESA
landcover dataset is disregarded and the Hawaii Statewide GIS Program agriculture dataset
determines the agricultural classifications in this study’s final WUI mapping analysis.
Pastureland is classified as vegetation cover due to its high flammability, while the other 10
agricultural land uses are removed from vegetation cover calculations, as is typical for
agricultural land.

67
Chapter 4 Results
This chapter describes the results of this thesis. The standard WUI map and the final WUI map
with additional Maui-specific subclassifications are shown. Then, the WUI mapping results from
the landcover dataset and agricultural data experiments that were conducted while developing the
final WUI mapping methodology are presented. Each WUI map is symbolized in accordance
with common practices from WUI literature and followed by a summary statistics table and pie
chart of the WUI cover in each map. Map figures are shown with 40% transparency on the WUI
feature layer to show the additional context of the terrain.
4.1 Final WUI Map
The WUI analysis in this thesis results in two WUI maps. The first map has four standard
WUI classifications: WUIF, WUIX, Non-WUI, and Non-WUI Wildlands. The second WUI map
builds off the first and adds a Forest subclassification to WUIX and Non-WUI Wildlands based
on their vegetation content. This addresses Maui’s unique multi-climate landscape, which
contains both grasslands and rainforests. The additional WUI subclassifications show the
nuances within the standard WUI classifications.
4.1.1 WUI Map with Standard WUI Classifications
Created with the ESA landcover dataset and improved with the Hawaii Statewide GIS
Program agricultural dataset, the standard WUI map with agricultural enhancements is shown in
Figure 32. Most of Maui’s area is covered by Non-WUI Wildlands. The only Non-WUI areas are
around the volcano, by the tip of the southern coast, around some agricultural plots in Central
Maui, and in small census blocks around the coast. The most densely developed areas are WUIF,

68
and WUIX is found in more rural areas around the island. This WUI map is highly intuitive and
meets assumptions about WUI on Maui based on its geography.
Figure 32. Standard WUI map with agricultural enhancements
The standard WUI classification results are shown in Table 4, compared to about 7% in
the ESA WUI map.
Table 10. The results indicate that WUIF covers under 3% of Maui’s area, while NonWUI covers about 8%. WUIX covers about 25% of the land, and Non-WUI Wildlands cover

69
about 65%. The number of census blocks in each classification is misleading in regard to area
coverage because urban census blocks are smaller than rural and wildland census blocks.
Table 4. Standard WUI classification outcomes
WUI Classification Area (square km) Percentage of Maui
(by area)
Number of Census
Blocks
Interface 48.3 2.56% 712
Intermix 469.8 24.91% 419
Non-WUI 148.9 7.89% 76
Non-WUI
Wildlands
1,219.3 64.64% 240
A pie chart of the WUI makeup of the standard WUI classifications is shown in Figure
33. WUIF and Non-WUI cover the smallest portions of Maui’s area, while WUIX and Non-WUI
Wildlands cover large swaths of land. Altogether, WUI areas cover a little over a quarter of
Maui.

70
Figure 33. Standard WUI classification makeup
4.1.2 Final WUI Map with Enhanced WUI Classifications
Enhancing the standard WUI map with vegetation cover information culminates in the
final WUI map with Maui-specific subclassifications, shown in Figure 34. The WUIX
subcategories are represented in two shades of orange, with the WUIX Forest classification
having a slightly greener tone. The Non-WUI Wildlands subcategories are shown in two shades
of green, with Non-WUI Wildlands having a lighter tone and Non-WUI Wildlands Forest having
a darker tone. Compared to the standard WUI map with only four classifications, the six
classifications shown in the final WUI map exhibit significantly more nuance across the island.
Rainforest coverage is distinct in East Maui and West Maui; the regional differences in
grassland and rainforest cover are now clear. Many previously WUIX census blocks are now
WUIX Forest, especially around the Non-WUI Wildlands Forest census blocks in East Maui, the
North Shore, and West Maui. Intuitively, the grassland-dominated census blocks of WUIX and
Non-WUI Wildlands classifications are grouped together, and the rainforest-dominated census
blocks of WUIX Forest and Non-WUI Wildlands Forest are also found together. This final WUI
map succeeds in incorporating the vegetation variations within Maui’s landscape, and the WUI
classifications reflect these differences accordingly.

71
Figure 34. Final WUI map with Maui-specific subclassifications
The WUI classification coverage outcomes for the WUI map with subclassifications are
shown in Table 5, subdividing the previous WUIX and Non-WUI Wildlands classification
coverages. WUIF and Non-WUI are not affected because these census blocks are not dominated
by vegetation; the WUIF and Non-WUI percentages of Maui’s land area remain at about 3% and
8%, respectively. WUIX, which overall covers almost a quarter of Maui, is divided into the new
WUIX, which covers about 16% of Maui’s land area, and WUIX Forest, which covers about 9%.
Non-WUI Wildlands is divided into the new Non-WUI Wildlands, which covers about 33% of
Maui, and Non-WUI Wildlands Forest, which covers about 31%. Both rainforest
subclassifications cover less land area than their grassland-dominated counterparts.

72
Table 5. ESA WorldCover WUI classification outcomes
WUI Classification Area (square km) Percentage of Maui
(by area)
Number of Census
Blocks
Interface 48.3 2.56% 712
Intermix 300.9 15.95% 263
Intermix Forest 168.9 8.95% 156
Non-WUI 148.9 7.89% 76
Non-WUI
Wildlands
625.9 33.18% 174
Non-WUI
Wildlands Forest
593.4 31.46% 66
This thesis’s final WUI classification outcomes are displayed in a pie chart in Figure 35,
which accurately represents the WUI coverage of Maui and its nuances. While wildlands cover
most of Maui, they are split almost evenly between grassland-dominated and forest-dominated.
The same is true for WUIX classifications.

73
Figure 35. Final WUI classification makeup
There are about 63,000 buildings and 154,100 people on Maui, and the pie charts in
Figure 36 reveal the proportion of buildings and people within the WUI. The pie charts show that
most buildings and people are in WUIF census blocks, followed by WUIX, then WUIX Forest.
The results of this thesis prove that over half of Maui’s population is in WUIF. Very few
buildings and people are in Non-WUI, Non-WUI Wildlands, and Non-WUI Wildlands Forest
census blocks. When viewed altogether, the pie charts reveal that while the WUI covers a small
part of Maui by landcover, almost all buildings and people in Maui are situated in the WUI.

74
Figure 36. Buildings and population by WUI classification

75
The percentages of buildings and population in the WUI, with distinctions between
Forest WUI and Non-Forest WUI, are shown in Table 6. Forest subclassification significantly
alters the distribution, and shows nuance within the land, buildings, and population in the WUI.
While 27.46% of Maui’s land area is in WUI, 18.51% is in Non-Forest WUI. Showing an almost
20% difference, 96.82% of buildings are in WUI, while 78.4% of buildings are in Non-Forest
WUI. Almost all people, 99.03%, are in the WUI, but this drops to 87.82% when WUIX Forest
is excluded. While almost all buildings and people are in WUI, 18.42% of buildings and 11.21%
of people are in WUI areas with less wildfire risk than their WUI counterparts because they are
within the rainforest ecosystem on Maui. Forested WUI covers about 9% of Maui’s land area.
Table 6. Land, buildings, and people in WUI
In WUI In WUI (Non-Forest) In WUI (Forest)
Land 27.46% 18.51% 8.95%
Buildings 96.82% 78.4% 18.42%
People 99.03% 87.82% 11.21%
4.2 Landcover Dataset Refinement WUI Maps
Multiple WUI maps resulted from comparing the effect of the vegetation and agricultural
landcover classifications from different landcover datasets on WUI mapping results, and then
adding a separate authoritative agriculture dataset. WUI classification results change
significantly based on vegetation cover designations. The difference in agriculture classification
is the most impactful aspect of vegetation cover for this study area.

76
4.2.1 LANDFIRE EVT Dataset WUI Map
The most stand-out quality of the EVT WUI map, shown in Figure 37, is that the region
connecting Central Maui and Upcountry Maui is overwhelmingly classified as Non-WUI.
Furthermore, many coastal areas in West Maui are also designated as Non-WUI. The large swath
of classified agricultural land in the dataset disqualifies many census blocks that may have been
classified as Non-WUI Wildlands by lowering their vegetation cover percentages, instead
leaving them as Non-WUI. The 2.4-kilometer wildland buffer affects WUIF classification in this
map. The buffer does not cover the whole island, missing most of Central Maui and a small
portion of West Maui’s coast. Hundreds of census blocks initially meet the WUIF definition by
vegetation cover and building density but are classified as Non-WUI because they are outside
2.4-kilometers from wildlands. For West Maui, Central Maui, and Upcountry Maui, their NonWUI classifications are inconsistent with expectations because these regions have dense
populations, frequent wildfires, high wildfire risk, and are generally regarded as WUI areas. The
barren qualities of the volcano and part of the southern coast are evident in this map, as they are
shown as Non-WUI. The coastal communities in East Maui are mostly WUIX, while the coastal
communities on West Maui are mostly WUIF until they transition to WUIX closer to Non-WUI
Wildlands. Upcountry Maui has an East to West pattern of Non-WUI Wildlands, WUIX, WUIF,
then Non-WUI.

77
Figure 37. LANDFIRE EVT WUI map
A summary statistics table of the EVT WUI map is shown in Table 7, and shows that
Non-WUI Wildlands cover about half of Maui. Per these results, Non-WUI Wildlands cover
about 53% of the island, while Non-WUI covers about 22%. WUIX covers about 15% and
WUIF covers about 10%. With almost a quarter of the island’s area, the EVT WUI map has
more Non-WUI areas than expected and its legitimacy is questionable.

78
Table 7. EVT WUI classification outcomes
WUI Classification Area (square km) Percentage of Maui
(by area)
Number of Census
Blocks
Interface 184.9 9.80% 691
Intermix 283.2 15.01% 108
Non-WUI 414.5 21.98% 563
Non-WUI
Wildlands
1,003.6 53.21% 85
4.2.2 LANDFIRE FVC Dataset WUI Map
The biggest differences in the FVC WUI map, shown in Figure 38. LANDFIRE FVC
WUI mapFigure 38, are the classification changes to Central Maui and Upcountry Maui. Central
Maui is shown as mostly WUI with Non-WUI Wildlands census blocks on the outskirts.
Upcountry Maui has Non-WUI Wildlands census blocks to the west of its WUI census blocks.
The newly classified Non-WUI Wildlands are in an area classified as Non-WUI in the EVT WUI
map. The 2.4-kilometer buffer also did not cover the entire island and affected WUIF
classifications. The classifications of the communities in East Maui, West Maui, and South Maui
appear nearly identical to the EVT WUI Map.

79
Figure 38. LANDFIRE FVC WUI map
As seen in Table 8, the largest shift in classifications from the EVT WUI map to the FVC
WUI map is from Non-WUI to Non-WUI Wildlands. Non-WUI Wildlands account for about
61% of Maui’s area, an increase from the 53% in the EVT WUI map. The WUIX classification
increases to about 17% of Maui, while Non-WUI decreases to about 11%. While the number of
WUIF census blocks increases drastically, the percentage of Maui does not vary significantly;
WUIF remains at about 10%.

