An official website of the United States government
Each year, wildfires ravage the western U.S. and change the lives of millions of inhabitants. Situated in southern California, coastal Santa Barbara has witnessed devastating wildfires in the past decade, with nearly all ignitions started by humans. Therefore, estimating the risk imposed by unplanned ignitions in this fire-prone region will further increase resilience toward wildfires. Currently, a fire-risk map does not exist in this region. The main objective of this study is to provide a spatial analysis of regions at high risk of fast wildfire spread, particularly in the first two hours, considering varying scenarios of ignition locations and atmospheric conditions. To achieve this goal, multiple wildfire simulations were conducted using the FARSITE fire spread model with three ignition modeling methods and three wind scenarios. The first ignition method considers ignitions randomly distributed in 500 m buffers around previously observed ignition sites. Since these ignitions are mainly clustered around roads and trails, the second method considers a 50 m buffer around this built infrastructure, with ignition points randomly sampled from within this buffer. The third method assumes a Euclidean distance decay of ignition probability around roads and trails up to 1000 m, where the probability of selection linearly decreases further from the transportation paths. The ignition modeling methods were then employed in wildfire simulations with varying wind scenarios representing the climatological wind pattern and strong, downslope wind events. A large number of modeled ignitions were located near the major-exit highway running north–south (HWY 154), resulting in more simulated wildfires burning in that region. This could impact evacuation route planning and resource allocation under climatological wind conditions. The simulated fire areas were smaller, and the wildfires did not spread far from the ignition locations. In contrast, wildfires ignited during strong, northerly winds quickly spread into the wildland–urban interface (WUI) toward suburban and urban areas.
more »
« less
This content will become publicly available on December 15, 2025
Firecast Zigzag Convolutional Network for Wildfire Prediction
Each year wildfires result in billions of dollars in property damage. Being one of the major natural hazards, wildfires nowadays are also a global affair whose negative impact is particularly devastating in developing countries. As wildfires are expected to become more frequent and severe, more accurate models to predict wildfires are vital to mitigating risks and developing more informed decision-making. Artificial intelligence (AI) has a potential to enhance wildfire risk analytics on multiple fronts. For example, deep learning (DL) has been successfully used to classify active fires, burned scars, smoke plumes and to track the spread of active wildfires. Since wildfire spread tends to exhibit highly complex spatio-temporal dependencies which often cannot be accurately described with conventional Euclideanbased approaches, we postulate that the tools of topological and geometric deep learning, specifically designed for non-Euclidean objects such as manifolds and graphs, may offer a more competitive solution. We validate the proposed methodology to predict wildfire occurrences in Greece and several regions of Africa. Our results indicate that the Firecast Zigzag Convolutional Network (F-ZCN) outperforms the current baseline methods for wildfire prediction and opens a path for more accurate wildfire risk analytics, even in scenarios of limited and noisy data records.
more »
« less
- PAR ID:
- 10639310
- Publisher / Repository:
- IEEE
- Date Published:
- Page Range / eLocation ID:
- 1943 to 1949
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Shaw, Rajib (Ed.)This study proposes measuring access to shelters and wildfire risks in tandem rather than in isolation to prevent wildfires from turning into human disasters. By leveraging a human-centered design approach in California, which has an active wildfire history and experience with some of the deadliest wildfires, three key findings are discerned. First, California experiences undesirable resource allocation where regions with a high risk of wildfire are surrounded by regions with a low level of access to emergency shelters, and regions with a low risk of wildfire are surrounded by regions with a high level of access to emergency shelters. Second, neither access to shelters nor wildfire risk is evenly distributed across space. This uneven distribution, however, discriminates against exurban areas. As one moves away from core cities, higher wildfire risk and comparatively limited access to emergency shelters are noticed, underscoring the heightened susceptibility of exurban areas to wildfires. Third, in contrast with existing research solely focusing on wildfire risk, it is revealed that the elderly, people with disabilities, and Hispanics are at a higher risk of experiencing high wildfire risk and low access to shelters. The findings suggest instilling equity into wildfire preparedness strategies while minimizing the gap in access to resources between disadvantaged and advantaged communities, given the trichotomy of exposure to the hazard (risk of wildfire), proximity to aid (access to shelters), and vulnerability to threat (community characteristics).more » « less
-
Abstract The growing frequency and size of wildfires across the US necessitates accurate quantitative assessment of evolving wildfire behavior to predict risk from future extreme wildfires. We build a joint model of wildfire counts and burned areas, regressing key model parameters on climate and demographic covariates. We use extended generalized Pareto distributions to model the full distribution of burned areas, capturing both moderate and extreme sizes, while leveraging extreme value theory to focus particularly on the right tail. We model wildfire counts with a zero‐inflated negative binomial model, and join the wildfire counts and burned areas sub‐models using a temporally‐varying shared random effect. Our model successfully captures the trends of wildfire counts and burned areas. By investigating the predictive power of different sets of covariates, we find that fire indices are better predictors of wildfire burned area behavior than individual climate covariates, whereas climate covariates are influential drivers of wildfire occurrence behavior.more » « less
-
Abstract Previous research has examined individual factors contributing to wildfire risk, but the compounding effects of these factors remain underexplored. Here, we introduce the “Integrated Human-centric Wildfire Risk Index (IHWRI)” to quantify the compounding effects of fire-weather intensification and anthropogenic factors—including ignitions and human settlement into wildland—on wildfire risk. While climatic trends increased the frequency of high-risk fire-weather by 2.5-fold, the combination of this trend with wildland-urban interface expansion led to a 4.1-fold increase in the frequency of conditions conducive to extreme-impact wildfires from 1990 to 2022 across California. More than three-quarters of extreme-impact wildfires—defined as the top 20 largest, most destructive, or deadliest events on record—originated within 1 km from the wildland-urban interface. The deadliest and most destructive wildfires—90% of which were human-caused—primarily occurred in the fall, while the largest wildfires—56% of which were human-caused—mostly took place in the summer. By integrating human activity and climate change impacts, we provide a holistic understanding of human-centric wildfire risk, crucial for policy development.more » « less
-
Background Existing fire spread models focus exclusively on wildland or urban fire simulation. Aims This study aims at an offline coupling of two fire spread models to enable a continuous simulation of a wildfire incident transitioning from wildland into wildland–urban interface (WUI) communities, evaluate the effects of wind input on simulation results and study the influence of building types on fire spread patterns. Methods The selected models are WRF-Fire, a wildland fire behaviour simulation platform, and SWUIFT, a model for fire spread inside the WUI. The 2021 Marshall Fire serves as the case study. A map of the fire’s timeline and location is generated using public information. Three simulation scenarios are analysed to study the effects of wind input resolution and building type on the predicted fire spread and damage. Key results The most accurate results are obtained using a high-resolution wind input and when incorporating different building types. Conclusions The offline coupling of models provides a reliable solution for fire spread simulation. Fire-resistant buildings likely helped limit community fire spread during the Marshall Fire. Implications The research is a first step toward developing simulation capabilities to predict the spread of wildfires within the wildland, WUI and urban environments.more » « less
