Subgrid Scale Physics in 1-Month Forecasts. Part I: Experiment with Four Parameterization Packages

Abstract

Four packages of subgrid scale (SGS) physics parameterization are tested by including them in a general circulation model and by applying the four models to 1-month forecasts. The four models are formulated by accumulatively increasing the elaboration and the sophistication of the physics. The first is the reference model (the A-physics); the second model (the E-physics) uses the Monin?Obukhov similarity theory for the fluxes of surface boundary layer, the turbulence closure scheme for the fluxes in the entire atmosphere, and subsurface soil heat conduction; the third model (the F-physics) replaces the cumulus parameterization by the Arakawa?Schubert method; and the fourth model (the FM-physics) enhances the SGS orography. One-month integrations are performed for eight January cases, with each case consisting of three different forecasts. Originally the forecast performance was expected to be a stepwise improvement with the elaboration of the SGS physics from the A to the FM, but the forecast results do not show up in such a simple way. The impact of these processes on the 1-month integration is subtle and yet significant. The superiority of the F-model over the A- and the E-models is evident in the last 10 days of the 1-month forecasts, though the performance of the E-model is consistently good, in comparison with the other models, in terms of root-mean-square (rms) error of geopotential height. It is likely that 80% condensation criterion in the E (instead of 100%) is at least partly responsible for the forecast deterioration in the last 10 days, compared with the F. The FM-model gives the lowest rms error, but the predicted transient eddies are extremely low, probably due to the excessively enhanced orography. The simulated global precipitation patterns are presented for the different models, and the drawbacks are discussed. The F- and the FM-models produce spatially smooth distribution of tropical rainfall. The 30-day forecast performance appears to be more sensitive to the initial conditions, rather than the SGS physics. The systematic errors in all of the models are substantial in magnitude, though they vary with the SGS physics.