ClickHere

for

FullArticle

Possible atmospheric origin of the 7 May 2007 western Black Seashelf tsunami event

Ivica Vilibić,1 Jadranka Šepić,1 Boyko Ranguelov,2 Nataša Strelec Mahović,3

and Stefano Tinti4

Received 12 October 2009; revised 15 January 2010; accepted 15 February 2010; published 8 July 2010.

[1] The paper examines the possibility that tsunami event recorded in the western BlackSea on 7 May 2007 was triggered by a traveling atmospheric disturbance. Asmeteotsunamis are favored by specific synoptic conditions, we inspected available groundand sounding observations, European Centre for Medium‐Range Weather Forecastsreanalysis fields, and satellite‐based products and compared them to the documentedMediterranean meteotsunamis. We found an atmospheric disturbance traveling toward30° (NNE) with amplitude of 2–3 hPa and propagation speed of about 16 m/s, passingthrough few tens of kilometers wide pathway over the region affected by the tsunami. Thisdisturbance occurred in the lower troposphere, but it was capped by instable convectivecell that preserved gravity disturbance’s coherence over a region at least 150 km long. Anocean modeling study showed that such a disturbance is capable of generating largetsunami waves and strong currents over the shallow regions, following the observationsover the region where maximum sea level oscillations have been documented. Therefore,this event has a potential to be classified as a meteotsunami, the first of such kind in theBlack Sea.

Citation: Vilibić, I., J. Šepić, B. Ranguelov, N. S. Mahović, and S. Tinti (2010), Possible atmospheric origin of the 7 May 2007western Black Sea shelf tsunami event, J. Geophys. Res., 115, C07006, doi:10.1029/2009JC005904.

1. Introduction

[2] Although tsunamis in the Black Sea are not as fre-quent and intense as in the nearby Marmara and AegeanSeas, they can be quite destructive especially if accompa-nied by a strong nearshore earthquake [Yalçiner et al.,2004]. Most of the Black Sea tsunamis are of seismic ori-gin and have been reported along the northern Crimean andCaucasus coastlines [Pelinovsky, 1999; Solov’eva andKuzin, 2005], and less along the Turkish [Altinok andErsoy, 2000] and Bulgarian [Ranguelov and Gospodinov,1995] coasts. Several destructive and tsunamigenic earth-quakes occurred off the western coastline during the lastmillenniums. Of these events the strongest tsunami wasreported in 543 A.D., causing large flooding in Varna Bayand the Balchik area with runup exceeding 2–4 m [Nikonov,1997]. Just a decade later, in 555 A.D., a nonseismic tsu-nami has been reported in the southwestern Black Sea by theancient Byzantine chronicler Nikomedius [Ranguelov,1996], giving reliability to a hypothesis that submarine

landslides or the atmosphere may create destructive tsunamiwaves in the Black Sea. In addition, paleostudies indicatethat major tsunamis may occur along the western Black Seaover a geological timescale [Ranguelov, 2003].[3] However, no tsunamigenic earthquake has been

reported to occur on 7 May 2007 in the Black Sea area,although tsunami waves of 2–3 m height have beenobserved in the morning hours along the northern section ofthe Bulgarian Black Sea coastline. Two hypotheses havebeen developed in order to detect the source of these non-seismic waves [Ranguelov et al., 2008]: (1) a submarinelandslide occurred about 50 km off the western coast, alongthe shelf break perimeter, generating the tsunami waveswhich were amplified along the double‐beam waveguide,and (2) an atmospheric high‐frequency disturbance traveledalong the shelf and generated long ocean waves through theProudman resonance mechanism [Proudman, 1929]. Theformer hypothesis has been explored by Ranguelov et al.[2008], who concluded that “submarine mass movementstaking place within a certain delimited source area offBulgaria may have generated tsunamis compatible with theobservations.” The latter hypothesis will be elaborated inthis paper—the tsunami waves generated through such amechanism are known as meteotsunamis [Monserrat et al.,2006], although these events have so far not been docu-mented to occur in the Black Sea giving such large effects.[4] Meteotsunamis ormeteorological tsunamis are observed

to occur regularly at certain places in the World Ocean, andeven have specific local names: “rissaga” in the Balearic

1Institute of Oceanography and Fisheries, Split, Croatia.2Geophysical Institute, Bulgarian Academy of Sciences, Sofia,

Bulgaria.3Meteorological and Hydrological Service, Zagreb, Croatia.4Department of Physics, University of Bologna, Bologna, Italy.

Copyright 2010 by the American Geophysical Union.0148‐0227/10/2009JC005904

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, C07006, doi:10.1029/2009JC005904, 2010

C07006 1 of 12

Islands [Ramis and Jansà, 1983; Tintoré et al., 1988];“marubbio” in Sicily [Candela et al., 1999]; “milghuba” inMalta [Drago, 2008], “abiki” in Nagasaki Bay, Japan[Hibiya and Kajiura, 1982], and “Seebär” in the Baltic Sea[Metzner et al., 2000]. These waves are also documented inthe Yellow Sea [Wang et al., 1987], the Adriatic Sea [Vilibićet al., 2004; Vilibić and Šepić, 2009], the Aegean Sea[Papadopoulos, 1993], the English Channel [Douglas,1929; Haslett and Bryant, 2009], the Great Lakes [Ewinget al., 1954], Florida [Churchill et al., 1995], the north-western Atlantic coast [Mercer et al., 2002], the Argentinecoast [Dragani et al., 2002], and the New Zealand coast[Goring, 2005]. Meteotsunami wave heights can be as highas 6 m (Vela Luka, Adriatic Sea; Daytona Beach, Florida) or5 m (Ciutadella inlet, the Balearic Islands; Nagasaki Bay,Japan) or less, and the waves can cause substantial damageto coastal infrastructures and produce human injuries andlosses [Donn and Ewing, 1956; Hibiya and Kajiura, 1982;Monserrat et al., 2006; Vučetić et al., 2009]. The most“problematic” regions are low‐tidal basins such as theMediterranean Sea, since the coastal infrastructure is notadapted to such large sea level oscillations.[5] The source of meteotsunami waves is a high‐

