Information on the past and present influences of human activity on fire regimes is valuable for fire management. The Madrean Ecoregion of the southwestern United States and northern México provides a unique setting to study spatial and temporal patterns of wildfire; mountains in the ecoregion are biogeographically similar, but the region is divided in half by the international border, where contrasting land uses and fire management approaches have developed over time and contributed to regional variation in fire regimes. These differences can have profound impacts on the functional connectivity of the many integrated, transboundary ecosystem services that sustain humans and wildlife. To better understand the influence of direct human interaction with ecosystems on patterns of burning, we first stratified the study landscape by anthropogenic biomes. Within each strata, we developed two boosted regression tree models to determine the relative degree to which the spatial gradient in severity (Landsat differenced Normalized Burn Ratio (dNBR)) of recent (1985-2011) wildfires was correlated with 1) bioclimatic variables for contemporary climate (1981-2010 normals) and 2) future climate (RCP 4.5) based on cross validation statistics (cv.mean, % deviance explained). We interpreted the results in relation to the description and geographic distribution of anthropogenic biomes in the region.
Results/Conclusions
Within the Remote Forests and Remote Rangeland anthropogenic biomes, where human populations are small, the gradient in severity had the weakest relationship to present or predicted climate (cv.mean <= 0.28). In contrast, relationships were stronger within other anthropogenic biomes, described as Populated (minor human populations), Residential (substantial populations) or Villages (densely populated), where people live and directly interact with ecosystems. Bioclimatic environment was an important influence on the gradient in burn severity within these strata (e.g., both Populated Forests and Populated Rangelands cv.mean = 0.40 for the normals and futures models). Winter precipitation and Precipitation as Snow emerged as influential factors in the models. In other populated strata, we also observed relatively higher correlations of severity with spatial variation in climate (e.g., Populated Rainfed Cropland cv.mean ≈ 0.55 and Pastoral Villages cv.mean ≈ 0.68 for the normals and futures models). Summer and Annual Heat Moisture Indices had the greatest relative influence in these models. The varied degree of influence by climate under different anthropogenic biomes suggests the need for climate adaptation and regional conservation planning that is carefully tailored to the challenges fire may pose for areas with a history of unique human-ecosystem interactions.