Spatial analysis of streamflow trends in burned watersheds across the western contiguous United States

Abstract

AbstractDespite increasing magnitude and frequency of wildfire, understanding hydrological processes contributing to changes in streamflow is not well examined for the entire western contiguous United States (WCONUS). This study provides insight into whether considering spatially varying watershed characteristics, including burn severity patterns, can better explain streamflow trends at broad spatial and temporal scales. Standard geographically weighted regression (GWR) and multi‐scalar (MS) GWR were benchmarked against ordinary least squares (OLS) regression to determine if spatially varying coefficients could better explain spatially varying patterns and processes of peak flow, low flow, center timing of flow, and flashiness trends in burned watersheds. In most cases, GWR reduced spatial autocorrelation and provided insight into regional factors explaining streamflow trends. Both OLS and GWR models did not find burn severity‐based metrics significant in explaining streamflow trends, except the peak flow trends model that included the spatial arrangement of high severity patches quantified using landscape shape index (LSI). Positive peak flow trends coincided with increasing LSI values in Northern California and negative peak flow trends with increasing LSI in the southern WCONUS. This research could guide regionally specific post‐fire response and inform the development of new spatial metrics that could better represent post‐fire hydrological processes.