In this section the climatology and recent changes in observed summer heat waves during the periods 1961 and 2010 and 1981 and 2010 are investigated. As shown in Table 2, the absolute Georgia-average for HWN (the number of heat wave events) amounts to 1.7 eventsyr-1 and shows a significant increasing trend of 0.4 eventsdecade-1. HWD (the length of the longest yearly heat wave event) measures 5.5 daysyr-1 in the Georgia-average and a significant positive trend of 0.9 daysdecade-1 is found. The Georgia-average for HWF (yearly sum of participating heat wave days) measures 10.4 daysyear-1. Similar to HWN and HWD, no decreasing station trend could be observed throughout Georgia (not shown). The Georgia average trend amounts to 2.9 daysdecade-1 (Table 2). These findings are in accordance with Perkins et al. (2012) and Perkins and Alexander (2013), demonstrating that the high trend magnitudes for HWF drive increases in HWN and HWD, as the number of heat wave days represents an influencing factor in the calculation of event length and occurrence. For HWex (extreme heat wave days) an absolute mean of 0.3 days and a significantly increasing trend of 0.05 daysdecade-1 are observed.

As Fig. 3 shows all mean absolute differences for the heat wave occurrence, intensity and duration were found to be positive. For HWN an absolute increase between 0 and 1 event can be observed during the period 1981 and 2010 (with respect to the period 1961–1990). A few stations in the south and northwest of Georgia show an increase of 1 to 2 events. HWD shows an increase between 0 and 3 days for the major proportion of stations analyzed. For 5 of 22 stations the duration of heat waves increased to 3 to 6 days between 1981 and 2010. The heat wave aspect HWF shows an increase of 8 to 16 days for 6 of 22 stations analyzed during 1981 and 2010. For HWex an increase between 0 and 2 days can be found for the highest proportion of stations. For seven stations a strong increase by 3 to 6 extreme heat wave days is found.

Comparing the trends in the heat wave aspects HWN, HWD, HWF and HWex over Georgia for the two analysis periods 1961–2010 and 1981–2010, a pronounced increase in the magnitude of all trends is observable. Significant trends are found for HWN, HWD and HWF. As Table 2 shows trends for HWN and HWD (1981–2010) double the trend magnitudes measured for the period 1961–2010. A trend of 0.8 eventsdecade-1 for HWN and 1.8 daysdecade-1 for HWD could be found. For HWF (and HWex), the trend magnitudes of 6.6 daysdecade-1 (0.15 daysdecade-1) for 1981 and 2010 are even two to three times higher the trend magnitudes observed between 1961 and 2010.

In this section the climatology and spatio-temporal changes of the heat wave predictand and selected predictors for the analysis periods 1961 to 2010 and 1981 to 2010 are presented.

As Table 3 shows, the climatological mean for Tmean95p averaged over Georgia was measured 20.4 ∘C. The absolute increase comparing the periods 1961–1990 and 1981–2010 amounts to 0.1 ∘C. Trends for the heat wave predictand Tmean95p were observed to be positive. During the analysis period 1961–2010 an insignificant warming trend of 0.2 ∘Cdecade-1 (significant at the 10 % level) could be found over Georgia. A significant trend for Tmean95p was observed between 1981 and 2010, which measures 0.4 ∘Cdecade-1. As shown in Fig. 4a spatial changes in Tmean95p, large areas with an absolute temperature difference of more than 0.6 ∘C from Eastern Europe to the Ural Mountains are detected with peaks of up to 1 ∘C, although no significant difference could be found. Also for Turkey and eastern Georgia an insignificant increase of 0.2 to 0.6 ∘C could be observed. However, in western Georgia an insignificant increase in Tmean95p of at most 0.2 ∘C was found.

