EQUITY IN WILDFIRE RISK MANAGEMENT: DOES SOCIOECONOMIC STATUS
PREDICT INVOLVEMENT IN FEDERAL PROGRAMS TO MITIGATE WILDFIRE
RISK?
by
RYAN S. OJERIO
A THESIS
Presented to the Department ofPlanning,
Public Policy and Management
and the Graduate School of the University ofOregon
in partial fulfillment ofthe requirements
for the degree of
Master of Community and Regional Planning
June 2008
"Equity in Wildfire Risk Management: Does Socioeconomic Status Predict Involvement
in Federal Programs to Mitigate Wildfire Risk?" a thesis prepared by Ryan S. Ojerio in
partial fulfillment of the requirements for the Master of Community and Regional
Planning degree in the Department of Planning, Public Policy and Management. This
thesis has been approved and accepted by:
.,....- ,.
Dr. Neil Bania, Chair of the Examining Committee
Date
ii
Committee in Charge:
Accepted by:
Dr. Neil Bania, Chair
Dr. Cassandra Moseley
Kathy Lynn, M.C.R.P.
..._--_ .._.~+-,------
Dean of the Graduate School
© 2008 Ryan S. Ojerio
iii
IV
June 2008
Ryan S. Ojerio
An Abstract of the Thesis of
for the degree of
Master of Community and Regional Planning
in the Department of Planning, Public Policy and Management
to be taken
Title: EQUITY IN WILDFIRE RISK MANAGEMENT: DOES SOCIOECONOMIC
STATUS PREDICT INVOLVEMENT IN FEDERAL PROGRAMS TO
MITIGATE WILDFIRE RISK?
Approved: _
Dr. Neil Bania
Currently, biophysical risk factors figure prominently in federal resource
allocation to communities threatened by wildfire. Yet, disaster research demonstrates that
socioeconomic characteristics including age, gender, poverty, race, culture, education and
political influence impact disaster risk and resilience.
Consequently, this thesis evaluates whether federal wildfire program resources are
reaching socially vulnerable populations. My hypothesis is that socially vulnerable
populations are less likely to be involved in such mitigation efforts because of the
emphasis on biophysical risk factors.
vTo evaluate this, biophysical and social vulnerability indicators were linked at the
Census Block Group level within the state of Arizona. Regression analysis was applied to
evaluate predictors of participation and inclusion in federally funded wildfire mitigation
efforts.
Findings indicate that resources are focused on areas of high biophysical risk,
without regard to social vulnerability. In fact, disadvantaged populations are less likely to
be involved in wildfire mitigation efforts than their more affluent counterparts.
CURRICULUM VITAE
NAME OF AUTHOR: Ryan S. Ojerio
PLACE OF BIRTH: Fort Collins, CO
DATE OF BIRTH: July 29,1972
GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:
University of Oregon
Michigan State University
DEGREES AWARDED:
Master of Community and Regional Planning, 2008, University of Oregon
Bachelor of Science, Biology, 1994, Michigan State University
AREAS OF SPECIAL INTEREST:
Collaboration in natural resource management and planning
Wildfire management and social vulnerability
PROFESSIONAL EXPERIENCE:
Graduate Research Fellow, Resource Innovations, Institute for a Sustainable
Environment, University of Oregon 2006-2008
Program Manager, Northwest Youth Corps, 2004-2006
vi
vii
GRANTS, AWARDS AND HONORS:
American Institute ofCertified Planners Outstanding
Student, 2008
Catalyst Award, 2007
Joseph M. Edney Memorial Scholarship for Environmental Planning,
2006
PlJBLICAnONS:
Glenn, T., Ojerio, R. S., Stephan, W., Braun, M. J., 1997. Microsatellite DNA
Loci for Genetic Studies ofCranes. Proceedings of the North American
Crane Workshop 7, 36-45.
Jacobson, D.J., LeFebvre, S. M., Ojerio, R. S., Berwald, N., Heikkinen, E., 1998.
Persistent, Systemic, Asymptomatic Infections ofAlbugo candida, an
Oomycete Parasite, Detected in Three Wild Crucifer Species. Canadian
Journal ofBotany 76, 739-750.
Jacobson, D.J., Ojerio, R. S., 1996. Host Specificity ofAlbugo candida on Wild
Crucifers from Two Geographic Regions. Phytopathology 86, 64.
Strong, M., Doerry, E. with illustrations by Ojerio, R. S., 2001. Glaciers! The Art
ofTravel and the Science ofRescue, A Falcon Guide. Globe Pequot Press,
Guilford, CT.
viii
ACKNOWLEDGMENTS
First and foremost I would like to thank Kathy Lynn for the intellectual
inspiration that made this thesis possible. Her support, encouragement, and advice have
been defining contributions to my graduate experience. Special thanks are also due to my
committee chair, Dr. Neil Bania, whose patience and guidance in statistical methods were
essential to this project. As well, I wish to acknowledge Dr. Cassandra Moseley for her
passion and enthusiasm for public lands management issues and social equity. Others
who contributed to the intellectual development of this project include Drs. Alexander
Evans, Ellen Donoghue and Victoria Sturtevant. The founders of the Catalyst Award also
deserve recognition for funding my travels to the Human Dimensions of Wildfire
Conference which also contributed significantly to the development of this thesis. I am
indebted to Glen Buettner and Cliff Pearlberg at the Arizona State Lands Dept. Forestry
Division for answering many questions and providing essential data for this project.
Finally, I am so thankful for the support and encouragement of my girlfriend, Britta
Torgrimson.
IX
TABLE OF CONTENTS
Chapter Page
I. INTRODUCTION 1
II. LITERATURE REVIEW 4
2.1 Wildfire Management Policy 4
2.2 Natural Disasters and Social Vulnerability 9
2.2.1 Dimensions ofSocial Vulnerability 9
2.2.2 Social Vulnerability in the Wildfire Context 11
2.2.3 Indicators ofSocial Vulnerability in the Wildfire Context 14
2.2.4 Economic Vulnerability in Rural Communities 15
2.2.5 Helping Agencies 16
2.3 Synthesis 18
III. MEASURES AND METHODS 19
3.1 Study Area 19
3.2 Unit ofAnalysis 20
3.3 Biophysical Wildfire Risk Factors 21
3.4 Indicators ofSocioeconomic Status 23
3.5 Wildfire Mitigation Activities 24
3.5.1 The Firewise Communities USA Program 25
3.5.2 Community Wildfire Protection Plans (CWPP) 26
3.5.3 State Fire Assistance (SFA) Grants 27
3.6 Limitations 28
3.7 Analysis 30
Chapter
x
Page
IV. FINDINGS 32
4.1 Correlations Between Socioeconomic Indicators 32
4.2 Frequency of Wildfire Mitigation Activities 35
4.3 Biophysical Factors 35
4.3.1 Land Hazard Rating 35
4.3.2 Structural Density 36
4.4 Socioeconomic Factors 37
4.4.1 Community Wildfire Protection Plans (CWPP) 38
4.4.2 State Fire Assistance (SFA) Grants 45
4.4.3 The Firewise Communities USA Program 51
4.5 Summary 57
V. DISCUSSION 58
5.1 Prioritizing Socially Vulnerable Populations 59
5.2 Next Steps 63
5.2.1 Involving Socially Vulnerable Communities in Planning and Implementation 64
5.2.2 Community Capacity 65
BIBLIOGRAPHY 68
Xl
LIST OF FIGURES
Figure Page
I. Relationship Between Land Hazard Rating and Probability of Involvement in
Three Types ofWildfire Mitigation Activities 36
2. Relationship Between Structural Density and Probability ofInvolvement in
CWPP's and SFA Grants 37
3. Probability ofInvolvement in a CWPP as a Function ofthe Average Land
Hazard Rating and Percent Poverty 41
4. Probability ofInvolvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent Non-White Residents 41
5. Probability ofInvolvement in a CWPP as a Function ofthe Average Land
Hazard Rating and Percent with a High School Diploma 42
6. Probability ofInvolvement in a CWPP as a Function ofthe Average Land
Hazard Rating and Percent English Speaking Households 42
7. Probability of Involvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent Unemployment 43
8. Probability ofInvolvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent Poverty 47
9. Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent Non-White Residents 47
10. Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent with a High School Diploma 48
II. Probability ofInvolvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent English Speaking Households 48
12.Probability ofInvolvement in an SFA Grant as a Function of the Average Land
Hazard Rating and Percent Unemployment.. 49
Figure
xii
Page
13. Probability ofInvolvement in the Firewise Program as a Function ofthe
Average Land Hazard Rating and Percent Poverty 53
14. Probability of Involvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent Non-White Residents 53
15. Probability of Involvement in the Firewise Program as a Function ofthe
Average Land Hazard Rating and Percent with a High School Diploma 54
16. Probability ofInvolvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent English Speaking Households 54
17. Probability ofInvolvement in the Firewise Program as a Function ofthe
Average Land Hazard Rating and Percent Unemployment 55
xiii
LIST OF TABLES
Table Page
1. Measures of Social Vulnerability 15
2. WildfIre Risk Assessment Criteria - Overall Rating 22
3. WildfIre Risk Assessment Criteria - Land Hazard Rating 23
4. Summary of Socioeconomic Status Indicators 24
5. SFA Grant Totals by Activity, Arizona, 2001-2007 27
6. Social Vulnerability Indicators Grouped into Factors Based on Collinear
Relationships 33
7. Pearson's Correlation Coefficients Between Indicators ofSocioeconomic
Status 34
8. Frequency of WildfIre Mitigation Activities by Census Block Group (CBG)..... 35
9. Logistic Regression Results for Multiple Models to Predict Involvement in a
CWPP 40
10. Logistic Regression Results for Multiple Models to Predict Involvement in
SFA Grant Funded Projects 46
11. Logistic Regression Results for Multiple Models to Predict Involvement in
the Firewise Communities USA Program 52
Map
xiv
LIST OF MAPS
Page
1. CWPP Plan Areas and Likelihood ofInvolvement by Census Block Group 44
2. SFA Grant Project Locations and Likelihood of Involvement by Census Block
Group 50
3. Firewise Recognized Communities and Likelihood ofInvolvement by Census
Block Group 56
4. Wildfire Mitigation Activities and Percent Poverty by Census Block Group 62
1CHAPTER I
INTRODUCTION
The increasing threat of wildfire across the United States is a symptom of
shortsighted forest management and increasing human development in regions where
reoccurring fire is a component of a naturally functioning ecosystem. With the increase in
intensity and frequency ofwildfires over the past two decades, there has been a
corresponding increase in suppression costs (Dombeck et a!., 2004). In areas where
periodic, small fires once cleared the under story of woody debris, decades of fire
suppression have yielded an overstock of forest fuel for a catastrophic wildfire (Dellasala
et al., 2004; Hessburg et al., 2005). Logging practices have also contributed to the
problem by altering stand density and structure (Dombeck et a!., 2004). Population
growth and urban expansion into forested areas is complicating the issue as more homes
and lives are put at risk. An area where homes and wildland fuels meet or intermingle is
commonly referred to as the Wildland Urban Interface (WUI). A recent study revealed
that developed acreage in the WUI, characterized by low density residential development,
has increased by approximately 50% since the 1970's. By 2030 the WUI is likely to
expand an additional 10% mostly in the Intermountain West (Theobald and Romme,
2007). With climate change projected to increase wildfire risk across much ofthe United
2States, the wildfire problem will continue to be a serious concern for communities and
public lands managers (Dale et al., 2001; McKenzie et al., 2004).
In response, federal wildfire management policy has evolved from a command-
and-control approach focused on fire exclusion and rapid suppression, to a more
decentralized, proactive approach. The current approach to wildfire management is based
on three main components: 1) a framework for creating Community Wildfire Protection
Plans (CWPP's), 2) grant programs for wildfire mitigation activities and 3) Firewise, a
national program to promote wildfire awareness and local initiatives to mitigate risk
through education, outreach and technical assistance.
There has been extensive research on the biophysical factors that contribute to
wildfire risk (Daniel et al., 2007). Vegetation, topography, weather, and historical
patterns ofwildfire ignition are widely used to measure wildfire risk and identify
communities-at-risk (Jakes et al., 2007a). Therefore it is not surprising that these factors
figure heavily in prioritizing and allocating resources to mitigation efforts. However,
research from a variety oftypes ofdisasters demonstrate an increase in vulnerability
linked to specific human dimensions such as, age, gender, poverty, race, culture,
education and political influence.
