Planetary Prediction Engine: Autonomous Geospatial Prediction via Intelligent Data Selection and Foundation Model Embeddings
PPE turns a natural-language geospatial question into data retrieval, feature fusion, model search, and prediction without the usual manual buildout.
The paper says the system pulls relevant covariates from Data Commons and Google Earth Engine, then combines them with geospatial foundation model embeddings including PDFM and AlphaEarth. It reports stronger results than expert or state-of-the-art baselines across US health, risk, and vulnerability indicators. In Nigeria food-security downscaling, PPE’s reported R² is 66.1% versus 31.5% for the baseline. For 2026 DRC Bundibugyo Ebola nowcasting, it identified 15 of 18 newly invaded health zones across five weekly forecasts. ArXiv · AI/CL/LG's note
The paper says the system pulls relevant covariates from Data Commons and Google Earth Engine, then combines them with geospatial foundation model embeddings including PDFM and AlphaEarth. It reports stronger results than expert or state-of-the-art baselines across US health, risk, and vulnerability indicators. In Nigeria food-security downscaling, PPE’s reported R² is 66.1% versus 31.5% for the baseline. For 2026 DRC Bundibugyo Ebola nowcasting, it identified 15 of 18 newly invaded health zones across five weekly forecasts. ArXiv · AI/CL/LG's note
score 4