Introduction

Based on observational data, the Intergovernmental Panel on Climate Change states that the warming of the climate system is unequivocal (IPCC 2013). The IPCC reports that human influence has contributed to observed global scale changes in the frequency and intensity of daily temperature extremes since the mid-twentieth century and has more than doubled the probability of occurrence of heat waves in some locations. These conclusions are supported by the inferred increased risk of heat extremes due to anthropogenic activities presented in a number of studies (Jones et al. 2008; Christidis et al. 2011, 2012). In the likely scenario of a 2 °C warmer world, almost every future summer will be as hot or hotter than the hottest that people have experienced in the recent past (National Research Council 2011). Increasingly, extreme events such as the recent European heat waves in 2003 and 2007, in Russia in 2010 and in the USA in 2012 offer glimpses into this warmer future. Especially the 2003 and 2010 cases are characterized as “mega-heat waves” (Barriopedro et al. 2011), being at variance with the 500-year seasonal temperature records over approximately 50 % of Europe.

The occurrence and intensity of heat waves are mostly related to large-scale general circulation patterns and certain atmospheric flow anomalies. Such features include adiabatic heating, warm air advection and positive radiation anomalies as a result of reduced cloudiness (Black et al. 2004; Meehl and Tebaldi 2004; Fink et al. 2004; Kenyon and Hegerl 2008; Grumm 2011). Factors that may amplify extreme heat events can also include soil moisture–atmosphere interactions related to dry conditions (Fischer et al. 2007; Seneviratne et al. 2010; Jaeger and Seneviratne 2011; Mueller and Seneviratne 2012; Zittis et al. 2014).

In the already environmentally stressed eastern Mediterranean and Middle East (EMME) region, significant positive trends of temperature extremes are indicated by a number of observation-based studies (Kostopoulou and Jones 2005; Maheras et al. 2006; Kuglitsch et al. 2010; Efthymiadis et al. 2011; Lelieveld et al. 2012; Tanarhte et al. 2012). Global and regional climate projections suggest that in this region heat stress intensification is very likely to continue throughout the twenty-first century (Sánchez et al. 2004; Giorgi 2006; Diffenbaugh et al. 2007; Fischer and Schär 2010; Hadjinicolaou et al. 2011; Lelieveld et al. 2013).

Despite that heat waves sometimes occur without significant consequences, their effects may be more widespread than other severe weather phenomena or natural hazards (Klinenberg 2002). For the rapidly increasing EMME population, the occurrence of such extreme events might have tremendous impacts. Increased hospital admissions and human mortality, failure of crops, higher water and energy consumption, are examples. In addition, the increasing urbanization (Grimm et al. 2008; United Nations 2012) and the associated locally produced or transported air pollution (Tressol et al. 2008; Solberg et al. 2008) or the urban heat island effect (Conti et al. 2005; Laaidi et al. 2012) might amplify these negative effects. Each of the societal aspects mentioned before can be differently affected by heat waves. For example, a short but extremely strong warm spell can dramatically reduce the quantity and quality of crop yields (IPCC 2007). On the other hand, a relatively cooler but prolonged heat wave can increase human discomfort, morbidity and mortality (D’Ippoliti et al. 2010; Anderson and Bell 2011). Moreover, long events that are usually connected with dry spells can increase water demand and forest fire risk (COPA-COGECA 2003).

