Abstract
An examination of the scale interactions in predictability experiments is made using the NCAR Community Climate Model Version 3 (CCM3) at various horizontal resolutions ranging from T42 to T170. Both identical-model and imperfect-model twin experiments are analyzed, and they show distinctive differences from the classical inverse cascade picture of predictability error growth. In the identical-model twin framework, error growth experiments using initial errors confined to long and short scales are compared and contrasted. In these cases, error growth eventually asymptotes to an exponential growth of baroclinically active scales. In the imperfect-model twin experiments, errors rapidly disperse from scales technically beyond model resolution to a small amplitude, spectrally uniform distribution of errors in resolved scales. The errors in resolved scales further amplify in a quasi-exponential growth of the baroclinically active scales. Finally, the implications of these growth mechanisms for the necessary resolution in short- to medium-range numerical weather prediction are given under the assumption that the accuracy of current initial state estimates of the atmosphere remain fixed at their present level.
Corresponding author address: Dr. J. J. Tribbia, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000. Email: tribbia@ucar.edu
