Spatio-Temporal CO₂ Flux Estimation from Remote Sensing

A method to estimate CO₂ emission flux on a high-resolution grid by combining remote-sensed and ground-level proxies. Satellite observation sparsity is resolved through learned imputation, and flux is allocated spatio-temporally with a deep architecture that attends across time.

Overview

Problem — direct CO₂ flux is unobservable at fine resolution; satellite proxies are sparse and noisy.
Approach — impute sparse observations to build high-resolution concentration maps, then allocate flux with a spatio-temporal, attention-based model.
Status — subject of a filed US patent (see Research).