Modelling pin-point cover data of complementary vegetation classes

Abstract

It is often ecologically meaningful to divide the vegetation into a number of complementary vegetation classes or functional types. Here, a method for modelling pin-point plant cover data for such complementary classes is presented. The joint distribution of pin-point cover data of complementary vegetation classes is modelled using a mixture distribution of the multinomial distribution and the Dirichlet distribution, where the Dirichlet distribution is used to model the effect of spatial aggregation. In order to demonstrate the method, the variation in cover with space or time is modelled using a hierarchical Bayesian approach, where the mean cover of each site at a specific time is modelled by a latent variable. The statistical modelling procedure is exemplified in a case-study of pin-point cover data of the two dominating species Calluna vulgaris and Deschampsia flexuosa, and the abundance of the complement species class of all other higher plants on Danish dry heathlands. The cover of C. vulgaris increased significantly with annual precipitation and the cover of D. flexuosa decreased significantly with annual precipitation. Furthermore, the mean cover of C. vulgaris and D. flexuosa within-sites was negatively correlated. There were no significant changes in the cover of the three complementary dry heathland vegetation classes from 2007 to 2012. The presented model allows information of complementarity to be incorporated and whereby increasing the statistical power. Furthermore, the spatial aggregation of the vegetation is modelled so that statistical inference tests will not be deflated due to pseudo-replication.