Taking the "wild" out of wildfires—harnessing information infrastructure data to predict, prevent, and prepare for the future

Wildfires are increasingly frequent as climate change increases heat and prolongs dry seasons¹. Globally, millions of hectares of trees were lost to wildfires between 2021 and 2025², endangering millions through displacement, property loss, and health issues from smoke exposure¹. Impacts on natural ecosystems and human infrastructure are exacerbated by ongoing environmental and human activities that continue to raise global temperatures³. Catastrophic events stemming from wildfires, amply documented in the western United States, Hawaii, Canada, and Spain, represent millions of dollars in lost infrastructure, human and animal life, and damage to local ecosystems^4,5,6.

Effective response to wildfires and smoke depends on clear, trusted, and timely communication⁷. Research indicates that inadequate messaging, timing misalignment, or a lack of comprehension can result in confusion and mistrust among the public and noncompliance with evacuation or health directives⁸. These issues disproportionately impact marginalized and vulnerable populations, such as the elderly, who exhibit lower preparedness levels stemming from confusion and variability in messaging⁷. Citizen science initiatives like Smoke Sense suggest that psychological preparedness, self-efficacy, and risk perception significantly influence protective behavior⁹. While both social media and traditional outlets are critical sources of information, they also propagate misinformation^7,9,10,11. Most wildfire research isolates either institutional response capacity or individual risk behavior, but rarely examines how these interact across contrasting governance and communication ecologies. Prior work also demonstrated that wildfire smoke harms respiratory health and that communication influences compliance; however, existing evidence largely assumes linear, top-down pathways from institutions to public behavior. Collectively, there is a necessity for audience-specific, coordinated, and evidence-based communication strategies to strengthen public resilience and health outcomes in wildfire-prone regions.

This study investigates three aspects: (1) how combinations of communication-related conditions influence wildfire preparedness and psychological resilience; (2) the relationship between individual preparedness, health outcomes, and perceptions of government performance as shaped by individual experience and sources of information; and (3) the impact of perceived evacuation safety, trust in authorities, and preparedness behavior on self-reported physical and mental health outcomes in wildfire-affected regions of Canada, Spain, Hawaii, and the USA. By focusing on both the psychological and physical dimensions of health and integrating quantitative and configurational methodologies, this study contributes to a more nuanced understanding of the intersection between communication and governance in determining health outcomes within wildfire-prone regions.

This study uniquely integrates multiregional survey data from Los Angeles, Hawaii, Canada, and Asturias with both Poisson GLMM and fsQCA to identify configurations of communication, preparedness, and governance that shape health outcomes and perceived safety. By explicitly modeling how formal systems (e.g., evacuation plans, air-quality monitoring, official alerts) combine with informal networks (social media, word-of-mouth), the study fills a critical gap in understanding how decentralized communication and psychological preparedness operate as functional substitutes or complements to institutional capacity across settings.

Community perceptions, health impacts, and emergency response

Across all regions surveyed, respondents displayed relatively low confidence in local government preparedness and communication about wildfires. Less than a third of respondents assigned positive ratings. Only 29% rated preparedness efforts as effective, and 22% rated communication efforts positively (Table S4). While 45% reported that local communication was timely and clear, only 13% felt that evacuation plans were highly accessible. Overall, the public’s perception of local wildfire management was also low, with only 10% agreeing that governments effectively raised risk awareness and 18% felt that they received sufficient individual preparedness support (Table S4). Communication efforts were rated inadequate by 39% of respondents (Table S4).

Participants identified key drivers of wildfires, with drought (63%) and high winds (74%) being most frequently cited as natural causes. Human negligence was considered very important by 62% (Table S5). Health consequences were widespread: 61% reported physical symptoms, such as throat irritation and coughing, and 58% experienced respiratory issues (Table S6, Fig. S2). Although 56% had evacuated at least once, only 44% found evacuation plans to be effective or communication to be sufficient (Table S7).

Regarding communication channels, both government alerts (56%) and social media (58%) were considered the most useful, while word-of-mouth and traditional media were moderately effective (Table S8). Participants rated direct interventions such as masks (46%) and air filtration (40%) as more effective than long-term strategies (Table S9). Detailed descriptive results are provided in the Supplementary Section.

