The SMC tree list generation database is comprised of a set of executable programs and a database of stand measurement data that are used to generate simulated stands or tree lists from a per-unit-area description of a forest stand. The tree list generation database uses a stand classification procedure based on a multidimensional histogram to store and reference (index) the actual stand measurement data that are in the tree list generation database. The stand measurement data stored in a tree list generation database consist of individual tree height and diameter at breast height measurements, and tree species for each sample plot or stand measurement that has been placed into a database.

Once a tree list generation database is available, i.e., populated with stand measurement data, a nearest neighbor stand selection procedure is used to create a canonical stand that is representative for a set of specified stand attributes, supplied by the user, and a simulated stand or tree list is drawn (generated) from this canonical stand. The overall procedure is very similar to the generation of pseudorandom numbers, except here random forest stands are generated.

The current version of the tree list generation database supports untreated and thinned stands of Douglas-fir (Psuedotsuga menziesii), western hemlock (Tsuga heterophylla), or combinations of these two species. The initial measurement data set in the tree list generation database spans Oregon, Washington, and southern British Columbia west of the Cascade Mountains.

Stand attributes that may be used for the stand classification and the nearest neighbor stand selection are site index, quadratic mean diameter (QMD), stand total age, stand type (pure Douglas-fir, pure western hemlock, etc.), etc., for all stands, regardless of treatment, and number of thinnings, percent of the basal area per-unit-area removed, pre-thin density in trees per-unit-area, etc., for thinned stands.

A validation of the tree list generation database methodology and implementation has been performed. For the validation, a simulated stand was generated to mimic the per-unit-area attributes of each actual stand. The actual and simulated stands were then compared to determine whether the simulated stands were distinguishable from the actual stands they were generated to mimic. The validation results indicate that at least 80% of the simulated stands, for both untreated (>5000 measurements) and thinned (>4000 measurements) stands, were statistically indistinguishable from the actual stands they were generated to mimic. The agreement between actual and simulated stands may even be greater than 90%. Thus, the tree list generation database methodology and implementation appear to work quite well.