Calibration of satellite measurements of river discharge using a global hydrology model

Measurements of river discharge and watershed runoff are essential to water resources management, efficient hydropower generation, accurate flood prediction and control, and improved understanding of the global water cycle. Previous work demonstrates that orbital remote sensing can measure river discharge variation in a manner closely analogous to its measurement at ground stations, and using reach flow surface area instead of stage as the discharge estimator. For international measurements, hydrological modeling can, in principle, be used to provide the needed calibration of sensor data to discharge. The present study tests this approach and investigates the accuracy of the results. We analyze six sites within the US where gauging station, satellite measurements, and WBM model results are all available. Knowledge is thereby gained concerning how accurately satellite sensors can measure discharge, if the signal is calibrated only from global modeling results without any ground-based information. The calibration (rating) equations obtained for the remote sensing signal are similar, whether based on gauging station or on model information: r² correlation coefficients for least squares fits at one example site (#524; White River, Indiana) are both .66 (n=144, comparing monthly daily maxima, minima, and mean, 2003-2006). Space-based 4-day mean discharge values for this site when using the model calibration are accurate to within ±67% on the average (n=1824; largest percent errors occur at low discharges), and annual total runoff is accurate to ±9%, 2003-2008. Comparison of gauging station versus modeled discharge commonly indicates a small positive model bias; the observed errors of satellite-observed annual runoff are also positive and could be improved by bias removal from the rating curves. Also, analysis of a large flood event, along the Indus River in 2010, shows that the model does not capture flood wave attenuation by overbank flow, and thus predicts faster flood wave celerity and higher peak discharge than was measured by the remote sensing. The incorporation of overbank processes would improve discharge estimation via modeling, and also facilitate more accurate satellite-based measurement of peak discharge. The analysis shows that existing and planned microwave sensors can usefully characterize global river discharge dynamics, and that water balance model-based rating curves provide acceptable calibration of remote sensing signal to discharge.