80
Table 8. FVC WUI classification outcomes
WUI
Classification
Area (square km) Percentage of Maui
(by area)
Number of Census
Blocks
Interface 188.3 9.98% 961
Intermix 325.3 17.25% 131
Non-WUI 215.4 11.42% 202
Non-WUI
Wildlands
1,157.2 61.35% 153
Demonstrated by the results, usage of the FVC dataset for WUI mapping is an
improvement because Central Maui is shown as WUI, and the Non-WUI Wildlands designation
fits the vegetative landscape between Central Maui and Upcountry Maui more accurately.
However, there are still concerns with agricultural accuracy. The large swath of Non-WUI
census blocks in West Maui is also still note for concern.
4.2.3 ESA WorldCover Dataset WUI Map
The ESA WUI map classification results are significantly different than the WUI maps
produced by the LANDFIRE landcover datasets, and the ESA WUI map is shown in Figure 39.
Upcountry Maui is largely WUIX, rather than WUIF. This is due to the higher vegetation cover
assignments within these census blocks because less land is classified as developed.
Furthermore, the number of census blocks classified as Non-WUI significantly decreases. This
reduction is most noticeable in West Maui and Central Maui. Since more land is classified as
grassland instead of agriculture, the vegetation cover in these areas rises. Also, the 2.4-km
wildland buffer extends to cover almost the entire island. Therefore, no census blocks that met
the WUIF definitions for building density and vegetation cover are assigned to Non-WUI for
falling outside the buffer, as occurred in the LANDFIRE WUI maps.

81
Figure 39. ESA WUI map
As seen in Table 9, WUIF and Non-WUI classifications are at their lowest area coverage,
while WUIX and Non-WUI Wildlands are at their highest. WUIF coverage drops from about
10% of Maui’s area in the LANDFIRE WUI maps to about 2%. Non-WUI covers about 7%, a
decrease from the 10s and 20s. WUIX covers about 25% of Maui, when it was less than 20% in
the LANDFIRE WUI maps. Non-WUI Wildlands cover about 66%, an increase from the low 50s
and 60s.

82
Table 9. ESA WUI classification outcomes
WUI
Classification
Area (square km) Percentage of Maui
(by area)
Number of Census
Blocks
Interface 38.1 2.02% 702
Intermix 480.0 25.45% 429
Non-WUI 125.6 6.66% 66
Non-WUI
Wildlands
1,242.6 65.87% 250
The ESA WUI map product proved the most intuitive of the three landcover dataset
experimentation results. Few census blocks on Maui are classified as Non-WUI, communities in
West and Central Maui fall under WUI classifications, communities in Upcountry Maui are
WUIX interwoven with Non-WUI Wildlands, and Non-WUI Wildlands coverage is high. Each
of these observations is consistent with patterns in Maui’s development and vegetation. Although
ultimately enhanced with Hawaii Statewide GIS Program agricultural data, these classification
distributions are similar to the standard WUI map.
4.2.4 Landcover WUI Comparisons
The WUI mapping experimentation with different landcover datasets exhibits that
vegetation and agriculture cover landcover designations drastically alter WUI classification
results. The area, percentage of Maui, and number of census blocks in each WUI classification
vary between the three WUI maps created from the EVT, FVC, and ESA landcover datasets. The
percentage of Maui’s WUI makeup by area is shown in a pie chart for each of the three
landcover datasets in Figure 40. WUIF coverage ranges from 2-10%, WUIX ranges from 15-
25%, Non-WUI ranges from 7-22%, and Non-WUI Wildlands ranges from 53-66%. The
distribution of WUI classifications is most evenly spread in the EVT WUI map. In the FVC WUI

83
map, the proportion of WUIF, WUIX, and Non-WUI are the most similar. The ESA WUI map
shows a more unequal WUI classification distribution, with each classification covering a
distinctly different proportion of Maui; this WUI map has the largest range of area coverages,
from 2-66%.
Figure 40. WUI classification makeup per landcover dataset

84
4.2.5 ESA WorldCover and All Hawaii Statewide GIS Program Agricultural Data WUI Map
The WUI map created from the ESA landcover dataset and Hawaii Statewide GIS
Program agricultural dataset, which removes all 11 classes of agriculture from vegetation cover
calculation, is shown in Figure 41. Pastures are over 90% of the agricultural land in the
agricultural dataset and cover over a quarter of the island, so removing the pastures from
vegetation cover calculations affects WUI classification calculations drastically, compared to the
previous ESA WUI map. The differences are especially noticeable in Upcountry Maui and
around the southern coast. While the tip of the southern coast has Non-WUI classification in the
ESA WUI map, the number of Non-WUI census blocks in this map increases drastically and
extends into South Maui, Upcountry Maui, and East Maui. More Non-WUI census blocks appear
in Central Maui and the North Shore as well. In every region, many census blocks classified as
WUIX in the ESA WUI map are switched to WUIF. The trend is most apparent in Upcountry
Maui, where there appears to be an evenly distributed mix between WUIF and WUIX.

85
Figure 41. ESA and all agriculture WUI map
The WUI classification outcomes from the WUI map using ESA and all Hawaii
Statewide GIS Program agricultural data are shown in Table 10, and this is the first WUI map in
which Non-WUI Wildlands covers less than half of the land area of Maui. While the Non-WUI
Wildlands cover 2/3 of Maui in the ESA WUI map, the Non-WUI Wildlands in this map
decrease to about 46%. The WUIF coverage increases to about 12%, and the WUIX coverage
decreases to about 15%, significantly closing the 23% gap between WUIF and WUIX in the ESA
WUI map. Non-WUI increases greatly to about 27%, compared to about 7% in the ESA WUI
map.

86
Table 10. ESA and all agricultural data WUI classification outcomes
WUI Classification Area (square km) Percentage of Maui
(by area)
Number of Census
Blocks
Interface 218.6 11.59% 717
Intermix 290.2 15.38% 343
Non-WUI 505.2 26.78% 237
Non-WUI
Wildlands
872.4 46.25% 150
A pie chart of the ESA and all agricultural data WUI classifications is shown in Figure
42, and these classifications are very misleading. These classifications show a significant
proportion of census blocks as Non-WUI instead of Non-WUI Wildlands, although many of
them are highly flammable. The large amount of land removed from vegetation cover
calculations also causes the shift to higher WUIF classifications, which is inaccurate in
Upcountry Maui because the communities are interwoven with flammable wildlands.
Figure 42. ESA and all agricultural data WUI classification makeup

87
Chapter 5 Dashboard Creation
This chapter describes the creation of a publicly available ArcGIS Dashboard built around this
thesis’s final WUI map with Maui-specific subclassifications. The Dashboard is the final product
of this thesis. The Dashboard combines the findings of this thesis with supplemental WUI-related
and wildfire management-related data to promote a well-rounded perspective on WUI and its
relation to other measures of wildfire risk. Existing WUI web applications are discussed, the
Dashboard-building methodology is described, and then the Dashboard product is shown. The
Dashboard can be accessed at the following link:
https://uscssi.maps.arcgis.com/apps/dashboards/c6be46ed9b5549bd89c32a47137cd5b4.
5.1 Web-Sharing Related Literature
Many WUI maps are shared as web applications to communicate WUI information with
the public in an accessible and engaging manner for exploration. The global WUI map created by
Schug et al. (2023), discussed in Chapter 2.3.2, is viewable on a web application. On the web
application, the WUI layer transparency can be adjusted. The WUI data can be viewed over other
basemaps and data, including the Global Human Settlement data and the ESA WorldCover data.
From the products of Radeloff et al. (2023a), discussed in Chapter 2.3.42.2.2.2, the 2020 zonalbased WUI map of the 48 conterminous US is shared by the US Forest Service as an ArcGIS tile
layer map service (Radeloff et al. 2023b). An ArcGIS StoryMap supported by the USDA further
introduces readers to the WUI and explores the data from Radeloff et al. (2023b) on the WUI
from 1990-2010 from different angles (Mockrin et al. 2023). The findings of Ketchpaw et al.
(2022), described in Chapter 2.2.1, are shared in an ArcGIS Dashboard, where the different WUI
map products of Montana that are calculated by different methods can be explored side-by-side.

88
5.2 Dashboard Methods
An ArcGIS Dashboard is used for this product because the ArcGIS platform seamlessly
integrates maps from ArcGIS Online. The interactive Dashboard allows viewers to explore the
multiple facets of WUI classification, as well as assess their relation to other wildfire-related
content. The methods are overviewed, the data is described, and the steps of the methodology are
detailed.
5.2.1 Dashboard Workflow Overview
A workflow diagram of the steps for this thesis’s Dashboard creation is shown in Figure
43. Creating the Dashboard product from this thesis’s final enhanced WUI map entails four
steps: Step 1 is adding more attributes to the census blocks in the final WUI map for additional
context, Step 2 is publishing all map layers and creating a Web Map from them, Step 3 is
creating the Dashboard and centering it on the Web Map, and Step 4 is creating the widgets for
the Dashboard. The Dashboard is configured with dynamic and static widgets that highlight the
various elements of the WUI analysis and summarize its final results, as well as present other
wildfire-related information.

89
Figure 43. Workflow diagram creating ArcGIS Dashboard
5.2.2 Dashboard Data Description
Data used in WUI analysis, data created from this thesis, and additional wildfire-related
data are used in the Dashboard and are portrayed in Table 11. Most of the data presented in the
Dashboard is used in WUI analysis; the Maui building footprints, census blocks, and agricultural
land use are used and described in Chapter 3. Further utilizing the census block data, housing
unit data and population data within census blocks are used in the Dashboard. The 2023 wildfire
extents are shown in Chapter 1.2.2 and are included in the Dashboard. The final enhanced WUI
map created from this thesis, shown and described in Chapter 4.1.2, is presented as the focal
point of the Dashboard. The remaining datasets to be described in this section are the Maui fire
station locations and the State of Hawaii-defined CARs.

90
Table 11. Dashboard data descriptions
Data Source Year Resolution Use Data Description
Building
Footprints
Maui County GIS
Dept.
2020 N/A,
vector
Context for building
distribution
Building footprint
polygons for Maui
Census
Blocks with
housing unit
and
population
census data
Hawaii gov,
sourced from US
Census Bureau
2020 N/A,
vector
Census data for
population and housing
unit counts, click on
census blocks to update
dynamic widgets
Census block polygons
Extents of
August 2023
wildfires
Maui County
GIS Dept.
2024 N/A,
vector
Context for locations
of 2023 wildfires
The extent of the
wildfires
Agricultural
Land Use
Hawaii Statewide
GIS Program
2020 N/A,
vector
Context for agricultural
land distribution
Polygons of existing
agricultural lands
categorized by crop/use
Enhanced
WUI
This thesis 2020 10m Centrally featured,
contains many
attributes for each
census block, including
WUI classification
The final WUI layer with
six enhanced WUI
classifications from the
results of this thesis
Maui Historic
Wildfires
Pacific Fire
Exchange
2005-
2020
N/A,
vector
Summarize the number
of historic wildfire
points per census block
Wildfire ignition points
from 2005-2020
Maui Fire
Stations
Hawaii Statewide
GIS Program
2017 N/A,
vector
Context for fire station
distribution
Point data of the 10 fire
stations located on Maui
CARs Hawaii Statewide
GIS Program,
sourced from
Hawaii DLNR
2017 N/A,
vector
Compare WUI
classification results
with State of Hawaii’s
CAR assessments in
Dashboard
Polygons of communities
at risk of wildfires, with
classifications of
Low/Medium/High risk
Maui fire station locations are included in the Dashboard for additional context and are
shown in Figure 44 (Hawaii Statewide GIS Program 2021). The ten fire stations on Maui are
spread across the island. Only one fire station is in East Maui. Fire stations are more
concentrated in West Maui, Central Maui, and South Maui. Exploring the relationship between
fire station locations, historical wildfires, WUI classifications, and other map layers is insightful
for viewers. Many relationships can be observed between the various map layers provided in the
Dashboard.

