Lawrence Berkeley National Laboratory

Large fires in the western United States become highly probable when dry conditions surpass critical thresholds of vapor pressure deficit (VPDt). VPDt likely differs between human- and lightning-ignited fires, potentially leading to ignition-type varied responses of fire weather risk to natural variability and various anthropogenic forcings, yet a comprehensive quantification remains lacking. Here, through fire observations with ignition types and a machine learning method, we found that human-ignited large fires had consistently lower thresholds (VPDt) across western US ecoregions. Consequently, the annual number of flammable days (when VPD > VPDt) for human-caused large fires was 93% higher on average and increased 21% more rapidly than those caused by lightning during 1979-2020. Through robust statistical detection and attribution of Earth System Models, we found that the anthropogenic greenhouse gas (GHG) emissions predominantly (81%) controlled the human-related flammable day increases, which was 18% greater than the effect of GHGs on the increases in lightning-related flammable days. Such ignition-type varied fire weather risk indicates more large fire-prone conditions for human-regulated fire regimes when GHG emissions are enhancing and ignitions are not limited by fuels.