Worldviews more than experience predict Californians’ support for wildfire risk mitigation policies

California must adapt to increasing wildfire activity concurrent with climate change and expanding housing development in fire-prone areas. Recent decades have seen record-breaking fire activity, economic costs, and human health impacts. Residents more frequently face home evacuations, prolonged periods of unhealthy air quality, and power shut-offs. Understanding how these experiences influence support for risk mitigation policies is essential to inform action on climate and fire adaptation. To better understand linkages between experience and policy support, we surveyed California residents (n = 645) about their wildfire-related experiences, risk perceptions, and support for 18 wildfire risk mitigation policies. To assess how the relationship between policy support and wildfire experience is modulated by preexisting worldviews, we measured the extent to which respondents are motivated by individualistic or communitarian values as proposed in the cultural theory of risk. We surveyed residents across a gradient of wildfire impacts, spatially stratifying residences based on wildland-urban-interface type and proximity to large 2020 wildfires. Support was generally high for most policies, though most respondents opposed incorporating future risk into insurance rates and coverage. Policy support models showed that communitarian worldviews were more consistently associated with greater support for diverse wildfire mitigation policies than were measures of recent experience with wildfire. These results suggest that California residents within our sample regions already support many wildfire risk mitigation strategies, and preexisting societal beliefs are a stronger predictor of these views than personal experiences with wildfire. Policy-makers can utilize this understanding to focus on crafting policies and messaging that resonates with individualistic values.


Introduction
Climate change has led to a global increase in wildfire risk (Ellis et al 2022, Richardson et al 2022).While wildfire is a natural and inevitable component of the Earth system, increased wildfire activity and human settlement of fire-prone areas create greater risks for society (Schug et al 2023).In the United States, California in particular has become an epicenter of increasingly severe wildfire impacts, exacerbated by a disproportionate increase in the wildlandurban interface (WUI) in regions more vulnerable to climate-driven wildfire (Rao et al 2022).With climate change and expansion of the WUI projected to continue (Radeloff et al 2018), policies that promote mitigation and adaption to wildfire are essential to California's future.
As the frequency and severity of impacts due to wildfires continue to increase, these experiences may influence public willingness towards wildfire mitigation.Fire seasons from 2010 to 2020 in California were record-breaking in terms of length, frequency of extreme fire weather, economic costs, and human health impacts (Palinkas 2020).The 2020 wildfire season was particularly devastating, with five of the seven largest fires in California's recorded history amid a period of especially high aridity (Higuera andAbatzoglou 2021, Safford et al 2022).A lightning storm ignited many of the fires in mid-August, enveloping the state in smoke and causing an unprecedented period of prolonged and hazardous air quality (Navarro and Vaidyanathan 2020).Further, a policy of public safety power-shutoffs was newly implemented in response to the deadly Camp Fire of 2018, causing widespread power shutoffs in fire-exposed areas during extreme fire weather (Mildenberger et al 2022).Residents of California are now living through wildfire impacts that are outside the range of their previous experience and will likely continue to do so as climate change increases the frequency and intensity of hot and dry conditions that are linked to greater wildfire risk (Loehman et al 2020, Parks and Abatzoglou 2020, Ellis et al 2022).
Will experiences with wildfire increase public support for adaptive policies that would reduce future fire risks?In related research, the evidence is mixed as to whether experiences with extreme weather events influence climate change opinions and adaptive beha- Opinions and experiences with natural hazards are filtered through the lens of an individual's worldview (figure 1).The 'cultural cognition' thesis posits that individuals are psychologically disposed to perceive and evaluate events according to their personal values (Douglas and Wildavsky 1983).This conceptual framework groups beliefs surrounding risk and the role of government across two dimensions, including an individualistic-communitarian continuum that measures the extent to which people believe social goals should be accomplished through individual or collective effort (Kahan 2012, Rachlinski 2021).Cultural cognition is useful in modeling policy support because it helps to explain how people perceive risk and differentially evaluate their experiences with natural hazards (Newman et al 2018).In the context of western US wildfires, cultural cognition has been used to predict individual behaviors (Steel et al 2023), however it has not yet been applied to predicting policy support.
As extreme wildfire seasons increase in frequency, it is imperative to understand how novel experiences with wildfire may affect willingness to adopt mitigation measures.To address this need, we surveyed California residents about their self-reported experiences with wildfire in the 2020 fire season (e.g.air quality, power shut-offs, insurance costs), their cultural world views, and their support for a range of wildfire risk and climate change mitigation policies.
We used a spatially explicit sampling design that stratified by proximity to wildfire perimeters and WUI designation to survey across a gradient of wildfirerelated impacts.Our research objectives were to: (1) analyze biophysical trends in fire-prone ecoregions of California to quantitatively contextualize the 2020 fire season, (2) characterize support across a range of policies along a continuum from household-level to government mandates, and (3) assess the relative importance of experiences with wildfire and worldviews in driving policy support.

