Eastern Europe and Western Russia experienced a strong heat wave during the summer of 2010. Maximum temperatures exceeded 40°C in early August, resulting in over 15,000 deaths and many wildfires, inflicting large economic losses on Russia. The heat wave resulted from strong atmospheric blocking that persisted over the Euro-Russian region from late June to early August. This study investigates the predictabilities of extreme Euro-Russian blocking and of the blocking-induced extreme surface temperatures in the summer of 2010, using medium-range ensemble forecasts. The results show that the blocking in June–August (JJA) of 2010 was easily predictable, even for a lead time of +216 hr; however, the blocking that occurred from 30th July to 9th August showed a lower predictability in forecasts over +144 hr compared with other blocking occurrences in JJA of 2010. This low predictability resulted in the failure to predict the extreme temperatures associated with the mature blocking in early August. Most of the forecasts predicted a decay of the blocking earlier than that observed.
 Atmospheric blocking, a nonlinear phenomenon maintained by upscale enstrophy and cyclonic wave-breaking effects [Tyrlis and Hoskins, 2008; Shutts, 1983], is an important weather regime in mid-latitude weather and climate, as persistent blocking can induce extremely high or low temperatures and severe precipitation anomalies over the surrounding area [e.g., Black et al., 2004; Neiman et al., 2004]. In the summer of 2010, Euro-Russian blocking persisted for more than 1 month, producing an intense heat wave over Eastern Europe and Western Russia. Extremely high temperatures were recorded in many cities: Moscow (55.5°N, 37.4°E, Russia) recorded its highest temperature of 39°C on 30th July during the prior 90 years, Joensuu (62.4°N, 29.4°E, Finland) recorded 37.2°C on 30th July (the previous highest temperature in Finland, 35°C, was recorded in July 1914), Gomel (52.2°N, 30.6°E, Belarus) recorded 38.9°C on 7th August, and Jaskul (46.1°N, 45.2°E, Russia) recorded 42.2°C on 8th August (surface synoptic observations (SYNOP), available at http://www.ogimet.com/gsynres.phtml.en, and State of the Climate Global Hazards, available at http://www.ncdc.noaa.gov/sotc/). Also, the downstream trough of the Euro-Russian blocking induced catastrophic flooding over Pakistan in July 2010 [Webster et al., 2011].
 The heat wave was more intense and covered a wider area of the Euro-Russian region than the heat wave over Europe in the summer of 2003 [Stott et al., 2004; Schär and Jendritzky, 2004]. In 2010, the heat wave over Russia resulted in more than 15,000 deaths, including more than 1,600 drowning deaths as people entered the water in an attempt to escape the heat, and caused severe economic losses due to damage to crops such as wheat. More than 600 wildfires resulted in smog levels that were five to eight times higher than normal, leading to widespread illness [Gilbert, 2010] (State of the Climate Global Hazards).
 Ensemble forecasts have recently become a major component of operational numerical weather prediction (NWP) systems and are useful in investigating the predictability of high-impact weather and in estimating in advance the potential occurrence of high-impact weather. It is well known that general circulation models tend to underestimate blocking frequency in NWPs and in climate projections [Palmer et al., 2008; Pelly and Hoskins, 2003; D'Andrea et al., 1998]. Matsueda  revealed that model performance in simulating blocking is greatly improved in state-of-the-art medium-range NWP models. It would be valuable to investigate the predictabilities of the extreme Euro-Russian blocking and of the blocking-induced extreme surface temperatures in the summer of 2010, using operational medium-range ensemble forecasts: CMC (Canadian Meteorological Center), ECMWF (European Centre for Medium-range Weather Forecasts), JMA (Japan Meteorological Agency), NCEP (National Centers for Environmental Prediction), and UKMO (United Kingdom Meteorological Office).
2.1. TIGGE Data, Reanalysis Data, and Observation Data
 Table S1 of the auxiliary material summarizes the operational medium-range ensemble forecast data used in this study. The data was obtained from the TIGGE (THe Observing system Research and Predictability EXperiment (THORPEX) [World Meteorological Organization, 2005] Interactive Grand Global Ensemble [Bougeault et al., 2010]) data portal. The WMO began the THORPEX project in 2005 to accelerate improvements in the accuracy of 1-day to 2-week forecasts of high-impact weather for the benefit of society, the economy, and the environment. The TIGGE portal quasi-operationally provides 10 global ensemble forecast datasets. The use of TIGGE is a very effective way to rapidly respond to significantly high-impact weather. Ensemble forecast data from CMC, ECMWF, JMA, NCEP, and UKMO were used for the following reasons. Matsueda and Tanaka  reported that ECMWF has the best performance over the Northern Hemisphere in the ensemble forecast and that CMC, JMA, NCEP, and UKMO have the second-best performance.
