Southern Appalachian Hardwood Forest Growth and Yield Models and Their Accuracy

Because mathematical models are based on experimentation in natural settings, there are many factors that cannot be controlled. Consequently, such models often contain a large error term that represents the collective uncertainty of those many sources of natural variability. Furthermore, most empirically derived models relate only two (or several) factors simultaneously for a relatively small time interval. For these models to be useful, however, they must be linked together so that they simulate a realistic set of landscape events. To do so, longer time intervals, multiple individual/populations, and more variables/relationships are integrated into a greater "whole." This is akin to the melding of actors, parts, stage sets, scenes, direction, etc. that are necessary for a complete and meaningful stage play. Computer simulations are the computational embodiment of those empirical or theoretical constructs, and allow us to perform a large number of complex calculations very quickly--something that would be largely impractical otherwise. By developing software that contains the established mathematical relationships, it becomes possible to recreate fairly realistic--albeit constrained--representations of the real world.

The accuracy of 10 growth and yield models for southern Appalachian hardwood forests forests was evaluated (Rauscher et al. 2000).

To be included, a model had to be publicly available and be able to forecast the growth of the appropriate species found in the region. Four of the models, CSTWIGS, NETWIGS, OAKSIM, and SILVAH, were developed outside the region of interest for this study. Their authors never intended these models to be used in the South. However, the species and stand structures for which these models were developed are similar to those in the southern region, and it seemed reasonable to test their utility in this region. Two more models, SETWIGS and GATWIGS, were developed from USDA Forest Service, Forest Inventory and Analysis data. NATPIS was developed from yield tables directly applicable to the Southern Appalachian region. The remaining three models, GHAT, MIXOAK, and YPOP, were all developed from permanent research plot data within the Southern Appalachian region.

Results indicate that GHAT and NATPIS, may be regarded as being operationally valid models for growth and yield predictions for Southern Appalachian yellow-poplar and mixed oak forests. However, the forests for which predictions are desired need to be have the same characteristics as those in the test data set. In brief, these characteristics are: even-aged, yellow-poplar or mixed oak dominated overstory, dry-mesic to wet-mesic moisture regimes, 20 to 100 yr in age, and 25 to 230 ft2/ac of basal area. For yellow-poplar dominated stands, the appropriate site index range is 74 to 138 ft. (base age 50) and for mixed oak the appropriate site index range is 60 to 110 ft (base age 50 for northern red oak). These results are not valid for: xeric oak forests, mixed-aged hardwood forests, heavily disturbed forests (e.g. those with high-grading), or so-called old growth forest conditions where overstory trees are approaching biological old age and experiencing significant dieback and top breakage.

Further details about the accuracy test of these 10 hardwood models may be found by following the links below:

• The test data set characteristics and the accuracy testing performance measures
• Results: overall model performance comparison
• Results: specific model performance evaluation

GHAT is available from the Department of Forestry, Virginia Polytechnic Institute and State University, Blacksburg, VA. It can be downloaded along with its users manual from www.fw.vt.edu/g&y_coop/models.htm

NATPIS is available from USDA Forest Service, Forest Health Monitoring, Research Triangle Park, NC.