91
Figure 44. Maui fire stations
State of Hawaii-defined CARs are described in Chapter 2.3.3 and CARs on Maui are
shown in Figure 45 (Hawaii Statewide GIS Program 2024). The high-risk CARs are in West
Maui, Central Maui, and South Maui. Medium-risk CARs are in Upcountry Maui and parts of
West Maui and Central Maui. Low-risk CARs are in East Maui, the North Shore, and Upcountry
Maui. Most of the land area of Maui does not receive CAR designation.

92
Figure 45. CARs on Maui
5.2.3 Adding Attributes to Census Blocks
While the final WUI map layer contains many attributes from WUI analysis, each census
block’s number of historic wildfires, agricultural area (not including pastures), and CAR risk
rating are added as attributes, following Step 1 of the workflow diagram. To obtain the number
of historic wildfires per census block, the Summarize Within tool is used. The Input Polygons
are the census blocks, and the Input Summary Features are the fire incidence points. Using the
Join Field tool, the output counts are added to the census block layer. The census blocks are the

93
Input Table, the summarized wildfire counts are the Join Table, and the Input and Join Fields are
the Geographic Identifier. The Transfer Field is the wildfire counts.
To calculate the agricultural land per census block, the Summarize Within tool is used.
The Input Polygons are the census blocks, the Input Summary Features are the agricultural plots
without pastureland, and the Shape Unit is square kilometers. Then, the Join Field tool is used to
add the agricultural area attribute to the census blocks. The Input Table is the census blocks, the
Join Table is the summarized agricultural areas, and the Input and Join Fields are the Geographic
Identifier. The Transfer Field is the agricultural area.
The Spatial Join tool is used to assign each census block its CAR risk rating. The Target
Features are the census blocks, the Join Features are the CARs, the Join Operation is Join Oneto-One, and the Match Option is Largest Overlap. The output is joined to the census blocks layer
with the Join Field tool. The Input table is the census blocks, the Join Table is the output from
the Spatial Join, the Input and Join Fields are the Geographic Identifier, and the Transfer Field is
the risk rating.
All census data attributes are removed except for block number, decennial population
count, and decennial housing count. Attributes from analysis that are kept are the Geographic
Identifier, WUI classification, census block area in square kilometers, building density (per
square kilometer), number of buildings in census block, percent vegetation cover, percent forest
cover, and percent grassland.
5.2.4 Preparing Web Map
All map layers to be featured in the Dashboard are added to a Web Map that becomes the
basis of the Dashboard, following Step 2 of the workflow diagram. The layers are first published,
and then added to a Web Map. From ArcGIS Pro, the final WUI feature layer is published as a

94
Web Layer to ArcGIS Online. Other layers that are published are fire stations, wildfires, building
footprints, the August 2023 wildfire extents, the State of Hawaii’s designated CARs, and
agricultural land use. All layers are added to the Web Map. The chosen basemap is the Terrain
with Labels basemap and the WUI classification layer has a transparency of 35% so the terrain is
visible through the data for geographic context. The Web Map is designed with the layers in the
following order: Fire stations, Wildfires 2005-2020, Buildings 2020, August 2023 Wildfire
Extents, Communities at Risk – per State of Hawaii, Agricultural Land Use 2020, and Maui
Wildland-Urban Interface. All layers are initially turned off, except for the WUI layer. The WUI
layer is last so other data layers can be viewed on top of it. Point layers are arranged above
polygon layers for optimal viewing.
5.2.5 Dashboard Structure Formatting
Creating a new ArcGIS Dashboard is Step 3 of the workflow diagram. ArcGIS
Dashboard design revolves around arranging elements. The ArcGIS Dashboard builder includes
11 element options, shown in Figure 46. The options include map, map legend, serial chart, pie
chart, indicator, gauge, list, table, details, rich text, and embedded content. A header and sidebar
are also options. The Dashboard for this thesis includes a map, map legend, pie charts, indicators,
details, as well as a header and sidebar. The pie charts and indicators are the most frequent
elements in this Dashboard; the pie charts are used for presenting summary information and the
indicators are used to present various WUI and wildfire-management related metrics. As seen in
Figure 46, elements can be placed on top, below, or to the left or right of other elements. The
arrangement of the Dashboard in this thesis project is designed to maximize the amount of
information presented at once, while still being easily digestible for viewers outside of the
scientific community.

95
Figure 46. ArcGIS Dashboard elements
The central element of the Dashboard is the Web Map that holds the WUI classification
layer and other data layers. The additional data layers are available to toggle and overlay on top
of the WUI layer to see their relationships with each other and with the WUI classifications. The
user has the option to change basemaps.
A panel is added to the far left of the Dashboard that explains the Dashboard and
provides context into Maui and its unique relationship with the WUI. The panel has the
following sections: Background, Purpose of the Dashboard, This Project’s WUI Definitions,
Maui-Specific Adjustments, and Additional Reference Data. The Background introduces what
WUI is and the Maui August 2023 wildfires. The Purpose of the Dashboard section describes the

96
Dashboard and WUI map, their novelty and contribution, and instructions on use. The panel
helps the reader to interpret the information presented by the Dashboard. The project-specific
WUI definitions overview the six classifications on the central WUI map. The Maui-Specific
Adjustments section overviews the agricultural adjustments made in the WUI analysis and the
forest subclassifications. The last section addresses the additional data used in the Dashboard and
provides links to the original data sources.
5.2.6 Static and Dynamic Widgets
The Dashboard has static information displays as well as interactive functionality.
Adding static and dynamic widgets is Step 4 of the workflow diagram. Static information is
gathered on the right side of the Dashboard’s central Web Map. Dynamic metrics are aligned
above the Web Map, across the top of the Dashboard. Supplemental elements are placed to the
Web Map’s left side, between the panel and the Web Map.
The static information on the map summarizes major findings of the WUI analysis. Pie
charts are shown statically and are always visible because their data encompasses the entire
island and provides critical insights into the WUI on Maui as well as the distribution of people
and buildings within the WUI. The WUI classification pie chart is sized largely because this
distribution is one of the most important results of the WUI analysis. Two smaller pie charts
showcase the portion of total buildings and population distributed within the WUI classifications,
respectively. The total buildings and population on Maui are shown as indicator numbers so
viewers grasp the content of the pie charts. The pie charts can be hovered over with a mouse to
reveal the percentage and number of buildings or population for each WUI classification.
The dynamic nature of the map allows metrics to update based on each census block’s
WUI classification and other attributes. The Dashboard allows users to click on any census block

97
in the WUI map of Maui, and the building density, percent vegetation cover, percent grassland
cover, percent forest cover, agricultural area, and number of historic wildfires in that census
block will appear. With these metrics, viewers can assess how each census block’s metrics relate
to the WUI definitions. These metrics change every time the viewer clicks on a different census
block. Of the metrics, building density and total percentage vegetation cover are presented most
prominently because they are the two most important factors of standard WUI mapping. The
metric for percentage of forest cover is displayed smaller because it is a Maui-specific metric and
not part of standard WUI definitions. The agricultural area in square kilometers for the census
block (which does not include pastureland) is shown because it is removed from vegetation cover
calculations and impacts the percentage vegetation cover in WUI analysis. This metric also
provides insight into the agricultural makeup of the different areas around Maui. The percentage
of grassland and the number of historic wildfires from 2005-2020 in the selected census block
are displayed relatively small because they do not affect WUI analysis but are included for
additional context. The percent grassland and percent forest cover do not always add up to the
total percent vegetation cover because shrubland also counts towards vegetation cover, but
shrubland is not given a dynamic metric on the Dashboard because of its relatively small
contribution.
To the left of the Web Map are supplemental elements, the WUI classification legend and
additional census block attributes. The legend is necessary to interpret the six final WUI
classifications and subclassifications: Interface, Intermix, Non-WUI, Non-WUI Wildlands,
Intermix Forest, and Non-WUI Wildlands Forest. The attributes are the WUI Classification,
Census Block Area (sq km), Building Density (sq km), Number of Buildings, Decennial Housing
Count, % Veg Cover, % Forest, % Grassland, Agricultural Area (sq km), # Wildfires 2005-2020,

98
and Fire Risk Rating, per HI Gov. Above the attributes, the title that pops up for each census
block is the block number and its WUI Classification. The attributes update dynamically each
time a new census block is clicked.
5.3 Dashboard Results
A screen capture of the Dashboard is shown in Figure 47. As the user explores the WUI
map and clicks on different census blocks, many metrics on the Dashboard update dynamically.
Dynamic metrics are shown from a selected WUIX census block in Upcountry Maui. The
selected census block is highlighted in blue. At the top of the Dashboard, the viewer can see that
the census block has a building density of 18.6 buildings per square kilometer and 95.4%
vegetation cover; these metrics explain why the selected census block is classified as WUIX.
Figure 47. Maui WUI ArcGIS Dashboard

99
Through the dynamic metrics, the Dashboard highlights the nuances in building density
and vegetation cover in census blocks that fall under the same WUI classification. The dynamic
portion, which stretches across the top of the Dashboard, is shown in Figure 48. The metrics in
this screenshot correspond to a different selected census block than the one in Figure 47,
showcasing the dynamic nature of these elements. The building density of 845 buildings per
square kilometer in this census block far surpasses the WUI threshold of 6.18, and the vegetation
cover is well below 50%, indicating that this census block is WUIF. The census block has 13.6%
forest cover and 8% grassland cover. There is no agricultural land in the census block, and there
were 10 wildfires from 2005-2020.
Figure 48. Dynamic metrics on Dashboard
Providing a summary overview of the WUI conditions on Maui, the static data on the
Dashboard is shown in Figure 49. The static portion of the Dashboard, on its right side, holds pie
charts from Figure 35 and Figure 36 that summarize findings from the final WUI map. The three
pie charts are discussed in Chapter 4.1.2. The large pie chart on top represents the thesis project’s
final WUI classification makeup of Maui. The total number of buildings and the total population
of Maui are shown as static indicator numbers; next to the indicator numbers are smaller pie
charts for building and population distributions within WUI classifications.

100
Figure 49. Static portion of Dashboard
The additional WUI-related and wildfire management data layers in the Dashboard are
shown in Figure 50, when the layers button is clicked on from inside the Dashboard’s central
WUI map. These layers contain valuable supplementary information that enhance the WUI
classification map and provide context into wildfire risk around Maui. All layers are initially all
turned off, but each can be toggled as the viewer desires.

