A new approach to extrapolate forest attributes from field inventory with satellite and auxiliary data sets

Abstract

We have developed a tool called “Field And SatelliTe for Ecosystem MAPping (FastEmap)” based on an innovative algorithm. The tool assimilates measured point-specific estimates of forest biophysical characteristics, satellite images, and auxiliary data sets to map forest parameters at a 30-m resolution. Four major components are integrated into an automated production line: (1) creation of additional “virtual” field plots; (2) using stepwise regression to identify significant explanatory variables and create a spatial prediction; (3) iteratively establishing groups that are composed of similar pixels and imputing a group with the weighted mean of the plots that are within the group; and (4) local interpolating and strata median filling. Taking the complex Southern California area as an example, we used this tool to map aboveground live biomass from 156 Forest Inventory and Analysis (FIA) plots. The accuracy evaluation from an additional 36 FIA plots produced an R2 of 0.78, a root mean square error of 28.3 tons/ha, and a mean absolute error of 22.7 tons/ha. Our results indicate that FastEmap is a promising tool to automatically map continuously varying forest attributes over large geographic extents. The next step is to test the performance of different settings and apply the tool to a broader area.