Changing only the initial surface conditions only does not have a reproducible effect on rainfall. However, the combination of initial surface and upper air fields do have a reproducible influence on rainfall. In this case, the effect of the 1988 SSTs appears to be consistent with the influence of 1988 initial conditions, in both reducing rainfall relative to 1987.
f)ECMWF: ensemble scores for 14 winter and spring cases
Fig 7 shows the ensemble scores for all 14 winter and 14 spring cases in terms of 500hPa height anomaly correlation over the northern hemisphere. It can be seen that in both winter and spring cases the correlations are predominantly positive, with some years such as 1982, 1988 having particularly high skill in winter. In spring, there is almost uniformly high skill, consistent with earlier studies.
3 Development of a coupled ocean atmosphere model for seasonal prediction
3.1The ocean model
The ocean model is a global version of the HOPE model. In the horizontal, a staggered Arakawa E grid is used with a resolution to match that of the atmosphere. The latter is a spectral model but uses a gaussian grid at T42 resolution for the surface fields which corresponds roughly to a resolution of 2.8 degrees in both latitude and longtitude. In the neighbourhood of the equator, however, the meridional resolution in the ocean is increased to 0.5 degrees. This asymmetry in the grid spacing between the meridional and zonal directions means that the effective zonal separation is also reduced from 2.8 degrees to an effective resolution closer to 1.5 degrees. There are 20 levels in the vertical, spaced from 20m to 25 m apart in the upper 200m. The parametrisation of horizontal mixing of momentum is by harmonic and biharmonic terms and by a shear dependent mixing term similar to that proposed by Smagorinsky. The horizontal mixing of heat and salt are by harmonic and shear dependent mixing. There is no biharmonic mixing of heat and salt. The harmonic mixing coefficient is larger, however: 2500m2/s compared with 200 m2/s for momentum. Vertical mixing is basically a Pacanowski and Philander 1981 Richardson number dependent formulation. It is well known that this does not provide sufficient mixing in the near surface however and so additional forms of vertical mixing are frequently used, such as a Krauss-Tumer mixed layer formulation. In the version of HOPE used in this study a simpler form, but with similar effect, is used. In the mixed layer, defined as the layer in which the temperature difters by 0.5 degrees of the surface temperature, mixing is enhanced, this enhancement being largest at the surface and dropping by 40% every model level lying within the mixed layer. The enhanced mixing applies equally to momentum, temperature and salt.
3.2The analysis system
The ocean initial conditions are obtained either using data assimilation, or not using any assimilation of sub surface data. The latter is known as the control. In the control run the ocean initial conditions are obtained by forcing the ocean model with the fluxes of momentum, heat and fresh water from the operational forecast system at ECMWF or from
