Application of Logistic Regression Tree Model in Determining Habitat Distribution of Astragalus verus

The relationship between plant species and environmental factors has always been a central issue in plant ecology. With rising power of statistical techniques, geo-statistics and geographic information systems (GIS), the development of predictive habitat distribution models of organisms has rapidly increased in ecology. This study aimed to evaluate the ability of Logistic Regression Tree model to create potential habitat map of Astragalus verus. This species produces Tragacanth and has economic value. A stratified- random sampling was applied to 100 sites (50 presence- 50 absence) of given species, and produced environmental and edaphic factors maps by using Kriging and Inverse Distance Weighting methods in the ArcGIS software for the whole study area. Relationships between species occurrence and environmental factors were determined by Logistic Regression Tree model and extended to the whole study area. The results indicated species occurrence has strong correlation with environmental factors such as mean daily temperature and clay, EC and organic carbon content of the soil. Species occurrence showed direct relationship with mean daily temperature and clay and organic carbon, and inverse relationship with EC. Model accuracy was evaluated both by Cohen’s kappa statistics (κ) and by area under Receiver Operating Characteristics curve based on independent test data set. Their values (kappa=0.9, Auc of ROC=0.96) indicated the high power of LRT to create potential habitat map on local scales. This model, therefore, can be applied to recognize potential sites for rangeland reclamation projects.