La fédération
Intranet
Interests
My interests are related to the modelling of uncertainties in the numerical representation of complex natural systems and the reduction of these uncertainties using observational information. The focus of my activities is on oceanographic applications, as illustrated below.
Uncertainties in model simulations
Performing ensemble simulations is like rolling ocean describing dice to see several possible outcomes of the experiment.
From Leroux et al., 2022. Ensemble numerical simulation of the Mediterranean Sea at kilometric-scale resolution. The figure shows snapshots of relative vorticity from two ensemble members (top and bottom) after 1, 20, 60 days (from left to right), focusing on a subregion south-east of the Balearic Islands. The divergence between ensemble members with time illustrates the intrinsic limitation in the predictability of the system. These simulations were produced by explicitly simulating location uncertainties in the model dynamics. This result was obtained using the probabilistic NEMO software.
Residual uncertainties after accounting for observational information
Observations do not cancel uncertainties completely. Ensembles are still necessary to describe the uncertainty that remains in the estimates.
From Brankart, 2019. Conditioning of a geophysical field to observational information. The figure shows three members of the prior ensemble (top panels), before the condition is applied, and three members of the posterior ensemble (middle panels), after the condition is applied. This is an idealized experiment in which the true field is known (bottom left panel), from which synthetic observations are simulated (bottom right panel). This result was obtained using the EnsDAM software.
Uncertainties in ocean analysis and forecast
Uncertainties increase again during the forecast. A reliable description of the ensemble of possibilities is needed to produce unbiased diagnostics and to help decision making.
From our activity in the SEAMLESS project. Ensemble statistical forecast of sea surface chlorophylle concentration for May 26, 2009. This is a 1-day forecast using ocean colour observations until May 25, 2009. It is obtained by applying the observation condition on a prior ensemble simulation performed with a probabilistic version of the NEMO ocean model (coupling physics and biogeochemistry). It is compared to the ocean colour observation image for May 26, 2009 (bottom right panel). This result was obtained using the probabilistic NEMO, EnsDAM and SeSAM softwares.