These results, particularly the performance of the PX / ACM2 combination, suggest that it is important to
consider the interactions between various model physical parameterizations, and that it is not sufficient to vary one at a time relative to a control run in order to determine the best overall combination.