Assessing the options for identifying critically important potential surface regions: Applications to nonadiabatic transitions