Climate Dynamics, 27, 69-82.
Deborah Herceg
Department of Applied Physics and Applied Mathematics,
Columbia University, New York, NY.
Adam H. Sobel
Department of Applied Physics and Applied Mathematics and Department of Earth and Environmental Sciences,
Columbia University, New York, NY.
Liqiang Sun and Stephen E. Zebiak
International Research Institute for
Climate Prediction, Lamont-Doherty Earth Observatory
of Columbia University, Palisades, NY
Abstract
The Big Brother Experiment methodology of Denis et al. (2002) is applied to test the downscaling ability of a one-way nested regional climate model. This methodology consists of first obtaining a reference climate by performing a large domain, high resolution regional climate model simulation - the Big Brother. The small scales are then filtered out from the Big Brother's output to produce a data set whose effective resolution is comparable to those of the data sets typically used to drive regional climate models. This filtered data set is then used to drive the same nested regional climate model, integrated over a smaller domain, but at the same high resolution as the Big Brother - the Little Brother. Any differences can only be attributed either to errors associated with the nesting strategy and downscaling technique, or to inherent unpredictability of the system, but not to model errors. This methodology was applied to the National Center for Environmental Prediction Regional Spectral Model over a tropical domain for a one-month simulation period. The Little Brother reproduced most fields of the Big Brother quite well, with the important exception of the small-scale component of the precipitation field, which was poorly reproduced. Sensitivity experiments indicated that the poor agreement of the precipitation at these scales in a tropical domain was due primarily to the behavior of convective processes, and is specific to the Big Brother Experiment on the tropical domain. Much better agreement for the small-scale precipitation component was obtained in an extratropical winter case, suggesting that one factor explaining the tropical result is the importance of convective processes in controlling precipitation, versus the greater importance of large-scale dynamics in the winter extratropics. In the tropical case, results from two ensembles of five three-month seasonal simulations forced by GCM output suggest a considerably greater predictability for the small-scale stationary component of tropical precipitation than did the Big Brother Experiment.