Land-use diversity can better predict urban spontaneous plant richness than impervious surface coverage at finer spatial scales.

Urban spontaneous plants, that are not intentionally propagated by humans and do not belong to the remnants of the natural habitats, not only occur in green spaces but are also distributed in diverse microhabitats in impervious surface areas. Impervious surface coverage is commonly used in studies on spontaneous plant diversity patterns in human-dominated landscapes; however, the role of habitat diversity (i.e., land-use diversity) has been overlooked. Here, we surveyed spontaneous plant composition and land uses (12 types) in 321 0.25 ha sampling sites on the Chongming District islands, Shanghai, to determine the role of land-use diversity in explaining species richness. We examined the linear relationships between species richness and land-use diversity, and quantified the importance of impervious surface coverage and land-use diversity using the random forest (RF) method. All these analyses were conducted for spatial scales from 0.25 to 5 ha in 0.25 ha increments. We found an overall positive relationship between species richness and land-use diversity, and the RF model predicted approximately 50% of the species richness variation at the smallest spatial scale. However, the positive relationship weakened with spatial scale increase, and a rapid decline in explanatory power occurred for all predictor variables in the RF model. Besides impervious surface coverage, both the vegetated and non-vegetated land-use diversity contributed substantially to the prediction of species richness at finer spatial scales. The findings clarify how land-use diversity, both in green spaces and impervious surface areas, affect urban spontaneous plant richness and should be considered in urban biodiversity conservation strategies at the neighborhood scale.