M. Türkeş, Department of Geography, Faculty of Sciences and Arts, Physical Geography Division, Terzioğlu Campus, Çanakkale Onsekiz Mart University, 17020 Çanakkale, Turkey. E-mail: email@example.com
The global mean surface temperature has increased rapidly and considerably over the last century due to the increase in the anthropogenic greenhouse gas emissions (IPCC, 2007). One of the major concerns related to increasing air temperatures is that it enhances the probability of the upper extremes of the temperature distribution. As a result, more recent analysis has focused on changes in frequency, duration and severity in extreme temperatures as critical factors on human health, air quality, the vegetation productivity, the phenology, hydrological systems and water resources/supplies problems as well as the occurrence of forest fires.
Recent surface temperature variability in most European regions and the Mediterranean Basin has been largely investigated (Kostopoulou and Jones, 2005; Moberg et al., 2006; Xoplaki et al., 2006; Diffenbaugh et al., 2007; Solomon et al., 2007; Kuglitsch et al., 2010; Rodríguez-Puebla et al., 2010 and so on). Many analyses of temperatures over long time series showed that cold extremes have been decreasing and warm extremes increasing during the last quarter of the 20th century in many parts of Europe and the Mediterranean region (Klein Tank and Können, 2003; Xoplaki et al., 2003a; Kuglitsch et al., 2010). Xoplaki et al. (2003a) detected that the summer (June–September) air temperatures over Greece and western Turkey were characterized by warm 1950s and 1990s (the warmest summer was in 1999) and rather cool 1960s, 1970s and early 1980s (the coolest summer in 1976). Rahimzadeh et al. (2009) has found increasing trends for summer-days, warm-days and tropical nights at most stations of Iran since 1970s. They also reported that increasing trends in the number of summer-days are statistically significant at the 0.05 level for half of the stations.
Several studies have also concluded that summer temperature variability and extreme temperatures in Europe and the Mediterranean Basin are closely tied to the North Atlantic atmospheric circulation patterns and SST variations over the Atlantic Ocean (Xoplaki et al., 2003b; Della-Marta et al., 2007; Folland et al., 2009). For instance, the most important summer warming pattern has been associated with blocking conditions, or cool summers have been attributed to the increase of northerly meridional circulation over the Eastern Mediterranean (Kutiel and Maheras, 1998; Xoplaki et al., 2003b). Some studies suggested a strong linkage between anomalous summer temperatures and the Atlantic Multi-Decadal Oscillation (AMO) and the North Atlantic Oscillation (NAO) (Della-Marta et al., 2007; Folland et al.2009). Hertig et al. (2010) have shown that extreme minimum and maximum temperatures exhibit an opposite behaviour between the Eastern and the Western Mediterranean basins, pointing to the concept of the Mediterranean Oscillation. However, some studies have emphasized other processes like soil moisture–temperature feedback mechanisms and urbanization leading to warmer conditions over the Mediterranean Basin (Türkeş and Sümer, 2004; Nastos and Matzarakis, 2007; Vautard et al., 2007; Wang et al., 2011).