As presented in Table 3, the Caspian Sea shows a higher climatological mean for SST (21.6 ∘C) than the Black Sea (20.2 ∘C). For the Caspian Sea a significant SST trend magnitude of 0.7 ∘Cdecade-1 during the period 1981 to 2010 could be observed, while for the Black Sea a significant magnitude of 0.6 ∘Cdecade-1 was found. The spatial distribution of SST absolute differences shows significant warming in the North Atlantic, The Mediterranean, Black and Caspian Sea (Fig. 4b). Largest differences of significant warming over Eurasia are found in the North and Baltic Sea and in the Kara Sea (0.6–0.8 ∘C).

As shown in Table 3 the climatology for SLP measures 1012 mb with an absolute increase throughout Georgia of 1.2 mb. While a significant trend magnitude for the period 1961–2010 of 0.5 mbdecade-1 could be observed, no change between 1981 and 2010 was detected. Figure 4c shows large areas of significant positive trends in the eastern Mediterranean, Western Asia and Northern Africa. A maximum of significant absolute change in SLP is found in northeast Turkey and stretching throughout Georgia measuring an increase of up to 2 mb.

For Z500 a mean climatology of 5753 m was detected over Georgia and an absolute increase of 15.9 m with respect to the period 1961–1990 (Table 3). A significant positive trend for Z500 of 8.7 mdecade-1 could be found during the period 1961 and 2010. However, the trend magnitude for the period 1981–2010 was smaller (2.3 mdecade-1) and insignificant. As shown in Fig. 4d, throughout southern Eurasia a significant absolute increase of Z500 of up to 20 m is observed, which corresponds to observations for Western Asia by Kuglitsch et al. (2010). These findings highlight a changing atmospheric circulation over Georgia, implying an increase in warm air advection from the Southwest, large-scale subsidence and surface heating, which might have enhanced the formation and persistence of heat waves in recent years. Moreover, the increase in Z500 is usually connected with the increase in stability and the inhibition of convection over the area.

For u-wind an absolute decrease of 0.6 ms-1 is observed comparing the periods 1961–1990 and 1981–2010. Georgia-averaged climatological mean measures 6.8 ms-1 (Table 3). Trends for the zonal wind are insignificant during both analyzing periods. The climatology for v-winds measures 3.7 ms-1 with an absolute decrease throughout Georgia of -0.5 ms-1. While a significant trend magnitude for the period 1961–2010 of -0.2 ms-1decade-1 could be observed, an insignificant change of 0.0 ms-1decade-1 between 1981 and 2010 was detected.

As shown in Fig. 5a OLR measures a mean climatological value of 246 Wm-2 and an absolute increase of 5.8 Wm-2, implying a decrease in cloudiness and an increase in maximum insolation. During the period 1961–2010 a significant positive trend of 2.9 Wm-2decade-1 could be observed. As all other predictors OLR shows an increase in the trend rate during 1981 and 2010 compared to the period 1961–2010 (3.1 Wm-2decade-1) (Table 3). The spatial distribution of OLR trends shows an increase throughout Western Asia with the highest absolute increase of up to 12 Wm-2 located across the Caspian Sea (including Eastern Georgia). Vertical Velocity at 500 mb measures an absolute increase of 0.01 Pas-1 comparing the periods 1961–1990 and 1981–2010 averaged over Georgia. The trend magnitude during the period 1961 to 2010 measures a significant increase of 0.01 Padecade-1 over Georgia. The trend for the period 1981–2010 shows a higher magnitude of 0.02 Padecade-1, but is insignificant. Regarding the spatial distribution of absolute changes during 1961 and 2010 a significant positive difference of up to 0.015 Pas-1 is located in Azerbaijan and northern Iran (Fig. 5b), suggesting an increase in atmospheric stability and subsidence. However, negative but mainly insignificant areas are located across the eastern Mediterranean and Black Sea as well as in the east of the Red and Caspian Sea and the Persian Gulf, implying an increase in instability and convergence. The location of these areas resembles the centers of zero trends in the OLR trend map (Fig. 5a), supporting the implication of local decreases in clear skies and insolation. Georgia shows a maximum absolute change of around 0.01 Pas-1 in the eastern part decreasing towards the western coast until approximately zero. As presented in Table 3 for RH500 a climatological value of 44 % could be observed. An absolute decrease of -4.8 % was found and a significant trend of -2.4 %decade-1 was measured during 1961 and 2010. The spatial distribution of absolute changes shows similar patterns. RH500 shows significant decreases of up to -0.06 % across the Caspian Sea (Fig. 5c) with lowest values in the eastern part of Georgia and higher insignificant differences of around -0.04 % to the West of Georgia. The reduction of air humidity is also found by Kuglitsch et al. (2010) detecting decreasing RH500 over West Asia accompanied by an increase in minimum and maximum temperature, Z500 and SLP.

