This study highlights an important and previously overlooked summer North Atlantic Oscillation (NAO) influence over the eastern Mediterranean. The featured analysis is based on a synergistic use of reanalysis data, satellite retrievals, and coastal and buoy meteorological observations. The physical mechanisms at play reveal a strong summer NAO involvement on the pressure fields over northern Europe and the Anatolian plateau. Especially during August, the summer NAO modulates the Anatolian low, together with the air temperature, meridional atmospheric circulation, and cloudiness over the eastern Mediterranean. Including the dominant action centers over Greenland and the Arctic, the identified modulations rank among the strongest summer NAO-related signals over the entire Northern Hemisphere.
The North Atlantic Oscillation (NAO) is a prominent planetary-scale phenomenon depicting a variety of teleconnection patterns extending over the Northern Hemisphere (NH; Wallace and Gutzer 1981). Depending on the season (i.e., winter–summer), NAO is represented by pressure centers that shift their action points between the latitudes 30° and 70°N over the Atlantic Ocean (Barnston and Livezey 1987; Hurrell et al. 2003; Folland et al. 2009). A number of studies have described NAO’s linkage to precipitation, storm tracks, ocean waves, and temperature patterns over the U.S. East Coast, the North Atlantic Ocean, and western Europe (Wallace and Gutzler 1981; Hurrell and van Loon 1997; Visbeck et al. 2001; Pozo-Vasquez et al. 2001; Trigo et al. 2002b; Hurrell et al. 2004; Hurrell and Deser 2009; Lionello and Sanna 2005). During the past decade, NAO has gained significant attention because of its suggested role in the NH warming (Hurrell et al. 2003; Folland et al. 2009). Although NAO is present throughout the entire year, the majority of the current literature attributes limited attention to its climatological characteristics during seasons other than winter (Folland et al. 2009). Hurrell and Folland (2002) point out that, during summer, NAO’s climatic variability is of particular importance since this season relates to environmental extremes such as heat waves and droughts. Folland et al. 2009 underscore that, during summer, NAO possesses less prominent spatial features compared to its winter counterpart (Hurrell et al. 2003; Hurrell and Folland 2002; Ogi et al. 2004, 2005; Feldstein 2007; ,Eshel et al. 2000).
During the summer season, the eastern Mediterranean is among the warmest regions over the NH midlatitudes. Its summer climatological characteristics are regulated by two pressure centers. To the east, the northwestern expansion of the Indian subcontinent monsoonal low combined with the intense heating over the Middle East, gives rise to the semipermanent cyclonic center also know as the Anatolian low (Fisher 1978). To the west, the Azores anticyclone expands toward the northeast and progresses as far as the northern Africa coasts. The aforementioned circulation features enforce a northerly flow over the eastern Mediterranean known as the Etesian winds (Greek word for year—Etos; Meltem in Turkish, Metaxas 1977; Maheras 1980; Alpert et al. 1990, 2004; Lascaratos 1992; Barry and Chorley 1992; Metaxas and Bartzokas 1994; Serreze et al. 1997; Lolis et al. 2002).
Based on a synergistic use of model (reanalysis) datasets, satellite retrievals, and coastal and buoy meteorological observations, this work provides a novel aspect of the summer North Atlantic Oscillation (SNAO) influence over the eastern Mediterranean. In section 2, this paper describes the employed datasets and numerical methods. In section 3, the authors offer their interpretation of the physical mechanism at work, while section 4 summarizes and further discusses the broader impact of the findings herein.
2. Data and methods
a. NAO index
There is no index that uniquely identifies SNAO dynamics and variability. Different index calculations rely either on the direct instrumental records from individual ground-based stations or on mathematical derivations such as the principal component analysis and consequent empirical orthogonal functions (EOFs; Hurrell 1995; Jones et al. 1997; Rogers 1997). Both approaches have their own unique advantages and disadvantages. For example, the station-based methods relate to an extended temporal coverage (e.g., in some cases back to mid-1800s), while the principal component analyses incorporate more robust spatial–temporal pattern characterization that is independent of season or the observing station’s geographical location (Hurrell et al. 2003). Although the gridpoint-based definitions can be valuable, an SNAO index time series is very useful when obtained from the patterns calculated from the aforementioned mathematical techniques. Small differentiations between the numerically retrieved SNAO indices exist and are due to the comparatively larger seasonal cycle of the NAO pattern shown in Hurrell et al. (2003). For example, the rotated EOF of the NH monthly 700-hPa patterns of Barnston and Livezey (1987), the unrotated covariance EOF patterns of Folland et al. (2009) for July and August, or the Hurrell et al. (2003) covariance EOF pattern for June to August are quite similar. Furthermore, the 700-hPa patterns are similar to those at surface because the SNAO pattern is nearly equivalent barotropic (Folland et al. 2009).
