A water-level based calibration of rainfall-runoff models using satellite altimetry data

Abstract

: Recent studies demonstrated the efficacy of calibrating rainfall-runoff models using continuous measurements of water level in rivers. The water-level based calibration, that implements an inversed rating curve function in conventional rainfall-runoff models and incorporates a small number of regionalized discharge indices in the calibration (hereafter referred to as IRC_reg method), has important implications for extending rainfall-runoff modelling to basins with no discharge observations. However, the method is applicable only to basins equipped with water level sensors if we rely on ground-based observations. In this work, we demonstrate the efficacy of using remotely sensed water level data collected by an altimetry satellite, Jason 2, to calibrate a rainfall-runoff model. The altimeter-based calibration is applied to five study catchments in Australia, resulting in Nash Sutcliffe Efficiency (NSE) values of 0.31-0.66 (excluding one outlier), which are comparable with NSE values of 0.66-0.87 (daily observations) and 0.22-0.62 (10-day observations) for ground-based calibration. The altimetry-satellite-based calibration performance is highly correlated with river width. Previous studies recommended that the cross-sections of rivers along the satellite tracks should be wider than 350 meters to enable Jason 2 to estimate accurate water levels (Dumont et al., 2009; Markert et al., 2019). However, all rivers in this study are narrow rivers with widths ranging from 7 meters to 85 meters, which influence the accuracy of altimetry-based water level measurements and the subsequent calibration performances. Also, the 10-day temporal frequency of the Jason 2 is expected to affect the calibration performance.