Biophysical trends
We quantified mean annual fire-season climate and area burned in wildfires from 1960 through 2020, as well as air quality from 2000 through 2020 for the five most wildfire-prone ecoregions of California (figure 2(a)).Ecoregion boundaries were derived from The Nature Conservancy Terrestrial Ecoregions dataset (The Nature Conservancy 2009).
Area burned annually was calculated from wildfire perimeter data and includes wildfires that burned on state, federal, and private lands (CAL FIRE 2020).Fires were assigned to one of the five ecoregions if their centroid was located within the ecoregion boundary, and fires less than 0.4 ha in size were removed from the dataset.To account for instances of fires mapped more than once in a given year, we aggregated all fire perimeters within a year using the R package sf (Pebesma 2018).
We assessed fire-season climate using monthly average climatic water deficit (CWD) at 1/24 • resolution (∼4 km) from TerraClimate (Abatzoglou et al 2018).CWD estimates the evaporative demand that is not met by precipitation while accounting for soil water-holding capacity and is considered a biologically meaningful proxy for the drought stress experienced by plants (Stephenson 1998).We calculated mean CWD from July through September of each year within each ecoregion.In order to reduce the influence of areas that are not prone to wildfires (e.g., urban centers), we only calculated mean CWD within the aggregated perimeters of all fires.The mean of all CWD pixels within an ecoregion was then calculated for each year.Data were accessed and processed through Google Earth Engine (Gorelick et al 2017).
Finally, we used PM 2.5 levels to assess trends in air quality because it represents a large component of wildfire smoke and is harmful to public health (Dennis et al 2002, Reid et al 2016).Data for PM 2.5 concentrations were accessed from the EPA Air Quality System (AQS) database, which collects air quality data from thousands of monitoring stations across the United States (https://epa.gov/aqs).We averaged the daily PM 2.5 concentrations for all monitoring stations (EPA parameter codes 88 101 and 88 502) located within a given ecoregion from July through September using the '24 h observation' values when available and the '24 h block averages' otherwise.
In order to compare biophysical trends across ecoregions, we calculated z-scores of each variable relative to the 60 year normal  for area burned and CWD, and the 20 year normal for PM 2.5 .Zscores represent how anomalous a variable was during a given year relative to its historic mean.We assessed ecoregion trends in each variable using the Siegel repeated median linear estimator (Siegel 1982), a more robust version of the Theil-Sen estimator implemented in the R package mblm.We additionally estimated the strength of correlation between annual area burned by ecoregion and the remaining two variables, CWD and PM 2.5 , using Pearson's r.

Survey sampling design
To obtain survey responses from individuals across a severity gradient of wildfire-related experiences, we used a stratified random sampling design to select residences to send recruitment postcards.We selected a subset of the three to four largest and most damaging 2020 fires that intersected each California ecoregion detailed above (North California Coast and Klamath Mountains were combined) and extracted their perimeters from the CAL FIRE wildfire perimeter data (CAL FIRE 2020) (figure 2).Three zones of proximity were then created from the edge of each fire: (1) 5-25 km, (2) 0-5 km, and (3) within the fire perimeter.In the final zone, our sample was drawn from only areas that were estimated to be unburned or burned at low severity, based on a composite burn severity mapping approach (Parks et al 2018).To sample residential areas facing different levels of exposure to wildfire, we further stratified our three proximity zones by WUI types based on designations from 2010 census and vegetation data (Radeloff et al 2018).These three zones included: (1) low-and medium-density WUI (6-741 dwellings km −2 ), (2) high-density WUI (>741 dwellings km −2 ), and (3) medium-and highdensity non-WUI (>49 dwellings km −2 ).A census tract was considered WUI if it was designated either 'interface' or 'intermix' (Radeloff et al 2018).
We generated random coordinates within each of these stratified zones and converted the points to street addresses using the Google reverse geocoding API through the R package ggmap (Kahle and Wickham 2013).We then subset the list of addresses for those labeled as a 'premise' (the Google designation of a dwelling unit) and removed duplicates, resulting in 29 510 unique addresses.Survey recruitment postcards were mailed to these addresses during the week of 15 March 2021 with a QR code and URL link to a Qualtrics web survey.A unique code was printed on each postcard allowing survey responses to be linked to the residence they were sent to.We sent an additional 10 000 reminder postcards to a subset of addresses that had not taken the survey approximately three weeks after the initial round was sent.

