Multiple Regression Inference of Yield for Douglas-fir Plantations in the Pacific Northwest

Abstract

Multiple regressions were formulated with predictor variables of latitude, longitude, and elevation, stand stocking, age, and site index for western Oregon, western Washington, and southwest British Columbia. Models were assessed for normality using plots of studentized residuals by fitted values, q-q plots, and fit was assessed by R2 value. Outliers were examined based upon Bonferroni p-values and Cook's distance. The developed seven multiple regression models were for response variables of cubic foot volume including top, cubic foot volume to top 4 and 6 inches in diameter, and Scribner board foot volume to top 4 and 6 inches in diameter. These models were used to examine the effects of early density controls and the effects of subsequent heavy first thinnings on volume yield. The assessment concluded that early density controls resulted in highest to lowest productivity in the order of 100 initial Trees Per Acre (TPA), 300 TPA, 500 TPA, and 700 TPA. In assessing effects from a first thin, the 300 TPA class was shown to have the most increased volume. The first thinning in the 500 and 700 TPA levels showed an overall minimal increase in volume and diameter size due to a longer delay in allocating growth from the crown to the stem. The models produced in this study can be of use in assessing and understanding predictions made by other regional growth models.