In the past years, there have been few applications of regional climate models (RCMs) to project with greater spatial detail the effects of anthropogenic global warming in the EMME region by the end of this century. These studies were mainly based on the SRES emissions scenarios and run at grid box sizes between 25 and 50 km. Önol and Semazzi (2009) used the RegCM3 RCM to downscale the NASA fvGCM fields and calculated a 3–6 °C temperature increase for summer over the whole region. Zanis et al. (2009) used a multi-model approach with several RCMs from the PRUDENCE project and estimated that for the summer, the Greek peninsula might experience an average 4.5 °C warming. The MM5 mesoscale model driven by the global output of the NCAR CCSM3 model produced a relatively small (2–4 °C) June to August warming over Anatolia, and larger in Mesopotamia (6–8 °C) and Iran (8–10 °C), although these results were based on 5-year simulations at the start and end of the twenty-first century (Evans 2010). Similarly, Tolika et al. (2012), analyzing results from various models of the ENSEMBLES project for the greater region of Greece, suggest an increase in temperature by the end of the twenty-first century combined with an overall drying (30–60 %). This warming is found most pronounced during summer and is in the range of 3–6 °C depending on the scenario.

More recently, Lelieveld et al. (2012) applied an RCM at 25 × 25 km resolution over an EMME domain including the Balkans, Caucasus, Iran and the Arabian Peninsula based on the SRES A1B emissions scenario, as part of an overall climate assessment of the region. The temperature projections for 2,071–2,099 indicated a “continual, gradual and relatively strong warming of about 3.5–7 °C.” In a follow-up paper, Lelieveld et al. (2013) showed that heat extremes (as expressed by a warm spell duration index) in the capital cities of the EMME will rise dramatically, and in conjunction with projected future ozone pollution extremes, emphasized the multiple environmental stresses in these urban areas. Kostopoulou et al. (2014), using the same model output as the previous two studies, calculated temperature- and rainfall-related indices of extremes based on threshold values. They found that the periods of hot days (maximum temperature >35 °C) and tropical nights (minimum temperature >25 °C) per year could increase by 1 month in the northern part of the EMME domain and by 3 months in the southern part.

Utilizing the available RCM output used in the latter studies, we extend them by presenting a comprehensive climatology of heat wave characteristics in the EMME region for the end of the twenty-first century, as compared with the period 1961–1990. We identify heat waves with a percentile-based definition therefore accounting for the different local climatic conditions of the region and examine a range of properties to consider both intensity and persistence. We refine the heat stress aspect by combining the threshold-based indices of hot days and tropical nights. We consider the spatial distribution of the modeled heat waves as well as representative locations and associate them with distinct synoptic conditions. Finally, we explore changes in the heat wave characteristics under three SRES emission scenarios (Nakicenovic et al. 2000) to obtain an ensemble of potential outcomes for a 30-year period at the end of the twenty-first century when the warming signal is the most significant.

Data

Since the coarse information derived from global models is not sufficient for impact studies and in most of the cases not representative for the complex EMME terrain (Fig. 1), we used the PRECIS (HadRM3P) regional climate model, developed by the Hadley Centre (Jones et al. 2004) (PRECIS is “Providing Regional Climates for Impact Studies”). In our simulations, the model has a horizontal resolution of 0.22° (≈25 km) and is driven by the HadCM3Q0 or HadAM3P General Circulation Models (see Table 1). The selected SRES scenarios cover a range of projected greenhouse gas emissions (Table 1), from an optimistic (B2) and an intermediate (A1B) to a more pessimistic estimation (A2) (Nakicenovic et al. 2000). For this study, the A2- and B2-driven runs share the same control (1961–1990) simulation (see Table 1), thus all results referring to these simulations and this period are consistent.

Fig. 1
figure 1

Topography of the eastern Mediterranean and the Middle East region as represented in the PRECIS (HadRM3P) model. The black dots indicate the location of the four stations used for the validation of the heat wave indices (A Athens, S Sivas, E Eilat, K Kerman)

Full size image
Table 1 Description of the global simulations and emissions scenarios
Full size table

The comparison between studies under SRES scenarios and the most recent Representative Concentration Pathways (RCPs) might be challenging, yet some remarkably similar patterns of changes in the climate can be identified (Knutti and Sedlacek 2013). In their study, using a multi-model approach, they highlight the robustness in large-scale features of climate change between simulations under the A2 SRES and RCP8.5 scenarios. Rogelj et al. (2012) show similarities in the median of the temperature increase between projections under the B2 SRES and RCP6, while the median of the projected warming for the end of the century under the A1B scenario is found between the values of the RCP6 and RCP8.5 scenarios (see also Fig. 3 in Zachariadis and Hadjinicolaou 2014).

