Advancing Fire Behavior Modeling For Post-Fire Hazards Assessments
Fire behavior is a complex and highly variable process in both space and time. The interdisciplinary PHIRE team includes USGS scientists from across the Mission Areas, research collaborators from universities and federal agencies, and federal, state, and tribal practitioners and stakeholders.
Post-fire Hazards and Impacts to Resources and Ecosystems (PHIRE)
Wildfire events can have long-lasting impacts on vegetation structures and hydrological processes, affecting wildlife, cultural resources, and post-fire hazard risk. Predicting the short- and long-term effects of wildfire on various natural, cultural, and economic resources remains a challenge for current empirical models. The USGS PHIRE (Post-fire Hazards and Impacts to Resources and Ecosystems) project seeks to enhance our understanding of how pre-fire conditions and fire behavior determine post-fire dynamics. We use stateof the art data collection methods, data products, and fire modeling to improve predictions of vegetation trajectories, wildlife impacts, and hydrological hazard risk, focusing on five wildfire events that occurred in California and Washington in the summer and fall of 2021. As part of that larger project, the fire modeling team focuses on integrating data from these fires into next-generation fire modeling to help predict risk of post-fire debris flows in currently unburned forests.
Fire Behavior and Debris Flow Risk Modeling
Fire behavior is a complex and highly variable process in both space and time. During a wildfire, interactions among weather, fuels structure, and topography – occurring from the scales of centimeters to kilometers, and milliseconds to days or years - affect how a fire spreads, the energy it releases, and its effects on ecosystems and resources. Post-fire hazards such as landslides, debris flows, and flooding are influenced by fire’s intensity (energy release) and its severity (impacts to vegetation and soils). Identifying landscapes that are vulnerable to debris flows and flooding is an important step for forest planning and community and resource protection, but predicting post-fire risk is difficult for forests that have not yet experienced wildfire. Given fire’s complexity, how can we model and predict fire behavior and its relationship to post-fire hazards?
Fire is integrated into predictive models of post-fire debris flows via fire severity information derived from remote sensing, but only for areas that have already burned. Linking predictive fire severity models with the potential severity of fire in currently unburned areas, which includes much of the western U.S., presents a challenge. One way to overcome this challenge is to simulate wildfires using next-generation, physics-based fire models, which produce realistic fire behavior and subsequent heterogeneity in fire outcomes. By working to connect fire behavior simulation to spatially explicit fire severity, we can conduct risk assessment for debris flow in unburned forests. Moreover, simulation models allow managers to investigate potential outcomes under different management strategies, expanding the use of debris flow risk modeling to a more holistic forest planning tool.
Simulating the Behavior and Outcomes of Focal Fires in the PHIRE Project
Because the outcomes of fire are dynamic and heterogeneous, spatially explicit models of fire behavior present a useful way to describe those potential outcomes. We are using QUIC-Fire, a coupled fire-atmospheric model that simulates the physical interactions among fire, wind, fuels, and topography in three-dimensional space. QUIC-Fire captures fine-scale (1-2m) fire dynamics, while being computationally lightweight enough to run on a laptop in near-real time. Using a combination of field and remotely sensed data and statistical and empirically based tools, we have developed a method for building 3D representations of fuels that serve as inputs to QUIC-Fire. This allows us to realistically represent any potential study area across the US, including pre-fire conditions in our focal fires. Canopy fuels are built using FastFuels, a software which uses satellite imagery and LandFire fuel classifications to impute detailed forest data from the Forest Inventory and Analysis Database (FIA), then generating 3D voxelized trees. These tree canopies are then processed by a software platform called DUET, which realistically distributes litter and herbaceous ground cover based on the locations, sizes and species of trees. Since these surface fuels are some of the main drivers of fire spread and behavior, we developed an additional method of adjusting the outputs of DUET to match field observations of fuel loading. When combined, these tools allow us to simulate fires under a range of weather conditions experienced during the wildfire event and obtain a variety of spatially explicit metrics of fire intensity and fuel consumption
Predicting Debris Flow Risk from Fire Severity
Differences in pre- and post-fire vegetation cover (difference in normalized burn ratio; dNBR) are an important input to current models of post-fire debris flow risk. These data are available for already-burned areas, but predicting post-fire debris flow risk before fires have occurred requires us to develop a method for simulating realistic dNBR distributions from QUIC-Fire outputs in currently unburned forests. This dynamic link will allow managers to test how different fuel treatments or levels of thinning or prescribed burning might mitigate severe fire and reduce risk of post-fire hazards. Current models of debris flow were parameterized based on study areas with extents between 25 and 400 ha, which also represents the range of operational domain sizes for QUIC-Fire. To assess the utility of fire behavior modeling in predicting debris flow, we are selecting test sites based on other factors that are important for debris flow. Attributes like steep topography and proximity to drainages can heuristically inform the best places for us to describe the statistical relationships between fire model metrics, fire severity, and debris flow risk.