frequency atmospheric gravity disturbance visible in theground air pressure series. Air pressure gradient is particu-larly important as it directly affects the ocean (enters directlyinto the ocean equations), and it may surpass 5 hPa over5 min during the strongest events [Vilibić et al., 2008].Moreover, each destructive meteotsunami event also needs anondissipative atmospheric disturbance traveling over ashelf with the speed similar to the speed of the tsunamiwaves (Proudman resonance). Then, the generated open‐ocean waves are amplified topographically close to thecoast, especially within a harbor or a bay with largeamplification factor [Monserrat et al., 2006]. However,nondissipative atmospheric gravity disturbances are occur-ring rarely and only during specific conditions, whichinclude ducting of the atmospheric disturbance in the lowertroposphere below an unstable layer [Lindzen and Tung,1976; Monserrat and Thorpe, 1992, 1996]. Such condi-tions are usually related to typical synoptic conditions,which include temperature inversion in the first kilometerabove the ground, overtopped by the flow of warm and dryair, and capped with an instable layer where the wind speedequals the speed of disturbance [Ramis and Jansà, 1983;Vilibić and Šepić, 2009]. The disturbance is more efficientlymaintained over large distances if capped not only byinstable layer itself but also by a convective system thatencompasses much of the upper troposphere [Belušić andStrelec Mahović, 2009; Šepić et al., 2009a].[6] In this paper all of these conditions will be evaluated

using available ground and sounding observations, satellite‐derived products, and atmospheric reanalysis fields, in orderto assess the possibility that the atmospheric processes wereresponsible for the generation of the strong tsunami wavesthat were observed along the western Black Sea coast on7 May 2007. These analyses are part of section 3, precededby an overview of the available material, data, and methodsin section 2. Section 4 deals with the results obtained from abarotropic numerical ocean model forced by a traveling

atmospheric disturbance, while further discussion and con-clusions are presented in section 5.

2. Material and Methods

[7] Not so many in situ ocean and meteorological mea-surements along the western Black Sea coastline wereavailable during the tsunami event of 7 May 2007, espe-cially concerning high‐frequency (a minute timescale) data.Ocean observations were available at a number of locations,including tide gauge measurements conducted at Ahtopoland Varna and eyewitnesses’ reports. A comprehensivereport on the observations is given by Ranguelov et al.[2008], and therefore only range values of the maximumand minimum sea levels will be taken into account here(Figure 1). Ranguelov et al. [2008] summarize the majorcharacteristics of the ocean observations as: (1) sea levelretreat was larger than sea level rise, (2) turbulence and mudcurrents were reported, (3) oscillations had 4–8 min period,and (4) maximum and minimum sea levels were +1.2 and−2.0 m, respectively. It is also worthwhile to say that noearthquake occurred in the region at that day.[8] As seismic origin of the tsunami was unrealistic due to

the absence of earthquakes, Ranguelov et al. [2008] assesseda possibility that a submarine landslide generated theobserved tsunami. They also introduced a possibility thatmeteorological forcing was the generator of the event. Asour study is focused on the latter option, we tried to analyzesynoptic situation and available ground, satellite, and radiosounding measurements and to compare it to the char-acteristics documented for other meteotsunamis [Monserratet al., 2006; Belušić and Strelec Mahović, 2009; Vilibić andŠepić, 2009]. For that purpose air pressure weekly chartshave been analyzed for a number of locations (Figure 1), buttheir resolution and quality did not allow for high‐resolutiondigitizing (e.g., at the 5 min resolution) and in‐depth anal-ysis on the minute timescale. However, sudden variations inair pressure are captured by a number of barograms, indi-cating the propagation of the atmospheric disturbance visi-ble in ground air pressure.[9] We also used the profile of the radio sounding carried

out at the Istanbul station (taken fromUniversity ofWyomingWeb site http://weather.uwyo.edu/upperair/sounding.html),measuring vertical profiles of wind, temperature, andhumidity (dew‐point temperature).Moreover, the Richardsonnumber Ri was derived from these measurements as

Ri ¼ N 2

ðdu=dzÞ2 ð1Þ

where N is the Brunt‐Väisälä (BF) frequency, u is the windspeed, and z is the height. The BF frequency was calculatedas the moist BF frequency [Durran and Klemp, 1982] on thelevels with high relative humidity (here taken as higher than90%) and as the dry frequency otherwise. A layer wasconsidered unstable if its Richardson number was lower than0.25. Unstable layers are favorable for the reflection andducting of the lower troposphere gravity disturbances [Šepićet al., 2009a], providing that the wind speed at the unstablelevel equals the propagation speed of the disturbance[Lindzen and Tung, 1976].

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

2 of 12

[10] Synoptic conditions have been analyzed using tem-perature, wind, relative humidity and geopotential values at500, 700 and 850 hPa from the European Centre forMedium‐Range Weather Forecasts (ECMWF) reanalysisdata. Further, the minimum Richardson number between300 and 600 hPa has been spatially examined, in order tomap the areas favorable for conductive tunneling of theatmospheric gravity disturbances. Notice that as for formula(1) the BF frequency was computed as the moist BF fre-quency in areas with high relative humidity and as dryfrequency elsewhere, the only difference being that here alower threshold was taken—70% opposed to 90%—toaccount for coarser vertical resolution data.[11] Meteosat Second Generation satellite images were

used for the recognition of convective clouds that passedover the western Black Sea area. For the detection of thecloud characteristics, 10.8 mm infrared channel images wereused. Cloud motion vectors, defining the speed and direc-tion of the convective clouds, were computed from thesatellite images, comparing similar‐looking equally sizedareas in the precursor and target images up to a certaindistance from the grid point, by using a standard cross‐correlation technique applied to rectangular targets [Kidderand Vonder Haar, 1995; Belušić and Strelec Mahović,