Daily maximum (TX) and minimum (TN) temperatures are used to identify summer heat extremes. Geopotential heights at the 500 hPa levels, derived from the model, are used to explore the synoptic conditions during heat waves. It must be noted though that the ability of regional climate models to represent large-scale atmospheric circulation relationships with climate extremes, especially precipitation, is limited and prone to substantial variability (Flocas et al. 2011). Extended evaluation of the PRECIS performance over the EMME region can be found in Lelieveld et al. (2012), where the model has been shown to be in general agreement with the CRU gridded observational dataset. Some cold biases during winter and slightly warmer conditions during the summer season were identified. In Lelieveld et al. (2013), an inter-comparison of nine regional climate models included in the ENSEMBLES project (www. ensembles-eu.org) indicates that the PRECIS projection, under the A1B scenario, for locations in the EMME, is somewhat warmer than the mean, but overall well within the ensemble range.

Daily ECA&D Footnote 1 (Klein Tank AMG et al. 2002) station data are used for testing the model results. Four stations were selected based on availability of time consistent datasets for the 1961–1990 period and the spread of the stations across EMME. These stations are Athens (Hellinikon) in Greece, Sivas in Turkey, Eilat in Israel and Kerman in Iran. The location of the stations is indicated in Fig. 1. A large part of the EMME, especially in the Middle East, may not be adequately covered from this selection of stations. This is due to the fact that consistent daily TX and TN observations in the region are not publicly available (or may not exist) throughout the period of study.

Methods

Definition of heat waves

For this regional study, it is preferred to not rely on a universal definition of heat waves. This is related to the complex orography, the distribution of land and major water bodies and the different climatic zones that are found in this atmospheric circulation crossroad between three continents. Therefore, we used a percentage-based definition. Following the WMO guidelines on the analysis of extremes (WMO 2009), we defined warm spells as the number of days in a period of at least 6 days where TX > TX 90th percentile (TX90). TX90 is calculated for a 5-day window centered on each calendar day of the control period (CTL 1961–1990). We considered as heat wave any warm spell that occurs in the extended summer season (May to September). A similar definition is used in Fischer and Schär (2010). For the future period (FTR: 2071–2099), we identified events according to the calculated thresholds for the control period. Hence, this represents the present-day understanding of heat waves.

Synoptic conditions

To identify the synoptic conditions related to extreme heat events, we created maps of the 500 hPa geopotential height averaged for the heat wave occurrence days and over the four selected ECA&D stations. Anticyclones that sustain heat waves can have a radius of more than 1,000 km; hence, these four stations are approximately representative of a domain of that magnitude. The Athens station characterizes conditions in Greece and the Balkan Peninsula, Sivas station represents the Asia Minor region, Eilat represents the conditions over the Levant and northern Arabian Peninsula, and Kerman signifies the southeastern most part of the EMME. The 500 hPa maps were created for both the CTL and FTR periods to investigate potential changes in synoptic conditions and try to connect circulation changes with heat wave characteristics.

