Atmospheric Research

Atmospheric Research 178–179 (2016) 535–549

Contents lists available at ScienceDirect

Atmospheric Research

j ourna l homepage: www.e lsev ie r .com/ locate /atmosres

Spatiotemporal investigation of long-term seasonal temperaturevariability in Libya

S.G. Elsharkawy ⁎, E.S. ElmallahBasic and Applied Sciences Department, College of Engineering and Technology, Arab Academy for Sciences, Technology and Maritime Transport, Abo Kir, Alexandria, Egypt

⁎ Corresponding author at: College of Engineering andSciences, Technology and Maritime Transport, Abo Kir, Al

E-mail address: [email protected] (S.G. Elshark

http://dx.doi.org/10.1016/j.atmosres.2016.05.0040169-8095/© 2016 Elsevier B.V. All rights reserved.

a b s t r a c t
a r t i c l e i n f o
Article history:Received 28 July 2015Received in revised form 17 April 2016Accepted 4 May 2016Available online 6 May 2016

Throughout this work, spatial and temporal variations of seasonal surface air temperature have been investigat-ed. Moreover, the effects of relative internal (teleconnection) and external (solar) forcing on surface air temper-ature variability have been examined. Seasonal temperature time series covering 30 different meteorologicallocations and lasting over the last century are considered. These locations are classified into two groups basedon their spatial distribution. One represents Coast Libya Surface Air Temperature (CLSAT), contains 19 locations,and the other represents Desert Libya Surface Air Temperature (DLSAT), contains 11 locations. Average temper-ature departure test is applied to investigate the nature of temperature variations. Temperature trends areanalyzed using the nonparametric Mann–Kendall test and their coefficients are calculated using Sen's slopeestimate. Cross-correlation and spectral analysis techniques are also applied. Our results showed temperaturedeviation from average within a band of ±2°C at coast region, while ±4°C at desert region. Extreme behaviorintensions between summer andwinter temperatures at coast region are noticed. Segmentation process declaredreversal cooling/warming behavior within temperature records for all seasons. Desert region shows warmingtrend for all seasons with higher coefficients than obtained at coast region. Results obtained for spectral analysisshow different short and medium signals and concluded that not only the spectral properties are different fordifferent geographical regions but also different for different climatic seasons on regional scale as well. Cross-correlation results showed that highest influence for Rz upon coastal temperature is always in conjunctionwith highest influence of NAO upon coastal temperature during the period 1981–2010. Desert region does notobey this phenomenon, where highest temperature–NAO correlations at desert during autumn and winterseasons are not accompanied with highest correlations for temperature–Rz.

© 2016 Elsevier B.V. All rights reserved.

Keywords:Mann–Kendal testTime series homogeneity testMediterranean climate changeLibya surface air temperatureTrend analysisCross correlationSpectral analysis

1. Introduction

The Mediterranean region has different critical characteristics thatmake its climate characterization at regional scale scientifically impor-tant. The Mediterranean Sea is regarded as the centre point of severalmillions of population. Nowadays, climate change has the potential toexacerbate more stresses in addition to stresses of water problems.Such problems introduce new threats to human health, ecosystems,and countries' national economies. The most serious impacts are likelyto be felt in North African and eastern Mediterranean countries.

Recently, different research papers have examined different param-eters concerning climate change on the Mediterranean regional scale.Philandras et al. (2015), studied the climatology of upper air tempera-ture in the Mediterranean region during the period 1965–2011. Juanet al. (2015), detected the areas in Valencia that were potentiallymore vulnerable to air temperature change. Jamil et al. (2015), analyzedthe effect of temperature differences between the Mediterranean Sea

Technology, Arab Academy forexandria, Egypt.awy).

surface and Syrian deserts on the formation of dust storms in the westof Iran. Juan and Inmaculada (2015) analyzed the possibility of usingsea surface temperature, SST, of the Atlantic Ocean to predict the re-cruitment of European eels. Elke et al. (2014), reviewed the climate var-iability of North Atlantic–European sector during the last century.Mingfang et al. (2014) discussed the North Atlantic Multidecadal SSTOscillation and the external forcing versus internal variability. Michaelet al. (2014) investigated the climate variability during warm and coldphases of the Atlantic Multidecadal Oscillation (AMO) during1871–2008. Elmallah and Elsharkawy (2011) investigated the tempera-ture variability of winter season in Egypt and its correlation to differentinternal and external deriving forces during the last century. Jorge et al.(2013) discussed the nonstationary interannual teleconnectionsmodu-lated by multidecadal variability. Della-Marta et al. (2007), in theirwork, showed that the western European climate has becomemore ex-treme.More researches have been performed on regional scale coveringwide diverse in northern and southern hemispheres: Kyung-Ae et al.(2015), Yellow Sea; Thelma et al. (2014), Philippines; Waqas et al.(2014), Pakistan; Michael et al. (2014), United States; Xuan et al.(2014), Lancang River Basins. Mohammad et al. (2014a, 2014b), NorthCarolina; Tong-wen et al. (2015), southern Altai Mountains in eastern

536 S.G. Elsharkawy, E.S. Elmallah / Atmospheric Research 178–179 (2016) 535–549

Central Asia; Fedorov et al. (2014), northeastern Eurasia; Feng et al.(2014), Changling Mountains, the southern edge of the Gobi Desert.

Attention must be directed to intensify climate research perspec-tives concerning the South Mediterranean–African sector. This isnoticed when compared to North Mediterranean–European sector.Such intensifications participate in scene completion at this region.Therefore, our investigation throughout this paper clarifies this direc-tion. Serious attention must be paid by decision makers to the manifes-tations of climate change analysis southern theMediterranean. This is inorder to adopt the results obtained and offer themost appropriate plansleading to most appropriate decisions in different fields.

Our aim throughout this work is to

• investigate seasonal temperature variations for coast and desert Libyausing average temperature departure method;

• investigate different seasonal trends during the last century using thenonparametric Mann–Kendall, M-K test, then calculate such trendscoefficients using Sen's slope estimates test;

• apply power spectral technique to detect the hidden signals withinthe seasonal temperature time series; and

• correlate seasonal temperature to terrestrial and extraterrestrialparameters, North Atlantic Oscillation (NAO) and sunspot number(Rz), respectively, as proxies, to investigate the driving forcesthatmight affect the seasonal temperature changes in such geograph-ical spot.

2. Location characteristics

Libya is situated in between 19° 51′ N to 33° 16′ N latitudes and 9°39′ E and 25° 14′E longitudes, where it extends over 1,759,540 km2

forming part of the North African plateau that extends from the Atlantic

Fig. 1. Location of meteorological sta

Ocean to the Red Sea. It is bounded to the north by the MediterraneanSea, the west by Tunisia and Algeria, the southwest by Niger, Chadand Sudan at south, and Egypt is located at its east. Libya's coastline isconsidered the longest among all African countries bordering theMediterranean. The Libyan Desert, which covers much of Libya, is oneof themost arid places on earth. The temperature there can be extreme.

3. Data and methods of analysis

3.1. Data set

Surface air temperature time series from thirty meteorological sta-tions with different time lengths are used. Our spatial classificationhas compiled the observing stations into two categories. The first one,located at coast region, contains 19 observation stations and titled“Coast Libya Surface Air Temperature.” The second category, located atthe desert region, contains 11 observing stations and titled “DesertLibya Surface Air Temperature.”

Fig. 1 shows the locations of themeteorological stations of coast anddesert regions. The specifications ofmonitoring stations are listed belowin Tables 1 and 2, respectively, with location's name, longitude, latitude,elevation, and the time periods. It can be seen from the tables that theshortest time series exist at Bir Al ghanam, Tharhuna, Gar Abulli, andCrispi and cover only 9 yrs. Also, the longest time series are recordedat Tripoli, Elkhoms, and Derna with lengths of 119 yrs, 99 yrs, 98 yrsduring the periods 1892–2011, 1912–2011, and 1913–2011, respective-ly. The rest of locations' records lie between 30 and 70 yrs. Also, theelevations from sea level are taking diverse from a few to about 600 mhigh. The four seasons for coast and desert regions, respectively, aredenoted and defined as follows: Autumn season, denoted as ACLSATand ADLSAT, includes September, October, and November. Winter sea-son, WCLSAT and WDLSAT, includes December, January, and February.

tions at coast and desert Libya.

Table 1Information of observation locations at coast of Libya.

Location's name Longitude Latitude Elevation (m) Period

1 Pisida 11.72 E 33 12N 10 1919–19382 Zuwarah 11.08E 32.88N 3 1954–20113 Bir Alghanam 12.65E 32.28N 35 1961–19704 Tripoli 13.20E 32.90N 112 1892–20115 Wheelus 13.28E 32.90N 84 1949–19676 Sidi Al Masri 13.22E 32.87N 25 1919–19707 Tarhonah 13.58E 32.43N 65 1961–19708 Al Aziziyah 13.02E 32.53N 72 1913–19519 Garabulli 13.75E 32.73N 42 1961–197010 Al khums 14.30E 32.63N 22 1912–201111 Misratah 15.05E 32.42N 32 1954–201112 Crispi 15.10E 32.20N 25 1951–196013 Banghazi 20.08E 32.13N 25 1891–193614 Benina 20.27E 32.10N 132 1945–201115 Shahat 21.85E 32.82N 625 1945–201116 Darnah 22.58E 32.78N 26 1913–201117 Ajdabia 20.17E 30.72N 7 1961–201118 Surt 16.58E 31.20N 14 1946–201119 Al Adem/Nasser 23.92E 31.85N 155 1945–1980

537S.G. Elsharkawy, E.S. Elmallah / Atmospheric Research 178–179 (2016) 535–549

Spring season, SCLSAT and SDLSAT, includes March, April, and May.Summer season, SUCLSAT and SUDLSAT, includes June, July, and August.Data series are provided by the European Climate Assessment &Dataset-KNMI Climate Explorer, through the link http://climexp.knmi.nl/start.cgi?id=someone@somewhere. All stations are land-basedobservation stations using mercury-in-glass thermometer having ascale marking with an increment of 0.20 °C. A support keeps it in a ver-tical positionwith its cylindrical shaped bulb at the lower end. The ther-mometer is placed in a standard wooden shelter at a height rangingbetween 1.5 and 2.0 m above ground level to ensure its position is attrue air temperature. This sheltering protects the thermometer fromprecipitation while all owing the free circulation of air around it andalso prevents accidental damage. The observed air temperature is repre-sentative for the free air surrounding the station's site. This is guaran-teed since the thermometer is freely exposed to sunshine, wind, andnot shielded by, or close to, trees, buildings, and other obstructionsover a circle having a diameter of about 40 m. The temperature timeseries of all stations are investigated on the basis of completeness ofrecords. Meteorological stations with less than 3% of missing valuesare only chosen to contribute into the whole data record taken intoconsideration here. Therefore, second-order degree polynomial inter-polation method is applied to replace such missing values with theinterpolated data (Munoz-Diaz and Rodrigo, 2006).

3.2. Methods of analyses

Average temperature departure method is applied here to differen-tiate between coast and desert temperature deviations during the20th century. After that, the nonparametric Mann–Kendall, M-K,

Table 2Information of observation locations at desert of Libya.

Location's name Longitude Latitude Elevation (m) Period

1 Ghadamis 9.50E 30.13N 347 1913–20112 Sabha 14.45E 27.02N 432 1945–20113 Hon 15.95E 29.12N 267 1954–20114 Jalo 21.57E 29.03N 60 1961–20115 Mizdah 12.98E 31.43N 150 1951–19706 Bani Walid 14.00E 31.80N 240 1951–19607 Yafran 12.55E 32.08N 691 1961–20118 Gharyan 13.00E 32.17N 400 1913–19709 Nalut 10.98E 31.87N 621 1961–201110 Gaghbub 24.53E 29.75N −1 1961–201111 Al Kufrah 23.30E 24.22N 436 1949–2011

statistical test is applied to investigate for the presence of monotonicchanges in trend. The nonparametric Sen's test is applied also for esti-mating the slope coefficients of linear trends (Waqas et al., 2014;Bocchiola 2014; Haijun et al., 2014). The frame of accuracy and signifi-cance limits are considered at 95% level. The M-K test is widely usedin climate time series trend analysis (Mavromatis and Stathis, 2011).Two advantages/disadvantages are concluded of using this test. Thefirst one is that this test is a nonparametric test that does not requirethe data to be normally distributed. Such tests are also calleddistribution-free tests since they do not assume a specific distributionfor the time series data. Therefore, additional investigationsmust be ful-filled to examine the time series distribution type. The second one isthat it has low sensitivity to abrupt breaks due to inhomogeneoustime series. Such properties for the nonparametric tests drive us to in-vestigate for the data homogeneity status of our temperature time se-ries. According to M-K test, the null hypothesis H0 assumes that thereis no trend, i.e. the data are independent and randomly ordered andthis is tested against the alternative hypothesis H1, which assumesthat there is a trend, Moberg et al. (2006). Further trend investigationis performed throughout applying the segmentation process upon theseasonal temperature time series at coast and desert regions. This isdone throughout picking different time windows that enable inchecking for partial temperature variations. Moreover, spectral analysistechnique is applied using Fast Fourier Transform, FFT, to detect differ-ent hidden signals within the temperature time series. Finally, cross-correlation technique with two-tailed T test, 95% significance, and (n-1) degrees of freedom is applied to measure the influence extent ofboth terrestrial and extraterrestrial parameters upon temperature. Sea-sonal averages have been calculated throughout taking averages overseasonal months every year for the whole period at each region as fol-lows:

Xjkr¼ < aik >jr

where Xjkr is the average value for seasonal temperature j and aik is thetemperature atmonth i corresponding to station k at location r. “i” takesvalues starting from 1 to 3 representing the month's number; “j” takesvalues from 1 to 4 representing the number of seasons, “k” takes valuesfrom 1 to n representing the number of observing locations; “r” takesvalues 1 and 2 representing the number of regions, coast, and desert.

3.3. Temperature time series homogeneity test

The temperature time series are considered homogeneous if theirvariability is related only to regional weather (Pandžić and Likso,2010). Artificial gradual change in the observation site environmentcauses an artificial gradual trend in air temperature, while observationsite relocation usually causes a significant abrupt change in the air tem-perature average in comparison with the average of rather long periodsbefore relocation. In some cases, relocation of weather stations tookplace, but significant homogeneity breaks were not indicated. Fortemperature time series which are not affected by urbanization, no in-homogeneity has been established. Data accuracy accommodates thatwhen the highest temperature change is up to 0.5 °C, then it is difficultto notice a quite difference between air temperature time series beforeand after homogenization. This is because such changes are small com-pared to natural variations of mean annual air temperature. Also, it isnoted that no homogeneity breaks have been discovered at weatherstations located at desert areas. Therefore, temporal smoothing of thetemperature time series emphasizes their inhomogeneities. Also,Moberg and Alexandersson (1997) found out that the spatial averageof mean annual air temperature is “immune” to the existing inhomoge-neities which appear in the mean annual air temperature time seriesconsidered for Sweden and Switzerland. Therefore, the spatial averag-ing acts as a kind of inhomogeneities rejection filter.


Standard normal homogeneity test, SNHT, had been developedby Alexandersson (1986). Suchmethods that employ the t test assumesa statistical model in which climate observations are statisticallyindependent and Gaussian. The effect of serial correlation is usually tomake comparisons of means liberal (the actual significance level isgreater than the specified significance level). Relative descriptivemethods should be adequate, when employed, to detect gross depar-tures from the Gaussian and red noise assumptions. Such methods aredescribed by plotting observations as a function of time, plottingestimates of the power spectrum, autocorrelation function, and plottingfrequency histograms of the data.

In this work, all temperature time series are tested using normaldistribution principle as well as serial correlation technique. We foundout that all temperature time series at coast and desert regions showGaussian-shaped distribution with reasonable fitting. Moreover, non-significant serial, auto, correlations are obtained for all temperaturetime series. The correlation coefficients obtained lies within 0.03–0.1with confidence level calculated at 95%. The results give evidence that

Fig. 2. Temperature departure patterns, ξ, for surface air temperature during autu

the serial correlation coefficients are much lower than the values ofthe 5% significance level.

