Page History

...

The system uses historical re-forecast runs on dates in past years relating to the date (i.e. month and day) of the current ensemble run.   Re-forecasts are based on an ensemble of forecast members ideally using the same model techniques and physics as the current model.   The re-forecast ensemble uses the appropriate reanalysis field for initialisation.   Perturbations are applied to all but the control.  This is similar to the operational ensemble, but does not involve any data assimilation.  The perturbations derive from singular vectors (SVs) plus geographical averages of ensemble of data assimilations (EDA).  The EDAs are perturbations that have been computed operationally over the most recent 12 months.  This approach means that the flow-dependence inherent in operational EDA perturbations is missing in the re-forecasts.  Stochastic physics are also used during the re-forecast runs, as in operational ensemble runs.  

The set of re-forecast ensembles is based on previous dates which can stretch back several decades.   They differ in number and detail according to the IFS model configuration and are the basis for deriving the corresponding model climates.  These are described in the relevant section for medium range M-climate, extended range ER-M-climate, and seasonal S-M-climate.

The procedures adopted for using re-forecasts allow for seasonal variations and model changes to be taken into account.  But note the model climates (M-climate, ER-M-climate, or S-M-climate) can nevertheless be different from the observed climate.

...

• the land surface scheme used by ERA-Interim differs from the current IFS model version.  An equivalent "offline land surface reanalysis" that is compatible with the current model version may be used instead. 
• reanalysis is not performed over open water surfaces and climatological reference data is used instead.       
• real-time IFS contains a lake model (FLake) while ERA-Interim did not.  Representation of temperature in the model climate may diverge in a systematic way from the actual changes within the forecast period.  These effects are particularly important over and around large lakes or inland waters (e.g. the Great Lakes - notably Lake Superior, Aral Sea, Caspian Sea and some others). 

Forecasters should consider possible deficiencies in model climates when considering extreme forecast index EFI data.  For an example of the effect, see Fig5.3.1 and Fig5.3.2.  Such effects have also appeared in the extended range ensemble and seasonal forecasts.

...

Fig5.3.1(left): Extreme Forecast Index (EFI) for 2m temperature for Days10-15, ensemble forecast run DT 00UTC 20 June 2017. 

Fig5.3.2(right): Cumulative Distribution Function(CDF) for 2 m temperature for Days10-15 in the middle of Lake Superior (red), with M-climate (black).  The initialisation techniques are different for real-time forecasts (using lake surface temperature observed by satellites), and for the re-forecasts (for which this information is not available).  This can lead to the model climate developing anomalously warm or cold lake surfaces and corrwspomdimg 2 m CDF temperature curve (black).   This affects subsequent extreme forecast index (EFI) and shift of tails (SOT) fields.   Here the realistic real-time forecast of 2 m temperature CDF (red) over Lake Superior is thus incorrectly flagged as having a strongly negative EFI value in Fig5.3.1.  

...

• Watch a comprehensive lecture on estimation of the model climate (re-forecasts).
  

...