OBSERVED CHANGES IN MAXIMUM AND MINIMUM TEMPERATURES IN TURKEY

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY, VOL. 16, (1996) OBSERVED CHANGES IN MAXIMUM AND MINIMUM TEMPERATURES IN TURKEY MURAT TURKES, UTKU M. SuMER AND GONuL KlLIC State Metemlogical Service, Department of Research, PO, Box 401 Ankara, Turkey Received 20 September 1994 Accepted 10 July 1995 ABSTRACT The aim of this study is to examine the nature and magnitude of the changes in mean seasonal daily maximum and minimum temperatures and diurnal temperature ranges for all of Turkey, its regions and the 59 stations during the period Maximum temperatures show significant cooling trends in many stations in summer and autumn. Most of the regional mean maximum temperatures have decreased in all seasons, except spring. ne of the observed trends in regions was significant statistically. Mean seasonal minimum temperatures have indicated significant warming trends in many stations in all seasons except winter. Majorities of the regional mean minimum temperatures have also increased markedly in all seasons. Warming trends of spring minimum temperatures were significant over the, and South-eastern Anatolia regions. Spring minimum temperatures for all of Turkey have shown a significant linear trend, with an increase of "C per decade. Diurnal temperature ranges have decreased significantly over most of the regions in all seasons, except partly in autumn and in so many stations in all seasons. Most of the decreasing trends in summer were significant at the 1 per cent level. It seems that changes of the diurnal temperature range in spring, summer and autumn were related closely to changes of maximum temperatures. Thus the decrease of the regional mean diurnal range was associated mainly with decreases of maximum temperatures in summer and autumn, and with slightly increased maximum temperatures in spring. Many stations that experienced a significant trend in diurnal ranges had a significant positive serial correlation. Our!%st results for Ankara have shown no significant trend in the mean monthly cloudiness and monthly cloudiness at 1400 hours local time, except a decrease in May cloudiness at 1400 hours. KEY WORDS: Turkey; maximum and minimum temperatures; diurnal temperature range; climatic trend and fluctuation. INTRODUCTION Climate is a variable phenomenon in that climatic conditions may change in response to external forcing and because of its own natural variability. With respect to the changes in surface temperatures, most common external and internal influences are changes in volcanic and solar activity, surface albedo, cloud cover, atmospheric concentrations of greenhouse gases and of sulphate aerosols. The world-wide interest in the greenhouse effect and urban heat island effect has led to detailed studies of long-term temperature data. According to the Intergovernmental Panel on Climate Change (IPCC, 1990), global mean surface temperatures have increased by 0.45 f 0.15"C since the late nineteenth century. This warming trend, however, has exhibited temporal and spatial variation. Similar variety appeared also in the years of 1990 and 199 1, which has been accepted as the two warmest years of the global record, and even in those years some regions indicated cooling (Folland et al., 1992). Global temperature anomalies were smaller in 1992 and 1993 than the peak value reached in 1990 (WMO, 1994). Some of this modest cooling has occurred due to the effects of the Mount Pinatubo eruption in 1991 (WMO, 1993; Bolin, 1994). Significant cold anomalies occurred also in 1992 over the most of Turkey (Tiirkeg et al., 1995). Various studies on surface air temperature data have shown that the observed warming during the several past decades has occurred mainly in association with the increase of daily minimum (night-time) temperatures rather than daily maximum (daytime) temperatures. The increase of night-time temperatures was apparent mainly over the USA, former USSR, China and Australia (Karl et al., 1991; Folland et al., 1992). It is considered that the marked increase in night-time temperatures may have occurred due to a combination of increased concentrations of CCC /96/ C 1996 by the Royal Meteorological Society

2 464 M. TURKES, U. M. SUMER AND G. KILIC anthropogenic greenhouse gases, increased cloud cover and urbanization. Using the gridded data set by Jones et al. (1 986), Nasrallah and Balling (1 993) have found that mean monthly temperature anomalies of the Middle East region exhibited significant warming trend. They have concluded that most of the increase has occurred in spring, although moderate amounts of warming have occurred in the summer and autumn seasons. Their results indicated that Turkey, with the highest sulphate ratios, had cooling trends in the west with -0.06"C per decade and in the east with -0.08"C per decade. In Turkey, studies conducted on long-term variations and trends in temperature data are very few. Turkes et al. (1992) analysed mean annual temperatures for all of Turkey by using departures of 21-year running means from the normal and concluded that annual temperatures increased by about 0.5"C from 1930 to Then a cooling trend was dominant by Using the temperature data of 18 stations, Kadioglu (1993) reported significant increasing trends in minimum temperatures, particularly in spring, and some decreasing trends in maximum temperatures. In a recent study, Tukes et al. (1 995) analysed mean annual temperatures from 7 1 stations in order to detect variations and trends by applying the non-parametric tests. They concluded that there was a general decrease in the mean annual temperatures in Turkey, and all coastal regions were characterized by cooling trends during the last two decades. Karaca et al. (1 994) investigated the effects of urbanization on the climate of Istanbul and Ankara, the two largest cities in Turkey. They found a significant increasing trend in the urban minimum temperatures of highly populated and industrialized southern part of Istanbul compared with the rural parts and pointkd out that there was not a significant urban heat island intensity in Ankara. However, regional and country-wide fluctuations and trends in maximum and minimum temperatures and the relationship between these changes and the decrease in diurnal temperature ranges were not analysed for Turkey. In this paper, we have analysed the maximum and minimum temperatures in order to document the nature and magnitude of the fluctuations and trends in Turkey, and to show whether these changes are similar to those that have occurred in other regions of the world. For this purpose, we have applied mainly the non-parametric statistical tests to the station-based data recorded at 59 stations during the period and the regional mean series for the period We have used the same tests on the diurnal temperature ranges (ATmax-min series) in order to document any possible changes in these series that may have occurred as a result of asymmetric changes or symmetric changes associated with the different size changes in the maximum and minimum temperatures. TEMPERATURE DATA We have used the seasonal means of daily maximum and minimum surface air temperatures recorded at 59 stations of the Turkish State Meteorological Service. For the station-based analysis, the study has covered the period of Stations had started to operate at different years up to the year The average record length of these stations was 57 years. Because these stations did not represent sufficiently the first 10 years of the regional mean series, we have based our regional analyses on the period The population of Turkey was 13.6 million in 1927 and reached 56.5 million in the 1990 census. During this period, the annual rate of growth has varied between approximately 2 and 3 per cent, except for the period Furthermore, it has been estimated that Turkey's population will be approximately 70 million by the year 2000 (DIE, 1991). The ratio of the rural population was 76 per cent in the 1927 census, but this ratio dropped to 41 per cent in the 1990 census. Therefore 59 per cent of Turkey's population are now living in urban areas. Most of the data used in this study have been obtained from stations located in the most heavily populated cities of Turkey and their populations have been growing with a highly increasing rate. Many stations located in the large cities of Turkey faced local and regional air pollution problems during the 1970s and 1980s. Maximum and minimum temperature records used in this study have been examined to get homogeneous records and as a proper geographical representation as possible. In principle, homogeneity of a record consists in the persistence of the measurement site, instrumentation and observation practice and nearby environmental conditions of that station. First, we selected a total of 80 stations for analysis. Our preliminary examination on the monthly data pointed out that a number of monthly records had missing observations. These missing values in the temperature records have been estimated by using data from a nearby reference station that has a high correlation (r > 0.8) and the same temperature regime as the base station. It was not possible, however, to estimate the missing data confidently for 15 stations and these were lefi out of the analysis. Secondly, we have checked the homogeneity

3 TEMPERATURES IN TURKEY 465 of our data by applying the non-parametric Kruskal-Wallis test for homogeneity of means to annual and seasonal series (Sneyers, 1990). Sample size of the subperiods and significance level of the test have been taken as nj = 10 and the lowest t~ = 0.05, respectively. We have also verified the homogeneity of the variances by using the same test. In this case, ranks of the absolute values of deviations from the general mean were used. Thirdly, we have made a subjective assessment of each statistically significant inhomogeneity by means of additional information available from the plotted graphs and our own station history file. Then the climatological significance of each has been assessed. Finally we have omitted six more stations from the analysis by this evaluation. When the station history files of the remaining 59 stations were examined, it appeared that 70 per cent of the stations had been moved and 79 per cent of these relocations had occurred between 2 metres and 5 km. Locations of the 59 climatological stations are shown in Figure 1 and number of stations used in calculating the regional mean temperature series is given in Table I. B L A C K S E A R e g i o n 9-0 'm GEOGRAPHICAL REGIONS OF TURKEY I 1 la' ib' 35 ' b' ib' 3's' i0' :i cbe Figure 1. Geographical regions of Turkey with locations of the climatology stations Table I. Number of the climatological stations in each region Geographical region Eastern Anatolia South-eastem Anatolia Turkey Number of station

4 and 466 M. -$, U. M. SuMER AND G. IULIC Time-series graphs METHOD OF ANALYSIS We have used the Gaussian filter, as a low-pass filter, to examine the long-term fluctuations visually in the regional mean series. In this method of filtering, the weighting of successive terms of a series varies symmetrically both backwards and forwards from a central weight (WMO, 1966). Sequential values of the statistics u(t) and u (t) from the progressive analysis of the Mann-Kendall test (Sneyers, 1990) have been plotted for selected stations. Without any trend, the graphical representation of the values u(t) and u (t) indicates curves that overlap several times. In the presence of a trend, intersection of these curves enables the location of the beginning of the trend or change to be approximated. Statistical tests We have applied the following two non-parametric tests to both regional mean and individual station data. The term of climatic trend is defined as A climatic change characterized by a smooth, monotonic increase or decrease of average value in the period of record. t restricted to a linear change with time, but characterized by only one maximum and one minimum at the end points of the record (WMO, 1966). We have used the Mann- Kendall test to detect any possible increasing or decreasing trend in the mean (Sneyers, 1990). When the original observations are replaced by their corresponding ranks ui, for each term yi, the number ni of terms yi preceding it (i >j) is calculated with (yi >yj> and the test statistic t is defined by i= 1 and its distribution function, under the null hypothesis, is asymptotically normal with mean and variance, n(n - 1) E(t) = ~ 4 n(n - 1)(2n + 5) v ~ t = 72 Using the two-sided test, the null hypothesis of absence of any trend is rejected for large values of I u(t) I with u(t) = [t - E(t)]/Jvart When the value of u(t) is significant at a desired level of significance, one can decide whether that is an increasing or decreasing trend depending on u(t) > 0 or u(t) < 0. The second test we have used is the Wald-Wolfowitz serial correlation. It was used to determine the randomness against the positive serial correlation (PSC) or abrupt change in series of observations (Sneyers, 1990, 1992). Using the one-sided test, the null hypothesis is rejected for large values of the test statistic u(r). The alternatives to randomness may also become an indicator of the presence of a low-frequency fluctuation or a sudden change of trend. RESULTS FROM THE REGIONAL DATA Changes in the regional mean maximum temperature series Most of the regional mean maximum temperatures apparently decrease in the summer and autumn seasons, but spring maximum temperatures tend to increase slightly in all regions. In winter, although there is a slight decrease over all of Turkey, maximum temperatures for all regions do not show any definite and common change in this season. Marked cooling trends in the summer maximum temperatures can be seen from the selected time-series graphs for the regions and for Turkey (Figure 2). Results of the statistical tests applied to the regional mean maximum and minimum temperatures are presented in Table 11. According to results of the Mann-Kendall test, none of the observed trends in mean maximum temperatures is statistically significant at the 5 per cent level. Decreases in the summer maximum temperatures are not linear. In fact, variations of summer maximum temperatures appear in a form of low-frequency fluctuation about a decreasing mean, as is observed particularly in

5 TEMPERATURES IN TURKEY (a) TURKEY SUMMER (b) MARMARA REGION SUMMER - 30? g295 i 29 z r (C) MEDITERRANEAN REGION SUMMER (d) CENTRAL ANATOLIA REGION SUMMER 32 I 31.5 I Y OE-Zr R*5.53W-2 CV(wpk.9 \I I I Figure 2. Annual variations of summer maximum temperatures over Turkey and selected regions with long-term average (. '), nine-point Gaussian filter (-), and linear fit to the trend (-) the and regions (Figure 2 (c and d)). Computed serial correlation coefficients for summer maximum temperatures are significant at the 1 per cent level in the and regions and all of Turkey and at the 5 per cent level in the Region. Changes in the regional mean minimum temperature series Majorities of the regional mean minimum temperatures show a long-term warming trend in the mean in all seasons except autumn. In spring, in spite of the marked cold anomalies in 1987 and 1992, minimum temperatures show prevailing increasing trends in all regions (Figure 3). Although spring minimum temperatures experience high year-to-year variability, warming trends in this season are characterized mostly by linear type persistence rather than abrupt changes or long-term fluctuations over all regions. Warming trends in the, and South-eastem Anatolia regions are significant at the 5 per cent level. Computed seasonal serial correlation coefficients are not significant except for summer. Summer minimum temperatures also tend to increase slightly in all regions except Eastern Anatolia (Figure 4). Magnitudes of the warming in the regional mean spring minimum temperatures are greater than those in the summer minimum temperatures. In summer, only a warming trend observed in the Region is significant at the 1 per cent level. Abrupt increases observed during the last decade in the and regions are significant. In autumn, the cooling trend in the Eastem Anatolia Region is significant at the 5 per cent level.

6 468 M. TURKES, U. M. SUMER AND G. KlLlC Table 11. Results of the Mm-Kendall (M-K) and Wald-Wolfowitz (W-W) tests for the regional mean series ( ). ( + ) Increasing trend; ( - ) decreasing trend; () no significant positive serial correlation; (*): significant at 5 per cent level; (**) significant at the 1 per cent level Geographical region Winter Spring Summer Autumn M-K W-W M-K W-W M-K W-W M-K W-W Turkey Eastem Anatolia South-eastem Anatolia Turkey Minimum temperature ("C) Eastern Anatolia South-eastem Anatolia + Maximum temperature ("C) Minimum temperature ("C) +* +* +* ** ** * ** ** ** ** * + Trend rates of the regional mean temperatures In spite of the high interannual variability and abrupt changes observed in the regional mean temperature series, we have calculated linear regression equations for regional series in order to determine seasonal warming and cooling rates per decade. Estimated linear regression lines are also shown in the seasonal time-series graphs of selected regions. In estimating linear regression lines, we have used the simple least squares solution with time as the independent variable and mean seasonal temperature as the dependent variable. Statistical significance of each estimated p coefficient has been tested by Student's t distribution with (n-2) degrees of freedom. The null hypothesis of absence of any trend has been rejected by using the two-tailed test for the large values of I t I. Table 111 shows the seasonal trend rates of regional mean maximum and minimum temperatures. Trend characteristics and significance levels of linear regression lines coincide perfectly with the results of the Mann- Kendall trend test. Daytime cooling trend rates are small or moderate and only two of them, the Region in winter and the Region in summer, are significant at the 5 per cent level. Regional mean minimum temperatures have increased over many regions during the last 54-year period, with a greater value in spring than the other seasons. It can be considered that rapid and intense urbanization in Turkey may have had an important role on this warming. Night-time urban warming trends in spring are marked and significant at the 5 per cent level in the,, and South-eastem Anatolia regions. Linear trend rates per decade for these regions are "C in the Region, "C in the Region, "C in the Region and "C in the South-eastem Anatolia Region. Along with the regions, spring night-time temperatures over all of Turkey also show a significant warming trend, with an increase of "C per decade. Summer night-time temperatures in the Region exhibit a marked warming trend, with "C per decade, which is significant at the 1 per cent level. In addition to warming trends, winter daytime temperatures in the central Anatolia Region, summer daytime temperatures in the Region and autumn night-time temperatures in the Eastern Anatolia Region show a significant cooling trend at the 5 per cent level. Trend rates per decade for these regions are "C in, "C in, and "C in Eastern Anatolia.

7 TEMPERATURES IN TURKEY 469 (a) TURKEY SPRING (b) BLACK SEA REGION SPRING F G8.5 [L 2 484%-2x R'=7 9017E-2 CV(%)=ll (c) AEGEAN REGION E7 5 L z 7 2 F U I $55 5 R' E-2 CV(%)=l3 68 ;O"l'ehS' '1'9%' 'ibsi' isso ' 'ibio' ibii' 'isz SPRING (d) SOUTHEASTERN ANATOLlA REGION ~-! SPRING -~ -a Rs7 4ME 2 cv(%) X ~T-T '1'&"1&5' '19?0"19%"19%' '1984 'l&' Figure 3. Annual variations of spring minimum temperatures over Turkey and selected regions with long-term average (... ), nine-point Gaussian filter (-) and linear fit to the trend (-) RESULTS FROM THE STATION-BASED DATA Results from application of the Manr-Kendall and Wald-Walfowitz tests to the seasonal temperatures of 59 stations are summarized in Table IV, and based on the original values of the test statistics (not given here) the following results were obtained. Changes in the maximum temperatures Station-based maximum temperatures show a low-frequency fluctuation and abrupt change with a significantly decreasing mean in many stations. In winter, three stations indicate a statistically significant cooling trend in the mean of the maximum temperature series. In spring, three stations experience a significant warming and cooling trend. In summer, 12 stations show a significant trend and 83 per cent of these are cooling trends. Summer maximum temperatures have been decreasing significantly at the stations of the,, central Anatolia and Eastern Anatolia regions. Fifty-six per cent of all stations also give significant PSC coefficients in summer. Seventy-nine per cent of significant PSC values coincide with the negative trends. Such a peculiarity can be seen from the time-series graph of the summer maximum temperatures plotted for Van, which records a rapid decreasing trend (Figure 5). The station of Van in the Eastern Anatolia Region indicates a significant cooling trend