101
Figure 50. Data layers for additional context

102
Chapter 6 Conclusions
This chapter discusses the contributions this thesis brings to WUI literature, the findings of this
thesis project, and the significance of studying WUI in Maui. The broad audience and
applications of the Dashboard product are examined. This thesis project is compared to existing
Maui WUI projects and the limitations of the project are addressed. Finally, steps to maintain the
Dashboard and future research opportunities are discussed.
6.1 Contributions
As an applied research project, this thesis first provides a WUI map as an intellectual
contribution to WUI literature and wildfire management body of knowledge for Maui. Then, this
work is applied and drives the development of the final ArcGIS Dashboard product. The
methodology and enhancements to standard WUI mapping practice and web-based WUI sharing
in this thesis contribute to theoretical and methodological practices, and to technological
applications.
6.1.1 Theoretical Contributions
This thesis advocates for WUI maps to be created for local fire-prone areas such as Maui.
Wildfire management strategies must be done on a local level, emphasizing the need for local
WUI mapping. Fire-prone areas should also develop WUI management plans based on shared
common WUI classifications, not other geographic or political boundaries. Maui is divided into
six community plan districts and seven Maui County Council districts, which loosely align with
the Maui geographical regions. However, each district and region contain a variety of WUI and
Non-WUI classifications that must be treated differently in terms of wildfire management. Areas
under the same WUI or Non-WUI classification should have similar approaches for wildfires due

103
to their comparable density and vegetation makeups. Furthermore, while many wildfire
management plans are made for community districts on Maui, a more collaborative approach to
wildfire management and land use planning can be taken by connecting communities across the
island that have similar WUI classification and similar wildfire histories.
This thesis project highlights the importance of scale-dependent WUI mapping in local
wildfire management and planning, particularly when analyzing small study areas. While this
thesis assesses WUI at a regional scale by calculating WUI summaries across Maui, this thesis
also analyzes WUI properties at the detailed census block level. Insights at multiple scales can be
gathered from this research. Small, local projects that identify locations for wildfire mitigation
interventions might find the census block level of analysis with detailed density and vegetation
attributes useful. Alternatively, larger community-wide and regional WUI summaries can be
useful for more regional and countywide land use policies. Accounting for local assessment is
impossible on national and international WUI mapping scales, so mapping WUI at local scales is
beneficial for local wildfire management and planning.
6.1.2 Contribution to Methodology
The WUI mapping enhancements in this thesis contribute to WUI scientific literature.
This thesis promotes testing input landcover datasets on WUI map results before choosing the
most appropriate for the study area, to improve accuracy in WUI results. Choice of dataset is
very impactful for the study area of Maui; WUI classification land coverage results changed by
up to 20% based on the datasets used. The WUI classifications from each landcover dataset were
compared to observations of Maui, as well as historical, developmental, and geographic context
about the island, to determine the best choice of landcover dataset. Landcover classifications
mostly erred in distinguishing agricultural land from grassland. The initial agricultural

104
classifications from the chosen landcover dataset were treated as vegetation cover. Then, the
appropriate agricultural plots, sourced from a local and authoritative agricultural dataset, were
removed from vegetation cover calculations. Accurate agricultural representation was also
determined by testing and comparing agricultural classifications with the vegetative conditions
on Maui, resulting in the separation of pastureland from other agriculture types. Other studies
can implement these landcover dataset testing strategies to ensure that the selected landcover
datasets and vegetation cover classifications used are the most accurate and representative for
their study areas.
This thesis develops zonal-based methods to classify ecosystem-specific vegetative
landcover for the study area, separating forest- and grassland-dominated WUI and Non-WUI
Wildlands, and interprets the impact of vegetation type for the study area of Maui. While
distinguishing forest- and grassland-dominated areas is precedented in Schug et al. (2023) in
point-based analysis, this thesis defines census blocks as forest-dominated if they have forest
cover ≥ 50%. Furthermore, reinterpreting the impact of vegetation type is critical for study areas.
Because the forest on Maui is rainforest, and not high-wildfire risk, woody forests, such as are
found most places around the world, the distinction between forest- and grassland-dominated
WUI is necessary and changes the interpretation of wildfire risk in Maui. In other places, forests
have higher wildfire risk than grasslands, but this is not the case in Maui. Based on the principle
that WUI areas and Non-WUI Wildlands are inherently flammable, this project does not rename
census blocks as “Grassland” WUI or Non-WUI Wildlands, because grassland in Maui matches
these assumptions. Census blocks receive the “Forest” modification because rainforests stray
from this principle. These methods support the WUI mapping practice of separating vegetation
types based on the varying wildfire risk of vegetation in study areas.

105
6.1.3 Contribution to Technology
The ArcGIS Dashboard platform summarizes WUI information, additional wildfire
management data, and communicates information through spatially enabled widgets. The
improvements to sharing WUI knowledge through the web that were developed in this thesis
support technical advancements in WUI literature. The design of the Dashboard promotes
combining WUI information with other wildfire-related information. Data such as locations of
historic wildfires and fire stations are not part of WUI mapping practices, but the supplementary
information gives viewers a well-rounded view of wildfire risk and wildfire management
alongside WUI. This thesis supports the notion that WUI map products are more helpful when
viewed with other wildfire management information than viewed in isolation, and that
Dashboards should be used to share WUI-related information broadly and publicly.
Furthermore, the Dashboard improves upon the standard mapping practice that areas
under the same WUI classification are presented equally. WUI-contributing factors of building
density and vegetation cover are different across census blocks, even when they fall under the
same classification. Maui has a very high population density, and WUI census blocks vary
greatly over the threshold of 6.18 buildings per square kilometer. The over/under threshold of
50% vegetation cover also holds significant ranges that are glossed over in static map products.
By creating a dynamic product that shows the individual building density and vegetation cover
components for each census block, the Dashboard discloses the black box of each census block’s
WUI-contributing factors. The WUI-related factors are presented with transparency so viewers
can see the differences within WUI zones, allowing them to make their own assessments on the
varied danger each census block faces. By being able to explore the components that go into a
census block’s WUI classification, viewers have a deeper understanding of the wildfire-related
issues that an area faces.

106
6.2 Findings
The statistics produced in this study allow the prevalence of WUI on Maui to be
compared to other locations and reveal the high prevalence of WUI on Maui, even relative to
other fire-prone areas. The results of this thesis project demonstrate that the prominence of
buildings, people, and land within the WUI in Maui far exceeds that of the fire-prone western
states or Oceania. Shown in Table 1Table 6, over a quarter of land in Maui (27.46%) is in WUI,
and almost all buildings (96.82%) and people (99.03%) are in WUI. As discussed in Chapter
2.3.4, Radeloff et al. (2023a) reveal that the average percentage of housing units (although
different from building density, compared here as a close proxy) for the conterminous US is
31.6%, the average for the fire-prone western states is 48.7%, and the highest in the western
states is 80.1%. Maui far exceeds these metrics, with 96.82% of buildings in WUI. The average
land area in WUI for the conterminous US is 9.4%, the average for the fire-prone western states
is 3.5%, and the highest in the western states is 8.2%. Again, Maui far exceeds these metrics with
27.46% of land in WUI, about triple the conterminous US average.
As stated in Chapter 2.3.2, Schug et al. (2023) provide continental summaries of its
worldwide WUI study, including Oceania. In Oceania, less than 5% of the total land area is
WUI, and over 60% of people are in WUI; Maui exceeds each of these proportions significantly.
Furthermore, the proportion of forest-dominated WUI is higher than grassland-dominated WUI
across Oceania, which contrasts the results in Maui, where grassland-dominated WUIX (15.95%)
is almost double the land area of forest-dominated WUIX (8.95%). The prevalence of WUI in
Maui is greater than other study areas it may be compared to in literature.
The distribution of land, people, and buildings within WUI classifications vary in
proportionality. Most development and population in Maui are concentrated in WUIF. As

107
gathered from the pie charts in the Dashboard, the buildings and people in both WUIX and
WUIX Forest are relatively proportional to their land area coverages, when compared to the
distribution of WUIF. WUIX covers about 16% of the land area, holds about 36% of buildings,
and 27% of people. WUIX Forest covers about 9% of land area, 18% of buildings, and 11% of
people. Contrarily, WUIF covers the smallest land area, less than 3%, and yet holds the highest
percentages of buildings and people, with about 43% and 61%, respectively.
The results show significant nuance in WUI coverage when including or excluding
forest-dominated areas. For example, 99.03% people are in WUI and therefore appear to be at
imminent wildfire risk. Without WUIX Forest, the percentage of people in the WUI is about
88%, revealing that the danger to people in the WUI is not as drastic as it first appears. The WUI
outcomes show that a large portion of land, buildings, and people in WUI are at lower wildfire
risk, being surrounded by rainforest, rather than grasslands.
6.3 Comparison to Existing Maui WUI Maps
This thesis builds on efforts from the SILVIS Lab and State of Hawaii to comply more
with national standards and create a Maui-specific product. When compared to the existing WUI
maps of Hawaii created by Schug et al. (2023) and the Hawaii DLNR, the WUI map and
Dashboard created from this thesis have many benefits.
6.3.1 Comparison to SILVIS Lab’s Global WUI
While the goal of this thesis is to accurately assess WUI at a local scale in an area that
recently experienced disastrous wildfires, and the goal of the global WUI analysis from the
SILVIS Lab is to observe WUI trends at global and continental scales, this thesis product shows
the benefits of using a WUI map created for a specific study area when studying a small location,
rather than using the product from a global analysis. The building and agriculture locations are

108
more accurate for Maui in this thesis’s WUI analysis than those of the worldwide WUI analysis,
because local datasets are used and landcover is adjusted, rather than using only global datasets.
This thesis uses an accurate building footprint dataset, an improvement on the built-up surface
estimate dataset used in Schug et al. (2020) that does not contain building footprints. The ESA
landcover dataset used in this thesis is one year more recent than the ESA landcover dataset used
in Schug et al. (2023).
The overall trends are similar; while the product of Schug et al. (2023) shows less WUI
area, Schug et al. (2023) claim to likely be a conservative estimate, so this is to be expected.
Both show WUIX Forest around forests in East Maui, WUIX in Upcountry Maui, and WUIX
and WUIF in West Maui, Central Maui, and South Maui. However, Schug et al. (2023) show
Non-WUI areas around Lahaina and in Kahului, which are both densely developed areas with
frequent fires and high wildfire risk. The WUI map created in this thesis is more accurate in and
around these communities because they receive WUI classification.
The Dashboard product has additional functionality when compared to the web map
product created from the results of Schug et al. (2023) because the attributes of census blocks are
revealed when clicked on. The analysis in this thesis is simpler than the analysis conducted in
Schug et al. (2023), and the calculations are transparent for viewers in the Dashboard.
Furthermore, the Dashboard and WUI map from this thesis use colors that symbolize grasslanddominated WUI to be more flammable than forest-dominated WUI, which is appropriate for the
vegetative conditions of the Maui study area. The opposite is true in Schug et al. (2023); the
global map symbolizes forest-dominated WUI as more dangerous than grassland-dominated
WUI because this condition is true in most other study areas that have woody forests, not
rainforests.