2009]. The optimal matcher is the one for which thecross‐correlation coefficient between the target and thematcher is maximal, and the shift between them definesthe cloud motion vector. It should be pointed out that thevelocity determined in this way is associated with the cloudtop movement, and not with the movement at the instabilitylevel relevant for the propagation of the surface air pressuredisturbance. However, since the atmospheric motion vectorsdefine the velocities of the features in the atmosphere, theycan be regarded as a proxy for the velocity of the pertur-bation if caused by the considered atmospheric system.[12] Finally, a two‐dimensional ocean numerical model

was applied to the western shelf of the Black Sea. Themodel was forced by a moving air pressure disturbanceonly, with speed and direction as determined from thesounding and satellite measurements. The details of themodel are given in section 4.1.

3. Atmospheric Observations

3.1. Synoptic Conditions

[13] The synoptic situation over Europe on 7 May 200712:00 UTC (Figure 2) reveals flow characteristics similar tothese observed during most of the Adriatic and Balearic

Figure 1. Geographical map of the western Black Sea with the position of the barograph stations (cir-cles) and ocean observation sites (diamonds, maximum and minimum sea levels are given in bracketsafter Ranguelov et al. [2008]). Sea level values at Ahtopol tide gauge are taken from the chart record.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

3 of 12

meteotsunamis. Ramis and Jansà [1983] listed them for theBalearic Islands as (1) the Mediterranean air mass is presentover the surface, with weak depression to the west of theBalearic Islands, (2) warm and dry air blows along the850 hPa level, and (3) an unstable layer is present at the topof the warm air. Such a vertical stratification is favorable for awaveguide mechanism [Lindzen and Tung, 1976;Monserratand Thorpe, 1996]. Similar conditions are satisfied duringmost of the Adriatic meteotsunamis [Vilibić and Šepić, 2009],although, at least one of such events was provoked by awave‐CISK (Convective Instability of the Second Kind,Belušić et al. [2007]) generated air pressure disturbance.However, Belušić and Strelec Mahović [2009] found thatconvective systems were present during all of the wave‐ductdestructive Adriatic meteotsunamis, indicating that convec-tive system’s instability, largely present in the upper tropo-sphere and overtopping the warm African air, is the bestconductor for the trapped gravity disturbances traveling inthe lower troposphere.[14] The mean sea level pressure field of 7 May 2007 at

12:00 UTC (Figure 2a) indicates a surface low pressureregion over the western Black Sea and eastern Balkans,stretching from a deep Scandinavian low toward the EasternMediterranean. The low, situated between the Azores and

Siberian Highs, traveled from the Adriatic Sea (6 May 2007,12:00 UTC, not shown) toward the northeast at the leadingside of a deep upper‐level trough (Figure 2c), reaching thewestern Black Sea on 7 May 2007. Simultaneously, theinflow of the warm African air, as seen in the 850 hPa leveltemperature field (Figure 2b), was extending over theEastern Mediterranean and Turkey, with a thermal frontover the western Black Sea margin. At this level a south-westerly wind was blowing with speed of 10–15 m/s, butwithout a pronounced jet‐like structure. However, strongsouthwesterly jet‐like winds with speed of 15–25 m/s maybe seen at the 500 hPa level (Figure 2c), stretching from theCentral Sahara up to the Black Sea. The mid‐troposphereinstability layer (Figure 2d) is encompassing western BlackSea, denoting the regions where meteotsunamis are allowedto occur [Šepić et al., 2009b]. Although the center of theBlack Sea instability layer is positioned more to the north-east of the region attacked by the tsunami, one should beaware that the tsunami appeared around 06:00 UTC (i.e., 6 hbefore the time of the plotted reanalysis fields). Therefore,the instability area moved for about 400 km during these6 h, which implies that the center of the instability was justabove the affected area at 06:00 UTC on 7 May 2007.

Figure 2. Synoptic charts as obtained from the ECMWF reanalysis fields of 7 May 2007 12:00 UTC:(a) surface air pressure, (b) temperature and winds at 850 hPa level, (c) geopotential height and winds at500 hPa level, and (d) minimum Richardson number Ri at heights between 600 and 300 hPa.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

4 of 12

Figure 3. Vertical atmosphere structure (air temperature, solid line; dew point temperature, dashed line;wind speed; and Richardson number Ri) obtained from radio sounding on 7 May 2007 in Bucharest at(a) 00:00 UTC and (b) 12:00 UTC, and in Istanbul at (c) 00:00 UTC and (d) 12:00 UTC.

Figure 4. Ground air pressure series at a number of eastern Bulgarian barograph stations as digitizedfrom barograms of 7 May 2007. Vertical line marks the disturbance which presumably generatedmeteotsunami waves.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

5 of 12

3.2. Sounding Profiles and Ground Observation

[15] Direct measurements of the vertical atmosphericstructure are unfortunately not available for the area affectedby the tsunami. However, two closest sounding stations,Istanbul and Bucharest, reveal the structure already foundduring the Adriatic and Balearic meteotsunamis [Monserratand Thorpe, 1992; Vilibić and Šepić, 2009]. Low groundwinds and temperature inversion may be found from 400 to900 m at Bucharest on 7 May 2007 both at 00:00 and12:00 UTC (Figures 3a and 3b). Dryer air may be found up to3000 m, overtopped by a thick layer of moist and instableair (3000–6000 m). The instability layer was accompaniedby a SSE wind with the speed of 20–22 m/s. Temperatureinversion on heights of 200–700 m at Istanbul on 7 May2007 (Figure 3c) was very strong before the tsunami event,vanishing 12 h later (Figure 3d). Dry air with relativehumidity between 20% and 50% extended up to 5000 m,where humid and unstable air may be seen, being charac-terized by negative Ri. Weaker S to SSW winds in the lowertroposphere (<10 m/s) at 00:00 UTC increased in the fol-lowing hours, reaching 16 m/s at the minimum Ri layer

