Comparison of the predictive performance of two species distribution models GAM and GBM for Thymus kotschyanus in Middle Taleghan Rangelands

In this study, the prediction of Thymus kotschyanus habitat distribution was investigated using two methods of generalized additive regression model (GAM) and Boosted regression trees (BRT) in central part of Taleghan rangelands. Data on vegetation and habitat factors such as topography, climate, geology and soil were collected. For data preparation, samples were taken from the field to record the presence/absence of the species using 735 plots of 1m2 and random-systematic method. Topography information’s, like slope, elevation and slope direction of sampling points were extracted from digital elevation map. In addition, on each site five soil samples were taken to the laboratory for the following tests: gravel content, soil texture, organic matter, caco3, phosphorus and potassium. The results showed that the probability of presence of T. kotschyanus increases as caco3, potassium and silt are more available and decreases when phosphorus and clay are increasing. Also, its relationship with elevation and gravel is gaussian. Results showed that both methods (GAM and GBM) are able to predict the distribution of T. kotschyanus habitat meaningfully.