Abstract: Habitat loss and fragmentation is one of the main threats to wildlife and biodiversity conservation. This study collected three years (2019—2021) elephant presence data across the entire Xishuangbanna State, built multiscale multivariable species distribution models to evaluate the impact of topography, land use, linear infrastructure and villages to the probability of elephant presence. We collected a total of 10 710 elephant presence locations, including 1 088 ones in dry season and 5 630 in rainy season. Nine out of 39 predictors influenced the probability of elephant presence most, including elevation, extensiveness of rubber plantation, cropland, mixed forest and landscape, patch density of bare land, cropland and dense forest, and road density. A total of 5 181 km² occupying 27% of Xishuangbanna state was predicted to be suitable for elephants, supporting approximately 347 (187—554) elephants. Suitable habitat mainly distributed in northern Xishuangbanna adjoining Pu’er and fragmented patches sat in southern Xishuangbanna surrounding Shangyong and Mengla reserves. However, the large area of cropland within the predicted suitable habitat was prone to be raided by elephants while the suitable habitats in Southern Xishuangbanna was at risk of being encroached by rubber plantations. There were small patches of habitat predicted to be suitable in Bulong and Yiwu reserves. This study provided science-based evidence for planning of the to be established Asian elephant national park and management of wild elephant populations in Yunnan Province.