Portal de Periódicos da CAPES

Modulation of Kuroshio transport by mesoscale eddies at the Luzon Strait entrance

2014; Wiley; Volume: 119; Issue: 4 Linguagem: Inglês

10.1002/2013jc009548

2169-9291

Ren‐Chieh Lien, B. Barry, Yu Cheng, Chung‐Ru Ho, Bo Qiu, Craig M. Lee, Ming‐Huei Chang,

Tropical and Extratropical Cyclones Research

Journal of Geophysical Research: OceansVolume 119, Issue 4 p. 2129-2142 Research ArticleFree Access Modulation of Kuroshio transport by mesoscale eddies at the Luzon Strait entrance Ren-Chieh Lien, Corresponding Author Ren-Chieh Lien Applied Physics Laboratory, University of Washington, Seattle, Washington, USACorrespondence to: R.-C. Lien, [email protected]Search for more papers by this authorBarry Ma, Barry Ma Applied Physics Laboratory, University of Washington, Seattle, Washington, USASearch for more papers by this authorYu-Hsin Cheng, Yu-Hsin Cheng National Taiwan Ocean University, Keelung, TaiwanSearch for more papers by this authorChong-Ru Ho, Chong-Ru Ho National Taiwan Ocean University, Keelung, TaiwanSearch for more papers by this authorBo Qiu, Bo Qiu Department of Oceanography, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, Hawaii, USASearch for more papers by this authorCraig M. Lee, Craig M. Lee Applied Physics Laboratory, University of Washington, Seattle, Washington, USASearch for more papers by this authorMing-Huei Chang, Ming-Huei Chang National Taiwan Ocean University, Keelung, TaiwanSearch for more papers by this author Ren-Chieh Lien, Corresponding Author Ren-Chieh Lien Applied Physics Laboratory, University of Washington, Seattle, Washington, USACorrespondence to: R.-C. Lien, [email protected]Search for more papers by this authorBarry Ma, Barry Ma Applied Physics Laboratory, University of Washington, Seattle, Washington, USASearch for more papers by this authorYu-Hsin Cheng, Yu-Hsin Cheng National Taiwan Ocean University, Keelung, TaiwanSearch for more papers by this authorChong-Ru Ho, Chong-Ru Ho National Taiwan Ocean University, Keelung, TaiwanSearch for more papers by this authorBo Qiu, Bo Qiu Department of Oceanography, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, Hawaii, USASearch for more papers by this authorCraig M. Lee, Craig M. Lee Applied Physics Laboratory, University of Washington, Seattle, Washington, USASearch for more papers by this authorMing-Huei Chang, Ming-Huei Chang National Taiwan Ocean University, Keelung, TaiwanSearch for more papers by this author First published: 14 March 2014 https://doi.org/10.1002/2013JC009548Citations: 81AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Abstract Measurements of Kuroshio Current velocity at the entrance to Luzon Strait along 18.75°N were made with an array of six moorings during June 2012 to June 2013. Strong positive relative vorticity of the order of the planetary vorticity f was observed on the western flank of the Kuroshio in the upper 150 m. On the eastern flank, the negative vorticity observed was about an order of magnitude smaller than f. Kuroshio transport near its origin is computed from direct measurements for the first time. Kuroshio transport has an annual mean of 15 Sv with a standard deviation of 3 Sv. It is modulated strongly by impinging westward propagating eddies, which are identified by an improved eddy detection method and tracked back to the interior ocean. Eight Kuroshio transport anomalies >5 Sv are identified; seven are explained by the westward propagating eddies. Cyclonic (anticyclonic) eddies decrease (increase) the zonal sea level anomaly (SLA) slope and reduce (enhance) Kuroshio transport. Large transport anomalies of >10 Sv within O(10 days) are associated with the pairs of cyclonic and anticyclonic eddies. The observed Kuroshio transport was strongly correlated with the SLA slope (correlation = 0.9). Analysis of SLA slope data at the entrance to Luzon Strait over the period 1992–2013 reveals a seasonal cycle with a positive anomaly (i.e., an enhanced Kuroshio transport) in winter and spring and a negative anomaly in summer and fall. Eddy induced vorticity near the Kuroshio has a similar seasonal cycle, suggesting that seasonal variation of the Kuroshio transport near its origin is modulated by the seasonal variation of the impinging mesoscale eddies. Key Points Kuroshio transport at its origin is directly measured Kuroshio transport anomalies are primarily due to westward propagating eddies Kuroshio transport is strong in spring and weak in fall 1. Introduction The Kuroshio is the primary poleward western boundary current of the western Pacific. It originates from the North Equatorial Current (NEC), passes Luzon Strait, and flows along the east coast of Taiwan, the continental slope of the East China Sea, and south of Japan [Nitani, 1972]. It then separates from the coast as a free inertial jet—the Kuroshio Extension [Qiu, 2002]. Because of its strong poleward transport of mass, heat, and salt, the Kuroshio plays an important role in large-scale ocean circulation and climate variability. As the Kuroshio flows past the South China, Philippine, and East China seas, there are many opportunities for exchange of water masses, especially in regions of strong turbulence mixing such as the Luzon Strait [Centurioni et al., 2004; Jan et al., 2008] and the lee of Green Island [Chang et al., 2013]. Direct measurements of Kuroshio velocity and transport are rare. During the World Ocean Circulation Experiment (WOCE), an array of moorings deployed east of Taiwan measured a Kuroshio transport of 21 Sv [Johns et al., 2001]. Similar transport values were obtained along the east coast of Taiwan using historical shipboard ADCP [Liang et al., 2003] and XBT/XCTD [Gilson and Roemmich, 2002] data. The Kuroshio transport upstream of the Taiwan coast has been derived primarily with hydrographic methods. East of Luzon, various estimates of the Kuroshio transport have been reported: 20 Sv by Sverdrup [1942], 30 Sv by Nitani [1972], 14 Sv by Qu et al. [1998], and 27.6 Sv by Yaremchuk and Qu [2004]. These estimates assume a geostrophic balance and are highly sensitive to the assumed level of no motion. Nitani [1972] reports that the estimate of Kuroshio transport east of Luzon is reduced to 80% if the level of no motion is changed from 1200 db to 800 db, and to ∼60% if the level of no motion is changed to 600 db. Qu et al. [1998] remark that repeated hydrographic sections are needed to minimize the effects of internal tides and eddies on the Kuroshio transport estimate. Kontoyiannis and Watts [1990] report that the ageostrophic component contributes to 30–60% of the Gulf Stream during periods of coalescence of cold-core rings. Johns et al. [1989], also studying the Gulf Stream, report that the velocity computed assuming geostrophic balance underestimates the current by 0.1–0.25 m s−1 in its core. Toole et al. [1990] show that the relative vorticity is significant in conserving potential vorticity in the Kuroshio, suggesting an ageostrophic balance too. Direct measurements are needed to provide a better estimate of the Kuroshio transport and to assess the assumption of geostrophy. Seasonal variations of Kuroshio transport off the Luzon coast have been studied previously. Using a linear time-dependent Sverdrup theory and a nonlinear reduced-gravity model, Qiu and Lukas [1996] report that the Kuroshio across 20°N has a seasonal minimum (maximum) in fall (spring) when the NEC bifurcation is at its northernmost (southernmost) latitude. The amplitude of the seasonal variation is 5.5 Sv. Combining hydrographic data, atmospheric climatologies, and drifter data, Yaremchuk and Qu [2004] also report stronger (weaker) Kuroshio transport across 18.5°N in spring (fall), with an amplitude of ∼3 Sv. Similar seasonal variation of Kuroshio transport was reported by Wyrtki [1961]. Chang and Oey [2012] define the Philippines-Taiwan Oscillation (PTO) as the difference between the wind stress curls averaged off the Philippines and Taiwan. During a positive PTO phase, the NEC shifts northward, eddy activity is enhanced, Kuroshio transport northeast of Taiwan increases, and westward transport into the South China Sea through Luzon Strait also increases. The later implies a weaker Kuroshio transport off Luzon following Sheremet [2001]. Because the PTO has a clear seasonal variation, maximum in winter and minimum in summer, a seasonal variation of Kuroshio transport off Luzon is inferred from the Chang and Oey [2012] study. Previous studies report that Kuroshio transport east of Taiwan is modulated strongly by the westward propagating eddies at a time scale O(100 days) [Yang et al., 1999; Johns et al., 2001; Zhang et al., 2001; Lee et al., 2013; Chang and Oey, 2012]. Anticylonic (cyclonic) eddies increase (decrease) Kuroshio transport. Hsin et al. [2013] investigate the eddy effect on the interannual variation of Kuroshio transport along the east coast of Taiwan. The relative intensity, rather than the number, of cyclonic versus anticyclonic eddies is found to be the major cause of the interannual variation of Kuroshio transport. Vélez-Belchí et al. [2013] demonstrate that the cyclonic eddies may also induce large Kuroshio intrusion onto East China Sea continental shelf. These eddies are generated along the North Pacific Subtropical Countercurrent (STCC) as a result of baroclinic instability between the surface eastward flowing STCC and the subsurface westward flowing NEC [Qiu, 1999; Roemmich and Gilson, 2001; Kobashi and Kawamura, 2002]. Except for direct mooring measurements reported by Zhang et al. [2001], Kuroshio transport in these studies is estimated from hydrographic surveys, sea level differences across the Kuroshio, numerical model results, or short-period shipboard surveys. Most of these studies also do not distinguish between Rossby waves and eddies. Barron et al. [2009] report that both linear Rossby waves and eddies in the western Pacific move westward. They both contribute as westward propagating anomalies in SLA time series analyses. However, the vast majority of eddies are nonlinear, defined by the eddy Froude number Fr = |Ueddy|/Ceddy > 1, where |Ueddy| is the current speed of the eddy and Ceddy is the eddy propagation speed [Chelton et al., 2007, 2011]. Linear Rossby waves do not carry fluid effectively. Because nonlinear eddies can transport isolated water for a long distance within its inner core [Early et al., 2011], their impingement and interaction could modify the Kuroshio water mass. Here we make the important distinction between linear Rossby waves and nonlinear eddies. Hwang et al. [2004], studying mesoscale eddies over the STCC using satellite altimetry data, report that there are more anticyclonic than cyclonic eddies. Anticyclonic (cyclonic) eddies prevail in summer (winter). These eddies often have an elliptical instead of circular shape. The effect of these eddies on the Kuroshio transport south of Luzon Strait has not been studied. Here we analyze direct velocity measurements taken by a moored array of the Kuroshio south of Luzon Strait. Estimates of Kuroshio velocity structure and transport from the moored array are presented in section 2.. A correlation of the Kuroshio transport time series with the Aviso Sea Level Anomaly (SLA) slope is presented (section 3.), showing a O(10) day variability, distinctly faster than the O(100) day variability reported by previous studies of Kuroshio transport variations east of Taiwan. Methods are presented to detect, track, and analyze eddies impinging on the eastern limit of the Kuroshio. Seven of the eight observed Kuroshio transport anomalies during the mooring observation period can be explained by the westward propagating eddies. Analysis of 20 years of Aviso SLA slope data (section 4.) reveals a seasonal maximum of Kuroshio transport in spring and minimum in fall. Eddy tracking analysis is applied to 20 years of Aviso data and a seasonal variation is found that is similar to the Kuroshio transport variation. A summary is given in section 5.. 2. Kuroshio Velocity and Transport Experiment and Measurements Six moorings were deployed along 18.75°N in a zonal section between 122°E and 122.87°E, each roughly 16 km apart, spanning ∼80 km at the Kuroshio entrance to Luzon Strait (Figure 1; red dots). Previous shipboard ADCP observations [Liang et al., 2003], drifter observations [Centurioni et al., 2004], and results from a high-resolution regional ocean model, the East Asian Seas Nowcast/Forecast System (EASNFS) of the U.S. Naval Research Laboratory [Rhodes et al., 2002] were used to determine mooring positions aimed at capturing the fullest extent of the Kuroshio transport. The Kuroshio path averaged between June 2012 and May 2013 is apparent in an Aviso map of absolute dynamic topography (AMDT), showing that the major portion of Kuroshio transport is within the span of the moored array (Figure 1c). Figure 1Open in figure viewerPowerPoint Positions of six moorings at the entrance to Luzon Strait (red dots), Aviso sea level anomaly (SLA) (color shading), and Aviso surface current anomaly (vectors) on (a) 10 May 2013 and (b) 5 June 2013. Velocity reference scale of 0.5 m s−1 is labeled in Figure 1a. Two eddies leading to Kuroshio transport anomaly events 7 and 8 are labeled in Figure 1b. Blue curves northeast of Luzon represent Seaglider tracks. Figure 1c shows the map of absolute dynamic topography (MADT) and Aviso surface current. Each mooring was equipped with one upward looking 75 kHz ADCP at 450 m nominal depth. The ADCPs took velocity measurements every 1.5 min and recorded averages every 15 min in 8 m vertical bins between 450 m and 45 m depth over the period June 2012–June 2013. The ADCP velocity measurements between the sea surface and 45 m depth were contaminated by the surface reflection of acoustic beams, but are estimated by two extrapolation methods (section 2.3.). Two Seagliders collected time series of high-resolution sections along the mooring line from November 2012 to May 2013 (sg124 from 3 November 2012 to 17 March 2013 and sg177 from 18 March to 31 May 2013). Seaglider is a buoyancy-driven, long-range, autonomous, underwater vehicle designed for oceanographic research [Eriksen et al., 2001]. Each Seaglider was equipped with a Sea-Bird CTD, which sampled temperature and salinity every 8 s while profiling to 1000 m at roughly 0.5 m s−1. The Seagliders surveyed a section between 121.8°E and 123.2°E nominally at 18.75°N along the mooring line. Strong currents associated with the Kuroshio sometimes deflected the Seaglider track from the mooring line. Fifteen transects (eight westbound, seven eastbound) were made. On average, westbound transects took 6 days and 28 dives, while eastbound transects took 12 days and 60 dives. Depth-average currents were calculated from the difference between GPS-derived surface positions at the start and end of a dive, combined with a hydrodynamic model of glider motion. Geostrophic shear normal to a section was calculated from along-section lateral density gradients, and turned into absolute geostrophic velocity by referencing to the glider-derived depth-average velocity. In this analysis, Kuroshio transport is computed mainly from direct mooring measurements. The geostrophic transport estimated from Seaglider is compared with the mooring measurements and is used to evaluate the horizontal span of the Kuroshio and to identify the possible Kuroshio transport missed by the moored array (section 2.3.2.3). Kuroshio Velocity and Vorticity Mooring velocity data are low-pass filtered at a 10 day interval to remove tides and high-frequency fluctuations. The monthly averaged meridional current is mostly northward in the observed depth range (Figure 2). In the following analysis, we define the Kuroshio as the low-frequency meridional current being northward. One striking feature is the weak northward current, <0.5 m s−1, in June 2012 (Figure 2b). As described in section 3., this is caused by the passage of a cyclonic eddy. The maximum northward current (i.e., the Kuroshio core) often exceeds 1 m s−1. The axis of the Kuroshio core tilts eastward with increasing depth. Except for June 2012, the zonal current is westward within the Kuroshio, on average about 0.3−0.6 of the intensity of the northward current. The prevailing direction of the Kuroshio is ∼120° counterclockwise from the east. The annual mean Kuroshio axis is located at 122.43°E, varying between 122.35°E and 122.52°E. The variation of the Kuroshio axis is surprisingly small, ±8 km, over an entire year. Figure 2Open in figure viewerPowerPoint Monthly mean (a) zonal velocity and (b) meridional velocity measured by the moored ADCP array between June 2012 and May 2013. The black curves are constant velocity contours at a 0.2 m s−1 interval. The white curves are 0.5 and 1.0 m s−1 velocity contours. Vertical tick marks at the bottom of Figure 2b label the mooring positions. Ratsimandresy and Pelegri [2005] demonstrate that the offshore shifting of the western boundary current axis with depth can be explained by the baroclinic structure of the density front. It is confirmed using geostrophic currents computed from an expendable bathythermograph (XBT) survey across a section of the Gulf Stream. However, western boundary currents may contain a significant ageostrophic component [Kuehl and Sheremet, 2009]. Therefore, a better confirmation is to use direct velocity measurements as we do here for the Kuroshio. We compare the baroclinic structure of the Kuroshio near its origin in depth and density coordinates. Density measurements taken during each of the 15 Seaglider transects along the moored array are used to transform the mooring velocity observations, averaged during the Seaglider transects, from the depth coordinate to the density coordinate. Indeed, the offshore shifting of the Kuroshio axis is reduced significantly in the density coordinate (Figure 3), in agreement with results reported by Ratsimandresy and Pelegri [2005]. Our results suggest that the offshore tilt of the Kuroshio axis with depth is caused primarily by its geostrophic component. We show (section 2.3.2.3) that the ageostrophic component of the Kuroshio is small except on the western side of the Kuroshio in the upper 150 m. Note that the baroclinic structure of the Kuroshio is better presented in the density coordinate on which internal wave heaving effects can be minimized. Figure 3Open in figure viewerPowerPoint Kuroshio baroclinic structure in depth and density coordinates. (a and c) Kuroshio meridional velocity baroclinic structure in depth coordinates measured by the moored array averaged over 19 November to 4 December 2012 and 1 January to 16 January 2013. (b and d) The corresponding baroclinic structure in density coordinates. Blue arrows illustrate the vertical alignment of the Kuroshio axis. Toole et al. [1990] report the relative vorticity of the Kuroshio across the 18.33°N section to be important in conserving the potential vorticity. Their relative vorticity ζ was estimated using geostrophic velocity and therefore constrained by the small Rossby number (Ro) assumption, i.e., ζ < f. We compute the relative vorticity, approximated by ζ ≈ ∂xv following Toole et al. [1990], using mooring velocity measurements, and therefore is not constrained by the small Ro assumption. Strong positive ζ exists on the western flank of the Kuroshio in the upper 150 m, often greater than 0.5f (Figure 4). It is sometimes greater than f in the upper 100 m, e.g., September, October, and December of 2012. The O(1) Rossby number suggests that the geostrophic assumption is inappropriate on the western flank of the Kuroshio. Negative ζ exists on the eastern flank of the Kuroshio, with a typical value of −0.2f in the upper 150 m, suggesting that a geostrophic balance might be appropriate in the vast portion of the Kuroshio especially deeper than 150 m east of the current's axis. Figure 4Open in figure viewerPowerPoint Monthly averaged relative vorticity ζ, normalized by the planetary vorticity f, approximated as the zonal gradient of the meridional velocity ∂xv measured by the moored ADCP array June 2012 to May 2013. The black curves are constant contours of ζ/f at 0.2 interval. The white curves are 0.5 and 1.0 ζ/f contours. Vertical ticks at the bottom of plots label the mooring positions. Kuroshio Transport To compute the total volume transport, the missing observations of velocity in the upper 45 m are filled using (1) linear extrapolation and (2) constant value of velocity measured at 45 m depth. The estimates of the northward transport computed from these two methods are in good agreement. We extrapolate velocity linearly from 450 m to 600 m. The maximum depth of the northward current at all six mooring positions increases from 440 m to 580 m toward the east. In this study, the Kuroshio transport is computed by integrating the northward current across the entire moored array from the surface to its maximum depth. The transports in the upper 45 m and below 450 m contribute ∼20% and 13%, respectively, to the total transport. Our mooring measurements miss a portion of the Kuroshio at its eastern limit. We use