Asymptotic models of species-area curve for measuring diversity of dry tropical forest tree species

Abstract

In a dry tropical forest, we examine the fitness and predictability of two non-asymptotic models (log-linear and power) of species-area curve, and the effect of sample location and scale on their regression-derived coefficients (c, z) for measuring tree diversity. Results indicate that, the log-linear model relatively better fits the data set, and yields better prediction of number of species on a small scale (i.e. predicted number of species for 3 ha using an equation based on 1 ha data). On the other hand, predictions from power function model for a larger area (i.e. predicted number of species for 15 ha using 1 ha and 3 ha equations) were closer to the observed values. The suitability of the model to fit the data was strongly influenced by the site and the scale of the plot size. The equations for the two models derived from data of small area (1 ha plot size) yielded inconsistent results, but those derived from a larger plot size (3 ha) consistently underestimated the number of species for 15 ha. The underestimation by power function model was lower compared to that by log-normal model for predicting the number of tree species. The study also shows that the coefficient z is site- as well as scale-dependent. The coefficient c can be used to predict α-diversity, and the number of species per individual can adequately describe the coefficient z. The results support discrete community concept for the dry tropical forests along a disturbance gradient and indicate that higher the z, greater would be the impact of harvest of individuals on biodiversity.