Survey content
We first asked survey respondents to confirm their residency in California at the address their recruitment postcard was sent to during the 2020 wildfire season, as well as whether they owned the property.The next set of questions asked the respondent about six different types of wildfire-related experience, including impacts to health (air quality and injury), disruptions to the household (evacuation and PSPS), and financial impacts (damages or loss of property and changes to insurance) (figure 1).We then asked respondents about their level of support for 18 wildfire risk reduction policies that either are currently being implemented or are under consideration in California (table 1).Support for each policy was scored as a four-level Likert-type item, including 'strongly oppose' , 'oppose' , 'support' , and 'strongly support' , as well as a 'do not know' option.Cultural cognition on the 'group' axis (individualism-communitarianism) was then assessed using the 'short form' version which includes six of the 17 total questions using a six-level Likert scale (Kahan 2012).See supplemental file 2 for the full survey.

Survey statistical analysis
Prior to modeling the effect of wildfire-related experiences and cultural cognition on support for wildfire risk mitigation policies, we converted each experience type to a four-level ordinal scale of impact (none, low, medium, and high).The exception to this was the injury category, which was converted to a binary variable (details on how each experience type was coded Table 1.Abbreviated descriptions of the 18 wildfire and climate change mitigation policies that were included in the survey.Respondents were provided more detailed policy descriptions with an accompanying image before being asked about their level of support for the policy.See supplementary file 2 for the exact wording of each question in the survey.2008).We report all survey responses that indicated their level of support for the 18 wildfire risk reduction policies (n = 645).We modeled support for each policy using Bayesian logistic regression in the R package brms (Bürkner 2017).Policy support was modeled as a function of experience, cultural cognition, and their interaction.Age, gender, and education level were included in all models to control for their effects.Both ordinal variables (experience and education level) were modeled as monotonic effects, which allow differences between adjacent categories (e.g.none to low vs. low to medium experience) to differ in their effect size, while still assuming an overall monotonic relationship (Bürkner and Charpentier 2018).Each model included only one experience type, resulting in 108 models (18 policies × six experience types).Total sample size varied between each policy, as we removed 'do not know' responses prior to fitting the model.Each model used the default parameters for the NUTS (no-U-turn sampler) Markov chain Monte Carlo (MCMC) algorithm as implemented in the brms package (Bürkner 2017), consisting of four MCMC chains with 1000 burn-in steps before sampling for an additional 1000 steps, resulting in 4000 posterior draws.All models achieved chain convergence with Ř values < 1.05, indicating agreement for between-and within-chain parameter estimates.A Cauchey distribution with a shape parameter of 2.5 was used as a weakly informative prior for all predictor variables (Gelman et al 2008).We excluded survey results from respondents that indicated they had a different primary address, those that did not answer all three of the demographic variables used in the models (age, gender, and education), and those that did not answer all cultural cognition questions.

