Modeling potential habitat of Astragalus verus Olivier for conservation decisions: A comparison of three correlative models

Abstract

Identification of potential habitats of endemic species can facilitate conservation planning and rehabilitation programs in rangeland ecosystems. This study aimed to generate potential distribution maps of Astragalus verus Olivier, a valuable endemic and medicinal plant, in a semi-arid part of central Iran (with an approximate area of 1000 km2) using logistic regression (LR), nonparametric multiplicative regression (NPMR), and ecological-niche factor analysis (ENFA) methods A stratified-random sampling approach was used to collect the presence-absence data, along with some environmental data, from 100 sites. Soil maps were obtained from 70 profiles, and climate maps were produced using data from 10 weather stations. Geostatistical methods were used to convert the obtained data to digital maps with 90 m × 90 m resolution. The species potential habitat maps were produced using ENFA, LR, NPMR, and an ensemble method. The accuracy of the models was evaluated using Cohen’s kappa coefficient and the area under the receiver operating characteristics curve (AUC) based on both calibrated and independent data sets. Although no method was superior under all circumstances, ENFA performed slightly better due to the very high anthropogenic impacts in the study area. Potential habitat maps and the species response curves revealed that A. verus habitats were characterized by a pH of 7.5, moderate to relatively heavy soil texture (44%–48% clay content), moderate slope (10%–20%), low electrical conductivity (<0.3 dS/m), and a soil saturation percentage of 45%–47%. These modeling approaches can be widely used to identify appropriate areas for rehabilitation and protection of endangered and valuable native plant species.