Suitability evaluation of habitat for ardeidae waterfowls based on logistic regression model

Abstract

Wetland ecological systems have recently suffered varying degrees of damage, significantly threatening wetland waterfowls and their living spaces. Considering Mai Po-Deep Bay Wetland as an example, the current study analyzed 14 independent variables that are closely related to ardeidae waterfowls. The actual data on ardeidae waterfowls in January 2003 were used as induced variables in a logistic regression model. Nine variable factors, including land use, normalized difference vegetation index, gradient, rainfall, TM4 vein, TM3 vein, road density, road distance, and habitat density, were obtained via screening. The precision of the model reached 0.743 via Nagelkerke R2 inspection with better fitting. The model result was used for the fast clustering for suitability classification of habitat. Classification result shows a good agreement with the actual data on ardeidae waterfowls within the same period, and the precision reached 77.4%. Finally, all variable factor data in January 2009 were collected to perform time-scale inspection on the regression equation. Moreover, the fitting precision with actual data on ardeidae waterfowls within the same period reached 75.8%. Therefore, the proposed model has better universality.