Heat wave characteristics

Following Fischer and Schär (2010), we calculated a range of indices characterizing the persistence and severity of heat wave events. These indices are: the average annual frequency of days (HWF90) meeting the heat wave criterion described in “Definition of heat waves” section, the average number of heat waves per summer (HWN90), the average peak temperature of the hottest heat wave per summer (HWA90) and finally, the average duration of the longest heat wave per summer (HWD90). Years without heat waves were excluded from the HWA and HWD calculations. Discomfort during heat waves is to a large degree attributed to high nighttime temperatures. During the night, the body should recover from the heat stress of the preceding day. For non-air-conditioned households, high nighttime temperatures can be critical for the health, especially of vulnerable individuals. This category includes elderly people, infants, people exposed to outdoor activities and in general socially or physically isolated individuals. Hence, in addition to the aforementioned heat waves properties, we calculated the average number of combined hot days and tropical nights (CHT). These are considered summer days with a maximum temperature exceeding 35 °C and a minimum nighttime temperature higher than 25 °C. Often tropical nights are considered to be those with a minimum temperature over 20 °C. Since this part of the world is already warm and these nighttime temperatures are common during summer, we used the 25 °C threshold. Heat wave occurrences and all related indices are calculated for each of the three different scenarios.

Results and discussion

Comparison with observations

To explore how well the regional model represents the heat wave characteristics that were discussed in the methods section, a comparison with station data was performed. The four selected stations cover a range of climatic zones ranging from a desert climate (Eilat) to a Mediterranean (Athens) or more continental type (Sivas, Kerman). The results of this comparison are shown in Table 2.

Table 2 Heat wave indices comparison between four ECA&D stations and the PRECIS (HadRM3P) simulations for the control period (CTL: 1961–1990)
Full size table

In general, the PRECIS model seems to reproduce the heat wave properties for the CTL period and the selected stations reasonably well. As discussed in the data section, the A2- and B2-driven runs diverge only in the FTR period and so the CTL part of the simulation is in common. The frequency (HWF) of heat wave occurrence days per summer and the annual average number of events (HWN) are in the range of the observations. However, a general overestimation of these two indices is identified, particularly for the Kerman station. Likewise, PRECIS overestimates the duration of events (HWD) by 1–2 days in most cases.

Peak temperatures (HWA) are generally underestimated in the A1B CTL simulation for all stations except Eilat. The A2/B2 CTL run overestimates heat wave peak temperatures for Athens and Sivas, while showing an underestimation for Kerman station.

Finally, the combined hot days and tropical nights are reasonably well reproduced for all stations except Eilat. This difference between the observed number of CHT days probably results from the difference in elevation related to the resolution of the model. We assume that this cancels out in the comparison between CTL and FTR simulations. More detailed evaluation of the model for both mean climate variables and extremes can be found in Lelieveld et al. (2012, 2013).

Climate change in the EMME

Some general results from the climate change projections in the region are presented in Figs. 2 and 3. We focus on changes during the extended summer (May to September) under the three greenhouse gas emission scenarios discussed in the data “Data” section. Changes in mean temperature between the FTR and CTL periods are depicted in Fig. 2. By the end of the century, the entire EMME region will experience a summertime warming ranging from 3 to 9 °C depending on the emission scenario. The mean temperature increase is the largest for the A2 scenario (Fig. 2d) and the smallest for B2 (Fig. 2c), while the warming pattern of the ensemble mean (Fig. 2a) is very similar to the intermediate A1B scenario (Fig. 2b). According to our simulations, this warming is not uniform across the region, but is consistently the strongest in the northern part of the domain and the regions of Italy, Balkans and Anatolia.

Fig. 2
figure 2

Projected change in summer (May to September) mean temperature based on PRECIS (HadRM3P) model results, a averaged over the thee emission scenarios, b for the A1B, c B2 and d A2 scenarios

Full size image
Fig. 3
figure 3

Same as Fig. 2 for precipitation

Full size image

In general, much of the region will experience a drying varying on average from 5 to 30 % of the precipitation of the control period (Fig. 3a). The drying is the strongest for the A2 scenario (Fig. 3d) and is projected to locally reach up to 60 %. The increase in precipitation projected over the southeast part of EMME (i.e., Arabian Peninsula, Gulf States and Iran) is a common feature of climate change projections indicating northward expansion of moist tropical weather influences (Seidel et al. 2008; Evans 2010; Collins et al. 2013) and is consistent with observed rainfall trends in these areas (Tanarhte et al. 2012). These increases might be large in percentage; however, they are small in absolute amounts due to the low precipitation during the warm period, especially in the Middle East.

