This paper presents an investigation into the record breaking winter and summer maximum and minimum temperatures of 2007 across the Greek region. Three summer heat waves were analysed representing the highest maximum temperatures of the last six to seven decades. Winter minimum temperatures (mainly during January) were also found to be exceptionally high and they were proven to be more “statistically extreme” than the summer maxima. During the heat waves, an extended horizontal transfer of warm air masses from northern Africa was detected. The comparison of the results with future projections derived from a regional climate model revealed that these temperature conditions are likely to occur more than 50% of the time by the end of the 21st century.
 For the majority of the eastern Mediterranean, the Balkan Peninsula and especially for Greece [Busuioc et al., 2007; Cheval et al., 2008; Founda and Giannakopoulos, 2009], 2007 can be characterised as a remarkably warm year. Three major heat waves were detected, one in each summer month, with maximum and minimum temperature values breaking many records since late 1940s. These extremely high temperature conditions were accompanied by quite severe social and environmental consequences, resulting in the death of seven people and the heat apoplexy of many others, serious problems in the electrical supply and the establishment of favourable conditions for forest fires (I. Pytharoulis et al., Characteristics and predictability of the intense heat wave that affected Greece in June 2007, paper presented at the 9th COMECAP 2008, European Social Fund, Thessaloniki, Greece, 2008). Katavoutas et al.  studied also the effects of this type of heat waves on the human body's thermal balance. In addition, abnormally high maximum (Tmax hereafter) and minimum (Tmin hereafter) temperatures, with large departures from their mean values, were also detected during winter 2007 (mainly January). According to Beninston , although this kind of extreme episode cannot be referred to as a heat wave, they are worth analysing due to the serious nature of their effect on agriculture and the environment.
 The main aim of this study is to examine, from a statistical point of view, the extreme temperature conditions of 2007, for Thessaloniki, the second largest city in Greece, situated in the north of the Greek region (central Macedonia). An attempt was made to investigate any common characteristics between the heat wave episodes and to analyse the dynamic-synoptic aspects that may have caused these extreme events. Finally, a comparison was made with simulated data from a regional climate model (RCM) to determine whether these extreme temperature conditions could represent the mean temperature conditions the study domain could experience by the end of the 21st century.
2. Data and Methodology
 The data utilised in the study consist of daily Tmax and Tmin time series. The data were recorded at the meteorological station located at the University of Thessaloniki (AUTh), and cover the time period 1946–2007. The station is located on the shallow closed Gulf of Thessaloniki which acts in the same way as a continental lake and presents a surface temperature of about 30°C during the hottest summer months. There are two main reasons for the use of this station for analysis. First, a long time series of homogenised temperature data was available for Thessaloniki. Second, within this time series, 94% of the days in July and also in January 2007 experienced Tmax that were higher than the mean Tmax values of the period of the available data (1946–2006). Regarding the reference period (1961–1990) this percentage was even higher reaching 97% in July and January 2007. Temperature data from several other meteorological stations in Greece were also employed to confirm the extreme temperature conditions, but due to space limitations their results have not been included in this paper. Suggestively it could be mentioned in the continental stations of Ioannina (altitude 484 m) and Larissa (in both these cities the metrological station is outside of the city) the maximum temperatures for the same days (as the ones analysed for Thessaloniki) reached the values 41.4°C and 45.0°C respectively. Also, for the island stations of Skyros and Skopelos in the northern Aegean Sea the maximum temperature reach the value of 40°C during the days of the heat waves.
 Future projections of Tmax and Tmin were provided by the Royal Netherlands Meteorological Institute (Koninklijk Nederlands Meteorologisch Instituut, widely known as KNMI) RCM [Lenderink et al., 2003; van den Hurk et al., 2006]. The model is developed under the framework of the European Union project ENSEMBLES (http://ensembles-eu.metoffice.com) and is forced with output from a transient run from the ECHAM5 global climate model under the SRES A1B greenhouse gas emissions scenario. The RCM data have a 25 km horizontal resolution, covering the study domain with 114 grid points in the longitudinal and 124 grid points in the latitudinal direction. The model uses a Semi-Lagrangian dynamical core and presents 31 vertical levels, corresponding to the ECMWF 31-level resolution, with the lowest level at 30 m. For the current analysis the nearest grid point to the meteorological station, located at a distance of only 8 km from Thessaloniki, was used.
 Following the methodology proposed by Beninston , a statistical analysis of the observed temperature data was undertaken initially; both on a daily and seasonal basis, and the heat waves were compared in order to examine the existence of common characteristics. Furthermore, in order to study the relationships between the heat waves and the atmospheric circulation changes, the fields of 500 hPa and the 1000–500 hPa thickness were examined on a basis of daily NCEP/NCAR data [Kalnay et al., 1996]. Finally, using the simulated data series from the RCM, several quantiles were computed and the results of the future projections for the last 30 years of the 21st century were compared with the 2007 observations.