For RR an absolute decrease of -0.7 mmday-1 observed comparing the periods 1961–1990 and 1981–2010. Georgia-averaged climatological mean for RR measures 5 mmday-1 (Table 3). Trends for RR are significantly decreasing by -0.35 mmday-1decade-1 regarding the period 1961 and 2010. The negative trend magnitude of -0.58 mmday-1decade-1 for the period 1981 to 2010 was even larger, but insignificant (Table 3). As implied by Fig. 5d, absolute differences in the spatial distribution of total precipitation shows a large area of significant decreasing rates by -1.2 mmday-1 predominantly in the east of Turkey, southern Georgia, Armenia, Azerbaijan and northern Iran. Similar to the RR, SM shows decreasing trends over the last five decades. As Table 3 shows, the climatological mean averaged over Georgia was measured 0.3 % and the absolute decrease amounts to -0.01 % comparing the periods 1961–1990 and 1981–2010. Trends for the heat wave predictor SM were found to be negative. During the analysis period 1961–2010 a significant decreasing trend of -0.004 %decade-1 could be found over Georgia. However, the trend magnitude for the period 1981–2010 was even larger (-0.007 %decade-1), but insignificant. Regarding the spatial distribution of SM change, significant decreasing rates are found in central and eastern Turkey, northern Iran and the Southern Caucasus. Towards northwestern Georgia the trend measures approximately zero. The spatial location of high reduction rates of RR, SM and RH500 have a strong resemblance of those measuring a high increase in SLP, Z500, OLR and O500, which supports the assumption, that the increase in surface heating, subsidence and warm air advection drive the reduction in air humidity, total precipitation and soil moisture.

As shown in Fig. 6a, western and central Europe and western Asia are dominated by negative SAT fields, associated with the surface cyclones over the area. However, Eastern Europe, the Mediterranean and Middle East as well as from Northern Africa to Western Russia warm centers are observed during major heat wave events over Georgia. Two centers of high SAT anomalies of up to 5 ∘C were found in Armenia and in the southern Ural Mountains. The SAT anomaly over Georgia stretches from 2.5 ∘C in the Northwest up to 5 ∘C in the Southeast, implying strongest heat-health impacts here. The centers of negative SAT anomalies with a maximum of -3 ∘C lie southeasterly and southwesterly of the positive heat anomalies below the middle level cyclones (see Fig. 7b).