(∼5500 m). Vertical wind gradient was quasi‐constant fromthe region of zero wind recorded at the surface up to thealtitude of the instability layer, which is a known favorablecondition for the ducting of lower‐troposphere gravity dis-turbances [Monserrat and Thorpe, 1996].[16] Ground air pressure was measured at synoptic sta-

tions by analog barographs having weekly charts. Figure 4displays air pressure series measured at a number of baro-graph stations on 7 May 2007, possessing enough qualityrecords to be digitized with 10 min resolution. Such a res-olution is adequate to follow eventual disturbances over thearea but is not sufficient to perform any in‐depth analysis, asthe processes are occurring on a minute timescale. One cansee that only weak high‐frequency air pressure oscillationshave been recorded in the southern part of the Bulgariancoastline (Ahtopol, Burgas) as well as in the inland Bulgaria(Shumen). However, strong negative air pressure oscillationof about 2 hPa may be seen at Emine between 5:10 and5:50UTC. Similar oscillation occurred at Varna between 6:10and 6:50 UTC, and at Kaliakra between 6:20 and 7:00 UTCbut with amplitude of about 3 hPa. Also, noteworthy oscil-

Figure 5. Color‐enhanced Meteosat 9 IR10.8 mm satellite images, measuring cloud top temperature,taken on 7 May 2007 every 30 min from 04:30 UTC (up) till 7:00 UTC (bottom).

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

6 of 12

lations on a minute timescale are found at the Kaliakrabarogram (not shown), introducing a noise in the record(significantly widening the pen‐recorded line); they mayindicate the occurrence of infra‐gravity waves that may alsogenerate resonant long ocean waves, if they possess coher-ence over an ocean area. Nevertheless, the latter cannot beproved without precise microbarograph measurements, withsubminute sampling resolution [Monserrat et al., 1998]. Theoscillation can be traced also in Karnobat and Shabla, butwith lower amplitudes. Rough estimates of the air pressuredisturbance speed may be assessed if assuming the propa-gation direction toward NNE (30°): the disturbance traveledthe distance of 90 and 130 km between Emine and Kaliakraand between Karnobat and Kaliakra in about 100 and130 min, respectively, yielding to the disturbance speed ofabout 15–17 m/s.

3.3. Satellite‐Based Analyses

[17] Color‐enhanced satellite images in 10.8 mm infraredchannel (Figure 5) recorded during the tsunami event clearlydepict the approach of the frontal zone to the western BlackSea around 04:00 UTC on 7 May 2007, with a number ofembedded convective cells traveling and developing overthe region. A particular attention should be drawn to thesecond convective cell that approached the area affected bythe tsunami at 06:00 UTC (the cells are marked by 1 and 2 inFigure 5). A narrow cloudless corridor oriented perpendicu-larly to the propagating direction may be seen, indicatingstrong mesoscale activity and a gravity disturbance cappedby the second convective system eventually preceded by agust front, which altogether could cause the sudden pressurechange above the sea.[18] Cloud motion vectors (Figure 6) indicate the cloud

top velocity of the second convective system are close to18 m/s at the beginning of the tsunami event (06:00 UTC).The direction of the cloud tops over the affected area wastoward NE (30°–35°). The speed was rising a bit in the nexthour, but remaining below 25 m/s—lower than in the rearpart of the frontal zone, where the convective cells traveledmore to the north, with speeds surpassing 25 m/s. Oneshould be aware that the velocities of the cloud tops arelarger than the velocities at the cloud instability layer (ca.5000 m), as measured at both Bucharest and Istanbulsounding stations (Figure 3). However, the propagationdirection of the convective cells as estimated from satelliteimages follows the estimates from the ECMWF reanalysisfields, and therefore it may be taken as relevant during theassessment of the atmospheric gravity disturbance propa-gation over the area affected by the tsunami.

4. Ocean Numerical Modeling

4.1. The Model

[19] A two‐dimensional nonlinear shallow‐water modelhas been used to reproduce the ocean waves generated overthe western Black Sea shelf. The model is based on themomentum equations containing the air pressure forcingterm and the continuity equation:

@u

@tþ u

@u

@xþ v

@u

@y� fv ¼ �g

@�

@x� guðu2 þ v2Þ1=2

C2ðhþ �Þ � 1

�

@P

@x; ð2Þ

@v

@tþ u

@v

@xþ v

@v

@yþ fu ¼ �g

@�

@y� gvðu2 þ v2Þ1=2

C2ðhþ �Þ � 1

�

@P

@y; ð3Þ

@�

@tþ @

@x½ðhþ �Þu� þ @

@y½ðhþ �Þv� ¼ 0; ð4Þ

where t is time, u and v are the vertically averaged velocitycomponents in the x and y directions, g is the acceleration ofgravity, z is the sea level elevation, h is the undisturbedwater depth, f is the Coriolis parameter, r is the waterdensity, P is the air pressure, and C is the Chezy’s frictioncoefficient:

C ¼ 18 log ð0:37h=z0Þ ½m1=2=s�; ð5Þ

Figure 6. Cloud motion vectors computed by using 15 minsubsequent satellite images and plotted every hour on 7 May2007 from 05:00 UTC (up) till 07:00 UTC (bottom).

