Testing the Accuracy of Growth and Yield Models for Southern Hardwood Forests
Abstract The accuracy of ten growth and yield models for Southern Appalachian upland hardwood forests and southern bottomland forests was evaluated. In technical applications, accuracy is the composite of both bias (average error) and precision. Results indicate that GHAT, NATPIS, and a locally calibrated version of NETWIGS may be regarded as being operationally valid growth and yield models for Southern Appalachian yellow-poplar (Liriodendron tulipifera) and mixed oak (Quercus spp.) forests that fall within the range of characteristics of the test data set. No publicly available growth and yield models specifically developed for southern bottomland hardwood forests exist. Four general models that contain most of the applicable species to predict growth of these forests were tested. SETWIGS was found to be the most accurate of the four models tested and is recommended for use if the reported level of accuracy is acceptable and the target stand characteristics fall within the range of our test data set. Results indicate that the growth and density dynamics of dense, young stands of both upland and bottomland hardwoods were poorly predicted by the models. Models predicted basal area and density changes in yellow-poplar stands more accurately than mixed hardwoods. Predictions for upland hardwoods were more accurate than those for bottomland hardwoods. Model accuracy uniformly decreases with increasing length of the projection period. South. J. Appl. For. 24(3):176-185.