Predicting tree recruitment with negative binomial mixture models

Abstract

Tree recruitment models play an important role in simulating stand dynamic processes. Periodic tree recruitment data from permanent plots tend to be overdispersed, and frequently contain an excess of zero counts. Such data have commonly been analyzed using count data models, such as Poisson model, negative binomial models (NB), zero-inflated models, and Hurdle models. Negative binomial mixture models (zero-inflated negative binomial model, ZINB; Hurdle negative binomial model, HNB) including NB model were used in this study to predict tree recruitments of Chinese pine (Pinus tabulaeformis) in Beijing. ZINB model and HNB model were suitable for dealing with excess zero counts, for which two equations are created: one predicting whether the count occurs (logistic function) and the other predicting differences on the occurrence of the count (NB model). Based on the model comparisons, the results showed that negative binomial mixture models performed well in modeling tree recruitment, and ZINB model was the best model of negative binomial mixture models.