Abstract
Using a 12-component spectral model derived from the barotropic vorticity equation, updating experiments were performed at intervals ranging from one to six time steps, where one time step was equivalent to about 2.4 hr. The Monte-Carlo and stochastic dynamic methods were used to determine a representative initial error growth for the model. In the first groups of experiments, it was found that when all 12 components were updated at intervals of five time steps or less, the rms vorticity errors eventually converged to zero; but when the updating interval was increased to six time steps, there was no tendency for convergence. In the second group of experiments the effect of not updating the smallest scales (sub-synoptic or mesoscale) was considered. The general result was that it was still possible to determine large-scale features rather well with fairly infrequent updating (three and four time steps) and model resolution to the intermediate (synoptic) scales, but intermediate-scale features could be recovered with good accuracy only when updating was done quite rapidly (every time step) or if smaller scale resolution was retained.
