If the previous method ignored
the outputs of the model at the fine scale and used the large scale
variables, with this one the opposite is done, the information provided
by the model at the large scale is ignored and the information
at the small scale is used.