SST anomalies suggest a close relationship between large-scale dynamic patterns as shown in Fig. 6b. While the Northern Atlantic and Eastern Mediterranean Sea show negative SST anomalies of up to -0.6 ∘C, positive SST anomalies are observed in the Caspian Sea of up to 0.7 ∘C. Lower positive SST anomalies can be found in the Eastern Black Sea (0.5 ∘C). Highest anomalies are found in the Barents and Kara Sea of up to 1 ∘C during heat wave events over Georgia. The SST pattern for the first CCA mode over Eurasia demonstrates a very close resemblance to the SST composite for observed heat wave events and accounts for approximately 52.3 % of the summer heat wave variability over Georgia (Figs. 6b and 9g). The squared correlation between the summer SST and heat wave pattern is 0.56. Although a close relationship between SSTs and heat waves could be observed, it does not necessarily mean that the SST anomalies are responsible for heat waves over Georgia. It is far more likely that SST anomalies are an accompanied phenomenon of extreme temperature events. Nevertheless, the possibility that an individual heat wave might be affected by local SST anomalies cannot be excluded. Composite and CCA fields have a strong resemblance with the SAT and Z500 fields suggesting enhanced local air advection, subsidence and maximum insolation over strong SST warming fields during heat wave events, which is in accordance with Feudale and Shukla (2011a). Moreover, the reduction of meridional winds over sea, mainly apparent over the southern Caspian Sea, prevents the generation of lee waves and cyclones (Fig. 7d). Hence, the cooling effect by wind-induced mixing is reduced, which further warms up the SST (Buzzi and Tibaldi, 1978). The simultaneous SST anomalies in the Caspian, Northern Barents and Kara Sea can be explained by a reduction of baroclinic instability between the Southern Caucasus and the Northern Barents and Kara Sea, as discussed by Feudale and Shukla (2011a) for the simultaneous SST anomalies in the Mediterranean and the Northern Atlantic during the heat wave 2003. The prevention of baroclinicity is caused by a diminishing land–sea temperature gradient, which is reflected by the negative anomalies of v-winds over the Northern Barents Sea and West Asia stretching from Northeastern Africa across the Southern Caspian Sea to the Kara Sea (Fig. 7d) resulting in a northward shift and intensification of the subtropical jet and a northward shift of the African ITCZ.

As shown in Fig. 7a, a deep surface trough is found over Southern Scandinavia extending across the Black Sea area towards the Persian Gulf and the Red Sea. The persistent surface lows are known as the “Persian Gulf trough” and “the Red Sea trough”, which govern a strong relation to heat waves and heavy precipitation in the Mediterranean (e.g. Ziv et al., 2004; de Vries et al., 2013). At the same time, pronounced anticyclones are located over the Ural Mountains, the Mediterranean Sea, North Africa and Iran, implying warm air attraction from the Mediterranean and Middle East to Georgia (Fig. 7a). The SLP composite and CCA mode over Eurasia show similar patterns during heat wave events (Figs. 7a and 9a). The first CCA mode captures 61.2 % of the summer heat wave variability. The squared correlation between the summer SLP and heat wave predictand amounts to 0.62. The observed heat wave patterns have a strong resemblance to those found for heat waves in Western Asia by Kuglitsch et al. (2010).

Composite and CCA patterns for Z500 demonstrate a large anticyclonic vortex with anomaly maxima of 90 m during major heat wave events. Negative anomalies of up to -60 m are located westerly over the British Isles, Southern Scandinavia and easterly over central Asia and central Russia. Over eastern Georgia positive anomalies of around 60 m are found, while towards the western coast the intensity decreases to 40 m. The first CCA mode shows a close resemblance to the composite and captures 61.6 % of the summer heat wave variability (Figs. 7b and 9b). The squared correlation between the summer Z500 and the heat wave predictand measures to 0.74. The location of the middle tropospheric troughs over Scandinavia and West Asia are in southeasterly and southwesterly location aside of the anticyclonic center resembling a typical west-oriented omega high, a nearly-stationary anticyclonic pressure field closely associated to high-impact heat waves (Fig. 7b). The observed pattern can be referred to the “RU (Russian) cluster” identified as one of six heat wave blocking patterns identified by Stefanon et al. (2012) or to the “Eurasia” region, one of three regions, in which the frequency of the extremely hot days per month homogeneously varies (Carril et al., 2008). Due to its location it is known as the Ural Blocking High (UBH) pattern and is associated with the 2010 Russian heatwave investigated by Barriopedro et al. (2011), Dole et al. (2011) and Grumm (2011). The blocking of westerlies is caused by a reduction of baroclinicity, which is typical for the mid-latitudes. In general, baroclinic instability enhances the blocking persistence, the interruption of the mid-latitude westerlies, the deflection of the west–east storm tracks, large-scale subsidence and incoming solar radiation. Baroclinicity usually limits the northern branch of the Hadley cell expanding to the North. Reduced baroclinic activity during heat waves leads to a northward shift of the descending branch of the Hadley cell and the African ITCZ, which is consistent with studies on heat waves over Eurasia by Cassou et al. (2005) and Carril et al. (2008). As shown in Fig. 8b, the precipitation composite over the central Sahel implies a northerly shift of the African ITCZ. This relation was first investigated by Rowell (2003), which observed an increase of rainfall over the western African Sahel and a northern shift of the African ITCZ during warm Mediterranean SST.

