Predicting Post-disturbance Changes in Forest Recovery Rates Using Statistical Models

The carbon balance of a forest ecosystem is fundamentally linked to its cycle of disturbance and recovery. Until now, our understanding of the climatic, topograhic or management controls on recovery rate is limited. In this study, we seek to examine variation in post-disturbance forest recovery trajectories. We assembled FIA biomass and tree height data for the North Carolina, South Carolina state (where re-measured data have been available since 2000 for each state), and retrieve a set of Landsat spectral and disturbance history parameters for each plot (time since disturbance, magnitude of disturbance, spectral change rate since disturbance, etc). These parameters are extracted and (as appropriate) averaged over the spatial extent of each FIA plot. We then use Random Forest modeling to relate the Landsat-based 'predictor' variables to the FIA structural attributes. The predicted changes are achieved with reasonable accuracy at these two states