Predicting Forest Regeneration in the Central Appalachians Using the REGEN Expert System

Abstract

REGEN is an expert system designed by David Loftis to predict the future species composition of dominant and codominant stems in forest stands at the onset of stem exclusion following a proposed harvest. REGEN predictions are generated using competitive rankings for advance reproduction along with other existing stand conditions. These parameters are contained within modular REGEN knowledge bases (RKBs). To extend REGEN coverage into hardwood stands of the Central Appalachians, RKBs were developed for four site classes (xeric, subxeric, submesic, mesic) based on literature and expert opinion. Data were collected from 48 paired stands in Virginia and West Virginia to calibrate the initial RKBs. Paired stands consisted of one mature uncut hardwood stand adjacent to a regenerating clear-cut stand with similar site characteristics that was harvested within the previous 20 yr. Data from 17 additional paired stands was collected a year later to validate the performance of REGEN. Predicted values were within 4 percentage points of measured values on average, and model error was typically less than 20 percentage points for species groups. These results confirmed the suitability of REGEN to predict the future species composition of stands regenerated using the clear-cut method in the Central Appalachians of Virginia and West Virginia.