Using Reinforcement Learning to Integrate Subjective Wellbeing into Climate Adaptation Decision Making

TL;DR

Subjective wellbeing is a central but under-integrated objective in climate adaptation policymaking, with floods directly reducing mobility and daily activities. The authors introduce a multi-modular integrated assessment model that links rainfall projections, flood dynamics, transport accessibility, and motility-based wellbeing within a reinforcement learning framework to discover policy pathways in Copenhagen under climate uncertainty. They implement four actionable interventions and compare two RL algorithms (PPO and IMPALA) to maximize long-run wellbeing, using Climate Atlas rainfall, SCALGO flood modeling, PCA-based motility, and a cumulative-link wellbeing model. The approach offers a scalable decision-support tool for exploring and prioritizing resilience strategies that sustain wellbeing in the face of uncertain floods and resource constraints, with potential transferability to other cities.

Abstract

Subjective wellbeing is a fundamental aspect of human life, influencing life expectancy and economic productivity, among others. Mobility plays a critical role in maintaining wellbeing, yet the increasing frequency and intensity of both nuisance and high-impact floods due to climate change are expected to significantly disrupt access to activities and destinations, thereby affecting overall wellbeing. Addressing climate adaptation presents a complex challenge for policymakers, who must select and implement policies from a broad set of options with varying effects while managing resource constraints and uncertain climate projections. In this work, we propose a multi-modular framework that uses reinforcement learning as a decision-support tool for climate adaptation in Copenhagen, Denmark. Our framework integrates four interconnected components: long-term rainfall projections, flood modeling, transport accessibility, and wellbeing modeling. This approach enables decision-makers to identify spatial and temporal policy interventions that help sustain or enhance subjective wellbeing over time. By modeling climate adaptation as an open-ended system, our framework provides a structured framework for exploring and evaluating adaptation policy pathways. In doing so, it supports policymakers to make informed decisions that maximize wellbeing in the long run.

Figures (5)

• Figure 1: Integrated assessment model using reinforcement learning to learn what the best set of policies are that maximize wellbeing.
• Figure 2: Average rainfall (and 95% confidence interval shaded) for 365 samples (equivalent to daily sampling per year) following different scenarios (RCP2.6, RCP4.5, and RCP8.5) and Climate Atlas dmi2023klimaatlas projections for rainfall in the Copenhagen municipality between 2023 and 2100.
• Figure 3: Flash flood maps in Copenhagen, Denmark for rainfall intensities of 0mm, 10mm, 50mm, and 100mm. Darker blue indicates a higher water depth at the specific location. Flash flood maps were generated and retrieved from SCALGO Live scalgo.
• Figure 4: Average relative accessibility loss for a 60 mm intensity rainfall event (approximately a 100-year return period) for driving, cycling, walking, and transit trips to hospitals, schools, gas stations, and eating locations.
• Figure 5: Preliminary wellbeing model diagram. Latent factors represented by circles, observed variables represented by rectangles. Items $y_1$ to $y_{17}$ represent the 17 survey items used for the motility components, ordered as reported in Table \ref{['tab:pca_loadings']}. For an in-depth look at the PCA components and their item loadings, see Table \ref{['tab:pca_loadings']}. For a look at the preliminary model linking the PCA components with life satisfaction, see Table \ref{['tab:clm_output']}.