Biophysical trends
Within the five ecoregions (figure 2), more than 1.4 million ha burned in 2020 (a total of 1.74 million ha burned in California (Safford et al 2022), with the difference due to the exclusion of fires outside the ecoregions in our analysis).Three of the five ecoregions (California North Coast, Central Coast, and Sierra Nevada) recorded their highest area burned in 2020, and four of the five ecoregions (all but the California South Coast) had their highest levels of mean fireseason fine particulate matter (diameter < 2.5 µm; PM 2.5 ) during the timeframes analyzed (60 years and 20 years, respectively) (figure 3).The 2020 fire season was also one of the driest between 1960 and 2020 across ecoregions as measured by CWD, ranking the driest year for the California South Coast, second driest for the North Coast, third driest for the Central Coast and Sierra Nevada, and sixth driest for the Klamath Mountains (figure 3).
Fire activity (area burned), fire-season aridity (CWD), and wildfire smoke-related air pollution (PM 2.5 ) levels increased broadly across ecoregions (figure 3).For the California North Coast, Klamath Mountain, and Sierra Nevada ecoregions, all three biophysical variables increased significantly over the years analyzed (p < 0.05).Though area burned in the California South Coast significantly increased, it was the only region where fire-season PM 2.5 has decreased since 2000, and no trend was identified in CWD.Finally, the California Central Coast showed a significant upward trend in CWD and area burned, but no trend in PM 2.5 .The correlation between CWD and area burned was moderate in the Klamath Mountains, Sierra Nevada, and California North Coast (r = 0.49, 0.36, and 0.32, respectively) and, while the direction of correlation was positive in the California Central and South Coast ecoregions, was non-significant.Fire-season mean PM 2.5 was more strongly correlated with area burned across ecoregions (r = 0.73-0.89)than was CWD and area burned, except in the California South Coast where there was no significant correlation.

Survey responses and policy support
We received 645 (2.2% response rate) completed survey responses from across the state (figure 2).Table 2 summarizes the explanatory variables used to model policy support, as well as the geographic locations of survey respondents.While our sample was older and more highly educated than would be expected from a random sample of California residents, an unrelated survey of property owners in the same regions yielded a similar socio-demographic profile (Tarver 2023), suggesting our sample may be representative of WUI residents in this geography.One-way analysis of variance (ANOVA) revealed that cultural cognition scores were significantly associated with education level (higher levels of educational attainment tended to correspond with more communitarian scores; F 4,537 = 10.64,p < 0.001; figure S9), however cultural cognition scores did not differ significantly with age or gender (respectively, figure S10 (F 7,534 = 0.685, p = 0.685) and S11 (F 1,540 = 2.221, p = 0.137)).
A wide range of wildfire-related impacts were reported by respondents (figure 4).In particular, 49% reported high levels of impact from mandatory evacuations, 42% reported high levels of impact due to poor air quality, and 61% reported moderate to high levels of impact from public safety power shutoffs (PSPS) (table S1).Our survey also included a broadly representative spectrum of responses from individuals classified across the cultural cognition axis from individualistic to communitarian (table 2).Approximately 32% were categorized as primarily individualistic, 31% as primarily communitarian, and 37% as some combination of both.ANOVA results revealed that cultural cognition scores were strongly associated with survey respondents' political affiliation; Republicans tended to be more individualistic, Democrats more communitarian, and independents fell in between (F 2,553 = 177, p < 0.001; figure S8).
Support for the 18 wildfire mitigation policies was widespread among survey respondents, with 10 of the policies receiving more than 80% support, including at least one in each of the six main policy categories (figure 5).Of the remaining eight policies, all but one still had greater than 50% support.Allowing insurance companies to consider future wildfire risk when pricing homeowner insurance rates was the only policy with the majority (53%) in opposition.

