INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 30: 89–109 (2010) Published online 23 February 2009 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/joc.1874 Recent severe heat waves in central Europe: how to view them in a long-term prospect? Jan Kyselý* Institute of Atmospheric Physics AS CR, Bočnı́ II 1401, 141 31 Prague, Czech Republic ABSTRACT: The study examines whether recent occurrences of severe heat waves in central Europe were exceptional in the context of past fluctuations, and estimates their recurrence probabilities under several climate change scenarios. Using data from a network of meteorological stations in the Czech Republic since 1961, it is found that 1994 was the year with the most severe heat waves over majority of the area. The other seasons with enhanced heat wave characteristics were 1992, 2003 and 2006. Analysis of the long-term temperature series at Prague-Klementinum reveals that the July 2006 heat wave, covering 33 consecutive days, was the longest and most severe individual heat wave since 1775. Probabilities of long and severe heat waves are estimated from daily temperature series generated by a first-order autoregressive model with a deterministic component (incorporating a seasonal cycle and a long-term trend). The model is validated with respect to the simulation of heat waves in present climate (1961–2006) and subsequently run under several assumptions reflecting various rates of summer warming over 2007–2100. The return period of a heat wave reaching or exceeding the length of the 2006 heat wave in Prague is estimated to be around 120 years in 2006. Owing to an increase in mean summer temperatures, probabilities of very long heat waves have already risen by an order of magnitude over the recent 25 years, and are likely to increase by another order of magnitude by around 2040 under the summer warming rate assumed by the mid-scenario. Even the lower bound scenario yields a considerable decline of return periods associated with intense heat waves. Nevertheless, the most severe recent heat waves appear to be typical rather of a late 21st century than a mid-21st century climate. Copyright 2009 Royal Meteorological Society KEY WORDS heat wave; climate variability; regional warming; stochastic modelling; probability estimates; central Europe Received 29 August 2007; Revised 9 January 2009; Accepted 18 January 2009 1. Introduction The area of western and central Europe has recently been affected by several long-lasting and severe heat waves, particularly in July–August 2003 and June–July 2006. A number of studies examined the 2003 heat waves as to their meteorological causes (e.g. Black et al., 2004; Fink et al., 2004; Fischer et al., 2007), climatological importance (including possible global warming relationships; Beniston, 2004; Schär et al., 2004; Schär and Jendritzky, 2004; Stott et al., 2004; Trigo et al., 2005; Chase et al., 2006), environmental effects (Jolly et al., 2005; Ciais et al., 2005; Rebetez et al., 2006) and human mortality impacts (Vandentorren et al., 2004; Johnson et al., 2005; Grize et al., 2005; Filleul et al., 2006), focusing primarily on western Europe. With an estimated death toll exceeding 30 000 over Europe, the August 2003 heat wave was the worst natural disaster in Europe during the last 50 years (De Bono et al., 2004), yielding a pervasive example of how seriously also high-income countries like France may be affected by adverse effects of climate change. * Correspondence to: Jan Kyselý, Institute of Atmospheric Physics AS CR, Bočnı́ II 1401, 141 31 Prague, Czech Republic. E-mail: [email protected] Copyright 2009 Royal Meteorological Society Less attention has been devoted to the manifestation of the 2003 heat waves over central-European countries (Hutter et al., 2007), partly as the areas of largest positive temperature anomalies in summer 2003 were centered over France and Switzerland (Beniston, 2004; Zaitchik et al., 2006). However, the 2003 heat waves received much attention from public as well as from mass media also in central Europe, and they were followed by another exceptional heat wave in July of 2006, the warmest month on record in several European countries, including Germany, Belgium, the Netherlands and the UK (EC, 2007), and the warmest July in Prague since the beginning of instrumental measurements (1775). One reason for the enhanced interest in the warm temperature extremes stems from a concern that frequency and severity of such events have increased recently, and that this trend is likely to continue due to global warming (Meehl and Tebaldi, 2004). The issues of climatological perspective and future likelihoods are timely also owing to the fact that the heat waves affected various sectors of human activities, with enormous socio-economic impacts. It is well established that impacts of extreme events are heavier when extreme conditions prevail over extended time periods (Rusticucci and Vargas, 2002; IPCC, 2007a). Heat–stress related mortality is a frequently examined example of harmful 90 J. KYSELÝ effects of heat waves, but there are many other fields in which such events influence (mostly negatively) human society; for example, the largest power failure in central Europe since the 1970s appeared during the 2006 heat waves, having severely affected industrial companies. A number of economic impacts of the hot summer of 1995 in the UK were summarized by Agnew and Palutikof (1999). The present study aims at supplementing current understanding of the recent heat waves in central Europe from the points of view of their long-term variability and possible future recurrence likelihoods. The primary issues that we address are (1) whether the recent heat waves were exceptional, and if so, in what sense; and (2) what are recurrence probabilities associated with the most severe heat waves observed in the area in the current and future climates. For that purpose, heat waves are examined in data from a dense network of meteorological stations in the Czech Republic since 1961, to reveal possible regional details in their variability and to obtain a comparison of the recent events with past occurrences of heat waves (Section 3). In addition, centennial-scale fluctuations in the frequency and severity of heat waves are studied using two long-term series of daily air temperature measurements (Section 4). Probabilities associated with the most severe heat waves are examined by means of a stochastic time series model (Section 5) including simulations assuming a warming trend (with a wide range of possible rates estimated from an ensemble of regional climate model outputs) towards future. 2. Data and methods 2.1. Data from a network of meteorological stations Daily air temperature measurements (maximum, minimum and average daily temperature) are available at 46 stations over 1961–2006 (Figure 1; Table I). The sites approximately evenly cover the area of the Czech Republic, and there are no missing values in the dataset over the period of May to September (when heat waves occur). The only exception is May 2006 in the records of the Kašperské Hory station; the daily temperatures for this month have been interpolated from neighbouring locations. This procedure does not introduce inhomogeneity in heat wave characteristics as heat waves did not occur in May 2006 at neighbouring (lower elevated) stations. No significant station displacements (exceeding 50 m in altitude) occurred since 1961; the only exception was the Ústı́ nad Orlicı́ station for which the series of daily temperatures has been homogenized, taking into account a station move in January 1971, by applying additive correction factors over 1961–1970 (estimated on the monthly scale from data at five neighbouring stations). At two highest elevated stations, Lysá hora Mt. (1322 m a.s.l.) and Churáňov (1118 m a.s.l.), heat waves (according to the definition in Section 2.3) do not occur, so the effective number of sites for the analysis of heat waves is 44. 2.2. Long-term data at the Prague-Klementinum and Milešovka Mt. stations Two other long-term series of daily air temperatures are examined. The Prague-Klementinum station (197 m a.s.l.) is the site with the longest uninterrupted series of daily temperature records in the Czech Republic, dating back to 1775. Homogeneity of the temperature series has recently been evaluated by Štěpánek (2005) who found no significant inhomogeneities in the period after 1840. The history of measurements is described in more detail, e.g. in Hlaváč (1937) and Brázdil and Budı́ková (1999). The series has been frequently employed in studies of climatic variability (Horová et al., 2003; Kyselý, 2007) including the analysis of heat waves over 1901–1997 in Kyselý (2002a). Herein, we examine heat waves over the whole period 1775–2006 although results in the late 18th and early 19th centuries must be interpreted with caution, due to a possible presence of inhomogeneities in the daily records before 1840. The issue of an urban heat island intensification resulting from the development of the city (the station is located in the historical heavily built-up center of Prague) was addressed by Brázdil and Budı́ková (1999) who reported that in summer, the urban Figure 1. Station network used for the analysis of heat waves over 1961–2006. Two mountain stations at which heat waves did not occur are labelled in italics. Regions and main orographic features are delineated in the right panel. Copyright 2009 Royal Meteorological Society Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 91 RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE Table I. The longest and most severe heat waves and record-breaking daily TMAX at individual stations over 1961–2006. Region Station name Altitude (m a.s.l.) Longest heat wave Most severe heat wave Highest TMAX Year Duration (days) Year TS30 (° C) Day TMAX (° C) Northwest Doksany Žatec Liberec Milešovka 158 201 398 833 2006 2006 1994 2006 33 32 20 12 1994 1994 1994 2006 77.2 62.7 41.6 18.4 13.08.2003 13.08.2003 31.07.1994 20.07.2006 38.3 38.6 36.2 35.0 Southwest Klatovy Nepomuk 430 465 28 28 1994 2003 54.3 48.7 27.07.1983 27.07.1983 40.0 39.2 Kralovice Cheb Aš Přimda 468 471 675 742 1976 1976 2003 1994 1994 2003 2003 22 21 14 18 1994 2003 2003 2003 37.1 32.9 20.4 34.3 27.07.1983 27.07.1983 