Identifying the determinant habitat characteristics influencing the spatial distribution of Ferula ovina (Boiss.) in semiarid rangelands of Iran using machine learning methods

Bashari, H., Tarkesh, M. and Besalatpour, A.A., 2021. Identifying the determinant habitat characteristics influencing the spatial distribution of Ferula ovina (Boiss.) in semiarid rangelands of Iran using machine learning methods. Ecological Complexity, 45, p.100909.

Abstract

Ferula ovina (Boiss.) is a valuable but vulnerable monocarpic perennial forb species from Apiaceae plant family whose habitat has been degraded over the years; hence identifying the factors controlling its distribution is important to assist range managers for the reclamation activities. The specific objective of this study was to investigate the quantitative relationships between soil properties, topographical features, climate factors and F. ovina distribution in a semiarid part of central Iran and to assess the relative importance of these factors in controlling its spatial variability. To discern these complex relationships, artificial neural networks (ANNs), support vector machines (SVMs), and decision tree CHAID algorithm were employed. Results from the ANN, SVM, and CHAID models indicated that the climate and topographic conditions should be considered more in explaining the variability in F. ovina occurrence and distribution. Factors such as slope, precipitation of warmest month, and minimum temperature of coldest month were identified by the ANN, SVM, and CHAID models as the determinant factors influencing the spatial distribution of F. ovina in central Iran, respectively. Furthermore, CHAID approach showed greater potential in predicting the F. ovinaoccurrence in the study area. This study provides a strong basis for identifying the most determinant habitat characteristics of F. ovina and other vulnerable or endangered plant species in semiarid rangelands of Iran; however, its general analytical framework could be applied to other parts of the world with similar challenges.