Clim-02 Evaluating Basis Risk in Index-Based Agricultural Insurance: A Comparative Study Using Satellite vs. Ground Weather Data PFE

Objective & Scope

• Evaluate basis risk in index-based agricultural insurance by comparing indices derived from satellite data (Sentinel-2/NDVI, CHIRPS) versus ground weather station data.
• Perform a comparative study linking remote-sensing indices to crop yield variations and to the performance of index insurance products in selected regions.

Technological Environment & Data Sources

• Languages & libraries: Python (pandas, matplotlib, geopandas), Scikit-learn for correlation and statistical modeling, Jupyter Notebook for interactive reporting.
• Remote sensing & weather data: CHIRPS (weather), Sentinel-2/NDVI (via Google Earth Engine), FAOStat (yields if available).
• Geospatial tools & storage: Google Earth Engine for remote sensing processing, GeoPandas and Folium for geospatial visualization, SQLite or PostgreSQL with PostGIS for data storage.

Tasks & Methodology

• Ingest and preprocess CHIRPS weather data and Sentinel-2 derived NDVI time series (GEE export), and match with available ground station records and yield data.
• Compute relevant indices (rainfall indices, NDVI-based drought proxies), quantify correlations, and estimate basis risk metrics between index triggers and actual losses/yields.
• Implement statistical models and ML approaches (e.g., regression, correlation analysis, possibly time-series models) to assess predictive power and spatial heterogeneity.
• Visualize spatial patterns of basis risk using GeoPandas/Folium and produce reproducible analyses in Jupyter notebooks.

Deliverables & Expected Outcomes

• A reproducible Jupyter Notebook workflow that ingests raw datasets, computes indices, runs comparative analyses, and visualizes results.
• A written report summarizing methodology, statistical results, spatial maps of basis risk, and recommendations for index design (including database schema for the storage solution used).
• Code and dataset exports stored in SQLite or PostgreSQL/PostGIS and documented steps to reproduce analyses.

Prerequisites & Supervision

• Required skills: Python data stack experience (pandas), basic geospatial handling (GeoPandas), familiarity with remote sensing concepts (NDVI) and statistics.
• Nice to have: experience with Google Earth Engine, basic SQL/PostGIS, and prior exposure to agricultural/climate datasets.
• Supervision expected from Hydatis research/engineering team with regular meetings and milestone reviews.

How to apply

• To apply, send your application referencing this project (Clim-02) to stages@hydatis.fr or apply via the company site: https://www.hydatis.com
• Include CV, cover letter outlining relevant projects/skills, and links to any demos or notebooks.