Synoptic conditions

As mentioned above, extreme heat events are usually connected with large-scale anticyclonic weather systems. Figure 4 depicts the mean synoptic conditions (500 hPa geopotential height) for heat wave occurrence days in four locations over the EMME region for the CTL period. Since the patterns for both CTL simulations (A1B and A2/B2) are very similar (not shown) their average is presented here.

Fig. 4
figure 4

Geopotential height of the 500 hPa level averaged for the heat wave occurrence days over. a Athens, b Sivas, c Eilat and d Kerman during the CTL (1961–1990) period (averaged for A1B and A2/B2 simulations)

Full size image

For all cases, heat waves are connected with high 500 hPa geopotential heights in the southern part of the region. For Athens (Fig. 4a), according to our simulations, heat waves typically occur when a ridge in approximately north–southwest direction is developed. This structure brings hot air from Africa into the region. Additionally, excess warming can be related to adiabatic heating connected with subsidence. The upper air circulation patterns during heat wave occurrences in Athens are in agreement with the atmospheric conditions during specific heat wave events case studies, described by Balafoutis and Makrogiannis (2001) and Founda and Giannakopoulos (2009). The heat wave pattern is similar for the Sivas station (Fig. 4b) while the ridge of high geopotential height is located more to the east. In these two cases, hot air advection is the key mechanism for the occurrence of heat waves. For the Eilat and Kerman stations (Fig. 4c, d, respectively), during heat wave days, the center of high geopotential heights is located directly overhead. Besides advection, strong adiabatic heating due to subsidence in the anticyclones center also contribute importantly. During the FTR period, the region is projected to experience enhanced anticyclonic conditions during summer, but the circulation patterns remain very similar to the CTL period (not shown).

Heat wave characteristics

The recent past (1961–1990) climatology of the HWF index for the two CTL simulations is presented in Fig. 5a, d. The mean annual frequency of days that meet the heat wave criteria in most areas does not exceed 4 or 5 days. Five HWF maxima over the EMME region are identified. They are located over south Iran, northern Saudi Arabia, eastern Turkey, Italy and the Balkans. The rest panels of the same figure depict the changes for the scenario period (2071–2099) for each of the three SRES emission scenarios and calculated using the control period threshold. The mean frequency of heat wave days per year will be dramatically increased in the whole EMME region. According to our simulations, these increases will range from approximately 20–130 days per year and will be the most pronounced over the Arabian Peninsula and the Gulf area. The increases in HWF index are stronger for the A2 simulation (Fig. 5b) and less striking for the B2 scenario (Fig. 5c).

Fig. 5
figure 5

Recent past climatology (a, d) and future changes in the heat wave frequency (HWF90) index, under A2 (b), B2 (c) and A1B (e) emission scenarios

Full size image

The index that describes the duration of summer extreme events is presented in Fig. 6. For the control period (Fig. 6a, d), according to the PRECIS simulations, heat waves in the region have an average duration that varies from 6 to 12 days. By the end of the century, these heat events will be several weeks up to months longer. The changes are again stronger over the southern part of our domain and the Arabian Peninsula, while their magnitude is consistently increasing from the optimistic B2 (Fig. 6c) to the intermediate A1B (Fig. 6e) and pessimistic A2 scenario (Fig. 6b).

