Postdoc: Weakly Supervised ML-Based Earth Observation for Climate Extreme Impact Quantification

Summary

Develops weakly supervised, annotation-efficient machine learning methods to detect, geolocate, and quantify physical and socio-economic impacts from climate extremes using multi-sensor satellite image sequences. Designs data-efficient change-detection modules built on frozen geospatial foundation models, integrates uncertainty quantification to flag out-of-distribution cases, and leverages noisy textual reports as distant supervision to produce actionable impact estimates for adaptation and decision-making.