Testing vertical root segregation in root neighbourhoods
Vertical root segregation was tested in two complementary ways using all the identified root morphotypes (i.e. a specific species composed of several root segments) belonging to 109 species sampled from 604 root neighbourhoods. First, we tested whether species were vertically segregated simply by exhibiting different root-placement patterns (i.e. root depth) within the 0-30 cm soil zones (Fig. 1a, 1c). Second, we sequentially tested whether species were vertically segregated by avoiding placing similar relative root abundance within the 0-10, 10-20 and 20-30 cm soil zone, respectively (Fig. 1b, 1d).
As each root morphotype was assigned a specific rooting type (Fig. 1) that was treated as a categorical functional trait, we constructed a trait dendrogram depicting the pairwise functional distance of all the identified and coded morphotypes based on a Gower Distance and determined the mean pairwise functional distances (MPD) of root neighbourhoods (Gower 1971; Laliberté et al. 2014; Swenson 2014). We then constructed 999 null communities for each root neighbourhood, from which the mean and the standard deviations of the 999 simulated MPDs were determined using the picante package (Kembel et al.2010). Null communities were constructed by randomizing the pairwise functional distances of the coded morphotypes by shuffling the tip names of the trait dendrogram, keeping species occurrences constant. Finally, we determined the standardized effect size of the functional MPD (SESMPD) to quantify the functional similarity of co-occurring species based on their root-placement patterns following the equation:
SESMPD =\(\frac{(\text{MPD}_{\text{obs}}\ -\ mean\ (\text{MPD}_{\text{null}}))}{sd\ (\text{MPD}_{\text{null}})}\),
where MPDobs represents the observed MPD. The mean (MPDnull) and sd (MPDnull) represent the mean and the standard deviations of the 999 simulated MPDs, respectively. Root neighbourhoods with a positive SESMPD indicated functional overdispersion (i.e. vertical root segregation) with values greater than 1.96 indicating significant segregation, whereas root neighbourhoods with a negative SESMPD indicated functional clustering (i.e. vertical root aggregation) with values below -1.96 indicating significant root aggregation (e.g. Swenson 2014). The absolute value of SESMPD indicated the magnitude of the difference between the observed and the simulated MPD. Similarly, we quantified root segregation based on species relative root distributions simply by replacing the root-placement-pattern-based trait dendrogram with a trait dendrogram constructed based on species relative root distribution. In addition to using MPD, we also used the standard deviation and variance of relative root distributions to evaluate the similarity of root depths among co-occurring species. For a specific soil zone, we constructed a trait dendrogram based on species relative root abundance in the soil zone.
Under the prediction of vertical root segregation, either by way of avoiding co-occurring or placing similar relative root abundance within the same soil zone, we would expect pervasive functional overdispersion in root neighbourhoods, which could be supported by the following two observations. First, there existed a higher proportion of root neighbourhoods that exhibited functional overdispersion than clustering. Second, and more importantly, that the difference between the observed and simulated MPD, measured as the mean of the SESMPD, was significantly higher than the expectation of zero when using a two-tail Student’st test (Yang et al. 2014).
All statistical analyses were performed in the R 3.6.0 statistical software (R Development Core Team, 2019).