Title: Small area estimation of biodiversity measures
Authors: Philip Rosenthal - Trier University (Germany) [presenting]
Jan Pablo Burgard - Trier University (Germany)
Stephan Feldmeier - Trier University (Germany)
Ralf Muennich - University of Trier (Germany)
Michael Veith - Trier University (Germany)
Abstract: Sustainability management and the protection of biodiversity are of increasing importance for evidence based policy. As policy can only be implemented in regions in reach of the policy maker good regional estimates for biodiversity measures are needed. Hill numbers provide a framework for such biodiversity measures. Because sampling of species is very costly and often restricted to certain regions, no traditional sampling is performed for many species. The data gathering process is usually not describable in the classical sampling theoretical framework. Observations are often sparse and scattered on the whole region of interest, leading to inaccurate outcomes of regional biodiversity measures when using classical estimation techniques. Model based small area methods may lead to more reliable estimates. We extend a multinomial logit model by a zero-inflated Poisson part. This model is used to construct a synthetic estimator for the Hill numbers for an arbitrary aggregation level of the observation grid. A parametric bootstrap is used for the mean squared error estimation. To validate the method a model based simulation is used to show the performances of the new estimator in different extreme scenarios. The new method is then applied to a grasshopper dataset to estimate their biodiversity on natural regions.