Plasticity of relative root distribution in response to root neighbours and edaphic heterogeneity 
Our hypothesis that relative root distribution would exhibit high plasticity in response to edaphic heterogeneity and root neighbours was not fully supported. For most species, relative root distribution was neither affected by edaphic heterogeneity nor by the presence of specific root neighbours (Fig. S3), suggesting that neighbourhood species composition is not a good predictor of a focal species’ root depth. Likewise, most species’ relative root distributions were unaffected by heterospecific root richness or abundance (Tables S5-S7; Figs. S3, S4), as observed in temperate forests and mountain grasslands (Valverde-Barrantes et al. 2015; Herben et al. 2018). Similarly, Cahill (2003) reported that grassland species did not exhibit substantial changes in root-shoot ratios with increasing below-ground competition, suggesting that increased heterospecific root mass may not necessarily be indicative of higher competitive pressure exerted on the focal species. However, these findings contrasted with previous studies reporting apparent neighbour-induced rooting plasticity (e.g. Cahillet al. 2010; Belter & Cahill 2015; Zhang et al. 2019). An explanation for the discrepancy might be that, in contrast to greenhouse experiments that have usually involved two species (Cahill et al.2010; Belter & Cahill 2015), individual species in natural communities experience high niche overlap and diffuse competition. Specifically, in competitive neighbourhoods composed of multiple species, the addition or removal of a single neighbour may not have a significant effect on the competition experienced by the target plant (Dyer & Rice 1997). Another reason might be that in species-rich communities, species exhibit varied rooting plasticity, both directionally and magnitudinally, in response to different neighbours, resulting in overall neutral adjustments of root depth (Litav 1967; Brisson & Reynolds 1994; Caldwell et al.1996; Schenk et al. 1999).
Root proliferation usually decreases as increasing heterospecific root abundance, presumably due to intensified competitive pressure and stronger chemical inhibitive effects (Brisson & Reynolds 1994). However, we found that the increase of one species’ root abundance did not reduce other species’ root abundance (Fig. 5). Instead, the pervasive positive associations of species pairs suggest concurrent root proliferation of co-occurring species into the same soil zones (Fig. 5b, 5c; Belter & Cahill 2015). Nevertheless, based on the positive association among two species, we could not conclude that such species can mutually and directly facilitate their root growth, as such correlations could simply result from the collinearity of species root abundance with other unaccounted-for factors. For instance, similar responses of root proliferation to limiting nutrients or soil microbes may lead to the co-occurrence of the paired species (Rog et al.2020). These uncertainties highlight the need for further manipulative experiments examining the direct and indirect paths of which root-to-root interactions play to better understand below-ground processes in complex forest communities (Belter & Cahill 2015; Zhanget al. 2019).
At the species level, the variation of relative root abundance in the 0-10 cm soil zone was high (CV: 38%-144%), suggesting that, in addition to the possible effects of edaphic heterogeneity and root neighbours, other factors also contribute to such variations. For instance, as root segments were randomly sampled, those sampled on the margin of the root system could be relatively shallowly placed, giving that root abundance generally declines with increasing distance from the parental tree stem (Jones et al. 2011). Such spatial arrangement of root systems could cause additional rooting variations independent of soil heterogeneity and root neighbours. However, prior studies also indicated that root-placement patterns were mainly affected by localized edaphic conditions, as root systems are modular, and each meristem can generate a potentially flexible and plastic root branches (Waiselet al. 2002). Nevertheless, the change of whole-plant root system arrangements still represents adaptive plasticity in response to root neighbours or soil conditions (Casper et al. 2003). Collectively, although there were substantial variations in relative root distribution left unexplained that could be attributable to tree ontogenetic stages or soil microbial interactions, our study suggests the relatively minor effect of interspecific interactions on a focal species’ root distribution.