Türkeş et al. (2002) analysed the trends and changes in the mean, maximum and minimum surface air temperatures of the 70 stations in Turkey for the period 1929–1999. Their study identified the general increasing trends in annual, winter and spring mean air temperatures particularly over the southern regions of Turkey, and the decreasing trends for summer and particularly autumn mean temperatures over the continental central and northern regions of Turkey. Türkeş et al. (2002) also indicated that the summer nighttime warming rates were generally larger than the nighttime warming in spring and autumn air temperatures, and the nighttime warming rates of spring and summer were generally stronger than those in the spring and summer daytime air temperatures. However, they did not investigate the trends and changes in the numbers of extreme warm/hot days or of tropical and summer-days in Turkey. Erlat and Yavaşlı (2009) examined the variation and trends in the annual numbers of tropical and summer-days at 10 stations of the Turkish Aegean region for the years between 1939 and 2008. They detected that the annual number of the summer and tropical days revealed weak increasing or decreasing trends between 1939 and first half of 1970s, but statistically significant increasing trend in all stations since the second half of 1970s. Especially the positive anomaly values that can be observed in all stations between the years 1998 and 2008 are remarkable. This showed that the daytime temperatures are increasing and thermal conditions are getting hotter in the Aegean region during the warm periods of the year. Kuglitsch et al. (2010) have used maximum (TX) and minimum (TN) temperature series of 246 stations across the Eastern Mediterranean region for determining changes in heat wave number (HWN95), length (HWL95) and intensity (HWI95). Hot summer daytime (TX95perc) and nighttime temperature (TN95perc), averaged over the whole area, have increased by 0.38 ± 0.04 °C/decade and 0.30 ± 0.02 °C/decade, respectively since the 1960s. The series of HWI95, HWL95 and HWN95 across the Eastern Mediterranean region have increased by a factor of 7.6 ± 1.3, 7.5 ± 1.3 and 6.2 ± 1.1, respectively. Hot spots of heat wave changes are identified along the Eastern parts of the Black Sea coastline of Turkey, in western, southwestern and central Turkey, and across the western Balkans.
However, long-term variability and trends in the annual numbers of summer and tropical days in Turkey as a whole have not been examined yet. Consequently, the main aims of the present study are listed as: (i) to investigate the climatology of the annual numbers of the summer and tropical days, and (ii) to detect the nature and magnitude of the long-term variability and trends in the annual numbers of summer and tropical days at the 97 major climatological and synoptic meteorological stations of Turkey by using nonlinear (monotonic) and linear trend detection tests for the period 1950–2010.
2. Data and methodology
The series of daily maximum air temperature observations used in this study were recorded at major climatological and synoptic meteorological stations of the Turkish State Meteorological Service. Essential quality and homogeneity controls of air temperature data set used in the present study were checked by Türkeş et al. (2002) for 70 stations in Turkey for the period 1929–1999. Information on the homogeneity and other time-series characteristics of Turkish air temperature data can be found in Türkeş et al. (2002) and Türkeş and Erlat (2008), respectively. We have updated the data set for the years 2000–2010 and added new stations by eliminating those series in case station's location was changed, and only those series that turned out to be 95% complete over the warm season of the year. The years before 1950 were not taken into account, because many daily observations are missing, which would cause altering the homogeneity of the set of observations. After having applied this procedure to 27 stations chosen for the analysis, the study contained continuous daily temperature data available at least 40 years of data per station. Therefore, the daily maximum air temperature series of 97 stations were analysed for this study. Sixty-one stations cover the period 1950–2010. Thirty well-distributed station records cover more than 50 years and six stations are available for a shorter period of 1968–2010. Figure 1 shows the geographical distribution of the 97 climatological and meteorological stations used in the study in Turkey.
Several climate extreme indices have been suggested to provide a uniform perspective on observed changes in climate extremes by the joint Working Group on Climate Change Detection of the World Meteorological Organization-Commission for Climatology (WMO-CCL) and the Research Programme on Climate Variability and Predictability (Peterson et al., 2001). Some of the extreme indices are based on percentiles such as the warm days (TX90) and cold nights (TN10), whereas some indices refer to counts of days crossing a threshold such as the numbers of frost days and summer-days (Peterson et al., 2001; Frich et al., 2002; Alexander et al., 2006; Moberg et al., 2006). In this paper, in order to focus on changes and trends in the extremes of warm and hot air temperatures, daily maximum air temperature series at each station were analysed as summer-days (the annual number of days with maximum temperature ≥ 25 °C) and tropical days (the annual number of days with maximum temperature ≥ 30 °C), respectively.
The coefficient of variation (CV) was used to investigate the spatial distribution pattern of year-to-year variability in the annual number of frost days at the 97 climatological and meteorological stations of Turkey. The statistic of CV is computed by expressing long-term standard deviation as a percentage of long-term average of the annual numbers of both summer and tropical days. The CV values would give a general indication of the probable percentage variation around the long-term average summer and tropical days at the stations. Thus, relatively less dispersed variables would have the lower CVs.
