Modelling of the Arthropod Population Densities in the Coastal Band of Lake Van Using Mixture Poisson Regression

Abstract

This study is based on periodical surveys of the coastal band of Lake Van. Twenty sampling sites were determined on the basis of 3 different characteristics: stream entrance point, settlement on the coastline, and natural coastline. Macro invertebrates were collected with a standard sweep net, and zooplankton samples were collected using plankton nets. Mixture Poisson regression was used for data analysis. In mixture Poisson modeling, the data set is assumed to have an unobservable heterogeneous structure, causing serious deviations in parameter estimations and standard errors. This heterogeneity is overcome when the data set is divided into homogeneous subpopulations. Thus, homogeneity is achieved for subpopulations but the heterogeneity to be put forward between the subpopulations is unclear. In this study, 20 zooplankton and 7 insect species were found and their populations were separated into 8 subpopulations according to the mixture Poisson regression model. The relationship between species densities and the physicochemical structure of water was statistically significant (p<0.01). In the subpopulations, it was determined that the densities of certain species in areas with the same characteristics were high. This indicates that these species are indicators for these areas.