Spatial analysis
Individual methylation rate and epigenetic variation may be associated with geographical distance. We used principle components of neighbourhood matrices (PCNM) implemented in the R package “vegan” (Oksanen, 2015) to model those relationships. Moran’s eigenvector mapping(MEM)was used to calculate spatial variables with R. Before calculating the semivariance, polynomial trend-surface regressions on each variable were performed using the method described in Chen (Chenet al ., 2016). Residuals from the trend-surface regressions were used to generate the semivariograms and the confidence envelopes in spatial autocorrelation analyses. In short, in each lag class, sample values were randomly selected and a semivariogram was calculated based on the selected data in 10,000 randomisations. A 95% confidence envelope was then obtained by calculating the 2.5 and 97.5 percentiles of the randomly generated semivariance values in each lag class.