two independent methods to assess the possible missing Kuroshio transport from our mooring measurements: (1) comparing mooring observations with numerical solutions of a theoretical model of western boundary currents described by Kuehl and Sheremet [2009] and (2) comparing mooring observations with the geostrophic current computed from Seaglider measurements. Following Kuehl and Sheremet [2009], the Kuroshio is represented by the quasigeostrophic (QG) potential vorticity advection-diffusion equation [Kuehl and Sheremet, 2009] where ψ is the stream function, J is the Jacobian operator, β = ∂yf, and Ah is the turbulent lateral diffusivity. This QG equation describes the balance of the nonlinear advection of the relative vorticity, the planetary vorticity, and the lateral diffusion of the relative vorticity. Note that we drop the bottom diffusion term from the Kuehl and Sheremet [2009] QG equation and assume that the diffusion of the Kuroshio can be approximated by the lateral diffusion. Kuehl and Sheremet [2009] introduce the inertia length scale and the lateral diffusion length scale , where describes the strength of the NEC in the far field from the western boundary. The above QG equation is solved numerically using a simple centered finite difference approximation coupled with a time equilibration method also described by Kuehl and Sheremet [2009]. The structure of the western boundary current depends mainly on the ratio of LI/LM [Sheremet et al., 1997]. Mooring observations averaged in the upper 200 m and over the entire measurement period are compared with numerical solutions of the above QG equation for different values of Ah and different ratios of LI/LM. Model solutions with Ah = 800 m2 s−1 provide the best estimate of the Kuroshio axis location. Profiles of numerical solutions are fitted to the Kuroshio profile observed by the moored array (Figure 5). The solution for LI/LM = 0 reproduces the Munk [1950] model result, a balance between planetary vorticity and lateral diffusion. On the western flank of the Kuroshio, solutions for different values of LI/LM do not differ significantly. On the eastern flank of the Kuroshio, solutions for LI/LM > 1 have a long exponential tail, whereas a recirculation is present in solutions for LI/LM < 1. Previous observations [e.g., Zhang et al., 2001; Liang et al., 2003] and our Seaglider observations suggest that the Kuroshio has a width of 100−150 km. Solutions for LI/LM> 1 have a long exponential tail with a Kuroshio width greater than 200 km and cannot be a proper representation for the Kuroshio. Compared with solutions for LI/LM = 0, 0.5, and 1.0, mooring measurements miss 9%, 14%, and 20% of the Kuroshio transport, respectively. Figure 5Open in figure viewerPowerPoint Comparison of the Kuroshio meridional current averaged between June 2012 and May 2013 and averaged in the upper 200 m using measurements taken by the moored array (magenta curve with yellow dots) with the theoretical Kuroshio model described by a quasigeostrophic potential vorticity equation, with different ratios of LI/LM, where LI is the inertia length scale and LM is the diffusion length scale defined in section 2.3.2.3. The shading shows the 95% confidence level of the measured Kuroshio northward current. Seaglider measurements cover a zonal section wider than the moored array and therefore provide a better estimate of the spatial extent of the geostrophic component of the Kuroshio. Individual sections of the geostrophic current computed from Seaglider measurements might suffer from the spatial-temporal aliasing problems and therefore contain strong spatial variability. Averaged over all 15 sections, the geostrophic current from Seaglider measurements, averaged in the upper 200 m, agrees very well with that from the mooring measurements over the zonal span of the moored array (Figure 6) and suggests that mooring measurements miss 9% of the Kuroshio transport on the eastern flank of the Kuroshio. Compared to model results from the QG equation, the solution for LI/LM = 0.5 agrees the best with the geostrophic current estimated from Seaglider measurements. Figure 6Open in figure viewerPowerPoint Comparison of the Seaglider geostrophic velocity, mooring observed velocity, and theoretical Kuroshio model. The thick black curve and shading shows the mean and 95% confidence interval of Seaglider derived geostrophic velocity averaged over the 15 Seaglider transects along the moored array and averaged over the upper 200 m. Yellow dots and vertical red lines represent the mooring observed meridional current averaged over the same period as Seaglider transects and averaged in the upper 200 m. The thinner colored curves represent the solutions of a theoretical Kuroshio model described by a quasigeostrophic potential vorticity equation with different ratios of LI/LM, as explained in Figure 5. On the western limit of the Kuroshio, the current may not be in geostrophic balance because of the large Rossby number and the close proximity to land. Therefore, on the western limit of the Kuroshio the geostrophic current computed by Seaglider measurements might not be representative for the Kuroshio. Combining both comparative analyses, we conclude that our mooring measurements miss less than 15% of the annual mean of the total Kuroshio transport. To estimate the temporal variation of the missing Kuroshio transport in our mooring measurements, we linearly extrapolate monthly averaged mooring velocity measurements and define the Kuroshio boundaries where the northward current vanishes. The missing northward Kuroshio transport varies from 26% in July 2012 to 4% in November 2012, with a mean of 14% and a standard deviation of 6%. Because of the abundance of eddies east of the Kuroshio, it is a nontrivial task to define Kuroshio's boundary at time scales shorter than O(month). In the following analysis, the Kuroshio transport is referred to that computed directly from mooring measurements, without correcting for the possible missing transport on the eastern and western limits. The annual average northward transport of the Kuroshio is 15 Sv with a standard deviation of 3 Sv (Figure 7c; red curve). This annual mean transport is close to that estimated by Qu et al. [1998], but significantly smaller than that estimated by Nitani [1972] and Yaremchuk and Qu [2004]. Most significantly, rapid changes of Kuroshio transport greater than 10 Sv were observed. For example, between 24 June and 4 July 2012, the Kuroshio transport increased from 7 Sv to 22 Sv. Similar transport variations of greater than 10 Sv on O(10 days) time scales were detected during May and June 2013, which is much faster than the O(100) day variations observed by Zhang et al. [2001] and Lee et al. [2013]. Figure 7Open in figure viewerPowerPoint (a) Sea level anomaly (SLA) and (b) its zonal gradient along 18°45′N. The magenta vertical line indicates the longitude and observation period of the ADCP moored array. (c) Zonal gradient of the SLA across the moored array between 122°E and 123°E (gray shading), and the Kuroshio transport measured by the moored ADCP array (red curve). Eight events with a transport anomaly exceeding 5 Sv are labeled in Figure 7c. We fit the observed velocity averaged in the upper 200 m to the analytical form of the Munk [1950] solution, equivalent to the QG solution for LI/LM = 0, and determine the Kuroshio maximum current axis. Note that the Munk solution has a similar shape to that of the QG solution for LI/LM = 0.5, which shows the best agreement with our observations (Figure 5). The position of the Kuroshio axis in the upper 200 m varies only ±8 km about 122.43°E. Therefore, we believe that the observed rapid changes in transport are not likely due to the meandering of the Kuroshio off the moored array. In fact, the observed rapid transport anomaly is due to the modulation by mesoscale eddies. 3. Mesoscale Eddy Effects on Kuroshio Transport Westward propagation of sea level anomalies (SLAs) in the western Pacific is observed by satellite altimeters [e.g., Roemmich and Gilson, 2001; Chelton et al., 2011]. The SLA and the zonal SLA slope at the latitude of our mooring array (18.75°N) show clear westward propagations with a speed of ∼0.14 m s−1 (Figures 7a and 7b). The altimeter products were produced by Ssalto/Duacs and distributed by Aviso, with support from CNES (http://www.aviso.oceanobs.com/duacs/). The westward propagations stall east of 123°E, i.e., east of t