The mechanisms driving the community spatial structure are central questions of community ecology. Previous studies of niche and neutral models have mainly focused on species richness area relationships or species abundance distributions separately, and at several discrete scales. However, considering only the number of species while ignoring information regarding species abundance, or considering changes in the relative abundance of species at only a small number of scales, is not sufficient to fully understand how diversity varies with scale, let alone to explore the mechanisms of diversity change. Multifractal analysis not only provides information regarding the relative abundances of each species and the diversity in groups of species with particular abundance but also offers a unique insight into investigating the community spatial structure. In the study, niche, neutral and combined processes were integrated into spatial point pattern models at multiple spatial scales. We assess the relative contribution of niche and neutral processes in shaping the community spatial structure. Our results show that the mechanism shaping community spatial structure strongly depends on the sensitivity of diversity measure to species frequencies (moment order q): for q < 1, it is dominated by the joint effects of dispersal limitation and habitat filtering; while it is driven by dispersal limitation only when q ≥ 1. The multifractal analysis succeeded in separating the mechanism under species richness scaling yet failed to distinguish the relative abundance scaling. The generality of our findings should be further tested in other types of forests with different habitat conditions variation. However, our work is sufficient to suggest that ecologists and policymakers should consider the species richness and relative abundance simultaneously across scales and sensitivity of diversity metrics to the contribution of rare and common species when testing the mechanism underlying the community spatial structure.