Abstract
In the first year of this study, we selected 32 rare orchid species by the IUCN criteria. Then a database including 433 locality records of these orchids was constructed. Moreover, 52 environmental layers of Taiwan were built at a resolution of 1*1 km. Using above mentioned data, we modeled species distributions of 24 orchid species by niche modeling tool MaxEnt. However, there were 8 species with too less localities to model by niche modeling techniques. Therefore, we proposed a novel method to predict their possible distributions by the theory spatial-autocorrelation. The result showed that several species' distributions were correlated with high spatial-autocorrelation, while few species with limited distributions were possibly rare because of habitat fragmentation. Interestingly, the result showed that spatial-autocorrelation seems positively related to endemicity but negatively related to species richness. Finally, we cultivated two species seedlings for future mode validation in growth chambers.