Nonseasonal sea level variations in the Japan/East Sea from satellite altimeter data

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi: /2004jc002387, 2004 Nonseasonal sea level variations in the Japan/East Sea from satellite altimeter data Byoung-Ju Choi and Dale B. Haidvogel Institute of Marine and Coastal Sciences, Rutgers University, New Brunswick, New Jersey, USA Yang-Ki Cho Department of Oceanography, Chonnam National University, Kwangju, Korea Received 16 March 2004; revised 4 October 2004; accepted 19 October 2004; published 24 December [1] A merged TOPEX/POSEIDON and ERS-1/2 altimeter data set is used to estimate nonseasonal variations of sea surface height (SSH) in the Japan/East Sea (JES) from October 1992 through February The altimeter data have a good correlation (0.95) with tide gauge data in the western JES. The nonseasonal variability of the SSH is higher in the southern warm region than in the northern cold region. The root-mean square value of the nonseasonal SSH is 5 10 cm in the southern area, whereas it is about 4 cm in the northern area. Empirical orthogonal function analysis of the nonseasonal SSH shows the first mode to be related to intraseasonal oscillations over the entire JES and the second mode to interannual path variations of the Tsushima Warm Current (TWC). The path change of the TWC is responsible for the high variability of SSH in the southern region. INDEX TERMS: 4223 Oceanography: General: Descriptive and regional oceanography; 4243 Oceanography: General: Marginal and semienclosed seas; 4275 Oceanography: General: Remote sensing and electromagnetic processes (0689); 4556 Oceanography: Physical: Sea level variations; 4572 Oceanography: Physical: Upper ocean processes; KEYWORDS: sea level, nonseasonal variations, Japan/East Sea, altimeter data, empirical orthogonal function analysis Citation: Choi, B.-J., D. B. Haidvogel, and Y.-K. Cho (2004), Nonseasonal sea level variations in the Japan/East Sea from satellite altimeter data, J. Geophys. Res., 109,, doi: /2004jc Introduction [2] The Japan/East Sea (JES) is a marginal sea located in the northwest Pacific Ocean. It comprises the Ulleung Basin and the Yamato Basin in the south, and the Japan Basin in the north (Figure 1a). The Yamato Rise is located in the middle. Water is exchanged through narrow channels with sill depths not exceeding 200 m while the maximum depth of the JES is about 3900 m. Twelve ground tracks of TOPEX/POSEIDON (T/P) pass over the JES with a distance between the tracks of about 250 km (Figure 1b) and an orbit repeat period of 9.9 days. ERS-1/2 have 41 ground tracks, a spacing of about 70 km, and a repeat cycle every 35 days. [3] The Tsushima Warm Current (TWC) enters the JES through the Korea (Tsushima) Strait (Figure 2a). Warm TWC water is confined to the upper few hundred meters in the JES owing to the shallow sill depths. After the TWC enters the JES, it generally divides into three branches: the Nearshore Branch (NB) along the Japanese coast, the East Korean Warm Current (EKWC) along the east coast of Korea, and the Offshore Branch (OB) into the south of the Ulleung Basin [Katoh, 1994; Hase et al., 1999]. The thickness of the warm water decreases with latitude until the permanent thermocline outcrops along the polar front. Copyright 2004 by the American Geophysical Union /04/2004JC The front along 38 N to 40 N separates the northern cold water region and the southern warm water region [Preller and Hogan, 1998; Tomczak and Godfrey, 1994]. [4] Sea surface steric height, which is calculated relative to 500 dbar using temperature and salinity climatological data from the Japan Oceanographic Data Center (JODC) and the Korea Oceanographic Data Center (KODC), is shown in Figure 2b. Geostrophic surface currents flow from the east coast of Korea to the Tsugaru and Soya straits with some meandering along the way. The NB is not identified in this figure because it flows along an isobath shallower than 200 m along the Japanese coast [Hase et al., 1999]. The EKWC and the polar front are smeared because the steric height is computed from long-term mean, 1 1 gridded hydrographic data. [5] Instantaneous composite dynamic sea level (DSL) will be estimated by combining the mean sea surface steric height calculated from the long-term mean hydrographic data and sea surface anomalies derived from the altimeters [Korotaev et al., 2003]. Composite DSL is the steric height relative to 500 dbar in this paper. Horizontal maps of composite DSL have the same physical meaning as those of dynamic height for the analysis of surface circulation. The composite DSL can be estimated by combining a mean sea surface height calculated by a three-dimensional numerical model and the sea surface anomalies from the altimeter data [Morimoto and Yanagi, 2001]. Although gradients of sea level across the polar front are in principle better 1of12

2 Figure 1. Bathymetry and satellite ground tracks in the Japan/East Sea. (a) Dashed line denotes the 200 m isobath and solid lines the 1000, 2000, and 3000 m isobaths. JB, YB, UB, UI, YR, KS, TS, and SS stand for the Japan Basin, Yamato Basin, Ulleung Basin, Ulleung Island, Yamato Rise, Korea Strait, Tsugaru Strait, and Soya Strait, respectively. (b) The ground tracks of TOPEX/POSEIDON (ERS-1/2) are shown as thick (thin) lines. Numbers from 1 to 7 along the coast indicate location of tide gauges at Sokcho, Mukho, Pohang, Izuhara, Hamada, Toyama, and Fukaura, respectively. maintained in a mean field from a numerical model than in historical climatological data, the locations of the front nonetheless vary depending on the details of the model simulation such as model used and parameters chosen. For this reason, we choose to use the mean field from long-term obervational data rather than from a numerical model. [6] Hirose and Ostrovskii [2000] found quasi-biennial variability in the southern Yamato Basin from along-track T/P data from 1992 to Since the along-track T/P altimeter data have a limited spatial sampling interval (250 km), the spatial structure of the quasi-biennial variability and mesoscale surface circulations could not be well resolved. Morimoto and Yanagi [2001] used 3.5 years of ERS-2 data and carried out empirical orthogonal function (EOF) analysis of composite DSL. Because seasonal variation of the mean surface circulation was dominant (93%) in the first two modes, other intraseasonal and interannual oceanic processes were buried and indiscernible. Analysis of 14-year-long tide gauge and hydrographic data sets reveals migration of the polar front, and identifies an eddy associated with intraseasonal and interannual variability with 43% of the total variance in the western JES [Kim et al., 2002]. This analysis was limited to the western JES because the hydrographic data did not have uniform data distribution in time and space. [7] This study aims to estimate nonseasonal variability of sea surface height in the JES using a merged T/P and ERS- 1/2 altimeter data set, and to find connections among the interannual variations at different locations. The merged altimeter data are described and validated against tide gauge data in section 2. Nonseasonal variabilities are examined in section 3. Major factors which induce sea level variability are discussed in section 4. Finally, we summarize our results with remarks in section Validation of Altimeter Data [8] The merged T/P and ERS-1/2 altimeter data at 1/3 degree spatial intervals and with a repeat cycle of 7 days were obtained from the Space Oceanography Division of Collecte Localisation Satellites (CLS) in Toulouse, France [Le Traon et al., 1998; Ducet et al., 2000]. Data from October 1992 through February 2002 are used in this study. T/P data are available for the whole period, but there are no merged T/P and ERS fields between January 1994 and March 1995 (ERS-1 geodetic phase). [9] Typically, tidal correction of altimeter data is carried out with the use of global tidal models such as Center for Space Research (CSR), Goddard Ocean Tide (GOT), and finite element simulation (FES) models [Eanes and Bettadpur, 1995; Ray, 1999; Le Provost et al., 1994]. However, it is difficult to remove the tidal signals completely from altimeter data in the marginal seas using the global tidal models because the global models do not have enough resolution in the marginal seas [Morimoto et al., 2000]. 2of12

3 Figure 2. (a) Schematic surface currents in summer and (b) mean sea surface steric height relative to 500 dbar. TWC, NB, OB, EKWC, NKCC and LC in Figure 2a stand for the Tsushima Warm Current, nearshore branch, offshore branch, East Korean Warm Current, North Korean Cold Current, and Liman Current, respectively. Mean surface currents flow northeastward and the contour interval is 2 cm in Figure 2b. Because the semidiurnal and diurnal periods of tides are much shorter than the sampling interval of any satellite altimeter, these tidal signals can appear in altimeter SSH data as aliased signals at periods much longer than the period of the original tides [Schlax and Chelton, 1994; Chen and Ezraty, 1997]. [10] The merged altimeter data were compared with tide gauge data at Ulleung Island. The altimeter data were obtained from the closest grid point to the tide gauge station. We expect the tide gauge data to properly reflect oceanic sea level variations because the island is small and the depth of the surrounding sea is over 1000 m. Atmospheric pressure effects were eliminated from daily mean sea level from the tide gauge using a barometric factor of 1 cm/mbar. To eliminate the signals shorter than a month in period, a lowpass filter (half-power period of 30 day) was applied. The merged altimeter sea level every 7 days (circles) is well correlated with the tide gauge data (line) in Figure 3a. The correlation coefficient between the altimeter and tide gauge data is 0.95, with a root-mean square (RMS) value of the difference of 3.0 cm. These results are comparable to or better than those of TOPEX in the tropical Pacific, where a correlation of 0.66 and an RMS difference of 4.3 cm are obtained [Mitchum, 1994]. Since seasonal variation of SSH is large at Ulleung Island, the mean seasonal cycle has been removed in Figure 3b. The mean seasonal cycle is defined by combining the time series of the annual and semiannual constituents obtained by harmonic analysis. The correlation coefficient between nonseasonal SSH from the altimeter and that from tide gauge is 0.93, with an RMS value of the difference of 2.7 cm. Simple averaging of the T/P data in time and space can reduce error estimates to the order of 2 cm [Cheney et al., 1994]. [11] In semienclosed marginal seas such as the JES and the Mediterranean Sea, sea level does not respond inversebarometrically to the atmospheric pressure at high frequencies (2 20 days) because the straits limit the water exchanges at high frequencies. Simple analytical models [Candela, 1991; Lyu et al., 2002], which take into account of friction in the straits, removes these atmospheric pressure-driven fluctuations better than the standard inverse barometer method. In the Mediterranean Sea, the RMS difference between the model correction and inverse barometer correction is about 2 cm [Le Traon and Gauzelin, 1997]. In the JES, the two corrections have an RMS difference of 2 3 cm [Nam et al., 2004]. Because the merged altimeter data set was processed using the inverse barometer method, the altimeter data set may contain about 2 3 cm of error from aliasing of the high-frequency fluctuations in the JES. 3. Results 3.1. Variability of Spatial-Mean Sea Level [12] The spatial mean of SSH over the JES has been calculated every 7 days with a latitude-dependent weighting function to ensure an equal-area average. The spatially averaged SSH variations reflect changes in total water volume in the JES. The time series is decomposed into a linear trend, a mean seasonal cycle, and nonseasonal components (Figure 4). The linear trend is obtained by least squares fitting of a straight line to the data. The three 3of12

4 Figure 3. Comparison of tide gauge data (line) and altimeter data (circles) at Ulleung Island. (a) SSH and (b) nonseasonal SSH. The daily sea levels are nontidal residuals that have been low-pass filtered to eliminate periods shorter than 30 days. Figure 4. Spatial mean of sea surface height from October 1992 to February (a) Mean SSH with the linear trend, (b) mean seasonal cycle, and (c) nonseasonal component. Note that the vertical scale varies from Figure 4a to Figure 4b to Figure 4c. 4of12

5 Figure 5. (a) The root-mean square (RMS) values of nonseasonal SSH and (b) percent of total variance in the nonseasonal variation. Contour interval is 1 cm in Figure 5a and 10% in Figure 5b. Gray color indicates values larger than 60% in Figure 5b. components account for 1.4%, 68.1% and 30.5% of the total variance, respectively. [13] SSH increases gradually from October 1992 to February 2002 with a slope of 0.58 cm/year, and is lowest in February 1996 and highest in October The range of the mean seasonal cycle is 13.8 cm and that of the nonseasonal component is 16.5 cm. The nonseasonal component shows both intraseasonal and interannual variability of total water volume. The time series of nonseasonal SSH is dominated by 50 to 200 day period fluctuations. Variancepreserving spectra of the nonseasonal SSH will be presented in section Spatial Structure of Nonseasonal Variability [14] SSH variability is not spatially uniform in the JES because of mesoscale activity such as the meandering of the TWC and migration of eddies. The spatial structure of the nonseasonal component is obtained by removing both the linear trend and the mean seasonal cycle from the altimeter data at every grid point. [15] A linear trend exists at most points and is presumably related to long-period oceanic processes. The mean seasonal signal mostly reflects seasonal heating and cooling of the surface layer. However, the mean seasonal cycle may include a seasonal portion of mesocale variability, which is enhanced or diminished by seasonal changes in stratification. As a result, when we remove the mean seasonal cycle from the data, the seasonal portion of the mesoscale variability may be lost. The nonseasonal component contains intraseasonal and interannual variations which are related to basin-wide oscillations, north-south migration of the polar front, meandering of the TWC, and eddy activities. [16] RMS values of the nonseasonal component are 5 10 cm south of 40 N and southwest of the Tsugaru Strait (Figure 5a). The largest variability, 10 cm RMS, occurs in the southern Yamato Basin (135 E, 38 N) where energetic quasi-biennial variability has been reported by Hirose and Ostrovskii [2000]. The Ulleung Basin, Yamato Basin and southwest Japan Basin (131.5 E, 39.5 N) also experience high nonseasonal variability. The shallow region between the Ulleung Basin and Yamato Basin has smaller variability than the surrounding deep sea. The nonseasonal component accounts for more than 60% of the total variance in the western interior, the central part along 39 N, west of the Tsugaru Strait, and offshore of Siberia (Figure 5b) EOF Analysis of Nonseasonal Variability [17] To examine statistically the temporal and spatial variability of the nonseasonal component of SSH, an EOF analysis was performed on the day cycles of 1/3 degree gridded nonseasonal SSH data. pffiffiffiffiffiffiffiffiffi The data were weighted with a weighting function cos q to ensure an equal-area treatment in the EOF analysis where q is latitude [North et al., 1982; Hendricks et al., 1996]. The EOFs were evaluated using singular value decomposition [Kelly, 1988], which separates the nonseasonal component of SSH into 488 orthogonal spatial modes and the associated amplitude time series. The orthogonal spatial modes were renormalized by the latitude-dependent weighting factor. We analyze here only the first two dominant modes, which are statistically independent and significant according to the criteria of Kelly [1988]. [18] Figure 6 shows the spatial patterns of the EOFs; the associated amplitude time series are plotted in Figure 7. The amplitude time series are normalized by their respective standard deviations. The spatial field plots correspond to a normalized amplitude time series value of +1, with units of centimeters. The SSH contribution of an EOF mode at any 5of12

6 Figure 6. The two empirical orthogonal functions (EOFs) describing the nonseasonal component of SSH. (a) Higher than 4 cm of spatial values are stippled in the first mode; (b) negative values are shaded in the second mode. given time is therefore found by scaling the spatial-field map by the appropriate amplitude time series value. Each dot in the amplitude time series represents a 7-day estimate. The amplitude time series of the first mode has intraseasonal and interannual variations while that of the second mode shows primarily interannual variation. [19] The first mode accounts for 38% of the total nonseasonal variance. Since the spatial values of the first Figure 7. The amplitude time series of the first two EOF modes. The time series have been normalized by their standard deviations and each dot represents a 7-day estimate. 6of12

7 Figure 8. Composite dynamic sea level (DSL) on (a) 6 September and (b) 4 October Sea level is raised about 11 cm from Figure 8a to Figure 8b over the entire Japan/East Sea. EOF are positive over the entire domain, the first mode extracts simultaneous sea level variation and represents total water volume change within the JES. Background values of the first EOF are about 3 cm. High spatial values of the first EOF are distributed over the transition zone between the cold and the warm water regions. Some regions have the first EOF larger than 4 cm (dotted area in Figure 6a): the Ulleung Basin, the southwest Japan Basin (132 E, 39.5 N), the southwestern Yamato Basin, and northern Yamato Basin. Thus, if the total volume of water increases (diminishes) within the JES, a large portion of the changed volume is added to (is subtracted from) these regions. The amplitude of the first mode (Figure 7a) is a normalized value of the spatially averaged nonseasonal SSH, so that there are intraseasonal variabilities (50 to 200 day periods) and interannual variations (300 to 700 day periods) in the time series as found in the time series (Figure 4c) and the spectra of spatial-mean SSH. [20] The second mode represents 8% of the total nonseasonal variance. The EOF of the second mode in Figure 6b shows a succession of three minima (gray areas) and four maxima (white areas) along 39 N. High values of this EOF are distributed only between 36 N and 40 N. The amplitude time series of the second mode (Figure 7b) has negative minima in July 1993, September 1995, September 1999 and October 2001, with positive maxima in October 1994, November 1996, September 1997 and September The magnitude of the extreme values increases from 1993 to Discussion [21] Physical interpretations of the two dominant modes are sought in this section using additional data including sea level from the coastal tide gauges and subsurface temperatures from concurrent observations. Detailed driving mechanisms are not pursued. Because interannual variations in the first mode are smaller than either interannual fluctuations in the second mode or intraseasonal variations in the first mode (Figure 7), the intraseasonal signals in the