Abstract:

The endangered shrub species Amygdalus pedunculata is distributed in Inner Mongolia Autonomous Region and Shaanxi Province, China. However, little is known about its resource quantity and conservation status. A field investigation was conducted to determine the natural distribution range of A. pedunculata. Eight environmental factors were chosen to build models with the genetic algorithm for rule-set prediction (GARP) model and maximum entropy (MaxEnt) model. We predicted the potential distribution of A. pedunculata in China. Using four model evaluation metrics (Kappa, true skill statistic (TSS), overall accuracy and area under the receiver operating characteristic curve (AUC)), we assessed the predictive performance of both models. The Jackknife method was used to investigate the most important environmental factors for the distribution of A. pedunculata. The results indicated that both models were effective for predicting the distribution of A. pedunculata, but MaxEnt performed better than GARP in terms of the four evaluation metrics. The species was predicted to have a broad suitable area, which ranged from the south of Daxing’anling to the east of Helan Mountains and the Ulan Buh Desert. Amygdalus pedunculata is mainly distributed in the middle regions of Inner Mongolia, including Mu Us Sandy Land, Kubuqi desert, Otindag Sandy Land, and the Wulashan and Daqingshan Mountains. Low suitable sites occurred in some regions of Liaoning, Hebei, Shanxi and Shaanxi. Besides, and there were some sporadic low suitable areas in the middle regions of the Ningxia Hui Autonomous Region and Gansu Province. Variable importance analysis showed that the variables relevant to precipitation had more significant effects on the geographic distribution of A. pedunculata. These results have important implications for resource conservation and ecology including the revegetation of semi-arid ecosystems.

Key words: Amygdalus pedunculata, potential distribution area, MaxEnt model, GARP model, species conservation