River discharge estimation from SWOT satellite using Hybrid Bayesian-Variational method

The Surface Water and Ocean Topography CNES and NASA mission, planned for launch in late 2022, will provide a global mapping of the water surface elevation, width and slope of rivers wider than 100m worldwide. The estimation of discharge using solely the SWOT-type observations together with the existing satellite data has received a noticeable attention recently. The attempts to solve this problem have expectedly confirmed that it is ill posed and additional data could be needed, such as the estimates of the mean, minimum and maximum discharge from the available global scale hydrological databases. However, taking into account the accuracy of such estimates and the issue of their relevance to the study period, the problem remains challenging. We suggest a new estimation method, designed specially to reduce the solution bias. It combines Bayesian and Variational approaches to improve the algorithm robustness and stability. In this method the likelihood function is computed, allowing a useful analysis of equifinality often encountered in discharge estimation problems for ungauged basins. The algorithm is designed for global and/or basin-scale applications given the multi-level structure of the methodology. The latter involves different levels of complexity in terms of the representation of the flow dynamics and therefore different computational requirements. Here, we investigate discharge and bathymetry estimation under strong uncertainties from SWOT simulations and their combination with optical (Landsat and Sentinel 2) and altimetry (Jason, ENVISAT, Sentinel 3) data. The global application involves a Python low-cost algorithm that will be run globally through the Confluence platform implemented by UMass with the SWOT Science Team.

Keywords: Data assimilation, Bayesian estimation, remote sensing, SWOT, hydraulic modelling, discharge estimation, altimetry, optical imagery, rivers