Important sources of predictability for the extended range forecasts are:

A major goal of Extended Range forecasting is prediction, well in advance, of persistent, anomalous large scale patterns that themselves can lead to severe weather events. For example strong and persistent large-scale high pressure systems are often associated with dry spells, and with heat waves in summer and cold spells in winter, and strong persistent quasi-zonal flow can be associated with very wet periods. Extended range, sub-seasonal forecasts cannot of course be expected to accurately represent day-to-day weather variations, but they should be able to capture large-scale circulation patterns that typically last longer than about a week, and also be able to roughly indicate the timing of a change from one circulation type to another.

It should be noted that biases have been reduced in 47R1.  In particular the lower stratosphere cold bias evident in 46R1 has a marked improvement in 47R1 (especially in summer).

Circulation patterns in the Euro-Atlantic Region

The large-scale circulation patterns that impact upon the European area can be categorised into four main classes:

Whilst the above is a very helpful conceptual framework, that is backed up by statistical categorisation techniques, it should also be stressed that:

  1. The amplitude of a given pattern will vary from case to case, and from day to day (e.g. coining the terms "strong NAO-" or "weak NAO-")
  2. Sometimes the atmospheric state lies between these regimes

This is discussed further below (see e.g. Fig.

Useful forecasts of the onset or cessation of anomalous, and even extreme surface weather over Europe may be made if the change from one circulation pattern to another, and the timing of that change, can be predicted well in advance of the event.  Additionally there needs to be an associated assessment of the confidence in any such transitions and their timing.  Errors in forecasting transitions from one type to another can lead to very large forecast errors in absolute (e.g. RMS error) terms. This is especially true in respect of the onset of blocking. 


Fig8.2.5.1: Geographical patterns of Euro-Atlantic climatological regimes (both anomalies and full fields) as used at ECMWF. North Atlantic Oscillation, positive phase (NAO+), North Atlantic Oscillation, negative phase (NAO­–), Blocking (BLO+), The Atlantic Ridge (AR) circulation pattern is a particular case of Anti-blocking (BLO–), inwhich the important feature is the trough over Scandinavia.  Geopotential anomalies (colour shading) and geopotential (contours) at 500 hPa are shown.

Fig: Anomalies in 2m temperature associated with the persistence (periods longer than 5 days) of the (a) positive North Atlantic Oscillation, NAO+, (b) negative North Atlantic Oscillation, NAO-, (c) Scandinavian Blocking, BL+, and (d) Atlantic Ridge (AR) (or Anti Blocking BLO-) regimes.

NAO-BLO diagrams

If NAO and BLO circulation systems are considered as orthogonal, a NAO-BLO phase space diagram may be used to investigate and illustrate the relationship between circulation type and other forecast or observed parameters.  The NAO–BLO phase space can offer the advantage of a simplified framework for assessing model performance in predicting temperature extremes.


Fig8.2.5.3: NAO–BLO phase space.  Well-defined circulation patterns lie towards the periphery of the diagram.  The central circle encloses a region where the circulation system is weak or cannot be confidently identified.

NAO–BLO phase space can be used to illustrate the relationship between severe cold European spells as in Fig8.2.13.


Fig8.2.5.4: Severe cold European spells, detected using the 2m temperature reanalyses, represented in the NAO–BLO space.  Colours have no special meaning.  The arrangement of the NAO-BLO diagrams corresponds to the arrangement of areas of Europe as shown on the map.

The analysis of severe cold spells in Europe, plotted on NAO-BLO diagrams for different areas is shown in Fig8.2.5.4.  The diagrams show severe cold spells in:

Fig8.2.5.5: Predictability distribution on NAO-BLO diagram.  Ensemble variance colour coded as scale.

Ensemble variance (spread) is indicative of predictability.  Thus NAO– has relatively high predictability (probably because it tends to be more persistent than other regimes), BLO+ has relatively lower predictability.

The NAO–BLO space explains about 30% of the daily winter variability over Europe.

Transitions between circulation patterns

A study of a large number of reanalyses (36 years of ERA-interim data) gives an indication of the frequency of transitions from one circulation type to another (giving a “climatology” of transitions).  Another study using available extended range re-forecasts (12 years of re-forecasts) gives an indication of the ability of the forecasts to capture similar transitions that occurred during the six-day period preceding various selected forecast lead times (Day11, Day16, Day21, Day 31).  The results are shown in Fig8.2.5.6.

Fig8.2.5.6: Frequency (in percentages) of transitions to a given regime; stacked bar colour denotes the previous regime.  Colours show transitions fromBLO+ (pale red), BLO− (purple), NAO+ (blue), NAO− (green), no clear initial circulation pattern (grey).  Reanalysis values are shown in the column on the far left of each section.  The other bars indicate the forecast values at Day11, Day16, Day21 and Day31, respectively.  Where the frequency is larger than 5% its value is indicated on the bar.

The results show:

In general:

The model statistics and relative frequencies, at all forecast ranges, compare well with those from the analysis, indicating that the IFS is well able to simulate transitions, and suggesting that model bias in this context is not a major problem.

NAO-BLO diagrams may be used to illustrate the sequence of transitions from one circulation type to another (e.g. NAO+ to BLO+ to NAO–).

Fig8.2.5.7: NAO-BLO diagram showing transition of circulation pattern with time.  Colours indicate the elapse of time.  The initial NAO+ circulation pattern becomes a BLO+ circulation pattern by T+72hr, and finally becomes a NAO– circulation pattern by about T+168hr.


A measure of skill is the anomaly correlation between the observed and the ensemble mean forecasts of the principle circulation patterns - i.e. components associated with westerly/easterly flow across the Atlantic (NAO+/NAO–), blocked/anti-blocked flow over Scandinavia (BLO+/BLO­–), and the bivariate correlation using both of these.

Fig8.2.5.8: Regime-based skill measures for ensemble mean fields from various global forecast systems. There is skill where correlation is above 0.5.

Forecasts may be considered to have skill where the anomaly correlation is above 0.5.  Extended range forecasts show predictive skill for:

The longer period of skill for NAO modes may be associated with the NAO modes being more persistent (notably NAO-), and the fact that models are correctly capturing that persistence.

Fig8.2.5.9: Continuous Ranked Probability Skill Score (CRPSS) for the four Euro-Atlantic Regimes for several forecast models. 

In Fig8.2.5.9, ECMWF (black) shows some skill for NAO-/NAO+ up to 20-23 days while for BLO+ (blocking) and AR (Atlantic Ridge) skill drops to zero at about 16-17 days.  In other words, the ECMWF extended range forecasts have more difficulty predicting episodes of "Blocking" and "Atlantic Ridge" than they do predicting episodes of NAO+ and NAO-.  Note that the plot is based on about 10 years of re-forecast data from all the models shown.  Every ENS forecast is represented on every panel - i.e. the plot does not just relate to questions such as "when NAO- was forecast did it happen?".

In general, the skill of extended range forecasts:

The skill in predicting heat waves or cold spells in the extended range may be limited by the ability of the forecast model to represent transitions to anticyclonic circulation regimes (BLO+, NAO-) over Europe.  However, once an NAO- circulation pattern has formed there is a tendency for it to persist in reality and in the IFS.

Uncertainty and Predictability

The ensemble variance or spread is an indicator of forecast uncertainty and normally increases with forecast lead time.  The rate at which the spread grows during the forecast can be used as an estimate of predictability.  Fig8.2.5.10 shows the change in spread with elapsed time.  Beyond day 3, forecasts with the ensemble mean first entering the NAO− sector have a lower mean ensemble variance than those with the ensemble mean entering any other sectors.  The differences between the mean ensemble variances could be associated with the fact that, by entering into a circulation pattern (NAO–) associated with higher predictability, the forecast uncertainty increases at a slower rate. 

Fig8.2.5.10: The mean ensemble variance as a function of lead time for all forecasts with the ensemble mean entering the BLO+ (red), BLO− (purple), NAO+ (blue), NAO− (green) sectors of an NAO-BLO diagram (e.g. Fig NAO- shows better predictability (less ensemble spread) than other circulation patterns.

In general:

The reliability of forecasts of cold conditions over certain parts of Europe (notably the north) is:


Circulation patterns like NAO are often associated with global teleconnections through propagation of Rossby wave trains.  El Nino-Southern Oscillation (ENSO) events, Sudden Stratospheric Warmings (SSW) and pronounced Madden-Julian Oscillation (MJO) events have been found to enhance the predictability and skill of forecast circulations in the North Atlantic/European area.

In particular, the presence of and the phases of significant MJO events can be linked to an increase of skill in forecasting NAO– circulation patterns  Fig compares the bivariate correlation (NAO & BLO) against analysis for forecasts initiated with and without an MJO event.  Forecasts initiated with an MJO event show higher skill (an improvement of the order of one day) between Day8 and Day15.  Correlations are significantly increased for Day11 and Day12 at a 90% confidence level, and for Day10 to Day13 at an 80% confidence level.


Fig8.2.5.11(left): Bivariate correlation (NAO & BLO) for forecasts with (red) and without (black) an MJO event in the initial conditions.  Colours refer to initial conditions for the forecasts:

Fig8.2.5.11(right): Brier Skill Scores (BSS) for NAO– predictions according to the initial MJO phase.  Colours refer to initial conditions for the forecasts:

The MJO influences skill in forecasts concerning NAO– circulation pattern:

The corresponding increases in skill for NAO+, BLO+ and BLO– are small.

Fig8.2.5.12: Variation of the MJO index bivariate correlation with forecast lead-time for two older ECMWF model cycles (40r1 and 40r3).  Beyond Day20 correlation falls below 0.7. By Day 27 it falls below 0.6, implying marginal skill.

Important Caveats for Forecasters

Forecasters must also be aware that high confidence in forecasts of a circulation pattern may not equate to high confidence in forecasts of the associated surface weather, in certain locations. For example, an exceptionally confident winter-time forecast of NAO- can be accompanied, over southern England, by exceptionally large spread in forecasts of 2m temperature. This is due mainly to the locations of strong atmospheric thermal gradients undergoing regime-related shifts away from their climatological positions. And in other scenarios/locations other synoptic factors will come into play.

MJO Teleconnection Example

The forecast based on 00UTC 25 Feb 2019 illustrates the tele-connection effect of tropical deep convection over the Indian Ocean upon subsequent downstream developments of NAO+ type over the North Atlantic/Europe. 

Fig8.2.5.13: Meteosat (IODC) IR (Channel 4) image DT 00UTC 25 Feb 19.  A large area of convection lies over an equatorial region of the Indian Ocean typical of Madden-Julian Oscillations (MJO

Fig8.2.5.14: MJO Wheeler-Hendon diagram for the monthly forecast based on DT 00UTC 25 February 2019.   The colours represent ENS forecasts at various lead times as given by the key above the diagram.  Initially the MJO lies within Sector1 (Western Indian Ocean) and is forecast to progress into Sector2 (Eastern Indian Ocean) by Day5 with a fairly limited spread among ENS members. The mean position of the ENS forecasts during the subsequent ten days show continuing westward progress although with an increasing spread before weakening (moving into the central circle) by Day20.

Fig8.2.5.15: Day15-21 forecast Mean Surface Level Pressure Anomaly verifying Day15-21 (11-17 Mar 19).  In the North Atlantic, the forecast suggests the pressure anomaly is likely to be 10-20hPa below ER-climate across areas in the north and 5-10hPa above ER-climate further south implying an anomalously strong westerly flow towards Europe (i.e. a NAO+ regime).

Fig8.2.5.16: Day15-21 forecast 2m Temperature Anomaly verifying Day15-21 (11-17 Mar 19).  The forecast suggests an anomalously mild period over central and western Europe where the 2m temperature is forecast to be 0°C-3°C above ER-climate.

Fig8.2.5.17: Day15-21 forecast Precipitation Anomaly verifying Day15-21 (11-17 Mar 19).  The forecast suggests an anomalously wet spell over western Europe where precipitation is forecast to be 0-10mm above ER-climate over western Europe, and 10-30mm over Ireland and the United Kingdom.

The pressure pattern with low pressure over the Iceland area and high pressure over the Azores are is representative of NAO+ type.  The forecast temperatures anomalies correspond to the NAO+ type temperature anomalies shown in Fig8.2.5.1 above. 

Additional Sources of Information

(Note: In older material there may be references to issues that have subsequently been addressed)

Updated/Amended 09/07/20 - Added comment on lower biases in 47R1. 

Updated/Amended 24/10/20 - amended links for open access.