QuickSCAT observations of extreme wind events over the Mediterranean and Black Seas during 2000–2008

Authors


Abstract

A total of 9 years (2000–2008) of QuickSCAT hi-resolution (12.5 × 12.5 km) surface wind observations are employed to identify seasonal means and extreme (gale-force) events over the Mediterranean and Black Seas. The Gulf of Lyon and the Aegean Sea are the regions with the highest sustained wind magnitudes throughout the year. Conversely, the lowest winds are found over the Tyrrhenian, the northern Adriatic and the eastern Black Seas. During winter, the Gulf of Lyon and the Aegean portray mean wind magnitudes ranging from 7 to 11 m/s. During summer, the Aegean Sea portrays the highest wind magnitudes over the entire study area. The lowest wind magnitudes during summer are encountered over the Tyrrhenian, northern Adriatic, eastern Black and eastern Mediterranean Seas (3–6 m/s). From the entire (2000–2008) observations, each QuickSCAT observation exceeding 20 m/s is considered an extreme event. For all the registered extreme events during 2000–2008, we calculate (1) 2-D maps of seasonal frequency occurrence and (2) the histograms of wind directions for ten sub-regions encompassed in the study area (Gulf of Lyon, Balearic Sea, Ligurian Sea, Tyrrhenian Sea, North African coast, Adriatic Sea, Ionian Sea, Aegean Sea, Eastern Mediterranean and Black Seas). Results pertaining to the frequency occurrence show coincidence between high (low) seasonal mean and extreme event frequency occurrence for the areas such as the Gulf of Lyon and the Aegean Sea (Tyrrhenian/southeastern Black Seas). In addition, it is shown that for several cases (Gulf of Lyon, Aegean Sea) the dominant wind direction of extreme events substantially differentiates from the seasonal mean wind directions. Finally we investigate the linearity of extreme event occurrence in terms of mean wind speed climatology over the sub-regions encompassing the study area. Copyright © 2010 Royal Meteorological Society

1. Introduction

Low-level circulation plays a critical role in controlling the climate by establishing the momentum and heat exchange between the Earth's surface and the atmosphere. This ability is manifested from the local (e.g. sea breeze) to the global/regional scale (e.g. coastal upwelling).

Among the world's oceans and semi-enclosed aqueous environments, the Mediterranean Sea is an exceptional case. Besides the presence of more than 2000 islands of various sizes, its coastline surrounds approximately 150 million inhabitants, mountain chains (e.g. Pindos, Alps, Pyrenees, Atlas, Taurus and Caucasus) and more than 30 seasonal named winds (HMSO, 1962; Cavaleri et al., 1991). The meteorological characteristics of the west Mediterranean basin are largely controlled by the (permanent) Icelandic and the Azores pressure centres (Barry and Chorley, 1992; Serreze et al., 1997). The topographic relief and the frequent land/ocean transition (i.e. islands) promote the development of thermal/pressure gradients and lead to the formation of localized air masses (e.g. Etesian winds, Maheras, 1980). To the east, the presence of the Eurasia and Middle East continental masses promote intense cooling/warming, which, depending on the season, favours the development of local thermal circulations with effects pronounced even over the Indian Ocean (Lionello and Sanna, 2005).

The Mediterranean basin counts numerous seasonal/all-year-round winds that frequently evoke gale-force surges. From west to east, we identify the easterly Levanter and its westerly counterpart Poniente. Over the Gulf of Lyon and Genoa (Ligurian Sea), the cold and dry Mistral blows from the north–northwest, frequently reaching the North African coasts and as far as the Ionian Sea. The northeastern influence of the Mistral into the Tyrrhenian Sea is regulated by the upper-level depression, a feature that fully develops over the Ligurian Sea and ranks the area as one of the most active cyclogenic centres in Europe (Flocas, 1988; Trigo et al., 2000). Both the Alps and the Pyrenees create the necessary topographic passages through which the ‘trans-mountainous’ (Tramontana) wind is developed (Estournel et al., 2003). During winter months, the northerly Bora (Greek word for North-Boequation image) dominates over the Adriatic and Aegean Seas (Bergamasco and Gacic, 1996; Poulos et al., 1997; Paklar et al., 2001). During summer and early fall, the pressure gradients developed between eastern Europe and Anatolia give rise to the Etesian (Greek word for Year-'Eτoequation image) winds, which further intensify over the Aegean Sea (Maheras, 1980; Lascaratos, 1992). The southerly (Scirocco) winds originating from the northern African coasts and other parts of the Mediterranean are observed during fall and spring and can also evolve into isolated high wind events (Hadjimichael et al., 2002). All aforementioned winds appear with a variety of names, adjusted to their local characteristics and linguistics (e.g. Etesian/Meltem, Scirocco/Juga, Bora/Vardar, etc—Figure 1).

Figure 1.

Highlighted study regions and dominant winds over the study area. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

The Black Sea is the next major semi-enclosed aqueous environment and a natural extension of the Mediterranean basin. The topography of the area imposes no barrier to the north, thus promoting the high pressure system over Eastern Europe as the major controlling meteorological factor. Towards the eastern Black Sea, the Caucasus Mountains provide the first (after the Ural mountain chain) topographic ‘bump’ after a relatively flat Siberian plateau.

A thorough climatological description of the Mediterranean Sea was obtained as early as 1962 (HMSO, 1962) and more recently by Lionello et al. (2006). May (1982) derived a 20-year climatological description of wind magnitudes from ships of opportunity, although the fact that naval routes exclude high wind areas has raised some concerns regarding possible underestimation. The space–time structure of wind fields in the Black Sea have been investigated by Efimov and Shokurov (2002), Maheras et al. (2001) and Trigo et al. (2000). The relative vorticity has been analyzed by Flocas et al. (2001), whereas the heat and buoyancy fluxes have been studied by Garrett et al. (1993) also employing observations from ships of opportunity. Over the Mediterranean, products such as model simulations, reanalysis and surface buoy observations of surface winds have been reported by Lavagnini et al. (2006). Satellite (QuickSCAT)-derived vorticities and comparisons with regional models over the Mediterranean were also reported by Accadia et al. (2007) and Zecchetto and De Biasio (2007). More recently, global scale wind curl/stress were calculated from QuickSCAT observations (Risien and Chelton, 2008).

The analysis herein presents an original effort to highlight seasonal mean and extreme surface wind features over the Mediterranean and Black Seas based on 9 years of hi-resolution QuickSCAT datasets. Although many of the identified features are referred to as ‘common knowledge’ (e.g. identification of Mistral, the effect of the Azores high, Etesian winds, etc.), the herein employed QuickSCAT dataset provides supplementary information absent from the current literature. Along these lines, Risien and Chelton (2008), based on 8 years worth of global QuickSCAT data, underscore that the sensor's capability of resolving fine spatial scale phenomena has a significant additive value compared to coarser model outputs from various sources (e.g. NCEP–NCAR, ECMWF, MEDAR/MEDATLAS, etc.). In their conclusions, the aforementioned authors point out that the above arguments are particularly valid for areas where topography and local characteristics play a major role.

In Section 2 we describe the employed QuickSCAT dataset and in Sections 2.12.2 we investigate the characteristics of mean wind magnitudes and extreme event spatial distributions and directional characteristics over the Mediterranean and Black Seas. In Section 2.3 we highlight areas where statistically significant trends between the mean and extreme values exist, while in Section 3 we summarize the findings herein.

2. The QuickSCAT dataset

Oceanic surface wind observations have been acknowledged as one of the major breakthroughs of the satellite era. Scatterometers are unique among the active remote sensors as they possess the ability to retrieve both wind magnitude and direction. In theory, the emitted microwaves are Bragg-scattered by the water surface, a medium highly responsive to surface wind changes. The radar's backscatter signal is a multi-parametric function that includes among others, sea surface roughness, the transmitted radar pulse and its incidence angle. The final products (i.e. zonal and meridional wind component) are retrieved via inversion of an empirical model function, linking surface roughness to surface wind.