Wildfire experience vs. worldviews
We modeled policy support with a subset of respondents who answered all demographic questions used in the model and who indicated their primary residential address was where they received the survey invitation (n = 542).Bayesian logistic regression coefficient estimates and their 95% credible intervals (CI) for all The y-axis shows the z-score for each variable and year, which is the number of standard deviations from the mean (60 year mean for area burned and CWD; 20-year mean for PM2.5) a given variable was that year (a value of 0 indicates the mean; positive values are greater and negative values are less than the mean).The Pearson correlation coefficient (r) between area burned and CWD (blue) and PM2.5 (black) are given for each ecoregion when significant (α < 0.05).Arrows in the upper right corner of each ecoregion indicate a significant increasing (pointed up) or decreasing (pointed down) trend in area burned (red), CWD (blue), or PM2.5 (black) across the timeframes analyzed.
18 policies and six wildfire experience types (108 total models) can be found in figure S7.
As support was high across most policies, we highlight the eight most controversial policy models (<75% support) in figure 6. Wildfire experience coefficients had a non-zero effect size (i.e. the CI did not overlap zero) in 11 of these 48 models, with PSPS experience correlating with support for four of the eight policies, experience of changes in insurance for three, evacuation experience for two, and damage and air quality experience for one (figure 6).In all but two of the models showing that experience had an effect, more severe levels of experience were related to decreased likelihood of supporting the given policy.The exception to this was greater experience with changes in homeowner's insurance (e.g.increased rates), which correlated with higher support for incentivizing defensible space and continued implementation of PSPS.
In contrast to wildfire-related experience, cultural cognition displayed a stronger and more consistent relationship with policy support (figure 6).Higher communitarian scores were predictive of increased policy support for six of the eight policies.For the remaining two policies, cultural cognition generally showed no effect, though one model for mechanical Table 2. Frequency counts and percentages of demographic and geographic traits of survey respondents.Total count includes all surveys that responded to the 18 policy support questions about wildfire risk mitigation.Model count only includes surveys that answered all three demographic and cultural cognition questions used to model support.Note that cultural cognition is grouped into categories for display and was modeled as a metric variable (−18-18).Respondents who indicated they rented rather than owned their residence were not asked about insurance-related impacts.For details about the criteria determining impact level see figures S1-S6.Policies with less than 75% support (dashed line) were selected for modeling the effect of wildfire experience and worldviews (cultural cognition) on support (figure 6).See table 1 for descriptions of each policy and supplementary file 2 for survey question wording.
thinning as a land management practice was the only instance in which higher individualistic scores predicted increased support.Among the demographic variables, respondent age was positively associated with support for four of the eight policies (figure 6).Education showed contrasting correlations with two policies, with higher educational attainment predictive of support for the most controversial policy (allowing insurance companies to consider future risk when pricing premiums), but decreased support for incentivizing maintenance of defensible space.Finally, men were more likely to support forest-thinning treatments and allowing insurance companies to consider future risk when pricing policies, while women were more likely to support financial incentives for maintaining defensible space, restricting new building in high-risk areas, and both policies related to Cap-and-Trade.

