Empirical modeling of remotely sensed data at regional to continental scales

A continental scale dataset was assembled to examine the drivers of greenness indices. Easily parallelized algorithms for ordinary least squares linear regression and a binary regression tree were implemented and used for the analysis. The most important drivers were found to be long and shortwave radiation, precipitation, elevation, and soil pH. This analysis shows that it is possible to perform empirical modeling with large datasets and thus the archives of remotely sensed data can and should be analyzed to shed light on models of large scale natural processes