M-climate is only re-created each Monday and Thursday.  This can have a small detrimental impact upon the extreme forecast index (EFI) and shift of tails (SOT) values, particularly in spring and autumn when day-to-day climatic changes are at their greatest.

In Fig5.3.5-1, the apparent decrease of the magnitude of the EFI during the sequence of forecasts for data times 00UTC 05 May to 12UTC 08 May is a response to a change in the M-climate with lead time.  Inland regions at higher latitudes warm quite rapidly in springtime and the M-climate 2 m temperatures rise steadily with increasing forecast lead time.  Therefore a forecast 2m temperature for five days ahead, though realistic, may appear significantly colder than the corresponding M-climate value for five days ahead and justify a high EFI.  However, the same forecast 2 m temperature for one day ahead will not appear quite so significantly colder than the cooler M-climate value for one day ahead.  Thus the EFI signal may become less negative and therefore less extreme even though the ensemble forecast remains identical from one day to the next.   The example shown is fairly extreme with SOT values nearing -1. 

Forecasts from data times of 00UTC and 12UTC use the same M-climate lead time for each day ahead.  The M-climate frames alter with every second frame of the EFI frame.

However, the final forecast in the sequence, DT 00UTC 09 May, used an updated M-climate as the basis for EFI and SOT computation.  So on the last two M-climate frames, the M-climate based on 4 May is replaced by that based on 8 May. This can be seen by the dark green boundary receding significantly to the north because the new M-Climate's focus is slightly later (and warmer) in the year.  This means that if the ENS forecast of 2m temp stays the same it will once again appear more extreme and this is shown with a return to a larger negative EFI value.

 

Fig5.3.5-1: The frame on the left shows a sequence of extreme forecast index (EFI) and shift of tails (SOT) charts.  These are from ensemble forecasts at 12hr intervals with data times from 00UTC 05 May to 00UTC 09 May 2017.  All verify for the same 24hr period 00UTC 09 May to 00UTC 10 May 2017.  Large negative EFI values (Dark Blue, Purple) give strong indication of a significantly cold event compared with corresponding M-climate climatology (shown in the frame on the right).  However, the large negative EFI values (Purple) from forecasts several days previously gradually become less extreme and less extensive (Dark Blues). Then return to large negative EFI values (Purple) on the last forecast.  This is because M-climate has been recalculated reflecting the progressively warmer springtime temperatures experienced in the past.  The forecast temperatures, though not significantly lower than previous forecasts, now appear significantly lower and more extreme compared with the latest M-climate. 


Fig5.3.5-2: The frame on the left shows Cumulative Density Function (CDF) of ensemble forecasts for northern Poland.  These are from ensemble forecasts at 12hr intervals with data times from 00UTC 05 May to 00UTC 09 May 2017.  All verify for the same 24hr period 00UTC 09 May to 00UTC 10 May 2017.  Each ensemble forecast has an associated colour. 

The black line is the M-climate for 24-48hr lead time. The frame on the right shows corresponding extreme forecast index (EFI) values for each ENS run.  The M-climate is reset each Monday and Thursday 00UTC forecasts.  This means the M-climate (black) curve will jump sideways on these days.  It jumps to the right (arrow B) during seasonal warming periods and to the left (arrow A) during seasonal cooling periods.

 At longer lead times (e.g. 5 days ahead) the M-climate will be warmer (i.e. to the right of the black line).  But it will gradually approach the indicated black line as the lead time reduces (direction of the arrow A).  The frame on the right illustrates the corresponding decrease of the magnitude of the EFI (i.e. less negative) between T+(96-120hr) to T+12-36hr) next to arrow A. 

The M-climate is reset before each Tuesday and Friday 00UTC forecasts.  Therefore the new M-climate curve will appear to the right of the M-climate shown in the diagram (direction of the arrow B).  Thus the EFI shows a jump to more extreme lower values next to arrows B in the right hand diagram.

In higher latitudes the most significant changes are in spring and autumn.  Changes are much weaker in summer and winter.


So in spring the EFI and SOT will undergo a slight decrease in magnitude (i.e. become less negative) with every other successive forecast.  However, when new M-climates are introduced into the Tuesday and Friday 00UTC forecasts there will be a jump towards lower (more negative) EFI values again.

A key conclusion then is that given a constant 2m temperature forecast, day-by-day M-climate warming will have the effect of reducing the EFI at progressively shorter lead times.  The opposite is true in autumn.

This issue has been partially addressed by changing the frequency of M-climate updates to twice per week; it was worse in the past when M-climate was only calculated once per week each Thursday.