Estimating Site Productivity from Non-Site Trees: A Site Index Based Approach

Site index (SI) is an indirect measure of potential site quality that is widely used in the Inland Northwest. However, a serious problem exists in applying site curves in the Inland Northwest due to a shortage of suitable \u27site\u27 trees. Site trees are explicitly defined for any given set of site index curves; generally the trees must be dominant or co-dominant and must exhibit characteristics that indicate the tree has been able to grow in height at its potential rate. Due to uneven-aged forest conditions and past selective harvest practices, it can be difficult to find trees that meet the site index criteria. Serious underestimates of SI will arise when non-site trees are used. The primary objective of this research was to develop models for predicting SI from non-site trees, using tree variables that represent vigor, competition, size, and social status. It was hypothesized that these variables would quantify the degree to which a tree is not a site tree. A secondary objective was to investigate whether a quantitative variable for species tolerance would allow analyses of pooled data in order to increase sample size. Stem analysis data from site and non-site trees was gathered on 100 sites located throughout NW Montana. Species included ponderosa pine, Douglas-fir, lodgepole pine, and western larch. At each sample site, site and non-site trees were sampled for as many species as were evident on the site. Regression analysis was used to build SI prediction equations for each species, all species combined, and all species combined using a derived tolerance variable. Results showed that species level models performed the best. These models had standard errors ranging from 5.0’ to 7.8’ and explained about 80% of the variation in observed SI with no bias. For comparison’s sake, soil/site models (no tree attributes used in prediction) typically have standard errors in the 6.25’ to 9.0’ range, while site index curves typically have standard errors less than 5.0’. While the calculated tolerance variable did not capture the species difference as much as hoped, further investigation of tolerance may prove to be fruitful. In conclusion reliable models for estimating SI from non-site trees can be constructed using tree and stand variables that represent vigor, competition, social status, and size. Such models will reduce bias and permit SI estimation where site trees are not available