Variable-shape stem-profile predictions for western hemlock. Part II. Predictions from DBH, total height, and upper stem measurements

Abstract

A general method of extending stem-profile prediction systems is developed. Deviations about a two-point system (based on DBH and total height) are described by the multivariate normal distribution. Once the distribution is characterized, it can be conditioned on any number of upper stem measurements using easily applied statistical relationships. The result is a smooth profile prediction that passes through all measurement points. Volume estimates are obtained by numerical integration. An analysis of a three-point application shows significant reductions in variance for most of the stem length. The variances of volume estimates are correspondingly reduced by factors of 2 to 10, depending on tree size.