Models (not just in climate science, also in most other fields of science...and engineering and business) generally include tunable parameters. When you have some unaccounted for factor that is not changing much over the time frame you are modeling, so that the unaccounted for factor has a fairly consistent affect, you can tune those parameters to in effect build the unaccounted for thing into the model.

If the unaccounted for factor starts changing significantly over the time frame you are modelling, this will not work well and then you have to understand that factor and build it in.

From the article, the impression I get is that we are now learning the trees and other plants affect climate more than we might have thought, but I don't see anything suggesting that the amount of the affect is changing much and so it would likely be the kind of thing that gets accounted for in the tunable parameters of the climate models.