The principal scientific basis for seasonal prediction in the atmosphere is that lower boundary anomalies, associated principally with SST but also with land-surface effects, are at least partially predictable on timescales significantly longer than that of daily weather. Moreover, even though the atmosphere is chaotic, these lower boundary anomalies can have a significant impact on atmospheric variability (changing, for example, the statistical probability of blocking over a given region of the atmosphere).
The impact of lower boundary forcing on atmospheric variability is, however, dependent on position, season and the strength of the boundary anomalies. The purpose of the first component of the ECMWF research programme, is to quantify this dependence more explicitly.
In order to do this, ensemble integrations have been run over seasonal timescales with the observed prescribed SST. Each ensemble is 120 days long and consists of 9 members, started from consecutive 12z analyses. The integrations are being run over the ECMWF reanalysis period (1979-1993) using reanalysis initial conditions and observed SSTs taken from the reanalysis project. These SSTs are based on the UKMO GISST analyses until October 1981, and on Reynolds 01 analyses after this date. For each year, ensembles are made for the four seasons.
Results will be shown for the first five winter periods (1978/79 to 1983/4), which included the strong El Nino event that occurred in 1982/83. In addition to their use as intial data, reanalysis fields are also used to verify the ensemble integrations.
It is planned that multi-model multi-initial-condition ensemble results will be produced since other modelling groups are involved in the project: Meteo-France (ARPEGE T42L3 1 cycle 12d), Electricite de France (ARPEGE T63L3 1 cycle 12d), UK Meteorological Office (Unified Model, 2.5x3.75 L19, version 3.4).
b)systematic error
It is important to judge the relative performance of a model with reference to the model's systematic error as it is well known that the extratropical response to El Nino depends strongly on the model's climatology, even if the response is expressed in terms of a deviation from the model's climate. Fig 1 shows the 500hPa height systematic error for the last 90 days of the 5 winter experiments (ie the mean error for JFM 1979-1983) illustrating a tendency for excessive westerlies in lower midlatitudes, and, in particular, a negative height error maximum to the west of the British Isles in the mid Atlantic.
c) response to 1982/83 SST winter anomalies
As mentioned above, the winter 1982/83 was associated with an intense El Nino event. The SST anomalies for this winter are shown in Fig 2b. For comparison, SST anomalies for the winters 1979/80 and 1983/84 are shown in Fig 2a and 2c respectively. In Fig 2 the