8 470 M. TURKES, U. M. SmER AND G. IULIC (b) AEGEAN REGION SUMMER l L.:g;:+!n:E-: , p ~O.zl", ,cvrr34;, Y,,,,, 1980,, rr (c) MEDITERRANEAN REGION SUMMER r-l- (d) CENTRAL ANATOLlA REGION SUMMER I I lg ~ E-2x R*l.SSSE-Z CV( ' '19iO' 'lbkb' '19%' 81k58 'lbibb'1b7bc'lib%' '10k' 'l'deb' Y=l C.B872-2x R44.Em4E-Z CV(%)= ,,I,,,,,,,,n,n,l, I I I, / / I, / I, 1 I I I I I,, I, I ~ I, Figure 4. Annual variations of summer minimum temperatures over Turkey and selected regions with long-term average (... ), nine-point Gaussian filter (-) and linear fit to the trend (-) and significant PSC at the 1 per cent level. In autumn, 10 stations indicate significant trends and 90 per cent of these are cooling trends. Changes in the minimum temperatures Minimum temperatures tend to increase significantly in so many stations in all seasons except winter. In winter, five stations indicate significant trends in the mean of minimum temperature series, most of which are decreasing. In spring, 26 stations show a significant trend and almost all of them are warming trends. It means that spring minimum temperatures are increasing at a considerable number of the stations over Turkey. Warming trends are apparent particularly in the and regions. For instance, an observed warming trend at Adana in the Region is significant at the 1 per cent level (Figure 6). In summer, 25 stations show a significant trend and 72 per cent of these are warming trends. Minimum temperatures of all stations having a significant trend over the Region of are tending upwards in summer. The value of the actual warming has been increasing since the year Summer minimum temperatures have significant PSC values at 76 per cent of the stations. Sixty-four per cent of significant PSC values coincide with the increasing trends. For instance, summer minimum temperatures of Manisa in the Region indicate a significant warming trend and positive serial

9 TEMPERATURES IN TURKEY 47 1 Table 111. Regional temperature trends for all seasons during the period. ( + ) Increasing trend; ( - ) decreasing trend; (*) significant at the 5 per cent level; (**) significant at the 1 per cent level Linear trend rate ( C decade- ) Season Region Maximum temperature Minimum temperature Winter Spring Summer Autumn Eastern Anatolia South-eastem Anatolia Turkey Eastern Anatolia South-eastern Anatolia Turkey Eastern Anatolia South-eastem Anatolia Turkey Eastern Anatolia South-eastern Anatolia Turkey * * I , * * * * 0.145* ** f * correlation, which are both significant at the 1 per cent level (Figure 7). In autumn, minimum temperatures of 23 stations show a significant trend and 61 per cent of these are downward. CHANGES IN THE DIURNAL TEMPERATURE RANGES Diurnal temperature ranges apparently decrease over almost all regions and all of Turkey in all seasons, except partly in autumn. Table V gives the results of statistical tests on mean seasonal diurnal temperature ranges for the regions. There is a strong decline in dry and hot summers over all of Turkey and all the regions during the last 54 years. Decreasing trends observed in the diurnal temperature ranges in summer are significant at the 1 per cent level over most of the regions, except those over the Eastern Anatolia and South-eastern Anatolia regions, and all of Turkey. A decreasing trend in wet and cold winters is significant at the 1 per cent level over the and regions, and at the 5 per cent level over the Eastern Anatolia Region. Almost all regional PSC coefficients for summer and most of the regional PSC coefficients for winter are significant at the 1 per cent level. If a simple correlation analysis is made between the mean seasonal diurnal temperature ranges and mean seasonal maximum and minimum temperatures, it seems that changes for most of the regional mean diurnal temperature range in winter are controlled greatly by changes in the minimum temperatures. For all of Turkey,

10 472 M. TURKES, U. M. SUMER AND G. KILIC Table IV. Number of the stations indicating significant trend andor positive serial correlation at the 5 per cent or 1 per cent level according to the Manr-Kendall (M-K) and Wald-Wolfowitz (w-w) tests Geographical region Winter Spring Summer Autumn M-K W-W M-K W-W M-K W-W M-K W-W Eastern Anatolia South-eastem Anatolia Total Maximum temperature ( C) Minimum temperature ( C) 2 4 I ) Eastem Anatolia South-eastem Anatolia Total changes in winter diurnal ranges are related more closely to changes in the minimum temperatures (r= -0.37, significant at the 1 per cent level) than to changes in maximum temperatures (r= -0.07, not significant). In spring, changes in almost all regional mean diurnal ranges are controlled largely by changes in maximum temperatures. Regarding all of Turkey, changes in the diurnal temperature ranges in spring are related more closely to slightly increased maximum temperatures (r = 0.66, significant at the 1 per cent level) than to changes of significantly increased minimum temperatures (r = 0.19, not significant). In summer, changes in all regional mean diurnal ranges are controlled largely by changes in maximum temperatures. For all of Turkey, changes of the diurnal temperature ranges in summer are related more closely to changes of maximum temperatures (r = 0.56, significant at the 1 per cent level) than to changes of minimum temperatures (r= -0.17, not significant). In autumn, changes in most of the regional mean diurnal ranges are controlled mainly by changes in maximum temperatures. For all of Turkey, changes of the diurnal ranges in autumn are related more closely to changes of maximum temperatures (r = 0.45, significant at the 1 per cent level) than to changes of minimum temperatures (r= -0.14, not significant). Therefore decreases of regional mean diurnal ranges in summer and autumn are associated mainly with decreases of maximum temperatures, whereas it is associated mainly with slightly increased maximum temperatures in spring. Results of the statistical tests applied to mean seasonal temperature ranges of 59 stations are summarized in Table VI, and based on the original values of the test statistics (not given here) we have found following. Diurnal temperature ranges decrease significantly in many stations in all seasons. In winter, 32 stations indicate a significant trend in the diurnal temperature ranges and 78 per cent of these tend downward significantly. Seventyfour per cent of the stations have significant PSC coefficients in spring. In spite of the increases observed in both maximum and minimum temperatures, diurnal temperature ranges apparently decrease because of increases in minimum temperatures are greater than those in maximum temperatures. In this season, 89 per cent of 27 stations having a significant trend experience a decreasing trend, which mostly is significant at the 1 per cent level. Thirty per cent of the stations have significant PSC values. In summer, 35 and 52 of stations indicate significant trends and significant PSC coefficients, respectively. Seventy-four per cent of these significant trends are downward and most of them are significant at the 1 per cent level. Decreases for the diurnal temperature range in summer are related to well-defined asymmetric trends in the diurnal maximum and minimum temperatures of this season. In autumn, 36

11 473 4 Figure 5. (a) Annual variations of summer maximum temperatures at Van (38"29' N, 43"23'E) with long-term average (... ) and nine-point Gaussian filter (-). (b) Sequential values of the statistics u(r) (-) and u'(t) (-) from Mant-Kendall test for Van and 37 of the stations show significant trends and significant PSC coefficients, respectively. Fifty-eight per cent of the significant trends are downward and most of them are significant at the 1 per cent level. TEMPERATURE-CLOUDINESS RELATIONSHIP IN ANKARA Clouds, with their negative and positive feedbacks have an important role on the climate system, but the characteristics and directions of feedbacks are complex. We have examined the relationship between temperature and cloudiness by using the monthly temperature and cloudiness data recorded in Ankara from 1930 to 1990, as a preliminary investigation. Regarding climatological application, total cloud cover is a fraction of the sky covered by all the clouds visible and cloud-cover scale varies between 0 and 10. Monthly series of mean cloudiness and cloudiness at 1400 hours local time do not show any significant trend except May cloudiness at 1400 hours. A decreasing trend in May is significant at the 5 per cent level. There is not a significant trend in May maximum temperatures. Graphical representation of the monthly correlation coefficients between the temperature and cloudiness series is given in Figure 8. Statistically significant negative relationships are dominant between mean monthly temperatures and mean cloudiness during relatively the warm and dry periods of the year (April-September) and between maximum temperature and cloudiness at 1400 hours local time from'march to October. Mean and maximum temperatures are relatively low when amounts of the mean cloudiness and of the 1400 hours local time cloudiness

12 474 M. m S, U. M. SuMER AND G. KILlC ADANA SPRING 14 H 13.5 L 'l I I,, 1, I I I I I, I I, I,,,,,,, 1 <, I,,,,,,,,,,,,,,,,,,,,,,,,d ls45 10Hl losd lobo' s lee0 Figure 6. (a) Annual variations of spring minimum temperatures at Adana (37'03' N, 35"21' E) with long-term average (... ) and nine-point Gaussian filter (-). (b) Sequential values of the statistics u(t) (-) and d(t) (- -) from the Mann-Kendall test for Adana are relatively high-in May, for example. On the contrary, significant positive correlation coefficients are seen between the minimum temperatures and mean cloudiness during relatively the cold and rainy period of the year (October-January). Low minimum temperatures are associated relatively with low cloud coverage in vember, as Palmieri et al. (1991) have pointed out for Rome. CONCLUSIONS We have based our analysis on mean seasonal daily maximum and minimum temperature data. This has allowed us to see and evaluate any differential behaviour between day and night. Results of the Mann-Kendall trend test have shown that maximum temperatures have decreased significantly in many stations in summer and autumn seasons. Most of the regional mean maximum temperatures have tended to decrease slightly over time in all seasons except in spring. Observed cooling trends in the regions, however, were not statistically significant at the 5 per cent level. On the contrary, considerable portions of the station-based minimum temperatures have indicated significant warming trends in all seasons except in winter. In autumn, however, the number of stations having a significant cooling trend in their minimum temperatures was more than those having a significant warming trend. Spring minimum temperatures have shown significant warming trends over the,, and South-eastem Anatolia regions, and approximately in half of the 59 stations used in this study. Least squares regression analysis

13 TEMPERATURES IN TURKEY 475 WISA SUMMER 1 16 JII,, I,,,,,,,,,,,,,,,,,,,,,,, lw lq80-2 ',,,,,,,,,,,,,,,,,,,~,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,-~~/, Q lee Figure 7. (a) Annual variations of summer minimum temperatures at Manisa (38"36' N, 27"26' E) with long-term average (... ) and nine-point Gaussian filter (-). (b) Sequential values of the statistics u(r) (-) and d(t) (- - -) from the Mann-Kendall test for Manisa Table V Results of the Mann-Kendall (M-K) and Wald-Wolfowitz (W-W) tests for regional AT,, - min series. ( + ) Increasing trend; (-) decreasing trend; () no significant positive serial correlation (*) significant at the 5 per cent level; (**) significant at the 1 per cent level A Tmax - min Region Winter Spring Summer Autumn M-K W-W M-K W-W M-K W-W M-K W-W Turkey - ** - * - ** ** - ** * - ** ** MallIl2U-a - - ** ** + * - ** ** - + ** - * - ** ** + - ** ** - * - ** ** - * Eastern Anatolia - * ** - ** + South-eastern Anatolia - ** - *

14 476 M. TURKES, U. M. SUMER AND G. IULIC Table VI. Number of stations indicating a significant trend andor positive serial correlation in series at the 5 per cent or 1 per cent level, according to the Mann-Kendall (M-K) and Wald-Wolfowitz (W-W) tests Region Winter Spring Summer Autumn Eastem Anatolia South-eastem Anatolia Total M-K W-W M-K W-W M-K W-W M-K W-W z has shown t t the observed warming trends of night-time urban temperatures in spring were significant over four geographica regions and all of Turkey. Linear trend rates per decade were "C in the Region, o"C in the Region, "C in the Region, "C in the South-eastern Anatolia Region and "C for all of Turkey. Mean seasonal diurnal temperature ranges have decreased over all of Turkey and most of the regions in all seasons except partly in autumn, during the period. Mean seasonal diurnal temperature ranges have tended to decrease significantly over all of Turkey and in most of the geographical regions of in dry and warm summers. Decrease in wet and cold winters were considerable as well. In winter, changes in the diurnal temperature range in most of the regions and over Turkey were controlled closely by changes of the minimum temperatures. For the spring, summer and autumn seasons, changes in the diurnal temperature range in the majority of regions and over Turkey were controlled closely by changes in maximum temperatures. Therefore, we have concluded that the decrease in the regional mean diurnal range is associated mainly with decreases of maximum temperatures in the summer and autumn seasons, and with slightly increased maximum temperatures in spring. Regarding all seasons, it was striking to see that 96 of 130 stations that were characterized by a significant trend have experienced a decreasing trend in the diurnal temperature ranges. Stations that indicate significant decreasing trends in winter and particularly in summer had a significant PSC coefficient. Our preliminary results on the relationship between temperature and cloudiness for Ankara have shown no significant trend in the mean monthly cloudiness and monthly cloudiness at 1400 hours local time, except for a Y ti!j 4.12 W (L K J F M A M J J A S O N D MONTH Figure 8. Correlation between monthly temperature and cloudiness in Ankara from 1930 to (-m-) Mean temperaturemean cloudiness; (-*-): minimum temperature-mean cloudiness; and (-A-) maximum temperaturcxloudiness at 1400 hours local time

15 TEMPERATURES IN TURKEY 477 decrease in May cloudiness at 1400 hours. Significant negative relationships were dominant between mean monthly temperatures and mean cloudiness from April to September and between maximum temperature and cloudiness at 1400 hours local time from March to October. Significant positive correlations were observed between minimum temperatures and mean cloudiness from October to January. The possible impact of the global warming on natural and socio-economic systems can vary from region to region in a large country such as Turkey. Therefore, an integrated impact assessment of the climate change should be developed not only for the whole of Turkey but also for its geographical regions. Without separating effects of the temperature changes between daytime and night-time, the largest response to warming in the winter and spring seasons is a decrease in energy demand for heating requirements. Warmer springs and summers can cause more energy demand for air conditioning and imgation services and an increase in water consumption. Human-induced pressure on water resources and degradation of water quality we have already faced will grow very rapidly. Therefore, findings on changes in the climate system and assessments on possible impacts of it should be taken into consideration by policy makers in Turkey and other countries of the region. Our findings on some asymmetric changes and symmetric changes with different sizes in the mean seasonal diurnal maximum and minimum temperatures and marked decreases in the diurnal temperature ranges have coincided generally with the results of studies in the USA, former USSR, China, and Australia, and in some other regions of the globe. However, we consider that a further analysis is required about effects of urbanization on country-wide temperature variations in order to decide whether rapid and intense urbanization and air pollution in the large cities of Turkey may have had an important role on night-time warming in summer and spring and on daytime cooling in summer and autumn. In conclusion, despite the several scientific studies we believe in that it is still difficult to distinguish a possible enhanced greenhouse effect signal or any change from the natural variability in the climate system. REFERENCES Bolin, B Report to the Ninth Session of the Intergovernmental Negotiating Committee for a Framework Convention on Climate Change (INC/FCCC), 7 February 1994, WMO/UNEP/IPCC, Geneva, 8 pp. DIE Census of Population in Turkey, Administrative Division (Summary Tables), State Statistics Institute of Prime Ministry, Ankara, 36 pp. Folland, C. K., Karl, T. R., Nicholls, N., Nyenzi, B. S., Parker, D. E. and Vinnikov, K. Y Observed climate variability and change, in Climate Change 1992: The Supplementary Report to the Intergovernmental Panel on Climate Change Scientific Assessment, Cambridge liniversity Press, Cambridge, pp IPCC Climate Change: The Intergovernmental Panel on Climate Change Scientific Assessment, Cambridge University Press, Cambridge, 365 pp. Kadioglu, M Climate change in Turkey and its possible impacts, Environ. Protection, Istanbul, 47, (In Turkish). Karaca, M., Tayancc, M. and Toros, H Effects of urbanization on climate of Y[aa]stanbul and Ankara, Atmos. Environ., Part B: Urban Atmosphere, in press. Karl, T. R., Kukla, G., Razuvayev, V. N., Changery, M. J., Quayle, R. G., Heim, R. R., Easterling, D. R. and Fu, C. B Global warming: evidence for asymmetric diurnal temperature change, Geophys. Res. Lett., 18, Nasrallah, A. H. and Balling, R. C Spatial and temporal analysis of Middle Eastem temperature changes, Clim. Change, 25, Palmieri, S., Siani, A. M. and D Agostino, A Climate fluctuations and trends in Italy within the last 100 years, Ann. Geophysicae, 9, Sneyers, R On the Statistical Analvsis of Series of Observations, WMO Technical te, 143, World Meteorological Organization, Geneva, 192 pp. Sneyers, R Use and misuse of statistical methods for the detection of climate change, in Climate Change Detection Project. Report on the Informal Planning Meeting on Statistical Procedures for Climate Change Detection, WCDMF, 20, pp Tiirkes, M., Siimer, U. M. and KiliG, G Protection of the Atmosphere and Climate Change, State Meteorological Service, Ankara, 110 pp. (in Turkish). Tiirkeq, M., Sumer, U. M. and Kilic, G Variations and trends in annual mean air temperatures in Turkey with respect to climatic variability, Int. 1 Climatol., 15, WMO Climatic Change, WMO Technical te 79, World Meteorological Organization, Geneva, 79 pp. WMO The global climate system in 1992, WMO Bull., 42(3), WMO The global climate system in 1993, WMO Bull., 43(3),