The mean is the average of the forecast values of the ensemble members at a given forecast time (i.e. the sum of the values divided by the number of ensemble members). The mean leans towards the values of a greater number of ensemble members and less weight is given to outliers. It is mostly used with medium range forecasts where the mean leans towards the most probable value.
The median is the middle value of the forecast values of the ensemble members at a given forecast time when sorted into a list (i.e. the same amount of values below and above). The median lies at the centre of the range of the ensemble members and can be more descriptive of the data set than the mean. It is mostly used with seasonal forecasts where the range of values can be quite large.
The ensemble spread is a measure of the difference between the members and is represented by the standard deviation (Std) with respect to the ensemble mean (EM). On average, small spread theoretically indicates high forecast accuracy, large spread theoretically indicates low forecast accuracy. The ensemble spread is flow-dependent and varies for different parameters. It usually increases with the forecast range, but there can be cases when the spread is larger at shorter forecast lead times than at longer. This might happen when the first days are characterized by strong synoptic systems with complex structures but are followed by large-scale fair weather high-pressure systems. The ensemble spread should reflect the diversity of all possible outcomes, in particular when the deterministic forecasts are “jumpy”, which might indicate that very different weather developments are possible. Two similar-looking forecast charts may display large differences in geopotential if they contain systems with strong gradients that are slightly out of phase. Conversely, two synoptically rather different forecast charts will display small differences if the gradients are weak. The spread refers to the uncertainty of the values of mean sea level pressure, geopotential height, wind or temperature, but not necessarily to the flow patterns. Such aspects are reflected in charts of ensemble spread.
The ensemble mean (or, on occasion, the ensemble median) forecast tends to average out the less predictable atmospheric scales. As the forecast proceeds the variation between the results of ENS members gradually increases. The ensemble mean, of course, will lie within the envelope of ensemble members throughout the forecast.
Clearly the error in a forecast increases through the forecast period and it is useful to have an idea of the likely magnitude and how it varies with forecast lead-time. The accuracy of the ensemble mean (EM) can be estimated by the spread of the ensemble; on average, the larger the spread, the larger the expected EM error. Assuming a gaussian distribution of ensemble results then the EM should also give an indication of the variability. An analysis of the relationship between root mean square error of the EM against lead-time shows a strong similarity with a measure of the spread of the ensemble members against lead-time. Thus the greater the spread, the greater the likely error. On average the spread increases with lead-time, but if less than normally seen at a given lead-time then the error is likely to be less than normally expected.
Fig8.1.2.2: The graph shows the error, on average, in 850hPa temperature for the Northern Hemisphere extratropics at various forecast-lead times. The relationship, on average, between ENS root mean square error (full line) and ENS spread (dashed line), shows a strong correlation. A low (or high) spread in the forecast implies low (or high) error on average (though at the same time any individual EM forecast may by chance be good or bad).
Special composite charts have been created to allow comparisons between ensemble mean (EM) and HRES. These charts normally show great consistency from one forecast to the next and can help forecasters judge how far into the future the ENS can carry informative value for large synoptic patterns. EM forecast values may be displayed (e.g. on ecCharts) together with the spread of the ensemble forecast values (Fig8.1.2.3). The coloured areas do not indicate the probability of the location of a feature, but merely indicate the magnitude of the uncertainty. Users should refer to HRES forecasts, Postage Stamp charts (example chart), Spaghetti Plot charts, or Clustering (example chart) to assess probability of departure from the EM before making forecast decisions.
Fig8.1.2.3: 500hPa ENS mean geopotential height (in red, 580dam isopleth crosses east Italy and north Greece) and spread of geopotential height among ensemble members (coloured according to the scale). Forecast for 12UTC 13 August 2017 T+120 from ENS data time 12UTC 8 Aug 2017. The greatest spread shows the areas of greatest uncertainty. The light green area indicates a spread of 4-5dam and this could be:
Fig 8.1.2.4: An example of forecast mean sea level pressure (taken from part of an ECMWF mean and spread chart) highlighting the difference between the HRES (Green) and the ensemble mean (EM, black). Absolute spread of ensemble members is shown by shading. The ensemble mean is the average over all ensemble members. It smooths the flow more in areas of large uncertainty (spread), something that cannot be achieved with a simple filtering of single forecast. If there is large spread, the ensemble mean can be a rather weak pattern and may not represent any of the possible states. The EM should always be used together with the spread to capture this uncertainty. Note in particular the small depressions forecast by the HRES near 35W (shown by arrows) and the additional uncertainy within the ENS nearby suggesting at least some of the ENS members show something similar to HRES although with timing and/or location differences.
The ensemble spread tends to show a strong geographical dependence. For geopotential and pressure there is generally little spread at low latitudes, but variability is greater at mid-latitudes and the spread is consequently rather higher. This latitude dependence tends to obscure the features of a given situation and a normalised spread or standard deviation (Nstd) is more useful. For this, the spread, measured by the standard deviation (Std) of ensemble member values at a given point and lead time, is normalised against the mean of the spread of the 30 most recent 00UTC ENS (Mstd) for 00UTC runs (or 12UTC ENS (Mstd) for 30 most recent 12UTC ENS runs) for the same lead-times and geographical locations.
