Abstract

This study has been carried out to analyse the precipitation and air temperature data in the Black Sea region of Turkey to aid the understanding of the effects of global indices. Connections between the temperature or precipitation data and global atmospheric indices such as the North Sea Caspian Pattern (NCP), the Southern Oscillation Index (SOI) or the North Atlantic Oscillation (NAO) were studied. Results of the cross-correlations between air temperature and the NCP/NAO showed a strong relationship, especially in the winter period. The seasonal and annual differences for the temperature and precipitation data during the negative and positive phases of the global indices were computed. According to this, the annual total precipitations are higher during the positive NCP index than the negative NCP index, while the annual total precipitations are higher during the negative NAO index than the positive NAO index. On the other hand, wavelet analysis showed that some short-term periodicities in precipitation and temperature data are connected with the NAO and the extreme phases of the SOI. The influence of the NCP should also be considered for the short-term periodicities of the temperatures.

INTRODUCTION

Variability of hydro-meteorological parameters (precipitation, temperature, etc.) is of great importance, as many scientists are researching these parameters to find the existence of climate change. Recent studies, focused on water management and planning, need accurate data on water resources.

Analysis of past meteorological signals has always been an important issue. Statistical methods such as spectral analysis assume that a signal is stationary, and that its statistical parameters do not change with time. However, hydrological and climatic time series are not stationary – their statistical parameters change with time (Drago & Boxall 2002). Therefore, wavelet transformation is a useful method for analysing these hydro-meteorological parameters (Pisoft et al. 2004). Wavelet analysis can be used to detect trends and long-term periodicals in a hydro-meteorological time series (Pisoft et al. 2004; Nalley et al. 2012; Partal 2012; Adarsh & Reddy 2014), for example, in defining inter-annual characteristics of streamflow, precipitation, or temperature (Penalba & Vargas 2004; Ling et al. 2013; Adamovski & Prokoph 2014; Szolgayova et al. 2014). Wavelet transforms are further used in order to define the relationship between the global indices and the hydrological data (Kulkarni 2000; Jiang et al. 2014). Sang (2013) presents a more detailed review on the applications of wavelet transformation in regard to the analysis of hydrological data.

The impact on the European continent (including the Black Sea region) of the main global index climatic modes, such as the North Atlantic Oscillation (NAO) or El Niño Southern Oscillation (ENSO), has been researched in various studies over the past years (Kahya & Karabork 2001; Polonsky et al. 2007; Iqbal et al. 2016; Ward et al. 2016); it was found that NAO and Southern Oscillation Index (SOI) are highly related to the hydro-climatic parameters in Europe (Burt & Howden 2013; Fendeková et al. 2014; Philandras et al. 2015). Furthermore, the inter-annual and multi-decadal variability of the hydro-climatic data in the Mediterranean–Black Sea region was also found to be related to NAO and ENSO (Voskresenskaya & Maslova 2011). These previous studies present that in the case where positive NAO is influential, increase in pressure gradient between Icelandic low and Azores high leads to strengthening of the storms. Accordingly, the northerly path of stronger storms causes the wetter and warmer weather conditions in northern Europe, while drier conditions in southern Europe (Bell & Visbeck 2009; Rousi et al. 2012). On the other hand, the negative NAO index phase shows a weak subtropical high and a weak Icelandic low. Due to decrease in pressure gradient, fewer and weaker winter storms occur, and winter storms tend to cross on a west–east pathway (Rousi et al. 2012). In this regard, these winter storms convey the humid air to the Mediterranean. Consequently, when the NAO index is negative, the Mediterranean is wetter, and northern Europe is colder (Bell & Visbeck 2009).

The influences of NAO and SOI on the hydro-meteorological data of Turkey were investigated by Karabork et al. (2005). In this regard, they found connections linking NAO to precipitation and streamflow data of Turkey; however, the effects of SOI and NAO on air temperatures in Turkey are negligible. In another study, Kahya & Karabork (2001) found that the monthly streamflow of northwest Anatolia (the West Black Sea region of Turkey) is associated with the extreme phases of SOI: El Niño and La Niña.

The North Sea Caspian Pattern (NCP) index is calculated from the normalised 500 hPa pressure difference between the averages of the North Sea and North Caspian. The NCP index has two phases, positive and negative. A negative NCP index implies an increased anomalous counter-clockwise circulation around the western pole of the North Sea and an increased anomalous clockwise circulation around the eastern pole of the North Sea (Kutiel & Benaroch 2002). This event causes an increased anomalous westerly circulation towards central Europe, and an increased anomalous easterly circulation towards Georgia and eastern Turkey (Ghasemi & Khalili 2008).