Fig. 6
figure 6

Recent past climatology (a, d) and future changes in the heat wave duration (HWD90) index, under A2 (b), B2 (c) and A1B (e) emission scenarios

Full size image

The strong increases in the HWF and HWD indices over the southern part of the domain and especially the Arabian Peninsula are illustrated using the probability density function of TX over these regions and the percentage-based definition of heat waves we used. In the aforementioned area, the 90th quantile is closer to the median and mean values (Fig. 7—red curve). Therefore, under warmer climate conditions, it is more likely to exceed the heat wave definition threshold (see “Definition of heat waves” section). In contrast, in the northern part of the domain, although the summertime warming is relatively stronger (Fig. 2a), the 90th quantile is more distant from the median and mean TX values (Fig. 7—black curve). As a result, the heat wave threshold is less likely to be exceeded in the FTR period resulting in relatively smaller changes in the HWF and HWD indices over the northern EMME. Therefore, a larger projected change in heat wave properties over certain geographical areas may not only be a direct response to mean warming, but also be related to the kurtosis of the respective TX probability distribution.

Fig. 7
figure 7

Probability distribution of maximum temperature for the CTL (1961–1990) period and gird points, located in Saudi Arabia (red curve) and northern Balkans (black curve)

Full size image

Heat wave events in the EMME region are expected to become a more frequent weather phenomenon. Figure 8a, d depicts the average number of events per year (HWN) for the A2/B2 and A1B CTL simulations. For most of the domains, HWs are a phenomenon with a 2-year return period as indicated by the 0.4–0.6 events/year. In general, up to eight more occurrences per year are projected for the end of the century. The changes are generally larger for the northern part of the domain and are more pronounced under the A2 scenario (Fig. 8b). This index is connected with the duration index (HWD) and the kurtosis of the TX probability distribution (Fig. 7). According to our simulations, in the southern part of the domain, heat waves will occur in a few but very long events. On the other hand, in the northern part, heat waves are expected to be higher in number, but of relatively shorter duration compared with the southern part of the domain.

Fig. 8
figure 8

Recent past climatology (a, d) and future changes in the heat wave number (HWN90) index, under A2 (b), B2 (c) and A1B (e) emission scenarios

Full size image

The heat wave negative impacts in some societal sectors are proportional to the peak temperatures. The heat wave amplitude index (HWA) describes the average peak temperature of the hottest event per year. HWA for the control period is depicted in Fig. 9a, d. As expected, peak temperature follows the orography and also depends on the latitude. For the future scenario period FTR, the increases in peak temperatures are larger in the northern part of the domain and especially over the Balkan Peninsula, while the changes are the strongest under the A2 emission scenario (Fig. 9b). This is in agreement with the projected changes in summer temperatures over the region depicted in Fig. 2 and also identified in Lelieveld et al. (2012) and Zittis et al. (2014). During the summer season, the warming is not uniformly distributed and is stronger in the northernmost EMME. This can be explained by the combined relatively intense drying and warming over this part of the domain (Fig. 2b). In the already dry southern part, sensible heat fluxes prevail, and therefore, evaporative cooling is negligible both in the control and future periods. However, in the northern EMME evapotranspiration, being a key cooling mechanism, will be likely reduced under the projected drier conditions (Zittis et al. 2014). For some locations, especially in the northern part of the domain, the increase in heat wave peak temperatures is much stronger than the mean summer temperature change presented in Fig. 2a.

Fig. 9
figure 9

Recent past climatology (a, d) and future changes in the heat wave amplitude (HWN90) index, under A2 (b), B2 (c) and A1B (e) emission scenarios

Full size image

Combined hot days and tropical nights

The CHT index, which describes the combination of hot days and tropical nights, is presented in Fig. 10. In most of the EMME region, under recent conditions, the number of CHT days is limited to less than 2 weeks in total per summer or year (Fig. 10a, d). Higher CHT numbers are found in the Mesopotamia region and North Africa. These regions differ from the surrounding non-vegetated areas where radiative cooling during the night reduces minimum temperatures to less than 25 °C. This number is projected to dramatically change by the end of the twenty-first century. In the entire EMME region, CHT are expected to become more frequent. These increases are strongest in the southern part of the domain and under the A2 and A1B emission scenarios (Fig. 10b and e, respectively), the latter two being very similar, where up to 100 additional CHT days are projected. Exceptions are the mountainous areas including the Dinaric Alps, Taurus, Caucasus and Zagros mountains. Despite the projected warming, nighttime temperatures over the elevated regions will not often exceed 25 °C, and therefore, the increases in the number of CHT are relatively small. Besides possible adverse effects on human health, these changes can be of great importance for the energy sector. Since air-conditioning during nighttime will probably be an essential need, energy demand profiles are likely to change. One may expect both aspects to be impacted in different proportions depending on societal adaptation to the emerging heat conditions.

