Hyperresolution hydrologic modeling in a regional watershed and its interpretation using empirical orthogonal functions

Hyperresolution (<1 km) hydrologic modeling of regional watersheds is expected to support a broad range of terrestrial water cycle studies, but its feasibility is still challenging due to process, data and computational constraints, as well as difficulties in interpreting the high-dimensional dataset of spatiotemporal model forcings and outputs. We address some of these modeling challenges by extending the application of a physically-based, distributed hydrologic model to the Río San Miguel watershed (3796 km2) in Mexico based on prior efforts that demonstrated the process fidelity at smaller spatiotemporal scales. Long-term (7 year) simulations are conducted at a hyperresolution (∼78 m) over the large domain using parallel simulation capabilities. To address data sparseness, we develop strategies to integrate ground, remotely-sensed and reanalysis data for setting up, forcing and parameterizing the model. Complementary tests with observations at individual stations and remotely-sensed spatial patterns reveal a robust model performance. After building confidence in the model, we interpret the spatiotemporal model forcings and outputs using empirical orthogonal functions (EOFs) analyses. For all model outputs, a large portion (58-80%) of the spatiotemporal variability can be explained by two dominant EOFs, which are related to model forcings and basin properties. Terrain controls on soil water accumulation have a marked impact on the spatial distribution of most hydrologic variables during the wet season. In addition, soil properties affect soil moisture patterns, while vegetation and elevation distributions influence evapotranspiration and runoff fields. Given the large outputs from long-term hyperresolution simulations, EOF analyses provide a promising avenue for extracting meaningful hydrologic information within complex, regional watersheds.