The Turkish Tsunami of 6 February 2023 in the Northeastern Mediterranean

A powerful earthquake struck southeastern Turkey on 6 February 2023 with magnitude Mw = 7.8, followed by a second earthquake Mw = 7.5 just nine hours later. In addition, a storm surge caused by the passage of a cyclone over the area coincided with the first earthquake. Field observations showed that the fault planes did not extend to the Mediterranean Sea; however, a weak tsunami was observed on several tide gauges in the northeastern Mediterranean. Historically, there were tsunamis in the Levantine Sea region caused by earthquakes on land within the East Anatolian and the Dead Sea Fault zones, having a complex cause, i.e., earthquakes and associated landslides. Hence, we studied the origin of the generated waves. We separated and estimated various components of sea level oscillations from 1 to 8 February 2023, using spectral and f–t analyses and sets of narrow-band and high-frequency filters. The earthquake with Mw = 7.8 on 6 February 2023 caused a weak tsunami with maximum amplitudes of 15–17 cm in Erdemli and Arsuz. We identified that the increase in sea level oscillations in Arsuz after 7 February was associated with the impact of the cyclone. The maximum amplitude of sea level oscillations, 22 cm, in Arsuz was observed on 7 February and is associated with infragravity waves caused by the influence of storm surge during the passage of the cyclone. In Erdemli, on the contrary, the influence of high-frequency waves of atmospheric origin is significantly lower, i.e., the amplitude of the sea level oscillations after 7 February was less than 10 cm.


Introduction
On 6 February 2023 at 01:17:35 UTC, the catastrophic Pazarcık earthquake of the Kahramanmaraş earthquake sequence, with a magnitude M w = 7.8 occurred in southeastern Turkey and northern Syria, 26 km to the east of Nurdagı (37.174°N, 37.032°E). The hypocenter depth was *10 km according to the United States Geological Survey (USGS). About nine hours later (10:24:49 UTC), the second strong earthquake (Elbistan earthquake) with a magnitude M w = 7.5 occurred in the same region, in the province of Gaziantep (USGS), 4 km SSE of Ekinözü (38. 024°N, 37.203°E). The number of aftershocks caused by these two earthquakes reached *6 040 as of 20 February 2023. By 10 April 2023 there were at least 50 399 casualties, *107 204 injured, and about 2 million people homeless; at least 105 800 buildings were destroyed or severely damaged, *229 550 buildings were damaged slightly or moderately, and 150 000 commercial infrastructures moderately damaged in 10 provinces across Turkey (IBC, 2023). In northwestern Syria, there were at least 8 476 casualties, *14 500 injured, and *5.37 million people homeless; at least 1000 buildings were destroyed and thousands damaged (IBC, 2023; Earthquake Response in Türkiye and Whole of Syria, 2023). In Lebanon, the earthquake damaged 16 200 buildings across the country.
During the last 36 centuries there have been 67 earthquakes with M [ 7, 133 earthquakes with 7 [ M [ 6, and at least 96 documented tsunamis in the Eastern Mediterranean (Yalciner et al., 2007). Onshore and offshore earthquakes occurring in the northeastern and eastern sectors of the Levantine Sea region have generated tsunamis in the past. Most of the events in the Levantine Basin are related to the boundary of the African, Arabian, and Anatolian tectonic plates. Six earthquakes: *1365 BC, 115,1404,1114,1752, and 1872 AD, are known tsunamigenic events within the East-Anatolian Fault zone (Altinok et al., 2011). According to (Heidarzadeh et al., 2017(Heidarzadeh et al., , 2021 since 2017 the Eastern Mediterranean Sea has experienced four tsunamigenic earthquakes that caused moderate damage to the coastal communities. A maximum tsunami run-up height of 3.8 m was observed in the central Aegan coast of Turkey for the 30 October 2020 Samos M w 7.0 earthquake. For the 20 July 2017 Bodrum-Kos M w 6.6 earthquake, a maximum run-up of 1.9 m was measured.
The main goal of the presented study was to analyze all available tide gauge records on the Levantine Sea, to evaluate the statistical and spectral properties of observed sea level oscillations, and to identify the tsunami. We revealed tsunami signals on the records of six tide gauge stations and separated the meteorological impact from these records.

