Background
The First Street Foundation has expanded its portfolio of peer reviewed, property specific, climate adjusted physical risk models with the launch of the First Street Foundation Wildfire Model, estimating the risk of wildfire on a property-by-property basis across the United States today, and up to 30 years into the future. This high-precision, climate-adjusted wildfire model provides insights for individual property owners of residential, commercial, critical, and social infrastructure buildings. These results are made available through Risk FactorT™, the first free source of high-quality probabilistic wildfire risk information at the property level available to the public.
The model was developed in partnership with researchers and wildfire experts from First Street Foundation and the Pyregence consortium, including Spatial Informatics Group, Reax Engineering, and Eagle Rock Analytics. This analysis follows the open science approach taken by First Street Foundation for climate-adjusted flood risk.
Methodological Overview
The First Street Foundation Wildfire Model integrates information on fuels, wildfire weather, and ignition into a Fire Behavior Model. The wildfire model requires data on the combustible fuels which may contribute to wildfire across the United States. The 2016 update, Version 2.0.0, of the canonical U.S. Forest Service (USFS) LANDFIRE (LANDFIRE, 2021) fuels dataset at the 30 meter resolution serves as a baseline of this fuels estimate, and that dataset is updated by including additional information of all known “disturbances” between 2016 and 2020 which could modify or change the fuels in a way not captured in the original dataset. These “disturbances” include activities such as recent wildfires, prescribed burns, harvests, and other forest management practices.
Another important and novel update included in the First Street Foundation Wildfire model is the reclassification of homes and other buildings from a “nonburnable” fuel type to a “burnable” fuel type. Typically, homes and other buildings are classified as nonburnable fuel types within LANDFIRE v2.0.0. In order to allow the wildfire behavior model to more accurately estimate how wildfire moves through the Wildland-Urban-Interface (WUI), properties within the WUI must be replaced by a burnable fuel type so as to not block the modeled wildfire spread.
To represent a wide range of possible weather-driven wildfire conditions across the landscape within the simulations employed here, the model utilizes a decade of NOAA weather data, the 2011-2020 Real Time Mesoscale Analysis (RTMA) dataset (NOAA/NCEP, 2022) augmented by data from Oregon State’s PRISM dataset (Parameter-elevation Regressions on Independent Slopes Model; PRISM, 2021).
These weather data include hourly surface wind, air temperature, relative humidity, and precipitation information at the 2.5 km horizontal resolution. This weather data supports a wide range of possible weather conditions, not to recreate any particular wildfire events, but to drive the wildfire behavior model millions of times in a Monte Carlo simulation scheme to derive 2022 wildfire hazard estimates.
Similarly, for 2052 the same weather time series was used to drive the simulations, but the air temperature, humidity, and precipitation were bias-adjusted to 2052 conditions following the CMIP5 RCP4.5 ensemble results. Rather than applying a bias-adjustment to the wind time series for the future climate, the same winds from the 2011-2020 time series were used to drive the 2052 simulations to reduce sensitivity of the model to highly uncertain future predictions of winds.
One of the primary indicators of where future wildfires will occur is informed through data on historical wildfire occurrences. These historical wildfires help to inform where wildfires may occur vis-à-vis the Fire Occurrence Database (FOD) developed by the USDA Forest Service (Short, 2014; Short, 2021). An open source wildfire behavior model was used, ELMFIRE (Eulerian Level Set Model of Fire Spread). This work does not develop new techniques for wildfire modeling, but rather implements computationally efficient and scalable modeling techniques at a high resolution based on existing science, wildfire probability, and hazard modeling paradigms. These scalable techniques make it practical to more easily conduct wildfire simulations at the 30 meter resolution across the entire country, enabling property and building specific assessments of wildfire risk.
For each 30 meter pixel across the country, information is recorded on the distribution and occurrence of burn incidence, flame lengths experienced, and the relative amount of embers which land in the pixel. These provide estimates of:
- Burn probability: the estimated likelihood of the area burning during any single year.
- Fire intensity: estimated flame lengths, including maximum, average, and sum of all flame lengths experienced.
- Ember exposure: the relative amount of embers which land in an area due to nearby simulated wildfires.
View more on the methodology here.
National Overview
Across the country, there are 49.4M properties with minor wildfire risk (with a cumulative burn probability below 1%); 20.2M properties with moderate risk (6% maximum cumulative burn probability); 6.0M with major risk (14% maximum burn probability); 2.7M with severe risk (26% maximum cumulative burn probability); and 1.5M properties with extreme risk (with cumulative burn probabilities of 26% and up). In total, approximately 71.8 million homes have some level of wildfire risk in 2022, growing by 11.1% to 79.8 million by 2050, owing to the impact of a changing climate.
Fire Factor’s range from (1-10) to describe a property’s aggregate thirty-year exposure to wildfire, informed by the parcel’s burn probability at a 30 meter resolution, taking into account not only the burn probability of the current time period but also how the parcel’s wildfire risk changes over the next 30 years with a changing climate.
The unpredictable nature of fire ember spread results in some randomness in the model. To account for the influence of ember spread, an “ember zone” of 300 meters surrounding any estimated ember landing location was used to determine indirect exposure by estimating how far isolated embers may get from dense ember landing areas or a wildfire line.
The properties with parcels which have only this “indirect exposure” are classified as “at risk” with a minor Fire Factor of 2. Meanwhile, properties that do not have this “indirect” exposure or any burn probability in the model are provided a Fire Factor of 1, which represents minimal risk. It should be noted that while properties with a Fire Factor of 1 do not show measurable risk within the model, this estimate may not translate fully into the burn probability of the property in real life.
In addition to ember zone exposure from the kernel function, properties will also be assigned a Fire Factor of 2 when their burn probability is less than 1% cumulatively over the 30 year period. The remaining range of scores are based on burn probability only. For example, for a Fire Factor of 3 the cumulative burn probability ranges between 1% and 3%.
Policy Implications
The hyper-local resolution of the model allows for an extremely granular understanding of wildfire risk, empowering communities, states, and national government actors to take steps to mitigate wildfire risk above and beyond wildfire suppression efforts.
Supporting wildfire suppression at the local, state, and federal levels is among the most expensive wildfire protection efforts, costing the federal government $2.0 billion annually across the U.S. today. Recent estimates from OMB suggest those costs could rise to $2.83 billion under conservative climate change scenarios by 2050, and perhaps to as much as $4.32 billion under higher emissions scenarios (Office of Management and Budget, 2022). States and communities that are capable of suppressing most destructive wildfires today may find their resources stretched thinner and their capacity further challenged by climate-fueled increases in wildfire occurrence.
Enhanced understanding of the specific nature and location of wildfire risk enables communities to more effectively lobby for funding for fuel treatments, prescribed burns, and other wildfire risk mitigation strategies that may be used to reduce risk to houses, businesses, and communities across the U.S., and could help constrain the costs associated with suppression activities. Individual homes and businesses can reduce their vulnerability to wildfires through a variety of actions and strategies (e.g. see insights from the Insurance Institute for Business and Home Safety) in the face of greater exposure to wildfires in the future. But first, this exposure and risk must be quantified and made available to enable such preventative measures to be planned.