Statistical associations and modeling

Participants (n = 156) who perceived the evacuation plan as safe were significantly less likely to report respiratory problems (χ² = 43.99, p < 0.05). The perceived effectiveness of air purifier distribution to vulnerable populations was associated with a reduction in respiratory symptoms (χ² = 20.97, p < 0.05) (Table 1).

Communication clarity was influential in shaping public behavior and trust. Participants who valued social media as a valuable information source were significantly more likely to perceive official communications as clear and straightforward (χ² = 23.00, p < 0.05). Community engagement (e.g., discussing with neighbors) was linked to a better understanding of wildfire messaging. The perceived usefulness of government communication and the usefulness of social media were significantly associated (χ² = 6.02, p < 0.05).

Participants who recognized climate change as a contributing factor to wildfires were more likely to view government initiatives related to health and air quality as effective (χ² = 36.75, p < 0.05). In addition, the Poisson GLMM revealed that the use of social media as an information source during wildfires was significantly associated with participants’ perceived wildfire impact (β = 0.030, p < 0.001) (Table 1), whereas attributing wildfires primarily to natural causes showed a small negative association with perceived impact (β = -0.009, p < 0.05) (Table 2). Perceived positive government performance in managing wildfire-related health and air quality had no statistically significant effect on perceived impact of wildfire-related health and air quality (β = −0.003, p > 0.05).

Qualitative comparative analysis results

A necessity assessment (Table 3 and S10) showed that effective wildfire governance was closely linked to a high degree of risk awareness and adequate, timely, and clear communication. These factors were especially critical to participants in Los Angeles and Canada, where high-consistency (e.g., 0.943 in Canadian configuration (1) and high-coverage (0.61) configurations indicated that risk knowledge and communication were key requirements for good governance (Table 3, S11–14). In contrast, Asturias faced institutional limitations and relied on informal preparedness networks, which restricted the availability of core resources and hindered governance efficiency.

Sufficiency analysis also highlighted key features of wildfire preparedness. Risk awareness, information availability, and timely communication (core conditions) consistently appeared in high-consistency designs in Los Angeles and Canada. The availability of personal preparedness resources and community-wide risk reduction efforts (peripheral conditions) also existed, but were not evenly distributed, suggesting they could act as contingent boosters rather than fundamental components. Hawaii and Asturias configurations relied less on core conditions and more on peripheral or mixed conditions (Tables S13 and S14), representing adaptive governance strategies in environments with low institutional consistency.

Los Angeles and Canada evaluated air quality monitoring systems, public health advisories and warnings, and clean air shelters differently, reflecting variations in institutional strategies and public health capacities (Tables 4, S15). In regions with strong institutional capacity (such as Canada and Los Angeles), public health and environmental systems were effective in mitigating wildfire-related adverse health effects. However, respondents from these sites agreed on primary institutional responses, including the distribution of air purifiers, government-funded research, and emergency response units with filtration systems (Tables S16–S17).

Peripheral solutions, such as limiting outdoor activities and the distribution of face masks, did not demonstrate universal presence but contributed to increased capacity. In Hawaii and Asturias, configurations suggested inadequate implementation in resource-scarce, less consolidated contexts.

In Canada and Los Angeles, official government channels, traditional media, and evacuation communication clarity, indicating evacuation plans and multi-platform communication, significantly improved preparedness and promoted behavioral compliance (Tables 5, S18).

Hawaii exhibited a distinct communication profile, where word of mouth and social media platforms attained the highest consistency scores (Table S18).

This study describes how communication, perception, and governance shape wildfire preparedness and health outcomes. Confidence in evacuation plans was significantly associated with reduced respiratory symptoms, indicating the health benefits of psychological preparedness. In areas like Hawaii and Asturias, where institutional and technological resources such as official alert systems and emergency planning infrastructure are relatively limited, informal communication, particularly social media and word-of-mouth, was considered more reliable than formal government channels. This suggests the need for decentralized emergency communication strategies. Although modeling did not identify a statistically significant correlation between perceived government performance and perceived effectiveness of wildfire response efforts, fsQCA findings indicated the importance of government-related conditions in shaping wildfire preparedness. This was particularly evident in regions such as Hawaii and Asturias. This suggests that the impact of government perception may be context-dependent, potentially becoming more pronounced in environments characterized by weak institutional frameworks.

