Research activities

The team’s research is organized around four areas :

• Area 1 – Natural stochasticity of ocean variability
• Area 2 – Sources of modeling errors, parameterizations, and quantification of uncertainties
• Area 3 – Model/observation synergies, inversions, assistance in the design of observation systems
• Area 4 – Lagrangian transport, ecological connectivity, and applications in low latitudes in relation to Southern countries

Area 1 – Natural stochasticity of ocean variability

Analysis of global (OCCIPUT, IMHOTEP) and regional (MEDIATION) ensemble turbulent simulations carried out by the team has shown that a significant proportion of ocean variability is random and spontaneously generated by the nonlinearities of its dynamics (eddies on the scale of basins, from a day to several decades), and that the phase of this variability is only partially controlled by the atmosphere. We propose to extend the study of this natural stochastic variability and its implications in two new directions : (i) characterize its multivariate structure at different scales in regions of interest such as the Mediterranean (MEDIATION context), coastal areas, or the southwest Indian Ocean (BRIDGES context) ; (ii) assess its impact on the predictability of ocean fluctuations (sea levels along coastlines, marine heat waves) and the uncertainties of Lagrangian indices (water mass transformation, connectivity).

In methodological terms, we plan to further explore the contribution of concepts from information theory and dynamic systems to better understand the stochastic variability of the ocean simulated via ensembles (collaboration with the University of Naples), and the use of this ensemble dimension to characterize current-turbulence interactions and scale interactions locally and instantaneously (collaboration with T. Uchida, L. Sun, Q. Jamet in the wake of the CONTaCTS project, see Uchida et al 2022).

The high cost of ensemble turbulent simulations will also motivate further collaboration on increasing the size of ensemble simulations a posteriori using an analog approach (collaboration with LOPS, IMT-Atlantique), and the exploration of ocean models such as Oceananigans that can be integrated on GPUs to significantly reduce the energy cost of ensemble turbulent simulations (collaboration with MIT). Collaboration with IGE researchers working on AI could eventually lead to the development of parameterizations that would make the non-turbulent ocean components of the simulators used today for climate projections stochastic.

Area 2 – Sources of modeling errors, parameterizations, and quantification of uncertainties

The natural stochasticity of the ocean inevitably amplifies uncertainties in the initial conditions of models, but many other uncertainties also affect them. Our understanding of the dominant sources of error in ocean models and their cumulative effects, as well as the systematic biases observed in observations, is still very partial, requiring further process studies and methodological developments of a fairly fundamental nature in order to move towards a more fundamentally “probabilistic” quantitative view of numerical modeling. The new ensemble/stochastic functionality that we recently introduced in CROCO opens up prospects for extending the offshore ocean approach to the regional/coastal level.

In the context of MEDIATION, we are collaborating with numerical scientists from INRIA (AIRSEA and ODYSSEY teams), which has notably made it possible to show the links between the Location Uncertainty framework (decomposition of velocity into a resolved component and an unresolved random component decorrelated in time) and location uncertainties due to grid perturbation (Clément et al., 2025). Uncertainties related to the parameterization of sub-mesh processes (meso- and sub-meso-scale turbulence, convective mixing) is another topic of common interest that could, for example, extend the work of Legay et al (2024) and Perrot et al (2025).

We want to strengthen these collaborations to better leverage multi-fidelity approaches (from LES continuum to eddy-permitting simulations). Work on statistical learning of “low-cost” emulators that would enable us to move towards large ensembles is another promising area of interest. We plan to resubmit the Noise Optimization for Models under Uncertainty in Location for Large Eddy Simulations of Oceanic Flows project, which was not selected under the PEPR MATH Vives program, in order to extend this research beyond MEDIATION (i.e., post-2028). The ongoing evolution of MANU towards a strengthened national program focused on interactions between mathematics and geosciences, the existing IMPT framework, and the GdR Défis-Théo-Climat provide a promising context for these activities, which should also encourage the strengthening of the team through permanent recruitment.

Area 3 – Model/observation synergies, inversions, assistance in the design of observation systems

Synergies between numerical approaches (modeling, assimilation, inversion) and satellite/in-situ observations are a recognized hallmark of the team : simulations help interpret observed variability and detect/attribute long-term trends ; synthetic data from simulations feed into OSSEs, guiding the optimization of observation systems and the analysis and assimilation of future data. OSSEs and ensemble synthetic data will increase these cross-benefits through their probabilistic dimension.

One of the stated objectives of the last five-year period was to build a more generic framework for solving inverse problems, with the dual aim of describing the uncertainty in the solutions produced and controlling calculation costs. The concept of 4D inversion using the MCMC method demonstrated in SEAMLESS opens up new prospects, particularly for the implementation of ensemble OSSEs on GPU computers.

We wish to pursue this approach in the context of the new challenges underlying the national OSSE roadmap recently decided by INSU (SNOOPI project), which aims to develop and consolidate digital methods and tools to optimize the design of national observation networks in the OA, SIC, and TS domains. One objective will be to consolidate the methodological framework and increase the flexibility and robustness of the tools needed for OSSEs, while making their application more systematic and less dependent on operational chains that are still not widely accessible to the academic world. These tools will also be shared with a scientific community of academic and non-academic users, potentially including DATLAS. This activity will enable new collaborations to be established with field observers in connection with certified in situ services (MOOSE in the Mediterranean), community codes (NEMO, CROCO), research infrastructures (Data Terra, ILICO, Euro-ARGO), and also with the space projects of the recent CNES space prospecting program (ODYSEA, SMOS-HR, geostationary water color).

Area 4 – Lagrangian transport, ecological connectivity, and applications in low latitudes in relation to Southern countries

One dimension that we have yet to explore in depth is the probabilistic simulation of Lagrangian transport of biotic and abiotic substances in the ocean. The ensemble framework provides an opportunity to revisit this aspect and fuel new applications such as the identification of ecological connectivity patterns at the regional scale, for which a probabilistic description is key to the design of effective marine protected area networks. The new context of the BBNJ international treaty, the launch of the BRIDGES PEPR in 2024, and our commitment to developing CROCO-based digital avatars for the South Indian Ocean region open up the opportunity for a “low latitudes” project stretching from South Africa to Ecuador, in line with the recommendation of the latest HCERES evaluation to focus our research more on the South.

Strengthening our team by recruiting an IRD researcher in this context would become much more natural, complemented by the recruitment of a “long-term” engineer planned to support the BRIDGES digital avatar at the IGE from 2027 onwards. The new European OPERA project, which is starting on ocean forecasting around Africa (coordinated by MOI), will be an opportunity to renew our link with operational oceanography, with new partners, new application issues, and training challenges in the South. BRIDGES will also open up new opportunities for multidisciplinary collaboration with socio-ecologists at the IGE, as well as with INEE laboratories located in mainland France and overseas (ENTROPIE in Réunion island).