Humans have changed the total number, seasonality, and spatial pattern of fires globally through increased ignitions, alteration of fuels, and climate change. While the importance of a changing climate for large lightning-started fires has been included in fire risk assessments, the direct role of humans in large fires through human ignitions has been overlooked. To identify large fires in the contiguous U.S., we used the Fire Program Analysis fire-occurrence database (3rd edition), which includes 1.73 million fires that were suppressed by a state or federal agency from 1992-2013. We excluded fires with unknown causes and split the remaining database into fires caused by humans vs. lightning. For each Level III ecoregion, we selected the largest 10% of fires regardless of ignition source (hereafter, ‘large fires’). We extracted the 22-year average monthly 100-hr fuel moisture and wind speed data, and LANDFIRE biophysical setting data that corresponded to each large fire. We compared the number of large human- and lightning-started fires, mean fire size, wind speed, fuel moisture, and vegetation classes by ecoregion. We calculated the total number of large fires that started each Julian day of the year, as well as the median day of year for human- and lighting-caused fires.
The size of a large fire varies across ecoregions by three orders of magnitude (mean fire size of 1-10 ha in New England vs. >1000 ha in the western U.S). From 1992-2013, there were ~4x more large fires caused by humans than by lightning across all ecoregions (human: n=129,883; lightning: n=30,635) and the percentage of large human-started fires was much higher in the eastern U.S. than the western U.S. There were 74,704 large human-caused fires below the 2.5th or above the 97.5th percentile of lightning-started fire dates, meaning that humans have added many large fires when lightning-started fires are rare, particularly in winter and spring. Large human started-fires have occurred in areas of significantly higher fuel moisture (p<0.0001) and higher wind speed (p<0.0001) than large lightning-started fires. There are more large human- than large lightning-caused fires in hardwood forests (22.5x), mixed hardwood/conifer forests (12.3x), and savannas (66.8x). We suggest that when assessing large fire risk in the U.S., fire management teams should include human ignition pressures. Furthermore, large fires should be evaluated in a regional context: the area burned is larger in the West, but large fires are correlated with greater population density and infrastructure in the East.