Seismotectonics of the Region
The earthquake of 6 February occurred along the East Anatolian Fault zone, which has a length of *450-700 km (Ç etin et al., 2023;Güvercin et al., 2022), and adjacent faults at the contact of the Arabian and Anatolian tectonic plates near their triple junction with the African plate ( Fig. 1). This fault generated large, destructive earthquakes with M [ 6.5 in the past. The right-lateral North Anatolian Fault and the left-lateral East Anatolian Fault are the main structures of regional importance in the recent tectonics of the Eastern Mediterranean. The Anatolian microplate is located between them and moves westward under the influence of internal deformation caused by the collision of the Arabian and African plates with the Eurasian plate (Emre et al., 2018). The southwestern end of the East Anatolian Fault zone transfers into the Cyprus Island Arc and the meridional Levant transform zone (Dead Sea Fault zone). In the northeastern part, the zone intersects with the North Anatolian Fault zone and is limited by the Bitlis-Zagros Fold and Thrust Belt. Late Quaternary active faults are mostly of a north-easterly strike and left-lateral strike-slip, occasionally with thrust and normal components. Slip values vary from 1 to 4 mm/year for western segments to 10 mm/year for segments located at the eastern end of the fault zone (Güvercin et al., 2022).
The field observations indicate that the Kahramanmaraş earthquakes sequence is the result of rupturing of the fault zones with a length of the main M w = 7.8 earthquake *300 km with major surface displacements on the order of 3-7 m (Ç etin et al., 2023). It corresponds to the Erkenek, Pazarcık, and Amanos segments of the East Anatolian Fault zone (Emre et al., 2018;Hancılar et al., 2023). For the Elbistan earthquake (M w 7.5), the total rupture length was around 160 km with major surface displacements on the order of 2-8 m. According to the USGS finite fault solution, as of 05 April 2023, the maximum slip value for the Pazarcık earthquake at the first segment, where the rupturing started, was 1.9 m. In the second segment, its maximum was 11.2 m, and for the third segment 8.2 m. The focal mechanism solutions of both earthquakes indicate a left-lateral strike-slip (USGS, 2023). The finite fault solution of USGS on 20 April 2023 consists strike 25°(USGS, 2023). The maximum slip value in the north more extended segment was 3.36 m, but this part of the slip distribution was located entirely on land. In the second segment, its maximum was 1.56 m. The epicenter of the second earthquake with magnitude M w = 7.5 was at a greater distance from the coast. Table 1 shows the main parameters and fault-plane solutions of these two earthquakes.
As a result of the earthquake with M w = 7.8, the maximum recorded peak ground accelerations (PGA) were 0.8 g for the horizontal component and 0.85 g for the vertical component. The PGA values computed using the relationships from updated NGA model (Chiou et al., 2014) are in the range from 0.013 g to 1.3 g for various epicentral distances. The estimated shaking intensity according to the scale (Bilal et al., 2014) ranges from 4.2 to 11.6 (Hancılar et al., 2023). The scale corresponds to empirical relationships between modified Mercalli intensity (MMI) and recorded peak ground-motion parameters developed for Turkey.
The earthquake sources located within the Eastern Mediterranean region have historically generated tsunamis of various intensities. In the tsunami catalog (Altinok et al., 2011) since *1365 BC, 30 tsunamigenic events have been recorded with a reliability of C 2 (questionable) for the northern, northwestern, and eastern sectors of the Levantine Sea (Altinok et al., 2011). Among them 18 of the tsunami events are located along the coast of the Levant: eight events are generated by onshore earthquakes (525 and 590 BC, 859, 1170, 1404, 1408, 1822, and 1872, eight events are related to offshore earthquakes (*1365(* and 140 BC, 115, 551, 1114(* , 1752 and two events in 1759 AD), and the other two events could have been generated by earthquakes with epicenter situated both on land and at sea (991, 1202 AD). For some events, the contribution of seismogenic landslides to tsunami generation is possible (115, 859, Figure 1 Map of the Anatolian region showing regional tectonic settings, the epicenters, and focal mechanisms of the catastrophic earthquakes that occurred on 6 February 2023 at 01:17:35 UTC (M w = 7.8) and at 10:24:49 UTC (M w = 7.5) (USGS), major fault zones (thick red lines) and minor active faults distribution (thin gray lines) (Bachmanov et al., 2017;Emre et al., 2018;Kasapoglu & Toksöz, 1983), regional earthquakes epicenters with M w [ 3 (pale yellow circles-for the 5 years before 6 February 2023, and bright yellow circles-for the period 6-10 February 2023) (USGS). Brown arrows indicate the general directions of movement of tectonic plates while red arrows-the direction of movement along the major fault zones (Emre et al., 2018;Kasapoglu & Toksöz, 1983). Blue circles indicate tsunamigenic earthquakes of the past according to (Altinok et al., 2011;Yolsal et al., 2007)  991, 1202, and 1759 AD). For three events, the contribution of seismogenic landslides to tsunami generation was established (551, 1404, and 1408 AD). Eight tsunami events are known for the Cyprus Island Arc zone, and all of them are associated with offshore earthquake sources. Four events are known for the southern coast of Minor Asia, three of which were generated by underwater earthquakes (68, 1403, 1743 AD). The location of the tsunamigenic earthquake epicenter is not well determined for the event 1036/1037 AD.

Analysis of Meteorological Processes
The earthquake coincided with a cyclone over the Anatolian Peninsula. We used the hourly mean sea level pressure (MSLP) and 10-m winds during the period 1-7 February 2023 for the analysis of the meteorological processes. These data were obtained from the Antalya and İskenderun 1 weather stations from RP5 portal (https://rp5.ru/) and ERA5 (fifthgeneration records of the ECMWF atmospheric reanalyzes of the global climate; Hersbach et al., 2020). The gyre was generated on 3 February from the Icelandic Low and the center of this atmospheric structure moved from west to east along the southern Turkish shoreline. The minimal pressure value in the cyclone was 991 hPa and wind speed (hereafter 10-m winds will be described) increased to 70 kmÁh -1 (19.4 mÁs -1 ) on its periphery. After separation, the gyre travelled over the northern part of the Northern Sea and crossed Central Europe in a southeastern direction from the Danish Straits to the Sea of Marmara (4 February, 21:00 UTC). Above this region, pressure rose to 1012 hPa and wind speed fell to 40 kmÁh -1 (11 mÁs -1 ). Three hours later, on 5 February, 00:00 UTC, atmospheric pressure over the Aegean Sea started to drop from 1012 hPa, to by 21:00 UTC it was around 997 hPa. The cyclone's center shifted to the east (Fig. 2a). Wind speed went up to 72 kmÁh -1 (20.0 mÁs -1 ) on the periphery over the central part of the Aegean Sea. In the following days such high values have not been observed.
Pressure dropped by another 1 hPa in the cyclone center a few minutes before the earthquake (6 February, 01:00 UTC, Fig. 2b). Wind speed and direction to the southeast of Cyprus played an important role in sea level oscillations at some ports. Here maximum speed value was 59 kmÁh -1 (16.4 mÁs -1 ), and the wind had a northeastern general direction, toward the Gulf of Arsuz. The wind was not so strong at the entry of the bay due to its geographic location, some features of coastal topography, and wind direction near the shore. Wind speed did not exceed 28 kmÁh -1 (7.8 mÁs -1 ). Pressure increased to 998 hPa in the gyre's center for the next 5 h (6 February, 06:00 UTC, Fig. 2c, e). Wind speed fell to 55 kmÁh -1 (15.4 mÁs -1 ), but in a similar direction to the southeast of Cyprus. At the same time, the wind became stronger on the sea boundary of the Gulf of İskenderun (36 kmÁh -1 , or 10.0 mÁs -1 ). Generally, wind speed gradually lowered over the eastern part of the Mediterranean Sea. The center of the cyclone was over the Gulf of İskenderun six hours later (6 February, 12:00 UTC, Fig. 2d, f). The pressure did not increase (998 hPa); however, wind speed rose again to the southeast of Cyprus (58 kmÁh -1 , or 16.2 mÁs -1 ). It might have caused a northward moving cyclone and the southern periphery was also transported. The wind regime remained the same. When the gyre was moving eastward, its area increased. During the following 12 h a low-pressure structure propagated to the eastern part of the Black Sea. As a result, a second core was generated over that region. Pressure evenly rose to 1002 hPa over the Gulf of İskenderun. Values increased to 1010 hPa on 7 February, 12:00 UTC. The cyclone continued its propagation to the east.

Analysis of Sea Level Observations
We used sea level records from six tide gauges on the Mediterranean Sea coast of southern Turkey for the detailed time series analysis. The stations are located westward from the earthquake epicenter: Arsuz, Erdemli, Taşucu, Bozyazı, Antalya, and Marmaris as shown in Fig. 3, with the wave parameters listed in Table 2. We obtained the 30-s sea level records from the Intergovernmental Oceanographic Commission (IOC, Flanders Marine Institute, 2023, https://www.ioc-sealevelmonitoring. org/) from 1 January to 7 February 2023. The observations are homogeneous, i.e., all data series have a time step Dt = 30 s, a vertical resolution is 1 mm, and time is the Greenwich Meridian time (UTC). We checked the series thoroughly and eliminated all spikes and shifts. We also filled in short gaps with linear interpolation.
Sea level oscillations (f tot ) may be considered as where f 0 is the mean sea level, f tid is the tide, and f res is the residual nonperiodic sea level oscillations. Furthermore, we analyzed the residual sea level records, which can be considered as the sum of lowfrequency (f LF ) and high-frequency (f HF ) sea level components; where f LF is a meteorologically induced sea level component that includes storm surges. Moreover, the high-frequency oscillations f HF can be represented as the sum of individual components of different origins as Here, f sei , f tsu , and f IG represent seiche, tsunami, and infra-gravitational (IG) wave components, respectively. Separating these components is not a trivial task.
Natural oscillations of gulfs and bays (seiches) can be excited by a wide range of different oceanic Figure 3 The location of the tide gauge stations (circles), the bathymetry of the region, and the calculated arrival time of the tsunami from a point with coordinates 36.8185 N°, 36.0496 E°(white isolines) from the time of the first earthquake (hours) using the wavefront orientation method (Fine and Thomson, 2013) Monserrat et al., 2006). IG waves are formed through the nonlinear interaction of wind waves and swell; the most intensive IG wave formation occurs in the coastal wave-breaking zone, where these waves are known as surf beats (Battjes, 1988;Munk, 1949). Typical periods of seiches are from a few minutes to a few hours, while conventional periods of IG waves are an order of magnitude shorter (30-300 s). The frequency bands of seiches and IG waves partly overlap, impeding their separation in our analyses. The spectral properties of the two processes are different (Medvedev et al., 2022), and the IG waves can be distinguished from atmospherically-induced seiches. The seiches are characterized by a series of resonant spectral peaks associated with the eigen frequencies of the corresponding basins, and the IG spectral component resembles white noise.

High-and Low-Frequency Filtering and f-t Analysis
We applied the approach described above to each station, i.e., carried out the time series analysis in several stages considering the components of sea level oscillations. The eastern part of the Mediterranean Sea is a micro-tidal area. The mean spring tidal range 2ðH M 2 þ H S 2 Þ, in this part of the sea, is about 0.3 m. The two earthquakes occurred during the spring tides, and the tidal range reached up to 0.38 m in Erdemli and up to 0.43 m in Arsuz ( Fig. 4a-b). First, we subtracted tides from the original sea level records using the least squares method (Pawlowicz et al., 2002).
Then, a high-pass filter was used with a 3-h Hamming window (Thomson & Emery, 2014) to separate f LF and f HF oscillations ( Fig. 4 and Fig. 5). Figure 4 shows f HF oscillations at Arsuz (Fig. 4c) and Erdemli (Fig. 4d) after the subtraction of tides (f tid ) and storm surges (f LF ). These records show sharp changes in the sea level character at 1:00-2:00 UTC on 6 February. The arrival of the tsunami corresponds to a low tide.
High-frequency records of sea level oscillations for all the stations are given in Fig. 5 and their f-t diagrams are presented in Fig. 6. We used the wavefront orientation method of Fine and Thomson (2013) based on Huygens' Principle (Shokin et al., 1987) for the calculation of tsunami arrival time (Table 2). This method is one of the most commonly used in modern tsunami travel-time calculations. Following Huygens' Principle, each of the points along a wavefront is a source for the tsunami. These points serve as start locations for travel-time computations to the next neighboring points. The calculated tsunami arrival time varies from 5 to 20 min in Arsuz to 2 h 35 min in Marmaris, which is the most distant from the seismic source among the analyzed gauges (Fig. 3).
The highest wave amplitudes were observed at Erdemli (17 cm) and Arsuz (15 cm), the two stations closest to the epicenter, with distances of *250 and 130 km respectively. The high-frequency sea level (f HF ) record at Arsuz shows significant noise before the earthquake. Sea level oscillations caused by the earthquake begin at Arsuz 20 min after the earthquake (Fig. 5). There is an immediate increase in the variance of f HF oscillations with a 12-min period for Arsuz and a 6-min period for Erdemli, as shown on the corresponding f-t diagram (Fig. 6). In Erdemli, the intensification of f HF oscillations persisted for slightly more than a day (Fig. 5) for the 6-min period.
