Fitting site index curves using mixed models for Eucalyptus urophylla plantations in Venezuela

Classification of stands through "site index curves" is useful for establishing and managing forest plantations because growth and yield is highly dependent on site quality. Most forestry data are taken longitudinally at the same sites and the variances often change with time, which makes unsuitable the application of the ordinary general linear model. Mixed models are good alternatives since they permit modeling covariance structures. In this work, linear and nonlinear mixed models were used to analyze the relationship of top height with age for Eucalyptus urophylla plantations in Venezuela, and to find the most appropriate "guide curve" for site classification. Data consisted of repeated measurements on permanent plots. The models fitted were Schumacher (linear), Chapman-Richards, Weibull, Logistic, Gompertz, and Monomolecular (non linear). Three different parametrizations were used: 1) original, assuming random asymptotes; 2) reparametrized so that the local factor is the asymptote; and 3) reparametrized so that the local factor is the parameter associated with the function's domain. In the last two cases the largest height at the base age has an associated random effect. Site classification was done using the concept of 'site's site index, interpreted as the most probable height at base age in all the hypothetical stands that could grow in that site. The best fitting guide curve was obtained with a mixed-model Chapman-Richards equation.