Soil fungal communities influenced complementarity effects
Six of the 16 mixed tree communities showed significant positive NE on productivity in the high water availability treatment and eight of 16 communities in the low water availability treatment (Table S6). Five communities showed significant positive CE in both treatments (Table S7). None of the communities showed significant SE (Table S8; Fig. S3).
The SEM that included fungal community composition explained 30% of the variability in NE compared to 13% for the null model (Fisher’s C = 38.462, P < 0.001 on 6 df, Fig. 2A, B; Table S9, S10). The SEM with fungal community composition explained 37% of variability in CE compared with 22% for the null model (Fisher’s C = 29.329, P < 0.001 on 6 df, Fig. 2C, D; Table S11, S12). The SEMs with fungal richness were not significantly better than the null models for NE or CE (NE: Fisher’s C = 3.404, P = 0.182 on 2 df; CE: Fisher’s C = 0.407, P = 0.816 on 2 df; data not shown).
For the models of NE and CE on productivity, functional diversity was positively associated with fungal NMDS axis 1 (std. coef. = 0.2, P < 0.001), tree functional identity (std. coef. = -0.59, P < 0.001) and species richness (std. coef. = 0.2, P = 0.004) affected NMDS axis 2, and NMDS axis 3 was independent of the exogenous treatment variables.
In the NE null model, tree species richness had a positive, direct effect (std. coef. = 0.33, P < 0.001) and functional identity had a marginally significant direct effect on NE (std. coef. = 0.14, P = 0.093; Fig. 2B). In the full model, species richness had a significant, but weaker direct effect on NE (std. coef. = 0.25, P = 0.002; Fig. 2A, Table S9). NMDS axis 2 was negatively (std. coef. = -0.31, P < 0.001) and the third NMDS axis was positively (std. coef. = 0.30, P < 0.001) related to the NE on productivity. Fungal NMDS axis 2 mediated negative indirect effects of tree species richness and functional identity on NE.
In the CE null model, tree species richness had a positive, direct effect on CE (std. coef. = 0.38, P < 0.001; Fig. 2D). In the full model, tree species richness had a significant, but weaker direct effect on CE (std. coef. = 0.32, P < 0.001; Fig. 2C, Table S10). Fungal NMDS axis 2 was negatively (std. coef. = -0.22, P = 0.024) and NMDS axis 3 was positively (std. coef. = 0.30, P < 0.001) related to CE. Fungal NMDS axis 2 mediated negative indirect effects of tree species richness and functional identity on CE.
The SEMs of NE and CE on productivity with plant pathogen and ECM richness provided a better fit than the null models (NE: Fisher’s C = 20.54, P < 0.001 on 4 df, CE: Fisher’s C = 15.81, P = 0.003 on 4 df, Fig. S4, Tables S13 and S14). In the NE and CE models, pathogen richness was negatively associated with functional identity (std. coef. = -0.4, P < 0.001) and ECM richness was driven by tree species richness (std. coef. = 0.2, P < 0.05). Pathogen richness alone was not associated with NE or CE (std. coef. = 0.1, P > 0.1). ECM richness alone was negatively related to NE (std. coef. = -0.36, P < 0.001) and CE (std. coef. = -0.29, P = 0.001). Pathogen and ECM richness had a negative, interactive effect on NE (std. coef. = -0.19, P = 0.014) and CE (std. coef. = -0.18, P = 0.025). Therefore, the interactive effect of pathogen and ECM richness mediated both positive and negative indirect effects of tree species richness on NE and CE.
There were no significant direct effects of water availability on the NE or CE on productivity in the null or full models. In the full models for these three variables, water availability had a significant, negative influence on NMDS axis 1 (std. coef. = -0.65, P = 0.02), but NMDS axis 1 had no significant direct effects on NE or CE (Fig. 2, Tables S9 and S11).