Next steps: Statistically Modeling Fire Severity from QUIC-Fire
Now that we have developed methods for simulating fire on realistic, three-dimensional fuel beds, our ongoing work is centered around identifying statistical relationships between the patterns we observe in simulations and those observed in post-fire monitoring. Because QUIC-Fire is a physics-based model, it does not make direct conclusions about fire severity (i.e., effects on vegetation and soils), so novel analyses must be conducted to understand these connections. Fire model outputs like fuel consumption, energy release, thermal radiation, and reaction rate can be modeled against observed dNBR distributions, a process that will involve attempting a variety of modeling techniques, including everything from generalized linear models to advanced machine learning. We anticipate that using spatial statistics to describe patterns of severity will be especially important in producing realistic severity predictions that can inform debris flow risk.
Funding for this project is provided by the Robert T. Stafford Disaster Relief and Emergency Assistance Act (42 U.S.C. 5121 et seq.) and supplemental funding acts for Federal disaster relief activities. Through this funding USGS supports recovery efforts in declared natural disaster areas, to aid recovery efforts from widespread wildfires, devastating hurricanes, prolonged volcanic eruptions, and damaging earthquakes. This enables USGS to repair and replace equipment and facilities, collect high-resolution elevation data, and conduct scientific studies and assessments to support recovery and rebuilding decisions.
Post-Fire Hazards Impacts to Resources and Ecosystems (PHIRE): Support for Response, Recovery, and Mitigation
Fire behavior is a complex and highly variable process in both space and time. The interdisciplinary PHIRE team includes USGS scientists from across the Mission Areas, research collaborators from universities and federal agencies, and federal, state, and tribal practitioners and stakeholders.
Post-fire Hazards and Impacts to Resources and Ecosystems (PHIRE)
Wildfire events can have long-lasting impacts on vegetation structures and hydrological processes, affecting wildlife, cultural resources, and post-fire hazard risk. Predicting the short- and long-term effects of wildfire on various natural, cultural, and economic resources remains a challenge for current empirical models. The USGS PHIRE (Post-fire Hazards and Impacts to Resources and Ecosystems) project seeks to enhance our understanding of how pre-fire conditions and fire behavior determine post-fire dynamics. We use stateof the art data collection methods, data products, and fire modeling to improve predictions of vegetation trajectories, wildlife impacts, and hydrological hazard risk, focusing on five wildfire events that occurred in California and Washington in the summer and fall of 2021. As part of that larger project, the fire modeling team focuses on integrating data from these fires into next-generation fire modeling to help predict risk of post-fire debris flows in currently unburned forests.
Fire Behavior and Debris Flow Risk Modeling
Fire behavior is a complex and highly variable process in both space and time. During a wildfire, interactions among weather, fuels structure, and topography – occurring from the scales of centimeters to kilometers, and milliseconds to days or years - affect how a fire spreads, the energy it releases, and its effects on ecosystems and resources. Post-fire hazards such as landslides, debris flows, and flooding are influenced by fire’s intensity (energy release) and its severity (impacts to vegetation and soils). Identifying landscapes that are vulnerable to debris flows and flooding is an important step for forest planning and community and resource protection, but predicting post-fire risk is difficult for forests that have not yet experienced wildfire. Given fire’s complexity, how can we model and predict fire behavior and its relationship to post-fire hazards?
