Interpolation of GEDI Biomass Estimates with Calibrated Uncertainty Quantification
The study addresses unreliable uncertainty quantification in estimating biomass from NASA's GEDI satellite data by introducing Attentive Neural Processes (ANPs). ANPs are a probabilistic model that learns to produce prediction intervals that adapt to the landscape, showing higher accuracy and better calibration than standard ensemble methods. This provides a more trustworthy way to map forest biomass.