The plight oflower-income citizens in the wake ofHurricane Katrina in 2005
underscore the differences in disaster vulnerability between those with economic and
political power and those without. Research on natural disasters suggests that such
disparity is evident in many types ofdisasters (Morrow, 1999) including wildfire (Haque
et al., 2007). This body ofresearch suggests that traditional planning modes, at least with
3regard to natural disasters, have failed to serve the least well offin society. Not
surprisingly then, a lack oftmst in public officials and institutions may prevent local
actors from engaging in planning processes. Yet, researchers in disaster planning and
management acknowledge the valuable expertise and contributions that even the most
disenfranchised can bring to disaster planning and response (Morrow, 1999).
Consequently, the purpose of this thesis was to evaluate whether federal wildfire
program resources that aim to involve local communities are reaching socially vulnerable
populations. In theory, resources should be going to the most at-risk populations. My
hypothesis is that socially vulnerable communities are less likely to be involved in federal
program efforts than less vulnerable communities of higher socioeconomic status.
4CHAPTER II
LITERATURE REVIEW
2.1 Wildfire Management Policy
First conceived in 1944, Smokey Bear and his message represent one ofthe most
successful government public relations campaigns ever. Today his message of fire
prevention is now recognized as a shortsighted and misguided policy attempt to manage
wildfire risk, but during the 1940's and 50's, attitudes towards public lands were
different. Many perceived public forests as sources oftimber to be managed and
protected from fire and there was wide support for employing a rapid, efficient,
command-and-control approach to fire suppression. Today, public attitudes are more
heterogeneous and our understanding of the importance of fire in natural ecosystems is
more sophisticated.
In a critique ofpublic forest management published in the mid-1980's, Allen and
Gould (1986) argue that U.S. Forest Service policy is misguided in attempting to apply
rational, scientific management decision processes to "wicked" public lands management
problems. More recently, several social scientists who study wildfire issues described the
development of the wildfire issue as the result of a "complex mix of physical, ecological,
economic, and social developments" (Carroll et aI., 2007, p. 239). They also point out
5that the wildfire issue spreads across jurisdictional boundaries and involves multiple
stakeholders. Solving this issue, they continue, will require an incremental, people-
centered approach rather than a single technocratic solution (Carroll et al., 2007).
Therefore an enduring strategy to solving the wildfire problem will require participation
from those communities-at-risk, particularly where those solutions impact the social,
economic and political fabric of the community.
In response to concerns about the rising costs of fire suppression, damage to
forests and losses to communities, the Clinton administration initiated an effort to revamp
federal wildfire management policy. That effort produced a report containing a series of
recommendations and lead to the development of the Western Governor's Association
(WGA) IO-Year Comprehensive Strategy for Reducing Wildfire Risk to Communities
and the Environment. These documents together are referred to as the National Fire Plan
(NFP) which describe the policy framework for reducing the threat ofwildfire by I)
improving fire prevention and suppression 2) reducing hazardous fuels 3) restoring fire
adapted ecosystems and 4) promoting community assistance (WGA, 2002). The
strategies outlined in the NFP represent a significant shift from a wildfire policy focused
solely on suppression to one that includes strategies for prevention and mitigation
through local community involvement (Steelman et aI., 2004).
Some critics ofenvironmental regulation claim that public lands management
policy bears some responsibility for the wildfire problem. The regulatory framework
imposed by the National Forest Management Act and the National Environmental Policy
Act (NEPA) makes it difficult for federal agencies to quickly plan and administer on the
6ground projects to reduce hazardous fuel buildup (Steelman and Burke, 2007). Therefore,
in 2002 the Bush Administration passed the Healthy Forest Initiative (HFI) which created
a class of"categorical exclusions" for qualifying fuels reduction projects, allowing such
projects to bypass the more lengthy NEPA analysis and review process (Steelman and
Burke, 2007). The following year Congress passed the Healthy Forest Restoration Act
(HFRA). HFRA outlines a framework for collaborative wildfIre planning and directs
communities to develop Community WildfIre Protection Plans (CWPP) to identify
critical risk factors, prioritize fuels reduction projects and establish the community's
Wildland Urban Interface (Will). HFRA also authorized $760 million annually in
funding for hazardous fuels reduction projects. The act instructs agencies to direct halfof
that funding to projects on private lands within the CWPP's identifIed Will (Steelman
and Burke, 2007). The intent ofthe CWPP process is to engage the community in a
leadership role in identifying priority areas for fuels reduction treatments. In developing
CWPP's communities are also encouraged to collaborate with state and federal agencies
(Newman, 2004). Community involvement and support for fuels reduction work on
private lands is critical because 89% of the Will acreage is privately owned (Theobald
and Romme, 2007).
During the past several years, many communities across the U.S. have completed
CWPP's, conducted fuels reduction projects using National Fire Plan (NFP) grant funds,
and completed other wildfIre preparedness activities (Jakes et at., 2007b). Although this
is encouraging, there is a lack of research to assess the effectiveness ofCWPP's.
7Similarly there is a lack ofconsensus regarding the effectiveness ofthinning to reduce
wildfire risk across different forest types (Daniel et al., 2007).
While expressing general support for the current direction of federal wildfire
policy, some see a need for greater emphasis on building community capacity to address
local wildfire issues. Steelman and Burke (2007) claim that fire suppression and fuels
reduction continues to be the top priority with significantly less funding being directed at
the other two goals: ecosystem restoration and community assistance. Without an
increase in both community economic and social capacity, communities will continue to
be dependent on federal dollars to mitigate wildfire risk. Steelman and Burke call on
Congress and land managers to measure progress on all the goals ofthe wildfire policy
(Steelman et al., 2004; Steelman and Burke, 2007). A 2004 report by the National
Academy ofPublic Administration found that federal programs do not explicitly address
the need to fund improvements to state and local capacities to plan and coordinate across
agency boundaries to accomplish landscape scale objectives (Wise and Yoder, 2007).
Participants in a series of focus groups including many stakeholders in the wildfire issue
called for more community involvement and emphasized building community capacity to
address wildfire risk mitigation (Bums et al., 2003).
In addition to funding through various NFP grant programs, communities can
access education and outreach materials and receive technical assistance through the
Firewise Program which was initiated in 2001. Publications, newsletters and educational
curricula are available through the program website as well as contact information for
statewide Firewise program coordinators. Firewise Communities USA is a specific
8component of the Firewise program that outlines a process by which participating
communities become 'Firewise Recognized' by meeting program benchmarks. Specific
activities vary across communities, but all recognized sites are required to create a
community wildfire plan, implement at least one community wildfire preparedness
project each year, spend $2 per capita annually on wildfrre projects and maintain an
active board of community volunteers to coordinate the plan. Recognition status is re-
evaluated annually. Although recognition status does not currently confer special
benefits, it could become a criterion for assistance grants or insurance coverage in the
future.
The first Firewise Communities USA pilot project was initiated in six states in
2001; to date there are 288 recognized communities in 36 states. The program depends on
homeowner commitment and local leadership. Although there is no size limit, in practice,
most Firewise communities are neighborhood organizations or home owner associations.
Arizona was one of six states to participate in the frrst year of the program
beginning in 2001. Since then 23 communities in Arizona have earned recognition status.
Yet there are many other neighborhoods, subdivisions, and towns that have not
participated who are also at risk.
Planning efforts, grant programs and the Firewise programs are available to all
local communities, but state agencies can also mediate the allocation ofprogram
resources helping to direct them to high priority communities-at-risk. Research on federal
funding allocation in Arizona, New Mexico and Colorado reveals that community access
to federal funds for fire mitigation activities is impacted by state program organization
9and prioritization based on biophysical risk factors. In New Mexico and Arizona
resources are directed to high-risk communities as identified by state agencies, whereas
Colorado has not prioritized particular communities and allocates a greater percentage of
federal dollars to statewide programs than New Mexico or Arizona (Steelman et af.,
2004).
2.2 Natural Disasters and Social Vulnerability
Approaches to disaster management have changed in the past few decades away
from a command-and-control top down reactionary approach to a more proactive
approach focused on mitigation and preparedness. Concurrently, the field ofdisaster
research expanded during the 80's and 90's recognizing the importance of political and
social conditions as factors in community capacity to prepare and respond to a disaster
(Cutter et af., 2000). Research has shown that the negative impacts are a function ofthe
social, political and economic environment as well as the natural processes that initiate
them (Fothergill and Peek, 2004; Haque et af., 2007). Indeed, disasters highlight a
community's weaknesses, both physical and social characteristics that contribute to
decreased capacity and resilience (Flint and LulofI, 2005).
2.2.1 Dimensions ofSocial Vulnerability
Researchers have identified multiple dimensions that contribute to a reduced
capacity to "anticipate, cope with, resist, and recover from the impact ofa natural hazard"
(Blaikie et af., 1994). The underlying factors that contribute to social vulnerability are
10
similar to those that produce social inequities - lack of access to resources, information,
political power, limited social capital and physical frailty (Cutter et a!., 2003).
Poor people are more likely to suffer negative impacts including property loss,
physical harm and psychological distress. Households with fewer financial resources are
less likely to take steps to prepare for a disaster and more likely to have difficulty during
the recovery phase (Fothergill and Peek, 2004).
At a community level, those that are well-prepared in terms ofboth economic and
social infrastructure are more adept and responding to and recovering from natural
disasters (Kumagai et a!., 2004a). The elderly are more likely to lack adequate economic
resources and physical ability to respond effectively and they are more likely to suffer
health consequences, physical harm and be slower to recover. Likewise children are more
vulnerable because oftheir dependence on family support (Morrow, 1999). People with
mental and physical disabilities are at increased risk because the will require extra
assistance (Morrow, 1999). Several researches have demonstrated cultural and ethnic
differences in risk perception and response (Buckland and Rahman, 1999). A lack of
education, literacy and language skills can cause disadvantages in responding to a
disaster when seeking information, applying for assistance or seeking post disaster
employment (Morrow, 1999). Gender has also been identified as a factor in vulnerability
(Cutter, 1995; Fothergill, 1996). The ability of a community to recover is related to its
capacity to engage in political processes, furthermore, the disadvantages posed by
income, language, ethnicity, race and political marginalization are compounded
(Morrow, 1999).
11
Because socioeconomic status is such an important factor in vulnerability,
effective emergency management needs to consider the different human dimensions as
well as the biophysical causes of disaster (Buckle et ai., 2000). In practice, measuring,
identifying and developing strategies to address social vulnerability is complex. Part of
the difficulty in measuring social vulnerability is due to the fact that rather than being
isolated units, populations of people are in fact members ofoverlapping units defmed by
geographic boundaries but also social and political relationships (King, 2001; Buckle et
ai., 2000). There is also a need for further research and comparative studies to illuminate
the interaction between social vulnerability and the impacts ofdifferent types ofdisasters
(Fothergill and Peek, 2004) to inform strategies relevant to the types ofdisasters that
communities face. Unfortunately our understanding of social vulnerability is very limited
compared to our understanding ofbiophysical vulnerability. This is due in part to the
difficulty in quantifying the social impacts of disasters (Cutter et ai., 2003). A better
understanding ofthe interactions between biophysical and social vulnerabilities at
multiple scales (local, regional, national) will improve our hazard assessments making
them more objective and less subject to "political whim" (Cutter et ai., 2003, p.258).
2.2.2 Social Vulnerability in the Wildfire Context
By comparison wildfires have received less attention in the field of disaster
research than hurricanes, floods, earthquakes and other catastrophic natural events; one
possible explanation is the (misguided) perception that wildfires are manageable through
suppression (McCaffrey, 2004). Consequently, social vulnerability in the wildfire context
is perhaps underestimated. Although interest in the social dimensions of risk management
12
has increased, it has not been fully integrated into the management ofwildfire. Much of
the early research was predicated on a rational theory approach. Basically, the theory
holds that once residents understand the risk they will be motivated to take action to
reduce their risk (Collins, 2005). But, the way that people perceive and measure risk
varies. Research has shown that attitudes towards government sponsored programs,
cultural beliefs about wildfire, and past experience with wildfire are important
determinants of involvement in wildfire mitigation activities. Although these findings are
important, Collins asserts that the socioeconomic barriers to mitigation action have not
received due attention. This is due in part to the assumption that WUI residents are
comprised primarily of "amenity migrants", those that chose to live in areas most at risk
to wildfire (Collins, 2005). But, many Will residents are not "amenity migrants" and
wildfire impacts can vary significantly between households within the same community.
