| description abstract | Prompted by the need for forecasts of changes in the length of day on 1-10 day time scales, this paper examines the quality of predictions of a proxy variable, namely the atmosphere's relative angular momentum (M) made by the medium-range forecast model (MRF)of the Nationa Meteorological Center during December 1985-November 1989. Skillful forecasts of M relative to persistence am produced by the MRF over its entire 1-10 day range, as found previously. Errors in the MRF are smaller than those of a damped persistence of anomaly empirical model only out to 8 dan however. Moreover, beyond about 3-4 dam MRF tommts of M for day N + 1 show no more skill than forecasts made by simply persisting the MRF prediction for day N, suggesting that significant room for improvement in dynamical forecasts of M still exists. Errors in the MRF foments of M are separated into their bias and nonsystematic components. Bias errors became especially prominent with the introduction of the most recent version of the MRF examined here, MRF88, whereas random errors in the M forecasts appear not to have been affected by model changes. Both types of errors in the M forecasts can be traced to problems with forecasts of the zonal mean zonal wind, [u], in the tropics. Bias errors in MRF88 forecasts of the globally integrated quantity M are large despite notable reductions in biases in [u] forecasts locally since the MRF was fim introduced. Evidence is offered that the pattern of bias in 10-day forecasts of [u] develops much earlier in the forecast cycle. | |