The nonparametric Mann–Kendall (M–K) rank correlation test (WMO, 1966) was used to detect any possible trend in the annual number of both summer and tropical days, and to test whether such trends are statistically significant or not. Before applying the test, original observations of xi were replaced by their corresponding ranks ki, such that each term is assigned a number ranging from 1 to N reflecting its magnitude relative to magnitudes of all other terms (ni). We continue this procedure of counting for each term of the series ending kN−1 and its corresponding number nN−1. Then the P statistic is computed as follows:
M–K rank correlation statistic τ is derived from N and P by the following equation:
Distribution function of τ is the Gaussian normal for all N larger than about 10, with an expected value of 0 and variance (τvar) equal to
The significance test (τ)t is then written as,
where tg is the desired probability point of the normal distribution with a two-sided test, which is equal to 1.960 and 2.58 for the 5 and 1 % levels of significance, respectively. Using two-sided test of the normal distribution, null hypothesis of absence of any trend in the series is rejected for the large values of |(τ)t| for the desired level of significance.
The least squares linear regression (linear regression, shortly) equations were also calculated to detect the trends rates (in° C/decade) in the summer and tropical day series, with time as the independent variable and the values of summer and tropical days as the dependent variable. The statistical significance of each estimated β coefficient was tested using the Student's t test with (n − 2) degrees of freedom (Türkeş et al., 2002). In using two-tailed test of the Student's t distribution, the null hypothesis for absence of any linear trend in time-series is rejected for large values of |t|.
3. Results of the analysis
3.1. Spatial distributions of average conditions
In general, Turkey is mainly characterized with a dry/warm summer and cold/rainy winter subtropical climate, or shortly with the Mediterranean macro climate (Türkeş, 2010, 2011). The geographical pattern of air temperatures over Turkey is controlled by air mass movement and circulation patterns. Generally, warm stable weather conditions are established, when the Azores anticyclone intensifies and extend towards the Eastern Mediterranean. This development takes place in late spring and summer, and normally ends by the late October, producing warm and stable weather in most of Turkey except the Black Sea coastal region and northeastern Anatolia sub-region. By mid-June, the Eastern Mediterranean basin is dominated by the Azores subtropical anticyclone from the mid-northeast Atlantic and a surface low-pressure trough that extends from the Asian monsoon through the Persian Gulf (Persian trough) at the lower atmospheric levels. At the upper level, subsidence over the Eastern Mediterranean during the summer season may be considered as part of the subtropical descending branch of the global Hadley cell, which shifts northward in summer (Barry and Chorley, 1998; Xoplaki et al., 2003b; Jones et al., 2006; Trigo et al., 2006; Türkeş, 2011). The anticyclonic character of large-scale circulation encouraging subsidence may cause high temperatures at any time of the period between March and October in Turkey. Especially from the beginning of summer conditions characterized mainly with continental tropical air masses, air temperatures rise in all geographical regions of Turkey, and regional contrasts seen in winter decrease.
Distribution of high temperatures over Turkey is closely related with not only atmospheric circulation pattern and weather systems, but also physical geographical features of the country including land–sea interactions, altitude and topographical effects. The highest average annual number of summer-days (ANSDs) is observed along the Mediterranean coast to the İzmir station on the Aegean Sea coast, reaching values 150 d, while these increases exceed 170 d in the southeastern Anatolia region and the Mediterranean coasts of Turkey. On average conditions, Adana (195.1) and Cizre (189.4) are the warmest two stations of Turkey with respect to the number of the ANSDs. Average ANSDs vary between 130 and 110 d over the coastal parts of the Marmara region and interior part of the Anatolia. ANSDs with less than 70 d are seen along the middle and Eastern Black Sea coastal belts. These number drops below 50 d at the stations in the northeastern Anatolia sub-region, such as in Ardahan (35.5) and Kars (51.8), where altitude is mostly over 1000 m (Figure 2(a)).
Spatial distribution patterns of interannual variability of summer and tropical days were investigated by CV calculated for the 97 climatological and meteorological stations of Turkey. CVs of the annual numbers of summer-days reveal the increasing variability from southwest to northeast. Spatial distribution of CV rates over Turkey range from 4.9% at Nazilli station in the Aegean region to 35.9% at Ardahan station in the northeastern part of Anatolia (Figure 2(b)).
