Abstract
Understanding the distribution pattern and maintenance mechanism of species diversity along environmental gradients is essential for developing biodiversity conservation strategies under environmental change. We have surveyed the species diversity at 192 vegetation sites across different steppe zones in Inner Mongolia, China. We analysed the total species diversity (γ diversity) and its composition (α diversity and β diversity) of different steppe types, and their changes along a precipitation gradient. Our results showed that (i) β diversity contributed more than α diversity to the total (γ) diversity in the Inner Mongolia grassland; the contribution of β diversity increased with precipitation, thus the species-rich (meadow steppe) grassland had greater contribution of β diversity than species-poor (desert steppe) grassland. (ii) All α, β and γ species diversity increased significantly (P<0.05) with precipitation, but their sensitivity to precipitation (diversity change per mm precipitation increase) was different between the steppe types. The sensitivity of α diversity of different steppe community types was negatively (P<0.05) correlated with mean annual precipitation, whereas the sensitivity of β and γ diversity showed no trend along the precipitation gradient (P>0.10). (iii) The α diversity increased logarithmically, while β diversity increased exponentially, with γ diversity. Our results suggest that for local species diversity patterns, the site species pool is more important in lower precipitation areas, while local ecological processes are more important in high precipitation areas. In addition, for β diversity maintenance niche processes and diffusion processes are more important in low and high precipitation areas, respectively. Our results imply that a policy of "multiple small reserves" is better than one of a "single large reserve" for conserving species diversity of a steppe ecosystem, and indicate an urgent need to develop management strategies for climate-sensitive desert steppe ecosystem.