Discussion
Citizens tend to support immediate governmental solutions to disasters more than future-oriented prevention strategies (Healy andMalhotra 2009, Dvir et al 2023).Severe impacts of disasters such as floods, tornadoes, or earthquakes tend to be confined to specific localities; however, millions of California residents have experienced inconvenience, discomfort, or loss in recent years due to wildfires.This may explain why we found Californians in the regions we sampled tend to support government policies designed to mitigate future wildfire risk to homes or communities-sometimes overwhelmingly so.Support was weaker for policies that could directly affect respondents' pocketbooks through fines, zoning restrictions, power shutoffs, or higher insurance rates.In a study conducted shortly before the devastating 2020 wildfire season, Hui et al (2022) similarly found that Californians preferred policies that allow people to make resilience decisions themselves but do not impose burdens on all residents.
Direct experience with natural hazards is personally traumatic as well as costly, and research has shown that persons who experience hazards are more likely to take actions to protect themselves from future risk of floods (Demski et al 2017) or wildfire (Wolters et al 2017).However, studies of the relationship between experience of natural hazards and broader policy support have shown mixed results.Flooding experience was linked to support for flood mitigation policies in the United Kingdom (Demski et al 2017); in contrast, research in California found that experiencing PSPS in 2019 did not increase willingness to pay for wildfire-mitigating reforms (Mildenberger et al 2022), nor did exposure to smoke predict support for public funding of some resilience measures (Hui et al 2022).In our study, some models indicated a significant relationship between policy support and wildfire experience, but in an unexpected direction: more severe levels of experience generally were related to a decreased likelihood of supporting a given policy.Our data do not provide a means to assess why experience suppresses support for mitigation policies, but previous studies suggest that perceptions of future wildfire risk may decline immediately after a wildfire (Larsen et al 2021) and that experiencing a wildfire may cause some people to believe it is not possible to mitigate wildfire impacts (McGee et al 2009).It is also possible that experience effects may differ among those most severely impacted-such as forced relocation out of the region due to extreme property damage-since our sample included only current post-wildfire residents.
The factor that most clearly influenced policy preferences was not individual experience, but rather individual values: specifically, respondents' scores on a cultural cognition measure that contrasted communitarian and individualistic worldviews.Persons who scored higher in the direction of communitarian values were more likely to support policies that can mitigate future wildfire risk, while those with individualistic values were more likely to oppose them.Cultural cognition (Kahan 2012) offers a way to measure the extent to which individuals form risk perceptions that reflect and reinforce their commitment to particular ways of life (Thompson et al 1990).An individual's political and cultural commitments shape how they process information about natural hazards and risk (Hartter et al 2020, Rachlinski 2021) and consequently their support or opposition to proposed policies to address hazards and risks (Kahan and Braman 2006, Boag et al 2018, Hamilton et al 2020).The communitarian-individualism continuum to some extent mirrors U.S. political platforms, with Democrats supporting solutions that call for government action, while Republicans prefer solutions promoting individual choice, and indeed the strong connection between party affiliation and the group axis of cultural cognition was reflected in our survey data (figure S8).Hui et al (2022) found a strong influence of party affiliation on support for wildfire adaptation policies; Democrats generally supported such policies while Republicans generally opposed them, although proximity to a recent wildfire was associated with greater Republican support for using public funds to encourage, but not require, homeowners to take risk-reduction measures.In our study, cultural cognition predicted the strength of support for both incentive-based and restrictive policies, as well as cap-and-trade policies that address climate change as an underlying cause.A cultural cognition lens helps policy makers not only to predict who will support particular policies, but to begin to explain the underlying reasons why.Armed with such understanding, policy recommendations can be crafted to emphasize individualistic benefits of proposed actions that might otherwise draw greater opposition (e.g., by emphasizing the role of PSPS in reducing future electric rate hikes to individual homeowners).
Although the generalizability of our survey findings may be somewhat limited by a low response rate (2.2%), our sample size (n = 645) was consistent with similar studies (e.g.Molina et al 2021, Shrestha et al 2021, Mildenberger et al 2022).Response rates to surveys about policy and politics have been broadly declining in the United States, leading to an overrepresentation of politically engaged individuals (Cavari and Freedman 2023), which could potentially bias our results towards Californians with stronger preexisting opinions on policy that are less influenced by personal experience.However, our sample included a balanced representation of individuals across the cultural cognition spectrum, which combined with the consistency of our findings that worldviews outweighed the influence of experience on policy support, suggests a broader inferribility of our key results.
While cultural cognition had a strong influence on support for policies to reduce future impacts, and experience of wildfire had little influence of its own, policy makers also should take notice that we found widespread support for almost all policies we asked about.Our findings that pre-existing societal beliefs were a stronger driver of these views than personal experiences with wildfire suggests that additional wildfire exposure is unlikely to substantially change public attitudes towards mitigation approaches.These results help advance understanding of beliefs surrounding risk and natural hazards, and can help guide California in making informed policy decisions regarding strategies to mitigate risks of wildfire.

Conclusion
The year 2020 was record-setting for wildfire in California's modern history, part of a trend of increasing wildfire activity across much of the state.Despite the significant impacts from those fires reported by the residents we surveyed, our results indicate that experience played little role in their support for policies to reduce future wildfire risks and, somewhat counterintuitively, decreased support in several cases.In contrast, preexisting worldviews influenced support more consistently and strongly, with communitarian views associated with greater support for the majority of wildfire risk mitigation policies compared to those with more individualistic beliefs.Lawmakers and advocates can capitalize on this knowledge to craft supportive messaging for policies that acknowledges individualistic worldviews.Regardless of the underlying mechanism, those surveyed were highly supportive of the majority of the 18 policies they were asked about, which included both incentive-based and restrictive policy types.Together, these results suggest that policy makers should not expect substantial shifts in constituent sentiment as trends in wildfire activity continue to increase.Rather, the support for broad action to reduce wildfire risk already exists.