The u-wind composite and CCA mode is consistent with reduced baroclinicity and shows reduced zonal wind activity throughout central Asia of up to -9 ms-1. An enhanced zonal wind flow is deflected by the anticyclonic center, implying a northward shift of the subtropical jet. Regarding Georgia, reduced wind speeds of -4 ms-1 are found in the eastern part and increase towards 1 ms-1 in the West, implying stronger influence of the anticyclonic blocking and associated subsidence in the Western part of Georgia (Fig. 7c). V-winds at 500 mb intensify over Western Russia by up to 9 ms-1, proving warm and dry air masses are attracted from south-southwest during major heat wave events. Negative anomalies are found from Northeastern Africa across the Southern Caspian Sea to the Kara Sea, implying a reduced meridional gradient between continental central Asia and the Northern Barents and Kara Sea (Fig. 7d). In western Georgia positive anomalies of over 7 ms-1 can be observed, while anomalies in the southeast of approximately zero can be found, implying a stronger relation of warm air advection and heat wave events are found in the West. The u- and v-wind composites and CCA modes show a strong resemblance (Figs. 7c, d and 8c, d). The squared correlation between the summer zonal wind and the heat wave predictand amounts to -0.77. The first CCA mode for u-wind captures 59.4 %, whereas the CCA pattern for v-wind accounts for 62.4 % of the summer heat wave variability (Fig. 9c and d). The squared correlation between the summer meridional wind and the heat wave predictand measures 0.76.

The analysis of vertical pressure velocity at 500 hPa reveals two opposing patterns affecting Eastern and Western Georgia differently (Figs. 7e and 9e). The first O500 CCA mode captures 55.5 % of the summer heat wave variability. The correlation between the summer O500 pattern and the heat wave predictand amounts to 0.81, implying a strong relation between subsidence patterns and the heat wave occurrence in Eastern Georgia. Similar to the CCA pattern, the composite detects intensified O500 of up to 0.7 Pas-1 located over the Caspian Sea and central Asia associated with local stability, subsidence, clear skies and maximum insolation during major heat wave events. Throughout Western Russia and the Barents Sea positive, but weaker anomalies are observed. These patterns mainly influence heat wave patterns over eastern Georgia, whereas a band of negative O500 anomalies (correlation fields) is found across the eastern Black Sea and West Georgia. Reduced vertical pressure velocity over Southern Scandinavia and central Asia as well as those stretching from northeast Africa to southwestern Russia are characterized by instability, convection and cloudiness, reflected by the decrease in enhanced surface heating and precipitation (Fig. 8a and c). High SATs combined with high SSTs lead to increased latent heat fluxes (Xoplaki et al., 2003a) and increased air humidity, hence convection and precipitation.

As presented in Fig. 7f, composite anomalies for relative humidity at 500 mb during major heat wave events show lowest values of up to -20 % over central Asia, eastern and southeastern of the Caspian Sea. Negative fields are also found northeastern of the Black Sea, over central Russia and the Barents and Kara Sea. The observed heat wave patterns have a strong resemblance to climate composites of major heat wave events in the eastern Mediterranean found by Kuglitsch et al. (2010). They also show a close relation to SLP, O500 and OLR patterns (Figs. 7a, e and 8a). Because of the atmospheric stability, vertical motion is restricted to the lower boundary layer, where solar radiation is maximized and the warm air is trapped. The spatial distribution of RH500 CCA fields shows similar patterns (Fig. 9f). The first CCA for RH500 mode captures 52.1 % of the summer heat wave variability. The squared correlation between the summer fields for RH500 and the heat wave predictand over Georgia amounts to 0.78.

