Thèse Climate Change Adaptation Modelling With ai And Spatially-Explicit Participatory Simulations In European Metropolises H/F

Doctorat_Gouv

Établissement : Institut Polytechnique de Paris École polytechnique
École doctorale : Ecole Doctorale de l'Institut Polytechnique de Paris
Laboratoire de recherche : CREST - Centre de recherche en économie et statistique
Direction de la thèse : SAMUEL RUFAT ORCID 0000000163561233
Début de la thèse : 2026-10-01
Date limite de candidature : 2026-04-14T23:59:59

The PhD thesis will leverage Artificial Intelligence (AI) to merge Agent-Based Models simulations (ABM) with spatially-explicit Geographical Information System (GIS) local data, and empirical data from local surveys on perceptions, decisions and adaptation behaviour, at multiple geographic levels, over the medium and long term. Climate disruption impacts will continue to drive an increase in extreme events, even in the most favourable scenarios, leading to greater risk over time, especially in large metropolitan areas across Europe. While adapting to climate change is a very real issue, anticipating people's actual behaviour over time and in space remains a major challenge. Current simulations are based on generic decision models rather than implementing empirical data tailored to each region and context, this remains a major knowledge gap. When designing policies and practices for behaviour change and climate change adaptation, it is essential to examine the mechanisms underlying human decision making in dynamic and complex environments. AI is expected to come into this as an accelerator, a way of emulating computationally expensive model components, integrating diverse data streams, and simulating social learning among local stakeholders.

The thesis will contribute to the development of AI-augmented spatially-explicit multi-level simulations, implementing the complexity of the interactions between households, businesses, public services, local stakeholders' perception of risk, and their short- and long-term behaviour. By applying AI and machine-learning to ABM and empirical survey data, with researchers and policymakers, this thesis will create dynamic, evidence-based simulations of climate adaptation in urban digital twins. This approach will identify effective interventions, anticipate unintended consequences, and engage stakeholders in planning for resilient futures. The thesis will build on empirical surveys results collected by a European network in several metropolises. Emerging spatial and data analysis methods will help detect the local predictors of perception and behaviour from the surveys. Stakeholders' engagement will help collect and implement the local special needs, cascading dependencies, strategies, etc. The thesis will leverage the empirical results from previous surveys to calibrate the models for Paris and Bucharest. Participatory methods involving all stakeholders, practitioners, decision-makers, and civil society will then be critical to locally fine-tune and subsequently cross-validate the Paris and Bucharest models. The in-silico experiments will also allow local practitioners in Paris and Bucharest to test the effectiveness of their different strategies and discover possible measures to better target public policies, communication and strategies across scales, to improve and customise resilience measures.

Environmental global changes are increasing the probability and intensity of extreme events, while the economic development of exposed areas is increasing the risk of disasters, particularly in Europe's large metropolises. With 2024 set to be the first year in which average global warming exceeds 1.5°C, the recurrence of flooding is one of the main risks associated with climate change in Europe. Climate disruption impacts will continue to drive an increase in extreme events, even in the most favourable scenarios, leading to greater risk over time, especially in large metropolitan areas across Europe. While adapting to climate change is a very real issue, anticipating people's actual behaviour over time and in space remains a major challenge. Current simulations are based on generic decision models rather than implementing empirical data tailored to each region and context, this remains a major knowledge gap. When designing policies and practices for behaviour change and climate change adaptation, it is essential to examine the mechanisms underlying human decision making in dynamic and complex environments. AI is expected to come into this as an accelerator, a way of emulating computationally expensive model components, integrating diverse data streams, and simulating social learning among local stakeholders.