Climate and fire historical data
May-October VPD measures the dryness of the atmosphere in peak wildfire season. It indicates the fuel dryness and, in regions of sufficient vegetation such as forests, the fuel load. In such regions the climate has a direct influence on wildfire events (McKenzie and Littell, 2017) and are referred to as aridity-limited or flammability-limited (Littell et al., 2018).
In non-forested regions that consist primarily of shrub, grasslands and/or desert, the climate has more of a facilitative influence on wildfire where for example the previous year’s precipitation determines the current year’s production of fine fuels. Such regions are known as fine fuel limited (Littell et al., 2018).
Seasonal VPD and precedent precipitation should therefore be robust predictors of annual burned area in aridity-limited and fine fuel limited regions respectively, and both variables together may be necessary in some regions. The precise relationships between these predictors and burned area will also depend on non-climatic factors such as geographic location, local vegetation, and local fire management.
To develop such relationships, daily VPD and precipitation datasets were obtained from the University of Idaho Gridded Surface Meteorological Data (GridMET) and analyzed. These are 4-km gridded datasets that span the continental U.S. for the period 1979–2020. These observational datasets blend spatial attributes of climate from PRISM1 with desirable temporal attributes from regional reanalysis (NLDAS-2) (Abatzoglou, 2011). Daily VPD is defined here as the difference between daily mean es (the vapor pressure at saturation – a function of air temperature) and actual daily mean vapor pressure ea, where mean es is the mean of the saturation vapor pressure for the minimum and maximum daily temperatures, and ea can be derived given the relative humidity and es.
Daily VPD and precipitation were first temporally aggregated to give monthly mean VPD and total precipitation grids. Monthly VPD and precipitation values were computed for each ecoregion by spatially averaging VPD grids and spatially accumulating precipitation grids. For each ecoregion, the average VPD over May to October was calculated to give a seasonal VPD timeseries (1984-2020) while the total precipitation over January to December was calculated to give annual precipitation timeseries (1983-2019). The seasonal VPD and annual precipitation timeseries were then scaled by the average 1984-2020 seasonal VPD and average 1983-2019 annual precipitation respectively. This scaled 1983-2019 precipitation timeseries is the precedent precipitation for the 1984-2020 wildfire seasons.
Scaled seasonal VPD and precedent precipitation datasets were also prepared using gridded observations (CRU TS Version 4.06) made available by the Climate Research Unit (University of East Anglia) and Met Office (Harris et al., 2020). These datasets helped to validate and select the GCMs used to create the future climate projections.
Burned area data were sourced from the Fire Occurrence Database (FOD) and Monitoring Trends in Burn Severity (MTBS). Both datasets include prescribed or planned fires. The FOD includes every reported fire, regardless of size, during the period 1992-2020 in the U.S (Short, 2022). The MTBS data includes all known large fires within the U.S. during 1984-2020 (MTBS, 2022), defined as at least 1,000 acres for the Western U.S. and at least 500 acres for the Eastern U.S. The original data were cleaned, and burned areas were aggregated by year for each ecoregion for both datasets. Annual burned area was then divided by the corresponding ecoregion area to obtain the annual burn fraction.