Eddy Activities of the Surface Layer in the Western North Pacific Detected by Satellite Altimeter and Radiometer

Transcription

1 Journal of Oceanography Vol. 52, pp. 457 to Eddy Activities of the Surface Layer in the Western North Pacific Detected by Satellite Altimeter and Radiometer SHIGERU AOKI 1 * and SHIRO IMAWAKI 2 1 Department of Earth System Science and Technology, Interdisciplinary Graduate School of Engineering Sciences, Kyushu University, Kasuga, Fukuoka 816, Japan 2 Research Institute for Applied Mechanics, Kyushu University, Kasuga, Fukuoka 816, Japan (Received 5 October 1994; in revised form 4 November 1995; accepted 12 January 1996) Geosat radar altimeter data during the first year (from November 1986 to November 1987) of its Exact Repeat Mission are analyzed to estimate the eddy kinetic energy and propagation characteristics of anomalies of sea surface dynamic topography (SSDT) for the western North Pacific. SSDT anomalies are compared with anomalies of sea surface temperature (SST) derived from NOAA satellite radiometer data. The eddy kinetic energy (K e ) is large in the Kuroshio stationary meander region and Kuroshio Extension region. In the downstream region of the Kuroshio Extension, K e is especially large on the upstream and downstream sides of prominent bathymetric features. In the interior region of the subtropical gyre is found a zonal tongue of large K e at around N. Westward propagation is dominant in the SSDT and SST anomaly field at mid-latitudes. Longitude-time lag correlation diagrams reveal the coincidence of SSDT and SST anomalies statistically, which fact suggests the baroclinic nature of the anomalies. Zonal phase speeds of SSDT anomalies are approximately equal to the theoretical speeds of baroclinic first-mode long Rossby waves, but the meridional variation of observed phase speeds does not follow the simple theoretical variation of decreasing speeds monotonously with increasing latitudes. 1. Introduction Since the activity of mesoscale anomalies in the ocean was recognized in late 1960 s, many observations and numerical modelings have been carried out to understand their life cycles and roles in the general ocean circulation. In the upper layer of the North Pacific, it has been known that mesoscale anomalies have a large energy in the Kuroshio and Kuroshio Extension region (Wyrtki et al., 1976; Emery, 1983). Various studies have been done for the Kuroshio region near Japan where routine observations have been done extensively. In the Kuroshio Extension region, many studies dealt with propagation characteristics of anomalies (Roden, 1977; Bernstein and White, 1977, 1981; Mizuno and White, 1983). Using 300 m depth temperature data, Bernstein and White (1981) derived westward phase propagation of 3.8 cm/s at N in E. Mizuno and White (1983) detected eastward propagation of cm/s in the western region (32 42 N, E) and westward propagation of 1 2 cm/s in the eastern region (32 42 N, E), also using 300 m depth temperature data. *Present affiliation: National Institute of Polar Research, Kaga, Itabashi-ku, Tokyo 173, Japan.

2 458 S. Aoki and S. Imawaki In the eastern boundary region of the North Pacific, fluctuations are generated and propagated into the open ocean. Annual baroclinic Rossby waves have been observed and studied extensively (Bernstein and White, 1974; Emery and Magaard, 1976; White and Saur, 1981). White and Saur (1981) detected the westward phase speed of cm/s east of 130 W (35 38 N) from the depth of 9 C temperature and the salinity at the surface. In the central North Pacific at N, northwestward propagation of 1 cm/s of annual baroclinic Rossby wave whose wavelength is about 300 km was reported by Kang and Magaard (1980). Mesoscale anomalies have been observed to exist in the interior region of the western North Pacific (Wyrtki, 1975; Roden, 1977). The seasonal and interannual variations of the Subtropical Countercurrent (Hasunuma and Yoshida, 1978), the subtropical front (White et al., 1978), and the upper layer temperature (Lie and Endoh, 1991) have been studied from fragmental observations and climatological data set, but the detailed distribution of the eddy energy and the generation and propagation characteristics of mesoscale anomalies have not been fully understood. Moreover, even in relatively well-observed region as the Kuroshio Extension region, there are some discrepancies among studies as seen in the case of the zonal phase propagation speeds. These limitations of understanding come mainly from the difficulty in obtaining data in high resolution over broader regions and longer time spans by conventional ship observations. Recently, satellite observations have made it possible to describe phenomena in high resolution both in time and space, and over a broad region. Satellite altimeter data have been used to estimate the eddy activities and the dominant space-time structures of anomalies in various regions. The wavenumber spectra of anomalies of sea surface dynamic topography (the sea surface height relative to the geoid and related to the oceanic flow; hereafter abbreviated SSDT) along subsatellite tracks were studied by Fu (1983) from Seasat altimeter data to describe the regional variation of the characteristics of anomalies. From Geosat altimeter data, Shum et al. (1990) showed global distribution of the eddy kinetic energy which was derived from the fluctuation component of the geostrophic velocity. For the North Atlantic, Halliwell et al. (1991) studied westward propagation of baroclinic anomalies at the Subtropical Convergence Zone (STCZ) from sea surface temperature (SST) and SSDT derived from satellite observations. In the Pacific, some studies (Tai and White, 1990; Qiu et al., 1991) have described the phase propagation of anomalies and distribution of eddy kinetic energy for the Kuroshio Extension region. Jacobs et al. (1993) studied the propagation of barotropic and baroclinic Rossby waves using SSDT anomalies estimated from Geosat altimeter data. In the western North Pacific, however, the detailed distribution of the eddy kinetic energy and dominant propagation characteristics of anomalies have not yet been thoroughly studied. Aoki et al. (1995) described the behavior of distinct cyclonic anomalies and the propagation characteristics of anomalies for the Kuroshio Extension, using Geosat altimeter data and NOAA Advanced Very High Resolution Radiometer (AVHRR) data. The present study deals with the distribution of eddy kinetic energy, statistical properties of SSDT and SST anomalies, and meridional variation of their zonal phase speed in the entire western North Pacific. Observed zonal phase speeds are compared quantitatively with theoretical phase speeds of Rossby waves. In this study, the accuracy of altimeter-derived SSDT anomalies is also evaluated with tide gauge data. 2. Data and Analysis Method Data and procedures to obtain time series of SSDT and SST anomalies are briefly described. Tide gauge data are also described.

3 Eddy Activities in the Western North Pacific Sea surface dynamic topography We used Geosat altimetry data during the first year (from 8 November 1986 to 17 November 1987) of its Exact Repeat Mission (ERM). Detailed description of the data set was given by Cheney et al. (1987). Standard data corrections were made using values provided in the GDR (Geophysical Data Record). The solid-earth tide correction, ocean tide correction, wet and dry tropospheric correction, and ionospheric correction were all done; the Schwiderski (1980) tide model was used for oceanic tide correction. The inverse barometer and electro-magnetic bias corrections were not applied, because the correction algorithms have not yet been well established (vandam and Wahr, 1993; Ponte, 1994; Glazman et al., 1994). The one-year data consist of 22 cycles, one cycle being days long. The data of both ascending and descending tracks were used after editing the data to remove doubtful or extreme values. We averaged the original data (sampled every one second) over 10 data points to reduce small scale variations as well as to reduce the number of data points, the computational burden; therefore, our data are distributed every 67 km along the tracks. The sea surface height was obtained by subtracting the measured distance between the satellite and sea surface from the satellite radial orbit height. We used recalculated radial orbits (Haines et al., 1990) based on the GEM-T2 gravity model. The SSDT can be obtained by removing the geoid from the sea surface height. The OSU86F geoid model of Ohio State University, which is spherical harmonic expansions to order and degree of 360 (Rapp and Cruz, 1986), was used as a preliminary mean field. The radial orbit height data are considered to contain an error of about 40 cm rms. We used the optimal interpolation (Bretherton et al., 1976; Wunsch and Zlotnicki, 1984; Imawaki et al., 1991) to reduce this error and extract the most reliable signal from the data. Details of the optimal interpolation used in this study are described in Ichikawa and Imawaki (1996). We give here an outline of its actual application. We need a priori knowledge of the statistical characteristics of the signal and error. The correlation function of signal W( r) was given by the Gaussian shape ( ) = w 2 0 exp r W r 2 L, 1 where w 0 is the amplitude of the signal, L the de-correlation space scale, and r the distance between the data points or between the data point and grid point to be estimated. w 0 was set to 0.2 m. L was set to 150 km, considering that the distance between neighboring tracks is km, and so anomalies of wavelengths shorter than 150 km were removed. Note that no a priori propagation tendency with time is assumed. The correlation function of error Φ( t) was set as follows; ( ) Φ t ( ) = σ 2 0 δ( t) + σ 2 1 exp t T 1 cos 2π t. 2 The first term on the right-hand side is the correlation function of measurement error including actual sea level fluctuations during 17-day repeat cycle as well as errors of various environmental correction terms. The amplitude of this white noise σ 0 was set to 0.2 m; δ denotes Dirac s delta. T 0 ( )