first mode and the interannual variations in the second are the focus of the following sections Intraseasonal Oscillations in the Japan/East Sea [22] Rapidly propagating barotropic long waves can simultaneously change sea level over a semienclosed basin of the dimensions of the JES. The intraseasonal variations of SSH in the first mode may thus be related to basin-wide oscillations of sea surface height. Such oscillations occur throughout the data record as shown in the amplitude time series of the first mode (Figure 7a). [23] As an example, there are intraseasonal oscillations from August 1995 to January During this period, the amplitude time series reaches a minimum in September and a maximum in October. Figure 8 shows the composite DSL distribution in these two months; sea level is uniformly raised about 11 cm from September to October with only slight changes of surface circulation. As the amplitude time series hits a minimum in November and a maximum in December, the sea level of the basin uniformly goes down 11 cm in November and up again 7 cm in December. [24] As we show next, tide gauges along the coast of the JES also record these intraseasonal variations. Tides whose periods are shorter than a month were removed from the hourly sea level data by harmonic analysis, and the inversebarometric component was eliminated. To eliminate signals shorter than a month in period, a lowpass filter with a halfpower period of 30 day was applied. Lastly, the mean seasonal cycle was subtracted from the time series. Each of the resulting time series is successively offset by 10 cm in Figure 9 to emphasize that the variations are nearly uniform in phase and amplitude. The time series represent nontidal and nonseasonal SSH, and their amplitudes are as large as 10 cm. Note that the nontidal and nonseasonal SSH time series at a given tidal station can differ from those at other stations because of local coastal effects such as wind setup and set-down, river runoff, or adjacent surface current variations. 7of12

8 Figure 9. Nonseasonal SSH from tide gauges along the coast of the Japan/East Sea. Locations of the tide gauges are Sokcho, Mukho, Pohang, Izuhara, Hamada, Toyama, and Fukaura (see Figure 1b) from top to bottom. [25] Estimates of coherence and phase, computed between pairs of tide gauge data sets, are shown in Figure 10. Coherence measures the linear time-invariant relationship between two time series and phase indicates whether one time series leads or lags the other. Sea level at Sokcho, Izuhara and Toyama are selected to represent the Korean coast, the Korea Strait, and the Japanese coast, respectively. For periods above 60 days, coherences between the time series are found to be higher than the 95% confidence level and the phase differences are near zero. [26] Since the amplitude of the first mode is a normalized measure of the spatially averaged nonseasonal SSH, we have compared the spatially averaged nonseasonal SSH from the altimeters with sea level from tide gauges in the straits. The nonseasonal component of spatial-mean SSH within the JES is correlated with sea level from a tide gauge (129.3 E, 34.2 N) in the middle of the Korea Strait with a correlation coefficient of 0.86 (Figure 11). The daily sea level data from the tide gauge were adjusted with daily atmospheric pressure and the adjusted sea level was filtered with a lowpass filter (half-power period of 30 day). Figure 10. (a) Coherences and (b) phases between the nonseasonal SSH time series from tide gauges at Sokcho, Izuhara, and Toyama. Phase is omitted if coherence is lower than the significance level. 8of12

9 Figure 11. (a) Nonseasonal components of SSH at Izuhara (thin line) and of spatial-mean SSH within the Japan/East Sea (thick line). (b) Coherence and (c) phase between the SSH at Izuhara and the spatialmean SSH within the Japan/East Sea. The nonseasonal component is departure from the linear trend and the mean seasonal cycle. Coherence analysis shows that they are highly coherent with near-zero phase difference above a 60-day period (Figure 11). By comparison, the nonseasonal component of sea level from tide gauges in the Tsugaru Strait has relatively less correlation (0.42) with spatial-mean SSH. The high coherence between sea level in the Korea Strait and spatial-mean SSH within the JES indicates that the barotropic intraseasonal variability of SSH may enter through the Korea Strait and modulate sea surface over the JES. Both power spectra of nonseasonal SSH at Izuhara (dashed line) and of the spatial-mean SSH within the JES (solid line) have dominant energy in the intraseasonal band from 50 to 200 day periods (Figure 12). There are three major peaks around 155, 100 and 80 days in both tide gauge and altimeter data. [27] Wind, atmospheric pressure, and Rossby or Kelvin waves have been known to cause intraseasonal variability