The study at hand employs the 700-hPa patterns derived in Barnston and Livezey (1987). These indices are readily available and up to date from the National Centers for Environmental Prediction (NCEP) and the Climate Prediction Center (CPC; http://www.cpc.noaa.gov/products/precip/CWlink/pna/norm.nao.monthly.b5001.current.ascii.table).
b. NCEP–DOE Global Reanalysis 2
The NCEP–Department of Energy (DOE) Global Reanalysis 2 project is employing a state-of-the-art analysis/forecast system to perform data assimilation using a multitude of datasets (surface, buoy, ship, satellites etc.) from 1979 onward (Kalnay et al. 1996; Kanamitsu et al. 2002; Saha et al. 2006). The present analysis employs thirty years (1979–2008) of monthly NCEP–DOE products of geopotential height (700 hPa), meridional wind (700 hPa), air temperature (700 hPa), and outgoing longwave radiation (OLR; represented as total).
The Microwave Sounding Unit (MSU) operating on the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites has been an important source of information in terms of tropospheric temperatures during the past three decades. The MSU is a cross-track scanner with microwave radiance measurements in four frequencies ranging from 50.3 to 57.95 GHz. Consequently, it retrieves the atmospheric temperature from an equal number of atmospheric layers from the earth’s surface to the stratosphere. As a series of follow-up spaceborne sensors, the Advanced Microwave Sounding Unit (AMSU) began its operation in 1998. The AMSU instruments are similar to the MSU, with the only difference being that AMSU possesses a higher number of microwave channels. The herein employed dataset (version 3.2—www.ssmi.com/msu) pertains to the temperature of the lower troposphere (TLT), retrievals that represent the averaged air temperature of approximately the lower five kilometers of the atmosphere (Mears and Wentz 2005).
Scatterometers are unique among the active satellite sensors, as they possess the ability to retrieve key meteorological parameters such as surface wind. In theory, the emitted microwave radiation is Bragg scattered by the water surface, a medium highly responsive to surface wind changes. The main instrumentation on board Quick Scatterometer (QuikSCAT) is a Ku-band radar. In principle, it is the radar’s backscattered energy that is proportional to the water surface roughness. The dataset used herein is the monthly surface (representative of 10 m) meridional wind component averaged from all satellite orbits spanning from 2000 to 2009 (QuikSCAT CP12 product with ground resolution of 12.5 km). After more than 10 years of successful missions, the QuikSCAT transmission officially ended as of November 2009.
e. HCMR-Poseidon and HNMS
The Hellenic Center for Marine Research (HCMR) Poseidon buoy network monitors a variety of meteorological and oceanographic parameters (e.g., sea level pressure, air–sea surface temperature, wind speed–direction, wave height–direction, salinity, dissolved oxygen, chlorophyll-A, etc). Poseidon became operational in late 1999 and calibrated datasets have been reprocessed for the period 2000–07 (Soukissian et al. 2002; Nittis et al. 2002; Kassis and Nittis 2008).
The Hellenic National Meteorological Service (HNMS) operates a network of ground-based meteorological observations over the Greek continental and islandic peninsula. The 6-hourly reports include among other datasets, air temperature, wind magnitude–direction, and cloud coverage.
