I am using the R gbm and dismo libraries to create boosted regression trees (BRTs) of landscape CO2 emissions.
I am obtaining maps of CO2 estimates using the raster predict function. For example, if Stacked is my raster stack of input predictor variables, and BRT1 is my boosted regression tree model, then my spatially referenced map of CO2 estimates is obtained by:
CO2 <- predict(Stacked, BRT1, type="response", n.trees = 1000)
I would like to have an associated map of CO2 standard errors. I know that for other models (e.g. GLM) I can add se.fit=TRUE to obtain the standard error + the predicted value. As this does not work for BRTs, I am wondering what other options are available.
Many thanks in advance!