of sea level in the Pacific and Indian Oceans [Enfield, 1987; Qiu et al., 1999]. The spatial-mean SSH within the JES is influenced by differences between the inflow and outflow transports through the straits. Hence the driving mechanisms of the spatial-mean SSH changes within the sea are closely related to those of the water exchanges through the straits. The possibility of barotropic oscillations induced by atmospheric pressure fluctuations of subinertial period (3 to 5 days) in the JES was proposed by Lyu et al. [2002]. Using a simple analytical barotropic model, Lyu [2003] suggested that monthly and interannual variations of the inflow through the Korea Strait and the mean sea level within the JES may be induced by sea level changes outside the Korea, Tsugaru and Soya Straits. Numerical model studies are presently underway to explore the effects of wind stress in Figure 12. Power spectra of the nonseasonal SSH at Izuhara (dashed line) and the spatial-mean SSH within the Japan/East Sea (solid line) in variance-preserving form. The spectra of the nonseasonal SSH at Izuhara (dashed line) is raised by 5 cm 2 for comparison with the other. PSD stands for power spectral density. 9of12

10 Figure 13. Flow patterns of the TWC with (a) amplitude maximum of the second mode and (b) with amplitude minimum. Solid, dashed and dotted lines are strongest surface currents derived from composite DSL on 6 November 1996, 17 September 1997 and 27 September 2000 for the positive maxima (13 September 1995, 22 September 1999, and 17 October 2001 for the negative minima). (c and d) The contour interval is 2.5 C and small dots represent temperature observation stations. the straits and sea level changes outside the straits on intraseasonal basin-wide variability Path Selection of the Main Tsushima Warm Current [28] Mean sea surface height from long-term hydrographic data has a north-south background gradient, i.e., SSH is high in the south and low in the north, so that the mean geostrophic surface current flows from the Korea Strait to the Tsugaru and Soya Straits (Figure 2b). As the normalized amplitude time series of the second mode approaches a positive maximum, the north-south gradient of SSH increases across the southern zero contour line in the second EOF by increasing SSH in the south and decreasing SSH in the middle, thus enhancing the north-south background gradient. As a consequence, a majority of the TWC flows along the southern zero contour line and warm water is generally confined to the southern region (southern white area of Figure 6b). As a result of the surface flow pattern, most of the warm water accumulates in the southern Yamato Basin and forms a large warm eddy. When the EOF amplitude approaches a negative minimum, the background north-south gradient of SSH diminishes across the southern zero contour line and warm water spreads over the region of negative EOF (gray area of Figure 6b) and forms several eddies in this region. In this case, the TWC meanders among the warm and cold eddies from the east coast of Korea to the Tsugaru Strait, and cold water fills the southern Yamato Basin to form a cold eddy or tongue (Figure 8). [29] From each instantaneous composite DSL map, the strongest horizontal gradient of DSL (strongest surface current) is traced by a line when the amplitude time series of the second mode has either a maximum (Figure 13a) or a minimum (Figure 13b). In most cases, the line coincides with a branch of the TWC and the submerged polar front. When the amplitude time series reaches a positive maximum, the strong surface current flows northeastward from the Korea Strait to the Tsugaru Strait. When the amplitude 10 of 12

11 time series hits a minimum, the surface current flows northward along the Korean coast, reaches about 40 N in the western JES, meanders to circumvent the southern Yamato Basin, and flows toward the Tsugaru Strait. [30] The amplitude time series of the second mode is a maximum in October 2000 and the main TWC flows along the dotted line in Figure 13a. Temperature observed at 100 m depth from September to October 2000 shows that warm water from the Korea Strait accumulates in the southern Yamato Basin. The temperature of the warm water is higher than 15 C in the southern Yamato Basin, and the cold water (<5 C) occupies the Ulleung Basin (Figure 13c). Temperature data were obtained from JODC and KODC. The amplitude time series is a minimum in October 1999 and the main TWC flows along the dashed line in Figure 13b. The subsurface temperature from September to October 1999 shows that warm water fills the Ulleung Basin (131 E, 37 N) and a cold eddy (tongue) intrudes from the north into the southern Yamato Basin (Figure 13d). Hence interannual path variations of the TWC mimic the subsurface temperature distribution and determine the location of the submerged polar front, which is the moving boundary between the cold and the warm water regions. [31] Hirose and Ostrovskii [2000] found strong interannual variation of SSH within the southern Yamato Basin and named it a quasi-biennial variation. They related the SSH variation to subsurface water density change using hydrographic data. Cold and fresh water appeared in July 1993 and in October 1995 while a warm and saline eddy developed in October 1996 in the southern Yamato Basin. They suggested that a weak summer monsoon wind stress curl field can excite the quasi-biennial variation. They also assimilated T/P along track data into a reduced gravity model and showed interactions between the path of the TWC and sea level anomalies of the southern Yamato Basin in October 1995 and October The events are well correlated with the amplitude time series of the second mode, i.e., the second mode of our analysis is consistent with their assimilated model results. [32] Our EOF analysis shows that the quasi-biennial variation is, in reality, related to path variations of the TWC: the TWC selects its main path, either the EKWC or the OB. When the main path of the TWC selects the OB, warm water accumulates in the southern Yamato Basin. Otherwise, the EKWC gains strength and warm water spreads over the western part of the sea. This interannual process occurs without accompanying changes in total water volume in the JES, so that it does not appear in the time series of spatial-mean SSH. [33] When the EKWC is strong and separates from the coast at about 38 N, the polar front aligns along 39 N to 40 N, the TWC meanders widely and its associated warm and cold eddies become evident in the western part of the sea (Figures 8 and 13b). The wavelength of the meandering is about 260 km estimated from the second EOF in Figure 6b. This is comparable to the magnitude (300 km) estimated from hydrographic data [Moriyasu, 1972]. 5. Summary and Concluding Remarks [34] Recent studies of sea level in the JES have found strong quasi-biennial variability in the southern Yamato Basin and significant interannual variation at Ulleung Island. The previous studies emphasized local processes and sampling intervals of the sea level were limited in time and space. Assimilation of T/P along track data into a reduced gravity model improved our understanding of connections among the interannual variations at different locations. [35] The merged T/P and ERS-1/2 altimeter data sets are appropriate to study mesoscale surface ocean circulation because of their small temporal (7 day) and spatial (1/3 ) sampling intervals. We used the merged altimeter data to examine variability of sea level and surface circulation from October 1992 to February The altimeter data correlates well (0.95) with concurrent tide gauge data at Ulleung Island, with an RMS difference of about 3 cm. [36] The nonseasonal component in spatial-mean SSH within the JES accounts for 30.5% of the total variance and has a range of 16.5 cm. RMS values of the nonseasonal SSH are about 4 cm in the northern cold water region and are 5 10 cm in the southern warm water region. The nonseasonal variation is dominant in the western side of the JES, into the eastern side along 39 N, and off the Tsugaru Strait. [37] EOF analysis has been used to decompose the nonseasonal SSH into 488 modes. The first two modes account for 38% and 8% of the total variance, respectively. The first mode isolates intraseasonal oscillations over the entire JES. Basin-wide oscillations of sea surface are thought to be responsible for the intraseasonal variations. The oscillations within the JES are correlated with sea level change in the Korea Strait. The second mode captures the interannual path selection of the TWC, which determines the strength of the EKWC, the location of the polar front, and the accumulation of warm water in the southern Yamato Basin. [38] Temporal and spatial variability of sea level in the JES have been quantified in this paper. However, the driving mechanisms of both intraseasonal oscillations and interannual path selection of the main TWC need further investigation in the future. Potential forcing processes include sea level change in the western Pacific Ocean, and interannual variation of monsoonal wind over the JES. [39] Acknowledgments. We appreciate support from the Office of Naval Research (ONR-N ), the National Science Foundation (NSF ), and the Institute of Marine and Coastal Sciences, Rutgers University. Y.