Modelling pin-point plant cover data along an environmental gradient

When plant cover is measured by the pin-point method the results are traditionally reported by the observed mean and the variance of the number of hits. However, in order to model plant cover data it would be advantageous to have a model that describes the stochastic ecological processes. Here a hierarchical stochastic model of zero-inflated generalised binomial distributed data is introduced. The model specifically represents two important characteristics of the distribution of plant species: plant species do not occur everywhere possible within their realised niche and the data will consequently be zero-inflated, and even if a plant species do occur in the area the number of hits typically will be overdispersed relative to the binomial distribution due to a non-random distribution of individual plants. The stochastic model is applied using different regression models with a varying degree of complexity on hierarchical pin-point cover data of 155 plant species at 10 different habitat types regressed on the measured carbon/nitrogen ratio. The plant cover data displayed a considerable amount of intra-class positive correlation with an average intra-class correlation of 0.31. The degree of intra-class positive correlation differed significantly among habitat types, and generally, it may be concluded that it is critical to take intra-class correlation into account when modelling pin-point cover data.