Note: The Ensemble Control Forecast is identical to Ensemble Control Forecast (ex-HRES) output in Cycle49 and HRES output in earlier cycles.  They have the same resolution and are scientifically, structurally and computationally identical.  Ensemble Control Forecast output, Ensemble Control Forecast (ex-HRES) output and HRES output are fully equivalent where shown in diagrams.  At the time of some older diagrams, HRES had resolution of 9km and ensemble members had a resolution of 18km.


Plumes

This product shows, for user-defined locations, the 10-day evolution of three parameters:

  • 850hPa temperature.
  • 6h precipitation (up to the time shown).
  • 500hPa geopotential height.

Lines show the evolution of the medium range ensemble (dotted for each individual ensemble member, solid for Ensemble Control Forecast ).  Shading (temperature and height only) denotes the probability the value will fall in a particular range at a given time.

 

Fig8.1.6-1:  A plume diagram for Paris based on medium range ensemble run DT 12UTC 21 June 2023.  Shading denotes the probability the forecast value will verify within the coloured range.  

  • light green shading: <10% probability.
  • mid-green: 10%-30% probability.
  • dark green: 30%-50% probability.
  • very dark green: 50%-100% probability.

Probability of 500hPa heights is shown similarly using a blue shading.  Higher probabilities are shown with darker shades.

The 850hPa temperature plume becomes wider beyond Sunday 25 June as the forecasts of ensemble members diverge.   Most ensemble members show a clear fall in temperature with darker shading.  A few, including Ensemble Control Forecast forecast, maintain higher temperatures rather longer.

The 500hPa height plume shows most ensemble members show heights remain above 578dam until Wednesday 28 June before becoming more uncertain later.  Note a few ensemble members fall below 570dam on Monday/Tuesday 26/27 June suggesting a possible, though less likely, passage of an upper trough.


In contrast to Meteograms, plumes can show bi-modal characteristics.

  • Large-scale bi-modality shows uncertainty in the ensemble evolution.  An example might be bi-modal 500mb height plume, perhaps accompanied by a bi-modal 850mb temperature distribution.  Part of the ensemble may favour transition to blocking; the rest may favour a zonal regime. 
  • Local bi-modality shows location or timing uncertainties in smaller features.   An example might be bi-modal 850mb temperature plume and a 500mb height plume that is not bi-modal.  Members may predict a front or minor low differently resulting in quite different local weather forecasts.

It is important to distinguish between these kinds of bi-modality.  Clustering products can also help in the detection of large-scale bi-modality.


Fig8.1.6-2 A plume diagram for Cluj-Napoca, Romania  based on medium range ensemble run DT 00UTC 21 June 2023.  An example of tri-modal characteristics within a plume.  On Tuesday 21 June the ensemble forecasts of 850hPa temperature splits into three main branches which persist for about two days: one rising to 17°C, the other with slightly more ensemble members falling to 7°C.  The branch lying between them should be viewed with some caution as for a period there is little consistency between the evolution of members.  The forecast 500hPa heights show a split with some showing a temporary significant dip in contour heights while others show minor rise.  Note also that on 30 June, Ensemble Control Forecast  forecast (yellow) shows greater 500hPa geopotential height than almost any member of the ensemble.



Fig8.1.6-3: The ensemble mean and spread for 500hPa heights from the same forecast verifying at 00UTC 21 June.  High standard deviation on right hand chart near 25ºE due to uncertainty regarding the timing and positioning of the major upper trough near 25°E.  The Normalized standard deviation on left hand chart shows greater spread (and hence uncertainty) near 25ºE  than recent ensemble results.


Fig8.1.6-4: Two plumes from ensemble for Southampton, Data Time 00UTC 18 January 2017 (left) and Data Time 00UTC 19 January 2017 (right).  Upper plumes show 850hPa temperature variations among ensemble, lower plumes show 500hPa geopotential height variations among ensemble for each of the two forecasts.  The ensemble mean (not shown) lies in the spread.  Any individual ensemble member will lie anywhere within the spread, but the unperturbed Ensemble Control Forecast is not used in the calculation.  Ensemble Control Forecast usually remains within the spread (eg plumes on the right), but on a few occasions (theoretically around 4% at longer lead-times) they extend outside of the plume (eg plumes on the left).  When this happens it is imperative to study the evolution of the atmosphere in some detail to decide the most likely evolution - and definitely not to assume that the unperturbed Ensemble Control Forecast results are necessarily the best.  In the case shown, Ensemble Control Forecast moved a small cold and potentially snowy vortex northwards across Southampton while the perturbed ensemble solutions moved it past the city either to the east or west.  The subsequent ensemble forecast showed much less uncertainty in temperature and depth of the vortex and it is questionable if the ensemble developed it at all.  Even so, because of the evidence from the earlier forecast, it would be unwise to dismiss the vortex altogether as it cannot be assumed the later ensemble forecast is necessarily the best even though it has benefitted from later observations.

 

Fig8.1.6-5: An example of uncertainty.  850hPa temperature plumes for Reading, England from ensemble runs of 00UTC 31 Aug, 12UTC 1 Sep, 00UTC 3 Sep 2018. The case is in association with uncertain forecasting of a cut-off low moving southeast across Britain.  There is large, even multi-modal, spread in the plume on the forecast based on 00UTC 31 Jan. The spread is much more narrow by the forecast based on 00UTC 3 Sep.

Cut-offs can be difficult for models to forecast. Consequently, users should consider the consistency of the model evolution. The blue lines are for cross-referencing; they denote the same dates and values on each panel.



(FUG associated with Cy50r1)