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

7 of 12

where z0 is the roughness scale. An explicit leapfrog scheme[Imamura, 1996] has been used to solve equations (2)–(4).[20] The model domain (Figure 7) encompassed the

western Black Sea shelf; the corresponding grid was takenfrom the General Bathymetric Chart of the Oceans (GEBCO)bathymetry database (available from http://www.gebco.net).This grid has 30″ resolution (i.e., 680 m × 925 m (for lon-gitude and latitude). The size of the computational domainwas 259 × 299 grid cells. According to a numerical stabilitycriterion, the time step was taken to be Dt = 1 s and thebottom roughness scale z0 was set to 0.001 m. A radiationcondition was used at all open boundaries.[21] The model was forced by a moving air‐pressure

disturbance that was introduced into the model in the fol-lowing way: (1) artificial traveling disturbance has beenconstructed, with linear increase of air pressure of 3 hPaover 6 min, and constant air pressure values before and afterthe disturbance, (2) speed and direction of the moving airpressure disturbance were set to be constant over the wholedomain with the values assessed from the measurements(i.e., 16 m/s and 30°), respectively, (3) the time intervalelapsed from the passage of the air pressure front above eachgrid point was calculated, and (4) the air pressure data wereinterpolated linearly in time at each grid point shifted in timeaccording to the determined time interval.

4.2. Model Results

[22] The response of the ocean to the moving air pressuredisturbance is noteworthy (Figure 8), with maximum sea

levels of about 1 m occurring west of Kaliakra up to Balchikand in Burgas Bay. The latter maximum is not realistic ashigh‐frequency air pressure oscillations were weak in thisregion, and therefore not introduced properly into the ideal-ized forcing. However, the first region of the modeled sealevel maxima is coinciding with the observations (Figure 1),except for the coastline northeast of Kaliakra, where note-worthy sea level oscillations have been observed but notmodeled in Shabla. The model also reproduced well someother observed properties of the tsunami waves, such as seawithdrawal rate being larger than sea level rise rate (Figure 9).Minimum sea level values have been modeled in Kavarna(−1.7 m), Dalboka (−1.3 m), and Balchik (−1.2 m), exactlyat places where minimum sea levels have been observed[Ranguelov et al., 2008]. Model is slightly underestimatingboth maximum and minimum values in that area, but this isexpected as the coastal bathymetry is too coarse for coastaltsunami modeling [Poisson et al., 2009] and the model isforced by idealized nondispersive disturbance propagatingwith constant speed and direction, which is far from the realsituation [Monserrat and Thorpe, 1992; Šepić et al., 2009a].Moreover, the model may underestimate the sea leveloscillations due to lack of infragravity air pressure forcing(i.e., on timescale of a few minutes and spatial scales of akilometer) that may generate resonant sea level oscillationson a periods of several minutes. Indeed these oscillationshave been reported (oscillations of 4–8 min, [Ranguelov etal., 2008]), but not reproduced well by the model in thisstudy.

Figure 7. Bathymetry and domain of the ocean numerical model. Isobaths of 50 and 20 m are markedwith dashed and dotted line, respectively.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

8 of 12

[23] Nevertheless, these “imperfections” did not preventthe model to reproduce qualitatively the observed oscilla-tions, indicating that the atmospheric traveling disturbancewas capable of exciting strong tsunami waves along thewestern Black Sea shelf. The last observed characteristics of

the tsunami event—turbulence, strong water currents, mudwaters, and foam in some sites (e.g., in Balchik andKavarna)—were reproduced as well, as the maximum cur-rents reached 100 cm/s in front of these places (Figure 8b).These currents were the strongest over the shallowestregion, which is extending southward of Balchik for about20–30 km (depths lower than 15 m, with a number of shoals).

5. Discussion and Conclusions

[24] The sudden occurrence of tsunami waves on 7 May2007 along the northern Bulgarian Black Sea coast withmaximal observed sea level range exceeding 3 m at someplaces, and a lack of tsunamigenic earthquake in the area,directed the research to the assessment of other possiblesources of this event. Ranguelov et al. [2008] evaluated thepossibility that a submarine landslide could have been thesource and found it plausible in terms of availableobservations. However, the observations are the most criticalpart of the analysis, as no observations or eyewitness reportswere documented in the area where maximum tsunamiwaves have been modeled (i.e., around Emine). Therefore,we evaluated the possibility that an atmospheric process isthe source of the observed tsunami waves (i.e., that this eventmay be classified as a meteotsunami). Meteotsunami wavesare normally generated by a high‐frequency air pressuredisturbance propagating over a shallow area (though travel-ing wind disturbance may be generator as well [de Jong andBattjes, 2004]), where the speed of the atmospheric distur-bance is equal to the speed of the long ocean waves, acti-vating a resonant amplification process of the ocean waves[Monserrat et al., 2006]. The atmospheric disturbanceshould be nondissipative in order to allow for significantgrowth of the ocean waves, which can be found only underspecific synoptic and mesoscale atmospheric conditions.[25] Analyzed synoptic conditions during the tsunami

event are found to be fully favorable for the generation andducting of an eventual traveling atmospheric gravity dis-turbance. The disturbance has been found as well, reachingmaximum amplitude of 3 hPa and traveling toward NNEwith the speed of 15–17 m/s (a rough estimate from thebarograph data). Moreover, the disturbance has not beenfound along the southern Bulgarian coastline as well as inthe inland area a few tens of kilometers to the west of theaffected area. That is an indication that the core of the dis-turbance was pretty narrow, presumably not wider than afew tens of kilometers, similarly as documented for the 20071st meteotsunami which occurred in the Adriatic [Šepić etal., 2009a]. A convective system passed over the regionwith velocities at cloud tops level of about 18–20 m/s,exactly at the same times when the ground disturbance hasbeen observed. The sounding data indicated that the speedand direction of the disturbance were about 16 m/s and 30°,respectively, and according to the ducting theory this is foundto be propitious for meteotsunami generation [Monserratand Thorpe, 1992, 1996; Vilibić and Šepić, 2009]. Theresearch was thus directed to the reproduction of the observedocean waves—the first study of such kind in the BlackSea—whereas similar studies were already conducted forthe Adriatic [Vilibić et al., 2004] and the Balearic [Liu et al.,2003; Vilibić et al., 2008] meteotsunamis.