109
Schug et al. (2023) create the global WUI map with the point-based approach, while this
thesis uses zonal-based analysis. The zonal-based results are more easily transferable to
government boundaries because they are based on census blocks. Although the results of the
point-based approach appear smoother, the results of the zonal-based approach in this thesis are
easier to interpret because census blocks divide the study area into manageable units.
Furthermore, additional attributes can be attached to the census block polygons, unlike pixels
from the point-based approach that cannot carry attributes other than classification.
6.3.2 Comparison to State of Hawaii’s WUI Definition
Compared to the WUI designations by the Hawaii DLNR, the WUI results of this thesis
show greater cohesion with WUI products created around the country and the world, as the WUI
calculations are based on Federal Register density and vegetation cover thresholds. By following
the census block practice, the WUI map created from this thesis is also more easily overlaid with
other GIS data and census-based information, while the community boundaries in the State of
Hawaii product do not align with census data.
The low levels of risk in the CARs correspond with the WUIX Forest designations in
East Maui and parts of Upcountry Maui in this thesis, further supporting the notion that these
WUI areas are at less risk than other types of WUI. The notion that CARs are at highest risk in
West, Central, and South Maui aligns with the results of this thesis as well. However, the Hawaii
DLNR defines WUI as a buffer around communities, while the WUI mapped in this thesis is an
improvement by including the communities themselves in WUI classification, giving greater
accuracy to the land area covered by WUI. For transparency and so viewers can make their own
assessments comparing the classification from this thesis and from the State of Hawaii, the

110
Hawaii CARs map is overlaid on the WUI map in the final ArcGIS Dashboard for visual
comparison.
6.4 WUI Significance
Increasing research into the 2023 Maui wildfire events supports the affected communities
and spreads awareness, respecting areas with so much historical and cultural significance.
Creating more WUI knowledge supports wildfire management. Spatializing the WUI on Maui
supports existing Maui County spatialization efforts.
6.4.1 Awareness
The one-year post-fire marker is a significant milestone for Maui. The wildfires are no
longer receiving media attention like they were in the weeks following the disaster, although the
affected communities, especially Lahaina, are nowhere near back to normal. The affected
communities on Maui need to not be forgotten by the public but remain in the public mind out of
respect for the great loss the communities suffered and for collective support in the recovery
process.
6.4.2 Supporting Maui and Maui County Mapping Efforts
This study contributes to the knowledge base of Maui decision makers. A significant
amount of data used in this thesis is sourced from the Maui County GIS Department. This thesis
can support Maui County’s efforts in spatializing information and using maps as a tool to
communicate with residents, as the final WUI and Dashboard products inform Maui residents
about the relationship between the communities they know and love and the surrounding
wildlands. Maui County has put significant efforts into creating map products to convey wildfire-

111
related information to the public, and this research will diversify and extend these mapping
efforts that are already being conducted.
The organization Maui Recovers is devoted to environmental protection and is assessing
many environmental qualities in Lahaina (Maui Recovers 2024a). The air quality is being
monitored because of the ash and dust released into the air, the coastline and watersheds are
being monitored because the debris from the wildfire must be stopped from entering runoff, and
soil quality is being measured carefully because of the negative effects of ash and other toxic
materials that leach into the ground after wildfires (Maui Recovers 2024a). Maui Recovers
provides mapping resources such as smoke maps, maps showing re-entry zones, the safety of
built structures, areas under unsafe water advisories, and more (Maui Recovers 2024b). The
community is very active and engaged in recovery efforts, yet there is room for further
assessments that provide spatialized insight into the relationship between Maui and the WUI.
6.5 Dashboard Audience and Applications
The Dashboard is a creative and innovative way to present data related to the WUI. The
Dashboard format presents a large variety of data in one place while being digestible and not
overwhelming to viewers. Gathered in the Dashboard, the Maui WUI overview and census-block
level WUI detail can aid a variety of stakeholders in Maui. The Dashboard is useful for finding
areas that are the least fire-prone, the Non-WUI Non-Wildlands, but due to the small land area of
this classification, the Dashboard results are also useful for stakeholders interested in
determining the relative safety of census blocks within WUI areas. Dashboard viewers can see
the relative building densities within census blocks, the number of housing units within each, see
their relative distance to wildlands, see their number of historic wildfires, and more. By
incorporating a Maui-specific WUI map with other wildfire-related factors in a shareable, public,

112
and interactive platform, the Dashboard has a broad audience and can be used by local
government and wildfire managers, as well as the private sector and the general public on Maui.
Both broad assessment of the WUI summaries, including the number of buildings and
people that fall in Maui’s WUI, and localized census block WUI assessment are useful for Maui
County’s wildfire management, urban planners, and fire departments. The WUI maps and
Dashboard created from this thesis can support future updates to the existing community and
wildfire protection plans on Maui, as well as building and landscape codes. The Dashboard can
support updating wildfire mitigation strategies, transportation and evacuation plans, utility
networks, firefighting resource allocations, and risk assessments across the island. Furthermore,
the Dashboard can help in choosing priority areas for existing strategies on Maui to enforce
defensible space guidelines, develop firebreaks, reduce fuels, and replace invasive grasses with
native plants. The Dashboard can inform decisions on targeting educational initiatives in highrisk communities, as well as decisions regarding the placement of new development.
Beyond the government sector, the audience of the Dashboard extends to the private
sector and general public on Maui. Homebuyers, homeowners, real estate developers, and
insurance agencies are invested in the safety of homes and buildings in relation to wildfires, and
the well-rounded Dashboard presents valuable information for them when assessing where to
buy, build, or insure on Maui. The general public can explore the Dashboard as an educational
resource and learn about WUI.
6.6 Limitations
While the thesis project makes great strides in researching WUI in Maui, studying WUI
is a complex process, and not all angles of WUI analysis can be addressed in one research
project. Furthermore, multiple methodological approaches exist for WUI mapping, and the

113
census block approach used in this thesis is limited due to the varying sizes of census blocks.
Limits on accuracy are also placed by the datasets used in this project.
6.6.1 Multiple WUI-Calculation Factors
WUI analysis contains many aspects; not all could be explored during this project. This
thesis reveals the impacts that vegetation cover datasets and agricultural datasets have on WUI
classification in Maui. However, other WUI research frequently tests the sensitivity of the
structural density threshold set nationally by the Federal Register (6.18 structures per square
kilometer) and vegetation cover threshold (50% vegetation cover) in WUI definitions for a study
area. These threshold sensitivities are not tested in this thesis. The national Federal Register
threshold definitions are maintained, which are not custom for Maui’s unique landscape. Using
these thresholds assumes that Maui’s building density distribution and vegetation cover generally
aligns with trends across the fire-prone areas of the US, while the island’s characteristics are
unique.
6.6.2 Zonal-Based Census Block Approach
Another limitation of the WUI results is inherent from using the census block analysis
approach; by design, census blocks vary greatly in shape and size. While they are the smallest
census division available for research, census block shapes are determined by natural and
manmade features, such as roads or streams (US Census Bureau 1994). Across Maui, census
blocks vary in size greatly, ranging from less than one square kilometer in highly developed
communities to 250 square kilometers in the rainforests of East Maui. Building density and
vegetation cover are often not evenly distributed throughout large census blocks, so metrics used
for WUI calculation can be skewed. For illustration, comparing the map of building locations to
the map of building density in Chapter 3.3.1 reveals that high-density metrics are often assigned

114
to low density areas, when high density buildings may only be in part of a census block. Most
noticeably, Upcountry Maui and East Maui have large swaths of mostly uninhabited land that
exceed the building density threshold.
Therefore, the WUI results in this thesis are an overestimate of WUI area, especially in
the larger census blocks deemed WUI. Likely, portions of the census blocks would not qualify as
WUI if the census blocks were assessed in smaller pieces, and not skewed by building densities
and vegetation cover relatively far away. While a fine-scale zonal-based approach based on
zones of relatively equal size or population is ideal, these kinds of zones are not available for
Maui or nationally, and census-block based WUI analysis is still a trusted and popular method.
6.6.3 Dataset Limitations
The two datasets with the most accuracy concerns in this project are the ESA landcover
dataset and the census block dataset. The 2021 ESA landcover dataset has 76.7% accuracy,
which is an improvement from the 74.4% accuracy in the 2020 ESA landcover dataset
(Tsendbazar et al., 2022) used in the SILVIS Lab worldwide WUI map. However, potential
misclassification in the 2021 dataset remains a concern. As mentioned in Chapter 3.2.2, the ESA
dataset appears to underestimate the coverage of developed land, especially in communities in
Upcountry Maui. These landcover classifications, although not extreme, skew the vegetation
coverage percentage metrics of WUI calculation higher, and can alter WUI classification. The
alignment of the ESA landcover dataset with census block outlines is also an accuracy limitation.
The outline of the island does not line up perfectly between the two datasets, although the
difference is not extreme. When census blocks around the coast extend into what the ESA dataset
classifies as ocean, vegetation cover percentages are lowered. The vegetation cover metrics
around the coast are not as accurate as the vegetation cover metrics in the inland census blocks.

115
Data update frequency is another concern. The datasets used in the final WUI analysis are
from either 2020 or 2021, so the WUI map created in this thesis is a few years outdated, even
though this is beneficial for seeing the conditions of Maui before the 2023 wildfires. The
agricultural data from 2020 is an update of 2015 baseline data, so updates may be several years
apart. The building footprint dataset is from 2020; there is a prior building footprint dataset from
2018, but future updates are uncertain. The 2021 ESA WorldCover dataset followed a 2020 ESA
WorldCover dataset, but no others have been released since, so the timeline of the next update is
uncertain. Census data updates are every 10 years, so WUI updates should occur no more than 10
years apart. The uncertain update frequencies are a limitation of the datasets, which is important
when considering updating the WUI map in the future.
6.7 Future Research
While this thesis project thoroughly explores WUI in Maui and produces a highly
accurate WUI map for Maui, many opportunities exist to expand WUI research in Maui and
enhance the Dashboard product. The Dashboard needs to be updated with new and improved
data over time. Also, input from Maui residents can improve the Dashboard, making the product
more useful for local viewers. This thesis project intentionally uses national WUI-calculation
thresholds so that the final map product can be compared to other standard WUI maps across the
US. However, WUI research literature reveals that WUI maps provide further insight into
specific regions when they are viewed in isolation with locally adjusted density and vegetation
cover thresholds. These adjustments can enhance Maui WUI map products in the future.
6.7.1 Future WUI Research
While this thesis has WUI sub-classifications that are Maui-specific, exploring Mauispecific building density and vegetation cover metric thresholds can increase the WUI

116
calculations’ appropriateness for Maui. Altering them may cause significant changes to WUI
classification results. The building density on Maui is much greater than the building density
threshold of 6.18 buildings per square kilometer, so increasing the building density threshold
may improve the limitation in this thesis project that large census blocks are skewed to classify
as WUI by their building densities. Furthermore, resolving the census block size issue on a
broader level by researching a zonal-based WUI classification approach that is not limited by the
various sizes of census blocks would not only improve WUI classification results on Maui, but
for the rest of the country as well.
Nevertheless, having used census block data for this thesis project opens a variety of
research opportunities regarding the incorporation of other census data. The plethora of
demographic data associated with each WUI classified census block can be explored, along with
discovered correlations between demographics and WUI classifications. Housing occupancy can
also be studied with WUI and wildfire patterns. Furthermore, the WUI map can be compared
with nationwide (48 conterminous states) census block-based WUI datasets.
Finding an alternative to either zonal-based or pixel-based WUI classification
methodologies that neither overestimates nor underestimates WUI is an endeavor worthy of
further research. The next step of interest for this thesis project is to assign WUI classification
attributes to building footprints themselves, to calculate specifically the human-impacted area of
WUI. The area of Maui covered by WUI buildings would be substantially less than the area of
Maui covered by WUI census blocks. Assessing WUI directly on a human-affected scale can
provide a more precise measurement of the true reach of WUI.
Automating WUI mapping analysis would make WUI mapping more accessible for other
government entities to replicate, rather than replicating and conducting the extensive analysis

