Thèse en informatique M/F : Development of a multi-fidelity and multi-model approach for headwater catchments

Job Description

Thesis Subject

Development of a multi-fidelity and multi-model approach for headwater catchments


Context — Headwater catchments (HWC) form the uppermost parts of river networks and represent a large fraction of the European hydrological system. Although they are relatively small (from a few to a few hundred km²), they play a disproportionate role in controlling water resources.

At the core of their functioning are the aquifers they host, which regulate how water infiltrates, is stored, and is released to streams. These aquifers sustain river flows during dry periods and strongly influence hydrological responses to climatic variability and extremes.

Yet, despite their importance, these systems remain poorly understood and difficult to predict. Their behaviour emerges from complex and highly variable interactions between geology, topography, and climate. Across Europe, this diversity leads to strong contrasts in groundwater dynamics, while most headwater aquifers remain sparsely monitored.

In a context of intensifying droughts and increasing pressure on water resources, this lack of understanding limits our ability to anticipate changes in water availability. Addressing this challenge requires new modelling approaches capable of representing groundwater processes across thousands of heterogeneous catchments.

To tackle this issue, the FutureFlow project proposes to transfer concepts from software engineering—such as multi-fidelity modelling and model-switching—to hydrology. The objective is to build adaptive modelling systems capable of selecting and combining models of different complexity, in order to better represent groundwater dynamics and improve large-scale predictions under climate change.

Objective — The PhD study will leverage existing numerical resources and mobilize advanced methods from software engineering to develop a robust and adaptable framework for manipulating and coupling different models. The resulting interface, models, and benchmarking strategies will provide the foundation for seamless interactions between models, supporting their automatic deployment from the catchment to the regional scale.

The main research question is: How can heterogeneous hydrological models be integrated, interfaced, and dynamically combined within a unified multi-fidelity framework?

Scientific challenge — Many computational models already exist, but they have generally been developed in isolation, with heterogeneous data structures and interfaces. As a result, replacing one model with another—referred to here as model-switch—remains difficult in practice.

This limitation hinders systematic comparison, combination, and adaptive use of models. The objective is therefore to enable model switching between models of different natures (e.g., physically-based vs. simplified or data-driven models) and different levels of precision, in order to support multi-fidelity approaches.

Methods — The work will first focus on designing a common interface to standardize interactions with HWC models. This includes harmonizing inputs/outputs, managing simulation workflows, and interfacing with calibration tools, while enabling seamless interaction between the models used in the project without modifying their source code.

It will involve developing specific adapters to unify access to the different hydrogeological models. Based on an analysis of their data structures and formats, the work will consist in defining standardized representations of simulation domains, hydrogeological properties, boundary and initial conditions, source/sink terms, stress periods, and outputs, as well as model-specific adapters enabling communication with the common interface.

In parallel, a unified modelling framework will be developed to manipulate data and orchestrate simulation workflows. Particular attention will be paid to ensuring consistency of parameterization across models, especially when dealing with different physical processes, spatial discretizations, and temporal resolutions.

Scientific contributions — Building on these developments, the work will contribute to the design of adaptive multi-model and multi-fidelity simulation systems. A central contribution will be the formalization and implementation of a model-switch mechanism enabling seamless transitions between models of different fidelity levels.

Beyond implementation, this work will address fundamental scientific questions related to the definition of model fidelity, the characterization of model validity domains, and the development of decision criteria for model selection. It will explore how heterogeneous models can be combined within a unified framework to optimize trade-offs between accuracy, robustness, and computational cost.

These contributions aim at advancing the understanding of adaptive simulation systems and providing a generic framework for multi-model integration beyond hydrology.

Implementation, validation & impact — The proposed concepts will be implemented in prototypes, including an online platform hosting the modelling framework, simulation models, and their adapters. These prototypes will support empirical evaluation, testing, and scientific dissemination.

Beyond their implementation, the expected outcomes have a dual impact. From a computer science perspective, the work will contribute to advancing methodologies for multi-model integration, adaptive simulation, and multi-fidelity systems. From an environmental perspective, the developed modelling framework will be used by a broader community of researchers and practitioners involved in water resources management and ecosystem protection, supporting improved assessment and anticipation of hydrological responses to climate change.

Environment — The PhD will be conducted in a multidisciplinary environment at the interface between computer science and hydrology, within an international collaboration between France and Switzerland. The work will involve close interactions between software engineering and environmental sciences, fostering cross-disciplinary skills.

A background in hydrology is not required, as the necessary domain knowledge will be progressively acquired during the PhD through collaboration with hydrogeologists and project partners.

References

Sallou, J., et al., 2020. Loop Aggregation for Approximate Scientific Computing. ICCS
2020 - International Conference on Computational Science.

Marcais, J., de Dreuzy, J.-R., et al., 2017. Dynamic coupling of subsurface and seepage flows solved within a regularized partition formulation. AWR 109, 94

Langevin, C.D., Provost, A.M., Panday, S., Hughes, J.D., 2022. Documentation for the MODFLOW 6 Groundwater Transport Model, Techniques and Methods. Reston, VA

Your Work Environment

About the Laboratory
=============
www.irisa.fr IRISA is currently one of France's largest research laboratories (with over 850 staff members) in the field of computer science and information technology. Organized into seven scientific departments, IRISA is a center of excellence whose scientific priorities include bioinformatics, system security, new software architectures, virtual reality, big data analysis, and artificial intelligence. Focused on the future of computer science and with a strong international orientation, IRISA is at the very heart of society's digital transition and innovation in the fields of cybersecurity, health, the environment and ecology, transportation, robotics, energy, culture, and artificial intelligence. Overview of the CNRS as an employer: https://www.cnrs.fr/fr/le-cnrs Presentation of IRISA as the host laboratory: https://www.irisa.fr/umr-6074 The position falls within a sector relevant to the protection of scientific and technical potential (PPST) and therefore, in accordance with regulations, requires that your appointment be authorized by the competent authority at the Ministry of Higher Education, Research, and Innovation (MESR).

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions