Drivers of endophyte richness
To test the effects of host identity (i.e. host species), abiotic variables (maximum and minimum temperatures, temperature standard deviation, light intensity, relative humidity and soil moisture) and biotic variables (tree basal area per BC, tree height and distance to the forest edge) on ASV richness, GLMMs, with a Poisson distribution and a log-link function were used (Zuur et al. , 2009). Interaction terms between host identity and all abiotic and biotic variables were fit within the model and BC identity was included in the model as a random variable (McCulloch, 1997). Best subset modelling, based on the lowest AIC-value, was used to assess which predictor variables from the global model should be retained (Burnham and Anderson, 2002). The model was overdispersed, therefore overdispersion was corrected by employing the observation-level random effects approach (Lawson et al. , 1999; Elston et al. , 2001). Marginal and conditional R2-values were calculated (Nakagawa et al. , 2017).