INTRODUCTION

Understanding the variation in species composition of ecological communities, commonly known as β-diversity, is a major challenge confronting ecologists (Sepkoski 1988; Tuomisto & Ruokolainen 2006; Anderson et al. 2011). Such researches can help to reduce the constant threat to biodiversity loss by contributing to the question of what factors control the distributions and abundance of organisms (Socolar et al. 2016). However, despite substantial efforts to disentangle the drivers of community assemblages (Legendre et al. 2005; Ferrier et al. 2007; Tuomisto 2010), scientists begin to realize that different ecological processes may result in very similar patterns of species composition (Myers et al. 2013).
The realization that the ecological processes are scale-dependent has assisted in uncovering essential features of the community assembly (McGill 2000; Barton et al. 2013; Chase et al. 2018). Processes of community assemblage are assumed to work as constraints with potentially varying strengths at different scales, which hierarchically determine the local community structures (Whittaker et al. 2001; Ricklefs & He 2016). For example, macroclimate (i.e. the overall climate of a large geographical area) is effective at large scales, while microclimate (i.e. the essentially uniform local climate of a small site or habitat) at small scales, (McGill 2000), and biotic interactions within neighborhoods are important at small scales (Huston 1999). Community structures are thus the result of hierarchical constraints at multiple scales (Scherrer et al. 2019).
Among the various ecological processes, environmental filtering and within-habitat spatial aggregation are considered to be the two major mechanisms invoking the spatial variations in species compositions (Plotkin et al. 2000; Seidler & Plotkin 2006; Laliberté et al. 2014; Kraft et al. 2015). Environmental filtering processes tend to support the survival of specific species in some habitats in preference to other species. Within-habitat spatial aggregations are clustering processes resulting from dispersal limitations, patchy extinction, or interactions among species within a given type of habitat (Kretzschmar & Adler 1993; Öhman et al. 1998; Cornell et al. 2007; Shen et al. 2009). One hypothesis regarding community assembly suggests that environmental filtering and spatial aggregation limit the survival and organization of potentially coexisting species (Myers et al. 2013; Xu et al. 2015). Specifically, environmental filtering processes progressively select the species which are best adapted to local conditions from the regional pool, first by large-scale environmental factors, and subsequently by fine-scale environmental factors (Diaz & Casanoves 1998; De Bello et al. 2013). Then spatial aggregation processes ultimately determine the spatial organization of the selected species (Cornell et al. 2007; Shen et al. 2009). However, to our knowledge, no research has been done to quantify the effects of environmental filtering and spatial aggregation on variations in species compositions across spatial scales.
According to Cornell et al. (2007), specific null models involving habitat features can effectively separate the effects of environmental filtering and spatial aggregation processes at a given spatial scale. On this basis, we divided the study region into different habitats at different spatial scales i.e. the region-zone, zone-area, area-district, plot-district and within-plot scales, ranging from the broad to fine, to examine the ecological processes. Specific null models were then applied by shuffling individuals among and within the various habitats at each spatial scale. The effects of environmental filtering which progressively select species from the species pool at the region to zone level, zone to area level, area to district level, and district to plot level, as well as the effects of spatial aggregation which ultimately shape the spatial organization of the selected species in the observed communities were hierarchically partitioned, as detailed in Figure 1.
With the exception of measuring the magnitudes of the ecological effects at each spatial scale, the focus was on the determination of how the effects of environmental filtering and spatial aggregation on β-diversity had vary along the latitudes. In addition, potential factors related to the environmental filtering processes were explored. According to the division criteria of the vegetation regionalization map of China (Zhang 2007), at the zone level, the species compositions were determined to vary mainly due to the changes in heat patterns from south to north, or along the elevational gradient; at the area level, the causes were mainly the changes in water-heat patterns with the medium geomorphic types; at the district level, local geomorphology was found to be the main reason. We thus related the effects of environmental filtering to a group of climate and plot variables.
Overall, the following hypotheses are formulated:H1 , the effects of the environmental filtering processes are stronger at the broader scales due to an increasing environmental heterogeneity with increasing spatial scale (Leboucher et al. 2019). H2 , environmental filtering at the broader scales increases with increasing latitudes where fewer species are often observed (Hillebrand 2004), which indicates stronger filtering effects from the regional species pool (Chu et al. 2019). Environmental filtering at the fine scales decreases with increasing latitude since the habitat heterogeneity at the lower latitudes is greater, which supports greater variation in species compositions (Xing & He 2019).H3 , the effect of the spatial aggregation at the within-plot scale decreases with increasing latitude. This is due to clustering caused by dispersal limitations and biotic competition which is assumed to be more dominant in species-rich communities when compared with species-poor communities (Myers et al. 2013).
In this study, we first quantified the hierarchical effects of environmental filtering and spatial aggregation on β-diversity patterns across five spatial scales. We also proposed one hypothesis to examine the relative importance of environmental filtering across scales, and two hypotheses to explain the variation of environmental filtering and spatial aggregation along latitudes at each scale. We showed that the pattern of β-diversity was mainly dominated by environmental filtering at the region-zone, district-plot scales, and spatial aggregation at within-plot scale. The environmental filtering at the broader scales had stronger effects at higher latitudes, while at the finer scales only at lower latitudes. The effects of spatial aggregation were more prominent at lower latitudes.