frankraDIUM/Automated-GIS-ML-System-for-Water-Infrastructure-Risk-in-Kenya

This project builds an automated, end-to-end GIS and machine learning pipeline to analyze water point infrastructure risk across Kenya, with a focused application in Nairobi County. # 🚰 Automated GIS-ML System for Water Infrastructure Risk in Kenya This project builds an automated, end-to-end GIS and machine learning pipeline to analyze water point infrastructure risk across Kenya, with a focused application in Nairobi County. Interact on Kepler . View on ArcGIS online: --- Map Preview --- ## 1. Project Objective The primary goal was to build a **scalable, automated geospatial and machine learning workflow** to assess water point infrastructure risks **across Kenya**, identifying patterns of non-functionality, underserved zones, and failure risk nationwide. Nairobi County was designated as a **high-priority focus area** for: - Targeted filtering of high-risk points - Visualization of priority zones - Actionable recommendations for urban WASH interventions Key deliverables: - Spatial SQL automation for service area coverage and clustering - Predictive ML model for water point failure risk - Interactive maps and GIS-ready exports - Prioritization insights for rehabilitation and policy ## 2. Data Sources & Ingestion **Datasets**: - WPdx Kenya water points (CSV): 21,953 points with attributes (`status_clean`, `install_year`, `water_tech_clean`, `pop_served_500m`, `is_urban`, etc.) Time period of the dataset:01 January 2011 - 01 December 2024. Modified: 14 December 2025 - WorldPop 2020 population raster (`ken_ppp_2020.tif`): ~100 m gridded population counts - GADM Kenya Admin Level 1 boundaries (shapefile): 47 counties, including Nairobi (`NAME_1 = 'Nairobi'`) **Ingestion**: - Water points → PostGIS table `water_points` (GeoPandas + `to_postgis`) - Population raster → `population_raster` (`raster2pgsql`, tiled, SRID 4326) - Admin boundaries → `admin_boundaries` - Nairobi subset created via `ST_Intersects` → `water_points_nairobi` (~11 points initially, expanded with buffer to ~25) All data processed in PostgreSQL/PostGIS. ## 3. Spatial Analysis & Automation (Kenya-wide) **Nationwide workflows**: 1. **Service area coverage** - PL/ …