Upstream satellite remote sensing for river discharge forecasting: Application to major rivers in South Asia

Abstract

In this work we demonstrate the utility of satellite remote sensing for river discharge nowcasting and forecasting for two major rivers, the Ganges and Brahmaputra, in southern Asia. Passive microwave sensing of the river and floodplain at more than twenty locations upstream of Hardinge Bridge (Ganges) and Bahadurabad (Brahmaputra) gauging stations are used to: 1) examine the capability of remotely sensed flow information to track the downstream propagation of river flow waves and 2) evaluate their use in producing river flow nowcasts, and forecasts at 1–15days lead time. The pattern of correlation between upstream satellite data and in situ observations of downstream discharge is used to estimate wave propagation time. This pattern of correlation is combined with a cross-validation method to select the satellite sites that produce the most accurate river discharge estimates in a lagged regression model. The results show that the well-correlated satellite-derived flow (SDF) signals were able to detect the propagation of a river flow wave along both river channels. The daily river discharge (contemporaneous) nowcast produced from the upstream SDFs could be used to provide missing data estimates given its Nash–Sutcliffe coefficient of 0.8 for both rivers; and forecasts have considerably better skill than autoregressive moving-average (ARMA) model beyond 3-day lead time for Brahmaputra. Due to the expected better accuracy of the SDF for detecting large flows, the forecast error is found to be lower for high flows compared to low flows. Overall, we conclude that satellite-based flow estimates are a useful source of dynamical surface water information in data-scarce regions and that they could be used for model calibration and data assimilation purposes in near-time hydrologic forecast applications.