-K. Cho was supported by the Meteorological Research Institute, Korea Meteorological Administration (ARGO program). The CLS Space Oceanography Division supplied the merged T/P and ERS-1/2 data. JODC and KODC provided temperature and salinity data. Air pressure data were obtained from the Korea Meteorological Administration and the NOAA-CIRES Climate Data Center. References Candela, J. (1991), The Gibraltar Strait and its role in the dynamics of the Mediterranean Sea, Dyn. Atmos. Oceans, 15, Chen, G., and R. Ezraty (1997), Non-tidal aliasing in seasonal sea-level variability and annual Rossby waves as observed by satellite altimetry, Annal. Geophys., 15, Cheney, R., L. Miller, R. Agreen, N. Doyle, and J. Lillibridge (1994), TOPEX/Poseidon: The 2-cm solution, J. Geophys. Res., 99, 24,555 24,563. Ducet, N., P. Y. Le Traon, and G. Reverdin (2000), Global high-resolution mapping of ocean circulation from the combination of T/P and ERS-1/2, J. Geophys. Res., 105, 19,477 19,498. Eanes, R., and S. Bettadpur (1995), The CSR 3.0 global ocean tide model, Tech. Memo CSR-TM-95-06, Cent. for Space Res., Univ. of Tex., Austin. 11 of 12

12 Enfield, D. B. (1987), The intraseasonal oscillation in eastern Pacific sea levels: How is it forced?, J. Phys. Oceanogr., 17, Hase, H., J.-H. Yoon, and W. Koterayama (1999), The current structure of the Tsushima Current along the Japanese coast, J. Oceanogr., 55, Hendricks, J. R., R. R. Leben, and G. H. Born (1996), Empirical orthogonal function analysis of global TOPEX/POSEIDON altimeter data and implications for detection of global sea level rise, J. Geophys. Res., 101, 14,131 14,145. Hirose, N., and A. G. Ostrovskii (2000), Quasi-biennial variability in the Japan Sea, J. Geophys. Res., 105, 14,011 14,027. Katoh, O. (1994), Structure of the Tsushima Current in the southwestern Japan Sea, J. Oceanogr., 50, Kelly, K. A. (1988), Comment on Empirical orthogonal function analysis of advanced very high resolution radiometer surface temperature patterns in Santa Barbara Channel by G. S. E. Lagerloef and R. L. Bernstein, J. Geophys. Res., 93, 15,753 15,754. Kim, K., Y.-K. Cho, B.-J. Choi, Y.-G. Kim, and R. C. Beardsley (2002), Sea level variability at Ulleung Island in the East (Japan) Sea, J. Geophys. Res., 107(C3), 3015, doi: /2001jc Korotaev, G., T. Oguz, A. Nikiforov, and C. Koblinsky (2003), Seasonal, interannual, and mesoscale variability of the Black Sea upper layer circulation derived from altimeter data, J. Geophys. Res., 108(C4), 3122, doi: /2002jc Le Provost, C., M. L. Genco, F. Lyard, P. Vincent, and P. Canceil (1994), Tidal spectroscopy of the World Ocean tides from a finite element hydrodynamic model, J. Geophys. Res., 99, 24,777 24,797. Le Traon, P. Y., and P. Gauzelin (1997), Response of the Mediterranean mean sea level to atmospheric pressure forcing, J. Geophys. Res., 102, Le Traon, P. Y., F. Nadal, and N. Ducet (1998), An improved mapping method of multisatellite altimeter data, J. Atmos. Oceanic Technol., 15, Lyu, S. J. (2003), Temporal variation of the transport from cable voltage across the Korea Strait and its mechanism, Ph.D. thesis, 136 pp., Seoul Natl. Univ., Seoul. Lyu, S. J., K. Kim, and H. T. Perkins (2002), Atmospheric pressure-forced subinertial variations in the transport through the Korea Strait, Geophys. Res. Lett., 29(9), 1294, doi: /2001gl Mitchum, G. T. (1994), Comparison of TOPEX sea surface heights and tide gauge sea levels, J. Geophys. Res., 99, 24,541 24,553. Morimoto, A., and T. Yanagi (2001), Variability of sea surface circulation in the Japan Sea, J. Oceanogr., 57, Morimoto, A., T. Yanagi, and A. Kaneko (2000), Tidal correction of altimeter data in the Japan Sea, J. Oceanogr., 56, Moriyasu, S. (1972), The Tsushima current, in Kuroshio: Its Physical Aspects, edited by H. Stommel and K. Yoshida, pp , Univ. of Tokyo Press, Tokyo. Nam, S. H., S. J. Lyu, Y. H. Kim, K. Kim, J.-H. Park, and D. R. Watts (2004), Correction of TOPEX/POSEIDON altimeter data for nonisostatic sea level response to atmospheric pressure in the Japan/East Sea, Geophys. Res. Lett., 31, L02304, doi: /2003gl North, G. R., T. L. Bell, and R. F. Cahalan (1982), Sampling errors in the estimation of empirical orthogonal functions, Mon. Weather Rev., 110, Preller, R. H., and P. J. Hogan (1998), Oceanography of the Sea of Okhotsk and Japan/East Sea, Coastal segment (11,S), in The Sea, vol. 11, edited by A. L. Robinson and K. H. Brink, pp , John Wiley, Hoboken, N.J. Qiu, B., M. Mao, and Y. Kashino (1999), Intraseasonal variability in the Indo-Pacific Throughflow and the regions surrounding the Indonesian Seas, J. Phys. Oceanogr., 29, Ray, R. (1999), A global ocean tide model from TOPEX/Poseidon Altimetry: GOT99.2, NASA Tech. Memo NASA/TM , Goddard Space Flight Cent., NASA, Greenbelt, Md. Schlax, M. G., and D. B. Chelton (1994), Aliased tidal errors in TOPEX/ POSEIDON sea surface height data, J. Geophys. Res., 99, 24,761 24,775. Tomczak, M., and J. S. Godfrey (1994), Regional Oceanography: An Introduction, 442 pp., Elsevier, New York. Y.-K. Cho, Department of Oceanography, Chonnam National University, Kwangju , Korea. B.-J. Choi and D. B. Haidvogel, Institute of Marine and Coastal Sciences, Rutgers University, New Brunswick, NJ , USA. (bchoi@ marine.rutgers.edu) 12 of 12