4. Results and discussion

4.1. Average temperature departure method

The average temperature departure method, anomaly like tech-nique, is performed to examine the seasonal temperature deviationextent from average. The calculations are performed as follows.

The deviation parameter, ξ, is calculated throughout differencingthe average seasonal temperature calculated at year j for all stationsin region r, Xjr, from average seasonal temperature, Xave, calculatedover all years for all stations in the region considered as follows:

ξ¼Xjr‐Xave

mn, winter, summer, and spring seasons at a. Coast region, b. desert region.


Thismathematical discussion leads us to expect that ξmight have ei-ther positive or negative results giving us to the following predictions:

ξ > 0 if Xjr > Xave ⇒ warming attitudeξ < 0 if Xjr < Xave ⇒ cooling attitude

Showing the extents with which the seasonal temperature, Xjr,varies from average.

Fig. 2 a and b show the temperature deviations, ξ, for all seasons atboth coast and desert Libya, respectively. Regarding the coast region,Fig. 2. a., positive ξ values are noticed on ACLSAT pattern prior to 1943while negative values are traced post to the same year. Passing 1987,ξ shows reversal positive/negative behavior with maximum negativeand positive values of −5.4 °C and 5.7 °C in 2000 and 2001. TheWCLSAT pattern shows maximum positive and minimum negative ξvalues of 2.7 °C and−2.3 °C in 1919 and 1944, respectively. SCLSAT pat-tern shows reversal variations with maximum positives and minimumnegative values of 4.4 °C, 4.3 °C, and −2.4°C during 1988, 1999, and1995, respectively. The values of ξ on SUCLSAT pattern lie within thebelt ±1.5 °C until 1980, after which a warming behavior appears andξ records highest positive departure within the belt +3°C. Maximumpositive and minimum negative values of 2.9 °C and −2.0 °C are regis-tered in 1982, 1987, 2002, and 1994, respectively. On Fig. 2. b., ADLSATpattern shows reversal negative/positive behavior for ξ until 1971.Starting in 1972, ξ takes ascending positive values showing rise in tem-perature until 1996 when the temperature has fallen below averageshowing a drop of−5.3 °C in 1999. In 2003, ξ takes positive maximumvalue of 3.5 °C. WDLSAT pattern shows negative to positive transitionbehavior with minimum negative and maximum positive ξ valuesof −5.7 °C and 3.9 °C in 1925 and 1997, respectively. SDLSAT patternshows negative ξ values until 1970 after which warming behavior istraced. Minimum negative and maximum positive ξ values of −6.6 °Cand 5.0 °C are noticed in 1925 and 1995, respectively. SUDLSAT patternshows negative ξ values since 1913 until early '70s. After 1973,almost all ξ values have taken positive results. Maximum positive and

Fig. 3. Temperature variability during autumn season at coast Libya, ACLSAT

minimum negative ξ values are 3.26 °C and −4.64 °C registered in1996 and 1941.

It is noted that Fig. 2.b shows higher ξ values at desert region whencompared to their corresponding at coast region, Fig. 2.a, especiallyduring winter time. This might be due to the continental nature of thedesert climate and the discovery of oil in Desert Libya since early1960s and its entry in the field of the world's largest producers of fossilfuel and its derivatives. Also, positive temperature deviation is noticedduring all seasons starting in early '70s until now. The temperaturedeviation from average at coast region lies within a band of ±2 °C,while±4 °C band is obtained for desert region. Also,maximum temper-ature departures are noticed during autumn season at both coast anddesert regions by the end of the 20th century.

4.2. Coast region seasonal temperature trend

The temperature time periods investigated at coast region for all sea-sons are as follows: ACLSAT whole period, 1892–2010, two segments1892–1965 and 1966–2010, respectively, represented by Fig. 3a, b,and c. WCLSAT whole period, 1889–2010, segments 1892–1950,1951–1980, and 1981–2010, respectively, shown on Fig. 4a, b, c, andd. SCLSAT whole period, 1892–2010, segments 1892–1965 and 1966–2010, respectively, illustrated on Fig. 5a, b, and c. SUCLSATwhole period,1892–2010, segments 1892–1944, 1945–1980, and 1981–2010, respec-tively, represented by Fig. 6a, b, c, and d. All temperature trends aretraced using the nonparametric M-K trend analysis and the trendcoefficients are calculated by Sen's slope estimate test.

Regarding ACLSAT patterns, negative, cooling, trend is found withcoefficient of −0.005 °C/yr giving a temperature fall of about−0.55 °C during the whole period. Reversal behavior of cooling (esti-mate slope of −0.015 °C/yr)/warming (estimate slope of 0.032 °C/yr)segments are traced during the periods 1892–1965 and 1966–2010,respectively. This behavior shows both temperatures drop and riseof −1.09 °C and 1.41 °C, respectively during these periods. WCLSATwhole period pattern shows negative, cooling, trend with slope

, during a. the whole period, 1892–2010, b. 1892–1965, c. 1966–2010.

Fig. 4. Temperature variability during winter season at coast Libya, WCLSAT, during a. the whole period, 1889–2010, b. 1892–1950, c. 1951–1980, d. 1981–2010.


estimate coefficient of −0.004 °C/yr giving a temperature fallof −0.47 °C. The period 1892–1950 shows negative, cooling, trendwith slope estimate of −0.016 °C/yr giving temperature fall of−0.93 °C/57yrs. Positive, warming, trend with slope coefficient of0.013°C/yr and hence a rise of 0.38 °C is traced during the period1951–1980. Warming trend with slope estimate value of 0.045°C/yr is

Fig. 5. Temperature variability during spring season at coast Libya, SCLSAT

traced giving a temperature rise of 1.31 °C during the period 1981–2010. SCLSAT patterns show warming/cooling/warming trendswith slopes estimates of 0.002 °C/yr,−0.007 °C/yr, and 0.025 °C/yr, re-spectively. Such slopes show total temperature changes of 0.24 °C,−0.51 °C, and 1.10 °C/45 yrs., respectively, during the periods underinvestigation. The SUCLSAT pattern shows positive, warming, trend

, during a. the whole period, 1892–2010, b. 1892–1965, c. 1966–2010.

Fig. 6. Temperature variability during summer season at coast Libya, SUCLSAT, during a. the whole period, 1892–2010, b. 1892–1944, c. 1945–1980, d. 1981–2010.


with slope coefficient of 0.013 °C/yr giving a temperature rise of 1.53 °Cfor the whole period. Positive, warming trend with slope estimate of0.011 °C/yr giving a temperature rise of 0.57 °C during the period1892–1944. Negative, cooling, trend with slope estimate coefficient of−0.03 °C/yr giving a temperature fall of −1.05 °C during 1945–1980.The last three decades, 1981–2010, show negative, cooling, trend withslope estimate of −0.016 °C/yr giving a temperature fall of −0.46 °C.

Our results, obtained above, revealed that both autumn and winterseasons showed cooling behavior while spring and summer showedwarming ones.We could predict that if such routineswere to be contin-ued, then, there would be a quite extreme divergence between summerand winter temperatures in the future. Although the expected behaviorfor temperature near the coast regions is to be mild during summertime, however, maximum warming that is found at coast region hap-pened during summer season. It is expected that the sea side shouldsuppress or decrease any temperature intentions to get raised. Thismight be due to the effect of extraterrestrial intrusion upon tempera-ture. Also, the detected reversal behavior in temperature patternsmight be described as one of the nonstationarities since it is consideredas an intrinsic part of natural climate variability (Hertig et al., 2015). As aresult of the effect of external forcing factor upon climate variability, thereversal behavior noticed mainly during winter season might be refer-enced to the long Gleissberg cycle (Ma, 2009). Also, it can be noticed,during winter season, that the warming trend coefficient exceeds itscounterpart of the cooling one with about 0.005 °C. The temperaturelevel continued to perpetuate for 65 yrs to reach its minimum. Thenstarted to rise until 2010 for another 45 yrs to overcome this effectthroughout performingwarming behavior. Until it could retrieve its nu-merical level that registered a century ago. This behavior might enforcethe idea proposed lately that the main factors affecting the climatechanges basically originate from nature (Willie et al., 2015). Furtherextensive discussions about the results obtained throughout this workcould be related to the combined effect of intensive volcanic and solaractivities (Alexander and Joan, 2015). In 1991, the Pinatubo volcanohas erupted and initiated cooling trend which impacts could be ob-served in the global temperature for a few years (Robock and Mao,1995; Barlyaeva et al., 2009). Our results showed that this phenomenonhas not affected regional climate. Else, it may had contributed, to some

extent, a decline in the temperature rise. This is in consistence withthe results obtained by Barlyaeva (2013) when regional climateresponses to external forcing are analyzed.

Comparative investigation is directed to the results obtained aboveand those obtained by Willie et al. (2015). Their results showedwarming trends during 1880s–1940s and 1980s–2000s while coolingtrends during 1950s–1970s. Same trends are obtained here for all sea-sons, except for summer, during the period 1980–2010while contradic-ting the results obtained for all seasons, except for summer, during theperiod 1892–1960. This comparison demonstrates some degree of sim-ilarity and hence compatibility between the analytical levels on bothglobal and regional scales. Therefore, attention should be directed toexamine the degree of dependence and reliability between such levelsof climate analyses. The question arises here is that “does the aggrega-tion of different regional climatic investigations converges with theresults obtained throughout global climate change studies?”

4.3. Desert region seasonal temperature trend

Desert region is characterized with the following seasonal timeperiods: ADLSAT whole period, 1923–2010, segments 1923–1970,1971–1990, and 1991–2010, respectively, represented by Fig. 7a, b, c,and d. WDLSAT whole period, 1914–2010, segments 1914–1959 and1960–2010, respectively, represented on Fig. 8.a, b, and c. SDLSATwhole period, 1913–2010, segments 1926–1939, 1942–1970 and1971–1998, respectively, represented on Fig. 9.a, b, c, and d. SUDLSATwhole period, 1913–2010, segments 1920–1969 and 1970–2010,respectively, represented on Fig. 10.a, b, c.

Regarding ADLSAT figures, all patterns show positive, warming,trends with slopes estimate of 0.028°C/yr, 0.003°C/yr, 0.107°C/yr, and0.275°C/yr, respectively. Giving a temperature rise of 2.44 °C, 0.14 °C,2.03 °C, and 5.23 °C, respectively. We can notice here that maximumwarming coefficient detected is during the period 1991–2010.WDLSAT shows that the temperature pattern had followed a dramaticrise since 1920. The trend line shows positive, warming, behavior withslope estimate of 0.063 °C/yr giving a temperature rise of 6.03 °C duringthe last century and the first decade of the newmillennium. The period1914–1959 that represents the pre-oil era shows positive, warming,

Fig. 7. Temperature variability during autumn season at desert Libya, ADLSAT, during a. the whole period, 1913–2010, b. 1923–1970, c. 1971–1990, d. 1991–2000.

Fig. 8. Temperature variability during winter season at desert Libya, WDLSAT, during a. the whole period, 1914–2010, b. 1914–1959, c. 1960–2010.


Fig. 9. Temperature variability during spring season at desert Libya, SDLSAT, during a. The whole period, 1913–2010, b. 1926–1939, c. 1942–1970, d. 1971–1998.


trend with slope estimate of 0.077°C/yr giving a temperature rise of3.47 °C during the 45 yrs. Fig. 8.c shows two different patterns for thetemperature period and the CO2 emissions during 1960–2010. On thisfigure, both curves are following positive routines. The trend coefficientfor WDLSAT curve gives 0.041 °C/yr yielding a temperature warming of2.05 °C during this period. Also, the CO2 emissions curve has been raisinto a value as six times as its original value that startedwithin 1960. After1960, the temperature records have been raisin above the level of 12 °Cto reach a temperature level above 15 °C with fluctuations higher thanthose obtained prior to 1960. It is hard in such situation to judgewhether the influence of the oil era is clear since the '60s until now.

Fig. 10. Temperature variability during summer season at desert Libya, SUDLS

The warming coefficient during 1960–2010 represents 59% of thattraced during 1914–1959. But, on the other hand, the new temperaturelimits after 1960s enforce us to pay more efforts to explain. On Fig. 9.a.,the whole period of SDLSAT, positive, warming trend with slope esti-mate of 0.06 °C/yr and hence temperature rise of 5.82 °C is obtained.Negative, cooling, trend during the period 1926–1939 is obtainedwith coefficient estimate of −0.33 °C/yr giving a temperature fallof −4.29 °C. Another negative, cooling, trend during the period 1942–1970 is traced with slope estimate of −0.029 °C/yr giving temperaturefall of−0.81 °C. The period 1971–1998 shows positive, warming, trendwith slope estimate of 0.084 °C/yr giving a temperature rise of 2.27 °C.

AT, during a. the whole period, 1913–2010, b. 1920–1969, c. 1970–2010.


Finally, SUDLSAT whole period shows positive, warming, trendwith slope estimate of 0.034 °C/y giving a temperature rise of 3.29 °C.Fig. 10.b. shows negative, cooling, trend with slope estimate of−0.017 °C/yr showing a temperature fall of −0.83 °C during the firsthalf of the 20th century. Positive, warming, trend with slope estimateof 0.058 °C/yr giving a temperature rise of 2.32 °C is obtained during1970–2010.

The third assessment report issued by the IPCC (2001) has classifiedthe global temperature change into three time zone domains. The firstdomain represented by the period 1910–1945 has warming signature.The second one during 1946–1975 shows cooling trend for globaland NH while slightly warming trends for SH. The third period, 1976–2000, is characterized by strong warming. Our results show dissimilar-ities between those mentioned in the IPCC report during 1910–1945.This is true since all seasons, except for summer coast, autumn, andwinter desert exhibit cooling trend during the same period. Also, simi-larities are shown between all results obtained throughout this workat both coast and desert regions and those obtained in the IPCC reportduring the period 1976–2010. Such comparative study tells that region-al climate changes should have limitations upon the property of copingwith the global trends. The IPCC (2007) fourth assessment report indi-cated that seasonal temperature trends during the period 1979–2005show quite large differences in regional terms. Warming was strongestover western North America, northern Europe, and China in winter.Summer has the highest warming found in the same period over

Table 3Trend analysis results for autumn temperature at coast and desert.

Period Time span(yrs)

Sen's slopeQ/yr

Sen'sslope

Trendtype

Trendeffect

Region: ACLSAT1900–2010 110 −0.005 °C −0.55 °C Negative Cooling1892–1965 73 −0.015 °C −1.09 °C Negative Cooling1966–2010 44 0.032 °C 1.41 °C Positive Warming

Region: ADLSAT1923–2010 87 0.028 °C 2.44 °C Positive Warming1923–1970 47 0.003 °C 0.14 °C Positive Warming1971–1990 19 0.107 °C 2.03 °C Positive Warming1991–2010 19 0.275 °C 5.23 °C Negative Cooling

Region: WCLSAT1892–2010 118 −0.004 °C −0.47 °C Negative Cooling1892–1950 58 −0.016 °C −0.93 °C Negative Cooling1951–1980 29 0.013 °C 0.38 °C Positive Warming1981–2010 29 0.045 °C 1.31 °C Positive Warming

Region: WDLSAT1914–2010 96 0.063 °C 6.05 °C Positive Warming1914–1959 45 0.077 °C 3.47 °C Positive Warming1960–2010 50 0.041 °C 2.05 °C Positive Warming

Region: SCLSAT1892–2010 118 0.002 °C 0.24 °C Positive Warming1892–1965 73 −0.007 °C −0.51 °C Negative Cooling1966–2010 44 0.025 °C 1.10 °C Positive Warming

Region: SDLSAT1913–2010 97 0.06 °C 5.82 °C Positive Warming1926–1939 13 −0.33 °C −4.29 °C Negative Cooling1942–1970 28 −0.029 °C −0.81 °C Negative Cooling1971–1998 27 0.084 °C 2.27 °C Positive Warming

Region: SUCLSAT1892–2010 118 0.013 °C 1.53 °C Positive Warming1892–1944 52 0.011 °C 0.57 °C Positive Warming1945–1980 35 −0.030 °C −1.05 °C Negative Cooling1981–2010 29 −0.016 °C −0.46 °C Negative Cooling

Region: SUDLSAT1913–2010 97 0.034 °C 3.29 °C Positive Warming1920–1969 49 –0.017 °C –0.83 °C Negative Cooling1970–2010 40 0.058 °C 2.32 °C Positive Warming

Europe and North Africa. This is similar to the results obtained herefor summer season at coast region. Moreover, over the past 100 yrs,1906–2005, the global average air temperatures have been risen. Mostof the global climate studies showed temperature increase in the desertareas between 1976 and 2000. They predicted a temperature increasebetween 1 °C and 7 °C for the desert region all over the globe in 2071–2100 (Kakakhel, 2006). Our results show that highest temperaturerise at desert is registered during winter season then spring andsummer season comes after. Also, the results obtained here declarethat desert region shows seasonal warming regime with temperaturerise between 2 °C and 6 °C over the past 100 yrs.

Mamtimin et al. (2011) investigated the temperature trend inLibyan hot Sahara using three monitoring stations for winter andsummer seasons during 1955–2005 and 1979–2005. They found outthat positive trend has prevailed in summer season. Their results werein accordance with ours where summer and autumn seasons have thehighest warming effect during the common period. Table 3 summarizesthe various events of cooling and warming for all seasons at both coastand desert areas. The table includes the period, time span, estimatedSen's slope values, the trend type and effect.

4.4. Seasonal temperature spectral analysis at coast and desert regions

The effect of different terrestrial and extraterrestrial factors on theclimate of the Mediterranean and their interaction with differentmodes deserves to be analyzed in further researches. Many modeshave been identified as the most important for explaining the climatevariability of different sectors of the Mediterranean region (Lionelloet al., 2006; López-Bustins et al., 2008). They could also explain the non-stationary nature of the link between NAO and climate (Jung et al.,2003; Beranova and Huth, 2007; Vicente-Serrano and López-Moreno,2008a). Various studies have demonstrated that the temperature re-sponse to winter NAO shows substantial temporal variability, resultingin very different correlation coefficients when different time spans areconsidered (Slonosky et al., 2001; Lu and Greatbatch, 2002; Huthet al., 2006; Vicente-Serrano and López-Moreno, 2008b).

Here, we apply spectral analysis technique using Fast Fourier Trans-form (FFT) upon seasonal temperature time series at both coast anddesert Libya over the last century to detect the periodicities hiddenwithin them. The existence, periods, strength of such cycles is of greatimportance when assessing the relative contribution of natural andanthropogenic contribution to modern climate change.

Fig. 11.a, b, c, and d show the power spectral patterns for ACLSAT-ADLSAT, WCLSAT-WDLSAT, SCLSAT-SDLSAT, and SUCLSAT-SUDLCAT,respectively. Maximum Gleissberg cycle has been detected onWCLSAT spectral pattern with length of 100 yrs (Gleissberg, 1944).Oscillationswithmediumperiods between 50 and 33.33 yrs are also de-tected during autumn, spring, and summer seasons. Such oscillationscan be referenced to the effect of Atlantic Multidecadal Oscillation,AMO. Periods between 30 and 7 yrs are detected over the four seasonsfor both coast and desert regions. These periods can be referencedto the effect of Hale and Schwabe cycles. Finally, short cycles between5 and 2 yrs are detected and can be referenced to the NAO and QBO(Garfinkel and Hartmann, 2010; Maravilla et al., 2004, 2008; Elmallahand Elsharkawy, 2011).

When the obtained spectral patterns are compared to each othersome interesting results are obtained. We noticed that ACLSAT,WCLSAT, and SUCLSAT spectral patterns have higher amplitudes thantheir correspondings at desert region. On the contrary, SDLSAT spectralpattern has higher amplitudes than SCLSATwhen short periods are con-sidered. Souza Echer et al. (2009) discussed the same phenomenon andconcluded that the spectral properties are different for different geo-graphical regions due to the complexity of the climatic system whichdepends on both external forcing as well as the local conditions. Herewe complete this sentence with not only that the spectral properties

Fig. 11. Power spectral patterns for seasonal temperature at both coast and desert regions.


are different for different geographical regions but also they are differ-ent for different climatic seasons.

4.5. Cross-correlation technique

We performed the cross-correlation calculations, CC, throughoutdouble phase regimes with two-tailed T test, 95% significance, and(n-1) degrees of freedom. This is because the sign of the correlationdepends upon the period studied, Georgieva et al. (2005).One phase isapplied to the whole temperature records with the correspondingseries of Sunspot number, Rz, and North Atlantic Oscillation, NAO(L. Mouel et al., 2008; Valev, 2006; Visbeck et al., 2001; Woollingset al., 2010). The other phase is concerned with data segmentationwhere temperature records are segmented into different time lengthsto check for the most active periods to be best correlated. Such tech-nique helps to evaluate the relative internal (teleconnection) and exter-nal (solar) forcing on the surface air temperature variability.

4.5.1. Temperature–Rz and temperature–NAO correlations at coast regionFig. 12. a, b, c, and d show different correlation patterns for seasonal

temperature with both Rz and NAO as follows: ACLSAT whole periodand segments 1900–1980 and 1981–2010 upper left and right patternsin Fig. 12.a. WCLSAT patterns are considered in Fig. 12.b. as the wholeperiod, outlier frame, segments 1900–1950, 1951–1980, and 1981–2010 upper left, right, and lower right patterns, respectively. Fig. 12.c.shows SCLSAT whole period, outlier frame, segments 1900–1965 and1966–2010 upper left and right patterns, respectively. SUCLSAT wholeperiod, shown by the outlier frame, and the segments 1900–1980 and1981–2010 upper left and right patterns shown in Fig. 12.d. All cross-correlation curves for ACLSAT–Rz andACLSAT–NAOpatterns are plottedin dashed and solid curves, respectively. The confidence level is plottedas dashed lines showing 95% confidence limit.

Fig. 12. a. shows that the curves representing the whole period andthe 1900–1980 period havemaximumsignificant positive/negative cor-relation coefficients hovering about 0.2 with different lags. Meanwhile,in the upper right pattern of the samefigure, 1981–2010, the correlationcoefficient rises up to reach 0.48 at lags 1 and 2 yrs for both ACLSAT–Rzand ACLSAT–NAO, respectively. TheWCLSAT–Rz curves, Fig. 12.b, showsuccessive peaks at different time lengths reflecting the length of a solarcycle with different coefficients and positive/negative maxima for thewhole and clustered periods. Maximum positive peaks are obtainedduring the periods 1951–1980 and 1981–2010 with coefficients of0.39 and 0.35 at lags−2 and −3 yrs, while negative maxima is foundduring 1981–2010 with coefficient of −4.2 at lag 2 yrs. The WCLSAT–NAO curves show peaks separated by 2–7 yrs. Maximum negativecoefficient of −0.45 at lag zero is obtained when the whole period isconsidered. Maximum negative coefficients of −0.58 and −0.7 at lagzero yrs are also obtained during 1900–1950 and 1951–1980, respec-tively. It is obvious that NAO has a remarkable effect on winter coastaltemperature. Once again, we can notice that higher temperature–Rzcorrelation leads to higher temperature–NAO correlation during1981–2010. On Fig. 12. c., regarding the whole period, the SCLSAT–Rzcurve shows positive and negative peaks spaced with about 11 yrs.The SCLSAT–NAO curve shows significant negative and positive correla-tions with maximum coefficients of −0.17 and 0.2 at lag zero yrs and−1 yrs., respectively. Successive peaks spaced by 2–3 yrs are obtainedwith different coefficient values. Regarding the period 1900–1965, thecurve SCLSAT–Rz shows insignificant peaks spaced with about 10 yrs.Maximum positive coefficient of 0.14 at−16 yrs lag is observed. Maxi-mum positive and negative coefficients of 0.23 at lag 3 yrs and−0.25 atlag zero yrs., respectively, are obtained for SCLSAT–NAO curve.Inspecting the period 1966–2010, maximum positive and minimumnegative peaks with values 0.26 and −0.31 at lags 8 yrs and 12 yrs arefound on the SCLSAT–Rz curve. The SCLSAT–NAO curve shows