In the present study, all surface observations are averaged in monthly values from 18 HNMS meteorological stations (indicated by triangles; Fig. 1) and 5 HCMR Poseidon buoys (indicated by circles; Fig. 1). The employed datasets cover the period 1980–2004 and 2000–04, respectively. Note that from the HCMR dataset several dates are void of data because of scheduled maintenance (the year 2003 is entirely missing except from the Athos buoy-1, and buoys at Lesvos-2 and Mykonos-3 also lack the year 2001, Fig. 1).
f. Numerical methods
The main analysis tools are composed of the Pearson linear correlation and the EOF/singular value decomposition (SVD). Both methods retrieve the spatial and temporal covariability between variables and possess individual advantages and disadvantages. In particular, the SVD analysis is a mathematical method that returns a set of spatial patterns expressed through a set of expansion coefficients that in turn represent the maximum possible covariance between the two variables in the space–time domain (i.e., dominant mode). This property can also act as a disadvantage since any dataset can create random–erroneous patterns. Bjornsson and Venegas (1997) offer an insightful discussion on EOF–SVD analysis. Conversely, the correlation analysis solely returns a single pattern and does not further distinguish between modes of dominant covariability. Neither SVD nor correlation analyses guarantee a cause–effect relationship. Such a deduction must be synergistic and is highly dependent on the information plethora available to the analyst.
In the remainder of this paper, the homogeneous correlation refers to the Pearson correlation (p) between the involved variable and its dominant mode expansion coefficients (i.e., Mode1, as these are derived by the SVD method) and the SNAO monthly indices. The heterogeneous correlation pertains to the Pearson correlation between the featured variable and the SNAO monthly indices. We use the 95% statistical significance level for all correlation calculations (two-tailed p < 0.05. Note that the two-tailed p significance testing imposes a stricter statistical significance condition since it does not assume a priori knowledge of the correlation sign).
3. Results and discussion
Figure 2 illustrates the heterogeneous correlation maps between NCEP–DOE air temperature (700 hPa) and monthly SNAO indices from 1979–2008 (June–September, Figs. 2a–d). Note that no color (white background) is assigned to values below the local statistical significance (p > 0.05). During August, the correlations over the eastern Mediterranean [−0.7 < r(air temperature, SNAO) < −0.5, 0.000 < p < 0.002] rank among the dominant SNAO spatial patterns over the NH midlatitudes. These correlations are indicative of cooler (warmer) air masses over the Greek peninsula, the Anatolian plateau and surrounding waters during relatively higher (lower) SNAO phases (Fig. 2c). During the remaining summer months the featured area relates to marginal (July: −0.5 < r < −0.4, 0.004 < p < 0.02, Fig. 2b) or no statistical significance (June: −0.3 < r < −0.2, 0.1 < p < 0.28, Fig. 2a, September: −0.2 < r < −0.1, 0.28 < p < 0.5, Fig. 2d). Folland et al. (2009) and Bladé et al. (2011) also demonstrate negative correlations between SNAO and air temperatures over the eastern Mediterranean for the months of July and August.
The previous findings are further investigated by the complementary employment of satellite retrievals and ground-based observations. The TLT is representative of the averaged air temperature between 500 hPa and the surface (the weighting functions of the involved atmospheric layers are extensively described in www.ssmi.com/msu and Mears and Wentz 2005). Note that the goal of this paper is not the NCEP–DOE MSU–AMSU comparison but rather the working hypothesis strengthening through the use of independent datasets. Figures 3a–d illustrate the heterogeneous correlation [r(TLT, SNAO)] maps for June–September. As previously stated, no color (white background) is assigned for values below the local statistical significance. The qualitative comparison of Figs. 2 and 3a–d identifies a strong spatial and numerical coincidence between the correlation values in terms of signs and magnitudes. As shown in Figs. 2c and 3c, during August, the SNAO–air temperature covariability over the featured study area is also identified as one of the prominent SNAO spatial patterns over the NH midlatitudes (Figs. 2c and 3c).