One of the most policy-relevant findings of this study is the high reliability of informal communication systems, particularly social media¹² and word of mouth, in less-resourced settings where institutional infrastructure is weaker. This contrasts with regions that have more robust infrastructure, where clear, timely communication enhances preparedness. In Hawaii, for example, peer-based communication outperformed formal systems, challenging conventional top-down emergency communication strategies. However, while social media platforms like X and Facebook enabled rapid updates during crises, they also created confusion, particularly with inconsistent access, as seen in Lahaina and Los Angeles^13,14. Furthermore, while these platforms and peer networks facilitate rapid information sharing and interaction within the community, they can also exacerbate misinformation and confusion, particularly within the context of decentralized communication. Reliance on informal networks in resource-limited areas may be more adaptive than stable¹⁵. However, decentralized, context-specific approaches that prioritize community trust and local engagement clearly provide important ways to bridge infrastructure gaps, which deserve further research. Another finding, that social media use heightened perceived wildfire impact, suggests its potential power in increasing wildfire awareness.

Our findings underscore the importance of psychological preparedness in shaping health outcomes during wildfire events. The positive associations between perceived institutional initiatives (e.g., safe evacuation, equitable air purifier distribution) and fewer respiratory symptoms suggest that, beyond the physical infrastructure of emergency systems, public confidence plays a crucial role in health protection. Preparedness strategies, transparent risk communication, participatory planning, and community-based drills may enhance trust and encourage protective behavior¹⁶.

We found widespread concerns about infrastructure resilience and personal safety, especially in high-risk evacuation areas¹⁷. In Los Angeles and Asturias, participants cited gaps in planning and insufficient preparedness resources. Respondents frequently criticized local governments for inadequate communication and limited support for individuals, echoing broader critiques that urban planning rarely prioritizes wildfire resilience¹⁸. While community-wide initiatives and individual preparedness efforts improved adaptability, they were not sufficient on their own¹⁹. Developing context-sensitive evacuation strategies should integrate formal systems with trusted informal networks, particularly in regions with limited institutional capacity²⁰.

Decentralized communication aids governance in low-resource settings but may reduce coherence, coverage, and reliability compared to institutional systems²¹. Our analysis revealed a counterintuitive but critical insight: perceived government effectiveness in managing wildfire-related health and air quality was not significantly associated with individuals’ sense of being impacted. Instead, subjective factors, including personal experience, social media, and interpersonal interactions, were more influential predictors of perceived impact, possibly due to the influence of personal narratives and emotionally charged content. These findings challenge the assumption that institutional performance alone mitigates perceived vulnerability²². They underscore the need for a dual approach that strengthens institutional response to address emotional and cognitive drivers of risk perception.

These findings align with existing literature identifying human activity and climate change as primary drivers of wildfire risk¹. Wildland-urban interface zones heighten the likelihood of property damage while presenting logistical challenges for emergency response and evacuation²³. We found that individuals who attributed wildfires to climate-related factors were more likely to believe government health interventions are effective, suggesting that environmental literacy can foster institutional trust. When people view wildfires within a broader ecological and climatic framework, they may interpret institutional actions as more legitimate and impactful³. Educational campaigns that clearly connect climate change to wildfire risk could enhance public engagement.

Our study found better concordance between direct and tangible protective measures (such as masks, air filtration, and clean-air shelters) than between long-term or systemic interventions. While systemic solutions are essential for long-term resilience, their benefits may not be readily visible to the public during acute crises²⁴. This discrepancy suggests a gap between public priorities and institutional investments. Bridging this perception gap requires better communication strategies that clearly link systemic interventions to perceived personal and community health outcomes.

Clear communication was essential for institutional trust and effective preparedness. Social media and neighbor-based discussions enhanced the acceptance of official messaging, highlighting the value of integrated, multimodal communication networks. Wildfire communication strategies that blend digital, interpersonal, and traditional channels may enhance reach and credibility across diverse populations²⁵.

The health impacts of wildfires were widely reported²⁶. Exposure to wildfire smoke is associated with respiratory and cardiovascular risks, especially among vulnerable groups²⁷. Post-wildfire runoff introduces contaminants into drinking water^28,29. Settings with stronger infrastructure could mitigate these effects through coordinated interventions (e.g., air quality monitoring, advisories, clean-air shelters), supported by institutionalized, multisectoral systems³⁰. In contrast, less supported areas depend on adaptive, informal or mixed strategies, which may lack resilience and scalability³¹.

