Regime Frequency Histograms

These show how the probabilities of each of the four regimes (NAO+, NAO-, Block, Atlantic Ridge (ATR), plus No Regime) evolve on a daily basis as the ensemble members progress through the extended range forecast period.  It gives an indication of the likelihood of a change of regime, and the potential timings of such a change.  Comparison with the cumulative histograms from recent extended range forecasts can also give insights into the consistency of the forecast extended range evolution and probabilities, but flip-flops and trends should be viewed with caution.  The histograms enable the user to assess rapidly the confidence to be given to forecasts issued to the consumer.

One of the four regimes is assigned only if:

  • the minimum distance between the forecast anomalies and any of the four regimes lies within an "average value", and
  • that distance is significantly different from the distance to other regime types.

If either or both criteria are not satisfied "No Regime" is assigned. 

Regimes and Regime Frequency Histograms for the Extended Range are derived using 4-regime scheme.


Real time example of Extended Range Weather Regime Frequency histogram.

Fig8.2.5-1: The plot shows the cumulative weather regime frequency from the extended range ensembles (DT 14 March 2019).  The coloured bars indicate the number of members forecasting a specific regime for each day of the forecast (Blue:NAO+, Red:Blocking, Green:NAO-, Purple:ATR).  Inclusion of an ensemble member within a regime type gives no indication of the strength of the solution within that type (e.g. a strong or weak Blocking regime).  Members that are not close enough to any of the four weather regimes are represented by the grey bars (members with no regime attribution).  Bars with diagonal lines represent the regime attribution of ensemble mean fields.



Fig8.2.5-2: An example of a somewhat complicated change of regime (from NAO+ (blue) on 24 March to Blocking (red) on 10 April) together with other inferences that may be made from a retrospective forecast cumulative weather regime frequency diagram (DT 21 March 2019).  The indicated verification (blocks with black borders low down) largely confirms the greatest forecast probabilities.  However an observed  interlude of NAO- (green) from 3 to 5 April was not really captured by the ensemble probabilities and the forecast change to Blocking (red) was somewhat premature.  The probabilities of each regime type tend to become less diverse with forecast lead time, which corresponds to increasing uncertainty further into the future.



Regime Projection or Phase Space diagram

These show the distribution of the forecast Euro-Atlantic flow type presented diagrammatically in NAO/BL phase space in two-dimensional probability density function format.   Three forecast periods are shown (days 5-11, 12-18, 19-25).  The x-axis is for NAO+/-, and the y-axis for BLO+/-.  An objective assessment of each ensemble member forecast, during each seven day period, supplies a measure of both the NAO and BL components for each member (i.e. 7 days x 51 ensemble members = 357 assessments in total).  The mean assessment for each member during the 7-day period can be plotted as an empirical orthogonal function (EOF) in NAO/BL space, with colouring subsequently applied to denote frequencies within the ensemble in this phase space.  A sequence of such diagrams for several lead times from a set of ensemble forecasts gives an indication of the changing probabilities of each forecast type.  See Fig8.2.5-3.

The regime Projection Diagrams are derived using Mirror 2-regime scheme.

Real time example of Extended Range Two-dimensional Regime diagram.  


 

Fig8.2.5-3: NAO/BL phase space diagrams for the periods: Days5-11, Days12-18 and Days19-25, from extended range ensembles with DT 14 March 2019 (same as in Fig8.2.5-1).  Colours represent the proportion of members that have a similar mean solution during each seven day period.  The shaded area gives an indication of the spread of regime types.  Colours represent different proportions (taken as probabilities) of a combination of regime types (Note the colours represent different probabilities on each diagram).  Within the central circle there is only a weak indication of regime type.

  • During Days5-11, the regime is mostly of NAO+ type with some element of BL+ type.  There is relatively little variation among ensemble members (relatively compact distribution with maximum values (dark blue) of 40%-55% of ensemble members).  However, there is an alternative of a BL- type with some smaller element of NAO+.  
  • By Days12-18, the regime type is more BL+, but rather weak (probabilities are rather low with the maximum value (dark purple) of 16%-27% almost within the central circle (where there is no strong preference for a particular regime) and with a large variation among ensemble members (large spread out area).
  • By Days19-25, the regime is unclear with a large variation among ensemble members with many within the central circle implying no clear regime type apart from a weak indication of BL+ type.

In Fig8.2.5-3 the extended range forecasts show a distinct shift in regime away from NAO+ regime type at the beginning of the forecast towards BL+ by the second 7-day period.   By the third 7-day period the distribution is more uncertain with all regime types represented (see histogram in Fig8.2.5-2) although the majority lie within the central circle on the phase space diagrams (i.e. no strong preference for a particular regime).  Note however a subsidiary area of higher probability slightly in the BL+ sector suggesting possible persistence of that regime type albeit with low probability.


See also the section regarding Medium Range Regime Charts which gives further information on regimes within the Extended Range.

Read more about the use of "Weather regimes" at ECMWF (2020 article)