Path analysis of relationships between conditioning and responding plant communities mediated via soil abiotic and biotic parameters
Dissimilarity matrices were calculated between all samples. We used Bray-Curtis dissimilarity for plants, fungi and bacteria with above described transformations, and Euclidean distance for the abiotic parameters using the ‘vegan ’ package (R Core Team 2018; Oksanenet al. 2018). For plant communities, all plant species present in less than three samples were removed prior to analysis in order to diminish the effect of rare plant species. Dissimilarity matrices during conditioning and responding phases were calculated separately. Mantel tests were carried out to explore the correlations between the distance matrices using Pearson’s correlation coefficients with 999 permutations using a Monte-Carlo permutation test to test the a prioriassumptions that conditioning plants will change soil communities (e.g., Heinen et al. 2018; Morrien et al. 2017) and abiotic conditions (e.g., Bezemer et al. 2006; Zhang, Van der Putten & Veen, 2016), and that this in turn alters the performance of responding plant communities (plant-soil feedback, Ehrenfeld et al. 2005; Van der Putten et al. 2013). We created a model that reflects these pathways separately (Figure 3), to visualize how the conditioning plant community affects the responding plant community.