Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Climate Adaptation with Reinforcement Learning: Experiments with Flooding and Transportation in Copenhagen

Miguel Costa, Morten W. Petersen, Arthur Vandervoort, Martin Drews, Karyn Morrissey, Francisco C. Pereira

TL;DR

The paper tackles increasing urban flood risk from climate change and its disruption to transportation by proposing a PPO-based reinforcement learning framework within an Integrated Assessment Model that links rainfall projections, flood dynamics, and mobility impacts. It defines a reward $R = \sum_{i \in \texttt{TAZ}} (\beta_{R} R_{i} + \beta_{D} D_{i} + \beta_{A} A_{i})$ with equal weights to learn when and where to elevate roads (1 meter) to minimize economic losses. Preliminary results show the RL policy outperforms random policies, achieving substantial reductions in direct damages, travel delays, and overall impacts by 2035 and 2100, albeit with higher adaptation costs earlier. The approach provides actionable guidance for proactive urban resilience planning and is intended to be publicly available for further use and extension.

Abstract

Due to climate change the frequency and intensity of extreme rainfall events, which contribute to urban flooding, are expected to increase in many places. These floods can damage transport infrastructure and disrupt mobility, highlighting the need for cities to adapt to escalating risks. Reinforcement learning (RL) serves as a powerful tool for uncovering optimal adaptation strategies, determining how and where to deploy adaptation measures effectively, even under significant uncertainty. In this study, we leverage RL to identify the most effective timing and locations for implementing measures, aiming to reduce both direct and indirect impacts of flooding. Our framework integrates climate change projections of future rainfall events and floods, models city-wide motorized trips, and quantifies direct and indirect impacts on infrastructure and mobility. Preliminary results suggest that our RL-based approach can significantly enhance decision-making by prioritizing interventions in specific urban areas and identifying the optimal periods for their implementation. Our framework is publicly available: \url{https://github.com/MLSM-at-DTU/floods_transport_rl}.

Climate Adaptation with Reinforcement Learning: Experiments with Flooding and Transportation in Copenhagen

TL;DR

The paper tackles increasing urban flood risk from climate change and its disruption to transportation by proposing a PPO-based reinforcement learning framework within an Integrated Assessment Model that links rainfall projections, flood dynamics, and mobility impacts. It defines a reward with equal weights to learn when and where to elevate roads (1 meter) to minimize economic losses. Preliminary results show the RL policy outperforms random policies, achieving substantial reductions in direct damages, travel delays, and overall impacts by 2035 and 2100, albeit with higher adaptation costs earlier. The approach provides actionable guidance for proactive urban resilience planning and is intended to be publicly available for further use and extension.

Abstract

Due to climate change the frequency and intensity of extreme rainfall events, which contribute to urban flooding, are expected to increase in many places. These floods can damage transport infrastructure and disrupt mobility, highlighting the need for cities to adapt to escalating risks. Reinforcement learning (RL) serves as a powerful tool for uncovering optimal adaptation strategies, determining how and where to deploy adaptation measures effectively, even under significant uncertainty. In this study, we leverage RL to identify the most effective timing and locations for implementing measures, aiming to reduce both direct and indirect impacts of flooding. Our framework integrates climate change projections of future rainfall events and floods, models city-wide motorized trips, and quantifies direct and indirect impacts on infrastructure and mobility. Preliminary results suggest that our RL-based approach can significantly enhance decision-making by prioritizing interventions in specific urban areas and identifying the optimal periods for their implementation. Our framework is publicly available: \url{https://github.com/MLSM-at-DTU/floods_transport_rl}.
Paper Structure (9 sections, 1 equation, 2 figures, 1 table)

This paper contains 9 sections, 1 equation, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Modelling Framework
  4. Rainfall Projection Model
  5. Flood Model
  6. Transportation Model
  7. Transportation and Mobility Impacts
  8. Reinforcement Learning
  9. Preliminary Experiments & Discussion

Figures (2)

  • Figure 1: Integrated assessment model using reinforcement learning to learn what the best adaptation measures are that minimize transportation infrastructure and mobility impacts.
  • Figure 2: Costs of floods impacts on transportation and mobility in Copenhagen in 2035. Top row shows results with random adaptation measures deployed over time and space, while bottom row shows impacts using optimal adaptations over time. From left to right: direct road infrastructure impacts, indirect impacts as travel delays, percentage of travel time difference for travel between TAZ, and where adaptation measures were deployed (red).