Loughborough University
Browse

A coupled human and natural systems (CHANS) framework integrated with reinforcement learning for urban flood mitigation

Download (18.55 MB)
journal contribution
posted on 2025-05-06, 09:55 authored by Haoyang Qin, Qiuhua LiangQiuhua Liang, Huili ChenHuili Chen, Varuna De-SilvaVaruna De-Silva
To address the challenge of escalating urban flood risk and the deficiency in effective flood emergency management, this study introduces a novel Coupled Human and Natural Systems (CHANS) modelling framework that employs hierarchical reinforcement learning to optimise mobile pump scheduling and placement for urban flood risk mitigation. The CHANS framework integrates hydrodynamic and agent-based models within a multi-GPU computing environment for high-resolution, real-time flood inundation modelling and risk assessment to enrich Reinforcement Learning (RL) training. In the application to Ninh Kieu District in Can Tho City, Vietnam, the new RL-enabled modelling framework is used to evaluate optimal mobile pumping strategies for concurrent pluvial flooding and post-flooding events against the traditional deployment approaches. Results demonstrate that RL-based strategies can significantly enhance flood risk reduction, outperforming traditional methods by achieving 2× and 4× improvements in the concurrent and post-flooding periods/events, respectively. Incorporating human factors and adapting to local conditions, the RL agent provides valuable insights into mobile pump scheduling and deployment strategies. Sensitivity analysis confirms the robustness of the CHANS modelling framework and underscores the role of RL in optimising mobile pump scheduling and placement where traditional rule-based strategies are challenging.

Funding

GCRF Living Deltas Hub

Natural Environment Research Council

Find out more...

PhD scholarship from the China Scholarship Council (No. 201908060298).

History

School

  • Architecture, Building and Civil Engineering

Published in

Journal of Hydrology

Volume

643

Publisher

Elsevier B.V.

Version

  • VoR (Version of Record)

Rights holder

© The Author(s)

Publisher statement

Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)

Acceptance date

2024-08-21

Publication date

2024-09-03

Copyright date

2024

ISSN

0022-1694

eISSN

1879-2707

Language

  • en

Depositor

Prof Qiuhua Liang. Deposit date: 4 December 2024

Article number

131918

Usage metrics

    Loughborough Publications

    Categories

    No categories selected

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC