The 2010 summer heat wave in western Russia was extraordinary, with the region experiencing the warmest July since at least 1880 and numerous locations setting all-time maximum temperature records. This study explores whether early warning could have been provided through knowledge of natural and human-caused climate forcings. Model simulations and observational data are used to determine the impact of observed sea surface temperatures (SSTs), sea ice conditions and greenhouse gas concentrations. Analysis of forced model simulations indicates that neither human influences nor other slowly evolving ocean boundary conditions contributed substantially to the magnitude of this heat wave. They also provide evidence that such an intense event could be produced through natural variability alone. Analysis of observations indicate that this heat wave was mainly due to internal atmospheric dynamical processes that produced and maintained a strong and long-lived blocking event, and that similar atmospheric patterns have occurred with prior heat waves in this region. We conclude that the intense 2010 Russian heat wave was mainly due to natural internal atmospheric variability. Slowly varying boundary conditions that could have provided predictability and the potential for early warning did not appear to play an appreciable role in this event.
 Questions of vital societal interest are whether the 2010 Russian heat wave might have been anticipated, and to what extent human-caused greenhouse gas emissions played a role. Exceptional heat and poor air quality due to wildfires led to large increases in deaths in Moscow and elsewhere in western Russia, despite international efforts to improve public health responses to heat waves [World Health Organization, 2009]. Russia's extreme heat commenced in July nearly coincident with the peak temperatures in the annual cycle, thereby exacerbating human and environmental impacts. During July, when daily temperatures (Figure 1, top) were consistently near or above record levels, the heat wave spanned western Russia, the Republic of Belarus, the Ukraine, and the Baltic nations (see Figure S1 in Text S1 of the auxiliary material). Despite record warm globally-averaged surface temperatures over the first six months of 2010 [National Climatic Data Center, 2010], Moscow experienced an unusually cold winter and a relatively mild but variable spring, providing no hint of the record heat yet to come (Figure 1, top).
 For the 2003 western European heat wave, human influences are estimated to have at least doubled the risk for such an extreme event [Stott et al., 2004]. Other boundary forcings also contributed to the 2003 European heat wave, including anomalous sea surface temperatures (SSTs) [Feudale and Shukla, 2010]. The goal of this study is to identify the primary causes of the Russian heat wave and to assess to what extent it might have been anticipated from prior knowledge of natural and human forcings and observed regional climate trends.
2. Data and Model Experiments
 The National Oceanic and Atmospheric Administration (NOAA) Land/Sea Merged analyses [Smith and Reynolds, 2005] are the primary surface temperature data used in this study. Results derived from this data set are compared with those obtained from three other observational temperature data sets (see Table S1 and references for these data sets in the auxiliary material). In the following analyses, western Russia temperatures are defined as area-averages over the region 50oN–60oN and 35oE to 55oE, the region of highest heat wave intensity and approximately centered over Moscow.
 Model simulations were performed to determine the potential for anticipating the Russian heat wave. First, the potential influence of increasing greenhouse gas concentrations, aerosols, and other natural external forcings on western Russian temperatures was assessed from simulations of 22 CMIP3 models [Meehl et al., 2007]. These models are forced by specified monthly variations in greenhouse gases and tropospheric sulphate aerosols for 1880–1999, and with the IPCC Special Emissions Scenario (SRES) A1B thereafter. About half of the models also include changes in solar radiance and the effects of volcanic eruptions for the period 1880–1999. Model time series of western Russia temperatures were normalized relative to the observed mean standard deviation for July from 1880 to 2009 so that the magnitude of interannual variability in all models was comparable with observed variability. Second, possible effects of specific boundary conditions observed during July 2010 were evaluated. For this purpose, 50-member ensemble simulations were performed for each of two atmospheric general circulation models, the GFDL AM2.1 [Delworth et al., 2006] and the middle atmosphere configuration of ECHAM5 (MAECHAM5) [Roeckner et al., 2003], using observed global SST, sea ice and atmospheric carbon dioxide concentrations for July 2010. Responses to 2010 forcings were determined through comparisons with two parallel 50-member control simulations that used 1971–2000 mean climatological forcings. Third, predictions generated in June 2010 with NOAA's climate forecast system model [Saha et al., 2006] were examined to assess the potential role of atmospheric, land, and ocean initial conditions in this event. These predictions were initialized with atmospheric, land, and ocean conditions in early (1–4) and late (27–30) June 2010.
