Spatial datasets of probabilistic wildfire risk components for the conterminous United States (270m) for circa 2011 climate and projected future climate circa 2047

The Large Fire Simulation System (FSim) simulates the growth and behavior of hundreds of thousands of fire events for risk analysis using geospatial data on historical fire occurrence, weather, terrain, and fuel conditions. It can be used to model the frequency and intensity of fires across large spatial and temporal scales. We simulated fire activity in FSim across the conterminous United States with a 2020 landscape (LANDFIRE) and under two sets of climate conditions: 1) using recent climate patterns from 2004-2018 and 2) with modeled future climate conditions for 2040-2054 to address how fire activity may change under future climate. The purpose of this research is to address how climate itself is expected to impact fire activity. Changes in climate will impact the number of days with conditions that promote burning and affect the intensity of burning conditions, which will impact ultimate fire activity and behavior. The data presented here represent modeled burn probability (BP) and conditional flame length probabilities (FLPs) for the conterminous United States (CONUS) at a 270-meter grid spatial resolution. Flame-length probability is estimated for six standard Fire Intensity Levels (FIL). The six FILs correspond to flame-length classes as follows: FLP1 = < 2 feet (ft); FLP2 = 2 - < 4 ft; FLP3 = 4 - < 6 ft; FLP4 = 6 - < 8 ft; FLP5 = 8 - < 12 ft; FLP6 = 12+ ft. Since they indicate conditional probabilities (i.e., representing the likelihood of burning at a certain intensity level, given that a fire occurs), the FLP data must be used in conjunction with the BP data for risk assessment. All calibration settings and input data used in this analysis, such as vegetation and fuels, were the same as those used in the prior 2020-landscape 2011-climate vintage national FSim run, Dillon et al. (2023) (referred to in the remainder of this document as either the "2020 national run" or the "2011 climate run"). The 2020 national run published here is distinct from the preceding one in two ways: 1) in the previously published version, burnable pixels that did not burn during any simulations were backfilled with the low burn probability of 0.00008 and were assigned flame length probabilities; in this publication we do not alter outputs in this way; and 2) a complete set of flame length and arrival time data are available for all simulated fires in this version. Because of stochasticity in FSim, the two 2020 national runs have minor differences in some areas. The weather streams used in the circa 2047 climate simulation were updated to a 15-year climate period centered on 2047 using projected shifts in monthly temperature, precipitation, and relative humidity from an ensemble of six General Circulation Models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5). Burn probabilities and flame length probabilities from this circa 2047 run were compared to those from the 2020 national FSim run to assess expected changes at the scale of counties and pyromes, or areas of homogeneous fire regime.
National-scale assessment of wildfire risk offers a consistent means of evaluating threats to valued resources and assets, thereby facilitating investments in management activities that can mitigate those risks. We used a simulation system to estimate the probabilistic components of wildfire risk across the nation. These outputs have been generated to support a number of national planning and risk assessment efforts. Climate-conditioned runs (c2047) were generated to simulate the effect of expected climatic changes on fire activity. These data have direct importance for disaster preparedness planning at a national scale. These data may also address how drivers of fire impact simulated fire activity.
These data are a newer version of the Short et al. (2016, 2020) data publications. This modified version is based on circa 2020 landscape data, which were the most current LANDFIRE products available at the time of production. The methods used to generate these data generally followed the same process used in Short et al. (2016, 2020), with improvements made at specific steps. The process steps outlined in the Data Quality, Lineage section of this metadata document are expanded to more fully explain each step and provide additional details on methods for this version of the data. Beyond the newer input landscape data from LANDFIRE, we also used updated datasets for other inputs such as fire occurrence, observed gridded daily weather, and wind data from weather stations. To better capture recent climate conditions, we also shortened the time period of historical weather records used to inform the generation of simulated weather streams for simulation runs, using the most recent 15 years this time (2004-2018) rather than full record from 1992-2012 in the second edition (Short et al. 2020).