ASSESSING A GCMS SUITABILITY FOR MAKING SEASONAL PREDICTIONS

This study investigates the predictability of seasonal mean circulatio n anomalies associated purely with the influence of anomalous sea surf ace temperatures (SSTs). Within this framework, seasonal mean atmosphe ric anomalies on a case by case basis are understood to consist of a p otentially predictable boundary-forced component and an unpredictable naturally varying component. The predictive capability of an atmospher ic general circulation model (AGCM) for seasonal timescales should the refore be assessed in terms of the average skill over many cases, sin; it is only then that the boundary-forced predictable signal in observ ations can be identified. To illustrate, experiments for 1982-1993 usi ng two versions of an AGCM are presented. The models, referred to here as MRF8 and MRF9, differ in the parameterization of a single process. Each model is run nine times for the 12 years using different initial conditions but identical observed global SSTs. The nine-member ensemb le mean anomalies for each season in 1982-1993 are compared with obser ved anomalies over the Pacific-North American (PNA) region. Several di fferent measures of the impact of SST boundary forcing on the extratro pical flow suggest that MRF9 is a better model for seasonal prediction purposes. The two AGCMs have substantially different zonal-mean clima tologies in the Tropics and subtropics, with MRF9 significantly better . It is argued that the improved mean flow in MRF9 enhances its midlat itude sensitivity to tropical forcing. The results highlight the impor tance of continued GCM development and give reason to hope that an eve n better model would lead to further improved forecasts of seasonal an omalies over the PNA sector.