Risk Factors and Wildfire Mitigation Planning by Public Utilities in Washington State

Abstract

Some of the most catastrophic fire events that have occurred in the western US in recent decades, such as the 2018 Camp Fire in California, were ignited by electric utility infrastructure. As wildfires and fire seasons intensify across the western United States, policymakers and utilities alike are working to mitigate the risk of wildfire as it relates to utility infrastructure. We pose the following research question: Is there an association between risk factors such as wildfire hazard potential and social vulnerability, and the inclusion of various strategies in mitigation planning by public or cooperative electric utilities in Washington, such as PSPS provisions and non-expulsion fuse installation? By applying statistical tools including t-tests and logistic regression modeling to test these potential associations, our analysis reveals statistically significant relationships between risk factors and the inclusion of specific wildfire mitigation strategies. We find that the inclusion of PSPS provisions in mitigation planning is significantly and nonlinearly associated with wildfire hazard potential, while social and socioeconomic vulnerability in the utility service area are negatively associated. Additionally, the installation of non-expulsion fuses is negatively associated with socioeconomic vulnerability in service populations. Overall, understanding the factors associated with wildfire mitigation planning can assist policymakers and state agencies in the prioritization of resources and practical support for utilities that may have limited capacity to mitigate wildfire risk.

1. Introduction

Throughout the western United States, catastrophic wildfires and fire seasons have lengthened and intensified over recent decades [1,2]. Additionally, a portion of these increases can be attributed to anthropogenic climate change [3,4]. Simultaneous wildfire incidents are also growing more common across the western United States, straining the availability of resources for wildfire response [1]. While wildfires result from both natural and man-made causes, the majority of ignitions in recent decades were human-caused, a portion of which can be attributed to power grid infrastructure [5]. Transmission lines can ignite wildfires in multiple ways, including through nearby vegetation, power arcs, and molten particles [6]. High wind conditions, in particular, are an important risk factor in power system wildfire events [7]. Although they only make up a small portion of annual wildfire ignitions, powerline-related fires often lead to more severe or intense wildfires because these ignitions typically occur when fire danger is relatively high [8]. These ignitions can lead to power outages, resulting either from damage to infrastructure or the pre-emptive de-energization of power lines [7]. Such de-energization events, often caused by public safety power shutoffs (PSPS), have a greater impact on socially vulnerable populations in some contexts [9]. Electric utilities across the western United States have adopted strategies for mitigating risk through wildfire mitigation plans (WMPs). However, smaller or more rural utilities may have less capacity to actively mitigate risk. This analysis takes a convergence research approach to developing an information base for decision support around wildfire risk mitigation [10]. In our analysis, we test the hypothesis that wildfire risk and social vulnerability are associated with wildfire mitigation planning for Washington public utilities, specifically the incorporation of PSPS provisions and non-expulsion fuse installation.

We seek to answer the following research question: Is there an association between risk factors such as wildfire hazard potential and social vulnerability and the inclusion of various strategies in mitigation planning by public or cooperative electric utilities in Washington, such as PSPS provisions and non-expulsion fuse installation? To answer this question, we test this association using statistical methods of two sample t-tests and logistic regression. An accompanying 2 × 2 matrix further expands on the relationship between wildfire hazard potential and whether public utilities have developed PSPS provisions.

1.1. Background

In 2018, the Camp Fire, one of the deadliest and most destructive wildfires in the history of California, wrought devastation across a wide swath of the state. The California Department of Forestry and Fire Protection (Cal Fire) held Pacific Gas & Energy accountable for damages after determining that a transmission ignition related to their infrastructure was the cause [11,12]. Many electric utilities, either of their own volition or in response to direction from their respective state governments, have begun to develop wildfire mitigation plans to address the risk of damage to their grid infrastructure, as well as the risk associated with utility-caused wildfires. In 2021, the Oregon legislature passed Senate Bill 762, comprehensive legislation that requested that all Oregon electric utilities submit wildfire mitigation plans [13]. Washington State followed suit in 2021, mandating that all investor-owned utilities create individual wildfire mitigation plans [14].

