climate-disease-forecast

Climate-driven disease prediction with agent-based ensemble modeling, ERA5 reanalysis integration, and uncertainty-quantified forecasting.

---

Architecture

ERA5 Reanalysis          Disease Surveillance         Population Data
(temperature,            (incidence rates,            (demographics,
 precipitation,           case counts,                 mobility)
 humidity, wind)          mortality)
        |                       |                          |
        +----------+------------+--------------------------+
                   |
          DataAcquisitionAgent
          (fetch, validate, gap-fill, merge)
                   |
          EnsembleRoutingAgent
          (model selection, weight assignment, BMA)
                   |
     +-------------+-------------+-------------+
     |             |             |             |
   ARIMA       Prophet       XGBoost       LSTM
     |             |             |             |
     +-------------+-------------+-------------+
                   |
          Bayesian Model Averaging
          (uncertainty quantification)
                   |
          Prediction Output
          (mean, CI_lower, CI_upper)

Agent system

The agents/ package implements autonomous model ensemble routing:

• EnsembleRoutingAgent -- Dynamically selects and weights forecast models based on recent accuracy. Maintains a model registry with performance tracking and produces uncertainty-quantified predictions through Bayesian model averaging.

• DataAcquisitionAgent -- Autonomously fetches and validates climate data from ERA5 reanalysis. Quality checks for missing values, outliers, and temporal continuity with automated gap filling.

Data sources

Source	Variables	Resolution
ERA5 reanalysis	Temperature, precipitation, humidity, wind, pressure	0.25 degree, monthly
WHO GHO	Disease incidence and mortality rates	Country-level, annual
National HMIS	Subnational case counts and sentinel surveillance	District-level, weekly/monthly
WorldPop	Population density and demographics	1km grid, annual

Components

Module	Purpose
agents/ensemble_agent.py	Autonomous model selection and ensemble routing
agents/__init__.py	Agent package exports
models/	Individual forecast model implementations
data/	Data loading, preprocessing, and feature engineering
evaluation/	Forecast verification and skill scoring

Quick start

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Run ensemble forecast
python -m agents.ensemble_agent --country BGD --indicator malaria_incidence --horizon 6

License

MIT