Fig. 10
figure 10

Recent past climatology (a, d) and future changes in the combined hot days and tropical nights (CHT) index, under A2 (b), B2 (c) and A1B (e) emission scenarios

Full size image

Conclusions

This study identifies and attributes climate warming-related changes in heat wave characteristics with the aim to support policy making in the EMME region and motivate mitigation measures and development of adaptation strategies. According to our projections, heat waves in the EMME are expected to change from events with a return period of about once every 2 years on average to a common phenomenon with multiple occurrences per year. All scenarios and indices that describe the frequency, persistence and severity of these extreme events are projected to increase strongly.

Heat waves, as defined by present-day standards, will have extraordinary duration of several weeks (under the optimistic B2 scenario) to months (under the A2 scenario). In addition, much hotter heat extremes are expected by the end of the century. Changes in severity of heat waves in terms of peak temperatures will probably exceed by far the projected mean summer temperatures increases and will be the most pronounced for the A2 and to a somewhat lesser extent for the A1B scenarios. Besides the general global warming, these changes in heat wave peaks can be attributed to the stronger summer anticyclonic conditions projected over the EMME, resulting in stronger advection of warm air masses from lower latitudes in the northern locations and more pronounced adiabatic heating due to increased subsidence in the southern locations. The number of tropical nights is also projected to increase strongly in the whole EMME region. This climate change feature will probably increase human discomfort and energy demand and may cause serious health impacts.

Changes in the discussed properties of heat waves will not be uniformly distributed over the domain. Alterations in frequency and duration-related indices are expected to be the most pronounced in the southern part of our domain and especially over the Arabian Peninsula. This is related to the percentage-based definition of heat waves we use in this study and the leptokurtic nature of the probability distribution of TX in this region. On the other hand, changes in heat wave characteristics related to number of events and peak temperatures will be stronger in the northern part of the EMME. Changes in the number of events are connected to changes in duration. The north–south gradient in peak temperatures is consistent with the changes in mean temperature and is related to the stronger drying over the northern part of the domain, which reduces summer evapotranspiration and the associated cooling.

The current fossil fuel-based energy production in the EMME region may become more intense in order to provide the essential space cooling and refrigeration needs. This would translate into more greenhouse gas emissions, leading to a vicious cycle with enhanced greenhouse warming and heat extremes. Switching to carbon poor and renewable energy sources, improving energy efficiency and insolation, and implementing urban cooling techniques in the region can contribute to mitigation and adaptation to climate change.

Since future summer conditions might be hotter than the hottest the region ever experienced, the timely design and implementation of adaptation strategies in sectors such as agriculture, water supply, energy production, food security, human health, tourism and biodiversity is highly recommended (see other studies in this Special Issue). In addition, side effects of these changes that may partly be devastating may include human displacement (UNHCR 2009), which has to be seriously considered.

While our modeling results derive from a set of three simulations, uncertainties related to initial and boundary conditions, physics of the model and greenhouse gas emission scenarios cannot be neglected. Since most of the region of interest is not included in the large European projects that have addressed these uncertainties previously, we had to base our study on the presented PRECIS simulations. With the use of two global models and three emission scenarios, we have only partly accounted for the driving model and emission uncertainties. Uncertainties of the regional climate modeling can be addressed by a multi-model approach. A way forward is the analysis of output from several RCM–GCM pairs from current and planned initiatives that consider the EMME region (e.g., the CORDEX experiment).Footnote 2