A new Perfect-Prog forecast system for spot temperatures has been developed for the medium-range forecast needs of the Canadian Meteorological Centre. The development of the new Perfect-Prog makes use of the NCEP reanalysis database which provides long and stable time series of atmospheric analyses, which are necessary for building reliable regression equations. The 20 year-old operational Perfect-Prog was based on a composition of non-consistent low-resolution 22 year-long series of analyses available at that time, and required strong bias correction to be applicable on current model outputs.

The choice of the development of this new Perfect-Prog system was partly driven by the need of a model-independent statistical post-processing technique to be applied on ensemble model outputs. In effect, the maintenance of a model dependent post-processing system such as MOS for a large number of models having various model errors is prohibitive. Also, because of the continuous increase of spatial resolution of NWP model output while the training NCEP database was fixed (T62 : 325 km grid-spacing), compatibility of the NWP model output resolution with the NCEP reanalysis training database was achieved by spectrally filtering the NWP output with spherical harmonics. The strategy developed to ensure resolution consistency between the NCEP reanalysis data in development mode and the operational driving NWP model in forecasts mode will be explained. Verification scores comparing the old and the new Prefect-Prog system will be shown.