117
manually. The methodology steps in this thesis can be made into an ArcGIS ModelBuilder or
written in a script that automates the WUI mapping workflow in ArcGIS Pro. Automating the
process would save considerable time, especially when repeating WUI mapping multiple times
to experiment with different datasets. Mapping vegetation cover multiple times, to gather all
vegetation cover, all vegetation cover without agricultural land, forest cover, and grassland
cover, demonstrates the suitability for streamlining the WUI mapping process with automation.
An automated WUI mapping process can be exported to various entities that could benefit from
creating WUI maps.
Going forward, another endeavor to improve WUI research is to create predictive WUI
maps to forecast future WUI growth. Predictive WUI maps are not currently created in WUI
literature, and adding the time dimension to WUI can increase the uses of WUI. While the WUI
analysis in this thesis project is useful for retroactively assessing the WUI conditions in Maui
before the August 2023 wildfires, predictive WUI maps should be created to support land use
planning and controlling the growth of WUI into the future. Predictive landcover data is
becoming more popular, and while this data must be improved before proving useful for the
Maui study area, a proactive approach to WUI mapping can greatly benefit WUI research,
wildfire management, and land use planning.
6.7.2 Enhancing Existing Dashboard
This thesis has proved that slight data changes alter WUI classifications significantly, and
therefore the overall representation of community danger, development’s relation with wildlands,
and wildfire risk distribution on Maui. Any future WUI experimentations with adjustments to
building density and vegetation cover thresholds would yield additional WUI maps, which could
each be added as additional layers in the Dashboard. Along with the other existing layers in the

118
Dashboard, the added WUI maps would provide more transparency to viewers; additional WUI
maps would highlight and increase viewers’ awareness of the scientific subjectivity of WUI
classification.
To maintain accuracy and relevancy, the WUI map and Dashboard will need to be
updated with new data as it becomes available. Landcover, buildings, population, and
agricultural datasets should be monitored for new releases to create updated WUI maps in the
future. For the Dashboard, the supplemental data layers will need to be replaced over time as
they are updated, so the product is relevant. With strong GIS backing from census data, the ESA,
and the Hawaii Statewide and Maui County GIS Depts., the WUI map and Dashboard products
should be updated every few years.
Furthermore, past datasets and future dataset updates create the opportunity to make the
Dashboard time-enabled. While the WUI layer, Web Map, and Dashboard are currently not timeenabled, creating WUI maps from past datasets and future datasets would enable comparisons
over time, and a time-enabled product. A time-enabled WUI layer would provide insight for
Maui wildfire managers and stakeholders into the growth of WUI in Maui over time. Viewing
past datasets would allow users to see the progression of development, landcover, and other
WUI-related factors over time, deepening their understanding of the changes in WUI.
From a cartographic standpoint, adding features promoting accessibility is important for
future research. This thesis project uses red and green tones for symbology because these colors
are standard in WUI literature to represent fire and wildlands, but these colors are difficult to see
for red-green colorblind viewers. Experimenting with different color schemes that are colorblindfriendly and adding additional data layers that communicate the same information in a
colorblind-friendly manner would improve the accessibility of the WUI map and Dashboard.

119
As Dashboards gain more features in the future, adding a widget with geoprocessing
ability can improve the functionality and impact of the Dashboard. Although dependent on WUI
mapping becoming automated, a widget with a WUI analysis geoprocessing service would allow
users to run WUI analysis with their own datasets and add the map product to the Dashboard.
Allowing Dashboard users to run WUI analysis and create their own map products would
increase the learning potential of the platform, as well as extend its usefulness.
6.7.3 Maui County Feedback
The robust GIS department in Maui County was one of the inspirations for this thesis
project. The final WUI map and Dashboard products will be shared with Maui County GIS
Specialists as an information-sharing act and to receive feedback. Ideally, the Dashboard will
obtain recommendations from Maui County GIS on additional wildfire-related datasets that are
relevant and important to stakeholders. Incorporating feedback on matters of interest that Maui
residents or the Maui fire departments would like to see on the Dashboard will greatly benefit
this project. In the future, perhaps many local governments will create Dashboards informing
their stakeholders about the WUI and other wildfire-related data in their areas, customized for
the needs in their own communities with their own local data.

120
References
Averett, N. 2024. “After the Smoke Clears: Wildland-Urban Interface Fires and Residues in
Nearby Homes.” Environmental Health Perspectives 132, no. 7: 72001-.
https://doi.org/10.1289/EHP14770.
Balmes, J. R., and S. M. Holm. 2023. “Increasing Wildfire Smoke from the Climate Crisis:
Impacts on Asthma and Allergies.” Journal of Allergy and Clinical Immunology 152, no.
5: 1081–83. https://doi.org/10.1016/j.jaci.2023.09.008.
Bar-Massada, A. 2021. “A Comparative Analysis of Two Major Approaches for Mapping the
Wildland-Urban Interface: A Case Study in California.” Land 10, no. 7: 679.
https://doi.org/10.3390/land10070679.
Blum, J. 2024. “Maui Rain by Month and Area.” Accessed October 28, 2024.
https://www.mauihawaii.org/rain-maui-rainfall/.
Bond-Smith, S., D. Bond-Smith, C. Bonham, L. Bremer, K. Burnett, M. Coffman, P. Fuleky, B.
Gangnes, R. Inafuku, R. Juarez, S. La Croix, C. Moore, D. Moore, N. Tarui, J. Tyndall,
and C. Wada. 2023. “After the Maui Wildfires: The Road Ahead.” Accessed November
19, 2024. https://uhero.hawaii.edu/after-the-maui-wildfires-the-road-ahead/.
Brewington, L. 2020. “Transitions and Drivers of Land Use/Land Cover Change in Hawai’i: A
Case Study of Maui.” In: Land Cover and Land Use Change on Islands. Social and
Ecological Interactions in the Galapagos Islands. Edited by S. J. Walsh, D. RiverosIregui, J. Arce-Nazario, and P. H. Page. Springer, Cham. https://doi.org/10.1007/978-3-
030-43973-6_4.
Byren, R. W., and TESA Tech Team. 2023. “The Wildfire Evacuation Dilemma--How Not To
Become Lahaina.” Accessed November 19, 2024. https://eastshorealliance.com/wpcontent/uploads/Wildfire-Evacuation-Paper.pdf.
Calkin, D. E., K. Barrett, J. D. Cohen, M. A. Finney, S. J. Pyne, S. L. Quarles. 2023. “WildlandUrban Fire Disasters Aren’t Actually a Wildfire Problem.” Proc. Natl. Acad. Sci. U.S.A.
120 (51) e2315797120, https://doi.org/10.1073/pnas.2315797120.
Carlson, A. R., D. P. Helmers, T. J. Hawbaker, M. H. Mockrin, and V. C. Radeloff. 2022. “The
Wildland–Urban Interface in the United States Based on 125 Million Building
Locations.” Ecological Applications 32, no. 5: e2597-n/a.
https://doi.org/10.1002/eap.2597.
County of Maui. 2021. “Report on Wildfire Prevention and Cost Recovery on Maui.” Accessed
October 24, 2024. https://www.mauicounty.gov/DocumentCenter/View/129493/Reporton-Wildfire-Prevention--Cost-Recovery-on-Maui---Part-1-Report--Exhibits-A-B-33-
MB.

121
———. 2010. “County of Maui 2030 General Plan: Countywide Policy Plan.” Accessed
December 3, 2024. https://www.mauicounty.gov/DocumentCenter/View/11132/Final--
Countywide-Policy-Plan---Complete?bidId=.
———. 2008. “Maui Island History.” Accessed October 1, 2024.
https://www.mauicounty.gov/DocumentCenter/View/3231/History.
County of Maui GIS. 2024. “Maui Wildfire Extents August 2023.” Accessed October 22, 2024.
https://services3.arcgis.com/fsrDo0QMPlK9CkZD/ArcGIS/rest/services/Maui_Wildfire_
Extents__August_2023_/FeatureServer/layers.
———. 2020. “Building Footprints 2020 (Feature Server)”. Accessed September 3, 2024.
https://services3.arcgis.com/fsrDo0QMPlK9CkZD/ArcGIS/rest/services/MauiCountyBuil
dingFootprints2020/FeatureServer.
———. 2017. “Fire Risk Areas.” Accessed October 22, 2024.
https://geodata.hawaii.gov/arcgis/rest/services/Hazards/MapServer/7.
Data USA. 2022. “Maui County, HI.” Accessed October 1, 2024.
https://datausa.io/profile/geo/maui-countyhi#:~:text=In%202022%2C%20Maui%20County%2C%20HI,median%20household%20i
ncome%20of%20%2495%2C379.
Dejournett, T. 2023. “Why the Big Bog in East Maui Is Now Hawai’i’s Rainiest Spot.” Accessed
November 19, 2024. https://www.hawaiibusiness.com/big-bog-east-maui-hawaii-rainiestspot/.
Dolling, K., P. Chu, and F. Fujioka. 2005. “A Climatological Study of the Keetch/Byram
Drought Index and Fire Activity in the Hawaiian Islands.” Agricultural and Forest
Meteorology 133, no. 1-4: 17-27. https://doi.org/10.1016/j.agrformet.2005.07.016.
Earth.org. 2024. “Unprecedented Scale: Exploring the Largest Wildfires in US History.” 2024.
Crisis – Biosystem Viability. Accessed November 19, 2024. https://earth.org/worstwildfires-in-us-history/.
Encyclopedia Britannica. 2024a. “Lahaina.” Accessed November 19, 2024.
https://www.britannica.com/place/Lahaina.
———. 2024b. “Maui.” Accessed October 1, 2024. https://www.britannica.com/place/Mauiisland-Hawaii.
Esri US Federal Data. 2024. “U.S. Census Blocks.” Accessed October 3, 2024.
https://www.arcgis.com/home/item.html?id=d795eaa6ee7a40bdb2efeb2d001bf823.
Federal Emergency Management Agency. 2023. “Hawaii Wildfires.” Accessed October 25,
2024. https://www.fema.gov/disaster/4724.

122
Federal Register. 2001. “Urban Wildland Interface Communities Within the Vicinity of Federal
Lands That Are at High Risk from Wildfire.” Accessed November 12, 2024.
https://www.federalregister.gov/documents/2001/01/04/01-52/urban-wildland-interfacecommunities-within-the-vicinity-of-federal-lands-that-are-at-high-risk-from.
Fire Safety Research Institute. 2024. “Lahaina Fire Comprehensive Timeline Report Released by
the Attorney General of Hawai’i.” Accessed September 30, 2024.
https://fsri.org/research-update/lahaina-fire-comprehensive-timeline-report-releasedattorney-general-hawaii.
Frazier, A. G., J. L. Deenik, N. D. Fujii, G. R. Funderburk, T. W. Giambelluca, C. P. Giardina,
D. A. Helweg, V. W. Keener, A. Mair, J. J. Marra, S. McDaniel, L. N. Ohye, D. S. Oki,
E. W. Parsons, A. M. Strauch, and C. Trauernicht. 2019. “Managing Effects of Drought
in Hawai’i and US-Affiliated Pacific Islands.” In: Effects of Drought on Forests and
Rangelands in the United States: Translating Science into Management Responses.
Edited by J. M. Vose, D. L. Peterson, C. H. Luce, T. Patel-Weynand. Gen. Tech. Rep.
WO-98: 95-121. https://doi.org/10.2737/WO-GTR-98.
Fu, X., A. Lidar, M. Kantar, B. Raghavan, and J. Whalen. 2023. “Edible Fire Buffers: Mitigation
of Wildfire with Multifunctional Landscapes.” PNAS Nexus 2, no. 10: pgad315–pgad315.
https://doi.org/10.1093/pnasnexus/pgad315.
Fukumoto, G.L. and C. N. Lee. 2003. “Kikuyugrass for Forage.” Accessed November 15, 2024.
https://www.ctahr.hawaii.edu/oc/freepubs/pdf/LM-5.pdf.
Google Earth. 2023a. “Historic Imagery 2023-3-24.” Accessed December 5, 2024.
https://earth.google.com/web/@20.87283978,-
156.67780025,1.49654665a,3238.04710026d,35y,0h,0t,0r/data=ChYqEAgBEgoyMDIzL
TAzLTI0GABCAggBMikKJwolCiExeXEyLXh1eHpaTlppeE9jaE9zcml2UmZZQ2czcFl
rNy0gAToDCgEwQgIIAEoHCJ277m8QAQ.
Google Earth. 2023b. “Historic Imagery 2023-8-11.” Accessed December 5, 2024.
https://earth.google.com/web/@20.87283978,-
156.67780025,1.49654665a,3238.04710026d,35y,0h,0t,0r/data=ChYqEAgBEgoyMDIzL
TA4LTExGABCAggBMikKJwolCiExeXEyLXh1eHpaTlppeE9jaE9zcml2UmZZQ2czcF
lrNy0gAToDCgEwQgIIAEoHCJ277m8QAQ.
Harrison, T. E. 2020. “Using Targeted Cattle Grazing to Reduce Wildfire Risk in Hawaiʻi:
Stakeholder Values and Public Support.” Master's thesis, University of Hawai’i at
Manoa.
Hawaii Emergency Management Agency. 2018. “State of Hawai’i 2018 Hazard Mitigation
Plan.” https://dod.hawaii.gov/hiema/files/2018/11/State-of-Hawaii-2018-MitigationPlan.pdf.
Hawaii Dept. of Land and Natural Resources. 2023. “Serial Photos and Videos of all Four Fires
on Maui.” Accessed November 19, 2024. https://dlnr.hawaii.gov/blog/2023/08/11/nr23-
126/.

123
———. 2013. “Hawaii Coordinate Systems.” Accessed November 19, 2024.
https://dlnr.hawaii.gov/shpd/files/2018/01/HawaiiCooSys.pdf.
———. 2010. “Issue 3: Wildfires.” Accessed September 13, 2024.
https://dlnr.hawaii.gov/forestry/files/2013/09/SWARS-Issue-3.pdf.
Hawaii Invasive Species Council. 2024a. “Invasive Grasses in Hawaii and their Impacts.”
Accessed November 15, 2024. https://dlnr.hawaii.gov/hisc/info/species/invasive-grassesin-hawaii-and-their-impacts/.
———. 2024b. “Invasive Species.” Accessed November 15, 2024.
https://dlnr.hawaii.gov/hisc/info/.
Hawaii Statewide GIS Program. 2024. “Fire Risk Areas.” Accessed October 22, 2024.
https://geodata.hawaii.gov/arcgis/rest/services/Hazards/MapServer/7.
———. 2022a. “Agricultural Land Use – 2020 Update.” Accessed October 28, 2024.
https://geoportal.hawaii.gov/datasets/342ee6c7547f45ddbfc07caf4ca2887d/explore?locati
on=20.734874%2C-156.240509%2C10.73.
———. 2022b. “Annual Rainfall (in).” Accessed November 27, 2024.
https://geoportal.hawaii.gov/datasets/HiStateGIS::annual-rainfallin/explore?location=20.738832%2C-156.319666%2C10.66.
———. 2021. “Fire Stations (Statewide).” Accessed October 22, 2024.
https://geoportal.hawaii.gov/datasets/fire-stations-statewide/explore.
Hawai’i Tourism Authority. 2024a. “Maui.” Accessed October 1, 2024.
https://www.gohawaii.com/islands/maui.
———2024b. “Maui Fact Sheet.” Accessed October 1, 2024.
https://www.meethawaii.com/media-room/hawaiian-islands-fact-sheets/maui/.
Hawai’i Wildfire Management Organization. 2013. “West Maui – Community Wildfire Hazard
Assessments.” Accessed September 13, 2024. https://www.hwmo.org/resourcelibrary/west-maui-community-wildfire-hazard-assessments.
Hawai’i Wildfire Management Organization, and C. Trauernicht. 2023. “Pacific Island Wildfire
Data.” Accessed October 28, 2024. https://pacificfireexchange.org/resource/fire-data/.
Johnston, L. M., and M. D. Flannigan. 2018. “Mapping Canadian Wildland Fire Interface
Areas.” International Journal of Wildland Fire 27, 1-14.
https://doi.org/10.1071/WF16221.
Jones, M. W., J. T. Abatzoglou, S. Veraverbeke, N. Andela, G. Lasslop, M. Forkel, A. J. P.
Smith, C. Burton, R. A. Betts, G. R. van der Werf, S. Sitch, J. G. Ganadell, C. Santín, C.
Kolden, S. H. Doerr, and C. Le Quéré. 2022. “Global and Regional Trends and Drivers of

124
Fire Under Climate Change.” Reviews of Geophysics (1985) 60, no. 3.
https://doi.org/10.1029/2020RG000726.
Juliano, T. W., F. Szasdi-Bardales, N. P. Lareau, K. Shamsaei, B. Kosović, N. ElhamiKhorasani, E. P. James, and H. Ebrahimian. 2024. “Brief Communication: The Lahaina
Fire Disaster – How Models Can Be Used to Understand and Predict Wildfires.” Nat.
Hazards Earth Syst. Sci., 24, 47–52. https://doi.org/10.5194/nhess-24-47-2024.
Kerber, S., and D. Alkonis. 2024. “Lahaina Fire Comprehensive Timeline Report.” Fire Safety
Research Institute. https://doi.org/10.54206/102376/VQKQ5427.
Ketchpaw, A. R., D. Li, S. Nawaz Khan, Y. Jiang, Y. Li, and L. Zhang. 2022. “Using Structure
Location Data to Map the Wildland–Urban Interface in Montana, USA.” Fire 5, no. 5:
129. https://doi.org/10.3390/fire5050129.
Lahaina Town. 2024a. “Lahaina Maui.” Accessed October 25, 2024. https://lahainatown.com/.
———. 2024b. “Lahaina Timeline.” Accessed October 25, 2024.
https://lahainatown.com/lahaina-history.php.
LaPier, R. R. 2023. “Native Hawaiian Sacred Sites have been Damaged in the Lahaina Wildfires
– but, as an Indigenous scholar Writes, their Stories will Live On.” Conversation, August
11, 2023, https://theconversation.com/native-hawaiian-sacred-sites-have-been-damagedin-the-lahaina-wildfires-but-as-an-indigenous-scholar-writes-their-stories-will-live-on211401.
Li, S., V. Dao, M. Kumar, P. Nguyen, and T. Banerjee. 2022. “Mapping the Wildland-Urban
Interface in California Using Remote Sensing Data.” Scientific Reports 12, no. 1: 5789–
12. https://doi.org/10.1038/s41598-022-09707-7.
Link, E. 2024. “The NIST Wildland-Urban Interface Fire Case Study Approach and Outlook.”
Technical Note (NIST TN), National Institute of Standards and Technology,
Gaithersburg, MD. https://doi.org/10.6028/NIST.TN.2296.
Macrotrends. 2024. “US Population Density 1950-2024.” Population. Accessed October 25,
2024. https://www.macrotrends.net/global-metrics/countries/USA/Unitedstates/populationdensity#:~:text=The%20population%20density%20of%20U.S.%20in%202022%20was%
2034.41%20people,a%200.38%25%20increase%20from%202021.
Mahmoud, H. 2024. “Reimagining a Pathway to Reduce Built-Environment Loss during
Wildfires.” Cell Reports Sustainability 1, no. 6: 100121-.
https://doi.org/10.1016/j.crsus.2024.100121.
Mass, C., and D. Ovens. 2024. “The Meteorology of the August 2023 Maui Wildfire.” Weather
and Forecasting, 39, 1097–1115, https://doi.org/10.1175/WAF-D-23-0210.1.

125
Maui County Department of Planning. 2022. “West Maui Community Plan.” Accessed
September 24, 2024. https://westmaui.wearemaui.org/wp-content/uploads/2022/02/2022-
West-Maui-Community-Plan-Web.pdf.
Maui Emergency Management Agency. 2020. “Hazard Mitigation Plan Update.” Accessed
September 10, 2024. https://www.mauicounty.gov/DocumentCenter/View/125977/2020-
Maui-County-Hazard-Mitigation-Plan-Final.
———. 2019. “After Action Report & Improvement Plan for 2018-08-18 Tropical Cyclone
Lane.” Accessed November 16, 2024. https://d1l18ops95qbzp.cloudfront.net/wpcontent/2023/10/25100557/maui-2018-lahaina-firehurricane-lane-after-action-report.pdf.
Maui Police Dept. 2024. “Maui Wildfires of August 8th, 2023 Maui Police Department
Preliminary After-Action Report.” Accessed September 30, 2024.
https://www.mauipolice.com/uploads/1/3/1/2/131209824/pre_aar_master_copy_final_dra
ft_1.23.24.pdf.
Maui Recovers. 2024a. “Environmental Protection.” Accessed November 19, 2024.
https://www.mauirecovers.org/environmentalprotection.
———. 2024b. “For Maui Wildfire Survivors.” Accessed September 23, 2024.
https://www.mauirecovers.org/.
Mengote, F. 2024. “Paradise on Fire: 2023 Maui-Lahaina Wildfire Case Study.” Master’s thesis,
University of Arizona.
Mockrin, M. H., B. McGuinness, D. P. Helmers, and V. C. Radeloff. 2023. “Understanding the
Wildland-Urban Interface (1990-2020).” Madison, WI: U.S. Department of Agriculture,
Forest Service, Northern Research
Station. https://storymaps.arcgis.com/stories/6b2050a0ded0498c863ce30d73460c9e.
National Institute of Standards and Technology. 2023. “WUI Definitions.” Accessed August 27,
2024. https://www.nist.gov/el/fire-research-division-73300/wildland-urban-interface-fire73305/hazard-mitigation-methodology-9.
Pagadala, T., M. A. Alam, T. M. R. Maxwell, and T. J. Curran. 2024. “Measuring Flammability
of Crops, Pastures, Fruit Trees, and Weeds: A Novel Tool to Fight Wildfires in
Agricultural Landscapes.” The Science of the Total Environment 906: 167489–167489.
https://doi.org/10.1016/j.scitotenv.2023.167489.
Parsons, E., and C. Martin. 2023. “The Tragedy in Lahaina: How Invasive Grasses and Shrubs
are Fueling the Wildfire Crisis in Hawai’i.” Accessed October 24, 2024.
https://naisma.org/2023/10/10/the-tragedy-in-lahaina-how-invasive-grasses-and-shrubsare-fueling-the-wildfire-crisis-in-hawai%CA%BBi/.
Partyka, G., and B. Erdman. 2023. “Meteorologic Analysis of the August 2023 Maui Wildfires.”
Accessed November 19, 2024. https://gmao.gsfc.nasa.gov/research/
science_snapshots/2023/meterologic-analysis-maui-wildfires.php.

