A Hybrid of Optical Remote Sensing and Hydrological Modeling Improves Water Balance Estimation

Declining gauging infrastructure and fractious water politics have decreased available information about river flows globally. Remote sensing and water balance modelling are frequently cited as potential solutions, but these techniques largely rely on these same in-decline gauge data to make accurate discharge estimates. A different approach is therefore needed, and we here combine remotely sensed discharge estimates made via at-many-stations hydraulic geometry (AMHG) and the PCR-GLOBWB hydrological model to estimate discharge over the Lower Nile. Specifically, we first estimate initial discharges from 87 Landsat images and AMHG (1984-2015), and then use these flow estimates to tune the model, all without using gauge data. The resulting tuned modelled hydrograph shows a large improvement in flow magnitude: validation of the tuned monthly hydrograph against a historical gauge (1978-1984) yields an RMSE of 439 m3/s (40.8%). By contrast, the original simulation had an order-of-magnitude flow error. This improvement is substantial but not perfect: tuned flows have a one-to two-month wet season lag and a negative baseflow bias. Accounting for this two-month lag yields a hydrograph RMSE of 270 m3/s (25.7%). Thus, our results coupling physical models and remote sensing is a promising first step and proof of concept toward future modelling of ungauged flows, especially as developments in cloud computing for remote sensing make our method easily applicable to any basin. Finally, we purposefully do not offer prescriptive solutions for Nile management, and rather hope that the methods demonstrated herein can prove useful to river stakeholders in managing their own water.