Climate conditions throughout Europe and the Black Sea region can be affected by these circulations. Brunetti & Kutiel (2011) studied the impacts of NCP on air temperature in the European continent. They found that there is a significant decrease in the air temperatures of the Black Sea region during the positive phase of winter NCP. Kutiel & Türkeş (2005) found that the air temperature in central Turkey is warm during the negative phase of NCP and cold during the positive phase of NCP. Oguz (2005) determined a period from 1960 to the 1980s characterised by an increased mean sea surface temperature of the Black Sea during the negative phase of NAO. Nastos et al. (2011) scrutinised the relationship between NCP and air temperature in Greece and found a significant correlation for the winter season.

Recent studies show that the global indices NAO, SOI and NCP may impact climate conditions of the European continent. On the other hand, the specific effects of these indices in the Black Sea region have not been studied extensively. The present paper investigates the extent of the effect of the NAO, SOI and NCP on both the inter-annual and multi-decadal variability of the precipitation and temperature data from the Black Sea region. The methods of autocorrelation, cross-correlation, discrete wavelet transform and continuous wavelet spectrum were employed to portray the relationship between the climate data and the global indices; however, it should be noted that correlations are not proof of a causal relationship between hydro-climatic variables and the global index. There are unanswered questions about the influence these global atmospheric activities have on weather anomalies of the Atlantic–European region (including the Black Sea region). The main aim of this study is to investigate the contributions of global indices on the climate of the Black Sea region over inter-annual and multi-decadal periods.

CASE STUDY

The precipitation and air temperature data used in this study come from five stations situated along the Black Sea region of Turkey, with data collected from 1951 to 2005. The monthly average temperature and the monthly total precipitation data are collected and managed by the Turkish State Meteorological Service. The locations of the stations are shown in Figure 1, and the key features of the data are summarised in Table 1.

Table 1

Long-term averages of the seasonal and annual data at the five stations of the Black Sea coasts

  Winter
 
Summer
 
Annual
 
Stations Average daily temperature (°C) Average monthly total precipitation (mm) Average daily temperature (° C) Average monthly total precipitation (mm) Average daily temperature (° C) Average monthly total precipitation (mm) 
Giresun 7.9 114.3 21.7 78 14.3 104.2 
Rize 7.1 212.5 21.8 152.1 14.1 185.3 
Samsun 7.8 66.6 22.1 36.8 14.3 58.6 
Sinop 7.6 68.6 21.6 36.7 14 55.8 
Zonguldak 6.9 125.8 21 82.1 13.6 102 
Mean 7.4 117.5 21.6 77.1 14.0 101.1 
  Winter
 
Summer
 
Annual
 
Stations Average daily temperature (°C) Average monthly total precipitation (mm) Average daily temperature (° C) Average monthly total precipitation (mm) Average daily temperature (° C) Average monthly total precipitation (mm) 
Giresun 7.9 114.3 21.7 78 14.3 104.2 
Rize 7.1 212.5 21.8 152.1 14.1 185.3 
Samsun 7.8 66.6 22.1 36.8 14.3 58.6 
Sinop 7.6 68.6 21.6 36.7 14 55.8 
Zonguldak 6.9 125.8 21 82.1 13.6 102 
Mean 7.4 117.5 21.6 77.1 14.0 101.1 
Figure 1

The used meteorological stations.

Figure 1

The used meteorological stations.

The Black Sea region of Turkey is bordered by the Marmara Region to the west, the Central Anatolia Region to the South and the Republic of Georgia to the northeast. The Black Sea region has an oceanic climate; there are intense rainfalls and occasional flooding. The East Black Sea Region has the highest rainfall in Turkey. Along the Black Sea region, summers are warm and humid, and winters are cool and damp.

The mean monthly total precipitation in Turkey is 54 mm, however there are large regional differences. While the regional mean monthly total precipitation for the Black Sea region is 101 mm, and at Giresun station the average monthly total precipitation is 104 mm, Samsun station sees only 58 mm, contrasted by Rize station which records the highest rainfall at 185 mm (Table 1). On the other hand, the annual average daily temperatures have little variation across the region; the average temperature fluctuates between 7 °C in winter and 21 °C in summer.

Figure 2 presents the linear trend lines for 55 years of data sets for the monthly average of annual total precipitation data and annual average temperature data at the five stations. The precipitation data show no linear trend at any of the five stations. On the other hand, the annual average temperature data at the Sinop station shows a linear increasing trend, while Zonguldak station shows a linear decreasing trend. The remaining stations do not show any linear trend.

Figure 2

Linear trends in monthly mean of annual total precipitation and annual average daily temperature at five stations.

Figure 2

Linear trends in monthly mean of annual total precipitation and annual average daily temperature at five stations.

METHODOLOGY

Wavelet transforms

Wavelet analysis extracts a time–frequency relationship of a signal using a main function called wavelets (Daubechies 1990). Wavelet transformation is an effective method for time–frequency analysis of non-stationary signals such as hydro-climatic variables.

A wavelet function ψ (τ,s) is defined as below:  
formula
(1)
where t stands for time; s is a scaling factor; and τ is time factor in which the window function is iterated (Meyer 1993).
The wavelet function is obtained from a single main wavelet function called mother wavelet. In this study, the main function is generated by translation of the Morlet main wavelet which is known as non-orthogonal. The Morlet wavelet can be defined as below:  
formula
(2)
where is non-dimensional frequency parameter; and η is a non-dimensional time parameter. The Morlet wavelet was selected as the mother wavelet in this study as it is one of the most suitable mother wavelets for hydrological applications. Compared with other wavelet functions, the Morlet wavelet function closely describes the shape of hydrological signals, and provides a good balance between time and frequency localisation (Kang & Lin 2007). The Morlet wavelet also gives useful results when finding the connections of a signal in the spectral space (Coulibaly & Burn 2004).
The successive wavelet coefficient of x(t) is obtained as below:  
formula
(3)
where (*) indicates the complex conjugate function. By smoothly varying both s and τ, W(τ,s) presents a two-dimensional picture of decomposition of x(t) under different resolution levels. |W(τ,s)|2 shows the frequency of the peaks in the spectrum of x(t), and changes with the time scale of these peaks.

The lower scales use a compressed wavelet and help to obtain high frequency components of a signal. This allows us to see the short-term periodical changes of a time series. In contrast, the higher scales help to obtain low frequency components of the signal, and present slowly progressing occurrences in a time series.

The discrete wavelet transformation (DWT), used with wavelet coefficients at different resolution levels (scale), allows efficient and useful wavelet analysis. Two logarithmic scale (dyadic scales) is the simplest method. The DWT has the form as follows:  
formula
(4)
where m and n are integers that control the scale and time parameter; s0 is a specified fixed dilation step greater than 1; and τ0 is the location parameter. The most well-known choice for the parameters s0 and τ0 is 2 and 1, respectively.
For DWT, the power of two logarithmic scale is the simplest and most efficient solution (Mallat 1989). According to the Mallat algorithm, for a discrete time series xi, where xi occurs at a discrete time i (i.e., integer time steps are used), the DWT is presented as below:  
formula
(5)
where the scale parameter and location parameter. presents wavelet coefficients for the different scale levels. Thus, both detailed series and an approximation series for different scales can be obtained by thd Mallat algorithm.

The obtained decomposed wavelet components at different periods can be used for time series analysis. Each detailed component plays a different role in the observed data and the behaviour of each detailed component is distinct (Wang & Ding 2003). Approximation component (A) presents the high-scale and low-frequency component in a time series, while detailed components (D components) present the low-scale and high-frequency components in time series.

Multiple linear regression (MLR)

MLR is an approach for modelling the relationship between a dependent variable and one or more explanatory variables. The assumption of the model is that the relationship between the dependent variable yi and the predictor variable xi is linear. The MLR equation is defined as follows:  
formula
(6)
where α is called the intercept; and βj are called slopes or coefficients. The MLR equation can be used to make a forecast of the value of y with the appropriate values of x.

METHODOLOGICAL STEPS

The methodological procedures followed in this study are as below.

  1. Cross-correlations between the observed air temperatures/precipitations of Turkey and the global atmospheric indices were presented.

  2. The seasonal/annual average daily temperature differences and the seasonal/annual monthly precipitation differences between the negative and positive phases of the global indices were determined.

  3. The continuous wavelet transforms (CWT) were applied to the seasonal/annual precipitation and temperature data for understanding annual, decadal and multi-decadal changes.

  4. Each of the observed hydro-meteorological time series was decomposed into its periodical components by the DWT. Afterwards, the correlation coefficients between the periodic components of the hydro-meteorological data and the global atmospheric indices were presented.

RESULTS

Correlograms between the hydro-meteorological data and global atmospheric indices

First, cross-correlation analysis was carried out for examining the connections between air temperature and precipitation data and the global atmospheric indices. Cross-correlograms identify the time shift between phenomena (Fendekov et al. 2014). For this aim, Pearson correlation coefficients (r) were computed between global indices and hydro-climate data. Then, the results of these correlation coefficients were evaluated according to Student t-test at significance levels of α = 0.05. The boundary interval of the correlation coefficient is [r] ≥ 0.27 for α = 0.05. The independence between the global indices and the hydro-climatic data for the period 1951–2005, with lag up to ±15 years at an annual step, was tested. Then, every lagged correlation coefficient was evaluated at the 5% significance level (Linage et al. 2013).

The temperature cross-correlograms show a clear relationship between the NCP and the annual mean air temperatures of the Black Sea region (Figure 3). Some of these correlations are statistically significant at α = 0.05. There is a significant negative correlation between the NCP and annual temperatures at the Sinop and Samsun stations in lag 0. Similar results were also obtained between the NAO and annual mean air temperatures (Figure 4); there are statistically significant negative correlations between the NAO and the annual mean temperatures in lag 0. This means that in years with a high NAO or NCP, a decrease in the air temperatures on the Black Sea coasts can be expected.

Figure 3

Cross-correlation correlogram between the annual mean NCP and annual mean air temperatures (horizontal lines refer to boundary intervals of autocorrelation coefficients for α = 0.05).

Figure 3

Cross-correlation correlogram between the annual mean NCP and annual mean air temperatures (horizontal lines refer to boundary intervals of autocorrelation coefficients for α = 0.05).

Figure 4

Cross-correlation correlogram between the annual mean NAO and annual mean air temperatures (horizontal lines refer to boundary intervals of autocorrelation coefficients for α = 0.05).

Figure 4

Cross-correlation correlogram between the annual mean NAO and annual mean air temperatures (horizontal lines refer to boundary intervals of autocorrelation coefficients for α = 0.05).

In addition to the correlations at lag 0, cross-correlations were found when the signals were offset. Cross-correlations between air temperature and the NCP at the Giresun and Rize stations show significant positive coefficients at lag 7. This means that the NCP index correlates with the air temperature values from seven years later. The same was true of the NAO, although at lag 12; there were significant positive correlations between the NAO and the air temperatures in 12-year lag periods. Similar results were also found for the winter periods, however this is not shown here.

In terms of annual precipitation, no significant correlations were found at lag 0 (Figure 5). However, there were significant positive correlations at lag 12 between annual total precipitations and the NAO index.

Figure 5

Cross-correlation correlogram between the annual mean NCP/NAO and annual total precipitations (horizontal lines refer to boundary intervals of autocorrelation coefficients for α = 0.05).

Figure 5

Cross-correlation correlogram between the annual mean NCP/NAO and annual total precipitations (horizontal lines refer to boundary intervals of autocorrelation coefficients for α = 0.05).

These correlations at positive lag values mean that the present time NCP or NAO indices are influencing the future air temperatures and precipitations of the Black Sea region.

Figure 6 presents the MLR line between global atmospheric indices and annual average temperature/annual average precipitation data for Samsun station. R2 is the determination coefficient. The annual average daily temperature is negatively correlated with the NAO and NCP indices (R2 = 0.2297 for the NCP), while a positive correlation can be seen with the SOI index. The MLR equation was computed so as to investigate the dependence of global index effects on the precipitation and air temperature data at the Samsun station.

Figure 6

Linear regression line between global index and annual average temperature/annual average precipitation data for Samsun station. R2 is determination coefficient.

Figure 6

Linear regression line between global index and annual average temperature/annual average precipitation data for Samsun station. R2 is determination coefficient.

The MLR equation for annual average daily temperature is:  
formula
(7)
where T refers to annual average daily temperatures. The NCP has the strongest dependent coefficient. This shows that the NCP has more of an effect on the air temperature in the Black Sea region than NAO or SOI. Figure 7 presents the scatter diagram between MLR estimations and observed temperatures. The determination coefficient is 0.287 for the MLR estimations.
Figure 7

The scatter diagram beween the multiple linear regression estimations and observed data.

Figure 7

The scatter diagram beween the multiple linear regression estimations and observed data.

The MLR for annual precipitation is:  
formula
(8)
where P refers to the annual average monthly precipitation value. As shown, NCP and NAO have stronger dependent coefficients than SOI. Estimations obtained from the MLR equation are shown in Figure 7. The determination coefficient is 0.06 for the MLR estimations. The MLR equation shows that the NAO and NCP have more of an effect than SOI on Samsun hydro-meteorological data.

Average seasonal differences for the hydro-meteorological data during the negative and the positive phases of the global indices

For the NCP index, the months with above 0.5 value were defined as positive phase, and months with under 0.5 value were defined as negative phase (Kutiel & Benaroch 2002). Table 2 presents the daily average temperature differences between both phases of the NCP for the winter, summer and annual periods. The table shows that for the winter period, the average daily temperatures during the negative NCP index are almost 3 °C higher than during the positive NCP. These differences are 3.7 °C at Samsun and 3.4 °C at Giresun. The annual mean temperatures during the negative NCP years are nearly 1 °C higher than during the positive NCP years. There are no differences for the summer air temperatures in either phase. These results show a clear influence of the NCP on temperature data in the Black Sea region.

Table 2

The average seasonal differences for average daily temperatures (° C) during NCP(−) and NCP(+) periods from 1950 to 2005 at the five stations

  Winter
 
Summer
 
Annual
 
  NCP(−) NCP(+) NCP(−) NCP(+) NCP(−) NCP(+) 
Giresun 9.7 6.3 21.4 21.6 14.5 13.5 
Rize 8.8 5.7 21.9 21.5 14.6 13.5 
Samsun 9.7 22 21.9 14.7 13.7 
Sinop 9.3 6.1 21.5 21.4 14.4 13.5 
Zonguldak 8.7 5.1 20.8 20.8 14 12.8 
  Winter
 
Summer
 
Annual
 
  NCP(−) NCP(+) NCP(−) NCP(+) NCP(−) NCP(+) 
Giresun 9.7 6.3 21.4 21.6 14.5 13.5 
Rize 8.8 5.7 21.9 21.5 14.6 13.5 
Samsun 9.7 22 21.9 14.7 13.7 
Sinop 9.3 6.1 21.5 21.4 14.4 13.5 
Zonguldak 8.7 5.1 20.8 20.8 14 12.8 

While there is no clear influence of either phase on the seasonal precipitations (Table 3), the annual precipitations are higher during the positive NCP index. Previous studies have shown that the NCP index is active mainly from October to April and has a major influence on the temperature regime in the Anatolia Region and the Middle East (Kutiel & Benaroch 2002).

Table 3

The average seasonal differences for average monthly precipitations (mm) during NCP(−) and NCP(+) periods from 1950 to 2005 at the five stations

  Winter
 
Summer
 
Annual
 
  NCP(−) NCP(+) NCP(−) NCP(+) NCP(−) NCP(+) 
Giresun 115 105 71 77 92 110 
Rize 218 234 148 145 168 190 
Samsun 69 65 32 35 52 60 
Sinop 66 79 32 37 43 60 
Zonguldak 137 116 66 82 89 99 
  Winter
 
Summer
 
Annual
 
  NCP(−) NCP(+) NCP(−) NCP(+) NCP(−) NCP(+) 
Giresun 115 105 71 77 92 110 
Rize 218 234 148 145 168 190 
Samsun 69 65 32 35 52 60 
Sinop 66 79 32 37 43 60 
Zonguldak 137 116 66 82 89 99 

Table 4 presents daily mean temperature differences between the negative and positive phases of NAO for the winter, summer and annual periods. The results indicate that for the winter period, the average daily temperatures during the negative NAO index are nearly 1 °C higher than during the positive NAO, and the annual average daily temperatures during the negative NAO years are higher than during the positive NAO years.

Table 4

The average seasonal differences for average daily temperatures (° C) during NAO(−) and NAO(+) periods from 1950 to 2005 at the five stations

  Winter
 
Summer
 
Annual
 
  NAO(−) NAO(+) NAO(−) NAO(+) NAO(−) NAO(+) 
Giresun 8.4 7.7 21.6 21.7 14.7 14.1 
Rize 7.6 6.8 21.7 21.5 14.4 13.9 
Samsun 8.4 7.5 22.1 21.9 14.7 14.1 
Sinop 8.1 7.4 21.6 21.5 14.4 13.9 
Zonguldak 7.5 6.5 21 21 13.9 13.4 
  Winter
 
Summer
 
Annual
 
  NAO(−) NAO(+) NAO(−) NAO(+) NAO(−) NAO(+) 
Giresun 8.4 7.7 21.6 21.7 14.7 14.1 
Rize 7.6 6.8 21.7 21.5 14.4 13.9 
Samsun 8.4 7.5 22.1 21.9 14.7 14.1 
Sinop 8.1 7.4 21.6 21.5 14.4 13.9 
Zonguldak 7.5 6.5 21 21 13.9 13.4 

The annual precipitations during the positive NAO index are lower than during the negative NAO (Table 5), and the average monthly total precipitations in the winter season during the negative NAO index are slightly higher than during the positive NAO index (except at one station). This results in the Black Sea region seeing more rainfall under the effect of the negative NAO index.

Table 5

The average seasonal differences for average monthly precipitations (mm) during NAO(−) and NAO(+) periods from 1950 to 2005 at the five stations

  Winter
 
Summer
 
Annual
 
  NAO(−) NAO(+) NAO(−) NAO(+) NAO(−) NAO(+) 
Giresun 108 103 70 76 110 101 
Rize 199 197 157 147 200 181 
Samsun 67 58 27 37 63 56 
Sinop 65 65 31 39 66 53 
Zonguldak 125 121 57 108 105 95 
  Winter
 
Summer
 
Annual
 
  NAO(−) NAO(+) NAO(−) NAO(+) NAO(−) NAO(+) 
Giresun 108 103 70 76 110 101 
Rize 199 197 157 147 200 181 
Samsun 67 58 27 37 63 56 
Sinop 65 65 31 39 66 53 
Zonguldak 125 121 57 108 105 95 

Continuous wavelet spectrums of the temperature and precipitation data

In this section, the CWT was applied on the temperature and precipitation data. Figures 8 and 9 explain the variability over time of high periodic structures (multi-decadal) and long-term trends. A CWT figure shows us the variability of energy (continuous wavelet coefficients) when effective hydrological events occur in a time series. Large values of the wavelet transform coefficients are obtained when there is a high correlation between the main wavelet and the time series. The shading corresponds to the wavelet coefficient value, scaled between zero (darkest) and the maximum absolute value of the coefficients (brightest). It should be understood that periodic events with the most effect are seen as bright regions on the CWT figures.

Figure 8

Continuous wavelet spectrums for the winter seasonal average temperatures at Sinop (a), Samsun (c) and Rize (e) for the annual mean temperatures at Sinop (b), Samsun (d) and Rize (f).

Figure 8

Continuous wavelet spectrums for the winter seasonal average temperatures at Sinop (a), Samsun (c) and Rize (e) for the annual mean temperatures at Sinop (b), Samsun (d) and Rize (f).

Figure 9

Continuous wavelet spectrums for the winter seasonal precipitations at Sinop (a), Rize (c) and Samsun (e), for the annual precipitations at Sinop (b), Rize (d) and Samsun (f).

Figure 9

Continuous wavelet spectrums for the winter seasonal precipitations at Sinop (a), Rize (c) and Samsun (e), for the annual precipitations at Sinop (b), Rize (d) and Samsun (f).

Figures 8 and 9 show the results of the CWT spectrum on the precipitation and temperature data at the three stations. The inter-decadal periodicities are located at the 1–4-year periodicity. Brightness at this periodicity shows the probability of global indices having short-term effects such as increased rainfall or drought. The strong periodic events (bright regions) in the precipitation data can be seen clearly from the CWT spectrums.

The temperature data in Figure 8 show events (bright regions) between 1950 and 1955, and between 1973 and 1980 for the winter temperature data (highlighted by the circles in Figure 8). These periodic events are similar for all stations. The brightness in the multi-decadal scales, such as that at 32-yearly periodicity, presents variability over time of the long-term trend. The long-term periodicities at the 14- and 27-year scales (bright lane zone) are the most significant multi-decadal events, continuous through the time studied.

The precipitation data in Figure 9 show strong periodicals located at the 1–4-year scale level. Events are seen: between 1950 and 1955; at 1960; between 1965 and 1970; at 1983; at 1995; and at 2005 for the Sinop winter precipitation data (highlighted by the circles in Figure 9). These periodicities can be seen as weak in the annual precipitation data. These short-term strong periodicities, located at the 1–4-year scale level, repeat roughly every 10–11 years. This 11-year periodical is most likely associated with sun activity. Further periodic structures, at the 3–7-year scale, are visible between 1960 and the 1970s, as well as between 1980 and 1990 for the annual precipitation data for Sinop. Moreover, the long-term periodicities such as 14–27-year scales (bright zone) can be seen as continuous through the time studied.

The wavelet spectrums of the observed precipitation data show some evidence of an effect from both the SOI and NAO. The short-term periodicities in the precipitation data may be related to strong SOI (El Niño or La Niña) events. For instance, the brightness in 1–4-yearly scales of the CWT figures on the precipitation data can be referred to the 1953, 1969, 1982, 1994 and 2005 El Niño events. Furthermore, the very strong short-term periodicity in the 1968 winter precipitation data at Sinop and Samsun stations, which saw the most rainfall in the last 55 years, may be explained by the concurrence of the strong negative phases of the NAO with strong La Niña events.

Consequently, the efficiencies of the global index are clearly seen in the continuous wavelet spectrums. The intersection of multiple global indices is seen as strong light region on the continuous wavelet spectrum.

Cross-correlations between the wavelet periodic components and the global atmospheric indices

The observed data were decomposed into an approximation (A) and four detailed components by wavelet decomposition in this study. The detailed components represent 2-yearly periodicity (D1), 4-yearly periodicity (D2), 8-yearly periodicity (D3), 16-yearly periodicity (D4) and 32-year periodicity (D5) in the data. The A component represents the approximation component at the fifth level of decomposition.

The correlations between the annual average daily temperature/annual precipitation and their wavelet components were computed and presented in Table 6 for three stations. All of the correlations are statistically significant at the level of 0.05. However, D1 has the highest magnitude of correlation for all the stations. Figure 10 presents the calculated correlation coefficients (r) between the global indices and the wavelet components of the annual average temperature data for the period 1951–2005. The periodic wavelet components have significantly high correlations. The 32-yearly modes (D5) of the annual mean temperatures exhibit significantly high negative correlations with regard to NAO at all the stations. The observed temperature data at all stations have significant negative correlations with NAO, as seen in Figure 4. On the other hand, the short-term periodic wavelet components (D1 and D2) of the annual mean temperatures have significantly high negative correlations with the NCP. It can be said here that the influence of NCP on short-term variations in the temperatures of the Black Sea coast should be considered. According to Figure 10, the D2 and D5 components are well correlated with the SOI index, but some of these are not statistically significant. On the other hand, even if there are significant correlations in some stations, it can be said that there is no remarkable relationship between the global indices and the annual precipitations (Figure 11).