3.1. Statistical Analysis of Temperature Time Series
 From the AUTh daily time series it was found that mean Tmax and Tmin during January, for the time period 1946–2006, are 9.6°C and 2.3°C, respectively. Yet for 2007 the mean values rise to 14.4°C for Tmax and 4.7°C for Tmin, which represents a doubling of the latter as compared to the corresponding long term mean value. Concerning summer, the equivalent values for Tmax and Tmin are shown in Table 1. It is clear from Table 1 that the increase in summer Tmax is greater than the increase found in January's Tmax. June 2007 is noted as the hottest June in the last 62 years, when the absolute maximum values of both Tmax and Tmin were the highest recorded during the study period. The results for July are similar, and only the record of the highest Tmin value was not broken in 2007. Finally, regarding August, the mean Tmax and Tmin values for 2007 are higher than the mean values of the study period, but the absolute maximum values of both Tmax and Tmin were observed in different years.
Table 1. Mean Maximum and Minimum Temperature Values for the Time Period 1946–2006 and the Year 2007 for the Summer Months
Highest Tmax value
Highest Tmax value
(Highest Tmin value)
(Highest Tmin value)
 According to Figure 1, which illustrates daily Tmax for the three summer months of 2007, along with the mean daily Tmax values for the study period (1946–2006), the heat waves seem to have some common characteristics, as was also observed by Katavoutas et al. . These extreme maximum temperatures appeared during the last few days of each month, roughly between the 21st and 28th, a fact that is random since it could not be attributed to specific factors. More specifically the days with the greatest intensity of the heat waves were for June (25/6–28/6), for July (19/7–20/7 and 24/7–26/7) and for August (24/8–27/8). Also, the day with the highest Tmax value was very close in each month, the 27th for June, and the 25th for July and August. Moreover, it should be mentioned that in the case of June, and especially of July 2007, almost all days presented Tmax values that exceeded the mean daily Tmax for the period 1946–2006. An exception is the first week of August, which appears to be “colder” than the mean Tmax values. Analogous results were found for the rest of the meteorological stations in the study domain.
 Examining Tmax from a seasonal point of view, it was found that for all seasons, 2007 was one of the hottest during the study period for the station of AUTh, as well as for the whole country. More specifically, for the summer (JJA) season, seven years with the highest Tmax values are found at the end of the time period, with 2007 being the hottest. For winter, the hottest years were detected both at the beginning and at the end of the time period, while for the transition seasons (MAM, SON) the ten hottest years are more dispersed. Analogous results were also found for Tmin. In summer (JJA), the hottest years were found at the end of the time period. In spring (MAM) four out of the ten years with the highest Tmin values are found between 2000 and 2007.
3.2. Links Between the Heat Wave Episodes and Atmospheric Circulation
 An investigation into the synoptic conditions on the days of the heat waves was also conducted. The 500 hPa geopotential heights and the 1000–500 hPa thickness mean fields for the days of each heat wave of 2007 were determined and compared to the general mean monthly field of the whole study period (1946–2006).
 For all three summer heat wave episodes, the synoptic conditions at the 500 hPa level show that a prevailing ridge, with its axis oriented from southwest to northeast, was situated over the Greek region and the Balkan Peninsula. Hot and dry north African air was channelled across Italy, the Balkans and Greece, which resulted in the occurrence of light northerly continental winds over the Aegean Sea, a result found and observed clearly from the analysis of the equivalent SLP field (not shown). In addition, large positive 500 hPa anomalies dominated over the Balkan area. The centre of the anomalies is located, for the three cases over Greece, with similar magnitude. The equivalent patterns of the 1000–500 hPa anomaly fields presented the largest values for June's heat wave – a possible contributing factor for its greater intensity.
 With respect to the days with high temperatures during January 2007, synoptic conditions over Greece present many similarities to those found in summer. It should be mentioned that the mean and anomaly fields derived from the days with extremely high Tmin (which presented larger departures from the mean Tmin in comparison to Tmax). There was a transfer of warmer air masses mainly from the southwest to the southern parts of the Balkan Peninsula as well as to Greece, where an intense centre of positive anomalies (both at the 500 hPa level and in the 1000–500 hPa thickness field) is found.
 In general, it is evident that in all of the heat wave episodes or in the days with extremely high temperatures in the Greek region, the main and prevailing cause is not only the extended horizontal transfer of warm air masses from northern Africa but also the abrupt and intense subsidence of air masses in the middle troposphere into the boundary layer, resulting to the adiabatic heating of the air near the surface [Prezerakos, 1989].
 Finally, it appears that there is a strong link between the daily summer Tmax of 2007 and the daily 1000–500 hPa thickness values over Thessaloniki, derived from the closest grid points to the meteorological station. In Figure 2, we attempt to compare the day-to-day variations of the two time series. It is evident that during the days with extreme high temperatures, the thickness values also appear to be higher, presenting three peaks that match each of the three summer heat waves. Analogous were the results from the comparison of the daily summer Tmax with the equivalent 500 hPa gridded data (not shown).
