Estimation of river bathymetry and implications for regional scale flood inundation simulation on the Inner Niger Delta

Abstract:

Global inundation estimates are needed to answer major questions such as where, when, how often and for how long does freshwater inundate the continental land surface, along with what volume is stored. Over the last three years numerical models have been developed that facilitate global scale simulations of inundation dynamics at <1 km resolution. Fundamental to the accuracy of these models and our understanding of the non-linear transform between runoff and inundation is the bankfull discharge at which the floodplain inundates. However for most of the world’s rivers bankfull discharge is unknown because the necessary data on river geometry are extremely poor or absent. Methods that estimate bankfull discharge using model inversion, given remote observations of surface water dynamics, are therefore needed if simulation accuracy is to improve.

Here, a number of approaches for bankfull discharge estimation from ICEsat and Envistat altimeter data are evaluated using a hydrodynamic model of the Inner Niger Delta in Mali. Previous models of this site parameterised the river channel using downstream hydraulic geometry theory, which related the reach averaged channel width to depth via two calibrated parameters. This approach achieved at best a 1.2 m RMSE to 126 ICEsat overpasses. Directly estimating reach averaged depth for a number of reaches was shown to reduce these errors by over 50%, however it was not clear how this applied to bifurcating channels on the delta. Therefore, a novel approach where geometry was determined based on estimating bankfull discharge return-period was also evaluated.

After calibration to the altimeter data, simulated inundation from 2002-2009 was evaluated against optical imagery from Landsat. After which, historical gauge records were used to simulate the inundation history from the 1960’s and 70’s when peak discharge on the Niger was over twice that of the past two decades because of different land use practices upstream. The analysis significantly extends the inundation record that can be obtained from the remotely sensed archive alone.