Research and Publications
View article.
The escalating frequency and damage of catastrophic WU fires are accelerating faster than social systems can adapt, presenting disruptive and systemic risks. Among the most pressing: the destabilization of the insurance industry. This crisis stems from a failure to accurately capture and model risk in the built environment, including fire spread into development and the structure-to-structure nature of WU fire loss. To effectively translate wildfire hazard into a quantifiable built environment risk, policymakers and researchers need a comprehensive and collaborative systems approach that prioritizes advanced risk modeling, mandates changes to the built environment through stricter codes, and coordinates efforts across all levels of government and industry. How did we get here? A long-standing desire to reduce perceived threats and increase the value of residential development in fire-prone landscapes perpetuated a cycle of fire exclusion. Aggressively suppressing fire in the western United States has allowed excess flammable vegetation to accumulate, exacerbating hazard. Climate change made things worse by increasing fire activity and fire season duration. Landscapes loaded with highly combustible fuels, including the homes themselves, represent a debt that has become due.
View article.
We evaluated the causes and predictability of four extreme wildfire episodes from the 2024–2025 fire season, including in Northeast Amazonia (January–March 2024), the Pantanal–Chiquitano border regions of Brazil and Bolivia (August–September 2024), Southern California (January 2025), and the Congo Basin (July–August 2024). Anomalous weather created conditions for these regional extremes, while fuel availability and human ignitions shaped spatial patterns and temporal fire dynamics. In the three tropical regions, prolonged drought was the dominant fire enabler, whereas in California, extreme heat, wind, and antecedent fuel build-up were compounding enablers. Our attribution analyses show that climate change made extreme fire weather in Northeast Amazonia 30–70 times more likely, increasing BA roughly 4-fold compared to a scenario without climate change. In the Pantanal–Chiquitano, fire weather was 4–5 times more likely, with 35-fold increases in BA. Meanwhile, our analyses suggest that BA was 25 times higher in Southern California due to climate change. The Congo Basin’s fire weather was 3–8 times more likely with climate change, with a 2.7-fold increase in BA. Socioeconomic changes since the pre-industrial period, including land-use change, also likely increased BA in Northeast Amazonia. Our models project that events on the scale of 2024–2025 will become up to 57 %, 34 %, and 50 % more frequent than in the modern era in Northeast Amazonia, the Pantanal–Chiquitano, and the Congo Basin, respectively, under a medium–high scenario (SSP370) by 2100. Climate action can limit the added risk, with frequency increases held to below 15 % in all three regions under a strong mitigation scenario (SSP126). In Southern California, the future trajectory of extreme fire likelihood remains highly uncertain due to poorly constrained climate–vegetation–fire interactions influencing fuel moisture, though our models suggest that risk may decline in future. This annual report from the State of Wildfires project integrates and advances cutting-edge fire observations and modelling with regional expertise to track changing global wildfire hazard, guiding policy and practice towards improved preparedness, mitigation, adaptation, and societal benefit.
View article.
Climate change is impacting wildfires in the contiguous United States; thus, projections of fire danger under climate change have the potential to inform responses to changing wildfire risks. We calculate fire indices for 13 dynamically downscaled regional climate models, then count days exceeding relevant fire danger thresholds, and compare future changes for mid- and late-twenty-first century relative to a historical reference period. We then compare the responses of the fire indices to highlight areas of agreement and disagreement on the sign and magnitude of future change in fire danger days. Many regions in the domain experience increases in the number of days exceeding fire danger thresholds by the midcentury. The regions which exhibit agreement across the simulation ensemble on the sign of change, and the magnitude of that change, vary greatly between indices. The timing and frequency of fire danger days (defined as days exceeding fire danger thresholds) throughout the year change, both in the shoulder season and during existing peaks in fire danger. By the end of the century, most of the domain experiences statistically significant increases in the number of fire danger days. Complex interactions between input variables, and the sensitivities to inputs, affect the response of fire indices under climate change. The projected increase in fire weather risk could place greater demands upon fire management resources, pose elevated hazards for populations exposed to fire, and potentially disrupt landscapes and infrastructure more frequently.
View article.
Wildfire suppression is shaped by a complex interplay of environmental conditions, resource allocation and management strategies. Examining the containment of the 2021 Schneider Springs Fire in the Eastern Cascades of Washington State, USA, we emphasize critical roles of variable selection, representative sampling and suppression-specific factors. Using descriptive, predictive and causal models, we assessed the influence of weather conditions, terrain features, personnel availability, tree canopy cover, fire containment lines, and previously identified ‘best available’ containment features. High vapor pressure deficit and strong winds were consistently associated with declining containment success. Terrain features such as valleys and ridges facilitated suppression operations, while steep slopes posed challenges. Additional personnel improved containment outcomes, though with diminishing returns in descriptive and predictive models. Tree canopy cover breaks enhanced suppression effectiveness, but with declining utility during windy conditions. Containment lines played a pivotal role, whereas the role of pre-identified containment features was context-dependent, likely influenced by broader strategic decisions. Wildfire containment was influenced by multiple variables, and suppression strategies were situationally determined. Causal models provided valuable insights by isolating total effects of primary variables. Findings underscore adaptive fire management strategies that incorporate context-specific information. Future research should integrate fine-scale weather metrics and additional fire behavior drivers that guide effective decision-making during dynamic operations.