In Washington, there are four primary kinds of electric utilities. Investor-owned utilities (IOUs) are private companies and represent the four largest utilities in Washington. Municipal utilities are owned and run by local governments, and generally their service areas are confined to urban or suburban areas. IOUs generally have more resources and greater capacity to mitigate wildfire risk, while municipal utilities often face limited wildfire risk due to their urban/suburban locations. By contrast, public utility districts (PUDs) are community-owned and locally regulated utilities [15]. Similarly, cooperative or mutual electric companies serve customers on a non-profit basis with voluntary and open membership [16]. The result is an interesting patchwork across the state of utilities with widely ranging governance structures, regulations, customer bases, and resources.

In 2021, the Washington State mandate for WMPs only applied to IOUs. It was not until July of 2023 that the Washington House of Representatives passed legislation requiring all utilities to submit a wildfire mitigation plan to the Washington Department of Natural Resources (WA DNR) by the end of October 2024 [17]. WA DNR also provided a template to help utilities organize their respective plans [18]. While every utility in Washington was required to submit a wildfire mitigation plan, each approached the process differently, with different levels of resource support. PUDs and cooperative electric utilities are our focus, given that most are located in rural communities with fewer resources for wildfire mitigation. The 2023 legislation opened up a new source of information about wildfire mitigation planning among publicly owned utilities, which our analysis leverages.

1.2. Literature Review

Multiple studies have investigated the relationship between wildfire risk and social vulnerability generally [9,19,20,21,22]. Studies have shown that wildfire risk and concurrent housing density have both contributed to inequitable outcomes for socially vulnerable communities [19,21], while Davies et al. (2018) evaluate the relationship between wildfire and social vulnerability factors, highlighting the disproportionate impact of wildfires on disadvantaged communities [20]. Xie and Meng (2025) conduct a social vulnerability analysis of planned PSPS in California, finding that incorporating social vulnerability could lead to more equitable processes [9]. Pollack et al. (2025) assess the practicality of applying social vulnerability metrics to decision-making about PSPS in Texas, finding that current policy frameworks may fail to protect certain communities facing high wildfire risk [22].

This analysis contributes to a growing body of research on the risk posed by wildfire to electric utilities and potential mitigation strategies [23,24,25,26]. Muhs et al. (2020) categorize wildfire mitigation techniques into categories of fault or failure prevention, arc-ignition prevention, and wildfire impact mitigation, while Panossian and Elgindy (2023) offer a literature review of methods for quantifying and managing wildfire risks with tools such as wildfire hazard potential (WHP) [23,25].

Other research on the risk posed by wildfire to electric utilities has largely focused on the individual components of the wildfire mitigation plans of IOUs in California [27,28,29]. Vazquez et al. (2022) compare the strategies of California IOU wildfire mitigation plans, identifying key components including grid design and system hardening, vegetation management, asset inspections, situational awareness and forecasting, PSPS, and operational response [27]. For each of these components, the authors present examples from the existing literature to support utility managers and future research into wildfire mitigation planning. Mitchell (2023) develops a methodology for identifying the most cost-effective strategies for high-risk circuits in California [28]. The research specifically focused on the wildfire mitigation efforts of rural electric utilities that were at an early stage of development, e.g., examining wildfire risk reduction through avian protection plans [29].

The provisions and protocols for PSPS as a mitigation strategy have received focused attention in the literature. PSPS have been established as a critical wildfire mitigation measure, as they allow a utility to de-energize a portion of the grid to reduce risk of ignition during periods of high risk [30]. While de-energization does minimize ignition risk, PSPS also impose costs on customers due to extended power outages [31]. In Northern California alone, an annual 1.6 million person-days of de-energization are estimated based on recent historical climate conditions and locally stated PSPS protocols [32]. To mitigate the associated risks, recent research has investigated methods of optimizing PSPS operations [33,34]. Grid hardening measures provide an alternate approach to reducing wildfire risk, but must be balanced against the cost of implementation [23]. An important example of grid hardening is the installation of non-expulsion fuses on overhead power lines, a replacement for standard expulsion fuses that introduce ignition risk as a result of arcing electricity onto nearby vegetation [27,35,36]. However, to our knowledge, there has been no prior research comparing strategies in public utility wildfire mitigation plans based on wildfire risk and social vulnerability.

2. Materials and Methods

2.1. Data

Our analysis is based on data drawn from the wildfire mitigation plans submitted by 32 public utilities to the WA DNR. For this set of PUDs and cooperatives, electric utility retail service territories were pulled from the Homeland Infrastructure Foundation Level Database [37]. Using ArcGIS Pro 3.1, this dataset was filtered to include utilities whose dominant service area falls within Washington’s state boundaries. As described above, investor-owned utilities (IOUs) and municipal utilities are outside the scope of this analysis due to differences in operations, risk, and scale.

