3.2 Rare taxa were mainly responsible for microbial spatial scaling patterns
We first investigated whether abundant and rare subcommunities followed similar spatial scaling patterns, such as TAR and DDR (Fig. 2a and b). The typical community richness and Bray-Curtis similarity were respectively used to quantify TAR and DDR. As a result, clear TAR (z = 0.494, P < 0.001) and DDR (d = -0.242, P < 0.001) patterns were observed for the bacterial communities (Fig. 2a and b). As expected, stronger TAR pattern was found for rare subcommunities (z = 0.517, P< 0.001) than abundant subcommunities (z = 0.029,P = 0.034) (Fig. 2a). Similarly, rare subcommunities (d = -0.447, P < 0.001) harbored stronger DDR patterns than abundant subcommunities (d = -0.12, P < 0.001) (Fig. 2b). For both TAR and DDR, rare subcommunities even showed stronger spatial scaling patterns than the whole community. The results suggested that rare subcommunities were mainly responsible for the spatial scaling patterns followed by microbial communities.
To verify the major contribution of rare subcommunities to the spatial scaling patterns of microbial communities, we extended both alpha- and beta-diversity to Hill numbers to analyze TAR and DDR patterns at different diversity orders. As such, the ambiguous definition of abundant and rare taxa can be well resolved by giving continuously decreasing weight on rare taxa. By setting the diversity order qto different values (here 0 ≤ q ≤ 2), different weight is given to microbial taxa with different relative abundance. The higher diversity order q is, the lower weight is given to rare taxa. Taking alpha diversity for example, the Hill numbers equal to community richness when q = 0, indicating all microbial taxa are equally treated. When the order q is set to 1 and 2, the Hill numbers respectively equals to the Shannon-Wiener and Simpson diversity index (Supplementary Figure 4). In the case rare subcommunities were mainly responsible for the microbial spatial scaling patterns, decreased DARq and DDRq slope coefficients with increasing q values were expected. As a result, sharply decreased slope coefficients of DARq and DDRq were observed when the diversity order q increased from 0 to 2 (Fig. 2c). Such results confirmed that rare subcommunities were mainly responsible for the observed spatial scaling patterns followed by microbial communities.