EVALUATION OF FINITE MIXTURE MODELS FOR DESCRIBING THE STRUCTURE OF DISTURBED GMELINA STANDS IN OLUWA FOREST RESERVE, NIGERIA

Resumo

Finite mixture model is relevant for describing diameter distribution that is multimodal or heavily skewed. Irregular diameter distribution is partly caused by forest disturbance such as bush burning and illegal logging. This study evaluated five finite mixture models for describing the irregular diameter distributions of the disturbed Gmelina arborea Roxb stands in Oluwa Forest Reserve, Nigeria. Twenty plots of 0.04 ha size, five each from four stand ages (19, 24, 29 and 34 years) were used in this study. Five finite mixture models: gamma, Gompertz, log-logistic, lognormal and Weibull mixtures were considered. The quality of fits produced by the models were evaluated with five indices: Anderson-Darling, Cramer-von Mises, Kolmogorov-Smirnov, Akaike Information Criterion and Hannan-Quinn Information Criterion. Relative rank sum from the indices was analysed using One-way analysis of variance. The results showed that gamma mixture had the smallest indices and relative rank, but not significantly different from Weibull and log-logistic mixtures (p > 0.05). Lognormal and Gompertz mixtures performed poorly. The order of ranking was: gamma followed by Weibull, log-logistic, lognormal and Gompertz. Application of the mixture models provided good predictions of the forest stand volume.