Spatial distribution modeling of grapevine (L foetid-assa ferula) using Bayesian Belief Network (BBN) and Support Vector Machine (SVM) models

Abstract

This study was conducted with the aim of identifying the environmental factors affecting the distribution of Ferula gabrielli and modeling the spatial distribution of this species using SVM models and Bayesian Belief Networks (BBNs). For this purpose, data on the presence and absence of the species under study were collected using the stratified random sampling method (65 sites of presence and 70 sites of absence) located in Khor and Biyabank cities. Physiographic data layers such as slope height and direction were also extracted from the DEM map of the area. Using the SVM modeling method, the relationship between the distribution of Anghuzeh species and environmental factors was investigated and the species response curves were prepared to these factors and a species distribution potential map was prepared. The results of SVM and BBNs models showed that the environmental variables of annual rainfall, seasonality of temperature, average daily temperature, altitude and precipitation in the coldest season have a significant effect on the presence of this species. The cause and effect model of BBNs consisting of 32 different variables showed that environmental factors have more influence on the desirability of the Angouzeh species habitat than management factors because the exploitation and management conditions of this species are relatively favorable. The BBNs models were able to show the uncertainty of the prediction results in the results and management variables that could not be used in the SVM model were used. According to the prediction map obtained from the SVM model and the reaction curves, the highest probability of the presence of Anghuzeh species was in the areas where the average daily temperature was 15.5 degrees Celsius, the annual rainfall was 75 mm, the altitude was 1400 to 2000 meters, and the rainfall in the coldest season was 28 mm.