The outcomes of interest in this analysis are (i) the potential inclusion of a PSPS provision and (ii) the installation or planned installation of non-expulsion fuses by public utility districts and cooperative utilities as components of their respective wildfire mitigation plans. For each plan, we searched for keywords and sections dedicated to each strategy. After carefully reviewing each utility’s mitigation plan, we used binary variables to represent whether or not the utilities incorporated PSPS provisions or non-expulsion fuse installation (

Table A1. Summary Statistics.

Variable	Mean	s.d.	Median	Min	Max
WHP 84th Percentile	0.0972	0.1394	0.0182	0	0.4224
Social Vulnerability Index	0.5648	0.1698	0.5319	0.15	0.9109
Socioeconomic Percentile	0.5088	0.1454	0.5026	0.1693	0.8227
PSPS Provision	Count
0	20
1	12
Total	32
Non-Expulsion Fuses	Count
0	14
1	18
Total	32

).

We represent the level of wildfire risk faced by utilities using the wildfire hazard potential (WHP), an index created by researchers at the Rocky Mountain Research Station to inform prioritization of fuel mitigation needs (see Figure 1) [36]. The WHP is currently used by many Washington utility wildfire mitigation plans to estimate the level of wildfire risk that their infrastructure faces [18]. WHP is available in a 30 m raster, integrating burn probability and fire intensity datasets generated by the large fire simulation system (FSim). Vegetation and fuel data from LANDFIRE 2020 are the primary inputs into FSim. As such, wildfire risk derived from WHP is accurate to the 2020 landscape for each utility. FSim was also calibrated against historic wildfire occurrences to improve accuracy [38]. Using ArcGIS Pro, we reclassified WHP from raw index values to five classes of risk [39]. These categories include Very Low (<44th percentile), Low (44th percentile–67th percentile), Moderate (67th percentile–84th percentile), High (84th percentile–95th percentile), and Very High (>95th percentile). The class breaks were determined using raw index values for Washington State. For the utility service area shapefile, we used the zonal histogram tool to find the proportion of total pixels that fell into each WHP class by utility service area, shown visually in Figure 1. We calculated the percentage of each service area above the 84th percentile of WHP, roughly representing the wildfire risk to infrastructure that each utility faces and refer to this as high/very high WHP (Figure 2).

We included measures of social and socioeconomic vulnerability to better characterize the geographic area that each utility serves. As noted, vulnerable areas are more susceptible to the de-energization risks of PSPS, and so vulnerability metrics are important considerations for this analysis [9]. We used the CDC/ATSDR Social Vulnerability Index (SVI) from 2022 to represent the level of vulnerability in the population served by each utility [40,41]. SVI uses 16 U.S. Census variables from the 5-year American Community Survey across categories including socioeconomic status, household characteristics, racial and ethnic minority status, and housing type and transportation to help determine which communities may need support before, during, and after a disaster. Separately we considered the subset of metrics that constitute the socioeconomic portion of SVI, including a given census tract’s poverty rate, unemployment rate, per capita income, and percentage of population without a high school diploma or health insurance. These data are at the census tract level, which does not fit neatly into the utility service areas [40]. To remedy this, we generated an area-weighted average where the SVI or socioeconomic score for each census tract is weighted by the percentage of utility service area it represents. Research suggests that SVI is highly dependent on geographic context, and thus aggregation does risk oversimplification in assessing vulnerability [42]. Despite this, social vulnerability indices are often used for guiding infrastructure investment, and for this reason we retained them [22]. The distribution of both variables is displayed in Figure 2. Summary statistics for all variables are included in the Appendix A (Table A1).

2.2. Methods

In this analysis, we applied statistical approaches to assess whether there is an association between the inclusion of PSPS protocol and/or non-expulsion fuse installation in utility WMPs and their WHP or their social vulnerability. We used two sample t-tests to determine if there is a statistically significant difference in wildfire risk between utilities that implemented specific wildfire mitigation strategies and those that did not. This analysis assumes that the public utilities and cooperatives in our sample are representative of a broader population.

In addition to initial t-tests, we created four logistic regression models to further examine the relationship between risk factors related to wildfire and social vulnerability and the adoption of PSPS protocols or non-expulsion fuses. Model I assesses the relationship between PSPS protocol and three independent variables, i.e., the percentage of service area at a high/very high WHP, the square of percentage of service area at a high/very high WHP (representing a potential nonlinear effect), and the area-weighted SVI. Model II modifies model I, replacing area-weighted SVI with socioeconomic percentile. Model III assesses the relationship between whether a utility has installed or plans to install non-expulsion fuses with the predictors high/very high WHP, and the area-weighted SVI. Model IV modifies model III, replacing area-weighted SVI with socioeconomic percentile. Models III and IV were separately tested for nonlinear associations.

Additionally, we sorted and displayed the utilities into a 2 × 2 matrix for the visual inspection and consideration of risk and mitigation together. We categorized utilities based on whether they have adopted a PSPS protocol as one dimension of the matrix, and the percentage of their service area that falls within a high/very high WHP as the other dimension. Specifically, utilities with more than 10% of their service area above the 84th percentile WHP were classified as higher WHP, and utilities below this were classified as lower WHP. We acknowledge that using 10% of the service area as our threshold cut-off influences which utilities fall into the higher WHP category. We encourage readers to refer to Figure 1 for more granular WHP information and further insight.

3. Results

We present the results of six t-tests and four logistic regression models to assess relationships between the inclusion of PSPS protocols and non-expulsion fuses in WMPs and wildfire risk and social vulnerability.

3.1. T-Tests

Utilities with a PSPS provision in their WMP are associated with a statistically significantly higher exposure to wildfire risk, as indicated by the WHP (

Table 1. Two sample t-tests comparing wildfire hazard potential (WHP), social vulnerability index (SVI), and socioeconomic percentile determined via the inclusion of PSPS provisions and non-expulsion fuses into utility wildfire mitigation plans.

	PSPS Provision (Mean)	No PSPS Provision (Mean)	Diff	t-Stat	p-Value
High/Very High WHP	0.16	0.06	0.10	−2.14	0.04
Social Vulnerability Index	0.51	0.59	−0.08	1.37	0.18
Socioeconomic Percentile	0.47	0.53	−0.07	1.29	0.21
N	12	20
	Non-Expulsion Fuses (Mean)	No Non-Expulsion Fuses (Mean)	Diff	t-Stat	p-Value
High/Very High WHP	0.07	0.12	−0.05	1.00	0.32
Social Vulnerability Index	0.52	0.62	−0.10	1.79	0.08
Socioeconomic Percentile	0.46	0.58	−0.12	2.43	0.02
N	18	14

). The observed difference in percentage of service area at a high/very high WHP between utilities with a PSPS provision and those without is 10.4%. Utilities without a PSPS provision are also observed to have a higher average SVI and socioeconomic percentile, but these differences are not statistically significant. Utilities that include non-expulsion fuses in their WMPs have a statistically significant lower mean SVI and socioeconomic percentile score. Utilities without non-expulsion fuse installation planning also have a greater average percentage of service area at a high/very high WHP, but these results are not statistically significant.

3.2. Logistic Regression

3.2.1. PSPS Provision

Our models find a significant nonlinear relationship between high wildfire hazard potential and the odds of a utility including a PSPS provision in their WMP, while SVI and socioeconomic percentile have significant negative associations (

Table 2. Results for logistic regression models for PSPS provisions and non-expulsion fuses.

Input Variables	PSPS Provision				Non-Expulsion Fuses
	Model I		Model II		Model III		Model IV
	β (s.e.)	%	β (s.e.)	%	β (s.e.)	%	β (s.e.)	%
High/Very High WHP	−0.253 (0.169)		−0.260 (0.175)		−0.023 (0.028)		−0.025 (0.029)
(High/Very High WHP)2	0.009 * (0.005)		0.009 * (0.005)
Social Vulnerability Index	−0.064 * (0.033)	−6.2			−0.040 (0.025)
Socioeconomic Percentile			−0.069 * (0.037)	−6.67			−0.070 ** (0.034)	−6.72
Intercept	2.829 (1.849)		2.790 (1.911)		2.758 * (1.500)		4.085 ** (1.845)
Pseudo-R2	0.299		0.284		0.090		0.150

* p < 0.1, ** p < 0.05; N = 32. β relating input variable with log(outcome variable). (s.e. or standard error) for β. %, i.e., 100 ∗ (exp(β) − 1) is presented for variables with a linear relationship in the log of outcome.