For example, renters have fewer options than homeowners, especially those homeowners
with adequate insurance and the resources to rebuild or relocate (Carroll et al., 2005).
Several studies have found that financial constraints limit residents' ability to take
precautionary measures. An Australian study showed that people with mental or physical
disabilities and those suffering from poverty are more at risk to structural fires because
they are less capable of responding in an emergency and more likely to have substandard
living conditions (Rhodes and Reinholtd, 1998). A survey of households in a WUI
community in California found that cost was the most common barrier to taking steps to
reduce their home's ignitability (Collins, 2005). Similar results were found in a study of
residents living in Colorado's Front Range, where residents cited concerns about cost,
13
time commitment, and a lack ofphysical ability to implement mitigation actions (Absher
and Vaske, 2007).
The findings from these academic studies are corroborated in a report on wildfrre
and poverty in the Western United States by Niemi and Lee (2001). The study's authors
found that poor households are more likely to have inadequate self-protection for
housing, limited access to health care a greater proportion oftheir economic assets at risk
to wildfire, and decrease resiliency to recover from the impacts of a wildfire.
Poverty also has impacts at a community level. An analysis of fire district
protection capability and poverty found conducted across the state of Washington found
that poor households are more likely in fire districts with low response capacity (Lynn
and Gerlitz, 2005). A study in Florida that sought to associate socioeconomic variables
with wildfire intensity showed that counties with higher incidence ofpoverty had fewer
ignitions, but once ignited suffered larger, more intense fires. The researchers speculate
that a lack of suppression capacity may account for the fmding (Mercer and Prestemon,
2005).
In addition to decreased capacity to prepare, poor communities are less likely to
recover quickly from a wildfrre. Where community resources are scarce these disruptions
are likely to be more severe. Communities can be impacted by the disruption of social
process, changes in the allocation of resources towards restoration and reconstruction
projects at the expense ofother community developments (Jakes, 2007).
Community responses to wildfire threat fall into two types, structural and social.
Structural responses focus on biophysical aspects such as actions to reduce hazardous
14
fuels, apply land use regulations, enforce building codes and enhance fire suppression
capacity. Social responses refer to intangible processes including planning, management,
organization and decision making processes. There is a need to better understand the how
socioeconomic status impacts these intangible processes (Steelman and Kunkel, 2004).
2.2.3 Indicators ofSocial Vulnerability in the Wildfire Context
In 2000 Case et al. (2000) suggested using Census data to model community risk
to wildfire based on socioeconomic measures, specifically populations of the very old,
the young and those suffering from poverty. They argue that the total social impacts
would be reduced by taking a strategic approach to protecting those where the social risk
is greatest. To measure social vulnerability in the wildfire context, I generated a broader
list of indicators based on two previous efforts to describe social vulnerability.
Cutter et al. (2000) developed a social vulnerability index for emergency
managers to use as a tool to compare risk across the United States. They started with a
review of literature and through factor analysis reduced 42 unique metrics down to 11
principle factors including personal wealth, age, density ofthe built environment,
occupation, household stock and tenancy, single-sector economic dependence,
infrastructure dependence and three factors related to differences in race and ethnicity.
In another effort to build a construct of social dimensions related to wildfire risk
and resilience, Evans et al., (2007) developed an Index of Community Capacity for
Protection from Wildfires (ICCPW). They also conducted a review ofthe literature and
reference some of the same research as Cutter et al. Although their index seeks to
measure capacity, its inverse - lack of capacity - is closely related to social vulnerability
15
producing much overlap between the constructs. Like the social vulnerability index, the
ICCPW includes measures of age, wealth, and employment, and ethnicity.
Table 1 describes the socioeconomic measures I chose for this project with
relevant citations from the natural disaster literature. I included dimensions that were
common to both indexes developed by Cutter et al. and Evans et aI. (e.g. age, wealth,
employment and ethnicity). I also sought measures that were readily available local scale;
in this case the Census Block Group.
Table 1. Measures of Social Vulnerability.
"easure Description Reference
Percent Vulnerable Age Total youth «15 yrs) plus total elderly (Aptekar and Boore, 1990;(>64 yrs) divided by the total population Morrow, 1999; Ngo, 2001;)
Race Percent of population that is not one race (Bolin, 1986) Peacock et aI., 1997;
= white Pulido, 2000)
Single-Mother Percent of households headed by a (Cutter, 1995; Puente, 1999;
Households single-mother Morrow, 1999)
Percent of the population with a (Tobin and Ollenburger, 1993;
Physical Disability disability Rhodes and Reinholtd, 1998;
Morrow, 1999)
Education Percent ofthe population that has earned (Heinz Center for Science, 2000)
a high school diploma
Percent ofpopulation that speak only (Buckland and Rahman, 1999)
Language English or speak English ''very well" or
''well''
Median Income Household median income (Blaikie et aI., 1994)
Poverty Percent offamilies below the federal (Niemi and Lee, 2001; Fothergillpoverty limit and Peek, 2004)
Unemployment Percent unemployment (Mileti, 1999)
2.2.4 Economic Vulnerability in Rural Communities
Many ofthe communities at-risk to wildfire are also economically linked to the
use ofnatural resources on adjacent public lands. For example, The Rodeo-Chediski fire
in Arizona in 2002 impacted both tribal and non-tribal communities. The tribal
16
communities will likely suffer greater long-tenn economic impacts from the loss of
timber resources than the non-tribal communities (Carroll and Cohn, 2003).
Flint and Luloff (2005, p.400) suggest that researchers' perspectives on natural-
resource based communities and theories of social vulnerability to natural disasters both
overlook the role of endogenous initiative and capacity. They identify the "traditional
perspective" on natural resource-based communities which holds that they are more
vulnerable to environmental and social change, economically unstable and subject to
"unbalanced power relationships with external institutions and agents" But Flint and
Luloff (2005) also describe recent research that reveals community initiative and capacity
in developing the economic and non-economic benefits of surrounding resources (Bridger
and Luloff, 1999; Luloffet al., 2003). Similarly, according to Flint and Luloff (2005, p.
402) the "traditional" view in natural disaster research characterizes vulnerable
communities as helpless and dependent on external resources for disaster response and
recovery. But, as with economic empowennent, researchers are beginning to
acknowledge the importance of"local knowledge, action, participation, and control". In
conclusion Flint and Luloff (2005) call for more research that seeks to understand
community capacity and how communities act in response to perceived risks.
2.2.5 Helping Agencies
Federal, state, and local agencies have an important role in assisting communities
to prepare for, respond to, and recover from a disaster by in facilitating access to federal
resources for groups that have been historically marginalized (Bolin and Stanford, 1998).
However, for emergency managers to effectively address the peculiar needs ofvulnerable
17
populations, they need to have data on the types ofpeople within their communities and
what types of assistance they may require (King, 2001). Such information can come from
social vulnerability indicators and mapping exercises like those described above, but also
through partnerships and dialogue with socially vulnerable communities.
The challenge is that the level of social and economic development generally
correlates with community capacity to develop productive partnerships with government
agency disaster management efforts (Buckland and Rahman, 1999). In a disaster
management scenario, social capital, Le. social networks built on trust and reciprocity,
leads to more effective community response (Neal and Phillips, 1995). Localized wildfire
mitigation efforts that empower communities, such as the CWPP process facilitate
collaboration and can lead to increased social capital within a community (bonding
capital) and between local stakeholders and outside helping agencies (bridging capital)
(Jakes et a!., 2007a)
Research on community social reactions to wildfire highlight the importance of
both types of social capital. Conflict between local and non-local entities during and after
a wildfire event are more likely where there tensions between local and outside agencies
already exists (Jakes, 2007). Tensions can also result from the loss of community trust in
land managers' ability to mitigate wildfire risk. This trust is particularly vulnerable where
past management practices and policies have led to suspicion and controversy between
local and outsider interests (Mendez et a!., 2003; Kumagai et a!., 2004b) or where there
are difference in culture between disaster victims and assistance agencies (Morrow,
1999). A study ofcommunity response to the Rodeo-Chediski fire in Arizona in 2002
18
demonstrated how the event could foster cohesion and conflict between and among local
and non-local entities. Community characteristics including history, culture, and social
norms created both challenges and opportunities during the fire response and the
following recovery (Carroll et ai., 2005; Burchfield, 2007). Thus communities with high
social capital are more likely to respond and recover more efficiently and effectively.
2.3 Synthesis
In summary, federal policy provides opportunities for helping agencies to engage
communities in proactive efforts to reduce risk. The CWPP process, NFP grants and
Firewise Communities USA program are the three main components of this policy.
Research demonstrates that a suite ofsocioeconomic dimensions are correlated with
increased vulnerability. These factors include age, race, disability, gender, political
influence, poverty, education and employment. Despite this research, relatively little is
known about social vulnerability in the wildfire context. Rather, the focus has been on
understanding biophysical factors of risk and educating WUI residents to encourage
mitigation action. This thesis seeks to assess the extent to which socially vulnerable
populations are involved in each ofthe types ofwildfrre mitigation efforts. Findings from
this research will help to determine if federal resources are being allocated equitably and
highlight factors that may limit community capacity to engage in mitigation efforts.
19
CHAPTER III
MEASURES AND METHODS
Data on socioeconomic condition, wildfire risk, and mitigation activities were
obtained from multiple sources. The first phase ofthis project involved integrating these
data sources into a cornmon unit of analysis. This phase merged overlapping data maps or
layers into a single layer yielding a single data table with one record for each Census
Block Group (CBG). The second phase was a statistical analysis of the data set to identify
significant relationships among and between measures of socioeconomic condition,
wildfire risk, and mitigation activities.
3.1 Study Area
Arizona presented an interesting case and appropriate study area for several
reasons. First, the state has a diverse mix of communities including Native American,
Hispanic, and so-called amenity migrants that are predominantly white, more affluent and
often retirees. As well as racial and ethnic diversity, there are significant class and
economic disparities; some communities are very affluent and others having high rates of
unemployment and poverty. All of these communities have been evaluated by a statewide
comprehensive risk assessment and many are at-risk to wildfire. Second, Arizona was
one of six states to initiate the Firewise Communities USA program in 2001, a federally
20
funded program that recognizes community efforts to mitigate wildfire risk. As an early
participant in the program, communities in Arizona have had access to the program for
several years and 24 communities have participated in the program making Firewise
recognition status a useful measure of wildfire mitigation activity. Although it may not be
a perfect microcosm ofthe Western United States where wildfire management is most
acute, Arizona presents many of the same types ofcommunities and issues faced by other
states.
3.2 Unit of Analysis
This project uses the Census Block Group (CBG) as the unit ofanalysis. The U.S.
Census provides an extensive array of data types at the CBG level that are not available at
the Census Block level. Other larger units such as Census Tracts, Census Designated
Places or ad hoc aggregations ofCBG's could mask significant socioeconomic variation
within such larger units. But CBG's are not homogenous socioeconomic units either;
Where CBG's are large, they may include diverse populations. My assumption is that the
splitting or aggregation ofpopulations caused by the arrangement ofCBG boundaries is
not biased towards over or under representing populations of specific socioeconomic
characteristics.
The 2000 census divided Arizona into 3,554 CBG's. CBG's with a very low risk
to wildfire based on the Arizona Statewide Comprehensive Risk Assessment conducted
in 2004 were excluded from this analysis. This effectively excluded those CBG's in
urban areas or other inhabited places that lack vegetation to warrant a significant wildfire
21
risk. CBG's were also excluded where the census was incomplete which occurred where
the population count was zero or very small « 10), but also included one CBG with a
population of48. In total 14 CBG's were excluded on the basis of incomplete
information yielding a total data set of960 records. My assumption is that the excluded
CBG's represent such a small fraction ofthe data set that their exclusion does not bias the
findings.
3.3 Biophysical Wildfire Risk Factors
Data on the biophysical wildfire risk factors were obtained from the Arizona State
Land Dept., Forestry Division. To evaluate risk for communities throughout Arizona, I
considered two potential sources: the Federal Register List ofCommunities-at-Risk
(2001) and the Arizona statewide comprehensive risk assessment (2004).
The Federal Register List identifies 159 communities in Arizona and ranks each
as high, medium or low risk. The list is restricted to communities that are adjacent to
federal lands and identified as Census Designated Places. Many smaller, populated areas
throughout Arizona are not included on the Federal Register list.
