An overdue post, but you can find Dan's work on biases in global tree cover products here:
Dan's research, titled, "Identifying Biases in Global Tree Cover Products: A Case Study in Costa Rica" was published in Forests last fall. In short, global tree cover products can have slight to severe challenges in mapping tree cover in agricultural landscapes and along gradients of rainfall and topography. Any user of these products should read this paper--under- and overestimates ranged from 10% tree cover up to 80% tree cover.
Rather than re-invent the wheel, I will just point to this blog post on my collaborator's site:
and this excellent news article by UMBC's Sarah Hansen:
You can find a link direct to the open-access paper on the Publications page of this site.