Built Environment Barriers to Flood Early Warning Dissemination in Low-Income Urban Areas: Data and Code Repository

This repository contains processed datasets, trained machine learning models, analysis scripts, and figures supporting the paper: 'Built Environment Barriers to Flood Early Warning Dissemination in Low-Income Urban Areas: An AI-Assisted Spatial Analysis of Warning Accessibility Gaps in Ghana' (Obeng Junior and Arda, 2026). Contents: (1) 100m-resolution feature matrices for Greater Accra (373,345 cells) and Kumasi (35,252 urban cells) comprising 15 morphological, terrain and population features; (2) Random Forest and XGBoost model evaluation outputs and SHAP feature importance; (3) AI-predicted shadow zone maps in GeoPackage format; (4) all manuscript figures; (5) Python analysis scripts for all pipeline phases. Raw input data (OSM, WorldPop, NASADEM, GHSL, Sentinel-1, CHIRPS) are available from their respective public sources as documented in the paper.