The statewide risk assessment lists 902 unique places and rates each according to
several criteria described in more detail below. The assessment was produced through a
partnership that included the Arizona State Land Dept., Forestry Division, USFS, BLM,
NPS, FWS and BIA. Staffused digital ortho quads to identify developed areas and named
unique communities using several sources including USGS names, place names and
towns.
22
I elected to use the statewide risk assessment data because in included GIS data
identifying the geographic footprint of each community. Plus, it provided data on the
separate factors included in the assessment. For example, I was able to access
information about the topography, forest fuels, historic fIre occurrence and structural
density for each community. Furthermore, the data was detailed down to a 1 km grid. In
contrast, I was only able to obtain latitude and longitude coordinates identifying a point
for each community on the Federal Register list and its associated risk rating ofhigh,
medium or low.
The statewide risk assessment considers several biophysical factors that relate to
the probability of a fIre occurrence and likelihood ofdamage to populated areas. These
factors are weather, topography, fuels, historical fIre occurrence and the presences of
structures. Typically these factors are combined into an index and used to rank wildfIre
risk as an overall rating from low to high based on probable occurrence and likely
intensity (Table 2). The statewide risk assessment also produced a simplifIed land hazard
rating (Table 3). Rather than using the overall rating, I elected to analyze the land hazard
rating and structural density rating as separate variables to be able to observe independent
effects from these variables in the statistical analysis.
Table 2. Wildfire Risk Assessment Criteria - Overall Rating.
Fire Regime 25%
Risk
Structural Density
35%
20%
35%
23
Table 3. Wildfire Risk Assessment Criteria - Land Hazard Rating.
Slope 60%
Fire Regime 25% 70%
20%
Whereas the statewide risk assessment developed land hazard and structural
density ratings by community boundaries, I needed to calculate these values for each
Census Block Group (CBG). Using GIS, I integrated data from the statewide risk
assessment with a CBG data map layer obtained from Arizona Geographic Information
Council. Specifically, I calculated the weighted average land hazard rating and structural
density rating for the developed area within each CBG. Similarly I determined the
maximum land hazard rating and structural density rating for each CBG.
3.4 Indicators of Socioeconomic Status
I selected a suite of socioeconomic indicators from a review ofliterature
pertaining to social vulnerability. Data were obtained from the 2000 U.S. census and used
to calculate specific measures (Table 4). Information on age, household relationship, and
race was taken from Summary File I (SF-I) which is based on a 100% sample. Other
information on education, employment status, and income was obtained from Summary
File 3 (SF-3) which is calculated from a sample ofthe population.
Table 4. Summary of Socioeconomic Status Indicators.
\ lU'iable Description
Percent Vulnerable Age Total youth «15 yrs) plus total elderly (>64 yrs) dividedby the total population
Non-White Percent ofpopulation that is not one race = white
Single-Mother Households Percent ofhouseholds headed by a single-mother
Disable Percent ofthe population with a disability
Education Percent of the population that has earned a high schooldiploma
English Percent ofpopulation that speak only English or speakEnglish "very well" or "well"
Median Income Household median income
Poverty Percent offarnilies below the federal poverty limit
Unemployment Percent unemployment
3.5 Wildfire Mitigation Activities
There are many ways that households and communities could mitigate their
wildfire risk. Activities could include creating defensible space around homes by
reducing buildup of flammable vegetation and debris, fitting homes with fire-resistant
materials, developing evacuation plans, purchasing suppression equipment, purchasing
disaster insurance or educating residents about the risks. This thesis is focused on the
components of current federal wildfIre management policy and is therefore limited to
data on Community Wildfire Protection Plans, State Fire Assistance grant awards, and
participation in the Firewise Communities USA program.
There are other grant programs to aid communities in managing wildfire risk
besides the State Fire Assistance Grant program, but I was unable to obtain data on the
Volunteer Fire Assistance Program (VFA), Rural Fire Assistance Program (RFA),
Economic Action Program (EAP) and Community and Private Land Fire Assistance
Program (CPLFA). The RFA and VFA programs continue to be funded as of 2008, but
24
25
the EAP and CPLFA programs have not been funded since 2004. While SFA grants have
and continue to represent the majority ofNFP grant funds expended in the state, omission
of the data on the other grant programs may under-represent community involvement in
grant funded wildfire mitigation projects.
For simplicity, I use the term "involved" as a generic way to describe a population
that is either actively engaged in an activity, or potentially benefits from that activity such
as a neighborhood that benefits from an adjacent fuels reduction project or is within the
plan area ofa CWPP.
3.5.1 The Firewise Communities USA Program
I obtained data on communities that have participated in the Firewise
Communities USA program from the Arizona State Land Dept., Forestry Division. Staff
provided a spreadsheet list of communities, dates of initiation in the program, recognition
date and current status in the program. As of January 2008, 24 communities were
involved in the Firewise Communities USA Program. Ofthose, 22 were 'recognized' in
2008; one is inactive and one is in the process ofearning recognition status. Timber
Ridge, near Prescott, was the first community in Arizona to receive Firewise recognition;
it earned recognition status in 2002.
To determine geographic location, I attempted to match recognized communities
to the list ofcommunities-at-risk from the statewide risk assessment and the Federal
Register List. I was only able to match about halfof the recognized communities in the
data set. To locate the others, I conducted an Internet search using Google. By searching
using the community name and/or the name of the lead organization, often a homeowners
26
association, I was able to identify approximate locations for the remaining communities
and use Google Earth to obtain latitude and longitude coordinates to create point
locations in a GIS data map layer. Lacking information about the geographic footprint of
the group of residents that comprise the Firewise community, I calculated a % mile buffer
zone as an estimate. I then assigned Firewise involvement to CBG's that contained a
Firewise community or intersected with this % mile buffer. The Arizona State Firewise
coordinator inspected a series ofmaps for each community and confirmed that my
methodology produced a reasonable approximation for the location ofeach recognized
community.
3.5.2 Community Wildfire Protection Plans (CWPP)
As of January 2008 there were 16 CWPP's in place throughout Arizona. As well,
there were seven plans in various stages of development. These unfinished plans are not
included in this analysis. All of the CWPP's are variable in both geographic extent and
scope. The largest encompasses all ofGraham County and the smallest includes just a
few, small communities. The earliest plans were adopted in 2004.
I estimated the geographic boundaries ofthe CWPP plan area from a visual
inspection ofa map provided by the Arizona State Lands Dept., Forestry Division. Using
the map as a reference, I manually digitized CWPP boundaries into a GIS data map layer.
Then I overlaid CWPP layer with the CBG map layer. Then, I manually linked CBG's to
CWPP's where a majority ofthe developed area from the CBG fell within a CWPP plan
area. Each CBG was determined to be either in a CWPP plan area, or not in a CWPP plan
area.
27
3.5.3 State Fire Assistance (SFA) Grants
The Arizona State Land Dept., Forestry Division provided data on SFA grants
awarded from 2001 to 2007. Grants were awarded for fuels reduction work, outreach and
education, and planning. The data set included the community name and sponsoring
organization, awarded amount and a brief description ofthe purpose of the grant. In total
across the six year period approximately $19 million was awarded with the bulk of
funding for fuels reduction work (85.1%) (Table5).
Table 5. SFA Grant Totals by Activity, Arizona, 2001-2007
Grant Acth it~ Total \molJnt Percent
Education and Awareness
Fire Suppression Equipment
Fuels Reduction Projects
Planning
Restoration
Total
Source: Arizona State Lands Dept. Forestry Division
$ 1,904,385
$ 131,937
$ 16,272,369
$ 182,390
$ 628,798
$ 19,1l9,879
10.0%
0.7%
85.1%
1.0%
3.3%
100.0%
SFA grants are available to Western States on a competitive basis. SFA grants are
intended to support activities related to fuels reduction, education, and planning.
Applicants must demonstrate a 50:50 match which can be a hard cash match or through
in-kind contributions of labor or donated equipment. Grants are more competitive if they
will produce measurable outcomes, include collaboration, support an existing community
wildfire plan and are likely to be enduring.
Using GIS, I linked the communities-at-risk data map layer from the Statewide
Risk Assessment, the SFA grants data table. Some grants could not be joined to specific
community where the community was listed as an entire county or in a couple instances
as "statewide". These grants and a few others that could not be associated to a specific
28
community from the Statewide Risk Assessment were excluded. These excluded grants
represent approximately 20% ofthe total dollar amount awarded and could bias the
findings if such grants were more likely to benefit populations ofa particular
socioeconomic status.
To associate SFA grant awards with Census Block Groups (CBG's), I overlapped
the communities map layer and the CBG map layer and joined the data sets. Where a
community which had benefited from one or more SFA grants intersected a CBG, I
coded the CBG as being "involved" with an SFA grant project, all other CBG's were
coded "not involved". Limitations in the data set precluded a more precise methodology;
detailed geographic information to specific populations would reduce measurement error.
However, my assumption is that the methodology applied does not bias the findings
along socioeconomic dimensions.
3.6 Limitations
Limitations are inherent in this study due to the nature and quality of the data.
Perhaps the most significant, which has been mentioned already, is the omission of grant
data from other wildfire mitigation grant programs. Particularly the Volunteer Fire
Assistance Program (VFA) and the Rural Fire Assistance Program (RFA). Both ofthose
programs are targeted towards increasing the capacity of communities that lack adequate
resources for wildfIre suppression. Had these data been available, it might alter the results
as poor, rural communities might be more likely to be involved in the VFA and RFA
programs.
29
A second limitation is the difficulty in using Census Block Groups (CBG's) as the
unit of analysis. CBG's in Arizona vary widely in area and population. Therefore
measurement errors in calculating socioeconomic characteristics and biophysical traits
are more likely in the larger CBG's. A related limitation is the use of structural density
from the Statewide Risk Assessment. Density per developed area within a CBG is not the
same as the size of a community. Since CBG's vary so widely in size and most divide
rather than encompass cities or towns, the complexity of the task prohibited me from
creating a community size variable for each CBG. It is likely that the size of a community
or proximity to an urban center is a significant variable, but its effect will have to be
estimated qualitatively from the maps.
It is difficult to estimate the impacts ofpotential measurement errors, but I am
assuming that they do not bias the results as they are not likely to systematically shift the
measurements ofkey variables.
Lastly, data freshness could be an issue for this study. Data used during this study
were collected and accumulated over a period ofapproximately nine years beginning
with the data from the U.S. census and ending with the most recent update ofFirewise
recognized communities in January 2008. According to U.S. Census Bureau estimates,
between 2000 and 2006 Arizona's population increased by 20.2% compared to a growth
rate of6.4% for the U.S.
30
3.7 Analysis
The statistical analysis consists of two phases, first an inspection ofcollinear
relationship between indicators of socioeconomic status, then regression analysis to
explore relationships between socioeconomic status and involvement in wildfire
mitigation activities.
When two or more independent variables are highly correlated it is difficult to use
statistical methods to discern the relative influence ofeach on the dependent variable.
Therefore when using a set ofmultiple dependent variables it is common for researchers
to attempt to reduce their suite ofmeasures to some smaller number that still serves as a
proxy for the underlying factor of interest. Many indicators of social vulnerability are
highly correlated such as poverty and median income indicating they are measuring a
similar community characteristic. Others are less so, such as disability and language.
Using SPSS, I calculated Pearson's correlation coefficients for each of the possible
bivariate relationships within the suite of social vulnerability measures. I then used these
results to group indicators that were highly correlated and interpret the [mdings from the
logistic regression analysis.
A binary logistic regression analysis is used to assess the ability of an independent
variable predict the dependent variable when the dependent variable is dichotomous. For
this study the dependent variables are involvement in 1) The Firewise Communities USA
program 2) An established CWPP and 3) A State Fire Assistance grant funded project
during 2001-2007. In a binary logistic analysis the independent variable is labeled the
predictor and the dependent variable the outcome. Including multiple predictors in the
31
regression can reveal the effects ofmultiple variables to evaluate the relative influence of
different predictors and determine statistical levels of significance for these affects.
32
CHAPTER IV
FINDINGS
The findings section has two components. First, I report on correlations between
indicators of socioeconomic status. Then, I describe the relationships between
biophysical and socioeconomic characteristics and the likelihood of involvement with
wildfire mitigation activities. For each wildfire mitigation activity, I use a logistic
regression to determine ifbiophysical risk factors for wildfire and socioeconomic status
predict involvement in wildfire mitigation activities.