The first space-based scatterometers became operational in the late 1970s with National Aeronautic Space Administration (NASA) Seasat (Guymer, 1983) and a few years later with the European Remote Sensing Satellite Systems (ERS-1/2) (Offiler, 1994). More recently, efforts by Japan and the UnitedStates led to the launch of the Advanced Observing Satellite (ADEOS) and NASA's Scatterometer (NSCAT) respectively. NSCAT's exceptional retrieval accuracy evolved in the rapid deployment of Sea Winds, an instrument that served as the payload on the QuickSCAT mission. QuickSCAT was operational from June 1999 through November 2009.

The main instrumentation on board QuickSCAT is the Ku-band radar (∼2 cm wavelength), where, as previously highlighted, the radar's backscattered energy is proportional to the ocean's surface roughness. The QuickSCAT Level 2B employs the normalized radar cross-section measurements (namely Level 2A) and further generates grids of wind vector cells (WVC). The products are based on an empirical model function relating the backscattering measurements to wind magnitude and direction in each WVC. QuickSCAT products are available in resolutions of 25, 20 and 12.5 km.

Each QuickSCAT observation goes through a standard quality control based on the combined usage of ancillary variables (e.g. WVC_QUAL_FLAG, WVC_SELECTION, etc.). The overall quality of QuickSCAT retrievals depends highly on the presence of land, ice/snow/ precipitation as well as the orbital characteristics such as the satellite sub-pixel distance from the nadir (Bourassa et al., 2003; Tang et al., 2004; Bentamy et al., 2005). The aforementioned quality control automatically excludes observations ‘contaminated’ by land and hydrometeor scattering in deep convective clouds or where the return signal is severely deteriorated by the presence of precipitation. More details on the data quality and filtering processes, as suggested by the standard QuickSCAT data processing release, can be found at ftp://podaac.jpl.nasa.gov/ocean_wind/quikscat/L2B/doc/QSUG_v3.pdf (NASA, QuickSCAT Science Data Product, 2006). The authors acknowledge the fact that high wind magnitudes frequently relate to precipitating environments where QuickSCAT may contain additional errors. Presently, there is limited knowledge regarding the evaluation of QuickSCAT retrievals where precipitation is present; therefore, the results herein are based on the highlighted quality flagging as was provided by the JPL/NASA.

The main diagnostic variable used in the following analysis is the 12.5 km (namely CP12) QuickSCAT product swaths. These include the qualified observations from January 2000–December 2008 mapped on a 12.5 × 12.5 km2 regular grid. Because a given grid point may be ‘seen’ twice (ascending/descending orbit) during one day, the final daily grid product incorporates u/v wind component averaging. Hereinafter, this study will be frequently referring to specific regions of the encompassed study area. Figure 1 identifies ten colour highlighted regions: Gulf of Lyon, Balearic Sea, Ligurian Sea, Tyrrhenian Sea, North African coast, Adriatic Sea, Ionian Sea, Aegean Sea, Eastern Mediterranean and Black Seas. Due to the Sea of Azov's shallow bathymetry and low salinity (thus prone to freezing) and regardless of the QuickSCAT quality control (pass or fail), all calculations/discussions exclude this region entirely.

Admittedly, there is no objective criterion based on which the geographical cropping of Figure 1 is dependent. At the same time, the selected regions do not strictly encompass the area whose name is given. This step is necessary so that some regions are studied separately simply due to their spatial features (e.g. north–south elongated regions such as the Adriatic Sea constrained by the Italian and Balkan Peninsula or on the contrary, more open environments such as the eastern Mediterranean).

2.1. Wind (magnitude/direction) seasonal means

From the daily CP12 QuickSCAT daily grids we calculated the 9-year seasonal means of wind speed and direction (Figure 2(a–d) DJF, MAM, JJA and SON). Grey colour relates to grid points that did not pass the quality filtering (i.e. coastal areas). Note that for visual clarity purposes, the wind vectors that accompany Figure 2 are illustrated in lower resolution and are not magnitude dependent.

Figure 2.

Seasonal mean (2000–2008) wind magnitude/direction from QuickSCAT CP12 product (12.5 × 12.5 km) for (a) Winter (DJF), (b) Spring (MAM), (c) Summer (JJA) and (d) Fall (SON) for the Mediterranean and Black Seas. Gray areas represent disqualified QuickSCAT observations

During DJF (Figure 2(a)), both the Mediterranean and Black Seas have their respective highest mean values compared to any other season. Throughout the year, the Gulf of Lyon sustains wind magnitudes higher than 7 m/s (Figure 2(a–d)) and predominantly during DJF (8–11 m/s), it extends its influence to the northeast and over the Ligurian Sea. In particular, the Gulf of Lyon represents the exit point of the Tramontane and Mistral winds and because of their prevailing direction, their effect is additive rather than cancelling. The Aegean Sea represents the next most distinctive feature over the encompassed study area. The wind ‘lobes’ to the east and west of the island of Crete correspond to the respective exiting points of the northerly flow. These features are not only dominant during JJA (Etesian winds, Maheras, 1980, Figure 2(c)) but also pertain to the winds with consistently the highest sustained magnitudes over the entire Mediterranean and Black Seas during all seasons (Figure 2(a–d)). Also during JJA, the other prominent feature of the entire study area pertains to the the wind direction reversal over the Balearic Sea (north–westerlies—Figure 2(a) to north–easterlies—Figure 2(c), Dorman et al., 1995; Losada, 1999).

The relatively lower wind magnitudes over the Tyrrhenian Sea is an example of topographic forcing, an interaction that is promoted by the ‘sheltering’ from the dominant northwesterly flow provided by the presence of the Corsica/Sardinia (Figure 2(a–d)). Note that the Tyrrhenian Sea depicts mean magnitudes that rank among the lowest throughout the year (3–6.4 m/s). Over the Adriatic Sea, easterly/northeasterly flow is present during DJF, MAM and SON (Figure 2(a,b,d)), while during JJA, the dominant regime follows the generalized northerly flow of the eastern part of the Mediterranean basin (Figure 2(c)).

During DJF and SON, the Black Sea mean wind magnitudes are shown to be spatially stratified, with the lower values gradually increasing from the southeast to the northwest. In terms of seasonal wind direction, the Black Sea shows prevailing northeasterly winds to the west and northwesterly winds to the east of the Crimean peninsula (Sevastopol), observation mostly evident during MAM. The southeastern Black Sea registers wind magnitudes between 3–4.7 m/s during any given season, values that also rank among the lowest over the entire study domain.

2.2. Wind seasonal extremes

2.2.1. 2-D maps of extreme event frequency occurrence

From the previous compiled QuickSCAT daily grids, we account for the ‘extreme events’, i.e. where the wind speed value exceeds 20 m/s. Note that the term ‘extreme’ is arbitrary and is mainly used for convenience while it certainly does not reflect an absolute global extreme. From the employed dataset, 62 234 extreme events were observed during DJF, 22 038 during MAM, 5465 during JJA and 39 822 during SON over the Mediterranean and Black Seas. Note that the aforementioned observations pertain only to the Mediterranean and Black Seas (although Figures 2 and 3 include regions such as, e.g. Gulf of Biscay, Sea of Azov, these are excluded from any of the featured results or discussions). To relate the registered extreme events to the necessary spatial information, we proceed to the extreme event frequency calculation (hereinafter frequency) for every {i, j} grid point during each season based on the simple averaging formula equation image (Figure 3(a–d)). Since each {i, j} grid point is tied to a daily observation, the above ratio represents the percentage of days during which wind magnitudes exceeded the 20 m/s threshold value.

Figure 3.

Seasonal extreme event (>20 m/s) occurrence frequencies % (2000–2008) from QuickSCAT CP12 product (12.5 × 12.5 km) for (a) Winter (DJF), (b) Spring (MAM), (c) Summer (JJA) and (d) Fall (SON) for the Mediterranean and Black Seas. Grey areas represent disqualified QuickSCAT observations

Throughout Figure 3, it is DJF (Figure 3(a)) that presents most of the yearly variability (encompassing from the lowest to the highest values). The Gulf of Lyon is the region that consistently relates to frequencies exceeding 0.5% and reaching their respective maxima during DJF (2–8%). The high spatial resolution of the employed QuickSCAT dataset further reveals that the coastal region immediate to the Gulf of Lyon shows discernible differentiation with significantly lower (<1%) frequencies compared to open sea frequencies (2–8%). The latter could be related to the angle at which the Tramontana and Mistral winds blow over the region, such that it shelters the centre of the Gulf from the relatively higher wind magnitudes few kilometres to the southwest/southeast (Figure 3(a)). The regions neighbouring the Gulf of Lyon to the east (Ligurian Sea, Gulf of Genoa) show frequencies ranging between 0.5 and 2.0%, while the region to the west (Balearic Sea) barely exceed 0.25%.

The next readily distinctive feature is the Ionian Sea, with its greater part ranging between frequencies 0.5 and 1.5%. The Aegean Sea, despite the observational gap (i.e. presence of islands) shows frequencies ranging between 1 and 5%. The topographic particularities of the Aegean Sea such as wind funnelling through the numerous islands present can enhance the presence of extreme events (Kotroni et al., 2001). Moreover, the previously highlighted wind ‘lobes’ to the east and west of the island of Crete (Figure 3(a)) are also accompanied by frequencies ranging between 1 and 2%. It is worth noting that the traditional wintertime cyclogenetic centres of the Mediterranean (Gulf of Genoa, Aegean and the northeastern Mediterranean-Cyprus and Turkey-bay of Antalya) all relate to frequencies higher than 1% (Trigo et al., 2000; Maheras et al., 2002). The Black Sea portrays frequencies exceeding 0.75% to the west and north while relatively lower values to the east.

During MAM, the only distinctive feature is the Gulf of Lyon (>1.5%), whereas the remaining study area relates to frequencies around 0.5% (Figure 3(b)). During JJA, the Mediterranean and Black Seas also relate to frequencies less than 0.5%. A noticeable differentiation from MAM is that the northern Africa coast and the entire southeastern Mediterranean present a significant reduction of extreme events. The SON frequencies are very close to the frequencies observed during DJF in both terms of percentages as well as spatial distribution, although frequencies greater than 2.0% are exceeded only over the Gulf of Lyon (Figure 3(d)).

2.2.2. Regional characteristics of extreme event frequency occurrence

Next, we account for the extreme events registered at every {i, j} grid point encompassed in each of the highlighted regions (Figure 1). The extreme event counts are documented as follows: (Region: DJF/MAM/JJA/ SON—Table I); Gulf of Lyon: 22 672/9847/1064/13 941, Balearic Sea: 1506/1099/420/2613, Ligurian Sea: 3372/1218/179/2280, Tyrrhenian Sea: 3273/1320/312/ 3268, North African coast: 6111/1965/139/3848, Adriatic Sea: 1236/636/679/1555, Ionian Sea: 5409/937/266/3504, Aegean Sea: 9484/2357/166/3182, Eastern Mediterranean Sea: 5452/954/48/1994 and Black Sea: 5719/1750/2192/ 3637. From the above results, we note the following: for eight out of ten regions, the maximum number of extreme events is observed during DJF with SON ranked second (on average, DJF is 30–40% higher), while for the Tyrrhenian Sea, DJF/SON are approximately equal. Exceptions to the aforementioned behaviour are the Balearic and Adriatic Seas with their maximum observed during SON. Additionally, for eight out of ten regions, JJA has the lowest extreme event occurrence, excluding the Adriatic and Black Seas, where JJA ranks third (after DJF/SON). The maximum/minimum frequency ratio is a simple way to compare variability between seasons. The associated values are registered as 5452/48 ∼ 113 over the Eastern Mediterranean and 1555/636 ∼ 3 over the Adriatic Seas, respectively.

Table I. Total number of extreme events observed by QuickSCAT for each of the ten highlighted regions for all four seasons
 DJFMAMJJASON
Balearic Sea150610994202613
Gulf of Lyon226729874106413941
Ligurian Sea337212181792280
Tyrrhenian Sea327313203123268
Northern Africa611119651393848
Adriatic Sea12366366791555
Ionian Sea54099372663504
Aegean Sea948423571663182
Eastern Mediterranean5452954481994
Black Sea5719175021923637

In addition, for each {i, j} grid point, the registered wind directions of extreme events are categorized as N (337.5°–22.5°), NE (22.5°–67.5°), NW (292.5°–337.5°), E (67.5°–112.5°), W(247.5°–292.5°), S (157.5°–202.5°), SE (112.5°–157.5°) and SW (202.5°–247.5°). No averaging (i.e. u/v components) is performed if a grid point has registered more than one extreme event. Consequently, we construct a histogram for each region/season (Figure 4(a–d)). Hereinafter, we use several ‘house-keeping’ variables (e.g. ratios between northerlies versus southerlies, westerlies versus easterlies) that will help us pinpoint several unique as well as common climatological features of the highlighted regions. Note that for convenience we refer to northerly as the total percentage (%) of N/NW/NE, southerly the total % of S/SW/SE and so forth. Key observations pertaining to Figure 4 are outlined as follows: (1) dominance (>50%) of northerly winds during DJF for the Gulf of Lyon, Ligurian, Aegean and Black Seas; (2) for the geographical area encompassing the northern African coast, the Adriatic, Ionian and Aegean Seas, during all seasons the ratio of easterlies (%)/westerlies (%) is consistently greater than 1.0 and reaches its maximum (3.4) over the Aegean Sea and its minimum (0.17) over the Balearic Sea during JJA; (3) for the eastern Mediterranean and Black Seas, the opposite behaviour is observed, with the easterlies (%)/westerlies (%) ratio consistently lower than 1 (except DJF-Black Sea); (4) out of the ten highlighted regions and four seasons (total 40 cases), only 5 cases have the ratio northerlies (%)/southerlies (%) less than 1 (the Adriatic Sea—0.75, during MAM, the Thyrrenian and east Mediterranean Seas—0.96/0.66 during JJA and the Thyrrenian and Adriatic Seas—0.90/0.59 during SON (5) the Gulf of Lyon represents the region that registers the lowest easterlies (%) and the highest northerlies (%)/southerlies (%) than any other region during all seasons.

Figure 4.

Histogram of the extreme event wind directions for each highlighted sub-region. Wind directions are categorized as N (337.5°–22.5°), NE (22.5°–67.5°), NW (292.5°–337.5°), E (67.5°–112.5°), W(247.5°–292.5°), S (157.5°–202.5°), SE (112.5°–157.5°) and SW (202.5°–247.5°)

The question we address now is the following: does climatology dictate the direction of extreme wind events? For example, if over a given region the % of NE of extreme events accounts for 50% then how much does this differ from the mean (directional) climatology? The directional categorization (N, NE, NW and so forth) employed in Section 2.2.2 is repeated, although this time we account for all {i, j} wind observations and not particularly for the extreme events (>20 m/s). In the same manner, we construct a histogram similar to Figure 4 (not shown) from which we compute all the differences for all regions, seasons and directions (ten regions × four seasons × eight directional categories = 320) as: DIRclimat%− DIRextreme%. Note that the term DIRextreme% was previously computed and displayed in Figure 4. The resulted differences (units : percentages) are fitted in a normal distribution with µ ∼ 0.01 and σ ∼ 8.8. Table II reports the DIRclimat%− DIRextreme% differences exceeding the µ + /− 1.96σ (95%). The authors acknowledge the fact that had the computed differences included exclusively a particular region/season, the µ + /− 1.96σ reported threshold would have also been different. Nevertheless, the intention here is to underscore the major differences between climatology and extreme events; Gulf of Lyon: extreme event % from the NW exceed the climatology during DJF, MAM and JJA (−29.3, − 30.7 and − 17.7% accordingly). North Africa coast: climatology from the N leads the extreme % during JJA (20.9%) Adriatic Sea: extreme events % from the E lead the climatology during MAM (−22.0%) and SON (−21.2%) Ionian Sea: Similar to the Adriatic Sea, the extreme events from the E lead climatology % during MAM (−21.4%) and SON (−19.8%). Conversely, climatology from the N and NW leads the extreme during JJA by 17.6 and 25.2% accordingly. Aegean Sea: extreme events from the E lead the climatology % during JJA (−29.2%). Conversely, climatology from the NW leads the extreme % during JJA (38.4%). Eastern Mediterranean: similar to the Aegean case, extreme events from the E lead climatology (−24.0%) and climatology from the NW leads the extreme events % during JJA(21.6%) Black Sea: extreme events % from the W leads the climatology % during JJA (−20.6%).