Fig. 12. a. Cross-correlation patterns for ACLSAT–Rz and ACLSAT–NAOduring thewhole period, 1900–1980 and 1981–2010. b. Cross-correlation patterns forWCLSAT–NAO andWCLSAT–Rz during the whole period, 1900–1950, 1951–1980, and d. 1981–2010. c. Cross-correlation patterns for SCLSAT–NAO and SCLSAT–Rz during the whole period, 1900–1965 and 1966–2010. d. Cross-correlation patterns for SUCLSAT–NAO and SUCLSAT–Rz during the whole period, 1900–1980 and 1981–2010.


successive peaks spaced with about 2–5 yrs with maximum significantpositive coefficient value of 0.3 at −1 yr lag. This period showsmaximum effect for both Rz and NAO upon temperature during springseason. The Rz patterns of both the whole period and the 1900–1980period in Fig. 12.d show 9–11 yrs cyclic behavior, while for 1981–2010period, the Rz curve does not show the expected uniform pattern.Maximumpositive peakhaving coefficient of 0.34 at lag 14 yrs is obtain-ed on SUCLSAT–Rz curve during the period 1981–2010. Different posi-tive and negative peaks are obtained for SUCLSAT–NAO curve at thelevel of 2–6 yrs for all patterns. When regarding the coast region, max-imum cross-correlation between Rz and surface air temperature takesplace during the last three decades. This, in turn, is reflected upon theNAO–temperature correlation where higher coefficient values duringthese decades are also obtained. Higher correlation between Rz–tem-perature leads to higher correlation between NAO–temperature atcoast region. It is worth mentioning that both North Atlantic andNorth Africa regions are characterized with the most sensitive climaticzones to solar forcing on the 11-yr periodicity level (Barlyaeva, 2013).Further investigation that relates different sun states, depressed andhot, to the results obtained at coast region could be conducted whenthe maximum sun activity records are investigated.

Fig. 13 shows the pattern representing themaximaof sunspot valuessince 1760. We preferred to plot the whole records for maximumsunspots to have a wide vision and quite comparison between earlier

and later values. Regarding the solar maximum during the period1900–1950 where the 14th-to-18th solar cycles are considered. Twosolar cycles, 14th and 16th, are related to the cool, depressed, sun leav-ing the 15th, 17th, and 18th cycles into the region of hot, active, sun.This classification, in turn, reflects shadows upon the Rz–temperaturerelationship for all segmented periods, where lowest amplitudes are ob-tained. Also, the period 1951–1980 starts at the maximum of the 18thcycle passing by 19th, 20th, and finally the maximum of the 21stcycle. All these cycles lie in the active sun span and show higher sun-spots than earlier cycles. Actually, periods of high solar activity duringthe last 60 yrs are classified in type within the last 1150 yrs (Usoskinet al., 2003; Solanki et al., 2004). Therefore, the amplitudes obtainedfor both WCLSAT–Rz and WCLSAT–NAO curves during this period arehigher. This result might give us the impression of the solar activity'spower effect as a driving force upon climate elements. Finally, the peri-od 1981–2010 covers the 21st solar cycle which maximum is in 1979with 155 sunspots and ends at the beginning of the 24th solar cycle.All these cycles lie in the hot, active, sun classification. Although Rzvalues have performed a slight decline since early 90s, but the recentcounts have still remained higher than those registered before the1940s (Solanki et al., 2004). It is clear that the maximum sunspotsduring this period of time show highest values than ever before. Theamplitudes of both WCLSAT–Rz and WCLSAT–NAO curves during thisperiod are the highest among the whole three periods ensuring the

Fig. 13. Graphical representation shows the maximum events for the sunspot activities since 1760.


previous conclusions obtained about the influence of solar activity uponthe climate elements. This is in consistence with the results obtainedby Souza Echer et al. (2012), Badruddin (2014), Takuya et al. (2014),Solheim et al. (2012) andBarlyaeva (2013). Also, Samuli and Jari(2012) concluded that the NAO has acted as a mediator between thesun and Earth's climate. Also, our results are in agreementwith previous

Fig. 14.Cross-correlation patterns for all seasonswith bothNAOandRzduring thewhole period

work by Lukianova and Alekseev (2004), who showed that Rz–NAOconnection could have been at its strongest over the latter part of the20th century. Moreover, previous studies have suggested that NAO be-haves as a potential transmitter of solar forcing on regional climates(Veretenenko and Thejll, 2004; Boberg and Lundstedt, 2003; Shindellet al., 2001).

at desert Libya. a. Autumn season, b.winter season, c. spring season, andd. summer season.


4.5.2. Temperature–Rz and temperature–NAO correlations at desert regionThe desert seasonal temperature time series are considered as a

whole since no segmentation trials did work out in this part. Fig. 14. a,b, c, and d show the correlation patterns between ADLSAT, WDLSAT,SDLSAT, and SUDLSAT against each of Rz and NAO on dashed andsolid curves, respectively. The dotted line shows the confidence levelof 95% limits. The ADLSAT–Rz curve shows crests with longest timeperiod for solar activity cycle where the solar cycle's length shows atime length of 12–13 yrs. Maximum positive cross-correlation onWDLSAT–Rz curve with coefficient of 0.24 at lag 12 yrs is obtained.Also, this curve shows a time period between each two peaks of about9–10 yrs. Positive correlations on SDLSAT–Rz curve with coefficientsof ±0.2 at different lags are obtained. The ADLSAT–NAO curve showssignificant positive correlation coefficient of 0.42 with lag -1yr andmaximum negative trough is obtained for WDLSAT–NAO withcoefficient of −0.4 at lags zero. Consecutive nonsignificant peaks arealso obtained spaced with different time lengths. The correlation curveSUDLSAT–NAO shows peaks with coefficient values around ±0.2 atdifferent lags.

5. Conclusion

Throughout this work, seasonal surface air temperature at coast anddesert Libya has been examined. Both average temperature departuremethod and nonparametric Mann–Kendall test have been applied toinvestigate the temperature behavior and trend intensions. Also, Sen'sslope estimate test has been used to calculate the temperature trendscoefficients. Moreover, spectral analysis as well as cross-correlationanalysis is applied to investigate the driving forces that might influencetemperature behavior.

Our results showed similarities as well as dissimilarities with the re-sults obtained for meteorological analysis on both global and regionallevels. Such dissimilarities should acquire further examination for thedegree of dependence and reliability between such levels. Therefore,we recommend further investigations on regional scale and its strategicrelationship to the global one. The temperature deviation showedwarming behavior at both SUCLSAT and SUDLSAT after 1980 and1973, respectively. Such behavior is also traced for SDLSAT after 1970.Autumn season showed maximum temperature departures at bothcoast and desert regions by the end of the 20th century. The tempera-ture deviation belts are ±2°C and ±4°C at coast and desert regions,respectively. This wider desert temperature deviation belt confirmedthe strong variability behavior for desert climate especially duringboth winter and summer seasons.

Regarding coast region, the temperature profile exhibits cooling be-havior for autumn and winter while warming for spring and summerseasons during the last century. Based upon these results, we concludethat extreme temperature divergence scenario at coast region wouldtake place for future summer–winter seasons. Segmentation processdeclared the property of reversal cooling/warming trends within thetemperature records for most seasons. Giving the interpretation forthe potential influences of terrestrial and extraterrestrial patternsupon earth's climate. All seasons, except for summer, showwarmingbe-havior during the last four decades. This is also a reflection for the activeSun. signature during these decades upon temperature variability.

Desert region temperature showed warming trend over all seasonswith high coefficient values compared to those obtained at coast region.Winter season is getting hotter with maximum trend coefficient. Con-cluding that the warming trends at desert are accompanied with higherfluctuations than those obtained at coast. These results are gettingaligned with other different temperature investigations concerningthe desert and Sahara temperature. They expected that the temperatureat Sahara is getting higher and the range of temperature rise will beabove 2 °C during the 21st century.We conclude here that themain rea-son that might cause this behavior may not only be referenced to thehuman interventions but also to the intervention of natural terrestrial

and extraterrestrial parameters. The results obtained by spectral analy-sis technique declared the existence of different signals within thetemperature records. This also have crystallized, quantitatively andqualitatively, the impact of terrestrial and extraterrestrial parametersupon the mechanism of temperature variability. Moreover, the spectralinvestigations showed that not only the spectral properties are differentfor different geographical regions but also different with differentclimatic seasons on regional scale as well. The correlation resultsshowed that maximum correlations for both Rz and NAO parameterson temperature are found during the period 1981–2010 for all seasonsat coast region. This is in conjunction with the analysis of sunspotnumbers that showed maximum sun activity during this period. Also,this maximum correlations for Rz–temperature are accompanied withmaximum NAO–temperature correlations during the same period.Solidifying that NAO pattern behaves as a potential transmitter ofsolar forcing on regional climates as it acts as a mediator between thesun and Earth's climate.

Acknowledgments

The authors acknowledge here the KNMI Climate Explorer and theexperts supervising it. This explorer has offered us all the data usedthroughout this work. Also, we acknowledge the reviewers for their ef-forts throughout their beneficial comments. Finally, we acknowledgeMr. Karim Essmat for his help in preparing the Tiff figure for this paper.

References

Alexander, R., Joan, F., 2015. The Earth's climate at minima of Centennial GleissbergCycles. Adv. Space Res. 56, 1590–1599.

Alexandersson, H., 1986. A homogeneity test applied to precipitation data. Int. J. Climatol.6 (6), 661–675.

Badruddin, A.O.P.M., 2014. Study of the influence of solar variability on a regional(Indean) climate: 1901–2007. Adv. Space Res. 54 (8), 1698–1703.

Barlyaeva, T.V., 2013. External forcing on air-surface temperature: geographical distribu-tion of sensitive climate zones. J. Atmos. Sol. Terr. Phys. 94, 81–92.

Barlyaeva, T.V., Mironova, I.A., Ponyavin, D.I., 2009. Nature of decadal variations in theclimatic data of the second half of the 20th century. Dokl. Earth Sci. 425A, 419–423.

Beranova, R., Huth, R., 2007. Time variations of the relationships between the North At-lantic Oscillation and European winter temperature and precipitation. Stud. Geophys.Geod. 51, 575–590.

Boberg, F., Lundstedt, H., 2003. Solar wind electric field modulation of the NAO: a corre-lation analysis in the lower atmosphere. Geophys. Res. Lett. 30 (15), 1825.

Bocchiola, D., 2014. Long term (1921–2011) hydrological regime of alpine catchments innorthern Italy. Adv. Water Resour. 70, 51–64.

Della-Marta, P.M., Haylock, M.R., Luterbacher, J., Wanner, H., 2007. Doubled length ofWestern European summer heat waves since 1880. J. Geophys. Res. 112, D15103.

Elke, H., Christoph, B., Heinz, W., Jucundus, J., 2014. A review of non-stationarities inclimate variability of the last century with focus on the North Atlantic–Europeansector Earth-. Sci. Rev. 147, 1–17.

Elmallah, E.S., Elsharkawy, S.G., 2011. Influence of circulation indices upon wintertemperature variability in Egypt. J. Atmos. Sol. Terr. Phys. 73, 439–448.

Fedorov, A.N., Ivanova, R.N., Park, H., Hiyama, T., Iijima, Y., 2014. Recent air temperaturechanges in the permafrost landscapes of northeastern Eurasia. Polar Sci. 8 (2),114–128.

Feng, C., Yu-jiang, Y., Fa-hu, C., Shu-long, Y., Hua-ming, S., Tong-wen, Z., Rui-bo, Z., Li, Q.,2014. Reconstruction of spring temperature on the southern edge of the Gobi Desert,Asia, reveals recent climatic warming. Palaeogeogr. Palaeoclimatol. Palaeoecol. 409,145–152.

Garfinkel, C.I., Hartmann, D.L., 2010. Influence of the quasi-biennial oscillation on theNorth Pacific and El Niño teleconnections. J. Geophys. Res. 115, D20116.

Georgieva, K., Kirov, B., Bianchi, C., 2005. Long-term variations in the correlation betweensolar activity and climate. Mem. Soc. Astron. Ital. 76, 965.

Gleissberg, W., 1944. A table of secular variations of the solar cycle. Terr. Magn. Atmos.Electr. 49, 243.

Haijun, D., Yaning, C., Zun, S., Weihong, L., Huaijun, W., Shuhua, Z., Gonghuan, F., 2014.Dynamics of temperature and precipitation extremes and their spatial variation inthe arid region of Northwest China. Atmos. Res. 138, 346–355.

Hertig, E., Beck, C., Wanner, H., Jacobeit, J., 2015. A review of non-stationarities in climatevariability of the last century with focus on the North Atlantic–European sector. EarthSci. Rev. 147, 1–17.

Huth, R., Pokorná, L., Bochnıacuteček, J., Hejda, P., 2006. Solar cycle effects on modes oflow-frequency circulation variability. J. Geophys. Res. 111, D22107.

IPCC, 2001. In: Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., Dai, X.,Maskell, K., Johnson, C.A. (Eds.), Climate Change 2001: The Scientific Basis. Contribu-tion of Working Group I to the Third Assessment Report of the IntergovernmentalPanel on Climate Change. Cambridge University Press, Cambridge, United Kingdomand New York, NY, USA, p. 881.


IPCC, 2007. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B.,Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: the Physical Science Basis,Contribution ofWorking Group I to the Fourth Assessment Report of the Intergovern-mental Panel on Climate Change. Cambridge University Press, Cambridge, UK andNew York, NY, USA, p. 996.

Jamil, A., Shahram, K., Samira, Q., Bakhtiyar, M., 2015. Effect of the temperature variationbetween Mediterranean Sea and Syrian deserts on the dust storm occurrence in thewestern half of Iran. Atmos. Res. 154, 116–125.

Jorge, L.-P., Julianán, V., Iñigo, G., Teresa, L., Marta, M., Elsa, M., Irene, P., Belén, R., Roberto,S., 2013. Nonstationary interannual teleconnections modulated by multidecadalvariability. Física de la Tierra 25, 11–39.

Juan, C.G., Inmaculada, P., 2015. Is the Atlantic surface temperature a good proxy for fore-casting the recruitment of European eel in the Guadalquivir estuary? Prog. Oceanogr.130, 112–124.

Juan, J.M.P., María, J.E.N., Jorge, O.C., 2015. Statistical downscaling and attribution of airtemperature change patterns in the Valencia region (1948–2011). Atmos. Res. 156,189–212.

Jung, T., Hilmer,M., Ruprecht, E., Kleppek, S., Gulev, S.K., Zolina, O., 2003. Characteristics ofthe recent eastward shift of interannual NAO variability. J. Clim. 16, 3371–3382.

Kakakhel, S., 2006. Global Deserts Outlook First Thematic Assessment Report in the GlobalEnvironment Outlook (GEO) Series of the United Nations Environment Programme(UNEP).

Kyung-Ae, P., Eun-Young, L., Eunmi, C., Sungwook, H., 2015. Spatial and temporal variabil-ity of sea surface temperature andwarming trends in the Yellow Sea. J. Mar. Syst. 143,24–38.

L. Mouel, J.-L., Courtillot, V., Blanter, E., Shnirman, M., 2008. Evidence for a solar signaturein 20th-century temperature data from the USA and Europe. Compt. Rendus Geosci.340 (7), 421–430.

Lionello, P., Malanotte-Rizzoli, P., Boscolo, R., Alpert, P., Artale, V., Li, L., Luterbacher, J.,May, W., Trigo, R., Tsimplis, M., Ulbrich, U., Xoplaki, E., 2006. The Mediterraneanclimate: an overview of the main characteristics and issues. In: Lionello, P.,Malanotte-Rizzoli, P., Boscolo, R. (Eds.), Mediterranean Climate Variability. Elsevier,Netherlands, Amsterdam, pp. 1–26.

López-Bustins, J.-A., Martin-Vide, J., Sanchez-Lorenzo, A., 2008. Iberia winter rainfalltrends based upon changes in teleconnection and circulation patterns. Glob. Planet.Chang. 63, 171–176.

Lu, J., Greatbatch, R.J., 2002. The changing relationship between the NAO and NorthernHemisphere climate variability. Geophys. Res. Lett. 29 (7), 1148.

Lukianova, R., Alekseev, G., 2004. Long-term correlation between the NAO and solaractivity. Sol. Phys. 224, 445–454.

