INFLUENCE OF GEOPOTENTIAL HEIGHTS, CYCLONE FREQUENCY AND SOUTHERN OSCILLATION ON RAINFALL VARIATIONS IN TURKEY

INTERNATIONAL JOURNAL OF CLIMATOLOGY

Int. J. Climatol. 18: 649–680 (1998)

INFLUENCE OF GEOPOTENTIAL HEIGHTS, CYCLONE FREQUENCYAND SOUTHERN OSCILLATION ON RAINFALL VARIATIONS IN

TURKEYMURAT TU8 RKES*

State Meteorological Ser6ice, Department of Research, PO Box 401, Ankara, Turkey

Recei6ed 29 May 1996Re6ised 19 No6ember 1997

Accepted 20 No6ember 1997

ABSTRACT

Normalized rainfall of Turkey tended to decrease in many annual and winter series and to increase in some springand summer series during 1930–1993. Low-frequency fluctuation of regional winter rainfall series was generallydominated by cycles of 3–3.2, 6–7, 7–8.4, and 14–21 years. Spring rainfall series depicted high-frequency oscillationswith cycles of 2–2.2 years, and longer cycles of 4.2–4.7 years. Mean 700 and 500 hPa geopotential heights overTurkey generally experienced positive anomalies from late 1970s to early 1990s, and showed an upward trend inwinter and summer. Significant negative correlations were found between geopotential height and rainfall anomaliesin winter over most of Turkey. Cycles of 2–2.2 and 3.2–3.8 years in spring rainfalls appeared to be associated withsimilar oscillations of spring geopotential heights. Cycles of 13 years in both winter geopotential series reflected in asimilar cycle of 14 years in annual and winter rainfall. The number of depressions reaching Turkey tended to decreasefor about 10 years. Increased frequencies and intensities of dry conditions in the last ca. 20 years may have beenrelated to increased geopotential heights and decreased frequency of depressions over Turkey.

Signs of warm minus cold event winter anomalies during various stages of the Southern Oscillation revealed theexistence of some coherent regions without significant signals. Most of the selected 48 stations had a positive signanomaly during year −1 warm and cold events. The cold event rainfall means showed a coherent region ofsignificantly increased rainfall conditions over the central-west and central parts of Turkey. Slightly wetter thannormal warm event conditions during year +1 were observed in many stations. Most of warm and cold eventresponses were characterized by a decreased rainfall. Drier than long-term average conditions were significant at somestations during year +1 cold events. Warm minus cold event differences had an opposite signal between year −1and year 0 (+1) in many stations. Opposition of composite anomalies was evident in most of stations between year−1 and year +1 cold events. © 1998 Royal Meteorological Society.

KEY WORDS: Turkey; rainfall; geopotential height; depressions; Southern Oscillation (SO) extremes; trends; persistence; powerspectrum analysis; cycles

1. INTRODUCTION

For the greater part of Turkey, it is natural systems and human activities that are most at risk due to thelack of sufficient and regular water during the year. This is largely due to the higher year-to-yearvariability and high seasonality in rainfall. On the other hand, these activities in countries with a lowerrainfall variability and seasonality do not suffer as much from the irregularities of the climate.Historically, sufficient and timely rainfall in the autumn and spring months is vital to the agriculturalactivities over most of Turkey. Therefore, by also considering high population growth and rapidurbanization, societies living in Turkey have to use water resources more rationally throughout the yearby planning their water requirements. With respect to the rainfall variability and change, there is someevidence for a consistent decreasing trend in some regions of the world, e.g. the Mediterranean Basin andthe Sahel region. The general downward trends over most of Greece and Turkey with abrupt decreases

* Correspondence to: Turkish State Meteorological Services, PO Box 401 Ankara, Turkey.

CCC 0899–8418/98/060649–32$17.50© 1998 Royal Meteorological Society

MURAT TU8 RKES650

during the last 30 years and ca. 20 years, respectively, were striking (Repapis et al., 1989; Amanatidis etal., 1993; Turkes, 1996a,b). The 500 hPa geopotential anomalies over the Mediterranean Basin during1980–1990 were higher than the reference means (WMO/WCDP, 1987; WMO/UNEP, 1990, 1992).During this period, country-wide mean annual rainfall totals for Italy were below the 1951–1980 normalin almost all years (WMO/UNEP, 1992). Rossi and Somma (1995) reported that the presence of persistentanticyclones during the period 1980–1990 became especially critical from September to March in1988–1989 and 1989–1990. Furthermore, a climate model by the Hadley Centre (UKMO, 1995) predicteda decrease in mean annual precipitation for future changes up to 2050, over many parts of the subtropics.These decreased rainfall regions include the Middle and Eastern Mediterranean, and the western andsouthern margins of Turkey. According to a recent assessment for Europe (ECSN, 1995), a drasticdecrease was expected for winter precipitation over Turkey.

According to evidence from several studies of the extratropical response to the El Nino/SouthernOscillation (ENSO) events, the large-scale circulation of the mid-latitude atmosphere is closely associatedwith the atmospheric variations in the tropical Pacific Ocean, particularly in winter. By considering thevariability of the extratropical signal associated with the ENSO phenomenon, Emery and Hamilton (1985)found a definite tendency for the Aleutian low to be intensified during the winters corresponding with themature phase of an ENSO event. Ropelewski and Halpert (1987) identified some coherent regions withENSO-related precipitation in Australia, North and South America, the Indian subcontinent, Africa, andCentral America. The relatively large, coherent regions suggestive of ENSO-related precipitation were alsoshown over the northern Africa–southern Europe (NAS) and the Mediterranean Mid-east (MME). Theyalso stated that the implied ENSO relationships in the regions of NAS and MME were difficult tounderstand or attribute to any of the known ENSO-related atmospheric circulation changes. Hamilton(1988) noted that a major factor in determining the strength of the Northern Hemisphere extratropicalresponse was the sea surface temperature (SST) in the far western Pacific/Indonesian region; and thestrong extratropical teleconnection with the ENSO events occurred more frequently when the SST in thefar western Pacific was anomalously warm. Kiladis and van Loon (1988) showed that in the year beforethe development of positive sea surface temperature anomalies in the central and eastern equatorialPacific, a strong South Pacific high was associated with below normal surface pressure over Australia andthe Indian Ocean. They concluded that this occurred concurrently with a poleward displacement of thePacific convergence zones, with above normal air temperature and precipitation over the subtropicalPacific, and opposite conditions along the equator. Ropelewski and Halpert (1989) also examined therelationships between precipitation and the high index phase of the Southern Oscillation (SO) for 19regions of the globe, with low SO index–precipitation relationships. They revealed that most of thoseregions also indicated evidence of characteristic precipitation anomalies during the high index phase of theSO, and the high SO index–precipitation relationships had the opposite sign for the low index phase.Kiladis and Diaz (1989) found that large regions of coherent and significant signals existed for bothextremes of the SO, with warm event signals generally opposite to those during cold events, and climaticanomalies during the year −1 tended to be opposite to those during the year 0. Their results also revealeda well-defined region of drier than normal warm event winter anomalies over the MME and south-easternTurkey during the year 0; along with the wetter conditions over the Balkans including north-westernTurkey during the year +1.

Fraedrich (1990) analysed the cyclonic and anticyclonic European Grosswetter to determine theresponse of the synoptic climatology over Europe to the extreme events in the ENSO. He found thatbi-monthly ranked composites computed over 2-year warm (cold) episodes showed more days of cyclonic(anticyclonic) steering over Europe, which was largest in January and February following the year of awarm or cold event. In a more recent study for Europe, Fraedrich and Muller (1992) showed that thenegative pressure anomalies at western and central European stations during the warm event winters wereassociated with positive temperature and precipitation anomalies; the reverse signals were observed innorthern Europe. They also found the area from western and south-western parts of Europe to the BlackSea was drier during the cold events, due to the lack of rain-bearing frontal systems over Europe.Scandinavia and Eastern Mediterranean including southern Turkey was drier during the warm events,

© 1998 Royal Meteorological Society Int. J. Climatol. 18: 649–680 (1998)

RAINFALL VARIATIONS IN TURKEY 651

because of southward shift of the European part of the cross-Atlantic cyclone track, and of northwardshift of the Mediterranean cyclone track relative to the cold event, respectively.

As for the recent studies in Turkey, Turkes (1996a) found that area-averaged annual and winterrainfalls tended to decrease slightly in all of Turkey and most noticeably in the Black Sea andMediterranean rainfall regions. Turkes (1996b) showed those rainfall variations over most of the rainfallregions except the Black Sea region were closely linked to those over the rest of the country. Theconcurrence of the dry conditions between the rainfall regions and the rest of Turkey appeared generallyduring the early 1930s, the late 1950s, the early 1970s, around the 1980s and the early 1990s, whereas wetconditions were observed generally during the period 1935–1945, around the 1960s and the late 1970s.Severe and widespread dry conditions over Turkey occurred especially in the years of 1973, 1977, 1984,1989 and 1990 (Turkes, 1996c). Turkes (1997) also documented that there was a general tendency fromhumid conditions of the 1960s towards dry sub-humid climatic conditions, in the aridity index values forall of Turkey. A significant change from humid condition to dry sub-humid or semi-arid climaticcondition existed in some stations of the Marmara and the Aegean regions. However, it is believed thatthere are still too many gaps to be completed on rainfall variations in Turkey, relationship betweenrainfall variations and variations of atmospheric disturbances over Turkey, and Turkey’s rainfallclimatology. The main goal of this study is: (i) to give further information on Turkey’s rainfallclimatology by examining the major characteristics of the spatial distribution of mean annual andseasonal rainfall totals and of year-to-year variability of rainfall, and seasonality of the rainfall for 99stations; (ii) to show the nature and magnitude of the secular trends, long-period fluctuations, year-to-yearpersistence and periodicity in the rainfall series of 91 stations and geopotential height series, and ofpersistence and periodicity in the regional rainfall series; and (iii) to quantify the spatial and temporalrelationships between the rainfall variations and the upper atmospheric variations and frontal depressionactivities; and association of the rainfall anomalies with various stages of the warm and cold events in theSouthern Oscillation.

2. DATA

Regarding the time-series analysis, this study uses the monthly totals of the daily rainfall data recordedat the 91 stations of the Turkish State Meteorological Service (TSMS), during the period 1929–1993, witha length of observation record varying from 54 years to 65 years. The data of 1929 was only used for the1930 winter. Eight stations were added to these 91 stations solely for the study of rainfall climatology ofTurkey, in which three of those have relatively shorter records, and others show some inhomogeneity. Therainfall data was broadly discussed in an earlier paper (Turkes, 1996b). The locations of the 99 stationswith numbers used in this paper are shown in Figure 1. Classification of the stations was based mainlyon seasonality of rainfall and geographical control of it. The list of these stations with numbers on themap of location is presented in Table I by the rainfall regime regions; with negligible difference toprevious classification, arising mainly from the displacement of Ulukısla from the MED to CCAN region.For detecting the relationship between variations in rainfall and upper atmospheric conditions, monthlyrecords of 00:00 UTC time 700 and 500 hPa geopotential level heights measured at Istanbul (Goztepe)and Ankara radiosonde stations were taken from the TSMS. These records include data for the period1952–1993. Cyclone (extra-tropical depression) data for the area that covers Middle and EasternMediterranean, Balkans and Turkey were taken from Deniz and Karaca (1995). They derived the pathsand numbers of the depressions from the Bulletin of Meteorologische Abhandlungen, which has beenpublished by the Berlin University Meteorology Institute. They subjectively analyzed the surface pressureand 500 hPa standard level weather maps for the period 1979–1992, by taking into account the locationcentres of isobars and contours on the maps, respectively (A. Deniz and M. Karaca, personnalcommunication, 1997).


MURAT TU8 RKES652

3. METHOD OF ANALYSIS

The normalized rainfall anomaly (Asy) for long series of a given station is

Asy= (Rsy−R( s)/ss

where Rsy is the total rainfall for the station s during a year y (or a season); R( s and ss are the mean andstandard deviation of the annual (or seasonal) rainfall for that station, respectively. Area-averagednormalized anomaly (Ary) for a given rainfall region is formulated to construct the regional mean series,as

Ary= (1/Ns) %Ns

s=1

Asy

where Ns is the number of regional stations operating in the year y.The Mann–Kendall rank correlation test has been used to detect any possible increasing or decreasing

trend in the long series of normalized rainfall and geopotential height anomalies. Using the two-sided testof the normal distribution, the null hypothesis of absence of any trend is rejected for large values of thetest statistic �u(t)� (Sneyers, 1990). The non-circular lag-one serial correlation test has been performed todetermine the randomness against the persistence from year-to-year variations in the series of normalizedrainfall and geopotential height anomalies. Using the one-sided test of the normal distribution, the nullhypothesis of randomness is rejected for the large values of the test statistic (r1)t. The serial correlationcoefficients for the first few lags have also been examined to determine whether the persistence in thenormalized rainfall and geopotential height series is a simple Markov-type persistence. If the lag-one serialcorrelation (L-1SC) coefficient differs significantly from zero, and if those respecting lag-two and lag-threeapproximate to the square and the cube of the L-1SC, respectively, the series would be assumed to containa Markov persistence (WMO, 1966). By using the Blackman and Tukey approach (Blackman and Tukey,1958), as outlined in WMO (1966), power spectrum analysis has been applied to the series of non-smoothed normalized rainfall and geopotential height anomalies in order to detect the dominant cycleswithin the observed fluctuations. Maximum lag has been chosen as about one-third of the record length,which is 21 years for the area-averaged rainfall series, and 13 years for the geopotential heights. Final

Figure 1. Location of 99 stations over Turkey and their numbers used in the study (see Table I for rainfall regime regions)


RA

INF

AL

LV

AR

IAT

ION

SIN

TU

RK

EY

653

©1998

Royal

Meteorological

SocietyInt.

J.C

limatol.

18:649

–680

(1998)

Figure 2. Geographical distribution of the mean seasonal rainfall totals (mm) for 99 stations in Turkey

MURAT TU8 RKES654

Table I. List of the stations used for rainfall climatology, rainfall variation, and theSouthern Oscillation

Region Station

Number Name

Black Sea (BLS) 1 Rizea

2 Trabzona

3 Giresuna

4 Samsuna

5 Sinop6 Zonguldaka

7 Bolua

8 Adapazarı

Marmara Transition (MRT) 9 I: zmit10 Istanbula

11 Sile12 Florya13 Corlu14 Luleburgaz15 Kırklarelia

16 Edirnea

17 Tekirdaga

18 Bilecika

19 Bursaa

Mediterranean (MED) 20 Canakkalea

21 Biga22 Bandırmaa

23 Balıkesir24 Edremit25 Dikili26 Akhisara

27 Simav28 Salihli29 Manisa30 I: zmira

31 Aydına

32 Muglaa

33 Bodrum34 Fethiye35 Antalyaa

36 Manavgat37 Alanya38 Silifkea

39 Mersina

40 Adanaa

41 I: skenderun42 Antakya

Continental Mediterranean (CMED) 43 Islahiye44 Kahramanmaras45 Gaziantep46 Kilisa

47 Malatyaa

48 Elazıga

49 Mus50 Adıyaman51 Sanlıurfa52 Sivereka

53 Diyarbakıra



Table I. (Continued)

Region Station

Number Name

54 Mardin55 Siirta

56 Cizre

Mediterranean Transition (MEDT) 57 Kutahyaa

58 Usaka

59 Burdur60 Ispartaa

Continental Central Anatolia (CCAN) 61 Kastamonua

62 Merzifona

63 Amasya64 Sebinkarahisara

65 Coruma

66 Tokat67 Yozgat68 Sivasa

69 Eskisehira

70 Ankaraa

71 Sivrihisara

72 Afyona

73 Polatlı74 Kırsehira

75 Kayseri76 Ilgına

77 Aksaray78 Nigde79 Konyaa

80 Karamana

81 Ulukısla

Continental Eastern Anatolia (CEAN) 82 Ardahan83 Karsa

84 Sarıkamısa

85 Igdır86 Agrı87 Gumushane88 Bayburta

89 Erzincan90 Hınıs91 Vana

Other stations 92 Ordub

93 Dortyolb

94 Tatvanb

95 Bitlisb

96 Catakb

97 Hakkarib

98 Aksehirb

99 Erzurumb

a Stations that also used for the study of the SO, in addition to rainfall variation.b Stations that used only for the study of rainfall climatology.

spectral estimates have been found by smoothing the raw spectral estimates with a three-term weightedaverage of the Hanning method. In order to examine whether the mean distribution patterns of rainfalloccurrence areas are sensitive to the variations at the heights of 700 hPa geopotential level, mean annual


MURAT TU8 RKES656

Table II. List of year 0s of the warm and cold events in the SO during the 1930–1993period

Warm events

1930 1932 1939 1941 1951 1953 19571965 1969 1972 1976 1982 1986 1991

Cold events

1931 1938 1942 1949 1954 1964 19661970 1973 1975 1978 1988

and seasonal 700 hPa height values were normalized. Then, by using the Pearson’s correlation coeffi-cient (r), a correlation analysis was made between the normalized rainfall anomalies and the normal-ized 700 hPa height anomalies for the period 1952–1993. Trend and correlation analyses were alsocarried out to find the possible tendency in the series of depression counts, and to make an objectivecomparison between the country-wide original rainfall series and the series of depressions, even thoughlength of depression series was very short.

For analysing if the rainfall anomalies over Turkey are associated with the extreme SO events, 48stations were selected, having almost equal length of winter data (1929–1993), and used to constructthe mean winter rainfall anomaly series, respecting extremes in the SO with respect to long-termaverage of the overall record. Annual and seasonal mean rainfall anomaly series of all of Turkey(91-station average series) have been also studied. The cold and warm events in the SO have beenanalyzed separately by using a composite technique, previously used by Kiladis and Diaz (1989), toexamine the climatic signals associated with either extreme of the SO. The list of the year 0s of thewarm and cold SO events since 1930 is given in Table II, based on the 26 warm and 22 cold events ofthe SO listed in Fraedrich and Muller (1992), with an addition of the recent warm event 1991.Following Rasmusson and Carpenter (1982), the year 0 of a warm event was defined by Kiladis andDiaz (1989), ‘as the year when the Southern Oscillation Index (SOI) changes sign from positive tonegative, and when central and eastern equatorial Pacific SST anomalies become strongly positive’,and cold event year 0s, ‘as having the opposite characteristics.’ In the study, composite rainfallanomalies for the winter series of 48 stations and for the annual and seasonal area-averaged series ofTurkey have been computed separately for the year −1s, the year 0s and the year +1s of the warmand cold events since 1930.

The significance of the difference between the means of composite rainfall anomalies correspondingto the warm and cold SO events has been checked by a one-sided Student’s t-test, with the nullhypothesis being that rainfall anomalies do not differ between the warm and cold events. A compari-son of composite means of the extreme SO events with the long-term average rainfall has also beenmade by means of the two-sided Cramer’s tk-test, based on the null hypothesis of no significantdifference between the means of warm and cold events and the long-term average of whole period.The test statistic tk is distributed as Student’s t (WMO, 1966). Any signal from both tests has beenconsidered if it was significant at the 5% significance level of the t distribution. The percentage ofconsistent signals (PCS), defined as the percentage of events having anomalies of a sign consistent withthe composite anomaly (Kiladis and Diaz, 1989), have been calculated in order to address the questionof whether a signal is dominated by a few large anomalies.

The gridding method of kriging has been used in order to interpolate the randomly scattered datafor the production of isopleths in all spatial distribution maps. Application of the kriging method tothe stations’ data has been made by means of a mapping package.



4. RESULTS

4.1. Principles of Turkey’s rainfall climatology4.1.1. Spatial distribution of mean rainfall totals. Zonality of the long-term mean rainfall totals is most

pronounced for winter. Mean winter rainfall decreases from coastal belts to interiors by varying fromabove 650 mm along the Eastern Black Sea and the Western Mediterranean to below 100 mm overcontinental Central Anatolia and Eastern Anatolia regions (Figure 2(a)). The Western Mediterraneansub-region is the most rainy area of Turkey in winter, with about 754 mm mean rainfall total in thestation of Manavgat. Daily highest rainfall total was measured at the station of Marmaris, with 466.3 mmin December 1992. High rainfall amounts along the Black Sea and the Mediterranean coasts areassociated mainly with the North-eastern Atlantic originated mid-latitude depressions and the Mediter-ranean depressions, respectively. Orographic rains on the windward slopes of the Taurus Mountains andthe Northern Anatolia Mountains also contribute to increasing rainfall amounts. The low rainfallamounts in the interior regions are attributed to the fact that these regions are somewhat protected fromthe effects of the rain-bearing air masses. These masses lose their moisture and get drier adiabaticallyduring the passage over the high mountain ranges. The effect of regional high pressure conditions that aresupported by the Siberian high invading westwards during the course of winter, has also an important roleon the lower rainfall particularly in the eastern part of Turkey where topography is considerably high.

