Site index curves from mixed models for Cupressus lusitanica plantations in the Central Region of Costa Rica

Abstract

The Site Index is the most widely used technique to classify site qualities in monospecific plantations and predict their productivity regarding wood and other goods. In Costa Rica, cypress (Cupressus lusitanica Miller) is extensively planted for reforestation in highland regions due to its fast growth and excellent quality wood. Internationally, numerous site index works have been developed for cypress. However, nonlinear mixed models have only been used once for New Zealand and never for Costa Rica. These models have advantageous statistical properties that allow addressing the violation of the assumptions of the fixed effects regression model, obtaining more precise estimates of the parameters, and high reliability for hypothesis testing. In this work, site index curves were developed using a nonlinear mixed model from permanent plots (n=44) and stem analysis (n=21) measured for more than 40 years in the Central Region of Costa Rica. The data were partitioned into development (62.2%) and validation (37.8%) data. Three reparametrized nonlinear mixed models (Richards, Korf, and Prodan) were fitted with a random parameter associated with the site index, and violations of the regression model assumptions were corrected. The best model was selected based on information criteria, goodness of fit, and graphical analysis for the development and validation data. The best predictive model was based on the Korf function. The family of polymorphic curves was generated for every three meters of height (27 to 15 m). The maximum mean annual increment for the IS of 21.67 m was 1.16 m∙year-1 and occurred at 8 years of age. The model was compared with other models developed with different approaches and in various parts of the world.