In this context, this study highlights the strength of examining both institutional capacity and community adaptation. Regions with well-developed infrastructure demonstrated timely communication and robust public health systems that were perceived as effective wildfire governance. While regions with less-developed infrastructure did not, mitigating factors, including flexible informal networks, were perceived as valuable, underscoring the need for integrated, context-sensitive, multi-channel preparedness strategies. The cross-sectional design captures a timely snapshot of perceptions, revealing critical gaps and strengths in governance and community response.

Despite these strengths, the sample size varied across regions and relied on self-reported data. The convenience and snowball sampling methods may have introduced selection bias. Not all questionnaires were complete; although validated strategies were used to handle missing data, errors may have been introduced. Regional comparisons were exploratory due to the smaller sample size outside the Los Angeles region. Despite these constraints, the findings contribute context-specific knowledge to inform more adaptive and inclusive wildfire management strategies.

In conclusion, this study provides a multi-scalar, multi-method examination of how public perceptions, communication strategies, and institutional responses shape wildfire preparedness and health outcomes. Our findings underscore the urgency of integrating decentralized, context-specific communication, strengthening psychological preparedness, and aligning institutional interventions with public expectations. Given the increasing frequency and intensity of wildfires under climate change, policymakers must prioritize inclusive, adaptive, and community-informed approaches to emergency management. Investments in social infrastructure, particularly in communication networks, trust-building mechanisms, and participatory planning, are as critical as investments in physical infrastructure in mitigating the health and societal impacts of wildfires.

Participant information and sampling

This cross-sectional study collected data from participants across Ontario, Quebec, Alberta, and British Columbia (Canada: n = 13); Asturias (Spain: n = 15); Maui and Oahu (Hawaii: n = 13); and Los Angeles (n = 115). Wildfire experts were identified through websites, news articles, academic profiles, social media, and relevant organizations, with a focus on regions at risk of wildfire. Potential participants were invited via email and encouraged to share the survey within their networks. Further dissemination occurred on social media (Twitter (X), Facebook). In Asturias, individuals were approached directly through personal connections. The survey link was publicly distributed, making an accurate response rate impossible.

Questionnaire development and expert validation

The 77-item questionnaire was refined through review by wildfire content experts to ensure clarity, cultural relevance, and accuracy. It was designed to assess perceptions of various aspects of wildfires, including preparation, health impact, risk perception, government performance, and communication efficiency. The instrument included 57 closed-ended questions organized into thematic blocks: demographics, causes, evacuation experience, health and environmental effects, and institutional performance, incorporating a five-point Likert scale. The anchors for Likert scales varied depending on the question; for example, some items assessed agreement (from strongly disagree to strongly agree), while others measured effectiveness (from very effective to not effective). The questionnaire was available in English and Spanish, primarily distributed via Google Forms, with a printable version available on request. It was pre-tested to ensure relevance and accessibility.

Data collection

The 77-item questionnaire was designed to collect information on demographics, wildfire concerns and impacts, causes, evacuation plans, health implications, air quality, prevention and control, and open-ended reflections. Variables were selected for their theoretical relevance and alignment with study objectives and were transformed into fuzzy-set scores through the calibration of their qualitative anchors, followed by a configurational analysis to identify region-specific patterns.

Data analysis

Data were coded and organized using an Excel spreadsheet. Frequency distributions and percentages were used to identify patterns aligned with the study objectives.

Chi-square tests were used to examine associations between perceived evacuation safety, communication tools, health impacts, and government response effectiveness. Methodologically, Likert-scale responses were dichotomized (4–5 = “yes” or “favorable response; 1-3 = “no” or “unfavorable response”) to ensure accuracy in the context of imbalanced data across regions. Chi-square p-values were computed using a Monte Carlo simulation (2000 replicates) for all tests. Findings were deemed significant if the alpha threshold was not exceeded (p < 0.05). Missing values (<3%) were addressed through item-wise median substitution, as multiple imputation was not required because of the low proportion of missing values.