4.1 Correlations Between Socioeconomic Indicators
Pearson's correlation coefficients indicate that the indicators of socioeconomic
status in this data set cluster into one main factor that includes seven of the nine
indicators (Table 6). The main group includes the measures of single-mother households,
poverty, education, race, unemployment, median income, and language. Within the main
group, single-mother households, poverty and education are the most highly correlated
with each other and other indicators in the group. This indicates that one of these
measures would serve as the best proxy for the factor as a whole. Conversely, English
was the least correlated variable, but still highly correlated with education.
33
The percent ofpeople with disabilities and the percent vulnerable age population
each represent two additional separate factors. While there is a correlation between these
indicators and each other as well as some correlations with the indicators in the main
group, the coefficients are less indicating that they are measuring a different dimension of
the overall concept of social vulnerability.
Table 7 provides specific Pearson's correlation coefficients for each ofthe
bivariate correlations and two-tailed test for significance.
Table 6. Social Vulnerability Indicators Grouped into Factors Based on Collinear
Relationships.
Group Yariablc Internal Collinearit~
Single-Mother Households Strong
Poverty Strong
Education Strong
Factor 1 Non-White Moderate
Unemployment Moderate
Median Income Moderate
English Weak
Factor 2 Percent Vulnerable Age
-
Factor 3 Disabled
-
Based on these [mdings, I conducted multiple logistic regression analysis, each
using a different set of factors or variables. A comparison ofthese different models and
their significance is discussed in the next section.
Table 7. Pearson's Correlation Coefficients Between Indicators of Socioeconomic Status.
Single- Non- Unemploy- Median VulnerableMother Poverty Education English Disabled
Households White ment Income Age
Single-Mother 1 .723** -.617** .833** .602** -.512** -.367** -.205** 0.028
Households IJ value 0 0 0 0 0 0 0 0.391
Poverty .723** 1 -.654** .743** .628** -.638** -.451** -.071 * .170**
IJ value 0 0 0 0 0 0 0.027 0
Education -.617** -.654** 1 -.666** -.531** .645** .733** 0.029 -.292**
IJ value 0 0 0 0 0 0 0.374 0
Non-White .833** .743** -.666** 1 .620** -.497** -.422** -.138** 0.055
IJ value 0 0 0 0 0 0 0 0.09
Unemployment .602** .628** -.531** .620** 1 -.449** -.393** -.127** .098**
IJ value 0 0 0 0 0 0 0 0.002
Median Income -.512** -.638** .645** -.497** -.449** 1 .377** -.130** -.444**
lJ value 0 0 0 0 0 0 0 0
English -.367** -.451** .733** -.422** -.393** .377** 1 .075* -.067*
IJ value 0 0 0 0 0 0 0.019 0.038
Vulnerable Age -.205** -.071 * 0.029 -.138** -.127** -.130** .075* 1 .377**
IJ value 0 0.027 0.374 0 0 0 0.019 0
Disabled 0.028 .170** -.292** 0.055 .098** -.444** -.067* .377** 1
lJ value 0.391 0 0 0.09 0.002 0 0.D38 0
••. Correlation is significant at the 0.01 level.
•. Correlation is significant at the 0.05 level.
Vol
+:>
35
4.2 Frequency of Wildfire Mitigation Activities
State Fire Assistance (SFA) Grants were the most common type ofactivity within
this data set (Table 8). Over half (5 1.6%) of the Census Block Group's in the data set
were associated with at least one such project between 2001 and 2007. About a third
(27%) ofthe CBG's were within a CWPP plan area. However, participation in the
Firewise Communities USA program was very rare. Only 5.4% of the CBG's in the data
set were associated with a Firewise Recognized Community.
Table 8. Frequency of Wildfire Mitigation Activities by Census Block Group (CBG).
\\ ildfit'e 'litigation \cth it~ C BG's 1m 01\ ed Percent of Total*
cwpp
SFA Grant Project
Firewise Recognized Community
*n=960 Census Block Groups
4.3 Biophysical Factors
4.3.1 Land Hazard Rating
204
327
49
27.0%
51.6%
5.4%
The average land hazard rating variable was a consistent and substantial predictor
of involvement in each of the three wildfire mitigation activities. Across multiple logistic
regression analyses with different combinations of variables, the beta-l coefficients for
the average land hazard variable were stable. In every case an increase in the average
land hazard was positively correlated with an increase likelihood of involvement in the
wildfire mitigation activity. Figure 1 graphically represents how changes in the average
land hazard rating are correlated with probability of involvement for a hypothetical CBG
36
with mean values for each ofthe other variables in the analysis. Involvement in the
Firewise program was the most sensitive to the land hazard rating with a predicted 27-
fold increase across the range of land hazard ratings. The likelihood of involvement in a
Cwpp increased II-fold from the lowest to the highest average land hazard rating and
the likelihood of an SFA grant project increased by a factor of 3.8.
Figure 1. Relationship Between Land Hazard Rating and Probability of Involvement
in Three Types of Wildfire Mitigation Activities.
100%
... 90%
c
Gl 80%E
Gl 70%~
0
60%>
.5
.... 50%0
~ 40%
:c 30%
ta
.c 20%e
IJ. 10%
0%
4 6 8 10 12 14 16
--+-CWpp
___ GRANTS
---.-FW
Average Land Hazard Rating
4.3.2 Structural Density
Similar to the average land hazard rating, the structural density rating was
positively correlated with an increase in the likelihood of involvement in a CWPP and an
SFA grant project. However, the land hazard rating was not statistically significant in
predicting participation in the Firewise Communities USA program. Compared to the
land hazard rating, the probability of involvement was less sensitive to changes in the
37
structural density variable. Figure 2 graphically represents how the probability of
involvement changes across the range of average structural density ratings. From low to
high average structural density the likelihood of involvement in an SFA grant project
increases by a factor of2.7, but the predicted likelihood ofbeing involved in CWPP only
increases by a factor of 1.3.
Figure 2. Relationship Between Structural Density and Probability of Involvement in
CWPP's and SFA Grants.
100% ,---------------------,
... 90%
c
Ql 80%E
Ql 70%~
0 60%>
.5
.... 50%
0
~ 40%
:c 30%
III
J:l 20%0
..
/1. 10%
0%
--
-------
~ ~
...
-=------
..----
,j
, ,
I -+-cwpp
___ GRANTS
4 6 8 10 12 14 16
Average Structural Density Rating
4.4 Socioeconomic Factors
The analysis of the socioeconomic factors is more complex than the biophysical
factors because there are more variables, many of which are highly correlated. Recall that
the findings from the bivariate correlations between measures of social vulnerability
indicate three principle factors (Table 6). Factor 1 consists of seven measures that are
highly correlated. The other two factors are the percent vulnerable age variable and the
38
percent disability variable. I conducted multiple logistic regression analyses with
variables from each factor to identify consistent and substantial correlations between
socioeconomic status and likelihood of involvement in each wildfire mitigation activity.
4.4.1 Community Wildfire Protection Plans (CWPP)
Several socioeconomic variables were correlated with involvement in a CWPP;
the nature ofthe correlations indicates that socially vulnerable populations are less likely
to be involved in a CWPP. Table 9 lists the results of several different logistic regression
models; the beta-I coefficients indicate the direction of the correlation. Within the 1st
factor, percent single-mother households, percent poverty, percent non-white and percent
unemployed were negatively correlated with involvement in a CWPP (Table 9). CWPP
involvement was positively correlated with the percent ofthe population with a High
School Diploma and percent English speaking households. There was not a statistically
significant relationship with the median income variable. The percent ofvulnerable age
residents in the population was negatively correlated with involvement in a CWPP and
the 3rd factor, percent residents with a disability, did not have a statistically significant
relationship with CWPP involvement.
Figures 3-7 illustrate the disparity between populations with different
socioeconomic characteristics by isolating a single socioeconomic variable and depicting
the predicted likelihood of involvement in a CWPP as the average land hazard rating
increases. Three populations are depicted for each variable; a population with the mean
value, a population at plus one standard deviation and one at minus one standard
deviation for the variable. The graphs show that social vulnerability measured by poverty,
39
race, education, language or employment status is correlated with a decreased likelihood
of involvement in a CWPP.
Table 9. Logistic Regression Results for Multiple Models to Predict Involvement in a CWPP.
Predictor Variable Model
1 I 2 I 3 I 4 I 5 I 6 I 7 I 8
Single-Mother Beta - 1 0.102 -0.106
Households p value 0.004 0.000
Poverty Beta - 1 -0.020 -0.054
pvalue 0.163 0.000
Education Beta - 1 0.035 0.058
pvalue 0.016 0.000
Non-White Beta - 1 -0.047 -0.038
p value 0.000 0.000
Unemployment Beta- 1 -0.047 -0.168
lJ value 0.195 0.000
Median Income Beta - 1 -0.040 0.009
lJ value 0.000 0.106
English Beta - 1 0.058 0.167
lJ value 0.100 0.000
Percent Vulnerable Age Beta - 1 -0.040 -0.032 -0.035 -0.032 -0.038 -0.033 -0.039 -0.036
lJ value 0.000 -0.001 0.000 0.001 0.000 0.000 0.000 0.00
Disability Beta - 1 -0.024 -0.001 -0.010 -0.008 0.000 0.017 -0.007 -0.005
lJ value 0.108 0.938 0.368 0.525 0.980 0.134 0.519 0.68
Land Hazard (Avg) Beta - 1 0.428 0.460 0.487 0.410 0.398 0.411 0.454 0.459
1) value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00
Structural Density (Avg) Beta - 1 0.058 0.081 0.076 0.102 0.092 0.063 0.110 0.094
1) value 0.046 0.002 0.006 0.000 0.000 0.Q18 0.000 0.00
Constant Beta - 1 -8.118 -2.737 -2.254 -3.628 -19.191 -7.913 -2.455 -2.652
p value 0.011 0.000 0.000 0.000 0,000 0.000 0.000 0.00
Bold text indicates statistically significant correlatIOns.
~
o
Figure 3. Probability of Involvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent Poverty.
41
100%
.... 90%
c
Q) 80%E
Q) 70%>
"0 60%>
.5
.... 50%
0
~ 40%
:c 30%
I'll
.c 20%e
ll. 10%
0%
4 6 8 10 12 14 16
-+- Mean POlerty
___ Low POlerty
---.- High POlerty
Average Land Hazard
Figure 4. Probability of Involvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent Non-White Residents.
100%
.... 90%
c
Q) 80%E
Q)
70%~
0
60%>
.5
.... 50%0
~ 40%
:c 30%
I'll
.c 20%e
ll. 10%
0%
4 6 8 10 12 14 16
-+- Mean % Non-White
___ Low % Non-White
---.- High % Non-white
Average Land Hazard
42
Figure 5. Probability of Involvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent with a High School Diploma.
100%
... 90%
c
Q) 80%E
Q) 70%~
0 60%>
.5
.... 50%
0
~ 40%
.Q 30%III
.Q
20%0
...
l1. 10%
0%
4 6 8 10 12 14 16
-+- Mean % with HS Diploma1
___ High % with HS Diploma
---.- Low % with HS Diploma
Average Land Hazard
Figure 6. Probability of Involvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent English Speaking Households.
100%
... 90%
c
Q) 80%E
Q)
70%~
0
60%>
.5
.... 50%0
~ 40%
.Q 30%III
.Q
20%0
...
l1. 10%
0%
4 6 8 10 12 14 16
Average Land Hazard
-+- Mean English
___ High English
---.- Low English
43
Figure 7. Probability of Involvement in a CWPP as a Function of the Average Land
Hazard Rating and Percent Unemployment.
100%
.... 90%
c
Gl 80%E
Gl 70%~
0 60%>
.5
.... 50%0
~ 40% ,
:.c 30%III
.c 20%e
Q. 10%
0%
4 6 8 10 12 14 16
Average Land Hazard
--+- Mean Unemployment
___ Low Unemployment
----.- High Unemployment
Mapping the results of the logistic regression model highlights locations where
high biophysical risk to wildfire coincides with a low predicted likelihood of involvement
with a CWPP. Map I illustrates the probability of involvement by CBG based on a
statistical model that includes all ofthe biophysical and social variables (Modell, Table
9). Communities-at-risk that are not within a CWPP plan area are located in high fire
hazard areas in the northeast corner ofthe state on tribal lands in Navajo and Apache
Counties as well as the eastern edge ofGila County and southern tip ofApache County.
