Offre de thèse financée : Apprentissage statistique causale en contexte multirisque - Avec application aux risques naturels et environnementaux

About INRAE

The French National Research Institute for Agriculture, Food and Environment (INRAE) is a public research institution bringing together a community of 12,000 people, with 272 research, service, and experimental units, located in 18 centers across France. INRAE is among the world's leading players in agricultural and food sciences, plant and animal sciences. Its research aims to build solutions for multi-performance agriculture, quality food, and sustainable management of resources and ecosystems.

Work Environment, Missions, and Activities

Context and Challenges

Natural risks (climatic, hydrological, ecological, agricultural, etc.) often result from the combination of several factors that interact in time and space. For example, drought can weaken an ecosystem and amplify the impact of a heatwave or pest attack, or saturated soils can worsen the consequences of heavy rainfall.

These so-called multi-risk situations are at the heart of current challenges and are often likely to be amplified by climate change and land use changes. However, they remain difficult to analyze with classical approaches, which often consider hazards in isolation. It is becoming essential to better understand the chains of causality and interactions, and the cascading effects, in order to improve risk assessment and inform public decisions.

Thanks to the increasing availability of data and physical and biological models of the processes involved, it is possible to develop and use statistical learning methods to better understand causal links and possible confounding effects between the different variables involved in the occurrence of risks in complex systems.

Thesis Objectives

The objective of this thesis is to develop an approach to attribute impacts to their different causes in multi-risk systems by combining data from sometimes highly heterogeneous sources, from observation or model simulation, through developments such as the following:

• Establish and confirm links in causal chains involving climate, environment, exposure, and vulnerability, explicitly in space and time.
• Analyze and quantify possible confusions between different variables contributing to a risk (e.g., climate vs. bioaggressor vs. other factors).
• Adapt causal attribution to apply directly to the impacts of climate change (damages) rather than just climate variables.
• Conduct applications on real systems studied at INRAE. Examples: in agriculture, study the effect of the joint occurrence of several climatic events on yields (e.g., concomitance of drought and temperature peak, or drought followed by a violent storm); in forestry, analyze tree mortality and separate direct climatic effects (thermal and water stress) from indirect effects favoring pest epidemics such as bark beetles.

The ambition is to go beyond a purely "correlative" vision to propose tools capable of identifying cause-and-effect relationships. Among the methods to be studied and extended are causal graphs and Bayesian networks (static or dynamic), causal analysis, and attribution techniques.

References

• Allard, D., Curt, C., Evin, G., & Opitz, T. (2021). Analyse multirisque : concepts, méthodes et verrous – un état de l'art prospectif. Rapport technique. https://hal.science/hal-04185479/.
• Bastos, A., Sippel, S., Frank, D., Mahecha, M. D., Zaehle, S., Zscheischler, J., & Reichstein, M. (2023). A joint framework for studying compound ecoclimatic events. Nature Reviews Earth & Environment, 4, 333–350. https://doi.org/10.1038/s43017-023-00410-3
• Hannart, A., Pearl, J., Otto, F. E. L., Naveau, P. and Ghil, M. (2016) Causal counterfactual theory for the attribution of weather and climate-related events. Bulletin of the American Meterological Society, 97(1), 99-110. https://doi.org/10.1175/BAMS-D-14-00034.1
• Jézéquel, A., Bastos, A., Faranda, D., Kimutai, J., Le Grix, N., Wilson, A. M., Rufat, S., Shepherd, T. G., Stuart-Smith, R. F., Van Loon, A. F., Bevacqua, E., D’Andrea, F., Lehner, F., Lloyd, E. A., Moemken, J., Ramos, A. M., Sippel, S., & Zscheischler, J. (2024). Broadening the scope of anthropogenic influence in extreme event attribution. Environmental Research: Climate, 3(4), Article 042003. https://doi.org/10.1088/2752-5295/ad7527Perkins-Kirkpatrick SE, Alexander LV, King AD, Kew SF, Philip SY, Barnes C, Maraun D, Stuart-Smith RF, Jézéquel A, Bevacqua E, Burgess S, Fischer E, Hegerl GC, Kimutai J, Koren G, Lawal KA, Min S-K, New M, Odoulami RC, Patricola CM, Pinto I, Ribes A, Shaw TA, Thiery W, Trewin B, Vautard R, Wehner M and Zscheischler J (2024) Frontiers in attributing climate extremes and associated impacts. Front. Clim. 6:1455023. https://doi.org/%2010.3389/fclim.2024.1455023
• Renard B., Barbero R., Goukouni I., Vidal J.-P., Mimeau L., Furusho-Percot C., García de Cortázar-Atauri I., Aubry M., Opitz T., and Allard D. (2026) A Joint Space-Time Probabilistic Model for Agricultural Droughts, Hydrological Droughts and Fire Weather in France. doi.org/10.5194/egusphere-2026-1406
• Zscheischler, J., & Lehner, F. (2022). Attributing Compound Events to Anthropogenic Climate Change. Bulletin of the American Meteorological Society, 103(3), E936-E953.

Training and Skills Sought

• Master/Engineer (Bac+5)
• Master or engineering degree with a background in statistics or applied mathematics.
• Knowledge in probability, statistics, causal graphs, extreme value theory is appreciated.
• Programming in R.

Your Quality of Life at INRAE

By joining INRAE, you benefit (depending on the type and duration of the contract):

• Up to 30 days of leave + 15 RTT days per year (for full-time positions)
• Parenting support: childcare vouchers, leisure benefits;
• Skills development programs: training, career counseling;
• Social support: counseling and listening, social aid and loans;
• Vacation and leisure services: holiday vouchers, preferential rate accommodation;
• Sports and cultural activities;
• Collective catering.

How to Apply

Send your CV and cover letter.

People welcomed at INRAE, a public research institution, are subject to the provisions of the Civil Service Code, particularly concerning the obligation of neutrality and respect for the principle of secularism. As such, in the exercise of their duties, whether or not they are in contact with the public, they must not manifest their convictions, through their behavior or appearance, whether religious, philosophical, or political. > Learn more: fonction-publique.gouv.fr website.