CALIBRATION OF A MIXED-EFFECT STEM TAPER MODEL FOR TECTONA GRANDIS

Abstract

The assumption of independence between observations has been frequently violated in forest literature. A part of it is related to the fact that parameters estimated by least squares and predictions made by these models are impartial in the presence of autocorrelation. Since, the absence of autocorrelation between observations is one of the basic assumptions of regression analysis, this study aimed to assess the calibration of a mixed effect model to estimate diameter and volume of the stem of Tectona grandis trees. The log volumes of 509 trees were calculated using relative method, and initially the variable-exponent taper model was fitted (Kozak 2004). For mixed effect modelling, the trees were considered as random effect. A Bayesian calibration was performed on the diameters of 18 trees which were not part of the training data set, and for these trees nine height combinations were tested along the stem. The calibration results were assessed using root mean square error and graphic analysis of residuals autocorrelation. The calibration led to precise estimates of diameter and volume along the stem. The use of the diameter at breast height (DBH) as prior information on stem taper was efficient in reducing residual autocorrelation.