The influence of climate and land cover on wildfire patterns in the conterminous United States

Abstract: The occurrence of wildfires is greatly dependent on an ecoregions typical climate and land cover type. To investigate whether climate or land cover primarily lead to wildfires in the conterminous United States, wildfire events from 1999- 2010 are analyzed. Wildfires are divided into years and eco-divisions, a sub-form of ecoregions. To assess whether warmer and dryer divisions are more severely affected by wildfires, the characteristic fire size is determined for every division. Smaller to medium sized fires are found to contribute most to the area burnt but no relationship between the fire size and the climate could be found. The Nesterov Fire Index, Growing Degree Days and precipitation are calculated for all years at a 0.5° resolution and are pooled together with elevation, the fractions of land cover classes and the area burned per cell. Envelopes are created for all factors to assess the threshold from whereon fire bigger than 700m2 are not affected by fire anymore. Fire events seem to occur randomly, whereby evergreen forest and shrub land are identified to burn easily. A Pearson correlation is performed between the parameters but only weak correlations are found. A weighted logistic regression is then carried out to test if more significant results are present when applying a GLM-model. Only slightly better correlations are found whereby the Nesterov Index scores as the best factor. A model selection is then done to inspect which factors explain the occurrence of wildfires best. Again the Nesterov Index scores as the best predictor, followed by the "others" land cover class (infrastructure, barren land, water bodies), evergreen forest and the GDD. The impact of these factors is not strong enough to conclude that climate or land cover is determined to be the dominant factor causing wildfires. However, climate sets the frame on where fires might occur and where they certainly do not. More factors over a longer time period and on a smaller scale must be taken into account to predict the wildfire occurrence.