Table 6

Correlations between hydro-meteorological data and their wavelet components

Discrete wavelet components Samsun
 
Zonguldak
 
Sinop
 
Temp. Precip. Temp. Precip. Temp. Precip. 
D1 0.73 0.67 0.75 0.79 0.69 0.67 
D2 0.59 0.65 0.64 0.63 0.65 0.66 
D3 0.35 0.54 0.35 0.41 0.54 0.49 
D4 0.37 0.38 0.32 0.22 0.38 0.44 
D5 0.41 0.29 0.34 0.27 0.29 0.34 
Discrete wavelet components Samsun
 
Zonguldak
 
Sinop
 
Temp. Precip. Temp. Precip. Temp. Precip. 
D1 0.73 0.67 0.75 0.79 0.69 0.67 
D2 0.59 0.65 0.64 0.63 0.65 0.66 
D3 0.35 0.54 0.35 0.41 0.54 0.49 
D4 0.37 0.38 0.32 0.22 0.38 0.44 
D5 0.41 0.29 0.34 0.27 0.29 0.34 
Figure 10

The correlations between wavelet D components of the annual temperature data and the global indices for the period 1951–2005 (the dashed line is confidence limit of correlation coefficient for α = 0.05).

Figure 10

The correlations between wavelet D components of the annual temperature data and the global indices for the period 1951–2005 (the dashed line is confidence limit of correlation coefficient for α = 0.05).

Figure 11

The correlations between wavelet D components of the annual precipitation data and the global indices for the period 1951–2005 (the dashed line is confidence limit of correlation coefficient for α = 0.05).

Figure 11

The correlations between wavelet D components of the annual precipitation data and the global indices for the period 1951–2005 (the dashed line is confidence limit of correlation coefficient for α = 0.05).

CONCLUSIONS

In this paper, the effect of global indices on precipitation and air temperature in the study area of the Black Sea coast were presented.

According to the cross-correlation analysis, there are statistically significant negative correlations between the annual/winter NCP and the air temperatures within the one-year time frame. The same result can also be seen between the NAO and annual/seasonal average temperature data. There are also significant positive correlations between the NCP and the annual temperatures in lag 7. On the other hand, there are no significant correlations between the global indices and the annual precipitations. As shown through the MLR equation, the NAO and NCP have more of an effect than SOI on Samsun hydro-meteorological data.

The mean temperatures during the negative NCP index are nearly 3 °C higher than during the positive NCP for the winter season. Previous studies have shown that there is a considerable decrease in air temperature of the Black Sea coasts during the positive NCP index (Brunetti & Kutiel 2011). This paper confirms that in years with both a high NAO and NCP, a decrease in the air temperatures of the Black Sea coast of Turkey can be expected.

The annual total precipitations are higher during the positive NCP index than the negative NCP index, while the annual total precipitations are higher during the negative NAO index than the positive NAO index. Thus, the Black Sea coast is wetter during the negative NAO and the positive NCP indices. This paper demonstrates that precipitation in the Black Sea coast of Turkey increases during the negative phase of NAO, and decreases during the positive phase of NAO. The results found in this study are consistent with previous studies. From this and previous studies, the air temperatures of the Mediterranean–Black Sea region can be considered in connection with NAO.

The continuous wavelet spectrums of the temperature and precipitation data show very distinct short-term periodicity (1–4 years) repeating every 5–11 years. This periodicity is most likely associated with sun activity and the extreme phases (El Niño and La Niña) of the SOI. The very strong short-term precipitation event in winter 1968 is consistent with the strong negative phase of the NAO in winter 1968, and strong La Niña events in the same year. In other words, the short-term periodicities observed in the precipitation data of the Black Sea could be related to the activity of the SOI and NAO. It is important to say that the results of this study are similar to the results of Partal (2012), who found significant evidence of SOI influence to hydrologic data from the Aegean Region (Turkey).

The results also show that the D5 (32-year periodicity) exhibits significantly high negative correlations between the NAO and the temperature data. The 32-year periodicity is significant because it shows variability of the long-term events. The influence of the NAO on the long-term periodicities of the air temperatures of the Black Sea coast should be considered. Conversely, the D1 and D2 (short-term periodicities) components of the temperature data have a very strong negative correlation with the NCP index.

As a result, this study illustrates that global indices have a great effect on temperature data of the Black Sea region. Particularly, it is understood that the NCP has the greater effect on the hydro-climatic data of the Black Sea coast. Global indices also affect precipitation in the Black Sea region, specifically that rainfall is higher during negative NAO and positive NCP indices. However, the effect on the precipitation data is not as strong as on the temperature data, according to the correlation analysis and the difference analysis. Moreover, the results of the wavelet analysis on precipitation data reveal relationships between the global indices and the observed data, some of which could not be observed through correlation or difference analysis. This study reveals the substantial impacts of global indices on the hydro-climatologic behaviour of the Black Sea coast.

