Abstract: [Objective] Landscape fragmentation can influence species richness at various spatial scales. A better understanding of the response of plant species richness to various environmental factors and determining its variation among spatial scales, can provide help for the development of effective protection strategies focusing on species diversity.[Method] In a total of 38 natural vegetation patches within the Saihanba nature reserve, including 12 herbs, 11 shrubs and 15 natural forest vegetation fragments, sampling plots were established. We recorded a total of 360 species belonging to 196 genera and 54 families. Plant species were divided into 4 functional groups:annual herbs, perennial herbs, shrubs, and trees according to growth types. We calculated species richness based on the functional groups at patch and plot scale using first-order jackknife estimators and area weighted averages, respectively. We used a generalized linear model (GLM) to explore the relationships between species richness and abiotic factors, such as topographic conditions, soil, as well as factors describing landscape structure at the patch and plot scales. We used redundancy analysis (RDA) to test the relative effects of topographic, soil, as well as factors describing landscape structure on the richness of plant functional groups at the two spatial scales.[Result] Landscape structure, topographic, and soil factors explained more than 60% of the total variation in species richness of functional groups; The response of species richness to landscape fragmentation from strong to weak followed the order of trees, shrubs, perennial and annual grasses; At patch scale, richness of woody plant species (shrubs and trees) was influenced significantly by factors describing landscape structure, including patch area, shape index, isolation and patch type (P<0.05), whereas it was significantly related to soil factors including pH, organic matter, total P content and total K content at the plot scale (P<0.05). At patch scale, factors such as patch area, patch type, pH and total P content were strongly correlated with the richness of perennial grasses(P<0.01), while at plot scale, patch type and elevation were significantly associated with perennial grasses (P<0.05). The richness of annual grasses species at patch scale was significantly associated with patch type, total N content and total K content (P<0.01); species richness at plot scale showed highly significant relationship with the patch type and soil organic matter (P<0.01).[Conclusion] Species richness of four plant functional groups showed differences with respect to the response mechanism at patch and plot scales when landscape fragmentation was considered. Our study confirmed that the distribution of species richness was determined by the combined effect of different spatial scales and environmental determinants. Species richness at patch scale was mainly affected by the landscape structure factors and few soil factors, whereas it was significantly related to the soil factors and the patch type, a landscape structure factor, at plot scale. Therefore, it is necessary to clarify the conditions of specific functional groups and spatial scales in the elucidation of the response of species richness to the environmental factors. The mechanisms influencing the distribution of plant species richness can be revealed only when plant functional groups and relevant spatial scales are specified.

Key words: multi-scale, functional group, species richness, landscape fragmentation, generalized linear model, variation partitioning