44
Map 1. CWPP Plan Areas and Likelihood of Involvement by Census Block Group.
o
e
o 0
o
"
"
"
..
"0
..
..
• ..
0 0 0
o 0
o
0
""0
"
" "..
",0
;
,; 'b.,/0 ~ "
'0
,
"~ "
"
jO .. 00
' "
0
"
.. •
•
00
e 0
" "
" , III
0 0
0
" 0
0
0
~
Tuc~on
0
..
00
o
..
..
"
..
o
"/o.
o
o
o
I r 'Hl'i~
o
0 "
:)noeni>:
0
"II
0
<Sl
"
°19
0
'l>
•
o
..
/
•o
o 0
o
8
o
..
o
o
Communilies·at-risk
Likelihood ,)( Involvement in <1 CWpp
CreOle:! 1'1 ~. O,erio MJj 7, 200f,
1\IA.D HlEi3 HARN UTV Z8NE 1:?N
45
4.4.2 State Fire Assistance (SFA) Grants
Similar to the findings for involvement in a CWPP, socioeconomic status was a
significant predictor of involvement in an SFA grant project. Table 10 lists the results of
several different logistic regression models; the beta-l coefficients indicate the direction
ofthe correlation. Within the 1st factor, poverty, the percent unemployment and the
percent non-white residents were negatively correlated with involvement in a grant
project. Median income, the percent English speaking households and the percent with a
high school diploma were positively correlated with involvement in a grant project. The
second factor, percent of residents ofvulnerable ages, was negatively correlated, but
there was no statistically significant relationship between the percent ofthe population
with a disability and the likelihood of involvement in a grant project.
Figures 8-12 illustrate the disparity between populations with different
socioeconomic characteristics by isolating a single socioeconomic variable and depicting
the predicted likelihood of involvement in an SFA grant project as the average land
hazard rating increases. The graphs show that social vulnerability measured by poverty,
race, education, language or employment status is correlated with a decreased likelihood
of involvement in an SFA grant funded project. Compared to the [mdings from CWPP
involvement, there is less of a disparity along the socioeconomic dimensions. The
greatest disparity in predicted involvement is indicated by the percent non-white residents
(Figure 9).
Table 10. Logistic Regression Results for Multiple Models to Predict Involvement in SFA Grant Funded Projects.
Model
Predictor Variable 1 I 2 I 3 I 4 I 5 I 6 I 7 I 8
Single-Mother Beta - 1 0.094 -0.035
Households IJ value 0.000 0.004
Poverty Beta - 1 -0.009 -0.023
IJ value 0.425 0.000
Education Beta -1 0.019 0.021
IJ value 0.067 0.000
Non-White Beta - 1 -0.026 -0.016
IJ value 0.000 0.000
Unemployment Beta - 1 -0.006 -0.062
IJ value 0.826 0.002
Median Income Beta -1 0.004 0.015
IJ value 0.536 0.002
English Beta - 1 -0.028 0.023
IJ value 0.116 0.049
Percent Vulnerable Age Beta - 1 -0.044 -0.048 -0.050 -0.046 -0.048 -0.050 -0.051 -0.049
IJ value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00
Disability Beta -1 -0.001 -0.007 -0.010 0.000 -0.011 0.000 -0.010 -0.009
IJ value 0.926 0.481 0.276 0.967 0.240 0.996 0.292 0.31
Land Hazard (Avg) Beta- 1 0.246 0.237 0.246 0.232 0.209 0.208 0.228 0.235
p value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00
Structural Density (Avg) Beta - 1 0.178 0.203 0.203 0.206 0.211 0.201 0.215 0.211
pvalue 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00
Constant Beta - 1 -1.114 -1.812 -1.579 -2.996 -4.165 -3.689 -1.752 -1.828
IJ value 0.432 0.000 0.000 0.000 0.000 0.000 0.000 0.00
Bold text indicates statistically significant correlatIOns.
~
01
Figure 8. Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent Poverty.
100% I I
90% J
,
-
I
c ~ --+- Mean POlA3rtyCIl 80%ECIl 70%~ ___ Low POlA3rty0>.5 ---.- High POlA3rty....
0 I ~
~ =-~:c 30%III.c 20%0...a. 10%
O%C,
I
!
4 6 8 10 12 14 16
Average Land Hazard
Figure 9. Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent Non-White Residents.
47
100% I
,
- OO%j ~cCIl 80%E
CIl 70%~
0 60%>
.5
50% +-....0
~ 40%
:c 30%
III
.c 20%
I
~
a. 10%
0% =J
4 6 8 10 12 14 16
I--+-Mean Non-Wh~
I --- Low Non-White
I ---.- High Non-Wh_it_e__
Average Land Hazard
Figure 10. Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent with a High School Diploma.
48
100% T
-
90%s::::
Q)
80%E
Q)
70%~
0
> 60%
.5
... 50%
0
~ 40%
:c 30%
III
.c 20%0
...
Q. 10%
0%
4 6 8 10 12 14 16
-+-- Mean Education
____ High EducationL-.-Low Education
Average Land Hazard
Figure 11. Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent English Speaking Households.
-+-- Mean English
100%
-
90% -s::::
Q)
E 80%
Q)
~ 70%
0
> 60%
.5
... 50%
0
~ 40%
:c 30%
III
.c 20%
0
... 10%Q.
0%
4 6 8 10 12 14 16
____ High English
-.- Low English
Average Land Hazard
49
Figure 12.Probability of Involvement in an SFA Grant as a Function of the Average
Land Hazard Rating and Percent Unemployment.
100%
.. 90%c
G> 80%E
G>
> 70%
'0
> 60%
.5
.... 50%
0
~ 40%
.c 30%
III
.c 20%e
D. 10%
0%
4 6 8 10 12 14 16
-+- Mean Unemployment
____ Low Unemployment
--.- High Unemployment
Average Land Hazard
Using the statistical model with all of the biophysical and socioeconomic
variables (Modell, Table 10), I calculated the predicted likelihood of involvement in an
SFA grant project for each CBG in the data set. Map 2 illustrates the distribution of SFA
grant projects along with these results. The map highlights areas where high biophysical
risk coincides with a low likelihood of involvement in an SFA grant project. Such areas
include communities on tribal lands in the northeastern part ofthe state in Apache and
Navajo Counties and a few communities on the eastern edge ofGila County and southern
Tip ofApache County.
50
Map 2. SFA Grant Project Locations and Likelihood of Involvement by Census Block
Group.
"
0';1'
.
• 0
•
•
•
o
0
0
0
o. e
0
•
't. 0
"
.
• 0
0
• 0 0 00
o
.O·J'·
-0
o
: \
o
o
.. / ~
, 0
o~
o
o
" 0
,:
Creole:ll:, 'l O;erio MO:I 7. 20(;6
N~D 198:< H':'R" Ur.'i ZONE 12N
"
0
0
..
"
~
..., :) ~'j kll'!'i
-
-
o
8 0 •00
•
°0
0
0
"
0
0
•
"00
nkson
0
0
"
0
PhOeni>
•
"
o
o00
I
o
o
o
o
o
o
.
.
o08
E·: lo':'!:Jcec'
fl - ,,.;,\
COl1lrnunitles·1l1·risk
likelihood of Involvem~nt - SFA Grants
• E>.(llI(je~ Census 9'cclo. :}TOUJ)3 wherE- CefiS~~ dat(lls illcorlplete or :tlE> m;l,. l~"lnd ha=~rj r::lting is very 10'11:"-1)
51
4.4.3 The Firewise Communities USA Program
Several indicators of socioeconomic status were significant predictors of
involvement in the Firewise Communities USA program. Table 11 lists the beta-l
coefficients for several logistic regression models using different combinations of the
socioeconomic variables. These results show that poverty, the percent non-white
residents, percent single-mother households and percent unemployment were all
negatively correlated with involvement in the Firewise program. The percent with a high
school diploma, percent English speaking households and median income were positively
correlated with involvement in the Firewise program.
When all of the variables from factor 1 are included in the regression only the
percent non-white residents is statistically significant indicating that race is the most
substantial predictor of involvement in the Firewise program.
Figures 13-17 illustrate the results ofthe statistical models for each of five
socioeconomic indicators. In each instance the predicted likelihood of involvement in the
Firewise program increases with an increase in land hazard rating. However, those
communities with higher social vulnerability as indicated by poverty, race, education,
language and employment status are less likely to participate in the program compared to
populations that are less socially vulnerable. Note that the difference in predicted
involvement between populations ofhigh vs. low social vulnerability is much greater for
the Firewise program than with CWPP's and the SFA grant projects.
Table 11. Logistic Regression Results for Multiple Models to Predict Involvement in the Firewise Communities USA
Program.
Model
Predictor Variable 1 2 I 3 I 4 I 5 I 6 I 7 I 8
Single-Mother Beta- 1 0.055 -0.168
Households p value 0.456 0.000
Poverty Beta - 1 -0.011 -0.071
pvalue 0.733 0.000
Education Beta - 1 0.047 0.075
p value 0.110 0.000
Non-White Beta- 1 -0.089 -0.081
pvalue 0.005 0.000
Unemployment Beta - 1 -0.107 -0.261
D value 0.218 0.000
Median Income Beta - 1 0.000 0.029
D value 0.379 0.001
English Beta- 1 -0.093 0.132
p value 0.097 0.013
Percent Vulnerable Age Beta - 1 -0.007 0.011 -0.003 0.021 0.013 0.013 0.000 0.006
D value 0.648 0.392 0.828 0.106 0.326 0.270 0.981 0.63
Disability Beta - 1 0.007 0.009 -0.001 0.013 0.000 0.023 0.003 0.008
p value 0.708 0.652 0.975 0.508 0.997 0.157 0.859 0.65
Land Hazard (Avg) Beta - 1 0.335 0.376 0.349 0.385 0.342 0.333 0.370 0.389
D value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00
Structural Density (Avg) Beta - 1 0.008 0.015 0.027 0.028 0.034 -0.005 0.045 0.025
p value 0.869 0.732 0.563 0.520 0.435 0.918 0.314 0.570
Constant Beta - 1 2.331 -5.193 -3.931 -7.902 -18.606 -12.098 -4.639 -5.049
p value 0.662 0.000 0.000 0.000 0.000 0.000 0.000 0.00
VI
tv
Figure 13. Probability of Involvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent Poverty.
53
100%
.... 90%c
Gl 80%E
Gl
> 70%
"0
> 60%
.5
.... 50%
0
~ 40%
:c 30%
III
J:I 20%~
11. 10%
0%
4 6 8 10 12 14 16
---+- Mean PO\erty
____ Low PO\erty
---.- High PO\erty
Average Land Hazard
Figure 14. Probability of Involvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent Non-White Residents.
100%
.... 90%c
Gl 80%E
Gl
~ 70%
0
> 60%
.5
.... 50%
0
~ 40%
:c 30%
III
J:I 20%0
...
10%11.
0%
-/
/
./
~
~
----
/
----
~
...~
---+- Mean Non-White
____ Low Non-White
---.- High Non-White
4 6 8 10 12 14 16
Average Land Hazard
Figure 15. Probability of Involvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent with a High School Diploma.
54
100%
.. 90%c:
CIl 80%E
CIl
70%>
'0
> 60%
.5
.... 50%
0
~ 40%
.c 30%
l'lI
.c 20%0
... 10%a..
0%
4 6 8 10 12 14 16
Average Land Hazard
-+-- Mean Education
___ High Education
---t- Low Education
Figure 16. Probability of Involvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent English Speaking Households.
100%
.. 90%c:
CIl 80%E
CIl
70%>
'0
> 60%
.5
.... 50%
0
~ 40%
.c 30%
l'lI
.c 20%0
...
10%a..
0%
4 6 8 10 12 14 16
Average Land Hazard
-+-- Mean English
___ High English
---t- Low English
Figure 17. Probability of Involvement in the Firewise Program as a Function of the
Average Land Hazard Rating and Percent Unemployment.
55
100%
-
90%c
Q)
80%E
Q)
~ 70%
0
> 60%
oS
.... 50%
0
~ 40%
:c 30%
III
.c 20%~
Q. 10% -
0%
4 6 8 10 12 14 16
---+--- Mean Unemployment
___ Low Unemployment
---.- High Unemployment
Average Land Hazard
Using the statistical model with each biophysical and socioeconomic variable in
the analysis (Modell, Table 11), I calculated the predicted likelihood ofinvolvement in
the Firewise program for each CBG in the State. Map 3 illustrates the distribution of
communities that have participated in the Firewise program along with these findings.
Most of the Firewise communities are clustered around Flagstaff, Prescott and the
northern portion ofGila County - areas that are generally more affluent and have a
greater percentage ofwhite residents that elsewhere in the state. Similar to the [mdings
with CWPP and SFA grant involvement, the northeastern portion ofthe state and the
southern tip ofNavaho County have areas ofboth high fire risk and low predicted
likelihood of involvement in the Firewise program.
56
Map 3. Firewise Recognized Communities and Likelihood of Involvement by Census
Block Group.
•
o
0"
o
o
o
••
o
o
o
"o
•
o
8
..
o
•
.
o
0,
.0
..
o
o
•
o
o
.0
o
'to
, 0
8 00
0
o
o
o
o
v
o
o
•
o 0
•
o
...,=J)~,,:on ..
... .
~
•
o
o
P_nl..
0
0 : 0 •0 ...., 0 •
'"
0
0"
·0
0
0
0
0
Fin
"
I
'b , 0
0
.\.
.
•Tu,~cll
0
0 0
..
S ::-
'"
0
.\ 0
"
, f, 0
~ ,'" .to
"
.,;.:. 0) ~,:, "'I'!'s
- -
e6c" '1o
o ,
o
o 0
• 0
o
o 0
o
e
I J
o
o
o
o
o
o
00
o
o
oR
I.~ . ~'.
lIkelillOoci 01 Involvement in the Flrewl8e Prooram
~. ',0',
Communlties-llt-nsk
h1c:l<t(,)I.
H;n
CreOled hi' R OJerl:. M.."Iy 7. 20Ll8
~IAD 1963 HARN UTl-,~ ZONE 12N
57
4.5 Summary
Collectively, the findings for each ofthe three outcome variables, involvement in
the Firewise program, CWPP plans, and SFA grant funded projects, demonstrate that
these wildfire mitigation activities are focused on areas ofhigh biophysical risk to
wildfire. The land hazard rating is positively correlated with each activity and the
predictive effect is significant in every statistical model irrespective of socioeconomic
characteristics. The structural density variable is also positively correlated with
involvement with CWPP's and SFA grant projects but not Firewise involvement, though
the correlation is less than that of the land hazard rating.
In terms of social vulnerability, those populations that are typically disadvantaged
and marginalized are less likely to be involved in these wildfire mitigation activities.
Regardless ofwhich measure of social vulnerability is used from the suite of indicators
included in this research, there is a significant disparity between the likelihood of
involvement and level of socioeconomic status. The disparity is greatest in the Firewise
program, but evident in all three outcome variables. For each ofthe three wildfire
mitigation activities, the percent non-white residents is the most consistent and
substantial predictor of involvement. Communities in Arizona with a high percentage of
non-white residents are primarily Native American Communities living on Tribal Lands.
58
CHAPTER V
DISCUSSION
As competition for federal resources to mitigate wildfire risk becomes
increasingly acute, it is all the more important to allocate those resources efficiently and
equitably. The delivery ofresources to communities-at-risk to wildfire should be
strategic, providing assistance where it will result in the greatest marginal improvement
in disaster resilience. The equitable approach would seek to equalize the burden of risk
across individuals and communities. Disaster research shows that communities with low
socioeconomic status bear a disproportionately large risk burden given the same
biophysical risk factors as a more aftluent community. Thus the concept of social
vulnerability is an important factor in wildfire risk management.
The results ofthis study demonstrate that in Arizona, biophysical wildfire risk
factors are significant predictors ofcommunity involvement in wildfire mitigation
activities. However, the findings also demonstrate that traditionally disadvantaged and
marginalized segments ofthe population are less likely to be involved in mitigation
efforts throughout the state. Identifying these at-risk populations and understanding the
underlying mechanisms that create the disparity is an important issue for planners, policy
makers, community leaders, residents and others interested in the equitable and efficient
use of federal resources.
59
Incorporating social vulnerability indicators in wildfire planning is a pressing
issue because ofthe magnitude ofthe threat, but also because of the approach to risk
management currently promoted by federal policy. That policy emphasizes proactive
measures to reduce risk and empower communities to engage in wildfire planning and
implementation. Despite efforts to provide resources, tools and technical assistance, this
study demonstrates that socially vulnerable communities are less likely to be involved in
wildfire mitigation efforts.
5.1 Prioritizing Socially Vulnerable Populations
This study used nine measures of social vulnerability based upon a review of the
literature on natural disasters and wildfires. Results indicate that many ofthose measures
were consistent and substantial predictors of involvement in federal wildfire mitigation
program efforts - in general, socially vulnerable populations were less likely to be
involved in wildfire mitigation efforts. Based on this fmding, wildfire managers should
use socioeconomic indicators to identify and prioritize socially vulnerable populations in
an effort to increase the level of involvement in these communities. But which indicators
should be used? Poverty, race, education, language and employment status were
significant predictors across each ofthe mitigation activities. One possibility would be to
use an index ofeach measure, but since these measures are also correlated with one
another, selecting a single measure from the suite would likely be as effective and more
efficient. As a practical matter, using poverty has advantages including precedence as an
60
eligibility criterion in other government programs and availability ofdata. Poverty is also
more socially acceptable than race and language as eligibility criteria.
A social vulnerability measure, such as poverty, could be integrated into wildfire
protection programs and plans in a variety ofways. First, poverty could be included in
the comprehensive wildfire risk assessments along with the biophysical factors such as
weather, forest fuels and topography to identify priority areas for fuels reduction work.
The same risk assessment could be used to identify socially vulnerable communities and
include goals in the CWPP to focus efforts on reducing structural vulnerability in those
communities. At a larger scale, the Arizona State Lands Dept. Forestry Division could
target socially vulnerable communities within the state and strive to assist those areas in
developing CWPP's, pursuing grant applications for wildfire reduction activities and
participating in the Firewise Communities USA program. At a national level, if these
finders transfer to other contexts, the federal government should use poverty as an
indicator of social vulnerability to identify regions where at-risk communities need
additional assistance.
Similarly, a social vulnerability criterion, such as poverty, could be included in
the State Fire Assistance (SFA) grant application to help to focus resources on these at-
risk communities or the 50:50 match requirement could be lowered if it proves to be a
barrier to participation from socially vulnerable communities.
Ofthe three wildfire mitigation activities, integrating a measure of social
vulnerability into the Firewise program would probably be the most difficult to achieve
since the participation in the program is initiated by the community. Without knowing
why socially vulnerable communities in Arizona are less likely to get involved it is
difficult to prescribe effective solutions.
Map 4 illustrates the distribution of different wildfire mitigation activities, the
percent poverty by Census Block Group, and areas with a high land hazard rating. High
poverty communities that are also at high risk to wildfire are located in the northeastern
comer ofthe state in Apache and Navajo Counties, the eastern portion ofGila County,
and a few areas in the northern portion ofCoconino County. This map highlights those
areas were additional research could help explain why socially vulnerable communities
are less likely to be involved in federal program efforts to reduce wildfIre risk.
61
62
Map 4. Wildfire Mitigation Activities and Percent Poverty by Census Block Group.
'0
o
o
. ..
o
{)
0 'l>
Phoenl,
"
D 0
0
0
q, 11·11 :11 h
0
AJ
:II:~
-uc.son
o
o
o ,9
0% J~esc~~t 0000
~j!i! <b 0
• r 0
o
.... 0o
o
.. F re'h:~e- COO,ml-rlty
C:"Vpp PI:\11 A'tG.
Porcelll Poverty
o
00
o
o
o
rJc~l~rJI~
I-'i~h
COllllllunl tles-at.risk
L,)rj,...;.n"jR.:;t.'"JH",· -1(1
1·1~%
IE .,H%
_ ?'·~31~b
_""',J-F>
CreoTe:llt, R O;,rlo Moy 7,2008
NAD 19S3 HARN UT.'/ Z:JNE nN
63
5.2 Next Steps
Additional research is needed to replicate the methods from this study in other
states to see if similar disparities exist in other contexts. These studies should be paired
with qualitative research to identify underlying causes and solutions. It would be
particularly useful to conduct case studies of socially vulnerable communities that are
involved in wildfire mItigation efforts and disseminate those findings amongst socially
vulnerable communities and wildfire management practitioners.
With regard to participation in the Firewise program, I suspect that dispersed
settlement patterns in rural areas are less conducive to the type of community organizing
and grass roots projects that the program is geared towards; this could explain some of
the [mdings from this study. Native American communities and other traditionally
marginalized populations may also be less inclined to participate in government
sponsored programs.
If further study indicates that a lack of awareness about the Firewise program is
an issue, the Arizona State Lands Dept. Forestry Division, which coordinates the
program, could target outreach efforts to socially vulnerable communities. Another
barrier to participation in the Firewise program might be the requirement that the
community demonstrates an annual expenditure of$2 per capita on wildfire mitigation
activities. Ifso, financial assistance or a waiver ofthe requirement could help these
communities get involved in the program and perhaps over time build their capacity to
meet all the requirements.
64
5.2.1 Involving Socially Vulnerable Communities in Planning and Implementation
Additional research to clarify the causes and the solutions to the Lack of social
equity in wildfire management will take time. Including residents and representatives
from socially vulnerable communities in the CWPP process could improve current
wildfire planning and implementation. It is important to involve vulnerable populations
and those who understand their needs in developing strategies that are appropriate and
relevant (Rhodes and Reinholtd, 1998). The singular focus on vulnerabilities, however,
overlooks potential capacities within populations that emergency managers could
capitalize on to develop disaster resilience (Buckle et aI., 2000).
I suspect that social capital and effective community leadership is a critical
ingredient to mobilizing human resources in so called "low-capacity" communities. I also
am interested in the role that intermediaries play in engaging communities in these
efforts. Public lands managers, researchers, emergency management staff and others
involved in wildfire mitigation have an opportunity to build social bonds that bridge
boundaries of race, class, organizational affiliation and political persuasion. I suspect that
these relationships encourage the trust and reciprocity necessary for local actors to
capitalize on outside resources. Furthermore, these bonds lead to more effective wildfire
response and recovery (Carroll et al., 2005).
Although natural resource managers, foresters, and forest fuels specialists are well
trained in delivering technical solutions such as thinning fuels, community involvement
requires experience and expertise in education, outreach, and social mobilization (Brooks
et aI., 2006). Wildfire management practitioners may be building that experience, but a
continued effort is needed. Towards that end, state and local agencies need support and
resources from the federal government to continue to promote effective community
involvement in wildfIre mitigation efforts.
5.2.2 Community Capacity
The concept of community capacity is another avenue of research that should be
pursued. Researchers from many disciplines including public health, economic
development, natural resource conservation and disaster management have explored the
concept of community capacity. Typically, the concept is composed of several
dimensions that describe a community's assets and abilities such as social, cultural,
political and economic capital. Although there is little consensus on a precise defmition
ofthe concept, in the most general sense community capacity is the ability to respond to
challenges and effect change that captures opportunities and fulfIlls the needs of
community members (Donoghue and Sturtevant, 2007). In the wildfIre context,
community capacity can be defIned as the ability of a community to organize and
mobilize resources to prepare for, respond to and recover from wildfIre (Evans et al.,
2007).
Despite much interest in the topic, previous research on community capacity has
focused on clarifying defInitions, but there has been little work to validate potential
measures against specifIc outcomes and incorporate valid measures into planning and
program evaluation (Donoghue and Sturtevant, 2007). The lack oftools to evaluate
community capacity to engage in wildfire mitigation activities can hamper project goals
65
66
especially if those goals were drafted prior to learning about community's capacity and
history (Brooks et ai., 2006).
Although this study did not directly evaluate community capacity, the finding that
socioeconomic status is a predictor of involvement in wildfire mitigation activities
suggests a relationship between these factors and community capacity. However,
socioeconomic measures that represent levels ofphysical and human capital don't
necessarily correlate with the community capacity for management and decision making
(Buckland and Rahman, 1999), nor the quality of social networks and leadership which
comprise social capital. For example, regional community assessment efforts during the
1990's that incorporated measures of socioeconomic status and social capital found
positive correlations with community capacity. But high socioeconomic status did not
always predict high social capital and some communities rated highly in social capital
despite low scores on socioeconomic status (Donoghue and Sturtevant, 2007).
In short, community capacity is a complex topic deserving ofadditional research
to clarifY linkages between socioeconomics, social capital and capacity. The biophysical
factors and dynamics ofwildfire are also complex, yet CWPP's consistently include a
comprehensive wildfire risk assessment ofthese variables. Similarly, planners should
prioritize assessments of social factors to identifY and support community deficiencies
and build upon community assets and strengths.
In summary, findings from this research and other efforts to investigate social
vulnerability in a wildfire context support the following recommendations:
• Use indicators of social vulnerability in comprehensive wildfire risk assessments;
• Modify grant criteria to reduce match requirements for poor communities;
• Prioritize grants and technical assistance for socially vulnerable communities;
• Research and disseminate findings from case studies where socially vulnerable
communities successfully participate in wildfire mitigation activities; and
• Ensure that socially vulnerable populations are included in CWPP planning and
implementation.
67
68
BIBLIOGRAPHY
Absher, J., Vaske, J.1., 2007. Understanding Obstacles to Firewise Implementation. In the
Proceedings of the Human Dimensions of Wildland Fire Conference.
International Association of Wildland Fire, Fort Collins, CO.
Allen, G.M., Gould, E.M., 1986. Complexity, Wickedness, and Public Forests. Journal of
Forestry 84, 20-23.
Aptekar, L., Boore, J.A., 1990. The emotional effects ofdisaster on children: a review of
the literature. International Journal of Mental Health 19, 77-90.
Blaikie, P., Cannon, T., Davis, 1., Wisner, B., 1994. At Risk: Natural Hazards, People's
Vulnerability, and Disasters. Routledge, London.
Bolin, R, 1986. Disaster impact and recovery: A comparison of black and white victims.
International Journal of Mass Emergencies and Disasters 4, 35-50.
Bolin, R, Stanford, L., 1998. The Northridge Earthquake: Community-based Approaches
to Uoment Recovery Needs. Disasters 22,21-38.
Bridger, J.C., LulofI, A.E., 1999. Toward an interactional approach to sustainable
community development. Journal ofRural Studies 15,377-387.
Brooks, J.J., Brenkert, H., J.E., S., Champ, J.G., T., S., Williams, D.R, 2006. Integrating
Social Science Into Forestry in the WildlandlUrban Interface. Fire Management
Today 66, 35-43.
Buckland, J., Rahman, M., 1999. Community-based Disaster Management during the
1997 Red River Flood in Canada. Disasters 23, 174-191.
Buckle, P., Mars, G., Smale, S., 2000. New approaches to assessing vulnerability and
resilience. The Australian Journal ofEmergency Management 15, 18-14.
Burchfield, J., 2007. Community Impacts ofLarge Wildfire Events: Consequences of
Actions After the Fire. In: Daniel, T.e., Carroll, M.S., Moseley, e., Raish, C.
(Eds.), People, Fire, and Forests. Oregon State University Press, Corvallis, OR
69
Bums, S., Porter-Norton, Mosher, M., Richard, T., 2003. People and Fire in Western
Colorado. Office of Community Services - Fort Lewis College.
Carroll, M.S., Blatner, K.A., Cohn, P.J., Morgan, T., 2007. Managing Fire Danger in the
Forests of the US Inland Northwest: A Classic "Wicked Problem" in Public Land
Policy. Journal ofForestry 105, 239-234.
Carroll, M.S., Cohn, PJ., 2003. Inductive case study for Rodeo-Chediski apache-
Sitgreaves National Forests. RFQ R3-01-02-60.
Carroll, M.S., Cohn, PJ., Seesholtz, D.N., Higgins, L.L., 2005. Fire as a galvanizing and
fragmenting influence on communities: The case of the Rodeo-Chediski fire.
Society & Natural Resources 18,301-320.
Case, P., Banks, B., Butler, E., Gosnell, R, 2000. Assessing Potential Wildfire Effects on
People. Journal of Sustainable Forestry 11, 159-175.
Collins, T., 2005. Households, forests, and fire hazard vulnerability in the American
West: A case study ofa California community. Environmental Hazards 6, 23-37.
Cutter, S.L., 1995. The Forgotten Casualties: Women, Children, and Environmental
Change. Global Environmental Change 5, 181-194.
Cutter, S.L., Boruff, BJ., Shirley, W.L., 2003. Social Vulnerability to Environmental
Hazards. Social Science Quarterly 84, 242-261.
Cutter, S.L., Mitchell, J.T., Scott, M.S., 2000. Revealing the Vulnerability ofPeople and
Places: A Case Study ofGeorgetown County, South Carolina. Annals of the
Association of American Geographers 90, 713-737.
Dale, V.H., Joyce, L.A., McNulty, S., Neilson, RP., Ayres, M.P., Flannigan, M.D.,
Hanson, P.J., Irland, L.c., Lugo, A.E., Peterson, C.J., SimberlofI, D., Swanson,
FJ., Stocks, B.J., Wotton, B.M., 2001. Climate Change and Forest Disturbances.
Bioscience 51, 723.
Daniel, T.C., Carroll, M., c., M., Raish, C. (Eds.), 2007. People, Fire and Forests. Oregon
State University Press, Corvallis, OR
Dellasala, D.A., Williams, J.E., Williams, C.D., Franklin, J.F., 2004. Beyond Smoke and
Mirrors: a Synthesis ofFire Policy and Science. Conservation Biology 18,976-
986.
Dombeck, M.P., Williams, J.E., Wood, C.A., 2004. Wildfire Policy and Public Lands:
Integrating Scientific Understanding with Social Concerns across Landscapes.
Conservation Biology 18, 883-889.
70
Donoghue, E., Sturtevant, V., 2007. Social Science Constructs in Ecosystem
Assessments: Revisiting Community Capacity and Community Resiliency.
Society and Natural Resources 20,899-912.
Evans, A., DeBonis, M., Krasilovsky, E., Melton, M., 2007. Measuring Community
Capacity to Resist and Repair After Wildfires. The Forest Guild.
Flint, e.G., Luloff, A.E., 2005. Natural resource-based communities, risk, and disaster:
An intersection of theories. Society & Natural Resources 18, 399-412.
Fothergill, A., 1996. Gender, risk, and disaster. International Journal of Mass
Emergencies and Disasters 14,33-56.
Fothergill, A., Peek, L., 2004. Poverty and Disasters in the United States: A Review of
Recent Sociological Findings. Natural Hazards 32,89-110.
Haque, C.B., Etkin, D., 2007. People and community as constituent parts ofhazards: the
significance of societal dimensions in hazards analysis. Natural Hazards 41, 271-
282.
Heinz Center for Science, E., and the Environment, 2000. The Hidden Costs ofCoastal
Hazards. Island Press, Covello, CA.
Hessburg, P.F., Agee, J.K., Franklin, J.F., 2005. Dry forests and wildland fires ofthe
inland Northwest USA: Contrasting the landscape ecology ofthe pre-settlement
and modem eras. Forest Ecology and Management 211, 117-139.
Jakes, P., 2007. Impacts of Wildland Fire on Communities. Social Issues Fact Sheet,
USDA Forest Service, 19.
Jakes, P., Burns, S., Antony, e., Saeli, E., Nelson, K., Brummel, R., Grayseck, S.,
Sturtevant, V., Williams, D., 2007a. Critical Elements in the Development and
Implementation ofCommunity Wildfire Protection Plans (CWPPs). In. USDA
Forest Service Proceedings RMRS-P-46CD.
Jakes, P., Kruger, L., Monroe, M., Nelson, K., Sturtevant, V., 2007b. Improving wildfire
preparedness: Lessons from communities across the US. Human Ecology Review
14, 188-197.
King, D., 2001. Uses and Limitations of Socioeconomic Indicators ofCommunity
Vulnerability to Natural Hazards: Data and Disasters in Northern Australia.
Natural Hazards 24, 147-156.
71
Kumagai, Y., Carroll, M., Cohn, P., 2004a. Coping with interface wildftre as a human
event: Lessons from the disaster/hazards literature. Journal ofForestry. 102,28-
32.
Kumagai, Y., Daniels, S., Carroll, M., Bliss, J., Edwards, J., 2004b. Causal reasoning
processes of people affected by wildftre: Implications for agency-community
interactions and communication strategies. Western Journal of Applied Forestry
19, 184-194.
Luloff, A.E., Finley, J., Flint, C.G., 2003. Community-based forestry in America: A
framework and analysis. In, Proceedings ofthe International Symposium on
Society and Resource Management, Maddalena, Sardinia, Italy.
Lynn, K., Gerlitz, W., 2005. Mapping the Relationship between Wildftre and Poverty.
Resource Innovations and the National Network ofForest Practitioners.
McCaffrey, S., 2004. Thinking ofWildftre as a Natural Hazard. Society & Natural
Resources 17,509 - 516.
McKenzie, D., Gedalof, Z.E., Peterson, D.L., Mote, P., 2004. Climatic Change, WildfIre,
and Conservation. Conservation Biology 18, 890-902.
Mendez, S.R., Carroll, M.S., Blatner, K.A., Findley, A.J., Walker, G.B., Daniels, S.E.,
2003. Smoke on the hill: A comparative study ofwildfrre and two communities.
Western Journal ofApplied Forestry 18,60-70.
Mercer, D.E., Prestemon, J.P., 2005. Comparing production function models for wildftre
risk analysis in the wildland-urban interface. Forest Policy and Economics 7, 782-
795.
Mileti, D., 1999. Disasters by Design: A Reassessment ofNatural Hazards in the United
States. Joseph Henry Press, Washington D.C.
Morrow, B.H., 1999. Identifying and Mapping Community Vulnerability. Disasters 23,
1-18.
Neal, D.M., Phillips, B.D., 1995. Effective Emergency Management: Reconsidering the
Bureaucratic Approach. Disasters 19,327-337.
Newman, C., 2004. Wildftre Protection: Community WildfIre Protection Plans from Four
Angles. Journal ofForestry 102,4-7.
Ngo, E.B., 2001. When Disasters and Age Collide: Reviewing Vulnerability ofthe
Elderly. Natural Hazards Review 2,80-89.
72
Niemi, L., Lee, K., 2001. Wildfire and Poverty: An Overview of the Interactions Among
Wildfrres, Fire-Related Programs, and Poverty in the Western States. The Center
for Watershed Health and Community Health.
Peacock, W., Morrow, H.B., H., G. (Eds.), 1997. Hurricane Andrew and the Reshaping of
Miami: Ethnicity, Gender, and the Socio-Political Ecology ofDisasters.
University Press ofFlorida, Gainesville, FL.
Puente, S., 1999. Social Vulnerability to Disaster in Mexico City. In: Mitchell, J.K. (Ed.),
Crucibles ofHazard: Mega-Cities and Disasters in Transition. United Nations
University Press, Tokyo, Japan.
Pulido, L., 2000. Rethinking Environmental Racism: White Privilege and Urban
Development in Southern California. Annals of the Association ofAmerican
Geographers 90, 12-40.
Rhodes, A., Reinholtd, S., 1998. Beyond Technology: A Holistic Approach to Reducing
Residential Fire Fatalities. Australian Journal ofEmergency Management 13,39-
44.
Steelman, T., Ginger, K., Devona, B., 2004. Federal and State Influence on Community
Responses to Wildfire Threats: Arizona, Colorado, and New Mexico. Journal of
Forestry 102,21-27.
Steelman, T., Kunkel, G., 2004. Effective Community Responses to Wildfire Threats:
Lessons From New Mexico. Society and Natural Resources 17, 679-699.
Steelman, T.A., Burke, e.A., 2007. Is Wildfire Policy in the United States Sustainable?
Journal of Forestry 105,67-72.
Theobald, D.M., Romme, W.H., 2007. Expansion of the US wildland-urban interface.
Landscape and Urban Planning 83, 340-354.
Tobin, G.A., Ollenburger, J.C., 1993. Natural Hazards and the Elderly. University of
Colorado, Natural Hazards Research and Applications Information Center,
Boulder, CO.
WGA, 2002. A Collaborative Approach for Reducing Wildland Fire Risks to
Communities and the Environment. In, Conference Report for the Fiscal Year
2001 Interior and Related Agencies Appropriations Act.
Wise, C., Yoder, A.J., 2007. Policy and Institutional Arrangements in Federal Wildland
Fire Mitigation. In: Daniel, T.e., Carroll, M.S., Moseley, C., Raish, C. (Eds.),
People, Fire, and Forests. Oregon State University Press, Corvallis, OR.