13.08.2003 13.08.2003 37.2 37.0 34.1 36.6 Brandýs n/Labem Praha-Klementinum Praha-Karlov Semčice 179 197 232 234 38.5 37.8 38.5 37.0 Hradec Králové Praha-Ruzyně Ondřejov Central lowland Central highland Velké Meziřı́čı́ Havlı́čkův Brod Moravské Budějovice Přibyslav Kostelnı́ Myslová Bystřice n/Pernštejnem Nedvězı́ Svratouch 22 33 22 23 1994 2006 1994 1994 73.1 49.1 55.6 62.8 278 364 526 1994 2006 1994 1994 2003 1994 1994 1994 23 17 17 1994 1994 1994 74.6 37.1 28.0 27.07.1983 27.07.1983 27.07.1983 31.07.1994 13.08.2003 30.07.1994 13.08.2003 27.07.1983 452 455 457 530 569 573 722 737 1994 1994 1994 1994 1994 1994 1994 1994 19 19 22 19 19 19 13 18 1994 1994 1994 2003 2003 2006 1994 1994 42.3 30.5 53.9 23.3 26.9 28.1 9.8 15.3 27.07.1983 27.07.1983 27.07.1983 27.07.1983 27.07.1983 27.07.1983 27.07.1983 27.07.1983 37.0 36.7 38.0 36.0 37.1 35.1 34.4 33.9 37.8 37.0 37.0 Northeast Ostrava-Mošnov Opava Lučina Valašské Meziřı́čı́ Vsetı́n Ústı́ n/Orlicı́ M.Albrechtice-Žáry Červená Lysá hora 251 272 300 334 388 399 483 750 1322 1994 1994 1994 1994 1994 1994 1994 1994 – 19 29 19 19 19 19 18 14 – 1994 1994 1994 1994 1994 1994 1994 1994 – 64.5 51.8 42.9 55.4 36.5 37.4 27.5 11.6 – 10.08.1992 29.08.1992 01.08.1994 29.08.1992 11.08.1992 27.07.1983 01.08.1994 30.07.1994 08.08.1992 30.07.1994 36.9 37.1 36.2 36.2 35.9 35.8 34.8 32.6 28.8 South České Budějovice Tábor Nadějkov Kašperské Hory Churáňov 388 440 615 737 1118 1994 1994 1994 2003 – 22 22 18 14 – 1994 1994 1994 2003 – 37.8 60.6 18.0 15.2 – 27.07.1983 27.07.1983 27.07.1983 27.07.1983 27.07.1983 37.8 37.3 36.6 36.8 34.2 Southeast Strážnice Holešov 176 224 34 23 1994 1994 67.1 63.3 13.08.2003 13.08.2003 37.7 36.5 Olomouc Staré Město Brno-Tuřany Kuchařovice Stránı́ 225 235 241 334 385 1994 1994 2003 1992 1994 1994 1992 1994 37 33 32 25 20 1994 1994 1994 2003 1994 61.9 76.8 67.2 58.2 40.8 13.08.2003 13.08.2003 13.08.2003 13.08.2003 28.08.1992 36.4 37.5 37.0 37.9 35.7 TS30 stands for cumulative TMAX excess above 30 ° C. In all regions, stations are ranked with respect to increasing altitude. If a station was displaced during 1961–2006, altitude for a longer subperiod is given. Copyright 2009 Royal Meteorological Society Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 92 J. KYSELÝ warming (estimated to make up 0.01 ° C/10 years) is by far the least pronounced and insignificant. Another long-term series of daily temperature measurements available is the one recorded at the Milešovka Mt. (833 m a.s.l.), a unique mountain station located on summit of an isolated cone-shaped hill (overtopping the surrounding terrain by more than 400 m) about 70 km northwest to Prague (Štekl and Podzimek, 1993). The series of daily extreme temperatures extends back to 1906; daily observations in the months of May to July over 1930–1935 have been lost (Huth and Nemešová, 1995), leaving a gap in the time series. Details concerning the history of measurements are given in Brázdil et al. (1999). Owing to the higher altitude, heat waves are rather rare events here; within the dense network described in Section 2.1, it is the highest elevated site at which heat waves occur. Nevertheless, the data are very useful as the station is not much affected by human activities, particularly urbanization effects, and are used mainly for comparison with the temperature variability in Prague over the 20th century. 2.3. Heat waves The definition of heat waves proposed by Huth et al. (2000) and employed in recent European studies (Hutter et al., 2007) as well as the global study of Meehl and Tebaldi (2004) is made use of. Two thresholds, T1 and T2 are applied: A heat wave is defined as a continuous period during which (1) TMAX (daily maximum air temperature) is higher than T1 in at least 3 days; (2) mean TMAX over the whole period is higher than T1 ; and (3) TMAX does not drop below T2 . The threshold values were set to T1 = 30.0 ° C and T2 = 25.0 ° C, in accordance with a climatological practice which refers to the days with TMAX reaching or exceeding 30.0 ° C and 25.0 ° C as tropical and summer days, respectively. (The average TMAX in July–August 1961–2006 ranges between 20.2 and 25.3 ° C at the 44 stations, and the standard deviation of daily TMAX in July–August is around 4.6 ° C; at all stations, T1 exceeds the average by at least one standard deviation.) The lower threshold T2 corresponds approximately to the threshold value of TMAX at which excess mortality appears in the population of the Czech Republic (Kyselý and Huth, 2004). The definition allows two periods of extremely hot days separated by a slight drop of temperature to make up one heat wave but, on the other hand, two periods of hot days separated by a pronounced temperature drop (below T2 ) are treated as separate heat waves. To characterize heat waves, the duration, the cumulative TMAX excess above 30.0 ° C (TS30), and the peak temperature are used (the same set of characteristics was applied in Kyselý, 2002a). The variables measure various aspects of heat waves; cumulative TMAX excess TS30 is probably the most appropriate characteristics of their severity. Mean temperature conditions of a summer season are expressed by the mean July–August TMAX Copyright 2009 Royal Meteorological Society anomaly from the reference 1961–1990 mean. The twomonth period of July–August instead of the more common three-month June–August period is employed since in July and August, majority of heat wave days occur in central Europe. 2.4. Regionalization To examine temporal fluctuations in heat waves using the network of 44 sites (and to avoid discussing possible site-specific features and suppressing important spatial variations), seven regions consisting of between four and eight stations each have been delineated. They reflect spatial differences in geographical characteristics, as well as mean summer temperature and precipitation (Kyselý, 2006). Note that the area under study is characterized by a complex orography, with altitudes ranging from below 200 m a.s.l. in large lowlands up to about 1500 m a.s.l. in four main mountain ranges in the southwest, northwest, north and northeast; most of the area lies between 200 and 600 m a.s.l. (Figure 1). The central lowland and southeast regions cover the main agricultural areas where heat waves (particularly if associated with droughts) may have largest negative effects on plant development and yields. Another important feature captured by the regionalization is that the southern part of the Czech Republic is usually more affected with heat waves than the northwest and especially northeast regions, the latter being the coldest area in summer (if effects related directly to generally decreasing TMAX with increasing elevation of sites are eliminated; Kyselý, 2006). 2.5. Stochastic time series modelling One way to estimate probabilities of extreme events is to employ stochastic time series modelling; this approach is particularly advantageous in studying spells of extremes for which simpler ‘block maxima’ and/or ‘peaks-overthreshold’ models (Coles, 2001) are not suited (if characteristics other than the frequency and magnitude of peak exceedances are of interest). A first-order autoregressive [AR(1)] model is applied here to generate long artificial series of TMAX (from which recurrence probabilities of heat waves are estimated) at two stations, located in the two main lowland regions: Prague-Klementinum in the central lowland region, and Brno-Tuřany in the southeast region. Rather generally, AR(1) models provide characteristics of heat waves and temperature threshold exceedances that are in a good agreement with observations in mid-latitude areas (Mearns et al., 1984; Macchiato et al., 1993; Hennessy and Pittock, 1995; Colombo et al., 1999; Kyselý, 2002b). According to results of a non-parametric test on the model order based on autoregression rank scores (Hallin et al., 1997; Kalvová et al., 2000), the AR(1) model is the most suitable one for temperature series in the area under study, and higher-order AR models are rejected relative to the first order process. There are two basic approaches to modelling temperature series when only a few months of year are examined; first, the seasonal cycle is considered a deterministic part Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 93 RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE and only deviations from this cycle are simulated by the AR(1) model (Macchiato et al., 1993), or the deterministic part is not taken into account at all and the whole series, one year stuck to another, is treated as if it were merely a realization of the AR(1) process (Mearns et al., 1984). Herein, the former, physically more reasonable method is adopted since the seasonal cycle may play an important role in both supporting a heat wave development and imposing some limitations on the length of a heat wave. Another important reason is that we decided to model the long-term trend as a deterministic part, too (see below). The AR(1) model is based on three parameters of TMAX series, the mean (µ(t)), the variance (σ 2 (t)) and the first-order autocorrelation coefficient ((t)); t denotes time since seasonal cycles in all three parameters (and a long-term trend in the mean) are considered explicitly. Values of TMAX are determined according to the recursion formula (Mearns et al., 1984; Macchiato et al., 1993) TMAX (t) = µ(t) + (t)(TMAX (t − 1) − µ(t − 1)) + ε(t) The initial value for the series is drawn from normal distribution N (µ(1), σ 2 (1)) for the first day considered in each year, and random variable ε(t) is then generated for each day from N (µ(t), σε 2 (t)) distribution where the variance of ε(t) is σε 2 (t) = (1 − 2 (t))σ 2 (t). Box–Muller random number generator (Press et al., 1992) is used to obtain independent random variable with normal distribution. An estimate of (t) is computed according to Kendall and Stuart (1976) t+L−1 u(i + 1)u(i) i=t−L (t) = t+L−1 t+L−1 2 u (i) u2 (i + 1) i=t−L i=t−L where u(i) stands for a standardized variable (TMAX (i) − µ(i))/σ (i) and L is the half-width of the moving window (set to 30 days here). Time series of TMAX are generated for months of May to September. Before the model parameters are estimated from the observed data (over 1961–2006), the time series of TMAX were detrended using an average regional