 The various ensemble prediction systems differ in terms of the horizontal resolution of the numerical model, method of initial perturbation, forecast length, and ensemble size (Table S1). Note that the forecast length is different for each centre: the minimum length of 216 hr is for JMA and the maximum length of 384 hr is for CMC and NCEP. This study considers forecasts of 500-hPa geopotential height (Z500) and temperature at 2 m (T2m), initialised at 1200 UTC for the period from 20th May 2010 to 31st August 2010, valid on the period from 1st June 2010 to 31st August 2010. The data were interpolated to a common grid spacing of 2.5°. For Z500 and T2m, the analysis (observation) for each NWP centre was defined as the control run at the initial time.
 To calculate the daily climatologies of Z500 and T2m, and the frequency of climatological blocking, Japanese reanalysis (JRA) data for the period 1979–2003 were analysed [Onogi et al., 2007].
 Also used was the daily maximum temperature (Tmax) from E-OBS (version3) [Haylock et al., 2008]. The daily climatology for Tmax was calculated using E-OBS for the period 1979–2003. The E-OBS data cover the area of 25°–75°N, 40°–75°E.
2.2. Blocking Index
 This study employs the objective blocking index proposed by D'Andrea et al. . The 500 hPa Geopotential Height meridional Gradients, GHGS (South) and GHGN (North), are computed for each latitude as follows:
A specific longitude on a given day is locally defined as being blocked (“local blocking”) if both of the following conditions are satisfied (for at least one value of Δ):
This study focuses on the Atlantic–European–Russian region (60°W–90°E), which is prone to blocking in summer.
2.3. Ensemble-Based Occurrence Probability of Blocking
 Here, the ensemble-based occurrence probability of blocking is defined following Matsueda . The occurrence probability is defined at each longitude and is measured by the ratio of members that predict the occurrence of local blocking to the total number of members. For example, a probability of one (zero) for a specific longitude indicates that all (no) members predict an occurrence of local blocking for a specific longitude.
Figure 1 shows the climatological blocking frequency for June–August (JJA) and the observed blocking frequency in JJA of 2010. The climatological seasonal-mean blocking frequency shows the well-known blocking maxima over the Atlantic–European–Russian region (Figure 1c). The most pronounced peak, at around 30°E, corresponds to European blocking. The other peaks, at around 10°W and 60°E, correspond to Atlantic and Ural blockings, respectively. European blocking is observed throughout JJA, with a maximum in June, while Atlantic and Ural blockings are observed mainly in June and July, respectively (Figure 1a).
 In JJA of 2010, blocking was continuously observed between 20°E and 70°E from late June to early August (Figure 1b). From late June to mid-July, the European blocking was located at around 30°E; subsequently, it moved eastward in late July and remained at around 50°E, as Ural blocking, until the beginning of mid-August. The seasonal-mean European and Ural blocking frequencies for JJA of 2010 were twice as high as the climatological frequencies, whereas the observed Atlantic blocking frequency was comparable with the climatological frequency (Figure 1d). In particular, the observed frequency at around 50°E was about six times as high as the climatological frequency.
 The time evolution of blocking is also apparent in the Z500 anomaly field (Figures 2a–2i; note that E-OBS does not cover all of the area shown in Figure 2). The long-lived blocking resulted in extreme temperatures over Eastern Europe and Western Russia in JJA of 2010. The period from the 21st to 31st July saw the rapid development of the upstream trough of blocking, with a negative Tmax anomaly over Italy (Figure 2f). This was followed by a strong Z500 anomaly from 1st to 10th August (Figure 2g). Figure 2j shows the areal-mean Tmax (45°–60°N, 25°–50°E) for an area that includes Moscow, Gomel, and Jaskul. The positive Tmax anomaly was maintained, although varying somewhat, from 19th June to 19th August, changing in tandem with the strength of continuous blocking (Figure 1b). During this period, the maximum anomaly of the observed areal-mean Tmax was 11.3°C (temperature of 36.9°C) on 7th August. After 16th August, the Tmax anomaly rapidly dropped to below zero with the disappearance of blocking.
Figures 3b–3e shows the occurrence probability of blocking, as estimated by ensemble members for the +072 hr, +144 hr, +216 hr, and +288 hr lead times in JJA of 2010, respectively (see Figure S1 for a specific forecast, valid at 12UTC on 7th August). For a lead time of +072 hr (Figures 3b and S1a), the observed blocking was well predicted by all the models, generally with a high probability above 0.75. A high probability (>0.50) was seen for the +144 hr and +216 hr forecasts (Figures 3c and 3d). In general, the blocking in JJA of 2010 was easily predictable even for a lead time of +216 hr, as in the case of wintertime blocking described by Matsueda . In particular, a high probability for the +216 hr forecast by ECMWF in late July suggests a high predictability of the Pakistan flood, as Webster et al.  showed. However, the western part of the blocking (30°–45°E; herein termed “western flank”) that occurred from 30th July to 9th August showed a lower predictability for lead times greater than +144 hr than did other blocking occurrences in JJA of 2010. This indicated a difficulty in simulating the upstream trough of blocking (Figures S1b and S1c). In the +144 hr forecast, all the centres started to show a low probability (<0.25) for the western flank, while all the centres partly showed a high probability (>0.75) for the eastern part of the blocking (45°–70°E; herein termed “eastern flank”) that occurred from 30th July to 9th August.