The Normalised Spread is defined as: Nstd = Std/Mstd
where, for a given forecast lead-time and location:
Nstd is the Normalised Standard Deviation.
Std is the Standard Deviation of the latest ENS.
Mstd is the Mean Standard Deviation of the spread of the 30 most recent 00UTC or 12UTC ENS runs.
The Normalised Standard Deviation highlights geographical areas of unusually high or low spread, where the uncertainty is larger or smaller than over the last 30 days. If the spread in a particular area remains similar to previous spreads in that area then Nstd has a value near 1, irrespective of whether the spread is large or small. If it has greater spread that recently then Nstd >1, if it has less spread than recently then Nstd <1. The normalised spread shows the increase or decrease in spread at a location, not the magnitude of the spread, and therefore highlights relatively low or relatively high uncertainty, not the uncertainty itself.
ECMWF produces Mean and Spread charts and Normalised Standard Deviation charts for each ensemble run to aid understanding of the uncertainty of the forecast and whether the forecast is more or less uncertain in a given area at a given lead-time.
Fig8.1.2.5(Right): HRES PMSL (hPa) in blue and spread of the ensemble members (represented by their Standard Deviation, purple shading). Colour scale for spread in hPa shown above the chart.
Fig8.1.2.5(Left): Ensemble mean PMSL (hPa) in blue with Normalised Standard Deviation (coloured shading). Normalised Standard Deviation is calculated by dividing the Standard Deviation (right hand frame above) by a Mean Standard Deviation, a pre-computed mean of the standard deviations of the 30 most recent 00UTC (or 12UTC) ensemble forecastsfor the given lead time (this is also a function of location). Colour scale for Normalised Standard Deviation in hPa shown above the chart - uncoloured indicates a similarity with previous ENS mean values.
The panel on the right in Fig8.1.2.5 gives an assessment of the reliability of the absolute values of the contoured ensemble mean or HRES forecast fields. Relatively large/small absolute values of standard deviation tend to indicate relatively high/low uncertainty in forecasts. No colouring or the paler purples imply high confidence, brighter purples/magentas imply low confidence.
The panel on the left in Fig8.1.2.5, the normalised standard deviation, aims to put the standard deviation measure into the context of the general ensemble behaviour within the chart area over the last 30 days. It tells whether the most recent ensemble is showing greater or less spread (and hence uncertainty) than recent ensemble results. If the spread at Day5 of a particular set of ensemble forecasts (right panel) seems to be large, but has of late also tended to be equally large at Day5 in the same area, then the left panel shading will denote a value that is close to 1 (uncoloured). If the spread in a particular area at Day5 in the ensemble is greater/less than the spread that had recently been seen there at Day5, then the shading of the normalised standard deviation (left panel) indicates a value rather greater/less than 1 (purple/green shading). So although the forecast for (say) longer lead-times in the ENS (say days 8-10) will usually be of rather low confidence, there will be some occasions when one can be rather more confident than usual for this lead-time. The normalised standard deviation will tend to show this by green shading.
Comparing the run-to-run changes in Mean and Spread charts and the Normalised Standard Deviation charts can be informative and aid an assessment of confidence in the forecast.
Fig8.1.2.6: Mean and Spread charts data time 00UTC 8 September 2017, for T+120 verifying at 00UTC 13 September 2017.
Fig8.1.2.6 Right: HRES PMSL (hPa) in blue and spread of the ensemble members (represented by their Standard Deviation, purple shading). Colour scale for spread in hPa shown above the chart.
Fig8.1.2.6 Left: Ensemble mean PMSL (hPa) in blue with Normalised Standard Deviation (coloured shading, see Fig8.1.2.5). Normalised Standard Deviation is a function of lead time and of geographical location. Colour scale for Normalised Standard Deviation in hPa shown above the chart.
Fig8.1.2.7: Mean and Spread charts data time 00UTC 10 September 2017, for T+72 verifying at 00UTC 13 September 2017.
Fig8.1.2.7 Right: HRES PMSL (hPa) in blue and spread of the ensemble members (represented by their Standard Deviation, purple shading). Colour scale for spread in hPa shown above the chart.
Fig8.1.2.7 Left: Ensemble mean PMSL (hPa) in blue with Normalised Standard Deviation (coloured shading, see Fig8.1.2.5). Normalised Standard Deviation is a function of lead time and of geographical location. Colour scale for Normalised Standard Deviation in hPa shown above the chart.
Consider the charts for T+120 (Fig8.1.2.6) and T+72 (Fig8.1.2.7), both verifying at 00UTC 13 September 2017.
Over Scotland and northern England at T+120 (Fig8.1.2.6):
Over Scotland and northern England at T+72 (Fig8.1.2.7):
Elsewhere, comparing the charts for T+120 (Fig8.1.2.6) and for T+72 (Fig8.1.2.7):
General points: