Hybrid Neural Network – Variational Data Assimilation algorithm to infer river discharges from SWOT-like data

Abstract

Abstract. A new algorithm to estimate river discharges from altimetry measurements only is designed. A first estimation is obtained by an artificial neural network trained from the altimetry large scale water surface measurements plus drainage area information. The combination of this purely data-based estimation and a dedicated algebraic flow model provides a first physically-consistent estimation. The latter is next employed as the first guess of an advanced variational data assimilation formulation. The final estimation is highly accurate for rivers presenting features within the learning partition; for rivers far outside the learning partition, the space-time variations of discharge remain accurately approximated however the global estimation presents a potential bias. Indeed, it is shown that if the estimation is based on the hydrodynamics models only, the inverse problem may be well-defined but up to a bias only (the bias scales the global estimation). This bias is removed thanks to the ANN but for rivers in the learning partition only. For rivers outside the learning partition, any mean value (eg. annual, seasonal) enables to remove the bias. Finally, the present hybrid and hierarchical inversion strategy seems to provide much more accurate estimations compared to the state-of-the-art for the considered 29 heterogeneous river portions.

Covariance kernels investigation from diffusive wave equations for data assimilation in hydrology

“…• The 1 st pair (σ z b , σ Ks ) = (0.5h, 10) is consistent with the estimate suggested by Larnier and Monnier [2020].…”

“…2.3.1) and/or the inflow discharge in addition to the bathymetry in order to provide an effective model (see e.g. Larnier and Monnier [2020] for the inversion of the triplet and Honnorat et al [2009]).…”

“…However, alternative estimates are often discussed as mentioned in Mirouze and Weaver [2010] and references therein. In Larnier and Monnier [2020], which treats an inverse problem including the present ones, the correlation lengths are taken empirically.…”

Extraction of roughness parameters from remotely-sensed products for hydrology applications

“…A second case of application will consist in assessing the benefits of using the remote-data-fusion-based friction coefficients as a first guess for the assimilation of synthetic SWOT observations in the HiVDI discharge algorithm (Larnier et al, 2021;Larnier and Monnier, 2020). The performance of such applications will be compared to the current implementation that uses a rough estimation of friction coefficients from the river width.…”

“…These parameters comprise direct priors values of the missing parameters or at least a prior distribution, e.g a mean value and a standard deviation. Henceforth, there is an emerging literature focusing on developing reliable priors (Peirong et al, 2017;Andreadis et al, 2020;Brinkerhoff et al, 2020;Larnier and Monnier, 2020). At first, it seems easier to get an accurate knowledge of the unobserved bathymetry A 0 as it a physically measurable variable.…”