Figure 7g shows a vector wind composite during heat wave events and provides evidence, that anomalously strong, warm and dry winds collect moist over the warm eastern Mediterranean and Black Sea producing cyclones and strong convective rainfall events above western Georgia (Fig. 8c). While western Georgia is dominated by moist air masses, eastern Georgia is influenced by warm and rather dry winds from North Africa and the Middle East of much lower wind speed anomalies. Moreover, the Surami Mountain chain dividing western and eastern Transcaucasia holds moist air masses and leads to adiabatic heating by foehns in on the luv side bringing dry air to the East.

High Outgoing Longwave Radiation rates are usually well correlated with high surface temperature, low soil moisture and water vapor concentration in the planetary boundary layer (Fischer et al., 2007). Figure 8a shows highest OLR anomalies of up to 28 Wm-2 east of the Caspian Sea during major heat wave events. The spatial distribution of high OLR anomalies corresponds well with areas of high SLP, Z500 and O500 anomalies (Fig. 7a, b and e) as well as low anomalies of relative humidity and zonal winds at 500 mb (Fig. 7c and f), implying enhanced subsidence, clear skies and maximum radiation leading to surface heating. Over eastern Georgia positive anomalies of 8 Wm-2 can be observed, while west Georgia shows negative OLR anomalies of -8 Wm-2, which suggests enhanced cloudiness and atmospheric instability.

Heat waves are usually connected to adjacent rainfall and soil moisture deficits, which is in agreement with Ferranti and Viterbo (2006), Fischer et al. (2007) and Seneviratne (2010). According to precipitation and soil moisture composites illustrating the 20 days prior to the detected major heat wave days, adjacent soil dryness dominates the whole territory of Georgia (not shown). Precipitation patterns during heat wave events are highly influenced by both, the large-scale circulation and local orographic patterns. Large-scale precipitation composite (correlation) fields over Eurasia correspond well with SLP, Z500 and O500 patterns (Figs. 7a, b, e and 9a, b, e). Local precipitation fields are mainly found over mountainous and coastal areas due to moist air advection, instability and convection. Soil moisture fields resemble well the SAT and RR composite and CCA patterns and show negative composite (correlation) fields over Turkey across the Southern Caucasus to Southern Russia (Figs. 6a, 8c, and 9i). Over eastern Georgia simultaneous negative precipitation (-2 mm) and SM fraction anomalies (-0.04) are observed, which supports the evidence of warm and dry air advection, subsidence and surface heating. At the same time, positive RR anomalies of 0.15 mm and negative SM fraction anomalies of -0.03 are observed over West Georgia. Similar to the composite, the CCA pattern detects reduced precipitation over eastern Georgia. The first precipitation and soil moisture CCA mode captures 51.2 % (50.0 %) of the summer heat wave variability (Fig. 9i and j). The correlation between the summer precipitation (soil moisture) patterns and the heat wave predictand amounts to -0.81 (-0.76), implying a strong relation between precipitation and soil moisture deficits during heat wave days over Eastern Georgia. This finding is in accordance with Haarsma et al. (2009), who found that summer SAT warming is enhanced due to soil moisture depletion which limits the cooling of the land surface by evaporation. However, as shown in Figs. 7g, 8e and 9i the anomalous southwesterly subtropical winds bring moist air masses to the high mountain ridges of the Caucasus Mountains in Western Georgia, leading to local precipitation maxima at luv sides due to enhanced orographic convection (Kostianoy and Kosarev, 2008). Moreover, soil dryness during heat wave events in mountainous and coastal areas drive instability and air humidity due to meso-scale circulations leading to enhanced precipitation and cooling (Stefanon et al., 2013). This effect could amplify precipitation events and further reduce temperature in mountainous and coastal areas during heat wave events over Georgia.