REFERENCES

REFERENCES
Bell
I.
&
Visbeck
M.
2009
North Atlantic Oscillation
.
www.ldeo.columbia.edu/res/pi/NAO/ (accessed 15 June 2009)
.
Coulibaly
P.
&
Burn
D. H.
2004
Wavelet analysis of variability in annual Canadian streamflows
.
Water Resources Research
47
,
W03105
.
Daubechies
I.
1990
The wavelet transform, time-frequency localization and signal analysis
.
IEEE Transactions on Information Theory
36
(
5
),
961
1005
.
Drago
A. F.
&
Boxall
S. R.
2002
Use of the wavelet transform on hydro-meteorological data
.
Physics and Chemistry of the Earth
27
,
1387
1399
.
Fendeková
M.
,
Pekárová
P.
,
Fendek
M.
,
Pekár
J.
&
Škoda
P.
2014
Global drivers effect in multi-annual variability of runoff
.
Journal of Hydrology and Hydrometeorology
62
,
169
176
.
Ghasemi
A. R.
&
Khalili
D.
2008
The effect of the North Sea-Caspian pattern (NCP) on winter temperatures in Iran
.
Theoretical and Applied Climatology
92
,
59
74
.
Iqbal
M. A.
,
Penas
A.
,
Cano-Ortiz
A.
,
Kersebaum
K.
,
Herrero
L.
&
del Río
S.
2016
Analysis of recent changes in maximum and minimum temperatures in Pakistan
.
Atmospheric Research
168
,
234
249
.
Jiang
R.
,
Gan
T. Y.
,
Xie
J.
&
Wang
N.
2014
Spatiotemporal variability of Alberta's seasonal precipitation, their teleconnection with large-scale climate anomalies and sea surface temperature
.
International Journal of Climatology
34
(
9
),
2899
2917
.
Kahya
E.
&
Karabork
C.
2001
The analysis of El Niño and La Niña signals in streamflows of Turkey
.
International Journal of Climatology
21
(
10
),
1231
1250
.
Linage
C.
,
Kim
H.
,
Famiglietti
J.
&
Yu
J. Y.
2013
Impact of Pacific and Atlantic sea surface temperatures on interannual and decadal variations of GRACE land water storage in tropical South America
.
Journal of Geophysical Research
118
,
10811
10829
.
Mallat
S. G.
1989
A theory for multiresoluion signal decomposition: the wavelet representation
.
IEEE Transactions on Pattern Analysis and Machine Intelligence
11
(
7
),
674
693
.
Meyer
Y.
1993
Wavelets Algorithms & Applications
.
Society for Industrial and Applied Mathematics
,
Philadelphia, PA
.
Nastos
P.
,
Philandras
C.
,
Founda
D.
&
Zerefos
C.
2011
Air temperature trends related to changes in atmospheric circulation in the wider area of Greece
.
International Journal of Remote Sensing
32
(
3
),
737
750
.
Philandras
C.
,
Nastos
P.
,
Kapsomenakis
I.
&
Repapis
C.
2015
Climatology of upper air temperature in the eastern Mediterranean region
.
Atmospheric Research
152
,
29
42
.
Pisoft
P.
,
Kalvova
J.
&
Brazdil
R.
2004
Cycles and trends in the Czech temperatures series using wavelet transform
.
International Journal of Climatology
24
,
1661
1670
.
Polonsky
A. B.
,
Bardin
M. Y.
&
Voskresenskaya
E. N.
2007
Statistical characteristics of cyclones and anticyclones over the Black Sea in the second half of the 20th century
.
Physical Oceanography
17
,
348
359
.
Rousi
E.
,
Anagnostopoulou
C.
,
Tolika
K.
,
Maheras
P.
&
Bloutsos
A.
2012
ECHAM5/MPI general circulation model simulations of teleconnections indices over Europe
.
Advances in Meteorology, Climatology and Atmospheric Physics
2
,
709
716
.
Szolgayova
E.
,
Parajkab
J.
,
Blöschla
G.
&
Buchera
C.
2014
Long term variability of the Danube River flow and its relation to precipitation and air temperature
.
Journal of Hydrology
519
,
871
880
.
Voskresenskaya
E. N.
&
Maslova
V. N.
2011
Winter-spring cyclonic variability in the Mediterranean-Black Sea region associated with global processes in the ocean-atmosphere system
.
Journal of Advance Scientific Research
6
,
237
243
.
Wang
D.
&
Ding
J.
2003
Wavelet network model and its application to the prediction of hydrology
.
Nature and Science
1
,
67
71
.