Figure 1 .
Figure 1.Conceptual framework used in this study.Increasing wildfire leads to more frequent and severe experiences (e.g.health consequences, financial loss, disruptions to daily life), which may be direct or indirect.These experiences are filtered through an individual's preexisting beliefs (cultural cognition) to inform their support for policies to reduce wildfire risk.
viors (Howe 2021).Though some research has shown that exposure to extreme weather increases concerns about climate change (Brooks et al 2014, Lujala et al 2015, Konisky et al 2016, Bergquist and Warshaw 2019), other studies have found no effect (Albright and Crow 2019, Hughes et al 2020).A smaller body of literature has looked directly at the influence of experience with natural hazards on policy preferences (van Valkengoed and Steg 2019).Some recent research has found that proximity to wildfire is associated with increased support for climate adaptation policies (Zanocco et al 2019, Hazlett and Mildenberger 2020, Hui et al 2022, Wolters 2023), but other factors also drive policy support, such as attribution, political ideology, and preexisting beliefs.Similarly, adoption of fire risk-reduction practices can increase after a wildfire (Mockrin et al 2018, Larsen et al 2021), but public support for policy and planning tools to reduce risk varies with local social and political contexts (Hazlett and Mildenberger 2020, Mockrin et al 2020).Given the elements that influence risk perceptions and mitigation preferences are multifaceted and context-dependent, work that helps disentangle the relative importance of various factors is critical for more nuanced understanding that can aid adaptive planning and outreach as climate change hazards continue to increase.

Figure 2 .
Figure 2. Survey response locations (black dots) and the 2020 fire perimeters (red) used to spatially stratify random households (within 25 km) for survey invitations.(a) shows the five ecoregions within California for which biophysical temporal trends were analyzed (figure 3).(b) Inset highlights spatial distribution of survey responses for a heterogeneous area within the Central Coast ecoregion, with low-and medium-density WUI (light green), high-density WUI (dark green), and medium-and high-density non-WUI (dark gray) shown for context.

Figure 3 .
Figure 3. Temporal trends by ecoregion (see figure 2) in area burned (red) and climatic water deficit, a measure of aridity (CWD; blue) from 1960 to 2020.Trends in fine particulate matter (diameter <2.5 µm; PM2.5; black) are shown from 2000 to 2020.The y-axis shows the z-score for each variable and year, which is the number of standard deviations from the mean (60 year mean for area burned and CWD; 20-year mean for PM2.5) a given variable was that year (a value of 0 indicates the mean; positive values are greater and negative values are less than the mean).The Pearson correlation coefficient (r) between area burned and CWD (blue) and PM2.5 (black) are given for each ecoregion when significant (α < 0.05).Arrows in the upper right corner of each ecoregion indicate a significant increasing (pointed up) or decreasing (pointed down) trend in area burned (red), CWD (blue), or PM2.5 (black) across the timeframes analyzed.

Figure 4 .
Figure 4. Impact levels for six types of wildfire-related experiences as a percentage of all survey respondents (n = 645).Respondents who indicated they rented rather than owned their residence were not asked about insurance-related impacts.For details about the criteria determining impact level see figures S1-S6.

Figure 5 .
Figure 5. Percent support (left blue bars) and opposition (right tan bars) for 18 wildfire-related risk mitigation policies in six broad categories.Policies with less than 75% support (dashed line) were selected for modeling the effect of wildfire experience and worldviews (cultural cognition) on support (figure6).See table 1 for descriptions of each policy and supplementary file 2 for survey question wording.

