This is just simple horseshoe artefact, which arises from linear ordination analysis (PCA in this case) when applied on heterogeneous data. I used simulated community dataset, structured by one artificial “environmental” gradient, and applied principal component analysis (PCA) on it. The R code which I used to generate this figure is below:

library (vegan)
simul1.spe <- read.delim ('https://raw.githubusercontent.com/zdealveindy/anadat-r/master/data/simul1-spe.txt', row.names = 1)
#png (width = 128, height = 98, bg = NA)
par (mar = c (2,2,1,1))
ordiplot (rda (log1p (simul1.spe)), display = 'si', axes = F, type = 'n')
points (rda (log1p(simul1.spe)), bg= 'white', pch = 3, col = 'red', cex = .7)
title (xlab = 'PCA1', line = .5, ylab = 'PCA2', cex.lab = 0.8)
#dev.off ()

(you may remove # to get *.png file with the exact size of logo and without white background)