linear trend estimator of mean July–August TMAX over 1961–2006 (Section 3.1; 0.5 ° C/decade at both stations); the trend is added up to the artificial time series as a deterministic part at a final stage. The seasonal cycle of µ(t) is smoothed using 15-day running means, and σ (t) and (t) are estimated for moving 61-day windows (covering 46 years of detrended observed data). Alternative settings have little influence on properties of the artificial time series. The model is validated with respect to the reproduction of basic temperature- and heat wave characteristics in Section 5.1. Copyright 2009 Royal Meteorological Society 2.6. Regional climate model (RCM) simulations for control (1961–1990) and late 21st century climate Regional climate model outputs available within the PRUDENCE project (Christensen and Christensen, 2007) are used to estimate possible late 21st century changes in mean July–August TMAX and its variance over the area under study. Basic characteristics of the 10 RCMs (24 examined scenario runs and 16 control runs) are given in Christensen and Christensen (2007). The RCMs have a horizontal resolution of about 50 km; the only exception is the high-resolution runs of the HIRHAM model with a 25-km grid. A driving GCM for the RCM simulations, control ones as well as scenarios, is the Hadley Centre HadAM3 GCM (Pope et al., 2000); the RCAO RCM is driven also by the ECHAM4 GCM (Roeckner et al., 1996), and the ARPEGE model (a regional model over Europe interactively nested in a global model) represents a third type of boundary forcing in addition to the HadAM3 and ECHAM4. The RCMs are run under SRES-A2 and SRES-B2 emission scenarios (IPCC, 2001) except for CLM, RACMO, CHRM and REMO for which only SRES-A2 runs are available. The A2 (B2) emission scenario leads to a faster (slower) increase in concentrations of greenhouse gases in the atmosphere compared to the A1B scenario referred to in IPCC (2007b) as a baseline, thus representing another range of uncertainty in possible future changes. 3. Temporal and spatial variability of heat waves over 1961–2006 3.1. Annual characteristics Average regional series of heat wave characteristics and mean summer (July–August) daily maximum temperature (TMAX ) are examined in this section (Figures 2 and 3). Enhanced heat waves characteristics, in terms of both cumulative TMAX excess TS30 and annual duration, are conspicuous since the early 1990s in all seven regions. A single year with the most severe and longest heat waves was 1994 except for the southwest region with the highest values in 2003. In the eastern part of the country (southeast, northeast and central highland regions), the 1994 heat waves were closely followed in severity by those of 1992; 2003 and 2006 were the third and fourth years according to heat wave characteristics in that area. The 2003 and 2006 heat waves were more severe in the western part of the Czech Republic. If values over all 44 stations are averaged (bottom right panel of Figure 2), 1994 is clearly the year with the most severe heat waves; their duration and severity were similar in 1992, 2003 and 2006, and no other season of comparably enhanced heat wave characteristics appeared since 1961. Only 2 years (1980 and 1987), heat waves did not occur at any station. Similar features of temporal variability are present in Figure 3 which shows mean July–August TMAX since 1961. In all regions, 1992, 1994 and 2003 were the Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 94 J. KYSELÝ Figure 2. Annual heat wave duration (wide columns) and cumulative TMAX excess in heat waves (TS30, narrow columns) in individual regions over 1961–2006. Averages over all 44 stations are shown in the bottom right panel. This figure is available in colour online at www.interscience.wiley.com/ijoc three warmest July–August seasons, with temperature anomalies relative to 1961–1990 between +3 and +5 ° C. The 2003 summer was the warmest one in the western parts of the country while 1992 in the eastern parts. Note that the 2006 July–August season is not characterized by a pronounced positive temperature anomaly since temperatures were not above average in August (cf. also Figure 4 discussed below). Linear trend estimates of the mean July–August TMAX over 1961–2006 are given in Table II; the increasing trends range from 0.40 to 0.63 ° C per decade, being significant at the 0.05 level in all regions. The trend is particularly large in the northern parts and relatively less pronounced in the south region. The mean July–August temperature has been above the 1961–1990 mean in all Copyright 2009 Royal Meteorological Society years since 1988 except for 1996 if averaged over all stations (bottom right panel of Figure 3), and except for 1–3 years also in all individual regions. The trend magnitudes are large but entirely consistent with increases of the mean summer TMAX over Europe reported by Klein Tank and Koennen (2003). Nevertheless, it has to be realized that linear regression is a very simplified model reporting only a general tendency; the temporal behaviour of many characteristics is closer to a step-like change in the early 1990s (see also Section 6). 3.2. The longest and most severe heat waves At a large majority of sites, the longest single heat wave occurred in July/August 1994 (Table I); its duration exceeded 1 month in the southeast region, and its distinct Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE 95 Figure 3. Same as in Figure 2 except for mean July–August TMAX anomaly (relative to 1961–1990). This figure is available in colour online at www.interscience.wiley.com/ijoc feature was an extremely long uninterrupted period of tropical days (more details in Kyselý (2002a, 2002b)). The heat waves of July/August 2003 and July 2006 were the longest ones at six and four stations, respectively. If severity of individual heat waves is measured by cumulative TMAX excess TS30, the July/August 1994 heat wave was the most severe one at 33 out of the 44 stations; the July/August 2003 heat wave at eight stations; and the July 2006 heat wave at three stations (Table I). The severe heat waves that occurred in the 1990s and 2000s were not particularly distinct for extremely high one-day temperatures. The record-breaking daily TMAX in the area were observed in July 1983, i.e. in a year when heat wave characteristics were less pronounced and heat waves were confined to relatively short periods. On July 27, 1983, TMAX reached 40 ° C at a number of locations in the southwestern and central Copyright 2009 Royal Meteorological Society parts of the Czech Republic, with a maximum of 40.2 ° C recorded in a southeast suburb of Prague (Krška and Munzar, 1984). A peak TMAX during the more recent heat waves only exceptionally exceeded 38 ° C. The fact that individual stations’ record-breaking TMAX were not generally reached during the 2003 and 2006 heat waves is demonstrated in Table I: at 22 stations, the highest TMAX observed over 1961–2006 appeared on 27 July 1983; at seven and six stations, the absolute maxima were recorded in 1994 and 1992, respectively. Recordbreaking TMAX was observed at 12 stations only in 2003 (on August 13), and at one station in 2006. (At two sites, the highest TMAX was reached in 2 years, so the numbers do not add up to 46.) The most distinct feature of the recent occurrences of heat waves is the persistence of the heat. Relatively frequent cold front passages usually interrupt or terminate Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 96 J. KYSELÝ Figure 4. Course of TMAX and heat waves (plotted bold) at Prague-Klementinum during summers of 1983, 1992, 1994, 2003 and 2006. Horizontal lines show thresholds applied in the heat wave definition; full (empty) vertical arrows denote passages of strong cold (occluded) fronts. Table II. Linear trend estimates of mean July–August TMAX over 1961–2006 in individual regions and in the average series over 44 stations. Region Trend (° C/10 years) Northwest Southwest Central lowland Central highland Northeast South Southeast Average 0.62 0.51 0.50 0.53 0.63 0.40 0.55 0.54 heat waves in central Europe (passages of strong cold and occluded fronts during five summer seasons discussed above are denoted by arrows in Figure 4). That was also the case of summer 1983; although most of the fronts were weak or moderate in intensity, they interrupted the hot periods several times (Krška and Munzar, 1984). Cold fronts were relatively frequent (but mostly decaying over central Europe) in the summer months of 2003, too, causing a large number of slight coolings (Figure 4); they were much less frequent in 1992, 1994 and 2006 when Copyright 2009 Royal Meteorological Society long periods with high air temperature and low interdiurnal temperature variability were related to persistent highpressure systems influencing central Europe. Remarkable is the fact that passages of strong cold fronts over Prague (according to the classification of the Czech Hydrometeorological Institute) were completely absent over the whole June–August period in 2006, supporting also the extremely low mean interdiurnal temperature change during this summer season. It should be noted that changes in the frequency of front passages, which are related to other features of atmospheric circulation, are mentioned Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE here to illustrate differences between individual summers and not as one of the mechanisms that contributed to increases in the heat wave characteristics. The role of persistent circulation patterns on the heat wave development is discussed in Section 6. 