Predictive modeling of plant species habitat distribution using logistic regression (A case study in western Taftan, Khash City)

The present study aimed to predictive modelling of plant species habitats distribution and preparation of predictive distribution maps of plant species using logistic regression.Vegetation sampling was carried out by random-systematic method where along four transects in the length of 150- 200 m established With regard to the type of the species, the quadrat size was determined by the minimal area method which ranged from 2 to 25 m2 and sample size was determined 40-60 quadrats by considering the changes in vegetation and statistical method. In order to soil sampling,soil sampling was performed by digging of eight soil profiles and sampling 0-30 and 30-80 cm depths. Habitats distribution was modeled using logistic regression and SPSS software. Predictive maps of plant habitat distribution were prepared using relevant models. Based on predictive obtained model, soil texture, soil organic matter in upper soil, percent of gypsum in subsoil, lime percent in upper soil, soli acidity (pH) in subsoil, type of geological formation, degree of slope and altitude were most effective variables in habitat distribution of plant communities. Based on Kappa value the agreement of predicted and observed maps was exellent for the habitats of A. scoparia, A. aucheri, and for the habitats of H. persicum , Z. eurypterum and A. sieberi was good and poor respectively. Results show that logistic regression could provide high predictive accuracy model for species that have unique habitat condition such as A. scoparia and A. aucheri in comparison with species that have wide ecological amplitude such as A. sieberi.