4 460 S. Aoki and S. Imawaki The second term is for the orbit error; σ 1 is the amplitude, t the time lag between the data points, T 1 the de-correlation time scale, and T 0 the revolution period of the satellite orbit (101 min). σ 1 was set to 1.0 m. T 1 was determined to be 26T 0 from calculation of correlation function of the orbit error estimated approximately from the actual Geosat altimetry data. The signal at a certain point is estimated to have covariance with all the measurements in the form of Eq. (1), with minimizing its estimated error at the same time. Estimates are made on a grid for each repeat cycle by optimal interpolation. Since we also obtain the estimated error in the optimal interpolation, unreliable estimates can be removed out quite easily. The critical value of the estimated error was set to 0.14 m, which corresponds to a case where no available data from both ascending and descending tracks near the grid. 2.2 Sea surface temperature We used SST data derived from Tiros-N/NOAA AVHRR radiometer data. The data were provided as the Multi-Channel Sea Surface Temperature (MCSST) data set, where the atmospheric correction has already been made. SST estimates were obtained from the daytime data only and originally given weekly on a grid of about 18 km 18 km. We used the data from 5 November 1986 to 18 November 1987, which are of almost the same time period as that of the Geosat altimetry data. The temporal fluctuation of SST at a certain point is dominated by an annual signal of large spatial scale mostly due to the local heating and cooling. This feature is seen as a meridional trend of long wavelength. It is hard to detect mesoscale variation in a series of snapshots which are dominated by this large-scale feature. The SST fluctuation component is here divided into two parts: the north-south trend and departure from it. After excluding doubtful data, we averaged SST zonally to derive the trend component and subtracted it from the original data. Then we averaged the residual over the one-year observation period and subtracted the mean component from SST in each cycle; this final residual component is considered to represent the SST mesoscale variation. To compare the SST anomaly with the SSDT anomaly, we averaged values at 16 points (meridionally 4 points zonally 4 points) to eliminate short wavelengths; so the resulting data are located at about 72 km 72 km grid. This density of data distribution is comparable to that of 10-point-averaged along-track Geosat data. Estimates were made where SST data can be obtained at least at one point in these boxes. Then, the weekly SST anomalies derived above were linearly interpolated into the same 17-day periods as the individual Geosat repeat cycles. We made estimates at grids where at least one weekly SST datum was obtained in individual repeat cycles. Finally, estimates were made on the same grid as for the SSDT field using optimal interpolation. The correlation function of the signal was chosen as the Gaussian shape with amplitude of 1 C considering the rms amplitude of the signal. The precision of this data set is unknown, but McClain et al. (1985) compared Multi-Channel SST (though it is different from the present data set) with ship-derived SST and found rms error of C. Therefore, the correlation function of the error was set as the white noise with amplitude of 0.5 C. As a first step, the SST signal was obtained with the de-correlation spatial scale of 150 km. However, the largescale anomaly was found to have some energy even at this stage of processing. Therefore, we derived the remaining large-scale component by a second-step optimal interpolation using signal correlation function with decorrelation scale of 2,000 km. We subtracted the latter from the former to obtain the final estimate of the SST anomaly. Unreliable estimates whose estimated error exceeds 0.5 C are not used hereafter.

5 Eddy Activities in the Western North Pacific Tide gauge data To evaluate the accuracy of altimeter derived SSDT anomalies, we used the tide gauge data. The daily sea level data at 11 islands contained in Pacific and Indian Ocean Sea Level Data Set distributed by the National Oceanographic Data Center were used. We also used the data at Chichijima obtained by Japan Meteorological Agency. The inverse barometer correction was not applied for the tide gauge data as for the altimeter data. The tide gauge data are averaged in the same 17-days as each Geosat repeat cycle. Only the fluctuation component is used. 3. Evaluation of SSDT Anomalies To evaluate the accuracy of SSDT anomalies obtained from the altimeter data, we compared these anomalies with the sea level data from 12 tidal stations in the tropical and subtropical North Pacific; their locations are shown in Fig. 1. Only stations located on small islands are selected. The comparison is made between these tide gauge data and the altimeter-derived SSDT anomalies near that tide gauge station. Correlation between the SSDT and sea level is good at all 12 stations. For comparisons at each station, the reader should refer to Aoki and Imawaki (1994). Figure 2 is the scatter plot of Fig. 1. Analysis domain and locations of tide gauge stations for evaluating SSDT anomalies.

6 462 S. Aoki and S. Imawaki Fig. 2. Scatter diagram between the Geosat SSDT and tide gauge sea level anomalies at 12 stations whose locations are shown in Fig. 1. fluctuation between the altimeter and tide gauge data for all cycles and stations. It shows good correlation; the correlation coefficient is 0.76, and the rms difference is 5.5 cm. The rms of SSDT anomalies (7.2 cm) and that of sea level anomalies (7.9 cm) are comparable. 4. Results We examine the distribution of the eddy kinetic energy to understand where the anomalies are generated and grown for the western North Pacific. Then, we study the propagation characteristics of anomalies and their regional difference. For the Kuroshio Extension region, characteristics of SSDT and SST anomalies were described in detail by Aoki et al. (1995). 4.1 Eddy kinetic energy From the fluctuation component of the geostrophic surface velocity derived from SSDT anomalies, eddy kinetic energy (K e ) can be estimated. The distribution of K e is shown in Fig. 3. Large K e regions are found around the Kuroshio quasi-stationary meander region south of Honshu, Japan and in the Kuroshio Extension region. In the Kuroshio Extension region, there are two maximum K e regions (over 600 cm 2 /s 2 ) at 34 N, E and at 35 N, E. Compared with distribution of rms variability of SSDT anomalies shown by Aoki et al. (1995), K e is largest around the local maximum of the rms variability at 34 N, 147 E. In the downstream region, large K e (over 300 cm 2 /s 2 ) is also detected on the upstream and downstream sides of the Shatsky Rise at about 160 E and Emperor Seamounts at about 170 E (Fig. 4). To reveal the relation between K e and the location of the prominent bathymetric features, zonal distributions of meridional averages (over N) of K e and bottom depth are shown in Fig. 5. K e is large at 162 E (downstream side of the Shatsky Rise), and at 169 and 176 E (upstream sides of Emperor Seamounts and Hess Rise), and is slightly small just on those rises and seamounts. In W, K e is apparently weaker than west to 180 E (one fourth or less). Generally, axis of the maximum K e is shifted slightly southward east to 155 E. To clarify

7 Eddy Activities in the Western North Pacific 463 Fig. 3. Distribution of the eddy kinetic energy (in cm 2 /s 2 ) at the sea surface estimated from Geosat altimeter data. Contour interval is 100 cm 2 /s 2. Fig. 4. Bathymetry of the western North Pacific. Contour interval is 1,000 m. Letters S, E and H denote the Shatsky Rise, Emperor Seamounts and Hess Rise, respectively.

