Abstract
Simulation and model prediction of virtual species aim to establish baseline for assessing the projected contemporary and future distribution ranges of real species in mountainous areas. Fundamental niches and geographic ranges of 5 virtual species were defined in the diagram of principal components analysis based on a high-resolution climate dataset generated from meteorological data. Heterogeneity of the climate dataset had been validated to influence the relationships between species responses and suitable environments, consequently affecting the geographical distributions of virtual species. The performances of 11 algorithms were evaluated by the extracted fraction of shared presences (ESP), instead of TSS and AUC. ESP calculates the overlap between simulated suitable ranges and predicted current potential ranges of virtual species. According to ESP, ensemble modeling outperformed the 11 algorithms. A small sample size has significant effects on model performance due to the extremely low value of ESP, and the presence of only 5 sample points was evidently a limitation of model predictions. Furthermore, geographical distance among sample points provide signals of niches that will be identified through accurate predictions of ensemble modeling in our analyses. By the 2050s and 2090s, climate change may drive the range expansion of real species currently distributed in inland areas or on leeward slopes, while causing range restriction or local extinction of real species in coastal areas or on windward slopes. Our study can inform application of species distribution models to provide scientific support for conservation planning in mountainous areas and forecasts of species distributions under climate change.