Figure 8. Results of the ocean model simulation: (a) maxi-mal sea level heights and (b) maximal ocean currents.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

9 of 12

[26] Unfortunately, the quality of the high‐frequency airpressure data was unsatisfactory for the construction of arealistic forcing based on the air pressure series; therefore,we constructed a sharp pressure jump (3 hPa over 6 min,following the observations at some barographs) embedded inthe constant air pressure series and traveling over the regionwith constant speed and direction. The model reproducedmost of the observations, which include (1) minimum sealevels and sea withdrawal larger than maximum sea levelsand sea level rise, (2) strong currents reproduced at placeswhere observed, and (3) maximum sea level rise and low-ering similar to the observations at most of affected places.On the other hand, the model overestimated the oscillations inthe Burgas Bay and along the southern Bulgarian coast,where only a weak air pressure disturbance was observed (seebarograph records at Ahtopol and Burgas in Figure 4)—notfollowed by the model which was forced by a nonchangingsharp pressure jump across the whole computational domain.Therefore, we believe that the maximum sea level oscilla-tions modeled in the Burgas Bay are unrealistic, as generatedby an unrealistic traveling pressure disturbance. Suppor-tively, the modeled oscillations at Ahtopol overestimate theobservations by about three times; applying a similar ratiobetween modeled and observed values at the Burgas Baydramatically lowers their visibility to the eyewitnesses.Another interesting point is that the first tsunami waverecorded at Ahtopol was not the largest as observed in thenorthern part—the maximum wave occurred about 6 h afterthe major disturbance and may be related to the topographicor edge waves that come from the north or to another

weaker atmospheric disturbance that occurred at that timejust over southern part (see barographic record in Figure 4).Conclusively, our presumptions about the source char-acteristics, spatial coverage, and outreach of the atmosphericdisturbance should be additionally verified by reproducingthe event by means of a mesoscale atmospheric model.[27] Also, ocean waves were underestimated to the north-

east of Kaliakra, which may be due to the changes in energycontent, speed, and direction of the atmospheric disturbance,not embedded in the model forcing. Another serious deficitof the model forcing (i.e., of its temporal and spatial prop-erties which do not include any infragravity waves usuallyoccurring with periods up to a few minutes) is its incapacityto reproduce properly the eyewitnessed period of oscilla-tions which was reported to be 4–8 min [Ranguelov et al.,2008]. However, state‐of‐the‐art mesoscale atmosphericmodels are still not successful in reproducing properly theseprocesses [Belušić et al., 2007; Šepić et al., 2009a], andpresently no realistic forcing can be applied in such modelingstudies. The resolution of the bathymetry may also be blamedfor this model shortage; efficient bathymetry resolution forreproduction of coastal tsunami and meteotsunami dynamicsshould be an order of magnitude lower than in our compu-tations, even down to 10 m [Vilibić et al., 2008].[28] Summarily, we may say that the atmospheric pro-

cesses had the potential to generate the tsunami waves thatwere observed along the western Black Sea on 7 May 2007,and that the event may be classified as a meteotsunami. Thisis a new fact, not mentioned and never modeled before forthis region, which increases the probability of tsunami

Figure 9. Time series of the modeled sea levels at some locations along the western Black Sea coast.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

10 of 12

expectations and influences the tsunami risk evaluation tothe population and infrastructure. Another novelty in thepaper may be found in the analyses of meteorological satelliteimages, where a method of computation of cloud motionvectors was applied to the event, and resulted in estimates ofdisturbance velocities that may possibly be used in anoperative system in the future. This study may be a clue forassessing other less destructive tsunami‐like events whichare still not classified regarding their source, as lots ofcataloged tsunami events are still marked as unknown, and someof them have only recently been recognized as meteotsu-namis (e.g., the middle Adriatic meteotsunami whichoccurred on 21 June 1978 with 6 m waves observed in theVela Luka Bay was classified as a tsunami of “unknown”type in several tsunami catalogs [Tinti et al., 2004; Vučetićet al., 2009]). Nevertheless, submarine landslide is still anoption for the generation of tsunami waves in the Black Sea[Ranguelov et al., 2008].[29] A further element of discussion is the question whether

(1) possible Black Sea meteotsunamis are connected withchanges in the atmospheric circulation above Europe due torecent climate changes [Giorgi and Coppola, 2007], allow-ing for meteotsunamis to appear nowadays in this region, or(2) the Black Sea meteotsunamis are more infrequent thanthe Balearic and Adriatic meteotsunamis, possessing largerreturn period and therefore having occurred in the past, butnot being recorded by observations or historical chronicles.The understanding of the meteorological conditions and ofrarity of these events is a very intriguing question, whichmay be answered through long‐term examination of thereanalyzed atmospheric fields. However, such a study shouldbe accompanied with high‐resolution atmospheric modelscapable of reproducing very specific conditions in theatmosphere. In any case, meteotsunamis should be includedin any Black Sea tsunami assessment study, as being possiblefor the region.

[30] Acknowledgments. Sounding data were taken from theUniversityof Wyoming Web site (http://weather.uwyo.edu/upperair/sounding.html),while the synoptic fields were analyzed using ECMWF reanalysis fields.The comments raised by two anonymous reviewers are appreciated. Thework was supported by the Ministry of Science, Education and Sports ofthe Republic of Croatia (Grant 001‐0013077‐1122). The method for calcu-lating Cloud Motion Vectors was developed within the framework of theproject “CEI Nowcasting System” sponsored by the Austrian Ministry forEducation, Science and Culture.

ReferencesAltinok, Y., and S. Ersoy (2000), Tsunamis observed on and near the Turkishcoast, Nat. Hazards, 21, 185–205, doi:10.1023/A:1008155117243.

Belušić, D., and N. Strelec Mahović (2009), Detecting and following atmo-spheric disturbances with a potential to generate meteotsunamis in theAdriatic, Phys. Chem. Earth, 34, 918–927, doi:10.1016/j.pce.2009.08.009.

Belušić, D., B. Grisogono, and Z. B. Klaić (2007), Atmospheric origin ofthe devastating coupled air‐sea event in the east Adriatic, J. Geophys.Res., 112, D17111, doi:10.1029/2006JD008204.

Candela, J., S. Mazzola, C. Sammari, R. Limeburner, C. J. Lozano, B. Patti,and A. Bonnano (1999), The “Mad Sea” phenomenon in the Strait ofSicily, J. Phys. Oceanogr., 29, 2210–2231, doi:10.1175/1520-0485(1999)029<2210:TMSPIT>2.0.CO;2.

Churchill, D. D., S. H. Houston, and N. A. Bond (1995), The DaytonaBeach wave of 3–4 July 1992: A shallow water gravity wave forcedby a propagating squall line, Bull. Am. Meteorol. Soc., 76, 21–32,doi:10.1175/1520-0477(1995)076<0021:TDBWOJ>2.0.CO;2.

De Jong, M. P. C., and J. A. Battjes (2004), Low‐frequency sea waves gen-erated by atmospheric convection cells, J. Geophys. Res., 109, C01011,doi:10.1029/2003JC001931.

Donn, W. L., and M. Ewing (1956), Stokes’ edge waves in Lake Michigan,Science, 124, 1238–1242, doi:10.1126/science.124.3234.1238.

Douglas, C. K. M. (1929), The line‐squall and Channel wave of July 20th,1929, Meteorol. Mag., 64, 187–189.

Dragani, W. C., C. A. Mazio, and M. N. Nuñez (2002), Sea level oscilla-tions in coastal waters of the Buenos Aires province, Argentina, Cont.Shelf Res., 22, 779–790, doi:10.1016/S0278-4343(01)00096-6.

Drago, A. (2008), Numerical modelling of coastal seiches in Malta, Phys.Chem. Earth, 33, 260–275.

Durran, D. R., and J. B. Klemp (1982), On the effects of moisture onBrunt‐Väisälä frequency, J. Atmos. Sci., 39, 2152–2158, doi:10.1175/1520-0469(1982)039<2152:OTEOMO>2.0.CO;2.

Ewing, M., F. Press, and W. J. Donn (1954), An explanation of the LakeMichigan wave of 26 June 1954, Science, 120, 684–686, doi:10.1126/science.120.3122.684.

Giorgi, F., and E. Coppola (2007), European climate‐change oscillation(ECO), Geophys. Res. Lett., 34, L21703, doi:10.1029/2007GL031223.

Goring, D. G. (2005), Rissaga (long‐wave events) on New Zealand’s easternseaboard: A hazard for navigation, in Coasts and Ports: Coastal Living—Living Coast; Australian Conference 2005, edited byM. R. Townsend andD. Walker, pp. 137–141, Institution of Engineers, Barton, Australia.

Haslett, S. K., and E. A. Bryant (2009), Meteorological tsunamis in SouthernBritain: An historical review, Geogr. Rev., 99(2), 146–163.

Hibiya, T., and K. Kajiura (1982), Origin of the Abiki phenomenon (a kindof seiche) in Nagasaki Bay, J. Oceanogr. Soc. Jpn., 38, 172–182,doi:10.1007/BF02110288.

Imamura, F. (1996), Review of tsunami simulation with finite differencemethod, in Long‐Wave Runup Models, edited by H. Yeh, P. Liu, andC. Synolakis, pp. 25–42, World Sci., Singapore.

Kidder, S. Q., and T. H. Vonder Haar (1995), Satellite Meteorology: AnIntroduction, 466 pp., Academic, San Diego.

Lindzen, R. S., and K.‐K. Tung (1976), Banded convective activity andducted gravity waves, Mon. Weather Rev. , 104 , 1602–1617,doi:10.1175/1520-0493(1976)104<1602:BCAADG>2.0.CO;2.

Liu, P. L.‐F., S. Monserrat, M. Marcos, and A. B. Rabinovich (2003),Coupling between two inlets: Observation and modeling, J. Geophys.Res., 108(C3), 3069, doi:10.1029/2002JC001478.

Mercer, D., J. Sheng, R. J. Greatbatch, and J. Bobanović (2002), Barotropicwaves generated by storms moving rapidly over shallow water, J. Geo-phys. Res., 107(C10), 3152, doi:10.1029/2001JC001140.

Metzner, M., M. Gade, I. Hennings, and A. B. Rabinovich (2000), Theobservation of seiches in the Baltic Sea using a multi data set of waterlevels, J. Mar. Syst., 24, 67–84, doi:10.1016/S0924-7963(99)00079-2.

Monserrat, S., and A. J. Thorpe (1992), Gravity‐wave observations usingan array of microbarographs in the Balearic Islands, Q. J. R. Meteorol.Soc., 118, 259–282, doi:10.1002/qj.49711850405.

Monserrat, S., and A. J. Thorpe (1996), Use of ducting theory in anobserved case of gravity waves, J. Atmos. Sci., 53, 1724–1736,doi:10.1175/1520-0469(1996)053<1724:UODTIA>2.0.CO;2.

Monserrat, S., A. B. Rabinovich, and B. Casas (1998), On the reconstructionof the transfer function for atmospherically generated seiches, Geophys.Res. Lett., 25, 2197–2200, doi:10.1029/98GL01506.

Monserrat, S., I. Vilibić, and A. B. Rabinovich (2006), Meteotsunamis:Atmospherically induced destructive ocean waves in the tsunami fre-quency band, Nat. Hazards Earth Syst. Sci., 6, 1035–1051, doi:10.5194/nhess-6-1035-2006.