Table II. The percentage difference (DIRclimat%− DIRextreme%) for all regions/seasons where this exceeds the µ+ /− 1.96 σ rule (95%)
 DJFMAMJJASON
Balearic Sea
Gulf of Lyon− 29.3% (NW)− 30.7% (NW)− 17.7% (NW)
Ligurian Sea
Tyrrhenian Sea
Northern Africa20.9% (N)
Adriatic Sea− 22.0% (E)− 21.2% (E)
Ionian Sea− 21.4% (E)17.1%/25.2% (N/NW)− 19.8% (E)
Aegean Sea− 29.2%/38.4% (E/NW)
Eastern Mediterranean− 24.0%/21.6% (E/NW)
Black Sea− 20.6% (W) 

The common characteristic of the previously highlighted results is that during DJF (except the Gulf of Lyon), no (statistical, as herein defined) differentiation is noted. A possible reason for this is that during the winter season, the overall circulation patterns are extensively controlled by the semi-permanent pressure centres. During the rest of the year, other physical and locally characteristic mechanisms such as sea breeze, thermal lows, topography (e.g. funnelling) and transient depressions may be responsible for the observed differences. For example, during JJA and over the Aegean Sea and the eastern Mediterranean, the dominant wind regime is from the N/NE/NW (Etesian). Although quite infrequent during the summer, transient surface lows moving along a W–E/NW–SE track (Trigo et al., 2000; Maheras et al., 2002 for August) and combined with the seasonal thermal low over the Anatolian plateau may give rise to a strong E wind component, depicted in the aforementioned difference (−24.0%). The same argument could apply for other neighbouring regions such as the Adriatic and Ionian Seas.

2.3. Trends between seasonal extremes and means

The qualitative evaluation of Figure 2(a–d) and Figure 3(a–d) reveals something rather intuitive; extreme event occurrence is (comparatively) higher where the mean seasonal wind magnitudes are (comparatively) higher. The Gulf of Lyon is an example that corroborates the previous claim. However, although the Aegean Sea registers mean wind magnitudes comparable to DJF during JJA, the associated extreme event counts are by far different (Figures 2(c) and 3(c)). Can the aforementioned argument be generalized? Would an increase (or decrease) of mean wind magnitude over a certain region lead to an increased (decreased) extreme event frequency occurrence?

Each {i, j} grid point encompassed in any of the ten regions is tied to a single seasonal value (Figure 2(a–d)) although at the same time it is also tied to an extreme event occurrence (i.e. counts). For each {i, j} grid point included in the highlighted regions (Figure 1), we register (1) its seasonal mean wind magnitude and (2) extreme events counts. Between (1) and (2) and for each of the regions we perform a first degree polynomial trend analysis. One could argue that higher order polynomials might better describe (1) versus (2) although the explicit study of such mathematical implementation is beyond the scope of this paper. Depending on the degrees of freedom for each region (function of the total grid points with at least one extreme event registered), the t-test is computed (not shown). Only the statistically significant results at the 95% significance level will be reported. Figure 5 illustrates the scatter plots between all the regions for DJF and where statistical significance is reached, the slope, intercept and Pearson correlation are also reported. The featured results (Figure 5) illustrate two regions, which during DJF meet the desired significance level. These are the Gulf of Lyon and the Aegean Sea, with linear correlations computed as ρ = + 0.88 and ρ = + 073, accordingly. For MAM/SON, only the Gulf of Lyon produces statistically significant and approximately equal slopes/intercepts (−90.30/7.8, ρ = 0.86—not shown). JJA did not yield any statistical significance for any of the highlighted regions.

Figure 5.