Fire is integrated into predictive models of post-fire debris flows via fire severity information derived from remote sensing, but only for areas that have already burned. Linking predictive fire severity models with the potential severity of fire in currently unburned areas, which includes much of the western U.S., presents a challenge. One way to overcome this challenge is to simulate wildfires using next-generation, physics-based fire models, which produce realistic fire behavior and subsequent heterogeneity in fire outcomes. By working to connect fire behavior simulation to spatially explicit fire severity, we can conduct risk assessment for debris flow in unburned forests. Moreover, simulation models allow managers to investigate potential outcomes under different management strategies, expanding the use of debris flow risk modeling to a more holistic forest planning tool.
Simulating the Behavior and Outcomes of Focal Fires in the PHIRE Project
Because the outcomes of fire are dynamic and heterogeneous, spatially explicit models of fire behavior present a useful way to describe those potential outcomes. We are using QUIC-Fire, a coupled fire-atmospheric model that simulates the physical interactions among fire, wind, fuels, and topography in three-dimensional space. QUIC-Fire captures fine-scale (1-2m) fire dynamics, while being computationally lightweight enough to run on a laptop in near-real time. Using a combination of field and remotely sensed data and statistical and empirically based tools, we have developed a method for building 3D representations of fuels that serve as inputs to QUIC-Fire. This allows us to realistically represent any potential study area across the US, including pre-fire conditions in our focal fires. Canopy fuels are built using FastFuels, a software which uses satellite imagery and LandFire fuel classifications to impute detailed forest data from the Forest Inventory and Analysis Database (FIA), then generating 3D voxelized trees. These tree canopies are then processed by a software platform called DUET, which realistically distributes litter and herbaceous ground cover based on the locations, sizes and species of trees. Since these surface fuels are some of the main drivers of fire spread and behavior, we developed an additional method of adjusting the outputs of DUET to match field observations of fuel loading. When combined, these tools allow us to simulate fires under a range of weather conditions experienced during the wildfire event and obtain a variety of spatially explicit metrics of fire intensity and fuel consumption
Predicting Debris Flow Risk from Fire Severity
Differences in pre- and post-fire vegetation cover (difference in normalized burn ratio; dNBR) are an important input to current models of post-fire debris flow risk. These data are available for already-burned areas, but predicting post-fire debris flow risk before fires have occurred requires us to develop a method for simulating realistic dNBR distributions from QUIC-Fire outputs in currently unburned forests. This dynamic link will allow managers to test how different fuel treatments or levels of thinning or prescribed burning might mitigate severe fire and reduce risk of post-fire hazards. Current models of debris flow were parameterized based on study areas with extents between 25 and 400 ha, which also represents the range of operational domain sizes for QUIC-Fire. To assess the utility of fire behavior modeling in predicting debris flow, we are selecting test sites based on other factors that are important for debris flow. Attributes like steep topography and proximity to drainages can heuristically inform the best places for us to describe the statistical relationships between fire model metrics, fire severity, and debris flow risk.
Next steps: Statistically Modeling Fire Severity from QUIC-Fire
Now that we have developed methods for simulating fire on realistic, three-dimensional fuel beds, our ongoing work is centered around identifying statistical relationships between the patterns we observe in simulations and those observed in post-fire monitoring. Because QUIC-Fire is a physics-based model, it does not make direct conclusions about fire severity (i.e., effects on vegetation and soils), so novel analyses must be conducted to understand these connections. Fire model outputs like fuel consumption, energy release, thermal radiation, and reaction rate can be modeled against observed dNBR distributions, a process that will involve attempting a variety of modeling techniques, including everything from generalized linear models to advanced machine learning. We anticipate that using spatial statistics to describe patterns of severity will be especially important in producing realistic severity predictions that can inform debris flow risk.
Funding for this project is provided by the Robert T. Stafford Disaster Relief and Emergency Assistance Act (42 U.S.C. 5121 et seq.) and supplemental funding acts for Federal disaster relief activities. Through this funding USGS supports recovery efforts in declared natural disaster areas, to aid recovery efforts from widespread wildfires, devastating hurricanes, prolonged volcanic eruptions, and damaging earthquakes. This enables USGS to repair and replace equipment and facilities, collect high-resolution elevation data, and conduct scientific studies and assessments to support recovery and rebuilding decisions.