8 464 S. Aoki and S. Imawaki Fig. 5. Zonal distributions of the eddy kinetic energy (solid line) and bottom topography (broken line) averaged meridionally over N. Letters S, E and H denote the Shatsky Rise, Emperor Seamounts and Hess Rise, respectively. Fig. 6. Meridional distributions (in six different longitude bands) of the eddy kinetic energy (in cm 2 /s 2 ) averaged zonally over 10.

9 Eddy Activities in the Western North Pacific 465 the meridional shift of the position of K e maximum, meridional distributions of zonally averaged K e (over 10 longitude) are shown in Fig. 6. From 140 to 160 E, K e is largest at N, and from 160 to 180 E, it is largest at N. Another interesting feature seen in Fig. 3 is the zonal tongue of relatively large K e found along N. The region of large K e (over 200 cm 2 /s 2 ) is restricted in the Philippine Basin (west to 140 E), but the region of relatively large K e (over 100 cm 2 /s 2 ) extends from the western end of the North Pacific to 170 W. Note that estimated large kinetic energy south of 13 N may be mostly due to errors of SSDT anomalies, which result in large errors of geostrophic velocity anomalies because Coriolis parameter is small at low latitudes. 4.2 Propagation of anomalies at mid-latitudes To grasp propagation characteristics of mesoscale anomalies at mid-latitudes, we examine longitude-time lag correlation diagrams of both SSDT and SST anomalies. We divide the analysis domain into the western region (from 140 to 160 E) and eastern region (from 160 to 180 E) to detect the regional difference of characteristics. We choose the zonal extent of these subregions as 20 in order to capture mid-latitude Rossby waves and detect their changes as they are propagated. Lag correlation diagrams are examined at every 0.5 latitude, and here those on Fig. 7. Diagrams of longitude-time lag correlation of SSDT anomalies (from N). The abscissa is the zonal lag (in km) and the ordinate the time lag (in day). Positive values are shaded. a) is for the western region ( E), and b) the eastern region ( E).

10 466 S. Aoki and S. Imawaki Fig. 8. As in Fig. 7, but for SST anomalies. every 5 are shown in the following figures; diagrams at every 2 latitudes are shown in Aoki and Imawaki (1994). Diagrams (Fig. 7) of longitude-time lag correlation of SSDT anomalies clearly show westward propagations at mid-latitudes from 20 to 40 N in both the western region and eastern region. The wavelength can be estimated from the lag of the second peak of the correlation coefficient on the zero time lag. Typical spatial wavelengths are about km, and temporal scales are estimated to be about days as in the same way. In particular, westward propagations of wave-like features are clearly revealed at 35 N in the eastern region (Aoki et al., 1995). Diagrams are slightly different between west and east regions, although they have some similarities. Diagrams (Fig. 8) of longitude-time lag correlation of SST anomalies show similar westward propagations at N, though the correlation coefficient is generally smaller than that of SSDT anomalies. At 40 N in the western region, SST anomalies show no apparent propagation. To figure out the relation between SST and SSDT anomalies, longitude-time lag crosscorrelations of SSDT and SST anomalies are examined (Fig. 9). Diagrams of cross-correlation show that both anomalies are closely correlated with each other; the strongest correlation is found very close to the origin (zero lag) in most of the diagrams and inclinations of the high correlation axis are similar to those of lag correlation diagrams of both SSDT and SST anomalies (Figs. 7 and 8). Maximum correlation coefficients are and these values are significantly different

11 Eddy Activities in the Western North Pacific 467 Fig. 9. As in Fig. 7, but for diagrams of longitude-time lag cross-correlation between SSDT and SST anomalies. from zero at 95% confidence level. The diagrams show that high and low SSDT anomalies are related with warm and cold SST anomalies, respectively, which fact suggests that the observed mesoscale anomalies have the baroclinic nature. In here, we assume that those SST anomalies are not just showing temperature anomalies of the sea surface skin but representing temperature anomalies in the subsurface layer. Zonal phase speeds can be determined from the inclination of the high correlation axis in the longitude-time lag correlation diagrams of anomalies. We estimate phase speeds from the diagrams for SSDT and SST anomalies to understand the nature of propagation and clarify the difference between regions. The meridional variations of estimated zonal phase speeds in the western and eastern regions are shown in Fig. 10. Phase speeds estimated from diagrams for SSDT and SST anomalies give almost the same values. Generally speaking, westward phase speeds decrease with increasing latitudes in both regions. For the sake of comparison, theoretical westward phase speeds of baroclinic first-mode long Rossby waves Cp 0 are also shown in the figure; Cp 0 is given as Cp 0 = βa 2 e, ( 3) where β is the meridional gradient of Coriolis parameter, and a e the first-mode internal Rossby deformation radius. We use climatological mean a e estimated by Emery et al. (1984). Ap-

