Predicting potential distributions of Zygophyllum eurypterum by three modeling techniques (ENFA, ANN and logistic) in North East of Semnan, Iran

Abstract

The paper investigates the use of ‘Ecological niche factor analysis’ (ENFA) method for modeling Zygophylum eurypterum species geographic distributions with presence-only data and ‘Artificial Neural Network’ (ANN), ‘Logistic Regression’ (LR) methods for investigating of Zygophyllum eurypterum species distribution with presence-absence data in North East of Semnan province. Plant density and cover, soil texture, available moisture, pH, electrical conductivity (EC), organic matter, lime, gravel and gypsum contents and topography (elevation, slope and aspect) are the variables sampled using the randomized systematic method. Within each vegetation type, the samples were collected using 15 quadrates placed at an interval of 50 m along three 750 m transects. To map soil characteristics, geostatistical method was used. The back propagation Neural Network in MATLAB software was used to generate the ANN model with one input, one hidden and one output layer and the Logistic Regression analysis was done in SPSS software and based on obtained models (ANN and LR methods) predicted maps in Arc Map software were created. The accuracy of the predicted maps (prepared by ENFA, ANN and LR) were tested with actual vegetation maps. Kappa coefficients prepared by these methods show good accordance with actual vegetation map prepared for the study area. The results of ENFA method show that 25200 hectares or 34 percent of study site is potential habitat of Z. eurypterum. The results also revealed that maps generated using the LR and ANN models for Z. eurypterum species have a high accordance with their corresponding actual maps of the study area. This species is distributed in rangeland with alkali-saline soil, high of lime percent, silty-sandy texture and in 1000-2000 meters elevation.