Combining big-data remote sensing and global hydrologic modelling improves daily discharge estimates across an entire large watershed

Abstract:

Remote sensing has been gaining attention as a novel source of primary information for estimating river discharge, and the Mass conserved Flow Law Inversion (McFLI) approach has successfully estimated river discharge in ungauged basins. However, McFLI currently suffers from two major drawbacks: 1) Existing satellites lead to temporally sparse and spatially discontinuous discharge estimates, and 2) because of the assumptions required, McFLI cannot guarantee flow continuity reach to reach. Hydrological modeling has neither drawback, yet model accuracy is limited in some basins by a lack of discharge observations. We thus combine McFLI and models in a generic data assimilation framework applicable at the global scale. We establish a daily discharge baseline model for 28,998 reaches of the Missouri forced by recently published global runoff data. We estimate discharge via McFLI using ~1 million width measurements from 12,000 Landsat scenes. We then assimilate McFLI into the model and validate at 403 USGS gauges. Results show that assimilated discharges did not impair already accurate baseline flows and achieved median improvements of 24% NRMSE, 0.42 NSE, and 0.16 KGE where baseline performance was poor (defined as baseline negative NSE, 225/403 reaches). We ultimately improved flows at 84% of these originally poorly modelled gauges. Thanks to the hydraulic routing, these substantial improvements occur even though Landsat images only provide McFLI discharges at 1.5% of reaches and on 26% of simulated days. The methods provided in this study are applicable globally including to ungauged basins. This study is one possible template for future Surface Water and Ocean Topography (SWOT) discharge estimation via McFLI and hydrologic models, and our results suggest that this combination of McFLI and state-of-the-art hydrologic models can accelerate understandings of the global hydrologic cycle. SWOT should improve this further by providing additional assimilation variables and spatiotemporal frequency.