The range of f HF in Arsuz increased over time (Fig. 5), reaching a maximum at the end of 7 February. This was most likely due to the meteorological impact, since the tsunami from the initial disturbance was weak.
In Taşucu, the f-t diagram showed a f HF oscillation component with a period of 60 min that was before the earthquake. After the earthquake, this oscillation increased and the signal intensity was maintained for almost a day. There was also a fluctuation with a 15-20 min period, which occurred after the tsunami arrival time and quickly attenuated.
At Bozyazı, after the tsunami arrival time the oscillation, with a period of 8-10 min, increased. The intensification persisted for slightly less than a day. In Bozyazı, the maximum f HF range was 13 cm, while in Taşucu, it was 6 cm. The calculated tsunami arrival times for these two locations are the same, 1 h b Figure 6 F-t diagrams of the high-frequency sea level records at Arsuz, Erdemli, Taşucu, Bozyazı, Antalya, and Marmaris for the period of 5-7 February 2023. The solid vertical lines indicate the times of the M w = 7.8 (01:17:35 UTC) and M w = 7.5 (10:24:49 UTC) earthquakes and the dashed white line represents the calculated tsunami arrival time generated by the M w = 7.8 earthquake. The first arrival at Arsuz (dashed line) is not shown as it occurred a short time after the earthquake. The Bozyazı record is shorter because the tide gauge had large gaps and then stopped working on 7 February Vol. 180, (2023) The Turkish Tsunami of 6 February 2023 in the Northeastern Mediterranean 3185 39 min, but the character of the tsunami signal manifestation in the records and on the f-t diagrams differs. At the two stations farthest from the source, Antalya and Marmaris, it is impossible to distinguish a tsunami in the sea level records; the tsunami signal appears only on f-t diagrams (Fig. 6). In Antalya, the calculated tsunami arrival time is 2 h 7 min. The oscillations there intensified, with a period of 30 min, after the arrival of the tsunami. This period corresponds to the eigenperiod (seiche) of the bay, as the strengthening persisted there for two days and it was manifested there before the earthquake. However, after the arrival of the tsunami, the intensification became evident. At Marmaris, we observed a f HF oscillation with a period of 30 min, and the eigenmode is also visible on the sea level records. However, after the arrival of the tsunami (2 h 35 min), the variance of this f HF oscillation intensified. The maximum wave amplitude recorded in Marmaris was 5 cm, while in Antalya it was 2 cm. We also considered tide gauges located farther than Marmaris along the Turkish, Greek, and Italian coasts. However, the tsunami signal did not appear either on tide gauge records or f-t diagrams, likely due to a weak initial disturbance. There is no available data on the closer coasts, i.e., in Israel or Egypt, although the wave could have been observed there. Also, the tide gauge record at Gazimagusa (Famagusta, Northern Cyprus) was presented in Ç etin et al. (2023). The tsunami amplitude there reached 17 cm, and the waves arrived there 65 min after the earthquake. Any clear signature from the second earthquake (M w = 7.5) could not be seen in either the tide gauge records or the f-t diagrams.

Spectral Analysis
The sea level measurements were divided into two parts to study the spectral properties of the sea level oscillations during the earthquakes on 6 February: (1) The 7 days, 19-25 January 2023, corresponding to background sea level variations before the event; this period was selected because this week was calm without storms, and (2) The 24 h beginning 00:00 (UTC) of 6 February 2023, for sea level oscillations during the tsunami.
We applied spectral analysis by the Welch method based on the fast Fourier transform (Thomson & Emery, 2014). The spectra were calculated using the Kaiser-Bessel spectral window (Thomson & Emery, 2014) with a length of N = 1024 values of 30-s sea level data, with half-window overlaps prior to the Fourier transform (Fig. 7). The full length of the background sea level record (7-day period) was 20160 values, the tsunami record (24-h period) had a length of 2 880 values. The number of degrees of freedom for the background spectra was m = 76, and for the tsunami spectra was m = 8. A. Medvedeva et al. Pure Appl. Geophys. In addition, the source functions R x ð Þ were calculated (Fig. 8), which quantify the amplification of the event spectrum S ts x ð Þ relative to background conditions S bg x ð Þ (Rabinovich, 1997): All spectra show an increase in spectral density across the entire frequency range. However, the distant stations exhibit weak spectral density at high frequencies (Fig. 7). For instance, Antalya is characterized by an increase in energy relative to the background level at frequencies up to 10 cycles per hour (cph) (Fig. 7e).