). In nonlinear regression models, the effect of an independent variable may change sign across the range of values relevant to the variable. The term turning point is used to identify the threshold at which this change in sign occurs. In model I, the turning point in WHP is located at 13.24% of a utility service area at a high/very high WHP. For utilities with a high/very high WHP above 13.24%, increasing WHP has a positive effect on the odds of a utility adopting a PSPS provision. For example, for a utility with 15% of its service area at a high/very high WHP, and SVI held at its sample mean, a 1% increase in high/very high WHP is associated with a slightly more than 4% increase in odds of a utility including a PSPS provision. A 1% increase in SVI percentile is associated with a statistically significant decrease of 6.2% in the odds that a utility develops a PSPS protocol. Model II reaffirms the nonlinear statistically significant relationship between high/very high WHP (as a squared term) and the inclusion of a PSPS provision when socioeconomic vulnerability (i.e., a subset of the social vulnerability index) enters the model as a predictor. Model II reveals that a 1% increase in socioeconomic percentile is associated with a statistically significant 6.67% decrease in the odds that a utility develops a PSPS protocol, similar in magnitude and direction as the SVI impact.

3.2.2. Non-Expulsion Fuses

Our models find that SVI and socioeconomic percentile are negatively, and statistically significantly, associated with the odds of a utility including non-expulsion fuse installation in their WMP, while high/very high WHP is not a significant predictor (Table 2). Since 18 of the 32 utilities in the sample include non-exclusion fuse installation in their WMP, the overall odds are 1.29. While model III reveals that neither WHP nor SVI has a statistically significant effect on the odds of a utility incorporating non-expulsion fuses as a grid hardening measure, model IV finds a significant effect for socioeconomic percentile. Specifically, a 1% increase in socioeconomic percentile is associated with a 6.72% decrease in the odds that a utility includes non-expulsion installation in their WMP. We also tested quadratic terms for high/very high WHP, SVI, and socioeconomic percentile, but the addition of these quadratic terms did not improve the explanatory power of model III or IV and so we did not include them in the final models.

3.2.3. Robustness Checks

We conducted several additional robustness checks for our logit models. Tests for potential confounding factors including utility size, count of customers, historical fire count, and land cover were carried out with tests of these factors for inclusion as independent variables. None of these variables were found to have a statistically significant association with the inclusion of a PSPS provision or non-expulsion fuse installation, and inclusion was not warranted given the statistical cost of estimating additional coefficients as gauged by the lack of notable gains in pseudo-R2. This result was not unexpected given that historical fire count and land cover are both strongly correlated with WHP, since fuels and fire history metrics are incorporated into WHP. We also tested alternate model forms, including ordinary least squares and probit models. These alternate forms yielded similar and consistent results regarding the associations of the independent variables with the logit modeling presented here. In addition to testing for nonlinear effects through the inclusion of the square of percentage of service area at a high/very high WHP as a variable, we also tested potential nonlinear effects through the inclusion of quadratic terms for SVI and socioeconomic vulnerability. Including these quadratics did not increase the explanatory power of the logistic regressions. Lastly, we tested the inclusion of interaction terms to represent the possible associations between input variables. The inclusion of interaction terms in the logit models did not improve their explanatory power, as observed in the largely unchanged pseudo-R2, and thus we conclude that including interaction does not improve the model, nor does it outweigh the statistical cost.

3.3. 2 × 2 Matrix

We present a 2 × 2 matrix to explore the intersection between the adoption of a PSPS provision and wildfire risk by dividing the sample set of utilities into four groups (Figure 3). For illustration of this approach, utilities with a “higher WHP” are defined as those with more than 10% of their service area at high/very high WHP, while utilities below this are classified as “lower WHP”. Notably, five utilities in this sample (in the red quadrant) are categorized as higher WHP using the 10% service area threshold, but do not include a PSPS provision in their mitigation plan. This subset of utilities is then represented visually and spatially in Figure 4, revealing that four out of the five higher WHP utilities that adopted PSPS provisions are located east of the Cascades in Washington.

4. Discussion

Our analysis identifies statistically significant associations through logistic regression modeling for PSPS protocol development and non-expulsion fuse installation as they relate to wildfire hazard potential and social vulnerability. Specifically, we find that an increase in the percentage of service area at a high/very high WHP is associated with an increase in the odds of a utility having developed a PSPS provision, for utilities with about 13% or more of their service area classified as high/very high WHP. This result is aligned with our hypothesis that utilities at higher wildfire risk are more likely to include PSPS provisions in their wildfire mitigation plans. The inverse association was found for utilities with service areas at a lower wildfire hazard potential, as evidenced by the nonlinear influence of the quadratic term in our logit, although we would note that these utilities are of less concern for mitigation. Conversely, we find that SVI is negatively associated with the odds of inclusion of a PSPS provision, indicating that utility service areas characterized by more social vulnerability are less likely to have PSPS provisions mitigating wildfire risk. While there are additional factors also at play, one plausible interpretation is that regions of greater social vulnerability may not have the capacity to develop and maintain a PSPS protocol across an electrical grid.

We observe that utilities that have either already installed, or plan to install, non-expulsion fuses are more likely to have lower levels of social and socioeconomic vulnerability. Further we find that higher levels of socioeconomic vulnerability are associated with lower odds of non-expulsion fuse installation, which aligns with existing research on utilities balancing mitigation strategies with cost-effectiveness [23]. Practitioners can leverage these results in support of prioritizing technical assistance and funding to utilities with higher social vulnerability. For example, in 2023, Washington received a USD 23.4 million Grid Resilience State and Tribal Formula Grant from the U.S. Department of Energy to modernize the grid, focusing on resiliency in historically disadvantaged communities [43]. Considering social vulnerability and wildfire hazard potential can help prioritize local communities for grants such as this.

Our 2 × 2 matrix identifies utilities that may need additional support in mitigating wildfire risk and thus contributes to the decision-relevant information base for practitioners (Figure 3 and Figure 4). In our illustrative example, the matrix identifies utilities at higher wildfire risk that do not have PSPS provisions in their WMP, representing unmitigated risk. These utilities can serve as a focal point for policymakers and state agencies, as they may have limited capacity to develop comprehensive mitigation plans yet face greater risk of wildfires. Strategies including increased technical assistance or grants targeting public utility wildfire mitigation could increase capacity to improve grid resilience. It is possible, of course, that some utilities simply do not report PSPS provisions in their submitted plans, or alternatively face economic challenges around implementing PSPS. The matrix further identifies utilities whose service footprints are at lower wildfire risk but who have the capacity to incorporate PSPS protocols in their planning. Finally, utilities that have both greater wildfire risk and a PSPS protocol included in their wildfire mitigation plan are considered to represent mitigated risk. Notably, many utilities along the east side of the Cascades are in this category, characterized by high risk which is mitigated by PSPS provisions.

It is important to note that wildfire hazard potential is not uniform across each service area, but rather there is variability across geography and topography (see Figure 1). Further, the distribution of risk depends on the specific location of utility infrastructure within the service area. Also, setting the threshold for “higher” wildfire hazard (i.e., 10% of the service area falling above the 84th percentile of WHP) influences which utilities are assigned to each category. That being said, these visualizations are intended to represent overall wildfire risk for the purpose of illustrating our analytic approach and its potential applications.

In this study, we average census tract social vulnerability index scores to the utility service area level using area-weighting. As noted by other researchers, social vulnerability index considered in conjunction with WHP can identify and assess relative wildfire risk to vulnerable communities at the census tract scale [20]. Recent research has also investigated the relationships between PSPS and generalized social vulnerability indices. PSPS can effectively limit the risk of power-related wildfires, but the outages themselves pose significant countervailing risks to the public and thus are often locally unpopular [9,31,32]. As posited by Pollack et al. (2025), incorporating context-specific risks can help strengthen generalized vulnerability indices such as SVI for usage in infrastructural planning [22]. Our study builds on this research by looking at associations between SVI and wildfire mitigation strategy planning in the consumer-owned utility context. Model I builds on this result, revealing that social vulnerability and wildfire risk are both associated with the presence of a PSPS provision.

We acknowledge that our logistic regression models have moderate explanatory power, while the magnitude and sign of individual variable coefficients lends important insight. Although it is difficult to account for the many variables that may exert an influence on a utility’s wildfire mitigation plan, we focus on a statistically influential set. Beyond these, Washington’s climate and geography play a role in wildfire mitigation planning as well, with western Washington at a much lower overall risk of wildfire than the eastern side of the Cascades (Figure 1). We also note that the 32 public and cooperative utilities included in our sample range greatly in terms of total number of customers served, acreage of service area, and miles of transmission or distribution lines managed.

