Simulation outputs and RL pump scheduling data for urban pluvial flood mitigation using a CHANS-RL framework in Ninh Kieu District, Can Tho City, Vietnam

This dataset contains simulation outputs from a Coupled Human and Natural Systems (CHANS) modelling framework that integrates a GPU-accelerated hydrodynamic model (HiPIMS), Agent-Based Model, and hierarchical Reinforcement Learning (RL) to optimise mobile pump scheduling for urban pluvial flood risk mitigation. Data include: time-series flood risk scores per Higher-level Zone (HZ) for baseline, traditional, and RL-guided pump deployment scenarios; mobile pump scheduling decisions at 10-minute intervals for Concurrent Pluvial Flooding (CPF, 90 min) and Post-Rainfall Pluvial Flooding (PPF, 120 min) events; and sensitivity analysis results (OFRR and FRRR) under three rainfall scenarios (50-year, 100-year present; 10-year 2050 projected). The study domain covers the 15 highest-risk zones of Ninh Kieu District (Can Tho City), Vietnam, simulated at 3 m spatial resolution. Modelled rainfall event: 1-in-20-year design rainfall (95 mm/h) for training; sensitivity tested at 102, 107, and 138 mm/h.Hydrodynamic simulation using HiPIMS (GPU-accelerated 2D shallow water equations, Godunov-type finite volume scheme) at 3 m resolution, driven by uniform design rainfall inputs. Flood risk quantified by aggregating cell-level risk scores (sigmoid flood probability × Flood Value combining inundation extent, population density, and building financial losses). RL agents trained using Proximal Policy Optimisation (PPO) over ~5,700 episodes. Pump scheduling decisions made at 10-minute intervals.<br>This dataset is associated with the following publication:Qin, H., Liang, Q., Chen, H., De Silva, V. (2024). A Coupled Human and Natural Systems (CHANS) framework integrated with reinforcement learning for urban flood mitigation. Journal of Hydrology, 643, 131918. https://doi.org/10.1016/j.jhydrol.2024.131918<br>