The pattern of the annual number of tropical days (ANTDs) frequency closely resembles annual number of summer-days pattern (Figure 3(a)). ANTDs exceed 150 d in the southeastern Anatolia region due to the influences of the southerly tropical circulations and dry continental air masses in the warm season. It may be speculated that the low moisture and dryness play an important role in the intensification of high temperatures over this region. In the Marmara, Central Anatolia and East Anatolia regions of Turkey, ANTDs are found between 50 and 30 d. The number of tropical days reveal a sharp decrease toward northeastern and the Eastern Black Sea sub-regions of the Anatolian Peninsula. The lowest number is seen over the Eastern Black Sea coasts and northeastern sub-regions of Turkey, which typically have 5 d of tropical conditions per year; for example, Sinop 2.8 d, Ardahan 2.9 d, and so on.
Year-to-year variability in the annual number of tropical days increases from the coastal and southern regions to northeastern regions of the country. Lowest CV is prevalent over the southwestern Anatolia and southeastern Anatolia regions of Turkey, whereas high interannual variability dominates over the Eastern part of the Black Sea region and northeastern sub-region of the Anatolian Peninsula (Figure 3(b)).
3.2. Trends in annual number of summer and tropical days
The annual numbers of summer and tropical days and their anomalies (departure from the average of the period 1961–1990 or 1961–1990 normal) were calculated for each of the 97 stations used in the study. Mann–Kendall rank correlation test statistics showed that annual numbers of summer and tropical days are mostly characterized with a general increasing trend over Turkey for the entire period of 61 years.
The analysis suggests an upward trend in the annual number of summer (tropical) days at 92 (91) of the 97 stations over the 61 year study period from 1950 to 2010. For the study period, statistically significant increasing trends for summer (tropical) days are detected at 64 (71) stations, 51 (58) of which are at the 0.01 significance level. However, insignificant cooling tendencies that were found for summer (tropical) days are detected at five (six) stations. M–K rank correlation coefficients calculated for the annual numbers of summer-days' time-series range between 0.10 and 0.53 and of tropical days' time-series are between 0.10 and 0.57 over the period 1950–2010. Coherent regions with a significant warming appear mainly over the Eastern part of Black Sea region (Figure 4(a) and (b)).
The significant increasing trend in annual number of summer and tropical days in Turkey did not occur in a monotonic way. Therefore, the entire 1950–2010 period has been divided into two sub-periods, namely 1950–1975 and 1976–2010. On the basis of M–K results, decreasing tendencies can be detected for 1950–1975 sub-period both for regionally averaged ANSDs and ANTDs (Figures 5(a) and 6(a)). The decadal trend rates calculated for the first sub-periods of these regional average series are − 2.5 d dec−1 for ANSDs and − 3.4 d dec−1 for ANTDs. The trend is the highest and the most significant for 1976–2010 sub-period for both temperature indices. Linear warming occurs for ANSDs with a rate of 6.8 d dec−1 which is significant at the 0.01 level. The highest warming is found for ANTDs, with a trend rate of 7.2 d dec−1 that is significant at the 0.01 level of the significance (Table 1).
Table 1. Results of the Mann–Kendall rank correlation test and the Student t test for the significance of the β coefficient from linear regression equation for the annual numbers of summer- and tropical-day series averaged from 97 stations for the long period (1950–2010) and for two sub periods (i.e., 1950–1975 and 1976–2010)
Significance tests and linear trend rates
Tropical days (1950–2010)
Tropical days (1950–1975)
Tropical days (1976–2010)
Level of significance
Student t test for β
Level of significance
Trend rates (days per year)
Trend rates (days per decade)
Examination of anomalies in time-series of summer-days of Turkey as a whole, which was formed by averaging the data from 97 stations reveal that the annual number of summer-days was characterized with negative anomalies particularly for the period of 1963–1988, with the lowest values referring to the years of 1959, 1967 and 1976 (Figure 5(b)). After this period, runs of the anomaly showed a strong trend towards more summer-days, and particularly the years of 1994, 2007 and 2010 were detected abnormally warm at many stations of Turkey.
As in ANSDs, there was considerable decadal variability in ANTDs. Between 1967 and 1984, the number of tropical days was mostly lower than the 1961–1990 normal. There was particularly a strong decrease in the number of tropical days in Turkey in the years of 1967, 1976 and 1983 (Figure 6(b)). The largest number of tropical days occurred since 1989. The years of 2001, 2007 and 2010 were characterized with the greatest total number of tropical days. The hottest summers in Turkey suggested that the last decade stands significantly above any other decade since 1950. As previous studies described for the Mediterranean Basin temperature series (Jones and Moberg, 2003; Xoplaki et al., 2003b; Feidas et al., 2004), the coolest summer periods in Turkey were the 1960s and 1970s, with the coolest years in 1976 and 1967. There is an upward trend reversion in both annual numbers of summer and tropical day after 1976, and the warmest summers in both of two time series are of the same years: 2010 and 2007.
It would be worth noting that summer 2010 was the hottest one in Turkey that exhibited the highest positive anomalies in the annual number of summer-days over the whole study period beginning from 1950 to 2010. A glance at the spatial distribution of this year reveals strong positive departures on the northeastern Anatolia sub-region. The largest positive anomalies of summer-days occurred in series of the stations that are located on the Eastern part of Black Sea coasts, exceeding the normal by four standard deviations with respect to the period of 1961–1990 (Figure 7(a)). As in the ANSDs, the spatial distribution of ANTDs for summer 2010 indicates overall strong positive anomalies (higher than three standard deviations) with maximum values on the Black Sea coasts and Eastern and northeastern sub-regions of the Anatolia Peninsula. Turkey was affected by a warm period particularly in 15th and 16th weeks (7 August 2010 to 20 August 2010). On the same period, the record value of tropical days was observed over the Black Sea Region (Acar Deniz and Türkeş, 2011).
Year-to-year variability and decadal changes in extreme temperatures in the Mediterranean basin in summer seasons were found to be connected with both the atmospheric anomaly circulation patterns especially with strong geopotential blocking situation over the central Europe (Trigo et al., 2006), and atmospheric oscillations and Asian and African monsoons (Alpert et al., 2006). Previous works show that the positive temperature anomalies in summer months generally agrees well with 500–1000 hPa thicknesses anomalies from the southern North Atlantic to Russia. In this season, the movement of the monsoon low to higher latitudes over the Eastern and southeastern parts of Turkey leads to increased thicknesses and ridges at the 500 hPa level. These circulations contribute to isolate Westerlies and subsidence at the mid and upper troposphere and, imply warm and dry air advection from the Persian Gulf towards the Eastern Mediterranean Basin (Xoplaki et al.2003b; Tatli et al., 2005; Founda and Giannakopoulos, 2009; Kuglitsch, 2010). Another factor that could possibly have amplified temperature anomalies is the land-atmosphere interactions related to soil moisture deficits. For example, record-breaking positive temperature anomalies of the summer 2010 over the Eastern Europe and part of Russia associated with quasi-stationary anticyclonic circulation anomalies, and deficit of precipitation (Barriopedro et al., 2011).
Conversely, observed negative temperature anomalies in summer over the Eastern Mediterranean Basin are linked to the observed negative geopotential height anomalies over the region (Xoplaki et al., 2003b; Tatli et al., 2005). This leads to cold air advection in the upper levels associated with instability. For instance, the summer of 1976 was the coolest one according to the number of summer and tropical days averaged from 97 stations in Turkey during the last 60 years in Turkey relative to the 1961–1990 normal period as in the Eastern Mediterranean Basin (Figures 5(b) and 6(b)). However, the summer of 1976 was one of the hottest summers, with strong precipitation deficits in northern France and southern England (Fischer et al., 2007). The temperature anomalies over Europe and the Mediterranean Basin would likely be explained with anticyclonic conditions and pronounced stability over the Western and Central Europe, and cool air advection from higher latitudes via the Scandinavia towards the Eastern Mediterranean Basin (Xoplaki et al., 2003b). Therefore, increased convective instability and latent heat flux at the surface and upper troposphere levels, as well as enhanced cloud formation and precipitation would be very likely to occur in Turkey under these particular conditions (Türkeş, 2010).