There are known limitations to representing wildfire risk with WHP. WHP is not intended as an all-encompassing measure of wildfire risk, and is strengthened by combination with spatial data of high-risk assets such as power lines and substations [36]. Therefore, consideration of the percentage of a utility service area with a high/very high WHP may not directly translate to high risk if utilities’ assets are not present in those specific areas. A utility serving both high- and low-WHP regions may have a cumulative percentage of service area at high/very high WHP, which may lead to potential underestimation of risk in fire-prone zones or overstated risk in low-WHP areas. However, in the absence of spatial and ownership data for transmission and distribution lines, a more nuanced approach is not feasible. If powerline location data were publicly available, we could better capture wildfire risk, as the location of critical infrastructure in fire prone areas could be represented accurately whether the overall service area was categorized as high- or low-WHP. It is also important to note that because WHP is derived from 2020 LANDFIRE vegetation data, our model does not account for any fuels or vegetation management changes that have occurred since. Despite these limitations, many Washington public utilities and cooperatives incorporate WHP into their wildfire mitigation planning, and use it as a motivating factor for determining risk mitigating strategies; thus, our analysis is designed to align with these practices [18].

Applying generalized social vulnerability and aggregating from the census-tract level to utility service areas introduces several limitations as well. As mentioned by Cutter (2024), the aggregation of social vulnerability as a single score for larger geographies can lead to oversimplification [42]. Large utilities in more rural areas may only comprise a few census tracts, while smaller utilities in more densely populated areas may include a far greater number of census tracts. Public or cooperative utilities that consider social vulnerability for grid hardening planning may opt to use SVI at the census tract level in order to consider the social vulnerability of individual tracts with greater service territory. Even at the census tract level, smaller high-risk or vulnerable communities may be overlooked [22]. Additionally, our definition of social vulnerability comprises the 16 measures from the American Community Survey, excluding other factors that may define a population as socially or socioeconomically vulnerable. Despite this, social vulnerability is still a useful tool for the initial assessment of potential vulnerability within a utility service area, although this should be followed with a careful analysis of local community conditions.

While the main focus of this analysis is PSPS provisions and non-expulsion fuses, further research considering wildfire risk and social vulnerability as they relate to mitigation strategies such as vegetation management, workforce training, or other grid hardening measures could shed light on rural public and cooperative utilities that may receive less attention than larger IOUs.

In researching public utilities in Washington, the greatest barrier to analysis is data availability. Greater data transparency and accessibility would allow for further examination of this topic both in Washington and beyond. Future analyses could expand to include small public utilities in Oregon and California, which would allow for an exploration of the applicability of this approach under different political, legislative, geographic and climatic conditions. While a strictly analogous analysis in Oregon and California is not possible at this time due to both state-specific political context and data limitations, future research could investigate the generalizability of these findings. Additionally, future research could account for additional variables as utility-specific data become more available, such as the extent and location of underground distribution lines associated with each utility. This would, in turn, allow state and federal governments to target funding or tailor technical assistance programs to utilities that lack the resources and capacity to mitigate risk despite facing significant potential wildfire impacts. Our results also contribute to developing an approach for identifying the utilities that could most benefit from public programs and funding opportunities.

5. Conclusions

As an independent analysis of mitigation strategies in Washington relying solely on publicly available data, this study provides a broad overview of the wildfire mitigation planning landscape. We identify significant associations between factors such as wildfire risk and social vulnerability, and the inclusion of specific strategies in public and cooperative utility wildfire mitigation plans in Washington State. The most significant and impactful relationships found to exist are between wildfire hazard potential and social vulnerability, and whether a utility has adopted a PSPS provision or not, as well as between socioeconomic vulnerability and whether a utility has installed or plans to install non-expulsion fuses. In the future, our approach can be used and adjusted for other states to analyze wildfire mitigation planning for public electric utilities in other contexts. Combining this research with projections of climate change and fuels accumulation in Washington State may further support policymakers in future planning and prioritizing utilities or regions which face significant risk but are limited in their capacity to mitigate it. Our results support a proactive approach to wildfire mitigation in Washington state, and shine a light on public and cooperative electric utilities and their ongoing efforts to adapt to wildfire risk under a changing climate.