Application of modeling techniques for the identification the relationship between environmental factors and plant species in rangelands of Iran

The objective of the present research was to compare Ecological Niche Factor Analysis (ENFA) and Artificial Neural Networks (ANN) to determine the optimum threshold of plant species (Thymus kotschyanus Boiss and Hohen. and Dactylis glomerata L.) in rangelands of Ardabil province. Systematic random sampling of vegetation was performed, and an overall 111 sites were considered and divided into two groups, namely sites with plant species and sites without plant species. Five plots of one square meter were placed in each site. The size of the plots was based on the magnitude of one or double the average area of the most common plant species. Plots were positioned along a 40-m transect (from the bottom of the slope to high altitudes with a distance of 10 m between the plots). Soil sampling was done in a depth of 0-30 cm (selected based on the activity of root plant species). Maps of environmental factors were prepared in ArcGIS 10.4.1 software. The sensitivity and specificity were considered to specify the optimum threshold. The optimal threshold between plant species, determined by modeling methods, showed that the highest accuracy belonged to ENFA model in T. kotschyanus habitat (with optimum threshold = 0.62 and sensitivity = 0.58) and ANN model in D. glomerata habitat (with optimum threshold = 0.37 and sensitivity = 0.02) had the lowest accuracy.