126
Pickett, E., I. Grossman, and Hawai’i Wildfire Management Organization. 2014. “Western Maui
Community Wildfire Protection Plan.” Accessed September 13, 2024.
https://dlnr.hawaii.gov/forestry/files/2024/01/Western-Maui-CWPP.pdf.
Purdy, P. 2018. “Maui Fire Department Community Risk Reduction Program for Wildland
Urban Interface.” Accessed November 19, 2024.
https://www.mauicounty.gov/DocumentCenter/View/142946/-Maui-County-WildlandUrban-Interface-WUI-Program.
Radeloff, V. C., M. H. Mockrin, D. Helmers, A. Carlson, T. J. Hawbaker, S. Martinuzzi, F.
Schug, P. M. Alexandre, H. A. Kramer, and A. M. Pidgeon. 2023a. “Rising Wildfire Risk
to Houses in the United States, Especially in Grasslands and Shrublands.” Science
(American Association for the Advancement of Science) 382, no. 6671: 702–
7. https://doi.org/10.1126/science.ade9223.
Radeloff, V. C., D. P. Helmers, M. H. Mockrin, A. R. Carlson, T. J. Hawbaker, S. Martinuzzi.
2023b. “The 1990-2020 Wildland-Urban Interface of the Conterminous United States -
Geospatial Data.” 4th Edition. Fort Collins, CO: Forest Service Research Data Archive.
https://doi.org/10.2737/RDS-2015-0012-4.
Radeloff, V. C., D. P. Helmers, H. A. Kramer, M. H. Mockrin, P. M. Alexandre, A. BarMassada, V. Butsic, T. J. Hawbaker, S. Martinuzzi, A. D. Syphard, and S. I. Stewart.
2018. “Rapid Growth of the US Wildland-Urban Interface Raises Wildfire Risk.
Proceedings of the National Academy of Sciences - PNAS, 115(13), 3314–3319.
https://doi.org/10.1073/pnas.1718850115.
Radeloff, V. C., R. B. Hammer, S. I. Stewart, J. S. Fried, S. S. Holcomb, and J. F. McKeefry.
2005. “The Wildland-Urban Interface in the United States.” Ecological Applications
Volume 15, Issue 3 p. 799-805, https://doi.org/10.1890/04-1413.
Ramayanti, S., B. Kim, S. Park, and C. Lee. 2024. “Wildfire Susceptibility Mapping by
Incorporating Damage Proxy Maps, Differenced Normalized Burn Ratio, and Deep
Learning Algorithms Based on Sentinel-1/2 Data: A Case Study on Maui Island,
Hawaii.” GIScience & Remote Sensing 61 no. 1.
https://doi.org/10.1080/15481603.2024.2353982.
Ritchey, T. 2022. “An Assessment of Prescribed Burns in Hawai’i to Identify Training
Opportunities and Limitations to Mitigating Long-Term Damage of Wildfires to
Communities and Ecosystem.” Bachelor’s thesis, University of Hawaii at Manoa.
Romero, S., and S. F. Kovaleski. 2023. “How Invasive Plants Caused the Maui Fires to Rage.”
New York Times, August 13, 2023. https://www.nytimes.com/2023/08/13/us/hawaiiwildfire-factors.html.
Roode, A. F. de, and I. Martinac. 2020. “Resilience Hubs: A Maui Case Study to Inform
Strategies for Upscaling to Resilience Hub Networks across Coastal, Remote, and Island
Communities.” IOP Conference Series: Earth and Environmental Science 588, no. 5:
52050-. https://doi.org/10.1088/1755-1315/588/5/052050.

127
Roy, D. P., H. De Lemos, H. Huang, L. Giglio, R. Houborg, and T. Miura. 2024. “MultiResolution Monitoring of the 2023 Maui Wildfires, Implications and Needs for SatelliteBased Wildfire Disaster Monitoring.” Science of Remote Sensing 10: 100142-.f
https://doi.org/10.1016/j.srs.2024.100142.
Schug, F., A. Bar-Massada, A. R. Carlson, H. Cox, T. J. Hawbaker, D. Helmers, P. Hostert, D.
Kaim, N. K. Kasraee, S. Martinuzzi, M. H. Mockrin, K. A. Pfoch, and V. C. Radeloff.
2023. “The Global Wildland–Urban Interface.” Nature (London) 621, no. 7977: 94–99.
https://doi.org/10.1038/s41586-023-06320-0.
Slaton, D. A. 2023. “The Impact of Definition Criteria on Mapped Wildland-Urban Interface: A
Case Study for Ten Counties along the Oregon-California Border.” Master’s thesis,
University of Southern California.
Sommers, W. T. 2024. “The Emergence of the Wildland-Urban Interface Concept.” Accessed
November 29, 2024. https://foresthistory.org/wpcontent/uploads/2016/12/Sommers_emergence-of-wildland.pdf.
Synolakis, C. E., and G. M. Karagiannis. 2024. “Wildfire Risk Management in the Era of
Climate Change.” PNAS Nexus 3, no. 5: pgae151–pgae151.
https://doi.org/10.1093/pnasnexus/pgae151.
Trauernicht, C. 2019. “Vegetation—Rainfall Interactions Reveal How Climate Variability and
Climate Change Alter Spatial Patterns of Wildland Fire Probability on Big Island,
Hawaii.” The Science of the Total Environment 650, no. Pt 1: 459–69.
https://doi.org/10.1016/j.scitotenv.2018.08.347.
Trauernicht, C., and M. Lucas. 2016. “Wildfire Ignition Density Maps for Hawai’i.” Accessed
November 13, 2024. https://pacificfireexchange.org/wpcontent/uploads/2022/07/WildfireIgnitionDensitiesForHawaii.pdf.
Tsendbazar N., P. Xu, M. Herold, M. Lesiv, and M. Duerauer. 2022. “Product Validation
Report.” Accessed November 7, 2024.
https://worldcover2021.esa.int/data/docs/WorldCover_PVR_V2.0.pdf.
Tunison, J. T., C. M. D’Antonio, and R. K. Loh. 2001. “Fire and Invasive Plants in Hawai’i
Volcanoes National Park.” In: Proceedings of the Invasive Species Workshop: the Role of
Fire in the Control and Spread of Invasive Species. Edited by: K.E.M. Galley and T.P.
Wilson. Fire Conference 2000: the First National Congress on Fire Ecology, Prevention,
and Management. Miscellaneous Publication No. 11, Tall Timbers Research Station,
Tallahassee, FL.
University of Hawaii at Manoa. n.d. “Weather Patterns.” Accessed November 19, 2024.
https://manoa.hawaii.edu/sealearning/grade-3/earth-and-space-science/weather-patterns.
US Census Bureau. 2022. “American Community Survey 1-year estimates. Retrieved
from Census Reporter Profile page for Maui County, HI.” Accessed October 1, 2024.
http://censusreporter.org/profiles/05000US15009-maui-county-hi/.

128
———. 1994. “Census Blocks and Blocks Groups.” In: Geographic Areas Reference Manual.
Accessed November 7, 2024. Washington DC: US Dept. of Commerce, Economics and
Statistics Administration, Bureau of the Census.
US Drought Monitor. 2024. “Map Archive: Hawaii, August 8, 2023.” Accessed October 24,
2024. https://droughtmonitor.unl.edu/Maps/MapArchive.aspx.
US Fire Administration. 2024. “Preliminary After-Action Report: 2023 Maui Wildfire.”
Accessed November 19, 2024. https://www.usfa.fema.gov/blog/preliminary-after-actionreport-2023-maui-wildfire/.
Voda, J. A. 2023. “Maui Wildfires: A Case Study in Local Government Crisis Communication
Response.” Master’s thesis, City University of New York.
Yamane, Marisa. 2024. “One Year After the Kula Wildfire: Stories of Survival, Rebuilding, and
the Aloha Spirit.” IslandNews, August 7, 2024. https://www.kitv.com/news/mauiperseveres/one-year-after-the-kula-wildfire-stories-of-survival-rebuilding-and-the-alohaspirit/article_f9d1d2be-4fda-11ef-86a5-cb6e2efbdd0e.html.
Zhou, X. 2024. “Fire Spread Simulation and Probabilistic Regional Fire Loss Assessment at
Wildland-Urban Interface.” Ph.D. dissertation, University of California at Los Angeles.

Abstract (if available)

Abstract In August 2023, Maui experienced a series of massive wildfires, including one that destroyed most of the historic town of Lahaina. Climate change has increased the frequency and intensity of wildfires, and the Maui wildfires exhibit the dangers of living next to wildlands. Identifying an area’s wildland-urban interface (WUI) is a crucial part of wildfire management. To provide insight into the relationship between development and the surrounding wildlands in Maui, this thesis studies the spatial distribution of the WUI in Maui before the 2023 wildfires. This thesis creates the first census block-based WUI map of Maui. Standard WUI map creation entails determining census blocks’ structural density, vegetation cover percentage, and distance from wildland areas. In census block-based WUI maps, each census block is assigned a WUI classification. This thesis experiments with multiple landcover datasets and a local agricultural dataset to assess their effects on WUI classification and determine the most appropriate datasets for mapping WUI on Maui. For the final WUI analysis, this thesis utilizes 2020 US Census Bureau census block data, European Space Agency 2021 WorldCover landcover data, and Hawaii Statewide GIS Program agricultural land use data to map WUI. The final WUI map product shows the arrangement of Maui in relation to the WUI before the wildfires. The findings of the WUI analysis show that 27.46% of land, 96.82% of buildings, and 99.03% of the population on Maui are in WUI. The final WUI map is developed into an ArcGIS Dashboard that allows users to explore the WUI in relation to other wildfire-related data, while providing transparency into WUI calculation. The findings of this thesis are useful for wildfire management, urban planning, the private sector, and the general public by providing insight into the spatial arrangement of the WUI in Maui, where to target priority areas for wildfire prevention interventions, and the relative safety of buildings and homes in the WUI.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

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

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

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

Development of a historical urban land use web application for the city of Hong Kong

PDF

San Diego Wildfire Hazards Information Center mashup

PDF

Human suffering during wartime: a StoryMap of violations of international law during the Russo-Ukrainian War

PDF

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

PDF

Advancing Redwood City's bicycle infrastructure through a geodesign workflow

PDF

Enriching the Demographic Survey sampling for the Los Angeles County Annual Homeless Count with spatial statistics

PDF

From ruins to pixels: using remote sensing and GIS to analyze, document, and visualize archaeological sites and ancient roadways in Chaco Canyon, New Mexico

PDF

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

PDF

Using census data, urban-land cover classification, and dasymetric mapping to measure urban growth of the lower Rio Grande Valley, Texas

PDF

The role of precision in spatial narratives: using a modified discourse quality index to measure the quality of deliberative spatial data

PDF

Silicon Valley construction project web mapping application

PDF

Alaska Hike Search: designing a mapping application for Alaskan trails and user contributed hazards

PDF

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

PDF

Evaluating machine learning tools for humanitarian road network mapping

PDF

Commute GeoCalculator: a GIS server extension for comparing automobile and transit travel costs

Asset Metadata

Creator Birdwell, Amber Jean (author)

Core Title Maui’s wildland-urban interface: enhancements for the unique vegetative and agricultural landscape

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Degree Conferral Date 2025-05

Publication Date 01/16/2025

Defense Date 12/10/2024

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

Tag Maui,OAI-PMH Harvest,Wildfires,wildland-urban interface

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elisabeth (committee chair), Huang, Guoping (committee member), Qi, Yi (committee member)

Creator Email birdwell@usc.edu,birdwellamber@gmail.com

Unique identifier UC11399FD4V

Identifier etd-BirdwellAm-13766.pdf (filename)

Legacy Identifier etd-BirdwellAm-13766

Document Type Thesis

Format theses (aat)

Rights Birdwell, Amber Jean

Internet Media Type application/pdf

Type texts

Source 20250115-usctheses-batch-1236 (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.

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