4. Recent heat waves in view of long-term variability The analysis of heat waves in the longest available series, Prague-Klementinum, clearly shows that the recent period of enhanced heat wave severity and duration is unprecedented since the beginning of temperature measurements (1775). The year with the largest overall duration of heat waves was 2006, followed by 1994, 2003 and 1947. In terms of annual cumulative TMAX excess TS30, the 1994 heat waves were most severe, followed by those in 2006, 2003 and 1947 (Figure 5). In addition to the 1990s and 2000s, other periods with more frequent heat waves appeared around 1950, 1930 and 1865. (Moderately enhanced heat wave characteristics occur around 1805, too, i.e. in the period with less reliable data. Since the direction of the possible inhomogeneity, located to 1837 by homogeneity tests (Štěpánek, 2005), points to too high temperatures before 1837, summer TMAX as well as heat wave characteristics would be lower after the suggested correction is applied to data before 1837.) Similar features of long-term variability are manifested for mean July–August temperature anomalies (bottom panel in Figure 5), namely the quasi-oscillatory behaviour with a period of around 60–70 years and an increasing magnitude of the peaks in the 20th century. Spectral density functions of detrended series are dominated by low-frequency variability; enhanced variability 97 between 50 and 100 years appears in the periodograms of both TS30 and mean July–August temperature (Figure 6). The maxima around 2000 and 1950 coincide with maxima of the Atlantic Multidecadal Oscillation index (discussed in Section 6; Enfield et al., 2001; Sutton and Hodson, 2005); the warming observed since the late 19th century in Europe, interrupted with a period of cooling in the 1950s to 1970s, likely contributed to the fact that the peaks in the 20th century had a larger magnitude. The recent period is particularly notable owing to pronounced positive anomalies persisting since the late 1980s: after 1987, negative deviations relative to 1961–1990 occurred in 1 year only. The heat wave lasting from 1 July to 2 August 2006 (33 consecutive days) was the longest heat wave ever recorded at Prague-Klementinum; only two other heat waves since 1775 persisted more than 3 weeks, in 1992 (26 days) and 1994 (22 days; Kyselý, 2002a). The July 2006 heat wave was also the one with the largest cumulative TMAX excess TS30 and the number of tropical days (Table III); on the other hand, peak temperature was not extraordinarily high. The July 2006 heat wave was the longest and most severe at the Milešovka Mt. since 1906, too (Figure 7). The overall warming (in terms of mean July–August TMAX ) during the 20th century is even stronger at the Milešovka Mt., which confirms that the urban heat island development does not severely influence summer TMAX in Prague. This is in accord with the conclusions of Brázdil and Budı́ková (1999) concerning the insignificant role of the heat island intensification on summer temperature trends in Prague (see Section 2.2), and strengthens credibility of findings concerning centennial-scale variability of heat waves. Figure 5. Annual heat wave duration (left), cumulative TMAX excess in heat waves (TS30, right) and mean July–August TMAX anomaly (relative to 1961–1990, bottom) at Prague-Klementinum over 1775–2006. The smoothed curves were obtained by applying 9-year Gaussian filter. Copyright 2009 Royal Meteorological Society Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 98 J. KYSELÝ Figure 6. Periodograms and spectral densities of detrended series of annual cumulative TMAX excesses in heat waves (TS30, left) and mean July–August TMAX anomaly (right) at Prague-Klementinum over 1775–2006. 5-point Tukey–Hamming filter was used to calculate spectral density. Table III. Ten most severe heat waves at Prague-Klementinum over 1775–2006 (according to cumulative TMAX excess TS30). Year Start day End day 2006 1994 1957 1992 1921 1892 2003 1868 1952 1952 July 1 July 21 June 28 July 16 July 23 August 14 July 30 August 6 July 31 June 28 August 2 August 11 July 10 August 10 August 12 August 25 August 14 August 18 August 16 July 15 Duration No. of Peak TS30 (days) tropical TMAX (° C) (° C) days 33 22 13 26 21 12 16 13 17 18 21 16 8 12 11 11 11 10 10 8 35.3 36.0 37.6 35.8 34.7 35.9 36.8 34.5 35.5 35.6 49.1 47.6 34.2 33.3 31.4 28.4 26.8 24.8 21.9 20.3 The heat waves are listed in descending order of severity. 5. Probabilities of recurrence of severe heat waves in present and future climates 5.1. Evaluation of heat wave characteristics in TMAX series simulated by the AR(1) model for the 1961–2006 period The AR(1) model (Section 2.5) is used to generate 100 000 artificial time series of TMAX (corresponding to Copyright 2009 Royal Meteorological Society months of May to September and the 1961–2006 period) at two stations, Prague-Klementinum and Brno, taking explicitly into account the long-term increasing trend in mean TMAX and the seasonal cycles of mean, variance and the first-order autocorrelation coefficient of TMAX . Before probabilities of extreme heat waves are estimated, the model is validated with respect to the reproduction of basic temperature- and heat wave characteristics. Average, median, and the 5% and 95% quantiles of the distributions of mean annual heat wave- and TMAX characteristics over the 100 000 artificial time series are shown in Table IV. The model reproduces all heat wave properties (frequency, annual duration, cumulative TMAX excess TS30, location in a year) at both stations; the observed values lie within the 90% bounds of the simulated distributions. If average/median is concerned, the AR(1) model tends to underestimate mean annual frequency and duration of heat waves, and the mean annual number of tropical days; mean annual TS30 is modelled in good agreement with observations since simulated heat waves tend to peak at slightly higher temperatures (this concerns mainly Prague). The AR(1) model intrinsically captures three basic characteristics of TMAX (mean, variance and the firstorder autocorrelation coefficient). Regarding variance, the Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 99 RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE Figure 7. Same as in Figure 5 except for Milešovka Mt. over 1906–2006. Data over 1930–1935 are missing. Table IV. Comparison of mean annual heat wave- and temperature characteristics over 1961–2006 in observed data and in 100 000 simulations with the AR(1) model corresponding to the same 46-year period. a. Prague Observed AR(1), 5% AR(1), average AR(1), median AR(1), 95% Frequency of heat waves Duration of heat waves (days) Number of days with TMAX ≥ T1 Cumulative TMAX excess TS30 (° C) Mean interannual variability of July–August TMAX (° C) Mean interdiurnal TMAX variability in July–August (° C) 1.33 0.85 1.10 1.11 1.37 10.7 7.1 9.5 9.4 12.1 10.0 7.5 8.9 8.9 10.4 13.0 9.2 13.0 12.9 17.3 1.48 1.01 1.33 1.31 1.68 2.48 2.49 2.57 2.57 2.64 Frequency of heat waves Duration of heat waves (days) Number of days with TMAX ≥ T1 Cumulative TMAX excess TS30 (° C) Mean interannual variability of July–August TMAX (° C) Mean interdiurnal TMAX variability in July–August (° C) 1.30 0.87 1.11 1.11 1.37 10.3 7.2 9.6 9.6 12.2 9.3 7.7 9.1 9.1 10.5 11.8 9.5 13.3 13.2 17.4 1.47 1.01 1.33 1.32 1.67 2.48 2.54 2.62 2.62 2.70 b. Brno Observed AR(1), 5% AR(1), average AR(1), median AR(1), 95% Average value, median and the 5% and 95% quantiles of the distribution over 100 000 simulations are shown for the artificial samples. Details on the AR(1) model are given in Section 2.5. overall value is reproduced well but the model underestimates the interannual variability and overestimates the intraseasonal variability (the last two columns in Table IV), which is a typical feature of temperature series obtained by autoregressive models (Madden and Shea, 1978; Johnson et al., 1996; Hansen and Mavromatis, 2001; Mavromatis and Hansen, 2001). The interdiurnal TMAX variability is the only temperature characteristic that falls outside the 90% bounds of the simulated values Copyright 2009 Royal Meteorological Society in the observed data. However, the AR(1) model with a long-term trend, modelled explicitly, partly rectifies the ratio of the interannual/interdiurnal TMAX variability compared to an AR(1) model without a trend. The model reproduces also basic features of the distributions of lengths and cumulative TMAX excesses TS30 of individual heat waves (Figure 8; model’s curves were determined from 100 000 realizations of a 46-year period and are therefore smooth compared to observations). The Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 100 J. KYSELÝ Figure 8. Comparison of AR(1)-simulated and observed heat waves at Prague-Klementinum over 1961–2006. Left: distributions according to the length (top) and cumulative TMAX excess TS30 (bottom). Right: cumulative frequencies of heat waves longer than x days (top) and with TS30 exceeding x ° C (bottom). Table V. Frequencies of short, medium-length and long heat waves in observed data over 1961–2006 and in 100 000 simulations with the AR(1) model corresponding to the same 46-year period. Length of heat waves 3–7 days 8–10 days >10 days Prague observed AR(1) 82.6 57.1 26.1 26.1 23.9 27.0 Brno observed AR(1) 87.0 57.8 19.6 26.2 23.9 27.5 Frequencies of occurrence per 100 years are given. fact that heat waves lasting about 5 and 6 days are most frequent is captured; frequencies of short heat waves (3 to 7 days) are underestimated (Table V), which is related to a too large interdiurnal TMAX variability (discussed above) and leads to an overall slight underestimation of heat wave occurrence. Frequencies of mediumlength heat waves (8–10 days) and long heat waves (>10 days) are simulated in good agreement with observations in Prague and they are slightly overestimated in Brno (Table V). The ability of the AR(1) model to capture even very long heat waves is demonstrated by the fact that the model’s most extreme heat waves in the simulations based on current climate considerably exceed in lengths the record-breaking observed heat waves. Analogous results hold true also for the distribution of TS30 in individual heat waves (Figure 8 bottom); the AR(1) Copyright 2009 Royal Meteorological Society model underestimates frequencies of heat waves of low severity. The occurrence of severe heat waves is captured in a good agreement with observations; four heat waves in which TS30 exceeded 25.0 ° C were observed at both stations over the 46 years, and their mean frequency over the simulated 46-year periods is 4.3 (4.5) in Prague (Brno). 