Trends (DJF) between mean wind magnitude (x-axis) and the number of extreme wind events (y-axis). Slope and intercept are reported only for statistically significant results (t-test > 95%)

It could be argued that the preceded trend analysis is (1) region-dependent as the geographical regional-cropping (Figure 1) was arbitrary and (2) the selection of 20 m/s as a value for the extreme event threshold was also subjective (Section 2.2); hence it cannot provide a bulletproof representation of other underlying trends. Nevertheless, the regression analysis shows that some regions demonstrate statistically significant (positive) trends while others do not. In general, the larger the scattering of extreme event counts (Figure 5y-axis) for a given mean wind magnitude value (Figure 5x-axis) the lower the correlation. In an ideal case (ρ = 1.0) and under the positive trend postulation (i.e. increased mean wind speed leads to increased extreme event counts), each mean wind magnitude registers one unique value of extreme event counts. In our case, the significant correlation values shown in Figure 5 depict the ‘responsiveness’ of a region in terms of extreme event occurrence at a given mean wind speed change. The analysis herein can be also employed in a limited-area approach, where the correlations shown in Figure 5 could be substantially stronger. As an example, the nonsignificant correlation for entire eastern Mediterranean Sea (Figure 5) becomes significant (ρ = 0.85) when only the bay of Antalya is accounted for (not shown). Such information combined with long-term forecasts (e.g. downscaling techniques in regional climate models) could provide estimates of extreme event frequency. The authors are not fully ready to discuss the presence or absence of statistically significant trends as these may involve synergistic physical mechanisms such as topography, seasonality or more complex air–sea interaction processes such as local surface turbulence and local climatic influences (SST gradients, currents, etc.).

3. Conclusions

This contribution has compiled 9 years of QuickSCAT hi-resolution surface winds over the Mediterranean and Black Seas. The analysis depicted key seasonal characteristics of the extreme states of the Mediterranean and Black Seas as well as similarities and differences with respect to the dominant ‘climatology’. In addition, this paper shows that extreme winds cannot be considered a priori knowledge with the climatology being the only point of reference. Further corroboration of the findings herein can be highly complementary to climate and downscaling modelling, energy budgets during extreme wind events (i.e. latent heat) and renewable energy and risk assessment studies over coastal and open sea areas. Several ambiguities tied to the preceding analysis can originate from the unavoidable under-sampling that every non-geostationary satellite relates to, especially when the latter pertains to a highly variable observation such as the surface wind. Typical examples of potentially unaccounted extreme events relate to diurnal forcing (e.g. sea/land breeze, topographic complexity) or, simply put, any ‘extreme’ event occurring outside the satellite's orbit. Nevertheless, the study's temporal and spatial extent allows the acquisition of the several key characteristics tied to the extreme wind regimes.

Since November 2009 and after completing more than 10 years of a highly successful mission, QuickSCAT stopped the data transmission. Currently, the Advanced SCATterrometer (ASCAT) on the European Space Agency's (ESA) METeorological OPerational satellite (METOP) is the instrument dedicated to the oceanic wind surface observations. Although the ASCAT sensor has a less extended swath than its predecessor, it has enhanced capabilities of observing through precipitating clouds (C-band). ASCAT-based algorithms are currently being validated with the aid of off-shore buoy observations (Nittis et al., 2002; Kassis and Nittis, 2008) and will enable wind retrieval near complex coastlines (e.g. Aegean Sea, Vogelzang et al., 2009, personal communication). Alternatively, ultra hi-resolution oceanic wind-retrievals over open ocean and near coastline environments is an additional advancement currently provided by the ESA's ENVIronmental SATellite (ENVISAT) and its Advanced Synthetic Aperture Radar (ASAR) although its use in temporally extended studies is yet to be investigated.

Acknowledgements

Special thanks to Evdoxia Tsimika for the technical editing of the manuscript. Additional thanks for the QuickSCAT data availability as well as the IDL codes from podaac.jpl.nasa.gov/pub/ocean_wind/quikscat/. Finally the authors extend their sincere appreciation to the anonymous reviewers for their highly constructive comments.

Ancillary

Advertisement