I have a dataset with 300 panels for the period 1990 to 2015. My main variables are yield, temperature and a set of controls. I need to estimate a model by regressing yield on temperature and the controls. Then I need to estimate what would be the change in yield if the temperature rises by 2.6 percent (this figure is based on 2040 to 2060 prediction scenario of the IPCC under RCP4.5).
I first run an Arellano Bond using xtabond2 (because I include lagged yield as regressors). This is my code.

Code:
xtabond2 LnYield L(1/2). LnYield Temp i.YEAR Control1 Control2, gmm(L.LnYield L.Control1 L.Control2,eq(diff) laglimits(1 .)collapse) gmm(L.LnYield L.Control1 L.Control2, eq(level) laglimits(1 2)) iv(Temperature, equation(level)) twostep small
My issue is how to estimate an equation by incorporating the 2.6 increase in temperature. I read some posts and the predict, forecast and margins help files (i.e https://www.stata.com/stata-news/news29-3/forecast/) but got a bit confused and I still am not clear whether (1) I need to first generate counterfactual data for the temperature and then predict yield or (2) there is some other way?