Nikonov, A. A. (1997), Tsunami occurrence on the coasts of the Black Seaand the Sea of Azov, Izvestiya–Phys. Solid Earth, 33, 72–87.

Papadopoulos, G. A. (1993), On some exceptional seismic (?) sea‐waves inthe Greek archipelago, Sci. Tsunami Hazards, 11, 25–34.

Pelinovsky, E. (1999), Preliminary estimates of tsunami danger for thenorthern part of the Black Sea, Phys. Chem. Earth Part A, 24, 175–178, doi:10.1016/S1464-1895(99)00015-0.

Poisson, B., M. Garcin, and R. Pedreros (2009), The 2004 December26 Indian Ocean tsunami impact on Sri Lanka: Cascade modelling fromocean to city scales, Geophys. J. Int., 177, 1080–1090, doi:10.1111/j.1365-246X.2009.04106.x.

Proudman, J. (1929), The effects on the sea of changes in atmospheric pres-sure, Geophysical Supplement to the Monthly Notices of the Royal Astro-nomical Society, 2(4), 197–209.

Ramis, C., and A. Jansà (1983), Condiciones meteorológicas simultáneas ala aparición de oscilaciones del nivel del mar de amplitud extraordinariaen el Mediterráneo occidental (in Spanish), Rev. Geofis., 39, 35–42.

Ranguelov, B. (1996), Seismicity and tsunamis in the Black Sea, inSeismology in Europe, edited by R. Stefanson, pp. 667–673, EuropeanSeismological Commission, Reykjavik.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

11 of 12

Ranguelov, B. (2003), Possible tsunami deposits discovered on the Bulgar-ian Black Sea coast and some implications, in Submarine Landslides andTsunamis, NATO Sci. Ser., Ser. 4, vol. 21, edited by A. Yalçiner, et al.,pp. 237–242, Kluwer Acad., Dordrecht, Netherlands.

Ranguelov, B., and D. Gospodinov (1995), Tsunami vulnerability model-ling for the Bulgarian Black Sea coast, Water Sci. Technol., 32, 47–53,doi:10.1016/0273-1223(96)00046-7.

Ranguelov,B., S. Tinti, G. Pagnoni, R. Tonini, F. Zaniboni, andA.Armigliato(2008), The nonseismic tsunami observed in the Bulgarian Black Sea on7 May 2007: Was it due to a submarine landslide?, Geophys. Res. Lett.,35, L18613, doi:10.1029/2008GL034905.

Šepić, J., I. Vilibić, and D. Belušić (2009a), The source of the 2007 1stmeteotsunami (Adriatic Sea), J. Geophys. Res., 114 , C03016,doi:10.1029/2008JC005092.

Šepić, J., I. Vilibić, and S. Monserrat (2009b), Teleconnections between theAdriatic and the Balearic meteotsunamis, Phys. Chem. Earth, 34, 928–937, doi:10.1016/j.pce.2009.08.007.

Solov’eva, O. N., and I. P. Kuzin (2005), Seismicity and tsunamis in thenortheastern part of the Black Sea, Oceanology, Engl. Transl., 45,781–794.

Tinti, S., A. Maramai, and L. Graziani (2004), The new catalogue ofthe Italian tsunamis, Nat. Hazards, 33, 439–465, doi:10.1023/B:NHAZ.0000048469.51059.65.

Tintoré, J., D. Gomis, S. Alonso, and D. P. Wang (1988), A theoreticalstudy of large sea level oscillations in the Western Mediterranean,J. Geophys. Res., 93(C9), 10,797–10,803, doi:10.1029/JC093iC09p10797.

Vilibić, I., and J. Šepić (2009), Destructive meteotsunamis along the east-ern Adriatic coast: Overview, Phys. Chem. Earth, 34, 904–917.

Vilibić, I., N. Domijan, M. Orlić, N. Leder, and M. Pasarić (2004), Resonantcoupling of a traveling air pressure disturbance with the east Adriaticcoastal waters, J. Geophys. Res. , 109 , C10001, doi:10.1029/2004JC002279.

Vilibić, I., S. Monserrat, A. B. Rabinovich, and H. Mihanović (2008),Numerical modelling of the destructive meteotsunami of 15 June 2006on the coast of the Balearic Islands, Pure Appl. Geophys., 165, 2169–2195, doi:10.1007/s00024-008-0426-5.

Vučetić, T., I. Vilibić, S. Tinti, and A. Maramai (2009), The great Adriaticflood of 21 June 1978 thirty years later: An overview of the existingreports, Phys. Chem. Earth, 34, 894–903.

Wang, X., K. Li, Z. Yu, and J. Wu (1987), Statistical characteristics ofseiches in Longkou Harbour, J. Phys. Oceanogr., 17, 1963–1966,doi:10.1175/1520-0485(1987)017<1063:SCOSIL>2.0.CO;2.

Yalçiner, A., E. Pelinovsky, T. Talipova, A. Kurkin, A. Kozelkov, andA. Zaitsev (2004), Tsunamis in the Black Sea: Comparison of the histor-ical, instrumental, and numerical data, J. Geophys. Res., 109, C12023,doi:10.1029/2003JC002113.

N. S. Mahović, Meteorological and Hydrological Service, Grič 3, 10000Zagreb, Croatia.B. Ranguelov, Geophysical Institute, Bulgarian Academy of Sciences,

Acad. G. Boncev St., bld. 3 Sofia 1113, Bulgaria.J. Šepić and I. Vilibić, Institute of Oceanography and Fisheries, Setaliste

I. Mestrovica 63, Split, 21000 Croatia. (vilibic@izor.hr)S. Tinti, Department of Physics, University of Bologna, Viale Carlo Berti

Pichat 8, I‐40127 Bologna, Italy.

VILIBIĆ ET AL.: ORIGIN OF THE BLACK SEA TSUNAMI C07006C07006

12 of 12