 The July surface temperatures for the region impacted by the 2010 Russian heat wave shows no significant warming trend over the prior 130-year period from 1880 to 2009 (Figures 1, middle and 1, bottom). A linear trend calculation yields a total temperature change over the 130 years of −0.1oC (with a range of 0 to −0.4oC over the four data sets, see Tables S1 and S2 of the auxiliary material for comparison). Similarly, no significant difference exists between July temperatures over western Russia averaged for the last 65 years (1945–2009) versus the prior 65 years (1880–1944) (Table S2). There is also no clear indication of a trend toward increasing warm extremes. The prior 10 warmest Julys are distributed across the entire period and exhibit only modest clustering earlier in this decade, in the 1980s and in the 1930s (Figure 1, middle). This behavior differs substantially from globally averaged annual temperatures, for which eleven of the last twelve years ending in 2006 rank among the twelve warmest years in the instrumental record since 1850 [Intergovernmental Panel on Climate Change, 2007]. The absence of prior July warming also differs from antecedent conditions for the 2003 western European heat wave, where a strong regional warming trend was detected over the twentieth century (see long-term trend map in Figure 1, bottom), a significant fraction of which has been attributed to anthropogenic forcing [Fischer and Schär, 2010].
 With no significant long-term trend in western Russia July surface temperatures detected over the period 1880–2009, mean regional temperature changes are thus very unlikely to have contributed substantially to the magnitude of the 2010 Russian heat wave. Another possibility is that long-term trends in variability may have increased the likelihood of an extreme heat wave. To assess this possibility, standard deviations of July surface temperatures were calculated for the two 65-yr periods before and after 1945. The results (Table S2) indicate slightly higher variability in the later period, but this increase is not statistically significant based on a standard F-test. Western Russia temperature extremes simulated in the 22 CMIP3 models (grey shaded area in Figure 1, middle) also do not display discernible trends during 1880–2009. The temporal distribution of extreme heat waves in the model data normalized to correspond with observed variability shows two events of similar magnitude to the heat wave intensity of about +5°C departure observed during 1880–2009, with one event in the earlier half of the 20th Century (light gray shading in Figure 1, middle). For model runs that are not normalized, the frequency of >5oC extreme events occurring before 1945 is even greater and comparable in frequency to that seen in more recent decades (dark gray shading in Figure 1, middle). In summary, the analysis of the observed 1880–2009 time series shows that no statistically significant long-term change is detected in either the mean or variability of western Russia July temperatures, implying that for this region an anthropogenic climate change signal has yet to emerge above the natural background variability. This is in contrast to regions such as western Europe, but similar to other regions like the central United States, consistent with strong regional (and seasonal) differences in climate trends that are yet to be fully understood.
 The nature of this heat wave and its origins were intimately tied to the upper-level atmospheric flow. The 500 hPa July flow (Figure 2, top) was characterized by a classic “omega” blocking pattern [Dole and Gordon, 1983]. The highest July 2010 surface temperature anomalies (Figure 2, middle top) occurred near the center of the block, where northward displaced subtropical air, descending air motions and reduced cloudiness all contributed to abnormally warm surface temperatures. Severe drought occurred with the Russian heat wave, making it likely that land surface feedbacks amplified this heat wave's intensity, as has been observed in prior severe droughts [Atlas et al., 1993; Fischer et al., 2007]. To the east of the heat wave region, anomalously cool temperatures occurred in conjunction with an upper level trough and southward transport of polar air.
 Russia is climatologically disposed toward blocking events during summer [Tyrlis and Hoskins, 2007], and many of its prior July heat waves were associated with blocks. Consistent with this, a composite analysis of the average temperature anomalies and 500 hPa heights associated with the ten largest prior heat waves in this region since 1880 shows patterns similar to 2010 (cf. Figures 2, top and 2, middle top and Figures 2, middle bottom and 2, bottom), although features are weaker as expected from such a composite analysis. The distance between centers of the temperature anomalies is comparable to the scale for stationary upper-air Rossby waves [Held, 1983], consistent with the role of atmospheric dynamical processes in accounting for the persistence of this pattern.
 We have diagnosed additional model simulations forced by observed boundary conditions for this period to assess whether those may have produced a forced response consistent with the blocking pattern and associated heat wave. These boundary conditions reflect a mixture of both natural and human influences on the climate system. The observed global SSTs include positive anomalies in the Indo-west Pacific Ocean and tropical Atlantic and developing La Niña conditions in the east Pacific (see Figure S1). The observed Arctic sea ice extent in July 2010 was the second lowest in the satellite record [National Snow and Ice Data Center, 2010]. Figure 3 shows the model response based on the AM2.1 model. The ensemble-mean responses of the atmospheric circulation (Figure 3, top) and surface temperatures (Figure 3, middle top) are far weaker and their patterns are inconsistent with the observed blocking and heat wave (cf. Figure 2). A similar conclusion is drawn from the MAECHAM5 simulation whose response to July 2010 forcing is also very weak (Figure S2 of the auxiliary material). These findings suggest that the blocking and heat wave were not primarily a forced response to specific boundary conditions during 2010.
 Nor are there indications that blocking would increase in response to increasing greenhouse gases. Results using very high-resolution climate models suggest that the number of Euro-Atlantic blocking events will decrease by the latter half of the 21st century [Matsueda et al., 2009; M. Matsueda and T. N. Palmer, personal communication, 2010]. The horizontal resolution of climate models is an important consideration in simulating blocking accurately. Although the ensemble-mean AM2.1 and MAECHAM5 responses bear no resemblance to the observed event, both models are capable of producing blocking over this area. For example, individual members within each model ensemble show flow patterns (Figures 3, middle bottom and S2, middle bottom) and temperature anomalies (Figures 3, bottom and S2, bottom) that are qualitatively similar to observations. However, these patterns reflect internal atmospheric variability within the models rather than a systematic response to boundary forcing, and thus are not evidence of a predictable signal. With only 50 ensemble members in these simulations, a meaningful assessment of changes in the tails of the distributions is not possible.
 A third suite of model runs has also been considered which differs from the prior sets in that it is initialized with observed ocean-atmosphere-land conditions of 2010 in NOAA's operational coupled Climate Forecast System (CFS). Comparing predictions of July blocking in models initialized in early June versus in late June further clarifies the roles of boundary forcing and initial conditions and also addresses the potential for early warning capabilities. When initialized in early June 2010, the predictions show no evidence for a change in the probability of prolonged daily blocking during July 2010 over western Russia compared to the July hindcasts that were initialized in each June during 1981–2008. The model predictions do, however, show approximately a doubling of the average duration of daily blocking during July for runs begun in late June 2010, by which time blocking was already present in atmospheric initial conditions (see Figure S3 of the auxiliary material). This increase coincides with a shift of the probability density function of western Russian temperature anomalies towards warmer values by about +1.5oC. These results are consistent with the interpretation that the Russian heat wave was primarily caused by internal atmospheric dynamical processes rather than observed ocean or sea ice states or greenhouse gas concentrations.
4. Concluding Remarks
 Our analysis points to a primarily natural cause for the Russian heat wave. This event appears to be mainly due to internal atmospheric dynamical processes that produced and maintained an intense and long-lived blocking event. Results from prior studies suggest that it is likely that the intensity of the heat wave was further increased by regional land surface feedbacks. The absence of long-term trends in regional mean temperatures and variability together with the model results indicate that it is very unlikely that warming attributable to increasing greenhouse gas concentrations contributed substantially to the magnitude of this heat wave. Nevertheless, there is evidence that such warming has contributed to observed heat waves in other regions, and is very likely to produce more frequent and extreme heat waves later this century [Intergovernmental Panel on Climate Change, 2007]. To assess this possibility for the region of western Russia, we have used the same IPCC model simulations to estimate the probability of exceeding various July temperature thresholds over the period 1880–2100 (Figure 4). The results suggest that we may be on the cusp of a period in which the probability of such events increases rapidly, due primarily to the influence of projected increases in greenhouse gas concentrations. Uncertainty in timing is nonetheless evident (Figure 4, inset), due in part to different model sensitivities to greenhouse gas forcing. Understanding the physical processes producing heat waves will be important for improving regional projections, and may also provide an improved capability for predicting some extreme events. However, as in the case of the 2010 Russian heat wave, events will also occur that are not readily anticipated from knowledge of either prior climate trends or specific climate forcings, and for which advance warning may thus be limited.
 The study is supported in part by the NOAA Climate Program Office. The CFS forecast data were provided by Wanqui Wang of NOAA's Climate Prediction Center. The comments provided by William Neff and Klaus Wolter are gratefully acknowledged.
 The authors thank two anonymous reviewers for their thoughtful comments.