ECHOSAT: Estimating Canopy Height Over Space And Time

A collaborative research team from institutions including NASA JPL and the University of Bonn has published ECHOSAT, a breakthrough AI system for global forest monitoring. Unlike existing static tree height maps, ECHOSAT provides the first temporally consistent, global-scale canopy height dataset at 10-meter resolution, spanning multiple years to capture forest dynamics essential for accurate carbon accounting. The system addresses a critical gap in climate science by enabling tracking of both growth and loss over time, moving beyond single snapshots to continuous monitoring of forest health and carbon sequestration potential.

The technical core of ECHOSAT is a specialized vision transformer model trained on multi-sensor satellite data that performs pixel-level temporal regression. A novel self-supervised growth loss regularizes predictions to follow natural growth curves, capturing both gradual height increases and abrupt declines from events like fires or deforestation. Experimental evaluation shows the model improves state-of-the-art accuracy for single-year predictions while providing unprecedented temporal consistency. The resulting maps, already publicly accessible, are expected to significantly advance global carbon monitoring initiatives and disturbance assessment capabilities, providing scientists and policymakers with dynamic data previously unavailable at this scale and resolution.