New Models for Predicting Diameter at Breast Height from Stump Dimensions

Abstract

Abstract Models to predict dbh from stump dimensions are presented for 18 species groups. Data used to fit the models were collected across thirteen states in the northeastern United States. Primarily because of the presence of multiple measurements from each tree, a mixed-effects modeling approach was used to account for the lack of independence among observations. The heterogeneous error variance was described as a function of stump diameter, which allowed for more accurate representation of prediction intervals. Application of the mean response model (fixed-effects parameters only) to independent data indicated an average absolute error between 0.2 and 0.7 in. for most groups. An additional advantage is that random-effect parameters allow the model to be calibrated to local conditions if some additional data are available. An example is provided that indicates the local calibration results in a mean residual value that is closer to zero compared with the mean response model. Efforts in other locales to use stump information to inform dbh predictions can obtain the same advancements by adopting a similar modeling methodology.