Ma, L.H., 2009. Gleissberg cycle of solar activity over the last 7000 years. New Astron. 14,1–3.

Mamtimin, B., Et-Tantawi, A.M.M., Schaefer, D., Meixner, F.X., Domroes, M., 2011. Recenttrends of temperature change under hot and cold desert climates: comparing the Sa-hara (Libya) and Central Asia (Xinjiang, China). J. Arid Environ. 75, 1105–1113.

Maravilla, D., Mendoza, B., Jauregui, E., Lara, A., 2004. The main periodicities in the mini-mum extreme temperature in northern Mexico and their relation with solar variabil-ity. Adv. Space Res. 34 (2), 365–369.

Maravilla, D., Mendoza, B., Jauregui, E., 2008. Solar signals in the minimum extremetemperature records in the southern region of the gulf of Mexico. Adv. Space Res.42 (9), 1593–1600.

Mavromatis, T., Stathis, D., 2011. Response of the water balance in Greece to temperatureand precipitation trends. Theor. Appl. Climatol. 104, 13–24.

Michael, Q., Joe, W., Xianjun, H., 2014. Analysis of diurnal air temperature range change inthe continental United States. Weather Climate Extremes 4, 86–95.

Mingfang, T., Yochanan, K., Cuihua, L., 2014. North Atlantic multidecadal SST oscillation:external forcing versus internal variability. J. Mar. Syst. 133, 27–38.

Moberg, A., Alexandersson, H., 1997. Homogenization of Swedish temperature data. PartII: homogenized gridded air temperature compared with a subset of global griddedair temperature since 1861. Int. J. Climatol. 17, 35–54.

Moberg, A., Jones, P.D., Lister, D.,Walther, A., Alexander, L.V., Brunet,M., Chen,D., Della-Marta,P.M., Jacobeit, J., Luterbacher, J., Yiou, P., Klein Tank, A.M.G., Almarza, C., Auer, I.,Barriendos, M., Bergström, H., Böhm, R., Butler, J., Caesar, J., Drebs, A., Founda, F.W.,Gerstengarbe, D., Giusi, M., Jónsson, T., Maugeri, M., Österle, H., Pandzic, K., Petrakis, L.,Srnec, M., Tolasz, R., Tuomenvirta, H., Werner, P.C., Wanner, H., Xoplaki, E., 2006. Indicesfor daily temperature and precipitation extremes in Europe analyzed for the period1901-2000. J. Geophys. Res. 111, D22106.

Mohammad, S., Manoj, K., Ademe, M., Keith, A.S., 2014a. Subseasonal climate variabilityfor North Carolina, United States. Atmos. Res. 145–146, 69–79.

Mohammad, S., Manoj, K.J., Ademe, M., Keith, A.S., 2014b. Seasonal and annual precipita-tion time series trend analysis in North Carolina, United States. Atmos. Res. 137,183–194.

Munoz-Diaz, D., Rodrigo, F.S., 2006. Seasonal rainfall variations in Spain (1912–2000) andtheir links to atmospheric circulation. Atmos. Res. 81 (1), 94–110.

Pandžić, K., Likso, T., 2010. Homogeneity of average annual air temperature time series forCroatia. Int. J. Climatol. 30, 1215–1225.

Philandras, C.M., Nastos, P.T., Kapsomenakis, I.N., Repapis, C.C., 2015. Climatology of upperair temperature in the Eastern Mediterranean region. Atmos. Res. 152, 29–42.

Robock, A., Mao, J., 1995. The volcanic signal in surface temperature observations. J. Clim.8, 1086–1103.

Samuli, H., Jari, H., 2012. Spring temperature variability relative to the North Atlantic os-cillation and sunspots—a correlation analysis with a Monte Carlo implementation.Palaeogeogr. Palaeoclimatol. Palaeoecol. 326-328, 128–134.

Shindell, D.T., Schmidt, G.A., Mann,M.M., Rind, D., Waple, A., 2001. Solar forcing of region-al climate change during the Maunder Minimum. Science 294, 2149–2152.

Slonosky, V.C., Jones, P.D., Davies, T.D., 2001. Atmospheric circulation and surface temper-ature in Europe from the 18th century to 1995. Int. J. Climatol. 21, 63–75.

Solanki, S.K., Usoskin, I.G., Kromer, B., Schüssler, M., Beer, J., 2004. Unusual activity of thesun during recent decades compared to the previous 11,000 years. Nature 431,1084–1087.

Solheim, J.-E., Stordahl, K., Humlum, O., 2012. The long sunspot cycle 23 predicts a signif-icant temperature decrease in cycle 24. J. Atmos. Sol. Terr. Phys. 80, 267–284.

Souza Echer, M.P., Nordemann, D.J.R., E, Rigozo, N.R., 2009. Multi resolution analysis ofglobal surface air temperature and solar activity relationship. J. Atmos. Sol. Terr.Phys. 71, 41–44.

Souza Echer, M.P., Echer, E., Rigozo, N.R., Brum, C.G.M., Nordemann, D.J.R., Gonzalez, W.D.,2012. On the relationship between global, hemispheric, and latitudinal averaged airsurface, temperature (GISS time series) and solar activity. J. Atmos. Sol. Terr. Phys.74, 87–93.

Takuya, S., Michinobu, K., Kentaro, T., Yugo, N., Minoru, I., Masafumi, M., 2014. Solarforcing of centennial-scale East Asian winter monsoon variability in the mid-to lateHolocene. Earth Planet. Sci. Lett. 395, 124–135.

Thelma, A.C., Rosalina, G.d.G., Flaviana, D.H., David, M.W., 2014. Long-term trends andextremes in observed daily precipitation and near surface air temperature in thePhilippines for the period 1951–2010. Atmos. Res. 145–146, 12–26.

Tong-wen, Z., Yu-jiang, Y., Yi-cheng, H., Wen-shou, W., Hua-ming, S., Liping, H., Rui-bo, Z.,Feng, C., Shu-long, Y., Zi-ang, F., Li, Q., 2015. Early summer temperature changes inthe southern Altai Mountains of Central Asia during the past 300 years. Quat. Int.358, 68–76.

Usoskin, I.G., Solanki, S., Schüessler, M., Mursula, K., Alanko, K., 2003. A millennium scalesunspot number reconstruction: evidence for an unusually active sun since the1940’s. Phys. Rev. Lett. 91 (21), 211101.

Valev, D., 2006. Statistical relationships between the surface air temperature anomaliesand the solar and geomagnetic activity indices. Phys. Chem. Earth 31, 109–112.

Veretenenko, S., Thejll, P., 2004. Effects of energetic solar proton events on the cyclonedevelopment in the North Atlantic. J. Atmos. Sol. Terr. Phys. 66, 393–405.

Vicente-Serrano, S.M., López-Moreno, J.I., 2008a. The nonstationary influence of the NorthAtlantic Oscillation on European precipitation. J. Geophys. Res.-Atmos. 113, D20120.

Vicente-Serrano, S.M., López-Moreno, J.I., 2008b. Differences in the nonstationaryinfluence of the North Atlantic Oscillation on European precipitation under differentscenarios of greenhouse gases concentrations. Geophys. Res. Lett. 35, GL034832.

Visbeck, M., Hurrell, J., Polvani, L., Cullen, H., 2001. The North Atlantic Oscillation, present,past and future. Proc. Natl. Acad. Sci. 98, 12876–12877.

Waqas, A., Aamira, F., Usman, K.A., Arif, A., 2014. Analysis of long term meteorologicaltrends in the middle and lower Indus Basin of Pakistan—a non-parametric statisticalapproach. Glob. Planet. Chang. 122, 282–291.

Willie, S., Ronan, C., Michael, C., 2015. Re-evaluating the role of solar variability onNorthern Hemisphere temperature trends since the 19th century. Earth Sci. Rev.150, 409–452.

Woollings, T., Lockwood, G., Masato, G.C., Gray, L., 2010. Enhanced signature of solarvariability in Eurasian winter climate. Geophys. Res. Lett. 37, L20805.

Xuan,W., Peiyu, L., Chunhui, L., Feifei,W., 2014. Analysis of regional temperature variationcharacteristics in the Lancang River Basin in southwestern China. Quat. Int. 333,198–206.

Documents

Atmospheric Research