View factsheet.
Across all treatment types, forests that underwent fuel treatments had significantly lower fire severity in three of five metrics (crown scorch percent, crown torch percent, and torch height) and 3x more surviving trees.
View review.
As the use of risk products has grown, so has confusion and controversy about how they work, their strengths and limitations, and the distinction between wildfire risk and wildfire hazard. Wildfire risk products vary widely in their design, purpose, and technical details. They may use different definitions of risk, values, data sources, fire metrics and models. This diversity means that risk assessments for the same location can vary by design. This variability does not necessarily indicate that the products are flawed or inaccurate. Rather, it reflects that wildfire risk analysis is a diverse field with evolving techniques that serve different users and applications. As this comparative review shows, there is no universal formula for assessing wildfire risk. For potential users of wildfire risk products, the strength in understanding diverse approaches lies not in direct comparison but in recognizing the unique contributions each makes. It also means that users need to invest time into selecting the right risk product(s) for their specific application.
View article.
Anticipating plausible future ecosystem states is necessary for effective ecosystem management. We use climate analog-based impact models and a co-production process with land managers to project future vegetation changes for the state of Oregon, United States, (2041–2070, RCP 8.5) at a management-relevant spatial resolution (270-m). We explore multiple analog-based methodologies, evaluate analog model performance with contemporary validation, and leverage climate analogs to assess projection uncertainty by quantifying areas where multiple vegetation trajectories are plausible under a single climate scenario. We find that analog-based models performed well at reproducing landscape-level vegetation composition, and moderately well at reproducing vegetation at the pixel level. Our results suggest that 64% of the study area will experience future climate conditions that support different potential natural vegetation types and 59% will experience climates corresponding with different potential plant physiognomic types, compared to reference-period conditions. We project a 60% reduction of mesic conifer-dominated forests with transitions to mixed evergreen forest types. We also project losses to dry forests, cold forests and parklands, with commensurate expansions of shrublands, grasslands, and geographic redistribution of dry forest types. We find that in many areas, several vegetation trajectories are plausible under a single climate scenario. Finally, we provide guidance for using future vegetation projections and uncertainty outputs in management decisions using the Resist-Accept-Direct (RAD) adaptation framework.
View article.
Since the late 1800s, the US government has largely removed Indigenous fire stewardship practices from the landscape by implementing a top-down fire suppression system that criminalized traditional fire practices and denaturalized the role of fire in forested environments. A century of routine fire suppression produced dense, homogenous forests capable of sustaining high-intensity wildfire that exceeds the suppression capabilities of land management organizations in many regions, spurring federal leaders to modify management approaches. As part of this change, numerous federal policies and plans have advocated for further involvement of Native American tribes and incorporation of Indigenous knowledge within management decisions. These initiatives represent opportunities to simultaneously expand tribal burning rights and reduce wildfire risk, but imbalanced power dynamics stemming from the historic and ongoing colonization of tribal nations continue to limit successful collaboration. The nature of these power imbalances is multifaceted, and this paper interrogates the ideological forces that uphold the settler-colonial relationship. We conduct a Critical Discourse Analysis (CDA) to analyze the discourses and frames used by tribal and non-tribal wildfire protection plans (WPPs), noting how different narratives are used to reinforce or contest common perceptions of wildfire and, more broadly, the legitimacy of a fire management system built on wildfire suppression and anti-Indigenous ideologies.
View article.
Extreme wildfire events (EWEs) thereby pose new challenges and limits to managing disaster risk. This refers not only to response operations but also to “conventional” preventive measures such as the creation of buffer zones that may no longer be effective. This paper depicts several limits of conventional wildfire risk management measures towards EWEs and introduces the concept of disaster risk management tipping points (DRM TPs) as critical thresholds that necessitate a revised set of disaster risk management strategies.
Building on a bibliographic review, we depict the novelty of the concept and apply it to selected illustrative examples. We propose that this conceptualization is useful when developing “layered” or diversified risk management approaches for different types of wildfire events including extremes. It may also leverage and shift the discussion around responsibilities in managing risk in terms of public versus individual contributions, the distribution of investments as well as related aspects of justice.
View factsheet.
To understand trends in fire severity and area burned, researchers analyzed satellite images taken before and after each fire in western US forests from 1985 to 2022. They classified areas as high severity when at least 95% of the tree canopy was lost according to a satellite-derived fire severity metric called the Composite Burn Index. To understand climate’s role, they combined three key indicators during the summer fire season (vapor pressure deficit, maximum temperature, and climate water deficit) into a single metric called fire season aridity to capture how hot and dry each fire season was. This same method was used to model future fire conditions, projecting changes in total and high-severity burn areas through the mid-21st century under a 2°C warming scenario.