Investigation of the island-induced ocean vortex train of the Kuroshio Current using satellite imagery

Remote Sensing of Environment 193 (2017) 54–64

Contents lists available at ScienceDirect

Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse

Investigation of the island-induced ocean vortex train of the Kuroshio Current using satellite imagery Po-Chun Hsu a, Ming-Huei Chang b, Chen-Chih Lin a, Shih-Jen Huang a, Chung-Ru Ho a,⁎ a b

Department of Marine Environmental Informatics, National Taiwan Ocean University, Keelung, Taiwan Institute of Oceanography, National Taiwan University, Taipei, Taiwan

a r t i c l e

i n f o

Article history: Received 25 April 2016 Received in revised form 23 February 2017 Accepted 28 February 2017 Available online xxxx Keywords: Ocean vortex Kuroshio Current Satellite imagery Green Island

a b s t r a c t In this study, we used satellite imagery to conduct a statistical study of the ocean vortex train induced by the Kuroshio Current on the leeward side of Green Island, Taiwan. The spatial scale and characteristics of the ocean vortex train were analyzed using image datasets from five different high-resolution satellites, including optical imagery from the Satellite Pour l'Observation de la Terre and the Formosat-2 satellites and synthetic aperture radar imagery from the European Remote Sensing Satellite, Advanced Land Observing Satellite, and Sentinel-1. Satellite altimetry data and a moored acoustic Doppler current profiler (ADCP) were used to calculate the velocity of the Kuroshio Current. The ADCP data suggest that the velocity increases on the western side of the vortex train when it is formed on the leeward side of Green Island. Data from the moderate-resolution imaging spectroradiometer (MODIS) showed that the sea surface temperature of the recirculation water was over 2 °C colder and the chlorophyll-a concentration was two times higher than that of the surrounding waters. These phenomena suggested upwelling, mixing processes, and an island-mass effect. Wind forcing had a pronounced effect on the characteristics of the vortex train. High-resolution satellite images indicate that the averaged aspect ratio of the vortex train is 2.09 and the dimensionless width is 2.02 under southerly winds, compared to 1.91 and 2.76, respectively, under northerly winds. © 2017 Published by Elsevier Inc.

1. Introduction The flow pattern in the vortex of an obstacle depends on the Reynolds number (Re) of the flow. Periodic oscillation of the vortex occurs when the Re value is between 40 and 70. Above a limiting value of Re b 60–90, shedding of standing eddies behind the object occurs. This flow pattern is referred to as a von Kármán vortex street (Pattiaratchi et al., 1987). Vortex streets occur frequently in the atmosphere (Li et al., 2000; Nunalee and Basu, 2014) and ocean (Barton et al., 2000; Young and Zawislak, 2006; Dong et al., 2007; Zheng et al., 2008, 2012; Teinturier et al., 2010; Topouzelis and Kitsiou, 2015). Ocean vortices and island-induced ocean vortex trains (IOVTs) are types of smallscale or mesoscale ocean phenomena, which often enhance biological productivity via upwelling and turbulent mixing and are closely associated with fishing activities (Caldeira et al., 2002, 2005; Hasegawa et al., 2004, 2009). The Kuroshio Current, a western boundary current in the North Pacific, originates from the North Equatorial Current and flows northward from east of the Philippines to south of Japan. This current could provide ⁎ Corresponding author at: Department of Marine Environmental Informatics, National Taiwan Ocean University, 2 Pei-Ning Road, Keelung 202, Taiwan. E-mail address: [email protected] (C.-R. Ho).

http://dx.doi.org/10.1016/j.rse.2017.02.025 0034-4257/© 2017 Published by Elsevier Inc.

a source of energy because it is rapid and steady as it passes the east coast of Taiwan. For a potential Kuroshio power test site near Green Island, located at 22°35′N 121°28′E, 40 km off the southeastern coast of Taiwan, several factors have to be considered, such as the Kuroshio Current velocity, island wakes, and the seabed topography. According to insitu measurements and numerical model simulations, it has been clearly shown that a high potential power density distribution is located in the northwestern area of Green Island, which is close to the influence area of the island vortex street (Hsu et al., 2015b). Because the vortex street has an adverse effect on the stability and power efficiency of turbine generators and the related anchoring platform, an understanding of the origins of the Green Island vortex is therefore an important research topic for energy exploitation of the Kuroshio Current. Ship surveys and satellite altimetry data have reported that the Kuroshio Current east of Taiwan has a mean maximum velocity of ~1.2 m/s (Yang et al., 2015) and an average velocity of ~0.76 m/s (Hsu et al., 2016). The incoming Kuroshio Current forms an ocean vortex and IOVT behind Green Island (Chang et al., 2013; Liang et al., 2013; Huang et al., 2014; Zheng and Zheng, 2014; Hsu et al., 2015a). When the Kuroshio Current strikes Green Island directly, a recirculation develops leeward of the island, followed by a wavy tail resembling a weak vortex street and a cold eddy that likely originates downstream of Green Island. The IOVT may be affected by the Kuroshio Current

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

velocity; it has been simulated via numerical modeling (Huang et al., 2014) and measured by in situ data (Chang et al., 2013). Besides, the seasonal variation of Green Island IOVT affected by different wind directions has been mentioned by Hsu et al. (2015a) using a numerical model. They found that different wind directions may change the spatial scales of the vortex area. The seasonality and size of the IOVT have to be considered, because changes in the spatial scales of IOVT may affect the consideration on site selection of the Kuroshio power generators. However, numerical simulations and a one-time survey cannot provide a synoptic view of the IOVT. To have a synoptic understanding the seasonality and spatial scales of the IOVT and its relationship with the Kuroshio Current velocity, high-spatial resolution satellite images and geostrophic velocities derived from satellite altimeter data are applied to conduct a statistical study of the Green Island vortices. Satellite imagery from 2001 to 2015 and data from two ADCPs moored on the western side and in the lee of Green Island are used to investigate the spatial characteristics and seasonal variability of the vortex and IOVT. In this paper, we also discuss the atmospheric wind forcing which could affect the spatial configuration of an ocean vortex train on a short-time scale. The analysis of the spatial characteristic, variations of IOVT and wake extension distance by satellite imagery, and Kuroshio flow speed observed by ADCPs moored data will provide quite reference values for numerical simulation of the Green Island wake in the future.

55

Table 1 Summary of high-resolution satellite images used in this study. Satellite

Sensor

SPOT 1–4 SPOT-5 SPOT-6 & 7 Formosat-2 ERS-2 ALOS-1 & 2 Sentinel-1

HRV (High Resolution Visible) HRG (High Resolution Geometrical) NAOMI (New AstroSat Optical Modular Instrument) Panchromatic sensor and Multispectral sensor C-SAR (C-band Synthetic Aperture Radar) PALSAR (Phased Array type L-band Synthetic Aperture Radar) C-SAR (C-band Synthetic Aperture Radar)

Space Organization, Taiwan. It has a spatial resolution of 2 m for panchromatic and 8 m for multispectral imagery. The ERS-2 satellite was placed into orbit in 1995 and retired in 2011. This satellite was equipped with SAR with a spatial resolution of 20 m and a swath of 100 km at the C-band with VV polarization. Sentinel-1, which was launched in 2014, provided continuity following the retirement of ERS-2 at the end of the Envisat mission. Sentinel-1 is a two-satellite constellation with the prime objective of monitoring the land and ocean using a C-band SAR to provide continuous SAR images. The ALOS1 satellite was launched in 2006 and retired in 2011 and was followed by the ALOS-2 satellite in 2014; both were initiated by the Japan Aerospace Exploration Agency (JAXA). The ALOS series satellite payload included Phased Array-type L-band Synthetic Aperture Radar (PALSAR) for day-and-night and all-weather land observations.

2. Data 2.2. Satellite altimetry data 2.1. High-resolution satellite imagery Green Island is located at 22°35′N 121°28′E, 40 km off the southeastern coast of Taiwan. The water depth around the island varies from 50 m to 4000 m (Fig. 1). In this area, ocean vortices are formed on the leeward side of Green Island by the incoming bathymetry-deflected Kuroshio Current. The resulting vortex shedding, upwelling of colder water, and vortex trains are accompanied by higher bio-productivity. To examine the spatial scales of the ocean vortices and IOVT, image datasets from five different high-resolution satellites, including panchromatic images from Satellite Pour l'Observation de la Terre (SPOT) and the Formosat-2 satellite and SAR images from the European Remote Sensing Satellite (ERS-2), Advanced Land Observing Satellite (ALOS), and Sentinel-1 (Table 1) are used. The SPOT and Formosat-2 satellite data are in the form of panchromatic images. The first SPOT satellite (SPOT-1) was launched in 1986, with a 10-m panchromatic and 20-m multispectral imagery resolution capability. The latest (SPOT-7) was launched in 2014, with a 1.5-m panchromatic and 6-m multispectral spatial resolution. The SPOT series of satellites were initiated by the Centre national d'études spatiales (CNES). Formosat-2 was launched on May 20, 2004, by the National

Daily Absolute Dynamic Topography (ADT) data with 1/4°×1/4° grid spatial resolution provided by Archiving Validation and Interpretation of Satellite Data in Oceanography (AVISO) are employed in this study. ADT is deduced from the Sea Level Anomaly (SLA) using a Mean Dynamic Topography (MDT), i.e., ADT = SLA + MDT, where MDT is mean sea surface minus the geoid (SSALTO/DUACS User Handbook, 2016). 2.3. Moderate-resolution imaging spectroradiometer data The moderate-resolution imaging spectroradiometer (MODIS) is an instrument carried by the Terra and the Aqua satellites. MODIS sea surface temperature (SST) and chlorophyll-a (Chl-a) images with 250-m spatial resolutions were used to show the features of the ocean vortices and the vortex trains leeward of Green Island. The images were processed from MODIS Level 0 data for Level 2 data using the SeaDAS program version 7.3, which created a 250-m resolution image with replication for the 500-m and 1-km resolution bands and resampling using the nearest method from the swath to the Mercator projection grids. The Level 0 data were downloaded from NASA's OceanColor

Fig. 1. Locations of Green Island and two ADCP stations, shown as white dots.

56

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

Fig. 2. Kuroshio-induced ocean vortex on the leeward side of Green Island as shown on (a) a SPOT-2 color infrared image taken at 02:42 UTC on July 1, 2005. (b) and (c) are MODIS SST and Chl-a images, respectively, taken at 04:40 UTC on July 1, 2005. The SPOT image was provided by CNES/CSRSR. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Web, which is supported by the Ocean Biology Processing Group at NASA's Goddard Space Flight Center. 2.4. Wind data The wind data were obtained from the Quick Scatterometer (QuikScat) and the first Advanced Scatterometer (ASCAT) and were provided by Remote Sensing Systems (RSS). Both datasets are comprised of three-day time-averaged gridded data obtained by mapping the scatterometer orbital data to a 0.25° longitude by 0.25° latitude Earth grid. QuikScat is a SeaWinds instrument that was placed in orbit in June 1999 and operated until November 2009. ASCAT was launched on the EUMETSAT MetOp-A satellite in October 2006 and continues to operate today. The surface wind speed is equivalent to the neutral wind speed 10 m above the water surface, is derived from the surface roughness, and is roughly equivalent to an 8–10-min mean surface wind. The angle of the wind movement follows the oceanographic convention that a value of zero indicates wind movement toward the north, while a value of 90 indicates wind movement toward the east. In this study, we used QuikScat data for January 2001–May 2007 and ASCAT data for June 2007–December 2015.

streaming speeds. The mean velocities are computed during periods when the vortices are shown and not shown on available satellite images to understand how much the velocity increases when the vortices appear. Finally, the aspect ratio (Ar = a / b, the ratio of the along-street distance a and the cross-street distance between two vortices b) and the dimensionless width (Br = b / L, the ratio of b and the island diameter L) of the vortex train (Young and Zawislak, 2006) under different wind velocities and incoming flow velocities are discussed. 4. Results 4.1. Characteristics of the Kuroshio-induced vortices and IOVT Ocean vortices and IOVT are formed on the leeward side of Green Island by the incoming bathymetry-deflected Kuroshio Current and can be observed in satellite images. Before analyzing their characteristics, it is important to understand the dynamic regime of the Green Island vortices. Focusing on the Kuroshio-induced vortices, we extracted the

2.5. ADCP data We used two moored ADCP sets northwest (NW) of Green Island (the white dots in Fig. 1), one with a 307.2-kHz system frequency, with 50 bins 4 m in length and a total recorded-profile range of 200 m, and another with a 76.8-kHz system frequency, with 68 bins 8 m in length and a total recorded-profile range of 550 m. The 307.2kHz ADCP set was moored at 22.70°N 121.45°E from September 12 to December 23, 2014, and the 76.8-kHz ADCP set was moored at 22.70°N 121.42°E from December 22, 2014, to August 6, 2015. The ADCP provided profiles of the velocity of the Kuroshio Current and the streaming of the ocean vortex on the western side of Green Island. 3. Methodology The methodology of this study is described as follows. Firstly, the island-vortex parameters and Reynolds number are used to describe the periodic oscillation of the vortex train, followed by the relationship between the angle of the incoming flow derived from the satellite altimetry and that of the Green Island vortex derived from the high-resolution satellite imagery are discussed. Thirdly, profiles of the flow velocity from the two ADCP moored data and the satellite images acquired during the observation periods of the ADCP are used to analyze the vortex

Fig. 3. Relationship between the angles of incoming flow at 22.625°N 121.375°E, southwest of Green Island, and the generated vortex on the leeward side of Green Island.

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

Re and island-vortex parameters, both of which are dimensionless numbers that can be used to describe a vortex. The island–vortex parameter (P) is used to describe the formation of vortices in shallow coastal waters and the characteristics of a vortex formed behind an island (Wolanski et al., 1984). This parameter is given by   P ¼ UH 2 =ðK z LÞ;

ð1Þ

where U is the velocity scale, H is the water depth of an upper layer (72 m), Kz is the vertical diffusion coefficient, and L is the horizontal length. We used a velocity of U = 1.14 m/s, obtained from the ADCP

57

stations at 22.70°N 121.42°E northwest of Green Island, and an island width scale of L = 7 km and set Kz = 0.1 m2/s (Pattiaratchi et al., 1987). The estimate of P ≫ 1 meant that the friction would be negligible and that the vortex would be similar to that formed at a high Re (i.e., eddy shedding). The vortex formed behind an obstacle at various Re values is given by Re ¼ UL=γ;

ð2Þ

where γ is the horizontal eddy viscosity for oceanic flows. A value of γ = 100 m2/s was used in this study, as suggested by Pattiaratchi et al. (1987), Heywood et al. (1990), and Chang et al. (2013). The

Fig. 4. (a) Profiles of flow speed at 22.70°N 121.45°E on the leeward side of Green Island obtained from the 307.2-kHz moored-ADCP station from September 13 to December 23, 2014. The dot and cross symbols indicate if the vortex is visible or not, respectively, on the satellite images. (b) Profiles of the flow velocity. (c) and (d) are the EKEs at depths of 22 m and 50 m, respectively, and (e) is the vertical shear calculated over the upper 100 m.

58

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

estimated Re = 80 meant that periodic oscillation of the vortex would take place and that shedding of standing eddies behind the object would occur (Pattiaratchi et al., 1987). Island vortices can be current induced or wind driven. If the following three facts are true, the wind-driven condition can be dismissed in our study area: (1) the winds in the Green Island field are weak, (2) the Green Island vortices appear only northeast or north of the island (Chang et al., 2013), and (3) a wind-induced vortex is usually accompanied by a higher SST, due to weaker vertical mixing resulting from sheltering from the wind on the leeward side of the island (Arístegui et al., 1994; Barton et al., 1998; Caldeira et al., 2005; Zheng and Zheng, 2014). Fig. 2a shows an example of a Kuroshio-induced ocean vortex on the leeward side of Green Island observed by SPOT-2 with 20-m spatial resolution at 02:42 UTC on July 1, 2005. The vortex was observed over a distance of N21 km and an area of approximately 150 km2 on the leeward side of Green Island. MODIS imagery has demonstrated that it can provide high-spatial resolution of ocean-surface features (de Souza et al., 2006; Alpers et al., 2013) and phytoplankton blooms (Peñaflor et al., 2007; Isoguchi et al., 2009; Shang et al., 2012). The SST data from MODIS showed the features of the Kuroshio-induced vortex, in which the temperature on the leeward side of Green Island was 2.4 °C lower than that of the surrounding waters (Fig. 2b). A current-induced vortex is associated with upwelling or vertical mixing, which causes cooler water to rise (Caldeira et al., 2005; Chang et al., 2013), and this increases the amount of high-salinity water in the Kuroshio region (Chang et al., 2013). It was also observed that the Chl-a concentration on the leeward side of Green Island was 0.07 mg/m3 higher than that in the surrounding water (Fig. 2c). The MODIS images show the

full spatial area of the vortex, which formed over a distance of approximately 30 km. Chang et al. (2013) obtained measurements using shipboard ADCP on September 15, 2010, and determined that the recirculating water was 1 °C–2 °C colder in SST and 0.02–0.04 mg/m3 higher in Chl-a concentration than the surrounding waters. The upwelling and vertical mixing in this area were apparently weaker than those recorded on July 1, 2005. The angle of the vortex formed and its areal extent may be affected by the angle and speed of the incoming flow. Examples of the consistency of the flow angle between the incoming flow and the Green Island vortex are given in Fig. 3 of Chang et al. (2013) based on shipboard ADCP current data. To obtain the relationship between the angle of the incoming flow and that of the Green Island vortex, we calculated the angle of the Green Island vortex from high-resolution satellite images and the angle of incoming geostrophic velocity from gridded satellite altimeter data provided by AVISO at 22.625°N 121.375°E, southwest (SW) of Green Island. Fig. 3 shows the relationship between the directions of incoming flow and the generating vortex on the leeward side of Green Island. Linear regression indicates a correlation coefficient of 0.65, with a p-value of b 0.01. The AVISO dataset has low spatial resolution (0.25° × 0.25°) and cannot be used to obtain the vortex streaming on both sides of the island and the subsequent confluence on the leeward side of Green Island. According to Lien et al. (2015), the surface velocity processed by AVISO northeast of Luzon Island in the Kuroshio region has a magnitude 1.6– 2.0 times lower than the velocity measured by sea gliders, moorings, and drifters. Therefore, we used the ADCP moored at 22.70°N 121.45°E on the leeward side of Green Island to calculate the flow velocity.

Fig. 5. Images of Kuroshio-induced ocean vortices and IOVT of Green Island. (a) SPOT-4 panchromatic image taken at 02:04 UTC on September 13, 2014, (b) Sentinel-1 SAR image taken on November 5, 2014, (c) SPOT-5 color infrared image taken at 01:54 UTC on November 25, 2014, and (d)–(g) are Sentinel-1 SAR images taken on November 27, November 29, December 21, and December 23, 2014, respectively. The images were provided by CNES/CSRSR and ESA/CSRSR. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

Fig. 4a and b show profiles of the flow speed and direction, respectively, from the 307.2-kHz moored ADCP station for the period of September 13–December 23, 2014. The dot and cross symbols in Fig. 4a indicate if the vortex is visible or not, respectively, on the satellite images. A higher eddy kinetic energy (EKE) is found when a vortex occurred (Fig. 4c and d). As a vortex forms, the strong vertical shear of the currents drives mixing, which is associated with intense overturning and upwelling in the vortex, as shown in Fig. 4e. The average speeds of the current on the west side of Green Island when a vortex is visible on satellite images are 1.22 ± 0.37 m/s and 1.24 ± 0.25 m/s at water depths of 22 m and 50 m, respectively, while the average speeds when a vortex is not visible on satellite images are 0.59 ± 0.24 m/s and 0.80 ± 0.22 m/s at water depths of 22 m and 50 m, respectively. This implies that the island vortices occur during periods with stronger currents. Seven available highresolution satellite images taken during the time span of the ADCP measurements show the ocean vortex on the leeward side of Green Island (Fig. 5). The average speeds from the ADCP measurements at the time of the seven images are 1.14 m/s at 22 m and 1.30 m/s at 50 m. Chang et al. (2013) indicated that the incoming surface current velocity would increase the speed of streaming on both sides of Green Island from 1.2 m/s to 1.6 m/s. A similar structure is also reproduced by the numerical model of Zheng and Zheng (2014). Fig. 6 presents the profiles of the flow velocity from the other moored ADCP, the 76.8-kHz station, in the period of December 23, 2014–July 27, 2015. When a vortex is visible on the satellite images, the average current speed on the west side of Green Island is 1.28 ± 0.22 m/s. It is 0.85 ± 0.23 m/s when a vortex is not visible on the available satellite images. The ocean vortex on the leeward side of Green Island is not only visible on high-resolution optical and SAR satellite images but also on MODIS images. Fig. 7 demonstrated four MODIS images clearly showing a vortex on the leeward side of Green Island. The

59

recirculation covers an area approximately 2–3 times that of Green Island. Fig. 8 shows a schematic diagram of the incoming Kuroshio Current as it passes a circular island of diameter L and a von Kármán vortex street. Ar and Br are typical spatial dimensions of vortices and can be calculated from the along-street distance and the cross-street distance between two vortices, as shown in Fig. 8a. The high-resolution images in Fig. 8b–d are examples of IOVT on the leeward side of Green Island. Fig. 8b is a PALSAR image taken by ALOS at 02:15 UTC on June 4, 2008. The “a” and “b” values of the vortices were 22.1–23.4 km and 10.6– 10.7 km, respectively, the Ar was 2.14, and the Br was 2.13. Fig. 8c is a color infrared image from SPOT-5 taken at 02:29 UTC on August 18, 2011. The “a” and “b” values of the vortices were 16.3–24.2 km and 10.2–10.3 km, respectively, the Ar was 1.98, and the Br was 2.05. Fig. 8d is an SAR image from Sentinel-1 on May 26, 2015, with “a” and “b” values of 20.8–31.7 km and 13.0–13.2 km, respectively, the Ar was 2.00, and the Br was 2.62. Fig. 9 shows an IOVT forming on the leeward side of Green Island and includes details of the ocean surface temperature and Chl-a. Fig. 9a demonstrates the IOVT downstream of Green Island, taken by SPOT-7 with 1.5-m spatial resolution at 02:17 UTC on June 4, 2015. The vortex was visible over a distance of N 45 km downstream of Green Island. The SST data show the unique features of the Kuroshio-induced ocean vortex train (white circle in Fig. 9b), which on the downstream side of Green Island had a temperature up to 3.3 °C lower than that of the surrounding water (Fig. 9b). The vortex trains generated by upwelling or vertical mixing can be seen in the MODIS Chl-a image (Fig. 9c). The Chl-a concentrations in the four vortices of the IOVT were 0.24 mg/m3 , 0.22 mg/m 3 , 0.22 mg/m3, and 0.16 mg/m3 or nearly twice the values in the surrounding water (0.10 mg/m3).

Fig. 6. (a) Profiles of flow speed at 22.70°N 121.42°E obtained from the 76.8-kHz moored-ADCP station from December 23, 2014, to July 27, 2015. The dot and cross symbols represent if the vortex is visible or not, respectively, on the satellite images. (b) Profiles of the flow velocity.

60

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

Fig. 7. MODIS Aqua images of Kuroshio-induced ocean vortices and IOVT of Green Island at (a) 05:10 UTC on April 4, 2015, (b) 05:05 UTC on June 16, 2015, (c) 05:15 UTC on June 30, 2015, and (d) 05:05 UTC on July 2, 2015.

Table 2 summarizes the characteristics of the IOVT recorded in this study and in the two numerical simulations by Huang et al. (2014) and Hsu et al. (2015a) for comparison, and Table 3 summaries the information of images with the 52 cases of IOVT. Both numerical models used

shallow-water equations to describe the horizontal structures of the ocean dynamics with the same entering angle of the current in the computational domain but a different incoming flow speed, length for Green Island, eddy viscosity, and wind condition. The results from this study

Fig. 8. (a) Schematic diagram of the Von Kármán vortex street for the incoming Kuroshio flows and IOVT images taken by satellites: (b) the ALOS image, (c) the SPOT-5 color infrared image, and (d) the Sentinel-1 image. The satellite images were provided by CNES/CSRSR and ESA/CSRSR. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

61

Fig. 9. Interpretation of the Green Island vortex trains. (a) SPOT-7 image taken on June 4, 2015. (b) and (c) are MODIS SST and Chl-a images, respectively, taken at 05:20 UTC on June 5, 2015; the block dotted lines indicate the route of the IOVT.

show an IOVT stream-wise length of 24.1 ± 8.65 km and a transverse length of 12.51 ± 3.96 km. The IOVT Ar from the satellite images was 1.98 ± 0.49 and the Br was 2.50 ± 0.79, as shown in Table 2. In the 52 cases, the average wind movement at 22.875°N 121.625°E was 6.8 m/s and the average incoming flow velocity of the IOVTs calculated from the AVISO dataset was 0.61 m/s. The comparison between the numerical model results and the satellite observations highlights the differences in these two approaches to this topic. The horizontal eddy viscosity γ for oceanic flow can vary from 102 m2/s to 105 m2/s (Apel, 1987). A value of 100 m2/s was adopted by Pattiaratchi et al. (1987), Heywood et al. (1990), and Chang et al. (2013). The eddy viscosity value set of Huang et al. (2014) appears to be too small, and it may cause the transversal length (b) to be shortened. In the result provided by Hsu et al. (2015a), the streamlines and vortex train on the leeward side of Green Island are notably farther downstream and the streamwise length (a) and transversal length are two times higher than our results. Two possible reasons may explain this: (1) the incoming flow speed value (U = 1.25 m/s) is larger than the average speed, therefore the result may be biased toward extreme values; and (2) the

Table 2 Characteristics of Green Island vortex trains. References

Data

a

b

Ar

Br

This study

Satellite image Satellite image Satellite image Numerical model

24.19

12.51

1.98

2.50

Huang et al. (2014)

Hsu et al. (2015a)

Conditions

52 cases from 2002 to 2015 27.7 10.2 2.72 1.84 The case on 29 March 1999 17.9 10.2 1.76 1.92 The case on 30 July 2003 22.58 6.45 3.50 1.29 Re = 100, no winds 24.40 6.65 3.67 1.33 Re = 200, no winds 24.37 6.55 3.72 1.31 Re = 500, no winds U = 1.00 m/s, L = 5 km, eddy viscosity = 10, 25, and 50 m2/s Numerical 57.22 21.84 2.62 2.73 Re = 100, no winds model 55.69 24.32 2.29 3.04 Re = 100, NE winds = 12 m/s 57.66 21.84 2.64 2.73 Re = 100, SW winds = 10 m/s U = 1.25 m/s, L = 8 km, eddy viscosity = 100 m2/s

island width scale (L = 8 km) is too large, 8 km is the width of Green Island 200 m below the sea surface. This value is 1.6 times greater than the island width at the surface. In summary, the estimated results of the dimensional parameters differ between the satellite images and the models, and the model parameters may need to take into account additional details of the incoming flow speed and eddy viscosity.

4.2. Wind effect East of Taiwan, the Kuroshio Current velocity and the monsoon direction are different in the four seasons. The characteristics of the

Table 3 Summary of images with the 52 cases of IOVT. Date (YYYY/MM/DD)

Satellite/sensor

Date (YYYY/MM/DD)

Satellite/sensor

2002/04/22 2002/05/27 2002/10/14 2003/04/07 2003/07/21 2004/07/05 2004/11/22 2005/07/25 2006/02/20 2006/03/27 2006/04/02 2006/05/01 2006/07/10 2007/02/05 2007/03/12 2007/05/12 2007/05/21 2007/06/29 2007/07/20 2007/07/30 2007/08/29 2008/02/25 2008/03/31 2008/04/19 2008/05/05 2008/06/04

ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR SPOT-2/HRV ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR SPOT-4/HRV ERS-2/C-SAR SPOT-2/HRV SPOT-2/HRV ERS-2/C-SAR SPOT-4/HRV ERS-2/C-SAR ERS-2/C-SAR ALOS/PALSAR ERS-2/C-SAR ALOS/PALSAR

2008/06/09 2008/07/14 2008/08/18 2008/10/20 2008/10/27 2009/01/05 2009/05/25 2009/06/24 2009/08/03 2009/09/07 2010/05/10 2010/08/23 2011/07/05 2011/08/17 2011/08/18 2013/08/03 2014/11/27 2015/03/29 2015/04/22 2015/05/02 2015/05/26 2015/06/04 2015/07/01 2015/08/30 2015/09/13 2015/10/05

ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ALOS/PALSAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR SPOT-5 HRG ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR ERS-2/C-SAR SPOT-4/HRV SPOT-5/HRG SPOT-5/HRG SPOT-6/NAOMI Sentinel-1/C-SAR Sentinel-1/C-SAR Sentinel-1/C-SAR Sentinel-1/C-SAR Sentinel-1/C-SAR SPOT-7/NAOMI Sentinel-1/C-SAR ALOS-2/PALSAR Sentinel-1/C-SAR Sentinel-1/C-SAR

62

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

Table 4 Spatial characteristics of the Green Island vortex trains in the four seasons. Fall (Sep–Nov)

Table 6 Characteristics of Green Island vortex trains under different conditions of winds and incoming flow velocities.

Spring (Mar–May)

Summer (Jun–Aug)

Winter (Dec–Feb)

Ar

Br

Ar

Br

Ar

Br

Ar

Br

1.95

2.52

2.04

2.38

1.87

2.56

2.08

3.06

Incoming flow velocity

Table 5 The percentage of wind direction and average wind speed on the leeward side of Green Island in the four seasons.

Southerly wind (%) Average wind speed (m/s) Northerly wind (%) Average wind speed (m/s)

Spring (Mar–May)

Summer (Jun–Aug)

Fall (Sep–Nov)

Winter (Dec–Feb)

36.1 6.5 ± 2.0 63.9 8.1 ± 2.4

67.6 7.4 ± 3.3 32.4 6.3 ± 3.0

12.8 6.6 ± 4.1 87.2 9.2 ± 3.1

11.8 6.7 ± 2.2 88.2 9.9 ± 2.5

Northerly winds ≤7 m/s

Ar b0.5 m/s 0.5–0.75 m/s N0.75 m/s

Green Island vortex trains obtained from satellite imagery during the four seasons are given in Table 4. The averaged transversal length of the vortices in winter is approximately 3 km longer than that in summer. Even though there is a northerly in winter and a southerly in summer over eastern Taiwan, different wind directions may occur during its monsoon season (Table 5). Therefore, we consider the influence of the southerly and northerly winds on the vortex spatial scale in addition to the seasons. This is a more rigorous analysis and avoids misunderstandings in the seasonal analysis. The wind direction, wind speed, and incoming flow velocity at the date of each satellite image are shown in Table 6. We divided the wind direction and wind speed values into four categories and the incoming flow velocity into three categories. For strong wind speed (N 7 m/s), the spatial scale of the IOVT in the southerly wind case has a different state for the different incoming flows; the streamlines on the leeward side of Green Island appear to be pushed farther downstream with higher Ar. The Br decreased as the incoming flow velocity increased. In the case of northerly winds, however, the Kuroshio Current flows against the wind and the Br increases as the incoming flow velocity increases. Let the southerly/northerly wind speed value be positive/ negative, the correlation coefficients between Ar and wind speeds, as well as Br and wind speeds are 0.554 and −0.552, respectively during the strong wind cases (N7 m/s). The p values for both correlation calculation are b 0.01 that were considered to be statistically significant. This indicates that stronger southerly/northerly winds may cause higher/ lower Ar value and lower/higher Br value. The averaged Ar and Br are 2.66 and 1.72, respectively under strong southerly winds, compared to 1.94 and 2.75, respectively under strong northerly winds. Based on the numerical model results of Hsu et al. (2015a), the speed contours of the IOVT are packed close to the island and pushed further downstream in the southerly monsoon, because the flow direction of the Kuroshio Current is the same as that of the monsoon. The characteristics of the IOVT from our results are similar to those from numerical models: the Ar with southerly winds is larger than that with northerly winds and the Br with southerly winds is smaller than that with northerly winds. However, the real effects are at a much lower magnitude, presumably because the incoming flow velocity and eddy viscosity used in the numerical models are higher than those observed under real conditions and in the numerical model the northerly winds (12 m/s) and southerly winds (10 m/s) are too high. Only 15% of the northerly winds are N12 m/s, and 12% of the southerly winds are N10 m/s during the period of 2001–2015.

Southerly winds ≤7 m/s

N7 m/s Br

Ar

Br

1.54 2.30 2.49 1.90 2.05 2.18 2.77 1.59 1.69 2.00 No data Ar = 2.09 ± 0.54, Br = 2.02 ± 0.76

Ar

N7 m/s Br

Ar

No data 2.09 1.87 2.89 1.78 1.87 2.39 1.57 Ar = 1.91 ± 0.45, Br = 2.76 ± 0.77

Br 2.64 2.88 2.90

5. Discussion The characteristics of ocean vortex trains induced by an interaction between the Kuroshio Current and Green Island were observed using satellite imagery. When an island wake occurred, the averaged reference incoming speed derived from satellite altimetry was 0.56 ± 0.14 m/s (Re = 40–50, L = 7 km, γ = 100 m2/s), and it was 0.35 ± 0.14 m/s when the island wake did not occur (Re = 15–25). However, the averaged near-surface velocity measured by available Argo drifters around Green Island is 1.00 m/s and the surface geostrophic velocity provided by AVISO near Green Island is 0.45 m/s for 28 cases. According to Lien et al. (2015), the surface velocity recorded by the AVISO dataset northeast of Luzon Island in the Kuroshio region has a magnitude 1.6– 2.0 times lower than that measured by sea gliders, moorings, and drifters. Therefore, a value 1.8 times the geostrophic velocity from AVISO is more reasonable for the real current velocity. Two moored ADCP stations were set on the western side of Green Island. This was the first time long-term data on the boundary of the vortex was provided. Chang et al. (2013) used shipboard ADCP and underway surface CTD measurements from September 15, 2010. The streaming on both sides of the Green Island enhanced the current speed up to 1.6 m/s when a vortex formed; the recirculation covered an area 1–2 times that of the island, with water that was 1 °C–2 °C colder and 0.02–0.04 mg/m3 higher in chlorophyll-a concentration than the surrounding waters. In our results, the current speed at the vortex boundary increased to 1.22 ± 0.37 m/s at a depth of 22 m and 1.24 ± 0.25 m/s at a depth of 50 m in 22 cases; the average velocity at a depth of 72 m was 1.28 ± 0.22 m/s in the 76 cases when a vortex formed. If there is no vortex occurrence, the average velocity was 0.59 ± 0.24 m/s at a depth of 22 m and 0.80 ± 0.22 m/s at a depth of 50 m in 12 cases, and the average velocity at a depth of 72 m was 0.85 ± 0.23 m/s in 11 cases. From the MODIS image examples, the SST of the recirculation water was over 2 °C colder and the chlorophyll-a concentration was two times higher than that of the surrounding waters. The elevated chlorophyll-a concentration indicates enhanced phytoplankton productivity from the deep water, and the cold vortex likely induced Kuroshio water properties downstream and northward east of Taiwan. In this study, the spatial scale of the vortices was first analyzed and explained using satellite imagery data. The advantage of our approach is that we obtain examples of vortices and IOVTs that have indeed occurred and are not just numerical simulation results. Numerical simulations can, of course, simply adjust their parameters; however, a deeper understanding of the background information is required. Compared to the results of previous studies, the average transverse length in 52 cases (12.51 ± 3.96 km) is two times higher than that of Huang et al. (2014) and half that of Hsu et al. (2015a); the average stream-wise length (24.1 ± 8.65 km) is also half that of Hsu et al. (2015a). Different incoming velocities may explain these differences. Even though we obtain more accurate spatial scale analyzes than in models, the number of data is limited by satellite tracks and cloud blocking; however, this result is still of great help to improve numerical simulations.

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

The results also demonstrate that wind forcing has an impact on ocean circulation on short-time scales and could affect the spatial configuration of an ocean vortex train. The streamlines on the leeward side of Green Island appear to be pushed farther downstream with higher aspect ratio when the speeds of the southerly wind are larger than 7 m/s, and the dimensionless width decreases as the incoming flow velocity increases. In the case of northerly winds, however, the Kuroshio Current flows against the wind and the dimensionless width increases as the incoming flow velocity increases. The numerical simulations provide two possible states: the vortices are pushed and packed near the north of the island with a northerly wind and the vortices are pushed further downstream with a southerly wind (Hsu et al., 2015a). For selecting a Kuroshio power test site, several factors have to be taken into account, such as current velocity, island wakes, seabed topography, and geological features. The most important one is the current speed which is larger than 1 m/s (Hsu et al., 2015b). If the test site is selected at the variation range of the wake area, the changes of current velocity caused the wake may decrease the efficiency and safety of the power generators. The findings in this study probably provide a solution for consideration of selecting the Kuroshio power test site. 6. Conclusions Multi-satellite imagery and two moored ADCP stations were used to investigate current speeds, surface patterns, and the horizontal scale of Green Island vortices and the characteristics of IOVT. An ocean vortex and IOVT are formed leeward of Green Island by the incoming Kuroshio Current flow. The results for the island–vortex parameter and the Reynolds number suggest that periodic oscillation of the vortex and shedding of standing eddies behind the ocean vortex occur. When an ocean vortex forms, observations from the moored ADCP suggest that the velocity on the western side of Green Island increases. From the 307.2-kHz moored ADCP data, the current velocity of the vortex boundary increased to 1.22 ± 0.37 m/s and 1.24 ± 0.25 m/s at depths of 22 m and 50 m, respectively, when a vortex occurred, and the velocity slowed to 0.59 ± 0.24 m/s and 0.80 ± 0.22 m/s at depths of 22 m and 50 m, respectively, when a vortex did not occur. From the 76.8-kHz moored ADCP data, the current velocity of the vortex boundary at a depth of 72 m increased to 1.28 ± 0.22 m/s when a vortex formed, and the velocity slowed to 0.85 ± 0.23 m/s when no vortex formed. Vortex turbulence occurring in the Kuroshio Current may impact water properties and biology far downstream. In the case of the Green Island vortex, the MODIS data demonstrated that the recirculating water was over 2 °C colder and the Chl-a concentration was twice as high as that of the surrounding water. These phenomena suggest upwelling, mixing processes, and an island-mass effect. The SST decrease may cause the air–sea heat flux to be reduced and change the microclimate on the leeward side of Green Island; however, how far downstream this influence extends remains to be investigated. The IOVT showed variability under wind forcing. The aspect ratio of the IOVT obtained from satellite images was 2.09 under southerly winds and 1.91 under northerly winds. The dimensionless width of the IOVT from satellite images was 2.02 under southerly winds and 2.76 under northerly winds. The same trend was found in numerical models but at a much lower magnitude. This is assumed to be due to the higher parameter settings used in the model. The vortex kinetic energy and vertical current shears in the recirculation area require additional cases of shipboard ADCP measurements to resolve the dynamic regime of different types of vortices in continual time. Nevertheless, our measurement still captured multi-satellite images of the Green Island vortex to analyze significant spatial scale characteristics, on which further regional oceanographic and numerical studies can be based. In addition, an interesting topic for further research is whether the MODIS SST and Chl-a data can be used to reconstruct and predict the occurrence of an IOVT vortex leeward of Green Island.

63

Acknowledgements The authors express their thanks to the anonymous reviewers for their valuable suggestions and comments. The satellite images were provided by the Alaska Satellite Facility, NASA Distributed Active Archive Center and the Center for Space and Remote Sensing Research, National Central University, Taiwan. Altimetry data are provided by the Archiving Validation and Interpretation of Satellite Data in Oceanography (AVISO), the Centre National d'Études Spatiales (CNES) of France. MODIS data are produced with the NASA's OceanColor Web, developed and maintained by the Ocean Biology Distributed Active Archive Center (OB.DAAC). QuikScat and ASCAT wind data are provided by the Remote Sensing System, which is a scientific research company and supported by NASA and NOAA. This work was supported by the Ministry of Science and Technology of Taiwan through grants MOST 104-2611-M-019-020 and MOST 104-2221-E-019-033. This is also a cooperative project with the National Natural Science Foundation of China under project U1405233. References Alpers, W., Brandt, P., Lazar, A., Dagorne, D., Sow, B., Faye, S., Hansen, M., Rubino, A., Poulain, P., Brehmer, P., 2013. A small-scale oceanic eddy off the coast of West Africa studied by multi-sensor satellite and surface drifter data. Remote Sens. Environ. 129: 132–143. http://dx.doi.org/10.1016/j.rse.2012.10.032. Apel, J.R., 1987. Principles of Ocean Physics. (631 pp.). Academic Press, San Diego. Arístegui, J., Sangra, P., Hernández-León, S., Cantón, M., Hernández-Guerra, A., Kerling, J., 1994. Island-induced eddies in the Canary Islands. Deep-Sea Res. I Oceanogr. Res. Pap. 41 (10):1509–1525. http://dx.doi.org/10.1016/0967-0637(94)90058-2. Barton, E.D., Aristegui, J., Tett, P., Canton, M., Garcia-Braun, J., Hernandez-Leon, S., Nykjaer, L., Almeida, C., Almunia, J., Ballesteros, S., Basterretxea, G., Escanez, J., Garcia-Weill, L., Hernandez-Guerra, A., Lopez-Laatzen, F., Molina, R., Montero, M.F., Navarro-Perez, E., Rodriguez, J.M., van Lenning, K., Velez, H., Wild, K., 1998. The transition zone of the Canary Current upwelling region. Prog. Oceanogr. 41 (4):455–504. http://dx.doi. org/10.1016/S0079-6611(98)00023-8. Barton, E.D., Basterretxea, G., Flament, P., Mitchelson-Jacob, E.G., Jones, B., Aristegui, J., Herrera, F., 2000. Lee region of Gran Canaria. J. Geophys. Res. Oceans Atmos. 105 (C7):17173–17193. http://dx.doi.org/10.1029/2000JC900010. Caldeira, R.M.A., Groom, S., Miller, P., Pilgrim, D., Nezlin, N.P., 2002. Sea-surface signatures of the island mass effect phenomena around Madeira Island, Northeast Atlantic. Remote Sens. Environ. 80 (2):336–360. http://dx.doi.org/10.1016/S0034-4257(01)00316-9. Caldeira, R.M.A., Marchesiello, P., Nezlin, N.P., DiGiacomo, P.M., McWilliams, J.C., 2005. Island wakes in the Southern California Bight. J. Geophys. Res. Oceans Atmos. 110 (C11), C11012. http://dx.doi.org/10.1029/2004JC002675. Chang, M.H., Tang, T.Y., Ho, C.R., Chao, S.Y., 2013. Kuroshio-induced wake in the lee of Green Island off Taiwan. J. Geophys. Res. Oceans Atmos. 118 (3):1508–1519. http:// dx.doi.org/10.1002/jgrc.20151. De Souza, R.B., Mata, M.M., Garcia, C.A., Kampel, M., Oliveira, E.N., Lorenzzetti, J.A., 2006. Multi-sensor satellite and in situ measurements of a warm core ocean eddy south of the Brazil–Malvinas Confluence region. Remote Sens. Environ. 100 (1):52–66. http://dx.doi.org/10.1016/j.rse.2005.09.018. Dong, C.M., McWilliams, J.C., Shchepetkin, A.F., 2007. