Predictive distribution models for determination of optimal threshold of plant species in central Iran

Abstract

The aim of this study was determination of optimum threshold for predictive distribution models of plant species. For this purpose, vegetation sampling was carried out using random-systematic method. The plot size and sample sizes were determined using minimal area and statistical methods respectively. For sampling the soil at each habitat, eight holes was drilled and samples were taken from 0 to 30 and 30 to 80 cm depths. Plant distribution modeling was conducted using Logistic regression (LR), the Maximum entropy methods (MaxEnt) and Multi-layer perceptron of artificial neural networks (ANN). Threshold optimum was determined using sensitivity-specificity equal and maximum sensitivity approaches. Results indicate that in the LR model, Seidlitzia rosmarinus model was the poorest model (opp=0.3). However, the Artemisia sieberi model is the most accurate one (opp=0.7). The poorest and strongest of MaxEnt models were related to Halocnemum strobilaceum (opp=0.1) and Seidlitzia rosmarinus (opp=0.3). The poorest and most powerful models of ANN with 0.4 and 0.8 discrimination ability related to Seidlitzia rosmarinus and Tamarix passerinoides habitats respectively.