Improving the reliability of large-scale hydrological models with satellite observations

Over the past three decades, large-scale hydrological models have gained popularity due to the need to support water resources management at the regional and continental scales. One of the most challenging tasks for developing such models is the availability of data. The presence of human-water interactions, especially reservoir operations, can influence the model parameterization, while measured discharge and/or water levels along the rivers are necessary to the calibration purpose. However, such information is often unavailable. In particular, data on reservoir storage or river discharge are often not measured or shared between the riparian countries of transboundary rivers. A potential solution for this challenging task lies in satellite observations. Specifically, reservoir storage/release and river discharge/water level can be inferred from satellite images (Landsat/Sentinel-1/2) and/or altimetry data (Jason/Sentinel-3). In this study, we take advantage of remote-sensed data to improve the accuracy of a hydrological-water management model (VIC-Res) setup for the northern portion of the Mekong River Basin. Our modeling framework combines VIC-Res with an automated calibration procedure (based on a multi-objective evolutionary algorithm) that explicitly accounts for key water management decisions—inferred from satellite data—occurring within the basin. Results show that the use of such data largely improves the performance and reliability of the model.