12 468 S. Aoki and S. Imawaki Fig. 10. Meridional distributions of the phase speed of SSDT anomalies (dots) and SST anomalies (open circles). The abscissa is the eastward phase speed (in cm/s), and the ordinate the latitude. Theoretical values of the first-mode baroclinic long Rossby wave using the internal Rossby radii by Emery et al. (1984) are shown by the solid line. a) is for the western region ( E), and b) the eastern region ( E). proximately, meridional variations of the observed phase speeds are consistent with those of the phase speeds of theoretical long baroclinic Rossby waves. However, there are some discrepancies between the observed and theoretical phase speeds. In the eastern region, observed westward phase speeds are almost equal to the theoretical phase speeds, but slightly lower at around 20 N, while they are higher at around N and 35 N. The differences are about 1 3 cm/s. In the western region, observed westward phase speeds are higher than the theoretical ones everywhere except at around 20 N, and their differences are about 1 4 cm/s. These departures from the phase speeds of simple theoretical long Rossby waves are discussed in the next section. 5. Discussion We compare the results obtained above with other studies and discuss the difference between observed zonal phase propagation speeds and phase speed of theoretical long baroclinic Rossby waves. Comparisons of the altimetric SSDT with tide gauge data show that the rms discrepancy is 5.5 cm and correlation coefficient is Cheney et al. (1989) obtained the rms of 3.7 cm and correlation coefficient of 0.68 from Geosat data during the Geodetic Mission ( ). For

13 Eddy Activities in the Western North Pacific 469 about 4 years Geosat data ( ), Miller and Cheney (1990) obtained the rms of 5.4 cm and correlation coefficient of These results are comparable with the present results. These values are sufficiently small for analyzing mesoscale variability at mid-latitudes in the western North Pacific. The distribution of the eddy kinetic energy (K e ) in the Kuroshio Extension region was shown from two-year XBT (expendable bathythermograph) data by Nishida and White (1982). They showed the large K e west of the Shatsky Rise and relatively smaller K e to the east. The present study shows the same decrease of K e at about 155 E, but the latter also shows some peaks in the downstream region. K e is large on the upstream sides of the Emperor Seamounts and Hess Rise, and this fact suggests that such bathymetric features have some influences on generation and growth of anomalies. Comparison of the distribution of K e with other studies using Geosat altimetry data (Shum et al., 1990; Qiu et al., 1991) shows qualitative agreement, but the present study shows smaller K e in this region, which is probably due to the spatial smoothing for scales less than 150 km in the optimal interpolation in the present study. The distribution of K e over the entire Pacific was shown by Wyrtki et al. (1976) on the basis of ship drift data. In the interior region, the detailed distribution has been poorly understood. The present study shows, for the first time, the existence of the large K e tongue along N. The reason for this large K e has not yet been understood, but it should be pointed out that this region corresponds to the center of the North Pacific subtropical gyre (in the climatological mean), namely the region of the highest geopotential anomalies (Hasunuma and Yoshida, 1978). Longitude-time lag cross-correlation diagrams between SSDT and SST anomalies (Fig. 9) suggest baroclinic nature of the anomaly field. If carefully examined, however, the highest correlation is found for SST at km to the west of SSDT. This westward shift of SST anomaly can be explained qualitatively by advection of SST by the geostrophic current associated with the SSDT anomaly (e.g. Halliwell et al., 1991). Detailed characteristics of zonal phase propagation have not been fully observed in the North Pacific. In the western North Atlantic, Halliwell et al. (1991) detected the westward propagation (3 4 cm/s) of baroclinic waves whose wavelength and period are about 800 km and 200 days, respectively, in the STCZ (roughly at N), and higher westward propagation (6 8 cm/s) to the north and south of the STCZ. It is interesting to note that, in the western North Pacific, similar meridional variations are observed, namely the westward phase speeds are lower at N than to the north and south of those latitudes (Fig. 10). Observed westward phase speeds are near to phase speeds of the baroclinic first-mode long Rossby waves, but are relatively higher than the latter. Moreover, detailed meridional variations of phase speeds cannot be explained by a monotonously decreasing function with increasing latitude which is expected by the theory. Jacobs et al. (1993, 1994) also indicate that simple baroclinic first-mode Rossby waves can explain only a fraction of actual variability and that anomalies of longer wavelengths are dominant. To explain observed phase speed quantitatively, we first consider a simple reduced-gravity model which includes effects of finite wavelengths and Doppler-shift by mean current of the upper layer. According to Kessler (1990), zonal phase speed Cp x of the baroclinic Rossby wave in the mid-latitude quasi-geostrophic system can be written as follows; Cp x = ( ) ( ) β ( a 2 e + k 2 + l 2 ) + U k 2 + l 2 a 2 e + k 2 + l 2 ( 4)

14 470 S. Aoki and S. Imawaki where k and l are the zonal and meridional wavenumbers, respectively. The second term on the right-hand side of Eq. (4) denotes the Doppler-shift by the mean current U (positive eastward); the effect varies with wavelength of anomalies and vanishes in the long wave limit. Attempts are made to explain discrepancies of phase speed between observed anomalies and theoretical long Rossby waves quantitatively, using the climatological mean of surface geostrophic current U relative to the 1000 dbar (Teague et al., 1990). The zonal wavenumber k is estimated from the wavelength which was derived by the lag correlation diagram of SSDT anomalies in the previous section (Fig. 7). The meridional wavenumber l is set to zero. The results are shown in Fig. 11. The discrepancy that westward phase speeds observed are lower than the theoretical wave speeds at about 20 N and 27 N in the eastern region and at about 20 N in the western region, can be explained by the effect of finite wavelengths in the first term in Eq. (4). However, the discrepancy that phase speeds observed are higher than theoretical phase speeds at about 30 and 35 N in the eastern region and to the north of 23 N in the western region, cannot be explained at all. The effect of the Doppler-shift of the mean current (the second term in Eq. (4)) is generally small or eastward over the whole region, and therefore, it cannot explain the discrepancy at these latitudes; Doppler-shift is westward at about 30 N in the western region, but its absolute value is negligibly Fig. 11. Meridional distributions of discrepancies (dots) between the phase speeds observed and the phase speeds of the theoretical first-mode baroclinic long Rossby waves. Triangles denotes the effect of finite wavelengths and squares the Doppler-shift by upper layer mean current. a) is for the western region, and b) for the eastern region.

