تعیین رویشگاه بالقوه گونه گیاهی کما (Ferula ovina Boiss) با استفاده از مدل شبکه عصبی مصنوعی در منطقه فریدونشهر اصفهان

Species distribution maps have been widely developed based on ecological niche theory together with statistical and geographical information system in plant ecology. The current study aimed to evaluate Artificial Neural Network (ANN) in mapping potential habitat of Ferula ovina Boiss in Ferydunshar rangelands, Isfahan. This is known as valuable forage and medicinal species. Environmental data (independent variables) and species occurrence data (dependent variable) were required to determine potential habitat of a given species. Some physical and chemical soil properties, climate and physiographic variables were mapped for the entire studied area using krigging and inverse distance weighting methods. F. ovina occurrence data were collected from 278 sites including 137 presence and 141 absence sites. The relationships between the studied environmental variables and F. ovina occurrence data were explored using ANN method. According to the sensitivity analysis, occurrence of F. ovina mostly correlated with silt and sand percentage, elevation slope, and organic matter. Model evaluation based on Kappa coefficient (0.66) and Receiver operating characteristic (ROC=0.9) showed good model fitness in relation to reality on local scales. The ANN technique enables managers to identify appropriate areas for rehabilitation practices such as direct seeding and planting.