RISC: Regional Impact-based Shelter Coordination Framework

⚠️ Status Declaration for Peer Review

This repository provides a Minimal Working Example (MWE) of the codebase supporting our submitted manuscript. Please note that the full code and data are currently undergoing final formatting and preparation for public archiving. To comply with peer-review requirements while protecting sensitive data, this repository contains the core algorithmic logic to demonstrate the feasibility of the proposed framework.

Upon the acceptance of the manuscript, the complete, fully documented codebase and replicable datasets will be permanently deposited and publicly accessible.

📖 Project Overview

This repository implements the Regional Impact-based Shelter Coordination (RISC) framework, a probabilistic and impact-based approach for regional emergency shelter planning under tropical cyclone (TC) risks. The framework transitions shelter planning from hazard-based assumptions to impact-based coordination, and consists of three core modules:

• Module 1: Shelter Demand Prediction Model (SDPM): A data-driven model that predicts spatially heterogeneous shelter demand by learning hazard–impact relationships from multi-source datasets.
• Module 2: Long-Term Probabilistic Simulation: Embeds the SDPM within a large synthetic ensemble of physically plausible TCs to characterize long-term shelter-demand risk and reveal hazard–impact misalignment.
• Module 3: Cross-Jurisdictional Coordination: Utilizes the Impact Disparity Coefficient (IDC) and an evidence-based Base Spatial Scale to identify demand imbalances and optimize cross-jurisdictional capacity sharing.

📂 Repository Structure

Module 1:

• LOSO_training
• LOSO_test

Module 2:

• AllTrackImpactSimu-1kmGrid
• hazard-impact return period
• Multi-spatial scale return period analysis
• Multi-spatial scale spatial correlation analysis
• zj-return period curves

Module 3:

• Coordination based on IDC
• Coordination results-deltaS maps
• Coordination results-spatial maps