Hydraulic Model Calibration Using CryoSat-2 Observations in the Zambezi Catchment

Geodetic altimeters provide unique observations of the river surface longitudinal profile due to their long repeat periods and densely spaced ground tracks. This information is valuable for calibrating hydraulic model parameters, and thus, for producing reliable simulations of water level for flood forecasting and river management, particularly in poorly instrumented catchments. In this study, we present an efficient calibration approach for hydraulic models based on a steady-state hydraulic solver and CryoSat-2 observations. In order to ensure that only coherent forcing/observation pairs are considered in the calibration, we first propose an outlier filtering approach for CryoSat-2 observations in data-scarce regions using a simulated runoff produced by a hydrologic model. In the hydraulic calibration, a steady-state solver computes the water surface elevation (WSE) profile along the river for selected discharges corresponding to the days of CryoSat-2 overpass. In synthetic calibration experiments, the global search algorithm generally recovers the true parameter values in portions of the river where observations are available, illustrating the benefit of dense spatial sampling from geodetic altimetry. The most sensitive parameters are the bed elevations. In calibration experiments with real CryoSat-2 data, validation performance against both Sentinel-3 WSE and in situ records is similar to previous studies, with Root Mean Square Deviation ranging from 0.43 to 1.14 m against Sentinel-3 and from 0.60 to 0.73 against in situ WSE observations. Performance remains similar when transferring parameters to a one-dimensional hydrodynamic model. Because the approach is computationally efficient, model parameters can be inverted at high spatial resolution to fully exploit the information contained in geodetic CryoSat-2 altimetry.