Calibrating satellite maps with field data for improved predictions of forest biomass

• DOI: 10.48550/arxiv.2407.07134
• Publication Date: 2024-07-09
• Authors: Paul May; Andrew O. Finley
• Keywords: Methodology (stat.ME); FOS: Computer and information sciences; Applications (stat.AP); 0101 mathematics; Statistics - Applications; 01 natural sciences; Statistics - Methodology

Spatially explicit quantification of forest biomass is important for forest-health monitoring and carbon accounting. Direct field measurements are laborious expensive, typically limiting their spatial temporal sampling density therefore the precision resolution resulting inference. Satellites can provide predictions at a far greater density, but these often biased relative to exhibit heterogeneous errors. We developed implemented coregionalization model between sparse predictive satellite map deliver improved 1-by-1 km throughout Pacific states California, Oregon Washington. The accounts zero-inflation in errors predictions. A stochastic partial differential equation approach modeling applied handle magnitude data. detail rendered by much finer than would be possible with alone, provides substantial noise-filtering bias-correction map.