15 Eddy Activities in the Western North Pacific 471 small (less than 0.5 cm/s). In the Kuroshio Extension region in the western region, this model is inappropriate because there exists a distinct eastward upper layer current. To explain the westward discrepancy by this model, unrealistic westward currents are required almost all over the region. We consider other possible effects which may explain the westward excess of observed phase speeds. First, westward mean current in the lower layer can cause the westward shift of phase speeds. Qiu et al. (1991) suggested a possibility of this effect in the Kuroshio Extension region. Imawaki and Takano (1982) showed from 3-year long moored current meter records that the zonal deep mean current at 5,000 m depth is about 0.6 cm/s westward at 30 N, 147 E, and not so salient. From 2-year long moorings along 152 E, Schmitz et al. (1987) showed westward current of about 4 cm/s at 4,000 m depth at 34 N in the Kuroshio Extension region, but the mean current at N is weak. Therefore, in the interior region, the simple two-layer model may not explain the discrepancy though the actual effects of the deep mean current have not yet been fully understood. Second, the bottom topography can cause changes in the phase speed (Pedlosky, 1979). Actual bottom topography, however, shows no continuous inclination to some large extent in these regions, and it cannot explain the westward shift in most areas at those latitudes. As mentioned above, the simple linear theory cannot successfully explain details of observed phase speeds at the whole mid-latitudes. In connection with these discrepancies, Tokmakian and Challenor (1993) obtained the result that the observed phase speed is higher than the phase speed of theoretical baroclinic long Rossby waves using the internal Rossby radius given by Emery et al. (1984) in the North Atlantic. Further discussions including more complicated effects (e.g. nonlinearity due to the finite amplitude effect) are necessary. Here, we did not mention about meridional propagation of the anomalies at all, although we did some analysis about that; meridional propagation does not have such a unique propagation direction as the zonal propagation, and small changes in the choice of the lag correlation domain make correlation diagrams quite different. Careful consideration is needed to detect the meridional propagation. Finally, it should be mentioned that the tidal correction error can cause aliasing at 317 days oscillation in case of M 2 constituent (Cartwright and Ray, 1990; Jacobs et al., 1992; Schlax and Chelton, 1994a, b). We estimate the potential tidal correction error component which has the same period (317 days), wavelength (7.97 in longitude) and westward phase propagation feature as the aliased tide correction error for M 2 (Schlax and Chelton, 1994b) using the least square fit. The potential tidal correction error is largest at the downstream region of the Kuroshio Extension region (about N, E). When removing this component, peaks of meridionally averaged K e over N (Fig. 5) diminish from 200 cm 2 /s 2 to cm 2 /s 2. However, locations of K e peaks do not change. Moreover, a large part of this potential tidal correction error component is considered to be true oceanic signal (Aoki et al., 1995). Therefore, the locations of K e are probably not largely affected by the tidal correction error aliasing. As for the phase propagation, removing the potential tidal correction error component does not alter the basic structure of correlation diagrams in the Kuroshio Extension region (Aoki et al., 1995). In the interior region, the amplitude of the potential tidal correction error is very small and lag correlation diagrams change very little by removing it. The high correlation between SSDT and SST anomalies also confirms that the SSDT anomalies are not just due to the tidal correction error but the true oceanic signal.

16 472 S. Aoki and S. Imawaki 6. Summary Satellite radar altimeter data of Geosat during the first year (from 8 November 1986 to 16 November 1987) of the ERM are analyzed to estimate SSDT anomalies for the western North Pacific. They are compared with SST anomalies derived from NOAA AVHRR. Using these data sets, we study the distribution of the eddy kinetic energy (K e ) and the propagation characteristics of anomalies. To evaluate the accuracy of SSDT anomalies, they are compared with sea level data from 12 tide gauge stations in the tropical and subtropical North Pacific. The rms discrepancy between the two data sets is about 6 cm, and the correlation coefficient is These results indicate that SSDT anomalies have necessary accuracy in analyzing mesoscale variability. The eddy kinetic energy (K e ) is large in the Kuroshio stationary meander region and Kuroshio Extension region, in agreement with previous studies. In the downstream region of the Kuroshio Extension, K e is large on the upstream and downstream sides of prominent bathymetric features including the Shatsky Rise and Emperor Seamounts. These bathymetric features may have an important role on the generation and growth of anomalies. In the center of the subtropical gyre is found a zonal tongue of large K e (100 cm 2 /s 2 ) at around N from western end to the date line. To clarify the propagation characteristics of mesoscale anomalies in detail, we divide the analysis domain into the western region ( E) and the eastern region ( E). The longitude-time lag correlation diagrams of SSDT anomalies in both subdomains clearly reveal that SSDT anomalies are propagated westward. SST anomalies also reveal similar westward propagation at N. Lag cross-correlation diagrams of SSDT and SST anomalies show that high and low SSDT anomalies almost coincide with warm and cold SST anomalies, respectively. This fact suggests that these anomalies have the baroclinic nature. The zonal phase speeds of these anomalies are roughly equal to those of baroclinic first-mode long Rossby waves, but the observed phase speeds do not exactly follow the simple theoretical variation of phase speed decreasing monotonously with increasing latitudes. In the eastern region, the observed speeds can be explained by theoretical speeds adjusted by effects of finite wavelength and upper layer Doppler-shift at some latitudes, but cannot be explained at around 30 and 35 N. In the western region, the observed westward phase speeds are slightly higher than the theoretical speeds at all latitudes except for around 20 N. To explain these discrepancies, more complicated effects including nonlinearity should be considered. Acknowledgements We wish to express our gratitude for the encouragement we received from Norihisa Imasato. Kaoru Ichikawa s contribution to data processing and his valuable discussion are greatly appreciated. We thank Akira Masuda for the suggestion in interpreting results. We also thank Carl Wunsch for his valuable suggestions about the altimeter data analysis and Charmaine King for helping us in the data processing. Yoichi Fukuda provided the information about the geoid model. Keiichi Nishimura helped us in the calculation of the geoid. We are also grateful to Ya Hsueh for his useful suggestions. Chet Koblinsky provided the Geosat radial orbit height data based on the GEM-T2 gravity model. The MCSST data set was provided by the Physical Oceanography Distributed Active Archive Center of the Jet Propulsion Laboratory. The tide gauge data were provided by National Oceanographic Data Center and Japan Meteorological Agency. William J. Teague provided the

17 Eddy Activities in the Western North Pacific 473 climatological mean dynamic height anomaly data. The computation was done on FACOM computers at Kyoto University and Kyushu University using the scientific subroutine library SSL II. This work was supported in part by the Grant-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture, Japan. References Aoki, S. and S. Imawaki (1994): Mesoscale anomalies in the western North Pacific derived from satellite observation data. Bull. Res. Inst. Appl. Mech. Kyushu Univ., 76, Aoki, S., S. Imawaki and K. Ichikawa (1995): Baroclinic disturbances propagating westward in the Kuroshio Extension region as seen by a satellite altimeter and radiometers. J. Geophys. Res., 100, Bernstein, R. L. and W. B. White (1974): Time and length scales of baroclinic eddies in the central North Pacific Ocean. J. Phys. Oceanogr., 4, Bernstein, R. L. and W. B. White (1977): Zonal variability in the distribution of eddy energy in the mid-latitude North Pacific Ocean. J. Phys. Oceanogr., 7, Bernstein, R. L. and W. B. White (1981): Stationary and traveling mesoscale perturbations in the Kuroshio Extension current. J. Phys. Oceanogr., 11, Bretherton, F. P., R. E. Davis and C. B. Fandry (1976): A technique for objective analysis and design of oceanographic experiments applied to MODE-73. Deep-Sea Res., 23, Cartwright, D. E. and R. D. Ray (1990): Oceanic tides from Geosat altimetry. J. Geophys. Res., 95, Cheney, R. E., B. C. Douglas, R. W. Agreen, L. Miller, D. L. Porter and N. S. Doyle (1987): Geosat Altimeter Geophysical Data Record User Handbook. NOAA Technical Memorandum, NGS-46, 28 pp. Cheney, R. E., B. C. Douglas and L. Miller (1989): Evaluation of Geosat altimeter data with application to tropical Pacific sea level variability. J. Geophys. Res., 94, Emery, W. J. (1983): On the geographical variability of the upper level mean and eddy fields in the North Atlantic and North Pacific. J. Phys. Oceanogr., 13, Emery, W. J. and L. Magaard (1976): Baroclinic Rossby waves as inferred from temperature fluctuations in the eastern Pacific. J. Mar. Res., 34, Emery, W. J., W. G. Lee and L. Magaard (1984): Geographic and seasonal distributions of Brunt-Väisälä frequency and Rossby radii in the North Pacific and North Atlantic. J. Phys. Oceanogr., 14, Fu, L. L. (1983): On the wave number spectrum of oceanic mesoscale variability observed by the SEASAT altimeter. J. Geophys. Res., 88, Glazman, R. R., A. Greysukh and V. Zlotnicki (1994): Evaluating models of sea state bias in satellite altimetry. J. Geophys. Res., 99, 12,581 12,591. Haines, B. J., G. H. Born and G. W. Rosborough (1990): Precise orbit computation for the Geosat Exact Repeat Mission. J. Geophys. Res., 95, Halliwell, G. R., Jr., Y.-J. Ro and P. Cornillon (1991): Westward-propagation SST anomalies and baroclinic eddies in the Sargasso Sea. J. Phys. Oceanogr., 21, Hasunuma, K. and K. Yoshida (1978): Splitting of the subtropical gyre in the western North Pacific. J. Oceanogr. Soc. Japan, 34, Ichikawa, K. and S. Imawaki (1996): Estimating the sea surface dynamic topography from Geosat altimetry data. J. Oceanogr., 52, Imawaki, S. and K. Takano (1982): Low-frequency eddy kinetic energy spectrum in the deep western North Pacific. Science, 216, Imawaki, S., K. Ichikawa and H. Nishigaki (1991): Mapping the mean sea surface elevation field from satellite altimetry data using optimal interpolation. Marine Geodesy, 15, 31 46, Jacobs, G. A., G. H. Born, M. E. Parke and P. C. Allen (1992): The global structure of the annual and semiannual sea surface height variability from Geosat altimeter. J. Geophys. Res., 97, 17,813 17,828. Jacobs, G. A., W. J. Emery and G. H. Born (1993): Rossby waves in the Pacific ocean extracted from Geosat altimeter data. J. Phys. Oceanogr., 23, Jacobs, G. A., W. J. Teague, J. L. Mitchell and H. E. Hurlburt (1994): An examination of the North Pacific ocean in the spectral domain using Geosat altimeter data and a numerical ocean. (to be submitted). Kang, Y. Q. and L. Magaard (1980): Annual baroclinic Rossby waves in the central North Pacific. J. Phys. Oceanogr., 10,

