Variability and Predictability in an Empirically Forced Global Model

Abstract

Climatological and predictability features of a simplified moist general circulation model are described herein. The simplified model is driven by empirical forcings designed to eliminate systematic errors and maintain computational efficiency. Resulting perpetual January climatological features of forced variables such as the tropospheric heights and rotational winds, as well as unforced variables such as the velocity potential, compare well with the observations. Unforced temporal variations in the midlatitude 500-mb geopotential, as well as the tropical circulation's intraseasonal oscillation, are also simulated reasonably well. Short-range persistence and predictability features of this model replicate geographical persistence and predictability features from simpler models and from numerical weather prediction. The streamfunction is highly persistent in the extratropics, less so in the tropical regions; similarly, the streamfunction is predicted better in midlatitude regions than in the tropics. By contrast, the velocity potential is more persistent in tropical regions but, like the streamfunction, is still predicted best in extratropical regions for short-range forecasts. At longer forecast ranges, the velocity potential is better predicted in the tropics than in midlatitudes. Interestingly, during prominent tropical intraseasonal oscillations, the model consistently demonstrates lower tropical forecasting skill. Predictions are more skillful during stagnant tropical periods.