Figure 6 .
Figure 6.Results of Bayesian logistic regression models predicting support for eight wildfire risk reduction policies with support below 75%.For each policy, a separate model was run with one of the six wildfire experience types (y-axis), while all other predictor variables remained in the model.Coefficient estimates greater than 0 indicate support for the policy, while estimates less than 0 indicate opposition.Bars show the 95% credible interval of each coefficient's posterior distribution.The cultural cognition variable ranged from −18 (highly individualistic) to 18 (highly communitarian), with 0 representing a neutral value.For gender, positive coefficient estimates indicate greater support by men and negative estimates indicate greater support by women.
Harden New Require Policies that require newly built homes in high-risk areas to use fire-resistant materials Harden New Incentives Policies that provide financial incentives like rebates to use fire-resistant materials in new homes Harden Old Incentives Policies that provide financial incentives like rebates to update old homes with fire-resistant Regulations on whether insurance companies are allowed to consider increased future wildfire risk to homes when pricing policies Disclose Risk Regulations on whether property owners must disclose wildfire risk when selling a property Land Management Public land management actions aimed at reducing wildfire risk Thinning State and federal policies regarding the use of mechanical thinning to selectively cut trees and shrubs Prescribed Burn State and federal policies regarding the use of prescribed/intentional fires to reduce fuels Grazing State and federal policies regarding the use of controlled grazing in grass and shrublands to reduce fuels Howe et al McGee T K, McFarlane B L and Varghese J 2009 An examination of the influence of hazard experience on wildfire risk perceptions and adoption of mitigation measures Soc.Nat.Resour.22 308-23 Mildenberger M, Howe P D, Trachtman S, Stokes L C and Lubell M 2022 The effect of public safety power shut-offs on climate change attitudes and behavioural intentions Nat.The California Wildfires.Global Climate Change, Population Displacement, and Public Health: The Next Wave of Migration ed L A Palinkas (Springer) pp 53-67 Parks S A and Abatzoglou J T 2020 Warmer and drier fire seasons contribute to increases in area burned at high severity in western US forests from 1985-2017 Geophys.Res.Lett.47 e2020GL089858 Parks S, Holsinger L, Voss M, Loehman R and Robinson N 2018 Mean composite fire severity metrics computed with Google Earth Engine offer improved accuracy and expanded mapping potential Remote Sens. 10 879 Pebesma E 2018 Simple features for R: standardized support for spatial vector data R J. 10 439-46 Rachlinski J J 2021 What is cultural Cognition, and why does it matter?Annu.Rev. Law Soc.Sci. 17 277-91 Radeloff V C et al 2018 Rapid growth of the US wildland-urban interface raises wildfire risk Proc.Natl Acad.Sci.115 3314-9 Rao K, Williams A P, Diffenbaugh N S, Yebra M and Konings A G 2022 Plant-water sensitivity regulates wildfire vulnerability Nat.Ecol.Evol.6 332-9 Reid C E, Brauer M, Johnston F H, Jerrett M, Balmes J R and Elliott C T 2016 Critical review of health impacts of wildfire smoke exposure Environ.Health Perspect.124 1334-43 Richardson D, Black A S, Irving D, Matear R J, Monselesan D P, Risbey J S, Squire D T and Tozer C R 2022 Global increase in wildfire potential from compound fire weather and drought npj Clim.Atmos.Sci. 5 1-12 Safford H D, Paulson A K, Steel Z L, Young D J N and Wayman R B 2022 The 2020 California fire season: a year like no other, a return to the past or a harbinger of the future?Glob.Ecol.Biogeogr.31 2005-25 Schug F et al 2023 The global wildland-urban interface Nature 621 1-6 Shrestha A, Grala R K, Grado S C, Roberts S D and Gordon J S 2021 Landowner concern about wildfires and implementation of fuel reduction treatments J. For.119 251-65 Siegel A F 1982 Robust regression using repeated medians Biometrika 69 242-4 Steel B S, Lovrich N P and Pierce J C 2023 Cultural theory, wildfire information source, and agency public trust: a central oregon case study Environ.Manage.73 579-94 Stephenson N 1998 Actual evapotranspiration and deficit: biologically meaningful correlates of vegetation distribution across spatial scales J. Biogeogr.25 855-70 Tarver R D 2023 Cross-Boundary Stewardship in Protected Area Centered Ecosystems: Perceptions of Success and Characteristics of Cooperative Engagement Ph.D (Utah State University) (available at: www.proquest.com/docview/2820285283/abstract/BC893A4A7CD14D68PQ/1) The Nature Conservancy 2009 Terrestrial ecoregions of the world (available at: https://geospatial.tnc.org/datasets/b1636d640ede4d6ca8f5e369f2dc368b/about) Thompson M, Ellis R and Wildavsky A 1990 Cultural Theory (Westview Press) (available at: www.taylorfrancis.com/books/mono/10.4324/9780429501180/cultural-theorymichael-thompson-richard-ellis-aaron-wildavsky)van Valkengoed A M and Steg L 2019 Meta-analyses of factors motivating climate change adaptation behaviour Nat.Clim.Change 9 158-63 Wolters E A 2023 Homeowner firewise behaviors in fire-prone central Oregon: an exploration of the attitudinal, situational, and cultural worldviews impacting pre-fire mitigation actions J. Environ.Manage.327 116811 Wolters E A, Steel B S, Weston D and Brunson M 2017 Determinants of residential firewise behaviors in central oregon Soc.Sci.J. 54 168-78 Zanocco C, Boudet H, Nilson R and Flora J 2019 Personal harm and support for climate change mitigation policies: evidence from 10 U.S. communities impacted by extreme weather Glob.Environ.Change 59 101984 A A