A Poisson GLMM was applied (glmer function, lme4 package) to assess factors influencing individual wildfire experiences, using the Nelder–Mead optimization algorithm to ensure model convergence. Predictor variables included participant perceptions of government response to health and air quality challenges, social media engagement, and views on wildfire-related factors. A random country effect was included to control for regional response differences.

The research utilized a convenience sampling method, drawing respondents from four distinct regions. Despite the sample size being relatively small and disproportionate to the geographical distribution, a statistical correction was applied using Monte Carlo simulations for Chi-square tests and random country-level effects in GLMM. This approach enhanced the robustness of the findings. Consequently, the results are exploratory and aim to highlight indicative regional patterns rather than generalize to the broader population. Chi-square and Poisson GLMM analyses were performed on the full data set (n = 156). Quantitative analysis is described in the Supplementary Materials.

In GLMM, the composite variable representing perceived wildfire impact was constructed from four items measured on a Likert scale: perceived air quality, perceived stress, perceived health disruption, and perceived displacement risk. An average of the responses was calculated to create an overall index, which was subsequently dichotomized for use in regression analysis ( ≥ 4 = high perceived impact; <4 = low perceived impact). No control variables were included in this analysis, as the focus was on perceptual and behavioral factors related to the effects of wildfires rather than demographic factors.

Fuzzy-set Qualitative Comparative analysis (fsQCA) was performed separately for each region (Canada, Hawaii, Spain (Asturias), and USA (Los Angeles)) to capture regional-context insights. The outcome variable was perceived effectiveness of wildfire governance, while eight condition measures related to community preparedness, health impacts, and evacuation communication. For analysis, min-max normalization methodology was applied to raw values to generate fuzzy-set membership scores, ranging from 0 (complete non-membership) to 1 (complete membership). Scores of 0.5 were reassigned to 0.51 to eliminate ambiguity during calibration, in line with fuzzy set research best practices³². This calibration process was necessary to record the presence or absence of each condition and the extent to which each condition was affected, making the subsequent fsQCA specifically responsive to qualitative reasoning and quantitative detail³³. The fsQCA process was executed using R Studio with relevant QCA packages^34,35.

Every condition in the context of fsQCA is a contributory factor, and the outcome is perceived government efficiency in controlling wildfires in the current study. The technique identifies sets (patterns) of conditions that are adequate or necessary for the result, as opposed to estimating the impacts of individual variables, as in regression. In small-N areas (such as Asturias or Hawaii), the findings should be interpreted as tendencies rather than definitive conclusions. Consistency and coverage provide directional evidence but may be sensitive to the distribution of cases; therefore, the findings represent a configurational inclination rather than causal confirmation. The conditions and outcomes are shown in Table 6.

A necessity analysis was conducted for each research objective to determine whether specific conditions were consistently present when effective wildfire governance was observed. Consistency and coverage values were calculated for each condition and its negation. A condition was categorized as essential when its consistency score exceeded the standard 0.87³⁶, indicating a high degree of consistency in the presence of the condition across instances where the outcome is positive. Each country’s case study and objective were analyzed separately, allowing for the identification of cross-regional consistencies and context-dependent differences. Both the presence and absence of each condition were analyzed to determine whether the existence of a condition was conditionally necessary for achieving the outcome.

Sufficiency and solution configurations were assessed by generating and minimizing truth tables. Truth tables were constructed based on the calibrated fuzzy data for each case study, illustrating all logically possible combinations of conditions. A configuration was deemed adequate when it achieved a consistency of 0.87 threshold or higher³⁶. To drive immediate solutions, simplifying assumptions were applied whenever feasible, in accordance with directional expectations. In instances where assumptions could not be implemented (due to insufficient empirical variability or the emergence of mutually inconsistent arrangements), more complex solutions were generated. Solution types (intermediate and complex) were documented for each country independently to account for contextual variability in the associations between conditions and outcomes.

Ethical approval

This study was conducted in accordance with the Declaration of Helsinki. Ethical approval was obtained from the Ethics Review Board of COMSATS University Islamabad, Department of Biosciences, CUI/Bio/ERB/01-01, 25 January 2025, and the University of Alberta Health Research Ethics Board (Pro00150334). Informed consent to participate in the study was obtained from participants. No ethics approval waivers were required for this study.