 For a lead time of +216 hr, all the centres showed a low probability (<0.25) for the western flank. In terms of the eastern flank, ECMWF and UKMO predicted the persistence of blocking with moderate probability (0.25–0.75), whereas CMC, JMA, and NCEP predicted a discontinuous evolution of blocking. In the +288 hr forecast, as easily expected from Figure S1d, both the eastern and western flanks showed a lower predictability than did other blocking occurrences in JJA of 2010. The lower predictability for the western flank is also apparent in the areal-mean (30°–45°E) occurrence probability (Figure 3a). The predicted occurrence probability for lead times greater than +144 hr was lower in early August than in other periods from late June to late July.
 The poor forecast performances in simulating the western flank generally led to poor performance in simulating T2m. Figure 4 shows the observed and predicted time series of the T2m anomaly (Tano) averaged over the region 45°–60°N, 25°–50°E. The predicted time series are for ensemble mean forecasts of the Tano, initialised from 1st June to 31st August. The increase in the Tano during early August, with a maximum of around 10°C on 7th August, was difficult to predict for almost all the ensemble forecasts initialised from 21st to 31st July, even for lead times of +000 hr to +216 hr (blue line). Most of the ensemble mean forecasts and individual ensemble members (Figure S2) predicted a Tano similar to the initial one or a decrease in the Tano in early August, as described in the following paragraph. The ECMWF high-resolution deterministic forecast (TL1279L91) also failed to predict the increase in the Tano during early August, due to a failure of the blocking prediction (Figure S3). For forecasts initialised during other periods, the +000 hr to +216 hr forecasts generally performed well in predicting the Tano, whereas the +216 hr to +384 hr forecasts sometimes underestimated the Tano.
 For individual ensemble members initialised on 26th July (Figure S2d; 288 hr before 7th August), most members predicted a Tano similar to the initial one or a decrease in the Tano during early August, mainly due to the predicted disappearance of blocking earlier than that observed (Figure S1d). Although some JMA members appeared to predict an increase in the Tano, the JMA analysis of T2m tends to yield lower values than those estimated by the other centres. The JMA members still seem to underestimate the increase in the Tano. For the members initialised on 29th July (Figure S2c; 216 hr before 7th August), the performance in simulating the Tano on 7th August was improved for all centres, due to an improvement in blocking forecast (Figures 3d and S1c), although the increase in the Tano remained underestimated. Many of the members initialised on 1st August (Figure S2b; 144 hr before 7th August) performed well in predicting the maximum Tano on 7th August, due to a further improvement of blocking forecast (Figures 3c and S1b). Almost all the members, however, predicted a rapid decrease in the Tano in mid-August, earlier than that observed. For the members initialised on 4th August (Figure S2a; 72 hr before 7th August), all the centres performed well in simulating the maximum Tano and the rapid decrease in the anomaly, except for the +216 hr to +384 hr forecasts of the NCEP members.
 Eastern Europe and Western Russia experienced an intense heat wave in the summer of 2010, caused by strong atmospheric blocking over the Euro-Russian region from late June to early August. The observed blocking frequency was twice or more as high as the climatological frequency. This study investigated the predictabilities of extreme Euro-Russian blocking and of the blocking-induced extreme surface temperatures in the summer of 2010, using five operational medium-range ensemble forecasts: CMC, ECMWF, JMA, NCEP, and UKMO.
 The blocking that occurred during JJA of 2010 was generally easily predictable, even for a lead time of +216 hr; however, the blocking that occurred over the Euro-Russian region from 30th July to 9th August, especially the western part of the blocking (30°–45°E), showed a lower predictability for lead times greater than +144 hr, compared with other blocking occurrences during JJA of 2010. This indicated a difficulty in simulating the upstream trough of blocking. This low predictability resulted in the failure to predict the extreme temperatures associated with the mature blocking in early August. Most of the forecasts predicted a decay of the blocking earlier than that observed, even though the blocking was already present in the initial condition. Furthermore, the erroneously early decay of predicted blocking led to an early (compared with observations) decrease in the T2m anomaly during mid-August. This finding might suggest a difficulty in simulating the maintenance and decay of blocking, unlike previous studies [e.g., Matsueda, 2009] suggesting that the main difficulty is to simulate the transition from a zonal to a blocked flow, rather than the persistence. Although Matsueda et al.  reported that high horizontal resolution is required for successful numerical modelling of long-lived blocking, model resolution does not seem to be an important factor in simulating the extreme Euro-Russian blocking in summer 2010.
 This work was conducted under the framework of the “Projection of the Change in Future Weather Extremes using Super-High-Resolution Atmospheric Models” supported by the KAKUSHIN Program of the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan. I am grateful to the two anonymous reviewers whose comments led to substantial improvements in the manuscript.
 The Editor would like to thank the two anonymous reviewers for their assistance in evaluating this paper.