Mappers - ECMWF is just about to transfer its operational suite from IBM to CRAY machinery, and so we have had exactly the same model running in parallel on the two systems. The plot compares day 5 output from the two computers. The differences in HRES (right hand side) are quite large, notably In E Canada and S of Alaska, and probably 'above average' overall. Though there is nothing profound in this I just thought it was nice example, or salutary lesson, for anyone believing details in a day 5 deterministic run. We have compared the accuracy of the parallel runs over time and they are almost identical.

Tim