It shows that land cover based proxies provide a poor fit to primary data surfaces for biodiversity, recreation and carbon storage, and that correlations between ecosystem services change depending on whether primary or proxy data are used for the analyses.
Read article here (external link)