Because of the frequent topographic transitions (e.g., the mountainous Greek peninsula, the Aegean Sea, the mountainous Anatolian plateau, the Levantine Sea, and so forth) in addition to the coarse resolution of the NCEP–DOE reanalysis data, lower-tropospheric levels (e.g., 1000/900 hPa) are not included in any of the featured analysis (note that all the hitherto documented trends are also evident at surface level, not shown). Alternatively, we seek further corroboration at the surface level from the observational networks maintained and operated by HCMR and HNMS. Although these datasets encompass a different time period and are of limited spatial representation (i.e., only representative of the Aegean Sea, Fig. 1), they also constitute valuable information given the de facto scarcity of direct ground-based observations over the study area. The monthly surface air temperatures are computed from the averaging of the 6-hourly observations of all available datasets. Figure 4a illustrates the August surface air temperature averages (x axis) plotted against the monthly SNAO indices (y axis). The computed correlation (r = −0.63, p = 0.008) underscores that the strong negative SNAO–air temperature relationship documented in Figs. 2c and 3c is also evident at the surface level. In addition, from Fig. 4a it is indicative that the slope corresponds to an increase of 1.54°C in August temperature for each standardized SNAO unit reduction. The documented correlations become marginally significant (e.g., July: r = −0.35, p = 0.06) or nonsignificant during the remaining summer months (June: r = −0.13, p = 0.49, September: r = −0.17, p = 0.37).
The HNMS-HCMR subdaily reports provide further insight in the SNAO–surface air temperature reported modulation. Figure 4b illustrates the normalized frequency distribution (percent, y axis) for the August maximum daily air temperatures during positive (solid line: 1981, 1982, 1983, 1984, 1988, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2002, and 2005) and negative (dashed line: 1980, 1985, 1986, 1987, 1998, 2000, 2001, 2003, and 2004) SNAO phases. The visual inspection of Fig. 4b clearly identifies that the negative SNAO phase is shifted toward higher maximum air temperatures by approximately 1.5°C. In accordance with the previous findings, the remaining summer months portray an insignificant difference between positive and negative SNAO phases (not shown).
In search for the dominant physical mechanism
The particularity of the hitherto findings can be attributed to the following facts: 1) the documented SNAO–temperature effect peaks during August while it represents one of the dominant signals over the NH midlatitudes; and 2) August is among the warmest months for the eastern Mediterranean hence any possible SNAO involvement should be investigated in more depth. In the following section, basic meteorological variables are compiled to obtain further insight in terms of the physical mechanisms at play.
SVD and correlation analyses are employed between the NCEP–DOE air temperature (700 hPa) and geopotential (700 hPa) for all Augusts from 1979–2008. To account for other dominant SNAO features, each NCEP–DOE variable included in the SVD analysis is encompassed within an area between 30°–70°N, 10°W–50°E. The three leading modes (explaining 33%—hereinafter as Mode1, Mode2 26%, and Mode3 13% of the variance, respectively) account for 72% of the total covariance. The (temporal) correlation between the two variables’ Mode1 returns a statistically significant value [r(Mode1geopotential, Mode1air temperature) = 0.89, p ~ 0]. More importantly, both Mode1air temperature and Mode1geopotential correlate well with the SNAO monthly variability (r = +0.42, p = 0.01 and r = +0.36, p = 0.05, respectively, Figs. 5a,b). The maps of Figs. 5c and 5d illustrate the homogeneous [r(geopotential, Mode1geopotential)] and heterogeneous [r(geopotential, SNAO)] correlations. Note that the solid (dashed) contours refer to the positive (negative) and statistically significant correlations while the dashed-dot contours refer to either sign correlations below the local statistical significance level. Figures 5c and 5d portray two action centers located over 1) the Anatolian plateau (r < −0.4, p < 0.02) indicating the lower-atmospheric pressures that dominate during relatively higher SNAO phases and 2) the northeast Europe–Baltic Sea (r > +0.4, p < 0.02), indicating the higher atmospheric pressures that dominate during relatively higher SNAO phases. The identified SNAO–geopotential spatial patterns weaken during June (r = −0.33, p = 0.06) or become absent (i.e., no statistical significance) during July (r = −0.25, p = 0.36) and September (r = −0.09, p = 0.62, not shown).
Figures 5e and 5f illustrate the homogeneous [r(air temperature, Mode1air temperature)] and heterogeneous [r(air temperature, SNAO)] correlations. These reveal homogeneous and heterogeneous correlations similar to those shown in Figs. 5c and 5d, with lower air temperatures over Greece, the Anatolian plateau and surrounding waters during Augusts of relatively higher SNAO phases (r < −0.5, p < 0.004, Figs. 5e,f). The identified SNAO–temperature spatial patterns weaken (marginal statistical significance) during July (r = −0.33, p = 0.06) and become absent (i.e., no statistical significance) during June (r = −0.25, p = 0.18) and September (r = −0.09, p = 0.6, not shown).
