Author

Lichun Jiang

Date of Graduation

2007

Document Type

Dissertation/Thesis

Abstract

An integral approach to estimating stem green and dry weight for yellow-poplar in West Virginia was compared to traditional ratio equation methods. The data were based on stem analysis of 26 trees from the Central Appalachian Broadleaf Forest province in northern West Virginia and 18 trees from the Eastern Broadleaf Forest province in west central West Virginia. The proposed equation generally performed better for the whole tree, as well as for sections within the tree, based on the 9 relative height classes examined. A constant wood-density was superior to the use of a linear wood-density equation as a function of height above the ground. The proposed equation explained over 90% of the variation in stem weight and compared favorably with existing fixed merchantable top weight equations. Compatible taper, volume and weight equations were developed for planted red pine in West Virginia. The data were based on stem analysis of 26 trees from the West Virginia University Research Forest. A commonly employed segmented polynomial taper equation was chosen due to its balance between prediction accuracy and ease of use. Seeming unrelated regression (SUR) was used to simultaneously fit the system of equations for inside and outside bark data. A volume bias that increased with tree size was observed when existing published total stem volume equation were evaluated with this dataset. The residuals of the proposed model exhibited no unusual trends and explained over 96% of the variation in outside bark diameter and volume as well as inside bark diameter, volume and weight. Nonlinear mixed-effects model approaches were utilized in the fitting procedures for a taper function for yellow-polar and a dominant height growth model for red pine. The precision of the taper equation was increased when compared to least squares regression. By modeling heteroscedasticity and autocorrelation, the precision of the dominant tree height model in explaining data variation was increased.

Share

COinS