One of the most significant signals of the global and regional climate changes and variability is the changes and trends in the extreme air temperature climatology, in addition to the changes and trends in the long-term average temperature conditions. Consequently, the present study described and analysed the trend of the annual number of summer and tropical days in Turkey for the period 1950 to 2010. The main conclusions of the study are summarized as follows:
(1)Results showed a general warming trend in the annual series of both summer and tropical temperature days at most of the stations over Turkey. Between 1950 and 2010, 53% (58%) of the stations show an increasing trend for summer-days (tropical days) that is statistically significant at the 1% level. The annual number of tropical days has increased at a faster rate than that of the annual number of summer-days. The pronounced trends at the annual number of tropical days can be attributed mainly to the fact that increasing the heat wave intensity, length and number associated with specific atmospheric circulation patterns across the Eastern Mediterranean region since the 1980s (Kuglitsch, 2010). Only five for ANSDs and six stations for ANTDs show insignificant decreasing trends.
(2)Two stages of change in the annual numbers of summer and tropical days can be recognized: decreasing trends up to the late 1970s, 1976 of which was considered as the coldest summer of the examined period. The ANSDs and ANTDs present the strong positive trend reversion in the late 1970s at the majority of stations. The last two decades were also recognized as the warmest summers during the whole period in Turkey. These results are consistent with the findings of other studies, which showed a strong and continuous warming trend at maximum temperatures observed after 1976 at northern hemisphere especially Eastern Mediterranean (Türkeş et al., 2002; Jones and Moberg, 2003; Feidas et al., 2004; Nastos and Matzarakis, 2007; Founda and Giannakopoulos, 2009). However, Xoplaki et al. (2003b) found no significant linear trend in the averaged summer months and entire summer mean air temperature for the Eastern part of the Mediterranean over the period 1950–1999. Recent observational studies and our results suggested that summer air temperatures of the Eastern Mediterranean basin have become particularly warmer in the last 10 years (Diffenbaugh et al., 2007; Hertig et al., 2010; Kuglitsch, 2010).
(3)Turkey experienced the warmest summer of its instrumental history in 2010. The magnitudes of normalized anomalies of tropical days at 2010 exceeded four standard deviations with respect to 1961–1990 normal at 19 stations. The spatial pattern of the maximum-level anomalies reveals that the northeastern Anatolia sub-region was in the centre of the exceptional warmth at all the observation periods.
(4)Recent studies revealed that the summer climate is controlled by the large-scale atmospheric circulation and atmospheric oscillation patterns such as the NAO and the Arctic Oscillation (AO) (Hurrell and Folland, 2002; Folland et al., 2009; Hertig et al., 2010). Because of this, in a future study, we shall analyse these relationships to explain the observed variations of the numbers of summer and tropical days by considering the influences of the high (positive)/low (negative) index phases of the summer NAO, AO and monsoon variability.
(5)The results show that climatology and variability of the high temperatures have been changing in Turkey, and the observed changes in the annual numbers of summer and tropical days appear to be fairly consistent with a warmer climate. Moreover, climate change model projections for the Mediterranean basin, larger Mesopotamia region and Turkey generally suggest an increase in the occurrence of extremely high temperature events and inter-annual variability in the summer season during the second half of the 21st century (Giorgi and Lionello, 2008; Alpert et al., 2008, Altinsoy et al., 2011, 2012; Türkeş et al., 2011; Sen et al., 2012; Ozturk et al., 2012). Finally, it can be concluded that these large-scale predictions and the recent observations on summer temperature trends over the Eastern Mediterranean region indicate the increasing potential risks of extreme high temperatures and longer heat-waves and their catastrophic effects such as enormous crop losses, forest fires and widespread persistent power losses.