The general predictions of climate impacts on species shifts (e.g., upward shift) cannot directly inform local species conservation, because local-scale studies find divergent patterns instead of a general one. For example, our previous study found three shift patterns with elevation (strong down-, moderate down-, and up-slope shifts) in temperate mountain forests. The divergent shifts are hypothesized to arise from both multivariate environmental variations with elevation and corresponding species-specific responses. To test this hypothesis, we sampled soils and leaves to measure elevation variations in soil conditions and determined plant responses using discriminations against heavier isotopes, carbon ( 13 C) and nitrogen ( 15 N). Functional traits of the species studied were also extracted from a public trait dataset. We found that: (1) With low soil water contents at low elevations, only the leaves of up-shifters had lower 13 C discriminations at low vs. high elevations; (2) With low soil P contents at high elevations, only the leaves of moderate down-shifters had higher 15 N discriminations at high vs. low elevations; (3) The leaves of strong down-shifters did not show significant elevation patterns of the discriminations; (4) The contrasting responses among the three types of shifters agree with their functional dissimilarity, suggested by their separate locations in a multitrait space. Taken together, the divergent shifts are associated with the elevation variations in environmental conditions and contrasting plant responses. The contrasting responses could result from the functional dissimilarity among species. Therefore, a detailed understanding of both local environmental variations and species-specific responses can facilitate accurate predictions of species shifts to inform local species conservation.