irrigation-need-prediction
nazaryansurenn-amb/irrigation-need-prediction
ML decision support for a Water User Association in Armenia: predicts irrigation need per grid cell from operational and earth-observation data, and maps WAIT / SOON / NOW priorities so field crews know where to go first. Python, scikit-learn, Streamlit. Code only — operational data excluded.
Summary
A geospatial machine learning decision support system for irrigation management in Armenia. It predicts weekly irrigation priorities (WAIT/SOON/NOW) per 250m field cell using Sentinel-2 satellite indices and weather data. The system combines rule-based current-week priorities with a gradient-boosting forecast for next week, achieving 85% accuracy in summer tests. Includes a Streamlit web application for operational use by water associations.