Mahsa Ghazimoradi; Mostafa Tarkesh; Hossein Bashari; Mohammad Reza Vahabi
Abstract
In this study, the ability of Generalized Additive Model (GAM) for mapping potential distribution of Ferula ovina Boiss, and the description of species response curves to environmental variables were examined in Fereidonshar region with area of 1000 square kilometers located in West of Isfahan province. ...
Read More
In this study, the ability of Generalized Additive Model (GAM) for mapping potential distribution of Ferula ovina Boiss, and the description of species response curves to environmental variables were examined in Fereidonshar region with area of 1000 square kilometers located in West of Isfahan province. The presence and absence data of the species were collected from 278 sites (including 138 presence sites and 140 absence sites) by random systematic method, and 9 soil variables, 22 climate variables and 3 physiographic variables were mapped with pixel size of 72*72 square meters by using interpolation techniques (krigging, Inverse Distance Weighting) for the entire studied area. Then, the relationships between presence and absence of the species with the environmental variables were examined by GAM method. According to the results, the presence of Ferula ovina was inversely correlated with some environmental variables including soil silt and clay contents and also its distribution was directly correlated with slope, distance from sea level, organic matter content, soil saturation percentage and average annual temperature. The model evaluation conducted by the separate data gave Kappa coefficient equal to 0.64 and ROC area under curve equal to 0.86. According to the produced potential habitat map and the response curves of the species, the presence of Ferula ovina is more likely in the habitats with mean annual temperature: 9-11 degree centigrade, slop: 25-50%, elevation: 1950-3000 meter (Above sea level), CaCO3 content of soil: 10-30%, organic matter: 4-6%, silt: 10-30% and soil saturation percentage: 45-60%. The GAM enables managers to identify appropriate areas for rangelands rehabilitation and protection programs. The produced model has a suitable performance in identifying regions with high growth potential and rangelands rehabilitation and protection programs.