Mean spring rainfall amounts range from below 150 mm over the Aegean Coast with the NorthernMarmara and the Central Anatolia connected with the Middle Mediterranean Coast to above 300 mm onthe Eastern Black Sea Coast and the south-east corner of the country (Figure 2(b)). Most of the countrygets a rainfall ranging from below 150 mm to about 200 mm in spring. In addition to the frontal rains,higher rainfall amounts over the wetter areas with a 200–350 mm mean are attributed to high topographyand their exposures to the dominant airstreams, and the local convective activities, respectively. Distribu-tion of the mean summer rainfall is characterized by the one direction zonality that differs from otherseasons, which increases from below 5 mm over the Syrian border to above 450 mm on the Eastern BlackSea (Figure 2(c)). Approximately half of the country receives a mean rainfall less than 50 mm in summer.Higher summer rainfall over the Eastern Black Sea and the high North-east Anatolia is related closelywith the postfrontal and orographically induced rainfall, and the local convective showers with thunder-storms, respectively, in addition to the frontal rains. Mean autumn rainfall totals generally decrease fromcoastal areas to interiors, as in the winter (Figure 2(d)). Eastern Black Sea is the wettest area, with about795 mm seasonal mean rainfall at the station of Rize. Rainfall amounts about 100 mm take up much ofthe country, particularly continental interiors. With respect to amount and distribution pattern of rainfall,the mean autumn rainfall is quite similar to the mean winter rainfall over the Black Sea region, with astrong gradient along the Northern Anatolia Mountains. The frontal depressions originated in theNorth-eastern Atlantic and the northerly airstreams from the ridging high behind these fronts bring heavyprecipitation to this region. Mean autumn rainfall over the Mediterranean and the Aegean regions differfrom winter rainfall, because the Mediterranean-type frontal cyclones are not so active in autumn as muchas in winter. In addition to direct effects of the fronts in all seasons, the north-west and north-east coastbordering the Black Sea obtains rainfall from post-frontal ridging high pressure cells, or when stormtracks lie anomalously to the south, producing onshore flow there. On the other hand, much of the westand south-west coasts of Turkey obtain rainfall in pre-frontal conditions, particularly during the cool partof the year, when south-westerly flow is present and storms track further north. The mean annual rainfallcondition is not given here, because it was previously examined in Turkes (1996b,c).

4.1.2. Seasonality of mean rainfall totals. A map of the rainfall seasonality over Turkey is shown inFigure 3. Following the definition by Glantz (1987), seasonality index (SI) at a station is computed bysummation of the absolute deviations of mean monthly rainfall from the overall mean and dividing it bythe long-term mean annual. The index particularly highlights the contrast of the rainfall amounts betweenthe seasons over different regions of Turkey. Index values greater than 0.50 coincide with the rainy-winterrainfall regime particularly over the MED region. On the contrary, index values smaller than 0.35correspond with uniform rainfall regimes over the BLS region and the Northern Marmara, respectively.


MURAT TU8 RKES658

Figure 3. Geographical distribution of the seasonality index values for 99 stations in Turkey

The winter rainfall maximum is most characteristic over the western and southern parts of Turkey,whereas autumn and spring rainfall maximums are dominant over the BLS region and the continentalinterior regions, respectively. Erzurum-Kars (North-eastern Anatolia) sub-region receives the majority ofits annual precipitation during the spring and summer months. Contribution of the winter rainfall to theannual total is greater than 40% over the MED and CMED rainfall regions with about 64% country-widemaximum at Antalya and less than 20% over the North-eastern Anatolia with about 12% minimum atArdahan (Figure 4(a)). Spring rainfall contributes more than 30% of the annual rainfall total almost all

Figure 4. Geographical distribution of the percentage contribution of mean seasonal rainfall amounts to mean annual total for 99stations in Turkey



Figure 5. Geographical distribution of the year-to-year variability (CV in percentage) for seasonal rainfall totals of 99 stations inTurkey

over the continental interiors (Figure 4(b)). In summer, maximum rainfall is concentrated over theNorth-eastern Anatolia with about 40% maximum at Ardahan (Figure 4(c)). The summer rainfallcontributes only less than 5% of the annual rainfall total throughout the CMED region, with about 0.5%country-wide minimum at Cizre, and most of the MED region. The contribution of the autumn rainfallincreases from below 15% of the annual total, along the Turkey–Syria border, to above 30%, along theBlack Sea Coast (Figure 4(d)). Minimum contribution of the autumn rainfall is observed as about 14%of the annual total at Cizre.

4.1.3. Spatial distribution of rainfall 6ariability. Mean condition of the year-to-year variability of annualand seasonal rainfall totals over Turkey is examined by the coefficient of variation (CV) for 99 stations.The CV is calculated by taking the standard deviation as a percentage of the long-term mean. Thecoefficients of variation (CVs) of winter rainfall decrease from western, southern and eastern regions ofthe country to the north and the Central Anatolia by showing a complex distribution pattern rather thansimple zonation. The CVs are quite above 35% over most of the CEAN, MRT, MEDT and MED rainfallregions (Figure 5(a)). Year-to-year variability of the spring rainfall varies from above 40%, over most ofthe MED and CMED regions, to below 25%, over the Eastern Black Sea Coast (Figure 5(b)). The highestvariability is at Adıyaman with about 59% in the CMED region, whereas the lowest variability is seen atRize with about 22% on the Eastern Black Sea Coast. Variability of the summer rainfall is above 80%over the MED region except at Iskenderun district and the CMED region and is below 35% on theEastern Black Sea, by showing one direction zonation from south to north (Figure 5(c)). The highestvariability is at Fethiye with about 179% in the MED region, whereas the lowest variability is at Rize withabout 28%. Areas with coefficients higher than 100% are prone to prolonged and more intensive summerdryness, because of the low mean rainfall. Variability of the autumn rainfall ranges from greater than 60%on the Mediterranean Coast to lower than 30% on the Eastern Black Sea Coast as in the spring andsummer seasons (Figure 5(d)). Variability of the annual rainfall decreases from the southern part of thecountry, with a seasonal rainfall regime, to the Black Sea Coast, where a uniform rainfall regimedominates (map not shown). The CVs are above 25% over most of the MED and almost all over theCMED regions. The CVs smaller than 25% have the largest geographical extent by covering approxi-mately four-fifths of Turkey.


MURAT TU8 RKES660

4.2. Variations in rainfall series

4.2.1. Trends and fluctuations. Long-period fluctuations in the annual and seasonal normalized rainfallanomaly series of 91 stations are examined by means of the 9-point Gaussian filter to eliminate thevariations shorter than about 10-year periods. Numbers of the stations having a significant secular trendin the mean are also given in Table III. Results from the Mann–Kendall test for 91 stations and visualinterpretations on the time-series plots (not presented) are summarized as follows. (i) Annual rainfallseries of many stations show a low-frequency fluctuation with a decreasing mean over the study period,and are generally characterized by a longer run of the dry (relatively drier than long-term average)conditions from early 1970s (Turkes, 1996b). Observed downward trends are significant at the 15 stations.(ii) Winter rainfall series are generally dominated by a low-frequency fluctuation with a decreasing meanin many stations and in some stations with an unchanged mean. The wet (relatively wetter than long-termaverage) conditions during the 1960s are followed by a general downward trend dominated by longer runsof drier anomalies from the late 1960s in many stations and in some stations from the late 1970s. Fourteenstations are characterized by a significant decreasing trend. The significance of most of downward trendsin the annual and winter series is at the 1% level in the stations of the MED region. (iii) Spring seriesgenerally indicate a high-frequency variation with an unchanged mean in many stations and in somestations with a slightly increasing mean. A general upward trend from the early 1940s to the late 1960sin many stations and in some stations to 1970s was evident. This increased rainfall period was generallyfollowed by a downward trend. However, only four stations experience any significant trend. (iv) Summerrainfall series generally tend to increase in some stations and to show unchanged situation in manystations. The longer runs of negative anomalies are a dominant characteristic, even though individualhighest positive anomalies are also observed in this season. Seven stations, almost all of which take placein the CMED and CCAN regions, indicate a significant upward trend. (v) Autumn series exhibit arandom run of anomalies at different time periods in many stations, mostly with an unchanged mean, andgenerally increased during the last 10 years in many stations, except those over the MED and MEDTregions. There is no significant trend that characterizes this season.

4.2.2. Year-to-year persistence. It was already reported that significant positive coefficients from theWald–Wolfowitz serial correlation test are indicative of the existence of a low-frequency fluctuation in theannual and winter series of Turkey, while significant negative coefficients indicate the existence of a highyear-to-year variability in the spring series (Turkes, 1996a,b). The lag-one serial correlation (L-1SC)coefficient, r1, has been computed for 91 stations and rainfall regions to assess the statistical nature and

Table III. Number of stations indicating a significant trend and/or serial correlation in the normalized rainfallseries at the 5 or 1% level, according to the Mann–Kendall (M–K) and lag-one serial correlation (L-1SC)

tests

SpringWinterRegion AnnualAutumnSummer

L-1SC M–K L-1SC M–K L-1SCM–K L-1SC M–K L-1SC M–K

+ − + − + − + − + − + − + −+ − + − + −

BLS 1 11 1 22116MRT

1 2 1 1 6 414188MED4 4 1 1 1 2CMED 3 5 1

1MEDT 21 11611CCAN 2 4 6126

1 1 1 1CEAN 1 5 1 2

18 7 9Total 4 17 1716 31 4

+, upward trend from the M–K test and positive correlation coefficient from the L-1SC test; −, downward trend from theM–K test and negative correlation coefficient from the L-1SC test.



magnitude of the observed persistence in variations of the rainfall anomaly series from year to year. TheL-1SC is positive at most of the stations in winter, except at the stations of the BLS region. There is asignificant persistence of winter rainfall variations from year to year in 31 stations, eight of which are inthe stations with winter-maximum Mediterranean type rainfall regime (Table III). Annual rainfall seriesof 17 stations also depict a significant positive correlation. However, the L-1SCs are negative at almostall stations in spring, in which 18 of them are statistically significant. This statistic indicates that springrainfall series contains a high-frequency oscillation. Both negative and positive L-1SCs are found for thesummer rainfall series, in which only nine of them differ significantly from zero. The majority of theautumn rainfall series seems to be statistically random against the serial correlation.

A simple Markov-type persistence is found in approximately 11 stations of the rainfall anomaly seriesfor winter, when the exponential relationships are that of r2$r2

1 and r3$r13 are considered. The Markov

persistence is the dominant characteristic particularly in those winter rainfall series of the stations, whichexist over the colder and more continental north-eastern part of the CEAN region. Desired relationshipsfor the Markov persistence are also seen to be satisfied approximately in the summer series of fourstations, in one autumn series, and in the annual series of four stations. On the other hand, the significantnegative L-1SCs in the spring series are associated closely with a marked short-period oscillation ratherthan a persistence, which comes out in the spectral analysis.

4.2.3. Power spectrum. Interpretations of the time-series graphs and results of the serial correlation testshave indicated the presence of some particular fluctuation types in the station rainfall series, especially inthose of low-frequency fluctuations in winter and high-frequency oscillations in spring. Power spectrumanalysis has been applied to the series of non-smoothed annual and seasonal normalized rainfallanomalies in order to detect the dominant cycles in those observed fluctuations. A procedure of the testsof statistical significance proposed by WMO (1966) has been carried out to objectively assess the resultsof power spectrum analysis. However, only the results for the area-averaged normalized series of therainfall regions and all of Turkey have been given here in detail. Significant positive L-1SC coefficientsand the desired exponential relationships have been approximately found for the area-averaged winterrainfall series of the MRT, MED and MEDT regions, and all of Turkey. Appropriate ‘null’ hypothesiscontinuum to the spectrum has been assumed as Markov ‘red’ noise for those winter series. Other regionalseries having nonsignificant positive L-1SCs or no Markov-type persistence, and significant but negativeL-1SCs (e.g. spring series) have been formulated by a ‘null’ continuum that is of ‘white’ noise. This is ahorizontal straight line of the value of 0.046, with the values of 0.082 and 0.098, at the 90 and 95%confidence levels, respectively.

The appropriate ‘null’- continuum and its 90 and 95% confidence levels are plotted superposed on thespectrum and are shown in Figure 6 (a,b) for the winter and spring series. The 90% confidence level istaken into consideration in order to assess the spectral bands that are below 95% level but mostly formsome successive apparent cycles in addition to individual marked cycles. The estimated spectral peaks thatexceed the 90% and the 95% confidence limits of the appropriate ‘null’ continuum of the spectrum arepresented in Table IV. The major spectral peaks for the winter rainfall series occur generally within thespectral bands with the cycles of 3–3.2 years, 6–7 years, 7–8.4 years and 14–21 years. However only thespectral peaks, at the lag-13 and the lag-14 corresponding to the periods of 3.2 and 3 years, respectively,significantly deviate from the ‘red’ noise continuum at the 95% confidence level in the CCAN and CEANrainfall regions, and in all of Turkey. For the spring rainfall series, marked spectral peaks are seen mainlywithin the spectral bands with the cycles of from 2.0 to 2.2 years and from 4.2 to 4.7 years. The spectralpeaks with the cycles of 2.1 and 2.2 years in the CCAN region and all of Turkey and the 2.1-year peakin the CMED region exceed the 95% confidence limits of the ‘white’ noise continuum. The longer cyclesof 2.5 years, 3.2 years and 3.5 years in the BLS region and the cycle of 4.2 years in the MRT region arealso significant at the 95% confidence level. For the summer rainfall series, only the spectral peak at thelag-11 corresponding to a cycle of 3.8 years differs from the 95% confidence limit of the ‘white’ noisecontinuum of the MRT and MEDT regions, although there are several prominent spectral peaks withinthe spectral band with the periods of 3.8–4.2 years of many regional series. The 3.8-year cycle is also


MURAT TU8 RKES662

Figure 6. (a) Power spectrum of the winter area-averaged rainfall series. —, ‘null’ continuum with the 90% (- - -) and 95% (– – –)confidence limits of the spectrum. (b) As in (a) but for the spring area-averaged rainfall series



Figure 6 (Continued)


MURAT TU8 RKES664

Table IV. Significant periods with corresponding lags in the area-averaged normalized rainfall series for the rainfallregions and Turkey as a whole. Cycles in bold italic are significant at the 95% confidence level

Region Winter Spring Summer Autumn Annual

Lag Cycle (year) Lag Cycle (year) Lag Cycle (year) Lag Cycle (year) Lag Cycle (year)

BLS 20 2.1 12 3.5 14 3.0 16 2.613 3.2 15 2.8 17 2.517 2.5 18 2.3

MRT 2 21.0 9 4.7 10 4.2 19 2.2 3 14.03 14.0 10 4.2 11 3.8 20 2.1 11 3.8

MED 2 21.0 9 4.7 11 3.8 17 2.5 2 21.010 4.2 12 3.5 3 14.019 2.2

CMED 3 14.0 19 2.2 18 2.3 2 21.0 3 14.04 10.5 20 2.1 19 2.2 19 2.26 7.0 21 2.0 20 2.17 6.0

13 3.2

MEDT 14 3.0 11 3.8 10 4.2 9 4.7 2 21.011 3.8 3 14.015 2.8

CCAN 5 8.4 19 2.2 7 6.0 1 42.06 7.0 20 2.1 2 21.0

13 3.2 21 2.014 3.0

CEAN 5 8.4 9 4.7 9 4.7 10 4.213 3.2 10 4.2 10 4.214 3.0 19 2.2

TURKEY 13 3.2 19 2.2 11 3.8 10 4.2 3 14.014 3.0 20 2.1 14 3.0

evident in the summer series of Turkey. The autumn rainfall series contain several spectral peaks thatoccur at both shorter and longer wavelengths. However, only the spectral band with the cycles of 2.8–3years exceeds the 95% confidence limit of the ‘white’ noise continuum of the BLS region. The cycles of2.2–2.3 years, 2.5–2.6 years, ca. 4 years and 14–21 years seem as being major spectral peaks for theannual series. The spectral band with the cycles of 2.3–2.6 years in the BLS region and the individualpeaks with the cycles of 4.2 years, 14 years and 21 years in the CEAN, MED and MEDT regions aresignificantly different from the 95% confidence limit of the ‘white’ noise continuum. Spectral peakscentred on the lag-19 with the cycles of 2.2 years and the lag-3 with the cycles of 14 years are evident inthe annual rainfall series of Turkey.

4.3. Some atmospheric influences on rainfall 6ariations

4.3.1. Relationship with 6ariations of 700 hPa geopotential heights. Mean distribution of the rainfalloccurrence areas that could be sensitive to the lower tropospheric variations has been examined bycorrelating the rainfall anomaly series of 91 stations with the normalized mean 700 hPa geopotentialheights at Istanbul for the period 1952–1993. Annual rainfall anomalies are significantly correlated withthe variations of the 700 hPa height anomalies at most of the stations over the MED and CMED regions,where pre-frontal processes are effective at least in the cool half of the year (Figure 7(a)). Negativecorrelation coefficients are significant at the 1% level, at the south-west corner of the country. Winterrainfall anomalies are significantly and most strongly correlated with the 700 hPa height anomalies



(Figure 7(b)). Negative correlation coefficients are significant at the 1% level throughout Turkey exceptthe BLS region, where post-frontal processes are dominant, and the far eastern margin of the country.Spring rainfall anomalies are significantly but not strongly correlated with the 700 hPa height anomalies,over the southern MRT, the western and middle BLS and the MED regions, and over the middle of theCEAN region (Figure 7(c)). Correlation coefficients are stronger with a 1% significance, around the Gulfof Iskenderun and on the south-west coast of the country. Summer anomalies (not shown) have weakpositive correlations particularly over the central and eastern regions of Turkey. Significant but not strongcorrelations are seen over the western BLS region, and northern and southern parts of the MRT andnorthern part of the MED regions. Autumn correlation coefficients exhibit a complex distribution pattern(Figure 7(d)). Significant negative correlations could be considered as an indicator of the rainfalloccurrence areas mainly affected by the atmospheric disturbances (e.g. upper level lows and troughs),whereas positive and weak negative correlations could be assessed as an indicator of the rainfalloccurrence in areas, where physiographic factors (e.g. topography, exposure and continentality) are moreeffective than, or as important as, atmospheric conditions.

The series of the 700 hPa geopotential height anomalies of Istanbul are significantly correlated with the500 hPa geopotential height anomalies of Istanbul, at the 1% significance level (in winter, r=0.92; inspring, r=0.73; in autumn, r=0.64; and annually, r=0.75), except the summer that is at the 5% level.As a result of these strong relationships between two standard levels, similar relationships were generallyfound between the normalized rainfall anomalies of 91 stations and the normalized 500 hPa geopotentialheights of Istanbul, except for summer and autumn series (all not shown). Summer pattern of 500 hPadiffers from 700 hPa by having significant positive correlations over the CMED region, and the autumnpattern is different, with significant negative correlations over the central and northern regions of thecountry.

4.3.2. Variations in series of geopotential heights. Time-series plots of the annual and seasonal meangeopotential heights at the 700 hPa geopotential level are shown in Figure 8 and test statistics of the 700hPa and 500 hPa geopotential height anomalies are given in Table V for Istanbul. In winter, 700 hPageopotential level heights show a marked low- frequency fluctuation about an increasing mean at Istanbul(Figure 8(a)). Higher geopotential values occur during the periods of 1957–1961, 1971–1977 and

Figure 7. Distribution of the correlation coefficients between the variations of normalized 700 hPa height anomalies at Istanbul andthe variations of normalized rainfall anomalies of 91 stations


MURAT TU8 RKES666

Figure 8. Variations of the mean annual and seasonal 700 hPa geopotential heights at Istanbul, with long-term average ( · · · ) and9-point Gaussian filter (—) with padded ends

1987–1993, whereas lower values are observed during the periods of 1953–1956, 1963–1970 and1979–1986. Spring geopotential heights exhibit a high year-to-year variability with a slightly increasingmean (Figure 8(b)). Higher values are apparent recently in 1989–1990. Summer geopotential heightsindicate a significant upward trend at the 1% level (Table V). Almost all geopotential heights are wellabove the long-term mean in 1985–1993 (Figure 8(c)). In autumn, higher geopotential heights are evidentin 1957–1966, while lower heights generally dominate over the periods of 1967–1974 and 1985–1992(Figure 8(d)). As a result of the general trends towards the higher 700 hPa and 500 hPa geopotential levelheights in seasonal series except in autumn, annual geopotential heights of Ankara (not shown) and ofIstanbul tend to increase during the study period (Figure 8(e)). The upward trends in the annual series of700 hPa and 500 hPa geopotential anomalies at Istanbul are significant at the 5% level (Table V). Higher700 hPa geopotential heights are observed during the recent periods of 1977–1993 at Ankara and1987–1993 at Istanbul. A common feature of the annual and seasonal series is that positive anomaliesdominate over the period from the late 1970s to early l990s.



Table V. Results of the Mann–Kendall (M–K) and lag-one serial correlation (L-1SC)tests for the series of 700 hPa and 500 hPa geopotential height anomalies at Istanbul

Time 700 hPa 500 hPa

M–K (u(t)) L-1SC (r1) M–K (r1) L-1SC (u(t))

Winter 1.68 0.41** 1.35 0.25*Spring 1.17 −0.04 1.37 −0.11Summer 2.76** 0.32* 1.72 −0.04Autumn −0.07 −0.09 0.57 −0.18Annual 2.08* 0.07 2.02* 0.05

* Significant at the 5% level; ** significant at the 1% level.

The L-1SCs are significant in the winter series of 700 hPa and 500 hPa geopotential height anomalies(Table V). The 700 hPa geopotential series also contains a Markov-type persistence. This high positiveserial correlation is related closely with an observed low-frequency fluctuation in time-series plot (Figure8(a)). Because all the geopotential series except winter are random against the Markov-type persistence,‘null’ continuum of the spectrum for those series has been assumed as a ‘white’ noise. The estimatedspectral peaks that deviate from the 90% and 95% confidence limits of the ‘null’ continuum aresummarized in Table VI. Spectral analysis shows a marked periodicity in the winter series with the cyclesof 13 years (Figure 9). Only the 13-year peak within the 500 hPa spectrum is significantly different fromthe 95% confidence limit of the ‘red’ noise continuum. The cycles of 26 years are not taken into account,because the length of geopotential height series is not long enough for an interpretation on such a longeryear peak of the spectrum. The spectral peaks with the cycles of 2–2.2 years and 3.3–3.7 years show upin the spectrums for both spring series of geopotential levels. The marked oscillations with the period of3.7 years in both spring series exceed the 95% confidence limits of the ‘white’ noise continuum. In autumn,the marked spectral peaks of the 500 hPa standard level occur within the spectral band with the cycles of2.9–3.3 years, in which the 2.9-year peak is significant at the 95% confidence level. The cycles of 3.3 years,3.7 years and 13 years, which also show up at the spectrums of spring and winter series for bothgeopotential height levels, respectively, are apparent at the annual 500 hPa geopotential heights.

When a subjective comparison is made between the prominent periods of normalized rainfall andgeopotential series, it is revealed that the cycles of 2–2.2 years in the spring rainfall series and bothgeopotential heights series, and the cycles of 14 years and 13 years in the annual and winter series of

Table VI. Significant periods with corresponding lags in the series of the normalized 700hPa and 500 hPa geopotential height anomalies at Istanbul. Cycles in bold italic are

significant at the 95% confidence level

500 hPa700 hPaTime

Lag Period (year) Lag Period (year)

Winter 2 13.0 2 13.0Spring 7 3.7 3.77

3.3 88 3.32.2122.013

Summer 4 6.58 3.3Autumn9 2.9

Annual 2 13.07 3.7

3.38


MURAT TU8 RKES668

Figure 9. Power spectrum of the annual and seasonal 700 hPa geopotential height series at Istanbul. —, ‘null’ continuum with the90% (. . .) and 95% (– – –) confidence limits of the spectrum

rainfall and geopotential heights, respectively, should be given consideration. The biennial oscillationswith a 2-year cycle and the cycles of 2.1–2.2 years in the spring rainfall series, which exist particularly inthe continental type regions and all of Turkey, appear to be associated with the similar quasibiennialoscillations in the spring 500 hPa geopotential heights. It seems that the significant 3.3–3.7-yearperiodicity in both spring geopotential heights series exactly does not appear within the oscillations in thearea-averaged rainfall series. However, the periods of 3.2–3.5 years and 3.8 years in various regions, butnot in all of Turkey, may correspond with this 3.3–3.7-year periodicity of the 700 hPa and 500 hPageopotential heights. The low-frequency fluctuations with the cycles of 13 years in both winter geopoten-tial heights series and in the annual 500 hPa geopotential series seem to be reflected in a similar fluctuationbut with a 14-year periodicity in the annual and winter rainfall series of Turkey as a whole, and in theregional series having the Mediterranean type rainfall regime. Quasibiennial oscillations in both Turkey’s



spring rainfall and geopotential heights and 14-year and 13-year periodicity in annual and winter rainfalland geopotential heights, respectively, may have been linked with the quasibiennial and 13.2-year cyclicityin the monthly zonal index series of surface pressure for the zone 35°–65° N, which were reported byKozuchowski (1993). Kozuchowski found the quasibiennial and 13-year cyclicity appeared also inseasonal values of the zonal index.

4.3.3. Role of 6ariations in extra-tropical depressions acti6ities. In this paper, classification of thedepression tracks and the basic depression data were taken from Deniz and Karaca (1995). Main tracksof the mid-latitude and Mediterranean depressions that reached the Turkey region during the period1979–1992 are shown in Figure 10. Mean seasonal frequencies of the depressions according to the pathsand mean seasonal distributions of all depressions were also studied, but not given here in detail. Frontaldepressions make a major contribution to the rainfall occurrence and distribution over Turkey throughoutthe year, and control its spatial and temporal variations over a long time period. The mean number ofdepressions reaching Turkey is the highest in winter with ca. 32% contribution and is the lowest insummer with ca. 17%. Depressions following path III have the highest mean annual frequency with 49.4%(Figure 10). The frequency of path III depressions is also the highest in all seasons, with ca. 56% wintermaximum. The path I depressions mainly source from the North Atlantic and the Scandinavia, and reachthe Turkish region by passing the Black Sea through Eastern Europe and Russia. The path II depressionsoriginate over the North Atlantic and arrive at Turkey via Middle Europe and the Balkans. Northernareas of Turkey are also affected by the southern extension of the cold fronts of the mid-latitudedepressions, even when they pass over far north of the Black Sea Basin. Most of the Mediterraneandepressions, whether occurring over the Mediterranean Basin or developing by modification of mid-lati-tude depressions when they enter into the western and central basins or falling into the Gulf of Genoa andthe Adriatic, follow path III across Italy and Greece to Turkey. Frontal storms associated with path IIIaffect most of Turkey throughout the year, but not so much in summer, and cause intensive and abundantrain showers with thunderstorms particularly over the Marmara, Mediterranean and Continental Mediter-ranean regions. The path IV includes depressions that both enter into the western and middle basinsdirectly from the mid-Atlantic, via Spain and north-west Africa, and some from the Gulf of Genoa.During the period from June to October, mid-latitude cyclonic storms are generally ineffective and rare

Figure 10. Main tracks of the mid-latitude and Mediterranean depressions reached Turkey during the 1979–1992 (re-plotted on anew base map with some modifications from Deniz and Karaca, 1995). Mean annual percentage frequencies are given in brackets

on the map


MURAT TU8 RKES670

Figure 11. Variations of the annual and seasonal total number of the depressions reaching Turkey for the period 1979–1992

over the Mediterranean and the Turkish region, because of northward migration of the polar jet-streamto the polar front zone in the latitudes of 55–60° N and beyond it. In summer, maritime polar andMediterranean air masses connected to the Atlantic which originated moist airstreams are replaced by thecontinental tropical air masses connected with relatively dry airstreams from mid-north Africa andArabia. In addition to these dynamic-originated depressions, westward and north-westward extensions ofthe thermal-originated Asiatic monsoon depression, which in Turkish is referred to as the Basralow-pressure, also has some importance for Turkey from May to October. In summer months, when afresh cool (cold) airstream invades the Eastern Mediterranean, a weak frontal depression may developwithin the north-west extension of the Asiatic monsoon low to Turkey.

According to the Mann–Kendall test, annual total number of the depressions that reached the Turkeyregion during the period 1979–1992 shows a significant downward trend at the 5% level in conjunctionwith significant decrease in the depression numbers of winter and summer (Figure 11). There is not amarked change for the depression frequency in spring and autumn. It is also seen that annual numbersof depressions decrease in the paths of II, III and IV (Figure 12), but only path IV has a significantly



Figure 12. Variations of the annual total number of the depressions according to the paths

decreased frequency at the 5% level. The decrease in the number of depression activities is also significantat the 5% level in autumn for path III (Figure 13) and in winter for path IV. Decrease of cold seasondepression frequency was also evident for the lower latitudes of the Northern Hemisphere (Serreze et al.,1997). Serreze et al. (1997) reported that cold season cyclone counts for the zone of 30°–60° N decreasedfrom 1966/1967 to 1975/1976, but with generally high values in the subsequent decade, with low valuesfrom 1988/1989 onward.

Turkey’s 91-station average rainfall series are strongly correlated with the series of total number of alldepressions in winter (r=0.76) and in autumn (r=0.62), which is significant at the 1% and the 5% level,respectively. In summer, however, country-wide average summer rainfall is negatively, but not signifi-cantly, correlated with the total number of depressions with r= −0.24. Some series for the rainfall regimeregions also have significant positive correlations in winter, e.g. Mediterranean region (r=0.72). Alsotaking into account the negative correlation pattern between the rainfall anomalies and the geopotentialheight anomalies, particularly in winter, it seems that general decreasing trends with recent dry anomaliesin winter and annual series are closely related to the decreased depression activities and the generalupward trends at the lower tropospheric heights over the Turkish region.

4.3.4. Relationship with extremes in the Southern Oscillation. According to the Student’s t-test,differences between the warm event and cold event anomalies do not reveal the existence of any significantSO signal in the annual and seasonal rainfall anomaly series for all of Turkey (Table VII). Most of theindividual warm and cold event composite anomalies in these series have a positive sign particularlyduring the year −1. According to the Cramer’s tk test, any individual composite means during the SOevents also deviate significantly from the long-term average rainfall of overall record (normal rainfall).The opposition of the sign of the warm minus cold event composite rainfall differences (SCD) is evidentin all series except winter between the year −1 and the year 0. The SCD also tends to be opposite inwinter, summer and autumn series between the year −1 and the year +1.


MURAT TU8 RKES672

Figure 13. Variations of the seasonal number of path III depressions

Whole results of the SO analysis with the Student’s t- and Cramer’s tk-tests during the year −1, year0 and year +1 have been given in the Appendix (Table VIII), for the winter series of 48 selected stations.Results of the analysis for whether the rainfall anomalies related with the SO extremes have beendocumented by separately assessing the composite rainfall anomalies associated with the warm eventsversus the cold events, and composite mean rainfalls during the warm and cold events with respect to theirlong-term average. The warm minus cold event winter rainfall anomalies over Turkey during year 0 showa decreased rainfall at 54% of the 48 stations. The signal of the drier than normal warm event condition(negative difference) shows up well in the stations of BLS and CMED regions, but being as a coherentarea of the stations without significant signals, and generally in the MED region. The signal of the wetterthan normal warm event condition (positive difference) exists particularly over the MRT region. Anysignals of warm minus cold event winter anomalies during year 0, however, are significant at the 5% level.Stations with opposite signs between the year 0 warm and cold events amount to 54% of total stations.The number of stations with the composite anomalies of positive sign and of negative sign are equal withabout 50% for both the warm and cold events. Any composite means of warm and cold event rainfallsduring the year 0 significantly differ from the normal rainfall.

The signs of warm minus cold event winter responses during the year −1 show dry conditions at 58%of the stations. The drier than normal warm event conditions concentrate over the regions of MRT,MED, CMED, MEDT and CCAN. Wetter conditions dominate over the BLS and CEAN regions. Thereseems only one significant negative signal at the 5% level, which is the station at Afyon in the CCANregion. All of positive signals are not significant. Shares of the stations with positive-signed compositeanomalies in total are of 67 and 79% for the warm and cold events, respectively. The stations with theopposite signs between the warm and cold event anomalies account 50% of the stations during the year−1. Only one composite mean, that for the station of Trabzon, is significantly wetter than the normalrainfall during the year −1 warm events. Positive warm event anomalies, whether significant at the 5%level or not, are dominant responses of rainfall particularly over the northern and western regions ofTurkey, including the BLS and the MRT regions, and some stations in the west of the MED region and



Table VII. Results of the SO analysis for the annual and seasonal rainfall series of Turkey. Thecomposite mean (r) and anomaly (a), percentage of consistent signals (PCS) and number of theyears (n) for the warm and cold events; the sign of the warm minus cold event composite rainfalldifference (SCD) from the t-test, and the combined warm and cold event PCS (total PCS). The

rainfall amount is given in mm

Time Year Warm events SCD Cold events TotalPCS

r a PCS n r a PCS n

Annual −1 647.4 24.3 54 13 + 627.9 4.8 42 12 480 612.4 −10.6 50 14 − 638.4 15.4 67 12 58

+1 638.3 15.3 64 14 + 609.8 −13.2 58 12 62

Winter −1 258.7 9.8 62 13 − 266.4 17.5 58 12 600 244.5 −4.4 57 14 − 252.5 3.6 67 12 62

+1 243.3 −5.6 50 14 + 228.6 −20.3 67 12 58

Spring −1 167.5 2.7 46 13 − 175.6 10.8 75 12 600 171.3 6.5 71 14 + 168.0 3.2 67 12 69

+1 162.9 −2.0 43 14 − 169.1 4.3 58 12 50

Summer −1 61.3 −2.9 46 13 − 64.9 0.7 42 12 440 70.3 6.2 50 14 + 62.0 −2.2 58 12 54

+1 69.0 4.9 50 14 + 61.2 −2.9 67 12 58

Autumn −1 146.6 1.7 38 13 + 138.6 −6.2 67 12 520 141.4 −3.5 50 14 − 147.2 2.3 25 12 38

+1 143.9 −1.0 79 14 − 156.8 11.9 50 12 65

in the north of the CCAN region. Wetter than normal rainfall conditions are more pronounced for thecold events during the year −1. During this period, 12 of the 48 stations have a composite mean rainfallthat is significantly wetter than the normal conditions, at the 5% level. The later situation is more strikingin the MEDT and CCAN region stations.

The warm minus cold event winter responses during the year +1 exhibit increased rainfall at 71% ofthe stations. This tendency of above normal warm event rainfall is not significant at the 5% level, exceptthat in the station of Ilgın. There is also no significant signal for the below normal warm event rainfall.The opposite sign between the year +1 warm and cold event anomalies exist at 33% of the stations,which most of these stations having opposite signs extending over the BLS and MRT regions. During theyear +1, 75% of the warm event composite rainfall anomalies and 88% of the cold event rainfallanomalies are characterized by a negative sign. Drier than normal rainfall condition is significant at fivestations of different rainfall regions during the year +1 cold events. The consistency of the compositecold event anomalies is generally greater than those of composite warm event anomalies except those inthe BLS region.

On the other hand, when the sign of the warm minus cold event composite rainfall differences (SCD)between year −1 (+1) and year 0, and between year −1 and year +1 are considered, it is found that58% (71) of the 48 stations show an opposite signal between the year −1 and the year 0 (+1). Theopposition of the SCD is less pronounced between the year +1 and the year 0, by covering 54% of thestations. The reversal sign among the years −1, 0 and +1 is also assessed for both composite anomaliesof warm and cold events. In the warm events of the SO, composite rainfall anomalies tend to be oppositeat 50% (50) of the stations between the year −1 (+1) and the year 0, and again at 50% between the year−1 and the year +1. During the cold events, opposition of the signs appears at 50% (54) of the stationsbetween the year −1 (+1) and the year 0. The greatest rate of opposition is found at 88% of the stationsbetween the year −1 and the year +1.


MURAT TU8 RKES674

5. SUMMARY AND CONCLUSIONS

The Black Sea and Mediterranean regions are the wettest areas of Turkey. Approximately 40% and 10%of the 91-station country-wide annual rainfall total occur during the cold winter and the warm summer,respectively. Rainfall is more seasonal over the western and southern regions of Turkey with an above40% winter maximum, while it is generally uniform over the Black Sea region with a maximum in autumn.The spring rainfall maximum is widespread over the continental interiors, and the summer maximum ismainly characteristic of the north-eastern region of the country. The CVs of winter rainfall are well above35% over a large area of Turkey. Variability of the spring rainfall is generally above 40% over the MEDand CMED regions. In the autumn, areas with a variability over 40% take up the majority of the country,with coefficients greater than 60% on the Mediterranean Coast. In Turkey both timely and sufficientrainfall in the autumn and spring are vital for sustaining natural systems, water resources and agriculturalactivities, especially those utilizing older methods, i.e. without irrigation. Variability in these seasons,however, is higher than that in the winter. Variability of summer rainfall is above 80% over most of theMED and CMED regions, being related with the lower rainfall totals in this season.

For the annual and winter rainfall series, wet conditions generally occurred during the 1940s, 1960s, late1970s and early 1980s, whereas dry conditions generally dominated over the early–mid 1930s, early–mid1970s, mid–late 1980s and 1990s over most of Turkey. Spring rainfall series of many stations generallyindicated an upward trend from the early 1940s to the late 1960s or 1970s and this period was generallyfollowed by a downward trend. The longer runs of dry anomalies were dominant characteristics of thesummer. Autumn rainfall series generally show a random run of anomalies in many stations mostly witha stable mean, along with some degree of low-frequency fluctuation in some stations. Significantdecreasing trends showed up in the annual rainfall series of 15 stations and in the winter rainfall series of14 stations, mostly over the Mediterranean region. The summer rainfall series tended to increasesignificantly in seven stations, although the total number of depressions that reached Turkey, except pathI depressions, generally decreased and mean 700 hPa geopotential heights showed an upward trend overTurkey in this season. The annual and winter rainfall anomalies of 17 and 31 stations, respectively, hada significant positive L-1SC. Spring rainfall anomalies of 18 stations depicted a significant negative L-1SC.The Markov-type persistence showed up in the winter rainfall series of 11 stations.

The area-averaged winter rainfall series of the MRT, MED and MEDT regions, and all of Turkey werecharacterized by the Markov-type persistence. The power spectrum analysis revealed marked oscillationswith the spectral peaks of 3–3.2 years, 6–7 years, 7–8.4 years, and 14–21 years in the winter series.However, only periods of 3 and 3.2 years in the CCAN and CEAN regions, and in all of Turkey weresignificant at the 95% confidence level. In spring, the quasi-biennial oscillations with the periods of 2.1 and2.2 years in the CCAN region and all of Turkey, the 2.1-year period in the CMED region, the longerperiods of 2.5, 3.2 and 3.5 years in the BLS region and the 4.2-year period in the MRT region all exceedthe 95% confidence limit. In summer, only a cycle of 3.8 years differs from the 95% confidence limit inthe MRT and MEDT regions. In autumn, only the cycles of 2.8 and 3 years in the BLS region weresignificant at the 95% confidence level. For the annual series, the spectral band with the cycles of 2.3–2.6years in the BLS region and the cycles of 4.2, 14 and 21 years in the CEAN, MED and MEDT regionssignificantly differed from the 95% confidence limit. The 2.2-year quasi-biennial oscillation and the14-year long period fluctuation were also apparent in the annual rainfall series of all of Turkey.

The Mediterranean- and north-eastern Atlantic-originated depressions make a major contribution tooccurrence and variation of rainfall over Turkey. These extra-tropical depressions that reached theTurkish region during the period 1979–1992 tended to decrease from 1983 in winter and summer, andannually. The decrease in the number of depressions, particularly in the paths of II and III, has beenimportant for Turkey’s rainfall. The area-averaged Turkey rainfall was significantly correlated with thetotal number of depressions in autumn and strongly in winter. Winter rainfall variations of the MEDregion were also strongly related to variations of the depression activity.

Correlation analysis between the 700 hPa geopotential heights at Istanbul and the rainfall of 91 stationsrevealed that rainfall variations were controlled by inter-annual variations in the lower tropospheric



conditions, except summer, mainly over the western and southern regions of Turkey. In particular,decreasing pattern of winter rainfall variations was found to be significantly linked to variations of the 700hPa geopotential heights, which had a tendency towards the higher atmospheric heights, over most ofTurkey except the BLS region. Winter 500 hPa and 700 hPa geopotential series showed a significant13-year periodicity, at the 95 and 90% confidence level of the ‘red’ noise continuum, respectively. Thecycles of 2–2.2 years and 3.3–3.7 years dominated in the spectrums for both spring series of geopotentiallevels. The marked oscillations with the periods of 3.7 years in both spring series exceed the 95%confidence limits of the ‘white’ noise continuum. In autumn, the cycles of 2.9 years and 3.3 years wereapparent at the 500 hPa geopotential anomalies, in which the 2.9-year peak was significant at the 95%confidence level. The prominent cycles of 3.3, 3.7 and 13 years were found at the annual 500 hPageopotential heights. The biennial oscillations with the period of 2 years and the periods of 2.1–2.2 years,which existed in the continental rainfall regions and all of Turkey, appeared to be associated with thesimilar oscillations in the spring 500 hPa geopotential heights. The periods of 3.2–3.5 years and 3.8 yearsmay have been associated with the 3.3–3.7-year periodicity at both geopotential heights. The low-fre-quency fluctuations with the cycles of 13 years in both winter geopotential series and in the annual 500hPa geopotential heights seemed to be reflected in a similar fluctuation but with a cycle of 14 years in theannual and winter rainfall series of Turkey, and in the regions with the Mediterranean-type rainfallregime. This general similarity in the periodicity for the winter and spring series would have beenmeaningful and relevant to because the spatial and temporal variations of Turkish rainfall, especially inthe winter, were closely related to the paths and frequencies of the depressions and the variability at thestandard geopotential levels. Taking into account the decreased number of depression activities andgeneral upward trend of 700 and 500 hPa geopotential heights, it may be speculated that anticycloniccirculation may have affected the Turkish region more frequently during about the last 20 years.Significant downward trends with the abrupt decreases from the early 1970s, particularly in winterrainfall, could be attributed to possible further northward shift of the polar front than normal position,as a result of the more eastward extension of the drought-dominated subtropical anticyclones from Azoresto Eastern Mediterranean.

The composite warm event means of annual rainfall totals of all of Turkey appeared to be increasedduring the years −1 and +1 and to be decreased during the year 0 in comparison with the long-termaverage (normal) rainfall. The cold event means of both annual and winter rainfall of Turkey tended tobe wetter than the normal rainfall during the year −1 and 0, and to be drier than the normal rainfall year+1. As for the winter rainfall series of 48 stations, the warm minus cold event rainfall anomalies showeda signal of the drier than normal warm event conditions during the year 0 over the BLS and MEDregions. The wetter than warm event conditions dominated over the MRT region. However, any of thesesignals during the year 0 was significant at the 5% level. Composite means of the warm and cold eventrainfalls did not also significantly deviate from the normal rainfall. The drier than warm event conditionsduring the year −1 showed up over the western, southern and central parts of Turkey. Wetter conditionswere seen over the BLS and CEAN regions. Significant negative signal was only found at one station ofthe CCAN region. Of 48 stations 67% and 79% depicted a positive sign anomaly during the year −1warm and cold events, respectively. The composite warm event mean of one station was significantlywetter than the long-term average during the year −1. Positive warm event responses covered a largearea over the northern and western regions. Wetter than normal rainfall conditions were more pro-nounced for the cold events, by having conditions significantly above normal at 25% of the stations. Thesestations showed a coherent region of significantly wetter conditions over the central-west and central partsof the Anatolia Peninsula. The wet signal was found to be significant only at one station, although thewetter than normal warm event responses during the year +1 accounted for 71% of the stations. Duringthe year +1, 75 and 88% of the individual warm event and cold event responses, respectively, weredominated by a negative sign. Significantly drier than long-term average conditions showed up at fivestations during the cold events. The rainfall anomalies were generally more consistent for the compositecold event anomalies. The SCDs revealed that the opposite signal between the year −1 and the year 0(+1) existed at 58% (71) of the stations. The opposition occurred at about half of the stations between


MURAT TU8 RKES676

the year 0 and the year +1. The opposition was striking with 88% of the stations between the year −1and the year +1 of the cold events.

In this paper simple arithmetic averaging technique was used to produce time-series for the rainfallregime regions. This was performed mainly to complete the author’s previous study, with the analyses ofsome new atmospheric variables, along with additional analyses of the previous station-based andarea-averaged series for the rainfall regime regions. Future studies related with the regional rainfallvariations of Turkey will be carried out by using different regionalization and averaging methods toobtain regional rainfall series. The author considers that different techniques of regionalization and ofaveraging may produce different results.

ACKNOWLEDGEMENTS

The author would like to thank to Utku M. Sumer of the TSMS for programming of some computations,and to Ali Deniz and Mehmet Karaca of Istanbul Technical University for giving permission to use theirdepression data. The author also wishes to acknowledge the valuable comments and constructivesuggestions of two anonymous reviewers.

APPENDIX A

The results of the SO analysis for seasonal rainfall of the selected 48 stations in Turkey (Table VIII).

Table VIII. As in Table VII, but for winter rainfall of the selected 48 stations in Turkey. Significant warm and coldevent signals with respect to the long-term average of series are stated on r columns in bold italic

Year SCD Cold events TotalWarm eventsRegion StationPCS

nPCSr ar a PCS n

639.0 −38.3 67 12BLS 64Rize −1 740.6 63.3 62 13 +42125018.2695.50 680.9 3.6 36 14 −

705.3 28.0 50 12 62+1 616.5 −60.8 71 14 −−16.6 50 12Trabzon −1 264.0 38.4 54 13 + 209.0 52

12505.3 58230.90 212.6 −13.0 64 14 −8.3 42 12+1 210.7 −14.9 57 14 − 233.9 50

521258−17.9Giresun 341.0−1 360.2 1.2 46 13 +356.4 −2.5 50 12 500 349.6 −9.3 50 14 −

42 12 583.1362.0+1 322.1 −36.8 71 14 −5658 12Samsun −1 212.9 0.7 54 13 + 208.6 −3.6

−11.5 50 120 194.3 −17.9 64 14 − 200.7 58125014.3 58226.5+1 191.5 −20.7 64 14 −

20.1 46 11Zonguldak −1 410.3 23.4 58 12 + 407.0 52621155−11.4375.40 360.5 −26.4 69 13 −

371.4 −15.5 58 12 52+1 388.5 1.6 46 13 +5612426.4Bolu 173.4−1 186.0 19.0 69 13 +

67 12 620 169.8 2.8 57 14 + 153.9 −13.1501250167.2+1 0.2161.1 −5.9 50 14 −

287.5 17.2 50 12MRT 60Istanbul −1 300.0 29.7 69 13 +67 12 62−8.0262.40 263.6 −6.7 57 14 +

242.8 −27.5 67 12+1 65269.8 −0.5 64 14 +58116451.2Kırklareli 235.2−1 208.7 24.7 54 13 −

178.0 −6.0 58 120 189.6 525.6 46 13 +178.4 −5.6 58 12+1 188.0 544.0 50 14 +



Table VIII. (Continued)

Year Warm events SCD Cold events TotalRegion StationPCS

PCS nr a PCS n r a

126734.1 60Edirne 215.3−1 198.8 17.6 54 13 −183.7 2.5 50 12 500 194.1 12.9 50 14 +

581258−19.6161.6+1 189.3 8.1 57 14 +233.2 29.7 64 11Tekirdag −1 62218.7 15.3 62 13 −

−7.9 58 120 207.7 4.3 62 13 + 195.5 601250−19.8 58183.6+1 201.7 −1.7 64 14 +

30.6 58 12Bilecik −1 148.3 4.0 54 13 − 174.9 56541258−9.2135.10 149.0 4.7 50 14 +

133.3 −11.0 75 12 62+1 149.0 4.7 50 14 +58 12 5629.7Bursa 296.0−1 292.0 25.7 54 13 −67 62120 257.8 −8.5 57 14 + 256.4 −9.9

−22.5 67 12+1 271.5 5.2 50 14 + 243.8 58

5020.5 12MED 56Canakkale 293.1−1 294.6 22.0 62 13 +6.8 58 120 274.8 2.2 50 14 − 54279.4

581258−8.4264.2+1 260.4 −12.2 57 14 −333.3 30.6 55 11 54Bandırma −1 342.5 39.8 54 13 +

52125836.2338.90 307.4 4.7 46 13 −−32.4 75 12+1 282.0 −20.7 64 14 + 69270.3

52125854.9Akhisar 347.4−1 297.4 4.9 46 13 −253.8 −38.7 67 12 580 302.9 10.4 50 14 +

75 12 65−43.8248.7+1 291.3 −1.2 57 14 +379.5 4.8 50 12 52I: zmir −1 370.9 −3.8 54 13 −

5412505.6380.30 362.1 −12.6 57 14 −304.0 −70.7 83 12 65+1 375.3 0.6 50 14 +

52125822.1Aydın 366.0−1 324.8 −19.0 46 13 −−10.2 50 120 346.2 2.3 50 14 + 50333.7−38.8 67 12+1 326.7 −17.2 43 14 + 305.1 54

1250−13.5 52Mugla 688.7−1 785.0 82.8 54 13 +80.5 67 120 665.5 −36.7 50 14 − 782.7 58

541250−46.4655.8+1 692.3 −9.9 57 14 +693.8 24.4 50 12 52Antalya −1 667.6 −1.8 54 13 −

38124252.7722.10 698 3 28.9 36 14 −−28.5 75 12+1 638.0 −31.4 64 14 − 640.9 69

48125031.0Silifke 393.3−1 355.6 −6.7 46 13 −379.8 17.5 67 12 650 340.7 −18.3 64 14 −

67 12 62−40.4321.9+1 325.7 −36.6 57 14 +387.5 58.0 50 12 56Mersin −1 324.1 −5.4 62 13 −

651258−19.0310.50 322.6 −6.9 71 14 +284.6 −44.9 58 12 58+1 298.8 −30.7 57 14 +

50 12 4448.9Adana 366.9−1 326.2 8.2 38 13 −332.1 14.1 58 12 730 269.7 −48.3 86 14 −

−66.7 75 12+1 314.8 −3.2 50 14 + 251.3 62

6434.3 11CMED 61Kilis 304.6−1 277.2 6.9 58 12 −16.1 55 110 257.2 −13.0 62 13 − 58286.4

521250−12.4257.9+1 261.1 −9.2 54 13 +130.5 6.5 58 12 60Malatya −1 117.1 −6.9 62 13 −

581258−4.1119.90 122.2 −1.8 57 14 +113.6 −10.4 58 12+1 111.8 −12.2 64 6214 −

5411367.7Elazıg 143.8−1 135.2 −1.0 69 13 −141.6 5.5 50 12 480 139.2 3.1 46 13 −

75 12 73−23.7112.4+1 124.1 −12.0 71 14 +262.4 6.5 50 12 56Siverek −1 267.3 11.4 62 13 +

651258−5.8250.20 238.5 −17.4 71 14 −226.0 −29.9 58 12 54+1 241.1 −14.8 50 14 +


MURAT TU8 RKES678


Year Warm events SCD Cold events TotalRegion StationPCS

PCS nr a PCS n r a

12423.7 48Diyarbakır 216.7−1 225.6 12.6 54 13 +227.1 14.1 58 12 620 203.4 −9.6 64 14 −

621267−24.2188.8+1 198.1 −14.9 57 14 +316.9 18.8 58 12Siirt −1 307.7 569.6 54 13 −

29.8 50 120 304.1 6.0 36 14 − 328.0 42621275254.2+1 −43.8309.5 11.4 50 14 +

52.5 67 12MEDT Kutahya −1 225.7 7.8 46 13 − 56270.45012504.1222.00 210.7 −7.2 50 14 −

184.2 −33.7 75 12 58+1 221.3 3.4 43 14 +50 12 6439.6Usak 262.3−1 208.1 −14.6 77 13 −

219.4 −3.3 58 120 228.0 545.3 50 14 +−22.0 83 12+1 216.5 −6.2 64 14 + 200.6 73

125053.0 56Isparta 303.7−1 236.6 −14.1 62 13 −−0.7 50 120 266.0 15.4 29 14 + 250.0 38

651267227.0+1 −23.6230.1 −20.6 64 14 +

108.2 14.8 55 11CCAN 62Kastamonu −1 108.8 15.4 69 13 +561250−9.184.30 97.1 3.7 62 13 +

90.1 −3.3 50 12 58+1 85.2 −8.2 64 14 −3.4 50 12Merzifon −1 102.1 −2.6 46 13 − 108.1 48

12584.0 62108.70 95.5 −9.2 64 14 −−13.5 58 12+1 105.6 0.9 50 14 + 91.2 54

4812420.7Sebinkarahisar + 151.8−1 162.0 10.9 54 13

151.2 0.1 33 12 420 151.7 0.6 50 14 +651258−0.2150.8+1 139.0 −10.9 71 14 −

119.4 5.7 50 12 60Corum −1 116.7 3.0 69 13 −5.5 67 120 101.4 −12.3 64 14 − 120.1 65

1275−13.1 58100.6+1 112.7 −1.1 43 14 +−1.8 58 12Sivas −1 131.5 7.0 54 13 + 122.7 56

461242−0.1124.40 126.4 1.9 50 14 +128.2 3.7 58 12 50+1 127.1 2.6 43 14 −

60125027.2Eskisehir 152.7−1 122.2 −3.2 69 13 −132.8 7.3 33 12 460 121.7 −3.8 57 14 −

−11.8 75 12+1 123.1 −2.4 57 14 + 113.7 65125023.0 48Ankara 144.5−1 120.6 0.9 46 13 −

−13.8 67 120 127.3 5.8 50 14 + 107.7 58581250−1.2120.3+1 107.8 −13.7 64 14 −

154.7 24.3 67 12 56Sivrihisar −1 127.3 −3.1 46 13 −67 12 58−5.2125.20 127.8 −2.6 50 14 +

117.6 −12.9 58 12 54+1 127.4 3.0 50 14 +26.3 50 12Afyon −1 119.4 −11.1 54 13 −* 156.8 52

1250−2.0 50128.50 135.5 5.0 50 14 +−23.9 75 12+1 114.5 −16.0 64 14 + 106.6 69

5612504.8Kırsehir 141.4−1 131.9 −4.6 62 13 −143.7 7.2 67 12 690 125.6 −10.9 71 14 −

501258−5.8130.8+1 129.6 −6.9 43 14 −153.0 24.0 67 12 56Ilgın −1 134.1 5.1 46 13 −

−2.6 42 120 126.9 −2.1 71 14 + 126.4 5812100−29.9 7799.1+1 129.2 0.2 57 14 +*

120.0 9.1 58 12Konya −1 123.2 12.4 54 5613 +62126717.0127.90 110.5 −0.4 57 14 −

92.9 −18.0 67 12 58+1 110.1 −0.8 50 14 +




TotalStation Year Warm events SCD Cold eventsRegionPCS

r a PCS n r a PCS n

64126717.3148.0−136211.1141.8−1Karaman433.6 461250133.7 6.1136.8−140

126.3 −4.4 57 14 + 111.6 −19.1+1 75 12 65

5479.6 13CEAN Kars + 64.6 −6.5 75 12−1 648.50 72.1 1.0 50 14 + 5869.8 −1.3 67 12

+1 69.6 −1.5 71 14 + 61.4 −9.7 67 12 69541146−1.8111.8+13625.8119.4−1Sarıkamıs

0 114.2 0.6 46 13 − 124.9 11.3 50 12 48+1 122.3 8.7 50 14 + 101.0 12− 5458

12.61267−2.777.7+1346 568.589.7−1Bayburt

0 80.3 −0.1 64 14 + 6276.3 −4.1 58 12+1 72.5 −7.9 71 14 − 78.1 65−2.3 58 12

61Van −1 102.2 −2.2 67 12 + 101.9 −2.4 55 114411366.1110.4−12501.2105.50581164−6.897.5+1354−6.398.0+1

* Significant at the 5% level.

REFERENCES

Amanatidis, G.T., Paliatsos, A.G., Repapis, C.C. and Bartzis, J.G. 1993. ‘Decreasing precipitation trend in the Marathon area,Greece’, Int. J. Climatol., 13, 191–201.

Blackman, R.B. and Tukey, J.W. 1958. The Measurement of Power Spectra, Dover Publications, New York, 190 pp.Deniz, A. and Karaca, M. 1995. ‘Analysis of cyclone tracks over Turkey’, J. Istanbul Tech. Uni6., 1–2, in press (in Turkish).ECSN 1995. Climate of Europe: Recent Variation, Present State and Future Prospects, National Meteorological Services First

European Climate Assessment, European Climate Support Network (ECSN), Nijkerk, The Netherlands, 72 pp.Emery, W.J. and Hamilton, K. 1985. ‘Atmospheric forcing of interannual variability in the northeast Pacific Ocean: Connections

with El Nino’, J. Geophys. Res., 90, 857–868.Fraedrich, K. 1990. ‘European grosswetter during the warm and cold extremes of the El Nino/Southern Oscillation’, Int. J.

Climatol., 10, 21–31.Fraedrich, K. and Muller, K. 1992. ‘Climate anomalies in Europe associated with ENSO extremes’, Int. J. Climatol., 12, 25–31.Glantz, M.H. 1987. ‘Drought, famine and the seasons in Sub-Saharan Africa’, in Climate and Human Health, Proceedings of the

Symposium in Leningrad, Vol. I., WCAP-No. 1, World Meteorological Organization, Geneva, pp. 217–232.Hamilton, K. 1988. ‘A detailed examination of the extratropical response to tropical El Nino/Southern Oscillation events’, J.

Climatol., 8, 67–86.Kiladis, G.N. and van Loon, H. 1988. ‘The Southern Oscillation. Part VII: Meteorological anomalies over the Indian and Pacific

sectors associated with the extremes of the Oscillation’, Mon. Wea. Re6., 116, 120–136.Kiladis, G.N. and Diaz, H.F. 1989. ‘Global climatic anomalies associated with extremes in the Southern Oscillation’, J. Climate, 2,

1069–1090.Kozuchowski, K.M. 1993. ‘Variations of hemispheric zonal index since 1899 and its relationships with air temperature’, Int. J.

Climatol., 13, 853–864.Rasmusson, E.M. and Carpenter, T.H. 1982. ‘Variations in tropical sea surface temperature and surface wind fields associated with

the Southern Oscillation/El Nino’, Mon. Wea. Re6., 110, 354–384.Repapis, C.C., Mantis, H.T. and Metaxas, D.A. 1989. ‘A study of the climatic variations in the eastern Mediterranean during the

first half of the 20th century’, Proceedings of the International Seminar on Climatic Fluctuations and Water Management, Cairo,11–14 December 1989.

Ropelewski, C.F. and Halpert, M.S. 1987. ‘Global and regional scale precipitation patterns associated with the El Nino/SouthernOscillation’, Mon. Wea. Re6., 115, 1606–1626.

Ropelewski, C.F. and Halpert, M.S. 1989. ‘Precipitation patterns associated with the high index phase of the Southern Oscillation’,J. Climate, 2, 268–284.

Rossi, G. and Somma, F. 1995. ‘Severe droughts in Italy: characteristics, impacts, and mitigation strategies’, Drought Network News,University of Nebraska, 7, 3, 9–12.

Serreze, M.C., Carse, F., Barry, R.G. and Rogers, J.C. 1997. ‘Icelandic low cyclone activity: climatological features, linkages withthe NAO, and relationships with recent changes in the Northern Hemisphere circulation’, J. Climate, 10, 453–464.

Sneyers, R 1990. On the Statistical Analysis of Series of Obser6ations, WMO Technical Note 143, World MeteorologicalOrganization, Geneva, 192 pp.


MURAT TU8 RKES680

Turkes, M. 1996a. ‘Trends and fluctuations in annual and seasonal rainfall data in Turkey’, in Climate Variability and ClimateChange Vulnerability and Adaptation, Proceedings of the Regional Workshop in Praha, September 11–15, 1995, pp. 114–126.

Turkes, M. 1996b. ‘Spatial and temporal analysis of annual rainfall variations in Turkey’, Int. J. Climatol., 16, 1057–1076.Turkes, M. 1996c. ‘Meteorological drought in Turkey: A historical perspective, 1930–1993’, Drought Network News, University of

Nebraska, 8, 17–21.Turkes, M. 1997. ‘Desertification vulnerability of Turkey with respect to climatic variability’, XI. World Forestry Congress in

Antalya, October 13–22, 1997 (unpublished voluntary paper).UKMO 1995. Modelling Climate Change 1860–2050, United Kingdom Meteorological Office (UKMO), The Hadley Centre for

Climate Prediction and Research, 12 pp.WMO 1966. Climatic Change, WMO Technical Note, 79, World Meteorological Organization, Geneva, 79 pp.WMO/WCDP 1987. The Global Climate System—Climate System Monitoring Autumn 1984–Spring 1986, World Meteorological

Organization–World Climate Data Programme, Geneva, 87 pp.WMO/UNEP 1990. The Global Climate System—Climate System Monitoring June 1986–No6ember 1988, World Meteorological

Organization–United Nations Environment Programme, Geneva, 70 pp.WMO/UNEP 1992. The Global Climate System—Climate System Monitoring December 1988–May 1991, World Meteorological

Organization–United Nations Environment Programme, Geneva, 110 pp.


INFLUENCE OF GEOPOTENTIAL HEIGHTS, CYCLONE FREQUENCY AND SOUTHERN OSCILLATION ON RAINFALL VARIATIONS IN TURKEY

Documents