A Climate-Intelligent Early Warning System for Vector-Borne and Waterborne Diseases in Uganda
# 🦠Uganda Multi-Disease GeoAI Early Warning System
Research title:
Climate-Intelligent GeoAI for Multi-Disease Early Warning: Integrating Climate Data, Spatiotemporal Machine Learning, and Outbreak Detection in Uganda
- A Climate-Intelligent Early Warning System for Vector-Borne and Waterborne Diseases in Uganda
**Summary**
This project is a complete GeoAI-powered early warning system that predicts vector-borne (e.g., malaria)
and waterborne (e.g., cholera) disease risk at the district level in Uganda.
It integrates high-resolution climate data from Copernicus ERA5-Land, administrative boundaries, engineered spatiotemporal features,
and machine learning models to generate risk predictions and outbreak alerts.
The system is delivered through an interactive Streamlit dashboard with real-time forecasting and mapping capabilities.
---
Dashboard Preview
---
---
**Key Objectives Achieved**
- Fuse climate (ERA5-Land) and spatial data (OSM + HDX boundaries)
- Build separate predictive models for vector-borne and waterborne diseases
- Develop an epidemiological outbreak detection engine
- Create a production-grade interactive dashboard with AI recommendations
- Simulate realistic early warning workflows (risk scoring, alerts, mapping)
**Technical Architecture**
Data Sources:
- Climate: ERA5-Land Monthly Means (temperature, precipitation)
- Spatial: Uganda ADM2 districts (135 districts)
- Health: Advanced synthetic epidemiological simulation (population-adjusted, spatial spillover, rolling outbreak detection)
Core Technologies:
- Python, GeoPandas, xarray, rioxarray
- Scikit-learn (RandomForestRegressor + RandomForestClassifier)
- Streamlit + Plotly + Folium
- Joblib model persistence
- Modeling Performance (Test Set)
- Vector-Borne Model: R² = 0.861, MAE = 9.12
- Waterborne Model: R² = 0.738, MAE = 5.06
- Outbreak Classifier: AUC = 0.936
**Key Features Engineered**
- Temporal lags, rolling statistics, anomalies
- Climate-disease interaction te …