Seasonal climate and variability of the ECMWF ERA-40 model

The climate and variability of seasonal ensemble integrations, made with a recent version of ECMWF model (used for ERA-40 production) at relatively high horizontal resolution (T(L)159), have been studied for the 10-year period, 1980-1989. The model systematic error over the Atlantic-European region has been substantially improved when compared with the earlier model versions (e.g. from the PROVOST and AMIP-2 projects). However, it has worsened over the Pacific-North American region. This systematic error reduces the amplitude of planetary waves and has a negative impact on intraseasonal variability and predictability of the PNA mode. The signal-to-noise analysis yields results similar to earlier model versions: only during relatively strong ENSO events do some parts of the extratropics exhibit potential predictability. For precipitation, there is more disagreement between observed and model climatologies over sea than over land, but interannual variations over many parts of the tropical ocean are reasonably well represented. The south Asian monsoon in the model is severely weakened when compared to observations; this is seen in both poor climatology and interannual variability. Overall, comparing the ERA-40 model with earlier versions, there seems to be a balance between model improvements and deteriorations due to systematic errors. For the seasonal time-scale predictability, it is not clear that this model cycle constitutes an advantage over the earlier versions. Therefore, since it is not always possible to achieve distinct improvements in model climate and variability, a careful and detailed strategy ought to be considered when introducing a new model version for operational seasonal forecasting.