Model selection and fit of algebraic difference models for basal area of loblolly pine plantations.

On the basis of annual observations from 148 sample plots, we built Schumacher algebraic difference models for basal areas of loblolly pine (Pinus taeda) plantations under intensive management. Using a different modeling approach from the conventional, we first fitted basal area yield models, and then used the algebraic difference approach (ADA) to derive corresponding algebraic difference models. The method for identifying the standdependent coefficient is modelfitting based, and leads to new ADA basal area models. The restricted maximum likelihood estimation (REML) was used to estimate variance structure models, and the maximum likelihood to fit expectation models. AIC, BIC(Schwarz’s Bayesian Information Criterion), and log likelihood ratio test (LRT) are the model selection criteria/test. The results showed that basal areas under different management scenarios presented significantly different growth trajectories with respect to both model coefficients and model structures.