DISCUSSION

In this study, we compared and analyzed the variations in species composition in the observed and randomized communities at five different spatial scales. Our study is the first to quantitatively partition the hierarchical effects of environmental filtering and spatial aggregation processes on the variations in species compositions of forest communities. The results revealed that the community structure is a result of both the ecological processes of environmental filtering and spatial aggregation at varying strengths depending on the spatial scale. Specifically, three hypotheses regarding the variations in the ecological effects across spatial scales and along latitudinal gradients, as well as the key factors of environmental filtering were examined and discussed in the present study to reveal the mechanisms that determine the origin and maintenance of the studied local community assemblages.
Our study agrees with a growing number of studies which have assumed or verified the hierarchical nature of ecological effects (Whittaker et al. 2001; McGill 2010; De Bello et al. 2013; Chalmandrier et al. 2013; Scherrer et al. 2019). In related investigations, De Bello et al. (2013) tested the relative influences of multiple environmental filters on plant functional trait structures. The results confirmed that the large-scale environmental filters tended to select a pool of species adapted to a specific site, and then the finer-scale filters determined the species abundance and local species coexistence. Scherrer et al. (2019) quantified the relative importance of habitat filtering and limitations of similarity in species traits in grassland environments. It was found that habitat filtering was the dominant assembly process at the plot level, with diminished effects at the subplot level, whereas limiting similarity prevailed at the subplot level, with weaker average effects at the plot level. To be more concrete, our study partitioned the ecological effects layer by layer across the five examined spatial scales, ranging from the region level which covered the entire natural forest areas of northeastern China, to the plot level (0.1 ha) which was on a smaller scale but sufficient for tree community sampling (Kraft et al. 2011). The results indicated that the species compositions of the local communities were controlled by both the up-scale environmental filtering and the down-scale spatial aggregation. The effects of the environmental filtering and spatial aggregation processes on β-diversity at the five examined spatial scales were all significant. However, it was determined that the environmental filtering at the region-zone and district-plot scales, as well as spatial aggregation at the within-plot scale, were the most dominant effects.
Inconsistent with H1 , the effects of the environmental filtering did not vary monotonically with the spatial scales. The results suggest that the degree to which the environmental conditions filtered the species from a large pool to a small pool may varies with the spatial scale. The species compositions of each plot were mainly affected by the environmental filtering at the smaller district-plot scale, less than by that at the broader region-zone scale. These results are not in agreement with those obtained by Cornell & Hughes (2007), who demonstrated that the environmental filtering in coral community assemblages did not change appreciably with spatial scales due to the habitat differences being the same regardless of the scale. We speculate that the variations across the spatial scales in this study are due to the fact that the effects of the environmental filtering at the region-zone scale produced marked effects mainly at the higher latitudes, while the effects at the district-plot scale are conspicuous throughout the study area (Figure 4 and 5). The results highlight the fact that the environmental factors which filtered the species from the regional pool to the zone level as well as from the district to the plot level are crucial for the species compositions in northeastern China.
Consistent with H2 , it was found the environmental filtering at the broad scales had positive correlations with latitude, but negative correlations at the finer scales. The positive correlations at the broad scales indicates that fewer species had adapted to the environments at the higher latitudes (Qiao et al. 2015), which was linked to the theory of the conservation of cold tolerance across the species (Algar et al. 2009). This was found to be in line with the results presented by Freestone & Osman (2011), who found that the proportion of species from the regional species pool that were present at the local scale had increased from the tropics to the temperate zone, thereby demonstrating that higher-latitude communities may experience greater influencing effects of species filtering from the regional pool than communities at lower latitudes. At the fine scales, the negative relationships between the environmental filtering effects and the latitudes reveal that more pronounced environmental filtering had resulted from greater habitat heterogeneity (Pianka 1966), which subsequently contributed to higher β-diversity values at the lower latitudes (Xing & He 2019). Myers et al. (2013) found that among forest plots spanning local scales, environmental filtering explained a larger proportion of the variations in β-diversity (after correcting for sampling effects) in temperate forests when compared with those of tropical forests. Our results emphasized the fact that the β-diversity values along the latitudinal gradients had responded differently to environmental filtering at the different spatial scales. Therefore, hierarchical analyses are necessary in view of possible confounding effects across multiple scales.
Our study confirmed that environmental filtering effects are related to different variables at each spatial scale. Part of the results are consistent with the priori criteria of the vegetation regionalization map of China, in which temperature and precipitation factors are always significant variables that may have resulted from different regulators. For example, at the broad scale (e.g. the region-zone scale) the effects are mainly mediated by latitude and elevation regulating factors (Table 4). However, at the fine scale (e.g. the district-plot scale), the local topographical environmental conditions, such as slope and aspect, are the dominant regulators (Table 4). In addition, some further clarifying details were obtained, including that wind speed and solar radiation were important filtering factors at the broad scales. They may have influenced the abundance and distributions of species by altering species dispersal potentials (Tackenberg et al. 2003) and creating heterogeneous light environments (Parks & Mulligan 2010). We also found that soil depth was a significant variable at the region-zone scale, and litter thickness was a significant variable at both the zone-area and district-plot scales. These findings reveal that soil attributes also contributed to the environmental filtering processes. Overall, the identified variables were sufficient to explain the environmental filtering effects at the region-zone scale (approximately 70% of the variation). Additional variations could probably be explained if more soil variables would have been available (John et al. 2007; Jones et al. 2008), particularly at the zone-area, area-district, and district-plot scales where the explained variances (approximately 20 to 30%) were much lower than those at the region-zone scale.
Spatial aggregation can reflect interactions between individuals and their close neighbors, which emphasizes the fundamental importance of the local distributions of trees within communities. We found that the effects of spatial aggregation in this study were significant and remarkable, which provided support for the theory that spatial aggregation is one of the dominant factors underlying the variations in species compositions (Cornell et al. 2007; Flinn et al. 2010).
Consistent with H3 , our study showed that the effects of spatial aggregation decreases with increasing latitude, indicating that the spatial aggregation was more dominant in the high-diversity communities. These results are causing the negative gradients in the observed β-diversity values along latitudes (Figure 4 and 5). In previous studies, the index of beta-deviation was considered to be closely related to spatial aggregation (Kraft et al. 2011; Qian et al. 2013; Xu et al. 2015; Xing & He 2019). Kraft et al. (2011) showed no global latitudinal gradient of beta deviation in global extent; Qian et al. (2013) showed a negative gradient in New World North and China, but no relation in New World South; Xu et al. (2015) observed a positive gradient in New World South, but no relation in a global extent and New World North. Xing & He (2019) found a strong unimodal latitudinal gradient in North America. The reasons for the various relationships may have been due to the different spatial extents among the studies (Patrick & Yuan 2019). Therefore, considering that spatial aggregation effects along geographical gradients could be various (positive, negative, unimodal, or non-relationship), specific spatial extents should be noted in future studies.
In conclusion, the stepwise partitioning of the environmental filtering and spatial aggregation processes used in this study highlighted the scale dependencies of the ecological effects on community structures. Accordingly, we recommend that greater emphasis should be placed on the scale effects of ecological processes (McGill 2010). Our study also emphasized that the divisions of spatial scales should be optimized to match the environmental patches which maximize environmental heterogeneity (Viana & Chase 2019). This will be particularly important due to the fact that the specific patterns and definitions of spatial scales will affect the variations in the data, and thereby the conclusions (Jelinski & Wu 1996).