5.2. Scenarios of summer temperature changes during the 21st century Ensemble of 40 RCM simulations (24 scenarios and 16 control runs; Section 2.6) was employed to estimate the possible late 21st century increase in mean July–August TMAX with respect to the reference 1961–1990 period. The values were obtained as averages over gridboxes in the area; the spatial variability of the projected increase is relatively small (Figure 9 for two outputs of the HIRHAM model under SRES-A2 and SRESB2 scenarios, and two RCMs with the largest and the smallest rise of summer TMAX over central Europe) but the models substantially differ in the magnitude of the projected change (Table VI). According to the SRES-A2 scenario, the mean July– August TMAX change between 2071–2100 and 1961– 1990 is 5.7 ° C (averaged over the RCM runs), with the range from 3.3 to 9.4 ° C (Table VI). The SRES-B2 scenario yields a smaller increase of 4.2 ° C (on average), ranging from 2.9 ° C to 6.0 ° C. In fact, an increase of around 1.5 ° C relative to the mean over 1961–1990 has already happened (the trend being close to 0.5 ° C per decade in the examined area, Section 3.1). The magnitude Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 101 RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE Figure 9. Spatial patterns of projected changes in mean July–August TMAX between the late 21st century scenarios and control climate (1961–1990) in two runs of the HIRHAM RCM (top) and two RCMs with the largest and the smallest increase over central Europe (bottom). A2/B2 stands for the SRES emission scenarios. This figure is available in colour online at www.interscience.wiley.com/ijoc Table VI. Projected changes in mean July–August TMAX (mean) and the variance of daily TMAX in July–August (var) between the late 21st century scenarios and control climate (1961–1990) in individual model runs, averaged over all gridboxes in the examined area. RCM, SRES-A2 scenarios mean (° C) var (%) RCM, SRES-B2 scenarios mean (° C) var (%) HadRM # 1 HadRM # 2 HadRM # 3 HIRHAM # 1 HIRHAM # 2 HIRHAM # 3 HIRHAM, high-resolution CHRM PROMES RCAO (ECHAM) RCAO ARPEGE CLM RegCM RACMO REMO 7.4 7.6 8.0 5.1 5.2 5.1 4.8 4.8 5.4 9.4 6.2 5.6 4.4 5.0 4.6 3.3 8.1 20.4 10.8 20.6 −7.7 36.9 7.7 15.4 16.3 41.7 42.6 95.6 9.3 22.5 31.1 −6.3 HadRM HIRHAM PROMES RCAO (ECHAM) RCAO ARPEGE ARPEGE (ARPEGE/OPA) RegCM 6.0 3.2 4.8 5.6 4.6 3.2 2.9 3.4 18.8 4.2 16.2 19.2 28.1 40.5 16.4 7.3 Average (SRES-A2) 5.7 22.8 Average (SRES-B2) 4.2 18.8 # denotes ensemble member. A driving GCM is given in parentheses if different from the HadAM/HadCM model. of the summer TMAX change in central Europe is relatively large in the RCM outputs compared to the global mean projected warming (IPCC, 2007b), but the average scenario for the late 21st century (warming close to 5 ° C Copyright 2009 Royal Meteorological Society relative to 1961–1990) is entirely consistent with the trend estimated over 1961–2006 if extrapolated towards future. The projected increase is coupled to summer drying over western, central and southern Europe (Pal Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 102 J. KYSELÝ et al., 2004; Hagemann et al., 2004; Rowell and Jones, 2006; Zaitchik et al., 2006), manifestations of which have already been observed over the recent decades (Trnka et al., 2008). (Note that low soil moisture has also been demonstrated to play an important role in the 2003 European heat wave; Black et al., 2004; Fink et al., 2004; Fischer et al., 2007). Estimates of return periods associated with extreme heat waves in a future climate are based on 100 000 realizations of artificial TMAX series covering the period 1961–2100, generated by the AR(1) model under three assumptions that differ in the rate of the warming and the magnitude of the mean July–August TMAX change in 2071–2100 relative to 1961–1990: (A) Warming of 0.5 ° C/decade (mid-estimate, taking into account both emission scenarios and all RCMs); (B) Warming of 0.2 ° C/decade (represents lower bound of the uncertainty range based on RCM simulations and the SRES-A2 and B2 scenarios); (C) Warming of 0.9 ° C/decade (upper bound). The assumed magnitudes of warming capture a large range of possible future climate developments; since the mid-estimate coincides with the observed trend since 1961, the lower (upper) assumption means that the trend will become weaker (stronger) during the 21st century. We consider year 2007 (which is a somewhat arbitrary choice, making the interpretation of results easier due to a clear separation of ‘present’/observed (1961–2006) and future climates) as the starting point of the trend change in experiments B and C; no trend change is involved in experiment A. Constantly increasing linear trends over 2007–2100 are assumed in the simulations. Under these assumptions, the mean summer TMAX increase makes up 5.5 ° C (3.1 ° C, 8.7 ° C) in 2071–2100 relative to 1961–1990 in experiment A (B, C). Variance of daily temperatures and the first-order autocorrelation coefficients are kept unchanged in experiments A, B and C. However, since global warming is expected to be supplemented by a rise in interannual variability of summer temperatures over large parts of Europe (Schär et al., 2004; Klein Tank et al., 2005; Scherrer et al., 2005; Della-Marta et al., 2007b), which is driven by a decline in spring and summer precipitation and a strong land-atmosphere coupling (Seneviratne et al., 2006; Fischer et al., 2007), a variance increase of 2.5%/decade is assumed over 2007–2100 together with the mid-scenario of warming of 0.5 ° C/decade in another experiment A-var. The variance increase is achieved by adding a seasonal temperature anomaly, drawn from a normal distribution with rising variance over time (on the interannual scale), to all values in the particular realization of a given year [generated by the AR(1) model]. This assumption leads to a rise in summer temperature variance of about 21.5% in 2071–2100 relative to present climate, a value that is in good agreement with mean of the current RCM projections in the area (Table VI), Copyright 2009 Royal Meteorological Society and quite conservative compared to some modelling studies that predict up to 100% increase in the variance of summer temperatures over central Europe by 2071–2100 (Schär et al., 2004; Weisheimer and Palmer, 2005). 5.3. Return periods associated with severe heat waves in future climate Although it is obvious that heat waves become longer and more frequent in a warmer climate, quantitative estimates may be useful in some applications, and they illustrate effects of the changes in mean and variance of summer temperatures on extremes. Distributions of lengths and cumulative TMAX excesses (TS30) in heat waves are evaluated for two future time slices: 2031–2060 (‘mid21st century’) and 2071–2100 (‘late 21st century’). 5.3.1. Scenario A (warming of 0.5 ° C/decade) Sharp increases in probabilities of long and severe heat waves appear in the future time slices (Figure 10 top; Table VII). Heat waves with the duration and severity comparable to or exceeding those of the record-breaking heat waves observed over 1961–2006 become quite frequent: in Prague, heat waves lasting at least 33 days (the length of the 2006 heat wave) occur with a mean annual frequency of 0.12 (0.54) in 2031–2060 (2071–2100), and heat waves with TS30 exceeding 49 ° C have a mean annual frequency of 0.27 (0.92) in 2031–2060 (2071–2100). Analogous findings hold true for the Brno station and heat waves reaching or exceeding the most severe and longest heat wave (observed in 1994): heat waves lasting at least 32 days occur with a mean annual frequency of 0.14 (0.58) in 2031–2060 (2071–2100), and heat waves with TS30 exceeding 67 ° C have a mean annual frequency of 0.13 (0.62) in 2031–2060 (2071–2100). Nevertheless, it also appears that according to the mid-scenario A, heat waves with a severity and duration comparable to 2006 (Prague) and 1994 (Brno) may be expected once in around 4–8 years in the mid21st century, and not on a regular (annual) basis. Even at the end of the 21st century, there are likely to be summer seasons with heat waves less severe than in 2006 and 1994. Temporal changes in probabilities of severe heat waves over 1961–2100, based on the AR(1) model simulations, are shown in Figure 11; frequencies of events reaching or exceeding the record-breaking heat waves (over 1961–2006) are examined. The expected annual frequency of a heat wave lasting at least 33 days in Prague (32 days in Brno) was 0.0081 (0.0116) in 2006, corresponding to the return period of 123 (86) years. The return period associated with such events sharply increases towards more distant past; in 1980, it is estimated to be 1700 (900) years, so the probability has already increased by an order of magnitude over the recent 25 years. Although the model incorporates a number of simplified assumptions, the change clearly illustrates that estimates of return periods of the recent heat waves would be severely biased (overestimated) if based on the assumption of a stationary climate. Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE 103 Figure 10. Cumulative frequencies of heat waves longer than x days (left) and with TS30 exceeding x ° C (right) at Prague-Klementinum in observed data (1961–2006) and simulations with the AR(1) model for time periods 1961–2006, 2031–2060 and 2071–2100, for scenarios A (warming rate 0.5 ° C/decade), B (warming rate 0.2 ° C/decade), C (warming rate 0.9 ° C/decade) and A-var (warming rate 0.5 ° C/decade and variance increase 2.5%/decade). The longest and most severe heat waves observed over 1961–2006 are marked by vertical bars. The range of y-axis is kept constant in all graphs. This figure is available in colour online at www.interscience.wiley.com/ijoc From the present onwards, probabilities of long and severe heat waves rise at an increasing rate (Figure 11): the return period of the longest observed heat waves declines by a factor of 2 in 2016 (compared to 2006), and by a factor of 10 around 2040 at both stations. Around 2070, heat waves with the duration comparable Copyright 2009 Royal Meteorological Society to or exceeding the lengths of the recent record-breaking events should be expected to occur once in about 3 years, and after 2090 in 2 out of 3 years. Similar changes appear also for the return periods of the most severe heat waves according to TS30, which decrease by a factor of 2 (10) around 2016–2018 (2040–2050). Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 104 J. KYSELÝ Table VII. Mean annual frequencies of heat waves reaching or exceeding the length and severity of the record-breaking heat waves over 1961–2006, simulated by the AR(1) model under various scenarios of future summer temperature change. Scenario Trend in mean summer TMAX Trend in variance of summer TMAX Time slice A +0.5 ° C/decade – B +0.2 ° C/decade C +0.9 ° C/decade A-var Prague Brno Length ≥33 days TS30 ≥49 ° C Length ≥32 days TS30 ≥67 ° C 2031–2060 2071–2100 0.12 0.54 0.27 0.92 0.14 0.58 0.13 0.62 – 2031–2060 2071–2100 0.03 0.07 0.08 0.18 0.03 0.09 0.03 0.08 – 2031–2060 2071–2100 0.44 1.24 0.77 1.67 0.48 1.26 0.49 1.42 +0.5 ° C/decade +2.5%/decade 2031–2060 2071–2100 0.19 0.58 0.36 0.90 0.21 0.61 0.20 0.66 Figure 11. Temporal changes in cumulative frequencies of heat waves reaching or exceeding the record-breaking heat waves observed over 1961–2006 at Prague-Klementinum (left) and Brno (right) in simulations with the AR(1) model under scenarios A, B, C and A-var. This figure is available in colour online at www.interscience.wiley.com/ijoc 5.3.2. Scenario B (warming of 0.2 ° C/decade) 5.3.3. Scenario C (warming of 0.9 ° C/decade) An increase in the frequency and severity of heat waves is conspicuously smaller than in scenario A (Figures 10 and 11; Table VII). In Prague, heat waves lasting at least 33 days (the length of the 2006 heat wave) occur with a mean annual frequency of 0.03 (0.07) in 2031–2060 (2071–2100), and heat waves with TS30 exceeding 49 ° C have a mean annual frequency of 0.08 (0.18) in 2031–2060 (2071–2100). Similar figures hold true for future probabilities of the most severe heat wave in Brno. According to the lower bound scenario, the record-breaking recent events remain rare in the mid-21st century, and even in 2071–2100, their mean recurrence intervals are between 5 and 15 years. Huge differences between future probabilities of severe heat waves in simulations based on the upper bound and lower bound scenarios (Figures 10 and 11; Table VII) reflect large uncertainties in the magnitude of future summer warming projected by current RCMs. In scenario C, heat waves lasting at least 33 days occur in Prague with a mean annual frequency of 0.44 (1.24) in 2031–2060 (2071–2100), and heat waves with TS30 exceeding 49 ° C have a mean annual frequency of 0.77 (1.67) in 2031–2060 (2071–2100). This means that the frequency of such events is by an order of magnitude higher than that in the lower bound scenario B. Analogous results are obtained for heat waves in Brno (Table VII). Copyright 2009 Royal Meteorological Society Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 105 RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE 5.3.4. Scenario A-var (warming of 0.5 ° C/decade and variance increase of 2.5%/decade) If increasing trend in the variance of summer temperature is incorporated, frequencies of severe heat waves increase faster compared to the mid-scenario A (with variance kept constant), mainly during the first part of the 21st century (Figures 10 and 11, Table VII). In Prague, heat waves lasting at least 33 days occur with a mean annual frequency of 0.19 (0.58) in 2031–2060 (2071–2100), and heat waves with TS30 exceeding 49 ° C have a mean annual frequency of 0.36 (0.90) in 2031–2060 (2071–2100). Hence probabilities of such events are by 33–57% greater relative to scenario A in the mid-21st century time slice, but they become comparable in both scenarios in the late 21st century. The decreasing role of enhanced summer temperature variability towards the end of the 21st century (also Figure 11) is due to the fact that the severe heat waves (according to present climate) become quite usual in the second half of the 21st century, notwithstanding the variance changes, and if the variance increase is superimposed to the warming trend, relatively cold summers without severe heat waves tend to occur more often. On the other hand, rising variance clearly supports the development of heat waves with extremely enhanced duration and TS30: in the 2071–2100 time slice, heat waves covering at least 3 months occur once in 300 years in simulations based on scenario A while once in 20 years according to scenario A-var. 6. Discussion and concluding remarks 6.1. Temporal and spatial variability of heat waves Using data from a network of meteorological stations covering the area of the Czech Republic over 1961–2006, it is shown that 1994 was the year with the most severe and longest heat waves, and the 1994 heat waves were overtaken by those of 2003 only in the southwest region. The other two seasons with enhanced heat wave characteristics were 1992 (mainly in the eastern part of the country) and 2006 (in the western part, including central lowland region surrounding Prague). According to long-term temperature series at Prague-Klementinum, 2006 was the year with the largest overall duration of heat waves since 1775, followed by 1994, 2003 and 1947; the severity of heat waves measured by cumulative temperature excess was highest in 1994. The July 2006 heat wave, lasting 33 days, was the longest and most severe individual heat wave in Prague since 1775. Data from the Milešovka Mt. station over the 20th century confirm the finding concerning the 2006 heat wave and suggest that the enhanced heat wave characteristics at Prague-Klementinum in the 1990s and 2000s are unlikely to be associated with an urban heat island development. However, the regional differences in the temporal variability of heat waves in the station network in spite of a relatively small area point to the fact that conclusions concerning temporal variability of Copyright 2009 Royal Meteorological Society heat waves based on data from a single station may not be representative for a wider area. The recent increase in heat wave characteristics in central Europe appears to be related to several mechanisms in addition to the ‘global warming’ (IPCC, 2007b); they include long-term variability associated with the Atlantic Multidecadal Oscillation (Sutton and Hodson, 2005) and possibly other modes of climate variability; changes in the frequency and persistence of circulation patterns conducive to heat waves in central Europe (Kyselý and Domonkos, 2006); and changes towards increased variance of summer temperature (Della-Marta et al., 2007b). The roles of individual factors are difficult to be differentiated; some aspects are discussed below. Since all these mechanisms likely supported the occurrence of hot summers in Europe in the 1990s and 2000s, the increase in some heat wave characteristics was manifested rather as a step-like change in the early 1990s than a gradual upward trend. 6.2. Recent heat waves and persistent circulation patterns The main specific feature of the recent heat waves compared to past events was not the peak temperature reached but the persistence of the heat, related to highpressure systems influencing large parts of Europe for prolonged time periods. The importance of persistent anticyclonic conditions for the development of the 2003 heat wave over Europe was shown by Black et al. (2004) and Luterbacher et al. (2004), and for the 1992 and 1994 heat waves in Prague by Kyselý (2002a). Enhanced persistence of summer heat waves over western Europe (doubled length since 1880) was documented by DellaMarta et al. (2007b). Since increases in residence times of circulation types over Europe and in the frequency of circulation patterns that exhibit higher persistence were reported recently (Werner et al., 2000; Kyselý and Domonkos, 2006; Beniston and Goyette, 2007), the rising trends in warm temperature extremes over Europe (Klein Tank and Koennen, 2003; Moberg et al., 2006; Della-Marta et al., 2007b) may partly be linked to the higher persistence of atmospheric circulation. Figure 12 suggests that an upward trend in the frequency of the Hess-Brezowsky circulation types (described in Gerstengarbe et al., 1999) conducive to heat waves in central Europe, observed over the 20th century, has likely contributed to the increases in heat wave characteristics; enhanced persistence of the types since the late 1980s is another factor playing a role since air temperature anomalies tend to increase with the duration of most types that support the development of heat waves (Kyselý, 2008). 6.3. Anthropogenic versus natural forcings on summer temperatures and heat waves in Europe The recently observed changes in mean summer temperatures and heat waves appear to be at least partly due to human activities; a detectable anthropogenic influence on the frequency of extremely warm events over Europe was Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 106 J. KYSELÝ Figure 12. Changes in the relative frequency and mean residence times of the Hess–Brezowsky circulation types conducive to heat waves in central Europe (central-European high, south and east types) in summer seasons (JJA) over 1881–2006. 9-year running means are shown. demonstrated by Stott et al. (2004). However, it should be also pointed to the role of natural climatic variability, as long-term fluctuations of European summer temperatures are likely related to an SST mode with a monopole structure over the North Atlantic called the Atlantic Multidecadal Oscillation (AMO; Sutton and Hodson, 2005; Della-Marta et al., 2007a). The AMO is probably associated with the North Atlantic thermohaline circulation (Knight et al., 2005). During the 20th century, the AMO index reached its maxima in the 1940s/1950s and the late 1990s while minima in the 1910s and the 1970s (Enfield et al., 2001), corresponding almost perfectly to periods of increased (maxima) and decreased (minima) heat wave severity in Prague. The AMO index was found to be a possible predictor of the frequency of heat waves over western Europe at the decadal time scale (Della-Marta et al., 2007a). Since the AMO may weaken in the next 50 years (Knight et al., 2005), the AMO-related summer cooling may partially offset the rise in the severity of heat waves expected from anthropogenic influences, and the lower bound scenario of future summer warming may be the more realistic one over the next few decades. Other studies attribute a substantial part of the warming over the first half of the 20th century to solar and volcanic forcings while the increase in temperatures during the second half of the 20th century is largely attributable to anthropogenic influences (Stott et al., 2004; Klein Tank et al., 2005; IPCC, 2007b). The extent to which the AMO sharpens summer heat wave severity over Europe (as well as the possible physical mechanisms of the link) remains unclear. 