Enhanced ensemble-based 4DVar scheme for data assimilation

Abstract

Ensemble based optimal control schemes combine the components of ensemble Kalman filters and variational data assimilation (4DVar). They are trendy because they are easier to implement than 4DVar. In this paper, we evaluate a modified version of an ensemble based optimal control strategy for image data assimilation. This modified method is assessed with a Shallow Water model combined with synthetic data and original incomplete experimental depth sensor observations. This paper shows that the modified ensemble technique is better in quality and can reduce the computational cost.

High-resolution data assimilation through stochastic subgrid tensor and parameter estimation from 4DEnVar

“…4DEnVar (Liu et al, 2008;Buehner et al, 2010) features a 4D ensemble-based covariance field in a variational system. The strategy described here is a direct extension of the scheme discussed in Yang et al (2015). The extension of 4DEnVar for parameter estimation is done by the so-called augmented state vector technique.…”

“…However, these raw data suffer large areas of missing information. Therefore, the same image pre-processing as in Yang et al (2015) has been done in order to fix the issues related to the observation errors' spatial distribution and the background state. For the observation errors: in terms of a point located in the unobserved region, the observation error is set to be dependent on the distance.…”

“…The algorithm in Yang et al (2015) is characterized by an iterative resampling strategy of the prior ensemble and an update of the ensemble sensitivity matrix within the increment minimization steps. It adopts the form of a Gauss-Newton scheme with nested loops that show the relevance of this approach, in particular in the case of a badly known background information and incomplete observations.…”

“…Such underestimation of the ensemble spread is usually related to the unawareness of the uncertainties associated with the dynamic system. Recent studies show that the quality of the ensemble spread can be significantly improved by not only perturbing the initial condition but also some parameters related to the dynamics (Wei et al, 2013;Yang et al, 2015). Those facts advocate the set-up of physically relevant stochastic dynamical models coupling the random production terms and the dissipation term.…”

“…In this article, we propose to use an ensemble-based variational scheme which gathers the advantages of both the variational assimilation schemes and the ensemble-based sequential methods (Yang et al, 2015). In addition to the obvious opportunity of not requiring any adjoint dynamical models, ensemble-based schemes provide other benefits, especially when considering stochastic dynamics.…”