Improving Streamflow Predictions in the Arid Southwestern United States Through Understanding of Baseflow Generation Mechanisms

Understanding the factors controlling baseflow (groundwater discharge) is critical for improving streamflow predictions in the arid southwestern United States. We used an enhanced version of the Noah‐MP land surface model with advanced hydrological process options and the Routing Application for Parallel computation of Discharge (RAPID) to examine the impacts of process representation, soil hydraulic parameters, and precipitation data sets on baseflow production and streamflow skill. Model experiments combined multiple configurations of hydrological processes, soil parameters, and three gridded precipitation products: NLDAS‐2, Integrated Multi‐satellite Retrievals for GPM Final, and NOAA AORC. RAPID was used to route Noah‐MP‐simulated runoff and generate daily streamflow at 390 U.S. Geological Survey (USGS) gauges. The modeled baseflow index (BFI) was compared with USGS‐derived BFI. Results show that (a) soil water retention curve model plays a dominant role, with the Van‐Genuchten hydraulic scheme reducing the overestimated BFI produced by the Brooks‐Corey, (b) hydraulic parameters (Van‐Genuchten parameters and hydraulic conductivity) strongly affect streamflow prediction, a machine learning‐based Van‐Genuchten parameters captures the USGS BFI, showing a better performance than the optimized National Water Model (NWM) by a median Kling‐Gupta Efficiency of 21%, and (c) incorporating a ponding depth threshold into the land surface models that increases infiltration is preferred. Overall, models with more physically realistic hydrologic representations show a better performance in modeling BFI and thus a better skill in streamflow predictions than the optimized NWM in the dry southwestern river basins. These findings can guide future studies in selecting reliable schemes and data sets (before calibration) to achieve better streamflow predictions as well as water resource projections.