

Monthly forecast products 


At present, there are two types of monthly forecast product plots available on the ECMWF web site. The first shows monthly forecasts of surface atmospheric variables such as rainfall, temperature and surface pressure. In these cases, maps of ensemble mean anomalies, probability and tercile plots are made available, detailed descriptions of which are given below. These maps are created in the same way as in seasonal forecasting. The second type is an extension of EPS products such as plumes, stamp maps and weather regime clusterization to the 32day monthly forecast. a) Ensemble mean anomaly maps: These products are similar to seasonal forecasting charts, but with weekly means instead of monthly means. Over each point of the map, atmospheric variables such as 2metre temperature, total precipitation, mean sealevel pressure or surface temperature, have been averaged over a weekly period ( week 1: day 5 to 11, week2: day 12 to 18, week 3: day 19 to 25, and week 4: day 26 to 32) and also over the 51 members of the realtime forecast and the 60 members of the back statistics. The plots display the difference between the ensemble mean of the realtime forecast and the ensemble mean of the backstatistics. The product therefore displays the shift of the forecast ensemble mean from the estimated "climatological" mean (created from ensemble runs over the past 12 years). In addition, a WilcoxonMannWhitney test (WMWtest, see for instance Wonacott and Wonacott 1977) has been applied to estimate whether the ensemble distribution of the realtime forecast is significantly different from the ensemble distribution of the backstatistics. Regions where the WMWtest displays a significance less than 90% are blank. Regions where the WMWtest displays a significance exceeding 95% are delimited by a solid contour (blue or red depending on whether the anomaly is positive or negative respectively). The blanking of "nonsignificant" shifts does not mean that there is no signal in the blanked regions, but only that, with the particular sampling we have, we cannot be sure that there is a signal. For this reason, there are likely to be many areas where a signal is real but remains undetected. This product is currently limited to surface temperature, 2metre temperature, total precipitation and mean sea level pressure. The probability maps display the probability that the predicted anomaly is greater than zero. As for the ensemble mean anomaly maps, the anomalies are calculated by subtracting the climatological mean from the real time ensemble forecast. The maps are also based on weekly means. Only regions where the WMWtest displays a significance larger than 90% are shaded. Regions where the WMWtest displays a significance exceeding 95% are delimited by a solid contour (blue or red depending on whether the anomaly is positive or negative respectively). The regions where the significance is less than 90% are blanked. From the 60member ensemble of the back statistics, three equally probable domains are defined: below normal, normal and above normal. The probability of the backstatistics (model climatology) in each domain is 33%. The upper tercile map displays the probability of an anomaly above normal and the lower tercile displays the probability of an anomaly below normal. It is calculated from the number of ensemble members in the real time forecasts which display an anomaly within the abovenormal domain defined from the model climatology. If there is no signal (the ensemble distribution of the real time forecast is not significantly different from the ensemble distribution of the model climatology), then the tercile map will indicate a probability close to 33%. Therefore, the probability range 2040% has been blanked, in order to indicate regions where the model does not predict a significant shift of ensemble distribution from model climatology. This product is an extension of the EPS plumes to 32days. It displays the time evolution of the ensemble forecast of geopotential at 500 hPa, 12hour accumulated precipitation and temperature at 850 hPa over several European cities. Each perturbed member of the ensemble is represented by a magenta line, whereas the control forecast is indicated by a continuous cyan line. The dotted cyan line represents the control forecast of the EPS with the same starting date and time as the monthly forecast. The black contours represent the maximum and minimum values of the ensemble forecast. The interval defined by these two lines therefore represents the ensemble spread. Since the ensemble distribution may look noisy after about 10 days of forecast (most especially for precipitation), the time series are also displayed in the form of quartiles. Instead of showing the time evolution of each ensemble member, the quartiles display the time evolution of the ensemble median (red), 25%75% domain (cyan), and extremes (dark blue area). One of the 3 plots displays the time evolution of the precipitation accumulated since the start of the forecasts, instead of 12hour accumulated precipitation. The plumes are not biascorrected. In other words, they have been created from the raw data of the realtime forecast.
This product (also called stamp maps) is also an extension of an EPS product. It contains 51 stamps. Each stamp represents one ensemble member forecast of geopotential at 500 hPa or mean sealevel pressure. For mean sealevel pressure stamps, the 8 and 16 degree isolines of temperature at 850 hPa are also displayed as cyan and red lines respectively. The stamp maps display the instantaneous fields at days 5, 10, 15, 20, 25 and 30. Weekly means (defined as week1: day 5:11, week2: day 12:18, week3: day 19:25 and week4 day 26:32) are also displayed. As for the plumes, this product is not biascorrected. This product is based on a method of regime clusterization used in EPS (see ECMWF Technical Memorandum 317). Six predefined patterns for the Atlantic/Europe region of geopotential at 500 hPa have been determined using several years of ERA15 data. The patterns represent different equiprobable weather regimes. Each member of the ensemble realtime monthly forecast of geopotential at 500 hPa is assigned to the closest pattern for different time steps (5, 10, 15, 20, 25 and 30 days). The products display the patterns computed by averaging all the ensemble members that fall within the same cluster, along with the number of ensemble members that fall within each cluster.This product has also been extended to weekly means (week1: day 5 to 11, week2: day 12 to 18, week3: day 19 to 25 and week4: day 26 to 32). The variability in the population of each cluster from one week to the next is also displayed here. This product is not biascorrected. The Hovmoeller diagram displays the time evolution of the ensemble mean anomaly of geopotential height at 500 hPa, averaged over the latitude band 35N55N. The anomaly has been computed by averaging all the members of the realtime forecast and subtracting the ensemble mean of the backstatistics (model climatology). Therefore, the graphics display anomalies with respect to the climate of the past 12 years.The xaxis represents the longitude, and the yaxis represents the time evolution (from top to bottom). Since it is an ensemble mean, the structures are much more detailed in the first days of the forecast (top part of the graphics) than in the last days (bottom part). Contours are plotted every metre. Shaded areas represent the ensemble spread and are displayed only when the amplitude of the anomaly exceeds 2 metres. The first vertical axis on the right side indicates the mean value of the ensemble spread, averaged over the complete area. The value of the spread increases sharply during the first days of the forecast. The second vertical axis on the right side of the figure represents the value of the mean zonal flow. The mean zonal flow represents the average, over all the longitudes, of the difference of geopotential at 500 hpa between latitudes 55N and 35N. The value of the zonal mean flow gives an idea of the meridional variability of the 500 hPa geopotential patterns.
References: Wonacott, T.H. and R.J. Wonacott, 1977: Introductory statistics.John Wiley, 650 pp. 



