Application of generalized linear models to estimate height growth

Abstract

Height growth analysis presents great importance in forestry, as it expresses site production capacity. Its use is associated with lower adjustment error models to generate estimates to inference with precision and reliability. The present study examined generalized linear models in predicting height growth of Pinus taeda L. depending on the age and diameter at 1.30 m height above ground level in stands in the highlands of Santa Catarina State. The data were obtained from complete stem analysis of 25 trees with 8 years old, divided into diameter classes from Lages, SC. Data were processed in original form without variables transformation. The model with gamma distribution and identity link function presented the best fit, with superior performance criteria deviation (1.21), Akaike (255.39) and residuals homogenization, showing potential to generate estimates of the variable.