Predicting potential geographical distributions and patterns of the relic plant Gymnocarpos przewalskii using Maximum Entropy and Genetic Algorithm for Rule-set Prediction
利用最大熵模型和规则集遗传算法模型预测孑遗植物裸果木的潜在地理分布及格局

Aims Understanding the geographical distribution and patterns of endemic species is critical to biodiversity conservation and biogeographical history reconstruction of the area occupied by the species. The ecological niche models (ENMs) are useful techniques to explore the links between the species distribution and the environmental data. The displayed potential habitats for the species, in turn can enable the examination of the predictive abilities of various ENMs. We attempt to determine the potential geographic distributions of the Tertiary relic plant Gymnocarpos przewalskii based on the model of maximum entropy (MAXENT) and genetic algorithm for rule-set prediction (GARP). Methods Based on sixteen sampled localities and seven environmental layers (isothermality, maximum temperature, minimum temperature, annual precipitation, potential evapotranspiration ratio (PER), altitude and soil types), we conducted predictions of G. przewalskii using MAXENT and GARP models. The spatial distribution maps of G. przewalskii with the different environmental suitable values (MAXENT) or overlap index (GARP) displayed the distribution patterns clearly. Important findings The potential distributions of G. przewalskii with the highest environment suitability are predicted at first in the middle of Hexi corridor and the western Yumen of Gansu Province, the north of Ningxia-Hui Autonomous Region, and the part of the Wulate banner of Inner Mongolia in China. The others are mainly in northwestern Tarim Basin and small isolated areas in northwestern Qaidam Basin. Both MAXENT and GARP produced good predictions for G. przewalskii. However, GARP predicted larger and more continuous suitable habitats around the species’ locations and some isolated and fragmented spatial predictions where the species has never been found or collected before. MAXENT predicted a distribution that is a logical proportion of the study area and removed most of the unlikely isolated habitats.