We use data, AI, and computing to understand our planet and how we can live better on it.
Lab News
September 2025: How can we map crop types in places with little or no ground data? Our latest paper, led by postdoc Xin-Yi Tong, shows that certain satellite features are invariant across regions, meaning that — contrary to conventional wisdom — models trained in one country can be applied in another. This opens the door to globally-shared crop mapping models, making it possible to generate accurate crop type maps even where field surveys are scarce — a step toward scalable monitoring for food security and sustainability.
September 2025: New paper in Remote Sensing of Environment, led by PhD student Kerri Lu! Our study tackles a common issue in environmental research: scientists use satellite-derived maps in statistical models, but errors in these maps can lead to biased results. Using a framework called Prediction Powered Inference (PPI), we show how to account for these errors so that estimates remain unbiased and confidence intervals have the correct statistical coverage. This paves the way for remote sensing maps to inform rigorous science and evidence-based policy.
June 2025: Several members of our lab attended the Living Planet Symposium in Vienna! It was our first LPS and we had a great time. We presented work on scalable ground truthing for crop type mapping, multi-modal weather forecasting, an EO dataset for foundation model training, and more.