Spatial variables
We included spatial variables implemented as Moran’s eigenvector maps (MEMs, Dray et al. 2006). We calculated MEMs based on the GPS coordinates of the sampled host plants, with the assumption of positive autocorrelation. We estimated the significance of individual eigenvectors with a permutation test for the Moran’s I statistic and included all the eigenvectors with Moran’s I >0.7, as well as the last significant one, thus representing both coarse and fine scale signal of significant spatial autocorrelation. We used the R environment (R Core Team 2020) and packages ‘spdep’ and ‘adespatial’.