The spectral analysis results (Fig. 7) are in good agreement with the f-t analysis. Spectral peaks with 86, 39, and 16-11 min periods are observed at Arsuz (Fig. 7a), which is also well traced in the f-t analysis. Furthermore, an increase of spectral energy in the background spectrum at all frequencies was observed. At the Erdemli station (Fig. 7b), spectral peaks with periods of 20, 6.6, and 1.5 min stand out sharply. At the Taşucu station (Fig. 7c), oscillations of 60, 22, and 9.7 min are well-pronounced periods. The Bozyazı station (Fig. 7d) is characterized by oscillations with 13, 2.3, and 1.8 min periods. On the day of the earthquake, the Antalya station (Fig. 7e) did not experience strong changes in the background spectrum of sea level oscillations. However, the fluctuations with periods of 30 and 12 min showed themselves more significant than usual; an increase in energy is observed only in the low-frequency part of the spectrum. There are also weak changes at the Marmaris station (Fig. 7f); perhaps oscillations with periods of 37, 16, and 9.6 min showed themselves slightly more pronounced than usual.

Discussion
The earthquake coincided with a storm surge in the eastern Mediterranean Sea. Figure 9a-b show the time series of residual sea level oscillations (f res ) at Arsuz and Erdemli including, the low-frequency component (f LF ) of the sea level variations (with period T [ 3 h), which is the storm surge with the 0.2 m setup. At Arsuz, a low-frequency seiche with a period of about 3-4 h is seen (Fig. 9a). The first tsunami were observed just at the moment when the storm surge reached its maximum values. In Fig. 9cd, records of sea level oscillations f HF after high-pass filtering (T \ 3 h) are shown, similar to those shown in Fig. 5 but in a shorter time frame. The main difference between the records at Arsuz and Erdemli is the nature of the attenuation of sea level oscillations caused by the tsunami. At Erdemli, the amplitude of f HF significantly decreases within 12 h after the earthquake, and the standard deviation of r (f HF ) after 12:00 UTC on 6 February is only 1.6 times higher than the r (f HF ) for the recording interval before the earthquake, for example, from 12:00 on 5 February to 00:00 on 6 February. In Arsuz, the situation is quite different. The f HF energy after the arrival of tsunami not only does not die out, it continues to increase. Thus, the r (f HF ) before the tsunami was 2.1 cm, in the first 12 h of the tsunami it was about 5.3 cm, and after 12:00 UTC on 6 February, it increased to 5.9 cm. Figure 9e-f show the nonperiodic sea level oscillations f bp , selected using a narrow-band filter with periods from 6 min to 3 h. The sea level record for Erdemli is practically the same as the record in Fig. 9d, while the record for Arsuz differs significantly from Fig. 9c. Thus, the r (f bp ) for Arsuz was 1.9 cm before the tsunami, was about 4.7 cm in the first 12 h of the tsunami, and it decreased to 3.2 cm after 12:00 UTC on 6 February. The difference between Fig. 9c and Fig. 9e are fluctuations with a period of less than 6 min. These oscillations are IG waves and partly high-frequency modes of the bays' natural oscillations. Figure 10 shows the time series of high-frequency sea level fluctuations f HF from 31 January, 00:00 to 8 February, 00:00 for Arsuz. In the time series f HF , an enhancement in the following time frames can be distinguished starting from A1-31.01, 19:00 UTC, A2-1.02, 17:20 UTC, and A3-6.02, 1:00 UTC. All three selected areas of the time series are well manifested in the enhancement of f HF on the f-t diagrams (Fig. 10a). The maximum amplitudes * 24 cm (the height * 52 cm) of f HF occurred on the evenings of 1 and 7 February. The total high-frequency sea-level oscillations f HF (Fig. 10b) are divided into two components using a narrow-band filter: (1) Narrow-band signal f bp in the range of periods T from 6 min to 3 h (Fig. 10c). This part of the spectrum of sea-level fluctuations is formed by tsunami f tsu and seiches f sei . Moreover, natural oscillations of the sea level in bays could be excited both by the passage of a cyclone and by the arrival of tsunami.