Seasonal Prediction over North America with a Regional Model Nested in a Global Model

Abstract

Ensembles of winter and summer seasonal hindcasts have been carried out with an 80-km resolution version of the National Centers for Environmental Prediction, Environmental Modeling Center Eta model over the North American region. The lateral boundary conditions for the Eta model are prescribed from Center for Ocean–Land–Atmosphere R40 atmospheric general circulation model integrations, which used observed atmospheric initial conditions and observed global sea surface temperature (SST). An examination of 15 seasonal winter hindcasts and 15 seasonal summer hindcasts shows that the nested model reduces the systematic errors in seasonal precipitation compared to the global model alone. The physical parameterizations, enhanced resolution, and better representation of orography in the Eta model produces better simulations of precipitation, and in some cases, its interannual variability. In particular, the precipitation difference between the 1988 drought and 1993 flood over the United States was much better simulated by the nested model. The predictions of circulation features were generally as good or better than those from the global model alone. Estimates of external (SST forced “signal”) and internal (dynamics generated “noise”) variability were made for both the global model and the nested model predictions. Contrary to the expectation that a higher-resolution model would have higher internal-dynamics-generated variability, the signal, noise, and signal-to-noise ratios of the near-surface temperature and precipitation fields were generally quite similar between the nested model and the global model predictions. In the winter season the nested model had larger signal-to-noise ratios in both temperature and precipitation than did the global model alone.