3.3. Comparison With Future Model Projections
 Utilising the daily Tmax and Tmin data from the KNMI model (the closest grid point to the station of Thessaloniki), the monthly Tmax and Tmin means, as well as their corresponding percentiles (Q10, Q25, Q50, Q75 and Q90), for the 30-year future period 2071–2100, were computed. The results were compared with the observed mean monthly Tmax and Tmin for the time period 1946–2006 and with the monthly Tmax and Tmin values for the year 2007. In the case of the Tmax comparison (Figure 3a), summer 2007 Tmax appear to be equivalent to the 10% quantile of the RCM projections according to the A1B scenario. What is noteworthy is that the January 2007 Tmax exceed even the limit of the 75% quantile, while the respected values for February, March, April and May were just below the 50% quantile. Therefore, it seems that the winter and spring Tmax values are more exceptional than the equivalent summer values for the particular year.
 The results regarding the Tmin comparison are quite similar. The Tmin values for 2007 for the first three months of the year (January, February, March) reach the 75% quantile limit of the future scenario data, while the summer Tmin are almost equal to the 10% quantile. Finally, it was also found that the minimum temperatures of October, November and December, are as warm as the 50% quantile of the RCM data (Figure 3b).
 In the last step of the study a more thorough analysis of the daily summer and winter temperature data in comparison to the Tmax and Tmin future data was conducted. Thus, Figure 4a illustrates the daily summer Tmax values of 2007 and the daily Tmax time series of two quantiles (50% and 75%) of the future projections (2071–2100) according to the A1B scenario. A certain number of days were found to present higher Tmax values than the quantiles. In particular, nine days exceed the 50th percentile and five exceed even the 75th percentile: these are the days that represent the three summer heat waves. Regarding the winter Tmin results (Figure 4b), a much larger number of days appear to be “hotter” than the two quantiles. It was found that Tmin is higher than the limit of 50% for 57 days, and higher than the 75% quantile for 36 days. As mentioned in the previous paragraphs higher absolute Tmin values were observed during January.
4. Discussion and Conclusions
 As in 2003, which was exceptionally hot for many parts of central Europe [Schär et al., 2003; Beniston, 2004], similar “record breaking” extremely high temperatures were observed in the eastern Mediterranean during 2007. Three heat waves were recorded during the summer months, with days with Tmax exceeding 40°C. The first two extreme episodes (June and July) were far more intense than equivalent episodes in 1987 and 1988, despite the fact that during the heat wave of 1987 1300 human deaths were recorded in Athens and other large Greek cities [Giles and Balafoutis, 1990]. The remarkably high summer temperatures of 2007 were accompanied by minimal deaths due to the fact that the government, as well as the population, were better prepared as a result of their previous experiences. Winter temperatures were characterised by large departures from mean values. In fact, winter Tmin was found to be statistically more “extreme” than summer Tmax.
 These high temperatures, combined with very low rainfall totals in 2007 (the annual precipitation totals were 20% lower than the mean total for Thessaloniki), especially during the cold period, resulted in an early start to the growing period for certain plant species (short period of dormancy), the early fruitage of cultivations and a depletion in the quality of the harvest with serious consequences for the national economy. The meteorological conditions of 2007 played the main role in the increase in the number of forest fires. During the August heat wave a large number of fires were observed with catastrophic consequences, especially in the region of Peloponnesus, where 177265 hectares of the forest, urban and agricultural areas were destroyed and over 80 people died. Finally, the abnormal high winter temperatures, even in high altitudinal areas, had a serious impact on the hydrological balance as well as winter tourism, as a result of reduced snowfall and snow cover. Also, there were large losses for the agricultural and energy sectors due to water shortages for irrigation and cooling requirements during summer, respectively.
 The summer heat waves and the extremely high winter Tmin values are related strongly to the atmospheric circulation at the 500 hPa level and the 1000–500 hPa thickness field, presenting similar characteristics. A long-wave ridge was present at the 500 hPa level in all three cases, as well as an intense centre of positive anomalies over the study domain (Greece). The advection of warm air masses from northern Africa into the eastern Mediterranean was found to be one of the main factors for the appearance of the heat waves. Pytharoulis et al. (presented paper, 2008) mention in their study that the subtropical jet stream, located to the north of Greece, also contributed to the summer heat waves (this study refers to the first heat wave of June 2007) through its associated secondary ageostrophic circulation. According to Beniston  the winter warm spells of 2003 in Switzerland were also related to the occurrence of anticyclonic conditions.
 The comparison of the daily and seasonal 2007 Tmax and Tmin values from the AUTh meteorological station with the future projections of an RCM for the same region showed that the temperature conditions of 2007 are expected to occur more than 50% of the time in the future. Analogous results were found by Beniston  in his study for Switzerland based on the 2003 heat wave. It is evident that a possible increase in this kind of summer heat wave could have serious consequences for tourism and, subsequently, the Greek economy. The impact of extreme Tmax in 2007 on several environmental and economic sectors can illustrate the effects that 21st century climate changes may have on Greece. In addition, they might comprise a guideline for the new climate conditions in relation to adaptation strategies for agriculture, tourism, land use, water resources and, in general, for the economy and society as a whole.
 This study has been supported by the European Commission ENSEMBLES project (contract GOCE-CT-2003-505539).