Fig8.1.5.1A: To view meteograms:
Fig8.1.5.1B: Alternative way to view meteograms:
View directly the meteogram site.
The ensemble meteogram provides a probabilistic interpretation of the ENS for specific locations. It displays the time evolution of the distribution of several meteorological parameters from the ensemble by a box and whisker plot. All ENS meteograms have a title section, giving the name (unless overwritten by the user), the true height of the chosen location, and the co-ordinates of the grid point used based on the HRES or ENS resolution.
The sub-section “Selection of grid points for Meteograms” explains the method of interpolation of grid point forecast data for presentation for a given location.
Forecast distributions are displayed using a box and whisker plot (see Fig8.1.5.2) which shows the median (short horizontal line), the 25th and 75th percentiles (wide vertical box), 10th and 90th percentiles (narrower boxes) and the minimum and maximum values (vertical lines).
Fig8.1.5.2: The box and whisker plot used in the ECMWF 10- and 15-day ensemble meteograms.
Ensemble meteograms are available for:
Note:
Fig8.1.5.3: 10-day medium-range meteogram for Dublin from HRES and ENS data time 00UTC 12 May 2017. The HRES and ENS CTRL are included in the 10-day ensemble meteogram for reference (Blue lines are HRES, Red lines are ensemble Control). The red numbers above the precipitation panel are the greatest precipitation value reached by any ENS member. ENS extreme values cannot be ignored as the evolution of every ENS member is considered to be equi-probable. In this example an unstable regime moved over the area during 13th and 14th and heavy showers were likely although a passage precisely over Dublin was of course uncertain. Note: the forecast temperatures are at 00UTC, 06UTC, 12UTC, 18UTC each day (15-day meteograms show forecast maximum and minimum temperatures for each day). UTC is used exclusively in the meteograms and maxima or minima will occur according to the longitude (or local time) of the location in question.
Fig8.1.5.4A: 15-day medium-range meteogram for Dublin from ENS data time 00UTC 12 May 2017. The displayed values are for the 24hr period each day, with additionally the distribution of 10m wind direction. Note: the forecast maximum and minimum temperatures are shown for each day (10-day meteograms show forecast temperatures at 00UTC, 06UTC, 12UTC, 18UTC).
Fig8.1.5.4B: As Fig8.1.5.4A with the addition of M-climate data. M-climate data is shown by colours with percentiles similar to the box and whisker scheme. The median wind forecast for 15 May lies well above the M-climate values (above the 75th percentile of the M-climate) with the whisker extending above the 99th percentile of the M-climate.
An alternative display (currently only available within the ecCharts meteogram platform) has a circle divided clockwise into eight arcs, each arc representing 1/8 cloud cover. So, for example, the arc covering 45°-90° represents 2/8 cloud cover. The shading within each arc is proportional to the number of members that forecast this particular degree of cloud cover or more.
Total precipitation in the 10-day ensemble meteogram is the accumulated precipitation (sum of convective and large-scale) over 6hr periods (00-06UTC, 06-12UTC, etc). In the 15-day meteogram it is the accumulated precipitation over 24hr periods (00-24UTC). Note:
On the 10-day meteogram the box-and-whisker plot locations align with the end of the 6 hour period.
Probabilities for intervals longer than the 6hr and 24hr time intervals cannot be deduced from the ensemble meteogram.
Periods of probabilities >0% in every interval can be wrongly interpreted as uninterrupted rain.
Consideration of the median alone can be wrongly interpreted as protracted dry spells.
The HRES is generally better able to generate realistic precipitation amounts and capture extreme rainfall than the ENS because of its higher resolution.
The precipitation shown on the ensemble meteograms cannot be directly intercompared as the rainfall range (y-axis) varies from one location to the next and from one forecast to the next. The rainfall range is chosen separately for each ensemble meteogram so that 100% of the predicted values are covered for the 15-day ensemble meteograms, and at least 90% of the predicted values are covered for the 10-day ensemble meteograms (if the top of the distribution is beyond the scale maximum the largest 6-hourly total is shown at the top as red numbers).
The HRES meteorological fields are interpolated using the four HRES grid points nearest to the location of the selected ENS grid point.
Because HRES values are interpolated and ENS values are not, the tendency for HRES to sometimes deliver higher wind speeds or larger rainfall totals than ENS, as discussed above, will to an extent be mitigated against on meteogram displays because of the interpolation. In turn, the degree to which this happens will depend on how near the closest HRES grid point is, spatially, to the chosen ENS grid point.
At longer lead times, the ensemble mean and the ensemble median will tend to gravitate asymptotically towards the M-climate. This is most clearly seen when the first ten days of the forecast are anomalous (e.g. after an initial spell of cold and rainy weather, the ensemble tends to indicate a return to milder and drier conditions at longer forecast ranges). This follows logically from the fact that at an infinite range, when predictive skill is completely lost, a climatological value constitutes the optimal forecast.
It is necessary to assess critically the parameters shown on meteograms. Verification of previous forecasts, particularly recent forecasts within a similar meteorological regime, may allow an insight into whether the latest HRES or ENS is likely to be the better forecast.
Occasionally ENS produces forecasts that diverge into a bi-modal (or possibly multi-modal) distribution of forecast results in two (or more) distinct patterns (e.g. if there is model uncertainty regarding positioning of a cold front, a number of ENS members for a given location may show warm midday temperatures while others show much cooler temperatures). Bi-modal distribution of forecast results will not be shown by meteograms (as box-and whisker plots cannot do this - the effect would be just to stretch out the boxes). However bimodal distributions can be apparent on plume diagrams.
If a majority of ENS members forecast temperatures below 0°C and, at the same time, a large number of members forecast substantial precipitation, there is no way to determine the likelihood of snowfall from the standard meteogram diagram alone. It could be that the precipitating members might all have temperatures well above 0°C. Users are encouraged to use the ecCharts meteogram product, which shows ENS probabilities of precipitation types by category. Probability of combined events can only be calculated from the original ENS data. Several charts of combined probabilities are available on ecCharts.
The relative forecast spread may vary considerably between one parameter and another in the same forecast step. For example
The ensemble can only predict severe weather events of the kind that the resolution of the ENS can resolve. The HRES has a small advantage over the ENS with respect to rainfall rate or wind speed. The latest HRES should be considered together with the ENS as part of the ensemble. If the HRES deviates systematically from the ENS, forecasters have to use their experience or local knowledge to decide which information is the more realistic or representative and, if necessary, adjust one to the other. In such circumstances, it may be appropriate to give more weight to HRES.
Fig8.1.5.5: Ensemble Meteogram for Kontiolahti in eastern Finland, 12UTC 22 April 2011. The systematic difference between the HRES (blue line) and the ensemble Control forecast (red line) is around 5°C. Kontiolahti lies close to a small lake better captured in HRES than in the ENS.
When creating a meteogram for a specific location, the land-sea mask at the four surrounding ENS grid points is considered. If there is at least one land grid point within these four, then the nearest land point will be chosen and the meteogram title section shows "ENS Land Point" together with its location and ENS altitude. If only sea points are available then the nearest sea grid point will be chosen and the meteogram title section shows "ENS Sea Point" together with its location and ENS altitude of 0m. Data at the selected ENS point will have been calculated using HTESSEL and FLake according to the proportions of land and sea cover within the surrounding grid point box (see examples below, or the Land-Sea Mask section for details).
HRES values (meteorological fields and orography) are interpolated onto the ENS grid from the four HRES grid points surrounding the location of the selected ENS grid point and are added to the meteogram. Data at the selected HRES grid points surrounding the selected ENS grid point will have been calculated using HTESSEL and FLake according to the Land-Sea Mask defining the land and sea cover within their surrounding grid point boxes. Grid points for meteograms are selected in a complex manner.
Therefore systematic differences between HRES and ENS can occur in connection with strong gradients along coasts, small islands or in mountainous regions. Any such discrepancy is usually most clearly apparent during the first few days, when the spread is normally small. Some influences of the adjacent sea areas may be over- or under-represented by the ENS and/or HRES meteograms. Users should consider the impact of the grid point(s) relative to the land-sea mask upon the indication of the forecast parameter on the meteogram (temperature, wind, etc). Users should critically assess differences in meteograms for coastal, island or mountainous regions.
Note: the so-called land-sea mask processing (where the land or sea nature of the source and target points was used to adjust the interpolation weights) used by the old ECMWF interpolation software scheme (called EMOSLIB) is not used by default in the new MIR interpolation package that was introduced early in 2019.