The hitherto findings have shown (Figs. 5c,d) that the identified SNAO–air temperature modulation over the eastern Mediterranean can be attributed to the control that SNAO exerts on the Anatolian low. The latter corroborates the strong physical coupling dictated by the consequent cooling of ascending air over low pressure centers and vice versa. Next, we look for additional implications that arise from the SNAO influence on geopotential. The relative anticyclonic (northern Europe–Baltic Sea) and cyclonic circulations (Anatolian plateau) during higher SNAO phases (Figs. 5c,d) are potential candidates as the regional flow modulators over the study area. More specifically, it is during the summer that the Etesian winds are synonymous to the meridional circulation over the eastern Mediterranean and this is precisely what we investigate next.
The role of the Etesians
The Etesian winds are the dominant summer feature with the highest magnitudes over the entire Mediterranean Sea (Chronis et al. 2011). Because of their northerly origin they work as a “ventilation” system via their dry–cool air advection; this also promotes atmospheric stability conditions (Bartzokas and Houssos 2005; Hadjimichael et al. 2002; Ziv et al. 2004). In addition, their presence has been linked to surface temperature modulations over coastal Greece, Turkey, and the northern African coasts (Carapiperis 1951, 1956; Georgopoulos 2002). Along the same lines, Hadjimichael et al. (2002) and Brody and Nestor (1980) report that conditions of prolonged Etesian dominance emerges from a combination of upper-level troughs (ridges) over the Black Sea and the Anatolian plateau (northern Europe).
SVD and regression analyses are employed between the NCEP–DOE variables of geopotential (700 hPa) and meridional wind (700 hPa) for all Augusts from 1979–2008. The leading modes account for 73% of the total covariance (variance explained 33% in Mode1, 26% in Mode2, and 14% in Mode3, respectively). The correlation between the two variables Mode1 is statistically significant [r(Mode1geopotential, Mode1meridional wind) = +0.56, p = 0.001]. As previously underscored (Fig. 5a), the Mode1geopotential correlates well with the SNAO monthly variability (Fig. 6a, r = +0.36, p = 0.05, solid lines: Mode1geopotential, dashed line: SNAO). Conversely Mode1meridional wind versus SNAO shows marginal correlation (Fig. 6b, r = +0.31, p = 0.08, solid lines: Mode1meridional wind dashed line: SNAO). Figures 6c and 6d illustrate homogeneous [r(geopotential, Mode1geopotential)] and heterogeneous [r(geopotential, SNAO)] correlations that have been discussed in Figs. 5c and 5d. The solid (dashed) black contours refer to positive (negative) and statistically significant correlations while dashed-dot contours refer to either sign correlations below the local statistical significance level.
It is the meridional wind component that introduces new clues to our investigation. Both the homogeneous [r(meridional wind, Mode1meridional wind)] and heterogeneous [r(meridional wind, SNAO), Figs. 6e,f] correlation maps illustrate spatial features of a strong meridional wind component with its main action center located over the northeast Europe–Black Sea (Figs. 6e,f, r < −0.4, p < 0.02). Consistent with our hypothesis, its location can be linked to the relative position of the previously highlighted pressure centers (respective cyclonic and anticyclonic flow, Figs. 5 and 6c,d) that enhances the meridional circulation during Augusts of relatively higher SNAO phases. The latter is corroborated by the north–south and northeast–southwest marginal correlation patterns shown over the Aegean–Levantine Seas (Fig. 6f, r = −0.3, p = 0.1). At lower levels (e.g., 1000 hPa), the aforementioned correlations become more significant over the Aegean and the Levantine Seas (r = −0.4, p = 0.02, not shown). Conversely, the homogeneous [r(meridional wind, Mode1meridional wind)] correlation map (Fig. 6e) fails to capture the meridional wind spatial feature identified in Fig. 6f. Unlike the correlation analysis, the SVD method is sensitive to the encompassed area therefore a complete agreement between the homogeneous and the heterogeneous spatial correlations cannot be constantly achieved. For example, when a less extended area–dataset (e.g., 30°–60°N, 0°–40°E) is employed in the SVD mathematical derivation, the homogeneous–heterogeneous correlations between meridional wind and SNAO highly agree (not shown). This apparent inconsistency may be attributed to the geographical extent of the encompassed area that includes the dominant SNAO features (e.g., Azores–Greenland). For June, July, and September the previous homogeneous–heterogeneous correlations (Figs. 6e,f) become statistically insignificant (June: r = −0.2, p = 0.28, July: r = −0.29, p = 0.12, September: r = −0.17, p = 0.36).
At lower-tropospheric levels, the effects of topography variability (e.g., wind funneling) may lead to the local wind intensification (Kotroni et al. 2001). To investigate the above claims, we employ additional satellite retrievals and surface-based observations. Figure 7a illustrates the heterogeneous correlation map between QuikSCAT meridional wind component monthly averages (2000–09) and monthly SNAO indices for June through August, as these were described in section 2d. From Fig. 7a we observe that statistically significant correlation values (−0.5 < r < −0.7, 0.02 < p < 0.14) highlighting a characteristic “arch” that extends from the north (western Black Sea coast) to the southeast over the Aegean and the Levantine Seas. This pattern spatially and temporally coincides with the typical Etesian winds propagation path, also highlighted in Fig. 6f. These observations independently corroborate the hitherto working hypothesis according to which relatively higher (lower) SNAO phases relate to an enhancement (suppression) of the meridional circulation over the eastern Mediterranean.
As show in Fig. 7a, although QuikSCAT offers the regional characteristics of surface meridional wind, it is also void of data over areas with complex coastlines (e.g., Aegean Sea). To partially overcome this restriction we employ the surface observations from the HNMS/HCMR ground/buoy network (Fig. 1). Figure 7b illustrates the scatterplot between averaged meridional wind magnitudes from all Augusts (1980–2004 for HNMS and 2000–04 for HCMR) and the respective monthly SNAO indices. The computed correlation coefficient (r = −0.54, p = 0.006) also supports the argument according to which the enhancement (suppression) of meridional wind magnitudes during relatively higher (lower) SNAO phases. For the remaining summer months, the computed correlations become marginal (July: r = −0.33, p = 0.06) or weaken below the 95% statistical significance level (June: r = −0.15, p = 0.45, September: r = −0.12, p = 0.57).
Based on composite climatological analysis, we further investigate the particularities of the summer months that describe the SNAO involvement. Figures 8a–d, upper panel illustrate the mean meridional wind climatology while Figs. 8a–d, lower panel illustrate the respective meridional wind anomalies for June–September (NCEP–DOE, 1979–2008, units: m s−1), only during positive SNAO phases (see http://www.cpc.noaa.gov/products/precip/CWlink/pna/norm.nao.monthly.b5001.current.ascii.table). Reiter (1971) suggested a similar method for the identification of the Etesian onset. As a first observation we underscore that over the eastern Mediterranean, the meridional wind spatial distribution (Figs. 8a–d, upper panel) portrays a characteristic “horseshoe” pattern that relates to the northerly winds dominance (i.e., annotated by the blue color or the negative magnitudes). In terms of anomaly magnitudes, the previously reported meridional wind–SNAO correlation pattern identified over the northeast Europe–Black Sea (Figs. 6e,f) is the dominant negative anomaly during June, July, and August (Figs. 8a–c, lower panel). During June (Fig. 8a, lower panel), the aforementioned anomaly feature holds its highest magnitudes (<−1 m s−1) over the Black Sea. During July (Fig. 8b, lower panel) and August (Fig. 8c, lower panel) it propagates southwestwards approaching the eastern Mediterranean with relatively lower magnitudes (>−0.5 m s−1). The main differentiation among the featured summer months is that unlike June (Fig. 8a, lower panel), the months of July and August (Figs. 8b,c, lower panel) show consistent negative meridional anomalies over both the Aegean and the Levantine Seas. On the other hand, although the month of September (Fig. 8d, lower panel) shows consistent negative meridional anomalies over the study area, comparatively to July and August, weaker magnitudes are present. Hereby, it is expected that the featured differentiations would particularly favor the strengthening of the meridional circulation during the months of July and August, a hypothesis that so far has been extensively corroborated.
Although the Etesian winds traditionally relate to atmospheric stability and cloud free conditions, their persistent onset has been linked to frontal passage and altocumulus and orographic cloud formation over coastal Greece and Turkey (Met Office 1962; Metaxas 1977; Metaxas and Bartzokas 1994; Kotroni et al. 2001; Ziv et al. 2004). Hereby, the presence of summer clouds over a region such as the eastern Mediterranean can impose substantial surface radiative forcing. In the absence of long-term in situ shortwave radiation measurements, the HNMS dataset incorporates observations comprised of categories that describe the overall sky condition (ranging from 0 for clear sky to 8 for overcast). Figure 9a illustrates a positive and statistically significant (r = +0.51, p = 0.01) correlation value between the averaged cloudiness (x axis) and SNAO monthly indices (y axis) for all Augusts from 1980–2004. As previously documented, for the remaining summer months, the computed correlations are well below the local statistical significance (June: r = +0.17, p = 0.42, July: r = +0.22, p = 0.3, September: r = +0.14, p = 0.51).
The findings related to Fig. 9a are also corroborated when OLR is implemented as a cloud proxy. The heterogeneous correlation map (Fig. 9b) between OLR (NCEP–DOE, 1979–2008) and August SNAO indices reveals increased cloudiness during relatively higher SNAO phases over the Aegean Sea and as far as northeastern Africa (i.e., negative correlations, Fig. 9b). It is further evident that the previously identified meridional wind–SNAO correlation pattern (e.g., Figs. 6f and 7a,b) spatially agrees with the observed north–south OLR–SNAO correlation pattern shown in Fig. 9b. As previously documented in terms of air temperature (Figs. 2c and 3c), the SNAO influence on OLR over the eastern Mediterranean also ranks as one of the prominent features over the NH midlatitudes (Fig. 9b). Similar correlation patterns with in situ cloudiness data are also documented by Folland et al. (2009).
This study has uniquely identified an SNAO influence on critical climatological variables over the eastern Mediterranean. Higher SNAO phases relate to the relative cooling of the surface/lower troposphere, enhanced meridional circulation, and cloudiness over the Greek peninsula, Anatolian plateau, and the surrounding waters. During the summer months and particularly during August, the identified covariability is among the dominant SNAO-related signals over the NH midlatitudes.
The featured results raise various implications since they pertain to the NH’s warmest season of the year, over an area that has already portrayed “warmer” climatic trends in both terms of air and sea surface temperatures (Raitsos et al. 2010). Especially for the eastern Mediterranean, temperature is a regulatory factor for many socioeconomic aspects such as tourism, health risks (e.g., heat waves), and water resources. At the same time, the Etesian winds have a unique role not only for temperature but also for ocean dynamics (e.g., coastal upwelling), sea state (e.g., wind waves), and biological productivity (e.g., chlorophyll) regulation. Likewise, the identified SNAO–cloud modulation over a traditionally cloud-free area during the summer implies a radiative forcing mechanism that is yet to be quantified. Given the long-term SNAO forecasts by Dorn et al. (2003) and Bladé et al. (2011). or the particularly recent forecasts of the long-term trend toward SNAO positive values by the increasing greenhouse gases (Folland et al. 2009), the results of this paper are critical for the understanding of the eastern Mediterranean climate evolution.
Many thanks to the HCMR-Poseidon team for providing the surface buoy datasets. We further extend our gratitude to Drs. D. Georgopoulos and A. Theocharis for their financial support and constructive comments as well as Dr. David Fanning for the provided ITT-IDL routines on SVD. The authors extend their appreciation to the INFOREST RESEARCH O.C. for the IDL technical support and all the agencies that make the herein employed reanalysis and satellite data freely and readily available. The authors also extend a sincere thanks to Evdoxia Tsimika for the manuscript editing. Finally, the authors extend their appreciation to the three anonymous reviewers for their constructive suggestions–corrections.
“When in return I asked his leave to sail and asked provisioning, he stinted nothing, adding a bull’s hide sewn from neck to tail into a mighty bag, bottling storm winds; for Zeus had long ago made Aiolos warden of the winds, to rouse or calm at will.” (Homer, The Odyssey, Book X, Lines 19–24).