A generic hydrological forecasting system using existing and future altimetry assimilation: an OSSE study over the Niger basin

The main motivation for this study is to evaluate the use of real time observations from different sources for hydrological forecasting. The advent of new satellite missions providing high-resolution observations of continental waters has raised the question of how to use them, especially in conjunction with models. At the same time, the multiplication of extreme events such as flash floods points to the need for tools that can help anticipate such disasters. To do so, it is necessary to set up a forecasting system that is generic enough to be used with different types of data and to be applied to different basins. It is in this perspective that a platform named HYdrological Forecasting system with Altimetry Assimilation (HYFAA) was implemented, which encompasses the MGB large scale hydrological model and an EnKF module that corrects model states and parameters whenever observations are available. As a preliminary study towards operationnability, the platform was tested in offline mode, in the framework of Observing Systems Simulation Experiments (OSSEs). Discharge estimates from three different observing systems were generated, namely in-situ streamflow measurement stations, Hydroweb radar altimetry, and the future SWOT interferometry mission. In this study, we chose to assimilate these data separately in order to analyze the capacity of the system to adapt itself to different orbital characteristics, especially coverage and repetitivity. This also allows us to quantify the contribution of SWOT. The MGB model, developed within the large-scale hydrology research group of the University of Rio Grande do Sul (Brazil), is a physically based and distributed hydrological model, which was coupled to an externalized Ensemble Kalman Filter (EnKF) to give corrected estimates of the model state variables and parameters.

HYFAA is run on the Niger river basin over a reanalysis period and its performance against a control ensemble simulation (without data assimilation) is assessed to quantify the impact of assimilating observations from the different observing systems. The results show that data assimilation leads to significant improvements of NRMSE and KGE of the simulated discharge, everywhere on the basin and regardless of the observation system considered. Moreover, it is shown that the correction of the hydrodynamic parameters helps to improve the performance of the assimilation, in particular when observations are dense in space, probably due to the concomitant correction of forcing biases. The assimilation of SWOT data combined with a selection method provides the best correction of the discharge on the river itself as well as on its tributaries, giving promising perspectives for the prediction of flash floods. We therefore discuss limits and prospects for application in the framework of Observing System Experiments (using real observations).