6.4. Stochastic modelling of heat wave probabilities A first-order autoregressive [AR(1)] model with a deterministic component that incorporates explicitly the seasonal cycle and the long-term trend was employed to generate artificial series of TMAX from which probabilities of long and severe heat waves were estimated. The stochastic model was found capable of reproducing basic characteristics of heat waves in present climate Copyright 2009 Royal Meteorological Society (1961–2006), and was subsequently run under several assumptions reflecting various rates of summer warming towards future (2007–2100), estimated from an ensemble of RCM outputs. Under the mid-scenario, which assumes an unchanged rate of warming during the 21st century as the one estimated over 1961–2006 (0.5 ° C/decade), probabilities of long and severe heat waves sharply increase. However, heat waves with a severity and duration comparable to the record-breaking ones in 2006 and 1994 do not become common events before the late 21st century, as they may be expected to occur once in around 4–8 years in the mid-21st century, and even at the end of the 21st century there are likely to be summer seasons with heat waves not that severe as in 2006 and 1994. The estimated future frequencies of severe heat waves differ by an order of magnitude between the upper bound and lower bound scenarios, which reflects large uncertainties in future summer warming projections over central Europe. According to the AR(1) model simulations, the return period of a heat wave reaching or exceeding the length of the 2006 heat wave in Prague is estimated to be around 120 years in 2006. Owing to an increase in mean summer temperatures, probabilities of the very long heat waves have already risen by an order of magnitude over the recent 25 years, and they are likely to increase by another order of magnitude by around 2040 under the summer warming rate assumed by the mid-scenario. Even the lower bound scenario, which counts on a decline in the rate of warming over the 21st century compared to the recently observed one (assuming a mean summer temperature increase of 0.2 ° C/decade over 2007–2100), yields a considerable decline of return periods associated with the long and most severe heat waves. 6.5. Effects of the possible increase in summer temperature variance due to summer drying and land–atmosphere coupling on heat waves If a moderate increase in the variance of summer temperature, which has been observed over western and central Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE Europe in the recent decades and appears to be a likely future scenario according to climate model simulations, is taken into account in addition to the mid-scenario of the warming over the 21st century, the return periods of the record-breaking heat waves shorten to around 3–5 years in the middle of the 21st century, and increasing variance clearly supports heat waves with extremely enhanced characteristics (relative to the same warming rate with constant variance). It should be pointed out that central Europe is a region where the summer drying and feedbacks between the land surface and the atmosphere, leading to an increase in the interannual temperature variability, may play a particularly pronounced role in a future climate (Seneviratne et al., 2006), and where the observed increases in the variance of summer temperature were largest (Della-Marta et al., 2007b). Such possible future changes in variance would lead to greatly amplified changes in extremes (Katz and Brown, 1992; Della-Marta et al., 2007b). 6.6. Limitations of the stochastic model The main statistical drawback of the AR(1) model adopted for the simulation of daily temperature series in the present and possible future climates is related to the reproduction of the ratio between interannual and intraseasonal temperature variability, which is, however, unlikely to severely bias the results, as all other basic temperature- and heat wave characteristics are captured in a close agreement with observations in the present climate simulations. Other limitations stem from the simplified assumptions adopted in the scenarios (e.g. constant rates of warming during the 21st century, no long-term changes in autocorrelations of daily temperature) and large uncertainties associated mainly with the magnitude of the change in variance of summer temperatures. Future studies should refine the climate change scenarios and rectify some of the drawbacks, but the conclusion that the longest and most severe heat waves observed recently appear to be typical rather of a late 21st century than a mid-21st century climate will likely remain unchanged. A similar approach based on simulations with a stochastic time series model (with a seasonal cycle and long-term trend or variability modelled explicitly) may be a useful tool in estimating probabilities of heat waves and related extremes in other regions and under various possible future climate developments. Acknowledgements The author is indebted to several persons for helping with the paper: particularly to R.Beranová, Institute of Atmospheric Physics, Prague, for preparing and analyzing RCM outputs and drawing Figure 9; I.Nemešová, Institute of Atmospheric Physics, Prague, for comments on an earlier version of the manuscript; J.Picek, Technical University Liberec, for testing the order of the AR model; E.Plavcová, Faculty of Mathematics and Physics, Charles University, Prague, for preparing database of passages of atmospheric fronts; and P.Skalák and the staff Copyright 2009 Royal Meteorological Society 107 of the Czech Hydrometeorological Institute, Prague, for preparing observed temperature datasets. The study was supported by the Czech Science Foundation under project 205/07/J044. References Agnew MD, Palutikof JP. 1999. The impacts of climate on retailing in the UK with particular reference to the anomalously hot summer of 1995. International Journal of Climatology 19: 1493–1507. Beniston M. 2004. The 2003 heat wave in Europe: a shape of things to come? An analysis based on Swiss climatological data and model simulations. Geophysical Research Letters 31: L02202. DOI:10.1029/2003GL018857. Beniston M, Goyette S. 2007. Changes in variability and persistence of climate in Switzerland: exploring 20th century observations and 21st century simulations. Global and Planetary Change 57: 1–15. Black E, Blackburn M, Harison G, Hoskins B, Methven J. 2004. Factors contributing to the summer 2003 heatwave. Weather 59: 217–223. Brázdil R, Budı́ková M. 1999. An urban bias in air temperature fluctuations at the Klementinum, Prague, the Czech Republic. Atmospheric Environment 33: 4211–4217. Brázdil R, Štekl J, et al. 1999. Climatic conditions of Milešovka. Academic Press: Prague; 434 [in Czech]. Chase TN, Wolter K, Pielke RA, Rasool I. 2006. Was the 2003 European summer heat wave unusual in a global context? Geophysical Research Letters 33: L23709. DOI:10.1029/2006GL027470. Christensen JH, Christensen OB. 2007. A summary of the PRUDENCE model projections of changes in European climate by the end of this century. Climatic Change 81: 7–30. Ciais P, Reichstein M, Viovy N, Granier A, Ogée J, Allard V, Aubinet M, Buchman N, Bernhofer C, Carrar A, Chevallier F, De Noblet N, Friend AD, Fiedlingstein P, Grünwald T, Heinesch B, Keronen P, Knohl A, Krinner G, Loustau D, Manca G, Matteucci G, Miglietta F, Ourcival JM, Papale D, Pilegaard K, Rambal S, Seufert G, Soussana JF, Sanz MJ, Schulze ED, Vesala T, Valentini R. 2005. Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature 437: 529–533. Coles S. 2001. An Introduction to Statistical Modeling of Extreme Values. Springer Verlag: London, 208. Colombo AF, Etkin D, Karney BW. 1999. Climate variability and the frequency of extreme temperature events for nine sites across Canada: implications for power usage. Journal of Climate 12: 2490–2502. De Bono A, Giuliani G, Kluser S, Peduzzi P. 2004. Impacts of Summer 2003 Heat Wave in Europe. UNEP/DEWA/GRID-Europe Environment Alert Bulletin, vol. 2, 1–4. Della-Marta PM, Haylock MR, Luterbacher J, Wanner H. 2007b. Doubled length of Western European summer heat waves since 1880. Journal of Geophysical Research 112: D15103, DOI: 10.1029/2007JD008510. Della-Marta PM, Luterbacher J, von Weissenfluh H, Xoplaki E, Brunet M, Wanner H. 2007a. Summer heat waves over western Europe 1880–2003, their relationship to large-scale forcings and predictability. Climate Dynamics 29: 251–275. EC. 2007. The 2006 European Heat Wave. The European Commission; [available at http://ec.europa.eu/health/ph information/dissemination/ unexpected/unexpected 9 en.htm, accessed August 11, 2007]. Enfield DB, Mestas-Nunez AM, Trimble PJ. 2001. The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental US. Geophysical Research Letters 28: 2077–2080. Filleul L, Cassadou S, Medina S, Fabres P, Lefranc A, Eilstein D, Le Tertre A, Pascal L, Chardon B, Blanchard M, Declercq C, Jusot JF, Prouvost H, Ledrans M. 2006. The relation between temperature, ozone, and mortality in nine French cities during the heat wave of 2003. Environmental Health Perspectives 114: 1344–1347. Fink A, Brucher T, Kruger A, Leckebusch G, Pinto J, Ulbrich U. 2004. The 2003 European summer heatwaves and drought – synoptic diagnosis and impacts. Weather 59: 209–216. Fischer EM, Seneviratne SI, Luthi D, Schär C. 2007. Contribution of land-atmosphere coupling to recent European summer heat waves. Geophysical Research Letters 34: L06707. DOI: 10.1029/2006GL029068. Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc 108 J. KYSELÝ Gerstengarbe F-W, Werner PC, Rüge U. 1999. Katalog der Grosswetterlagen Europas nach Paul Hess und Helmuth Brezowsky 18811998 . Deutscher Wetterdienst, Offenbach a. Main [available at http://www.pik-potsdam.de/∼uwerner/gwl/welcome.htm]. Grize L, Huss A, Thommen O, Schindler C, Braun-Fabrlander C. 2005. Heat wave 2003 and mortality in Switzerland. Swiss Medical Weekly 135: 200–205. Hagemann S, Machenhauer B, Jones R, Christensen OB, Déqué M, Jacob D, Vidale PL. 2004. Evaluation of water and energy budgets in regional climate models applied over Europe. Climate Dynamics 23: 547–567. Hallin M, Zahaf T, Jurečková J, Kalvová J, Picek J. 1997. Nonparametric tests in AR models with applications to climatic data. Environmetrics 8: 651–660. Hansen JW, Mavromatis T. 2001. Correcting low-frequency variability bias in stochastic weather generators. Agricultural and Forest Meteorology 109: 297–310. Hennessy KJ, Pittock AB. 1995. Greenhouse warming and threshold temperature events in Victoria, Australia. International Journal of Climatology 15: 591–612. Hlaváč V. 1937. Die Temperaturverhältnisse der Hauptstadt Prag. Teil I. Prager Geophysikalische studien VIII . Prague, 111. Horová I, Zelinka J, Brázdil R, Bud’ı́ková M. 2003. Density estimate and its application to analysis of temperature series. Environmetrics 14: 87–102. Huth R, Kyselý J, Pokorná L. 2000. A GCM simulation of heat waves, dry spells, and their relationships to circulation. Climatic Change 46: 29–60. Huth R, Nemešová I. 1995. Estimation of missing daily temperatures: can a weather categorization improve its accuracy. Journal of Climate 8: 1901–1916. Hutter HP, Moshammer H, Wallner P, Leitner B, Kundi M. 2007. Heatwaves in Vienna: effects on mortality. Wiener Klinische Wochenschriften 119: 223–227. IPCC. 2001. Climate Change: The Scientific Basis. The IPCC Third Assessment Report. Cambridge University Press: Cambridge. IPCC. 2007a. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the IPCC Fourth Assessment Report. Cambridge University Press: Cambridge; [available on-line at http://www.ipcc-wg2.org]. IPCC. 2007b. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the IPCC Fourth Assessment Report. Cambridge University Press: Cambridge; [available on-line at http://ipcc-wg1.ucar.edu/wg1/wg1-report.html]. Johnson GL, Hanson CL, Hardegree SP, Ballard EB. 1996. Stochastic weather simulation: overview and analysis of two commonly used models. Journal of Applied Meteorology 35: 1878–1896. Johnson H, Kovats RS, McGregor G, Stedman J, Gibbs M, Walton H. 2005. The impact of the 2003 heat wave on daily mortality in England and Wales and the use of rapid weekly mortality estimates. Euro Surveillance 10: 168–171. Jolly WM, Dobbertin M, Zimmermann NE, Reichstein M. 2005. Divergent vegetation growth responses to the 2003 heat wave in the Swiss Alps. Geophysical Research Letters 32: L18409. DOI:10.1029/2005GL023252. Kalvová J, Jurečková J, Picek J, Nemešová I. 2000. On the order of autoregressive (AR) model in temperature series. Meteorological Journal 3: 19–23. Katz RW, Brown BG. 1992. Extreme events in a changing climate: variability is more important than averages. Climatic Change 21: 289–302. Kendall M, Stuart A. 1976. The Advanced Theory of Statistics, vol. 3. Charles Griffin and Co.: London; 585. Klein Tank AMG, Koennen GP. 2003. Trends in indices of daily temperature and precipitation extremes in Europe, 1946-99. Journal of Climate 16: 3665–3680. Klein Tank AMG, Koennen GP, Selten FM. 2005. Signals of anthropogenic influence on European warming as seen in the trend patterns of daily temperature variance. International Journal of Climatology 25: 1–16. Knight JR, Allan RJ, Folland CK, Vellinga M, Mann ME. 2005. A signature of persistent natural thermohaline circulation cycles in observed climate. Geophysical Research Letters 32: L20708. DOI:10.1029/2005GL024233. Krška K, Munzar J. 1984. Temperature peculiarities of the tropic summer 1983 in Czechoslovakia and in Europe. Meteorologické Zprávy 37: 33–40. [in Czech, with English summary]. Kyselý J. 2002a. Temporal fluctuations in heat waves at PragueKlementinum, the Czech Republic, from 1901–1997, and their Copyright 2009 Royal Meteorological Society relationships to atmospheric circulation. International Journal of Climatology 22: 33–50. Kyselý J. 2002b. Probability estimates of extreme temperature events: stochastic modelling approach vs. extreme value distributions. Studia Geophysica et Geodaetica 46: 93–112. Kyselý J, Huth R. 2004. Heat-related mortality in the Czech Republic examined through synoptic and ‘traditional’ approaches. Climate Research 25: 265–274. Kyselý J. 2006. Spatial variability of heat waves in the Czech Republic and summer temperature peculiarity of southwest Bohemia. Meteorologické Zprávy 59: 183–189. [in Czech, with English summary]. Kyselý J, Domonkos P. 2006. Recent increase in persistence of atmospheric circulation over Europe: comparison with long-term variations since 1881. International Journal of Climatology 26: 461–483. Kyselý J. 2007. Implications of enhanced persistence of atmospheric circulation for the occurrence and severity of temperature extremes. International Journal of Climatology 27: 689–695. Kyselý J. 2008. Influence of the persistence of circulation patterns on warm and cold temperature anomalies in Europe: analysis over the 20th century. Global and Planetary Change 62: 147–163. Luterbacher J, Dietrich D, Xoplaki E, Grosjean M, Wanner H. 2004. European seasonal and annual temperature variability, trends, and extremes since 1500. Science 303: 1499–1503. Macchiato M, Serio C, Lapenna V, LaRotonda L. 1993. Parametric time series analysis of cold and hot spells in daily temperature: an application in southern Italy. Journal of Applied Meteorology 32: 1270–1281. Madden RA, Shea DJ. 1978. Estimates of the natural variability of time-averaged temperatures over the United States. Monthly Weather Review 106: 1695–1703. Mavromatis T, Hansen JW. 2001. Interannual variability characteristics and simulated crop response of four stochastic weather generators. Agricultural and Forest Meteorology 109: 283–296. Mearns LO, Katz RW, Schneider SH. 1984. Extreme high temperature events: changes in their probabilities with changes in mean temperature. Journal of Climate and Applied Meteorology 23: 1601–1608. Meehl GA, Tebaldi C. 2004. More intense, more frequent, and longer lasting heat waves in the 21st century. Science 305: 994–997. Moberg A, Jones PD, Lister D, Walther A, Brunet M, Jacobeit J, Alexander LV, Della-Marta PM, Luterbacher J, Yiou P, Chen DL, Klein Tank AMG, Saladie O, Sigro J, Aguilar E, Alexandersson H, Almarza C, Auer I, Barriendos M, Begert M, Bergstrom H, Bohm R, Butler CJ, Caesar J, Drebs A, Founda D, Gerstengarbe FW, Micela G, Maugeri M, Osterle H, Pandzic K, Petrakis M, Srnec L, Tolasz R, Tuomenvirta H, Werner PC, Linderholm H, Philipp A, Wanner H, Xoplaki E. 2006. Indices for daily temperature and precipitation extremes in Europe analyzed for the period 1901–2000. Journal of Geophysical Research 111: D22106. DOI:10.1029.2006JD007103. Pal JS, Giorgi F, Bi XQ. 2004. Consistency of recent European summer precipitation trends and extremes with future regional climate projections. Geophysical Research Letters 31: L13202. DOI:10.1029/2004GL019836. Pope DV, Gallani M, Rowntree R, Stratton RA. 2000. The impact of new physical parameterizations in the Hadley Centre climate model HadAM3. Climate Dynamics 16: 123–146. Press WH, Teukolsky SA, Vetterling WT, Flannery BP. 1992. Numerical Recipes in Fortran. The Art of Scientific Computing (2nd edn). Cambridge University Press; 963. Rebetez M, Mayer H, Dupont O, Schindler D, Gartner K, Kropp JP, Menzel A. 2006. Heat and drought 2003 in Europe: a climate synthesis. Annals of Forest Science 63: 569–577. Roeckner E, Arpe K, Bengtsson L, Christoph M, Claussen M, Dümenil L, Esch M, Giorgetta M, Schlese U, Schulzweida U. 1996. The Atmospheric General Circulation Model ECHAM4: Model Description and Simulation of Present-Day Climate. MPI Report 218, Hamburg. Rowell DP, Jones RG. 2006. Causes and uncertainty of future summer drying over Europe. Climate Dynamics 27: 281–299. Rusticucci M, Vargas W. 2002. Cold and warm events over Argentina and their relationship with the ENSO phases: risk evaluation analysis. International Journal of Climatology 22: 467–483. Schär C, Jendritzky G. 2004. Climate change: hot news from summer 2003. Nature 432: 559–560. Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc RECENT SEVERE HEAT WAVES IN CENTRAL EUROPE Schär C, Vidale PL, Lüthi D, Frei C, Häberli C, Liniger MA, Appenzeller C. 2004. The role of increasing temperature variability in European summer heatwaves. Nature 427: 332–336. Scherrer SC, Appenzeller C, Liniger MA, Schär C. 2005. European temperature distribution changes in observations and climate change scenarios. Geophysical Research Letters 32: L19705. DOI:10.1029/2005GL024108. Seneviratne SI, Lüthi D, Litschi M, Schär C. 2006. Land-atmosphere coupling and climate change in Europe. Nature 443: 205–209. Stott P, Stone D, Allen M. 2004. Human contribution to the European heatwave of 2003. Nature 432: 610–614. Sutton RT, Hodson DLR. 2005. Atlantic Ocean forcing of North American and European summer climate. Science 309: 115–118. Štekl J, Podzimek J. 1993. Old mountain meteorological station Milešovka (Donnersberg) in central Europe. Bulletin of the American Meteorological Society 74: 831–834. Štěpánek P. 2005. Air temperature fluctuations in the Czech republic in the period of instrumental measurements. PhD thesis. Faculty of Natural Sciences, Masaryk University, Brno, 136 [in Czech]. Trigo RM, Garcia-Herrera R, Diaz J, Trigo IF, Valente MA. 2005. How exceptional was the early August 2003 heatwave in Copyright 2009 Royal Meteorological Society 109 France. Geophysical Research Letters 32: L10701. DOI:10.1029/ 2005GL022410. Trnka M, Kyselý J, Možný M, Dubrovský M. 2008. Changes in Central European soil moisture availability and circulation patterns in 1881–2005. International Journal of Climatology. (in press). DOI:10.1002/joc.1703. Vandentorren S, Suzan F, Medina S, Pascal M, Maulpoix A, Cohen J-C, Ledrans M. 2004. Mortality in 13 French cities during the August 2003 heat wave. American Journal of Public Health 94: 1518–1520. Weisheimer A, Palmer TN. 2005. Changing frequency of occurrence of extreme seasonal temperatures under global warming. Geophysical Research Letters 32: L20721. DOI: 10.1029/2005GL023365. Werner PC, Gerstengarbe F-W, Fraedrich K, Oesterle H. 2000. Recent climate change in the North Atlantic/European sector. International Journal of Climatology 20: 463–471. Zaitchik BF, Macalady AK, Bonneau LR, Smith RB. 2006. Europe’s 2003 heat wave: A satellite view of impacts and land-atmosphere feedbacks. International Journal of Climatology 26: 743–769. Int. J. Climatol. 30: 89–109 (2010) DOI: 10.1002/joc
© Copyright 2026 Paperzz