An integrated hydrological model to simulate terrestrial water storage in a large river basin: Evaluation using GRACE data

Abstract

Study region

Krishna River Basin, India

Study focus

Integrated hydrological models that simulate surface and subsurface processes, and their interactions in a single framework are essential for effective management of large-scale river basins. This research is aimed at developing an integrated hydrological model for the Krishna River Basin (KRB), India by tightly coupling a semi-distributed hydrological model (SWAT) with a fully distributed groundwater model (MODFLOW).

New hydrological insights for the region

For the first time, the coupled SWAT-MODFLOW model was applied to a large (>250,000 km²), conjunctively managed, complexly interacted river basin. SWAT simulated streamflows and evapotranspiration (ET) for the seven sub-basins agree with gauge flows (R: 0.45–0.85; NSE: 0.2–0.69) and MODIS-derived ET fluxes (R: 0.18–0.81; NSE: −3.18–0.66). However, MODFLOW simulated heads are in less agreement with observed groundwater levels (R: −0.6–0.6) due to spatial averaging. The coupled SWAT-MODFLOW is used to estimate terrestrial water storage anomalies (TWSA) at sub-basin level and validated with downscaled GRACE data at 0.25^° resolution. Model performance is good for the lower reaches (NSE >0.3; R >0.65) with high flow exchanges and is low for the upper reaches due to underestimation of individual fluxes (NSE <0.3; R <0.65). We observed a basin-wide decline in TWSA during the simulation at a rate of 1.45 mm/month. Our findings can help in developing contextual plans for effective management of water resources.