Continental-scale prediction of hydrologic signatures and processes

Abstract. Understanding how dominant hydrologic processes and their drivers vary across diverse continental-scale landscapes is critical for hydrologic modeling and water management applications. Our research addresses this question by synthesizing large-sample watershed datasets, Caravan and GAGES-II, and developing random forest models to identify patterns in hydrologic behavior. We assessed dominant processes by examining hydrologic signatures—summary indicators of watershed behavior derived from hydroclimatic time series and random forest models across 14,146 gauged U.S. watersheds. The results reveal clear continental-scale gradients in hydrologic processes, including baseflow, overland flow, storage, and water balance losses. Our map of dominant processes highlights, for example, the transition from baseflow to fast responses and back to baseflow along the elevation gradient from the Appalachian spine, through the Piedmont, to the Eastern Coastal Plain; a distinct outer ring around the Great Lakes region; and sharp contrasts between coastal and inland processes in the West. Variable importance analysis from random forest models show that processes in the western U.S. are primarily controlled by climate, whereas in the eastern U.S., soil, geology, and topography play larger roles, with distinct human influences apparent in urban areas. Our estimates of dominant processes and their drivers provide a framework to extend process knowledge from research watersheds to the continental scale, assess current hydrological understanding, and evaluate hydrological model structures.