18 474 S. Aoki and S. Imawaki Kessler, W. B. (1990): Observations of long Rossby waves in the northern tropical Pacific. J. Geophys. Res., 95, Lie, H. J. and M. Endoh (1991): Seasonal and interannual variability in temperature of the upper layer of the northwest Pacific, J. Phys. Oceanogr., 21, McClain, E. P., W. G. Pichel and C. C. Walton (1985): Comparative performance of AVHRR-based multichannel sea surface temperatures, J. Geophys. Res., 90, Miller, L. and R. Cheney (1990): Large-scale meridional transport in the Tropical Pacific Ocean during the El Niño from Geosat. J. Geophys. Res., 95, Mizuno, K. and W. B. White (1983): Annual and interannual variability in the Kuroshio current system. J. Phys. Oceanogr., 13, Nishida, H. and W. B. White (1982): Horizontal eddy fluxes of momentum and kinetic energy in the near-surface of the Kuroshio Extension. J. Phys. Oceanogr., 12, Pedlosky, J. (1979): Geophysical Fluid Dynamics. Springer-Verlag, New York, 624 pp. Ponte, R. M. (1994): Understanding the relation between wind- and pressure-driven sea level variability. J. Geophys. Res., 99, Qiu, B., K. A. Kelly and T. M. Joyce (1991): Mean flow and variability in the Kuroshio Extension from Geosat altimetry data. J. Geophys. Res., 96, 18,491 18,507. Rapp, R. H. and J. Y. Cruz (1986): Spherical harmonic expansions of the Earth s gravitational potential to degree 360 using 30 mean anomalies. Report No. 376, The Ohio State Univ., 22 pp. Roden, G. I. (1977): On long-wave disturbances of dynamic height in the North Pacific. J. Phys. Oceanogr., 7, Schlax, M. G. and D. B. Chelton (1994a): Detecting aliased tidal errors in altimeter height measurements. J. Geophys. Res., 99, 12,603 12,612. Schlax, M. G. and D. B. Chelton (1994b): Aliased tidal errors in TOPEX/Poseidon sea surface height data. J. Geophys. Res., 99, 24,761 24,776. Schmitz, W. J. Jr., P. P. Niiler and C. J. Koblinsky (1987): Two-year moored instrument results along 152 E. J. Geophys. Res., 92, 10,826 10,834. Schwiderski, E. W. (1980): On charting global tides. Rev. Geophys. Space Phys., 18, , Shum, C. K., R. A. Werner, D. T. Shandwell, B. H. Zhang, R. S. Nerem and B. D. Tapley (1990): Variations of global mesoscale eddy energy observed from Geosat. J. Geophys. Res., 95, 17,865 17,876. Tai, C.-K. and W. B. White (1990): Eddy variability in the Kuroshio Extension as revealed by Geosat altimetry: Energy propagation away from the jet, Reynolds stress, and seasonal cycle. J. Phys. Oceanogr., 20, Teague, W. J., M. J. Carron and P. J. Hogan (1990): A comparison between the generalized digital environmental model and Levitus climatologies. J. Geophys. Res., 95, Tokmakian, R. T. and P. G. Challenor (1993): Observations in the Canary basin and the Azores frontal region using Geosat data. J. Geophys. Res., 98, vandam, T. M. and J. Wahr (1993): The atmospheric load response of the ocean determined using Geosat altimeter data. Geophys. J. Int., 83, 113, White, W. B. and J. F. T. Saur (1981): The source of annual baroclinic long waves to wind forcing in the eastern subtropical North Pacific. J. Phys. Oceanogr., 11, White, W. B., K. Hasunuma and H. Solomon (1978): Large-scale seasonal and secular variability of the subtropical front in the western North Pacific from 1954 to J. Geophys. Res., 83, Wunsch, C. and V. Zlotnicki (1984): The accuracy of altimetric surface. Geophys. J. Roy. Astron. Soc., 78, Wyrtki, K. (1975): Fluctuations of the dynamic topography in the Pacific Ocean. J. Phys. Oceanogr., 5, Wyrtki, K., L. Magaard and J. Hager (1976): Eddy energy in the Oceans. J. Geophys. Res., 81,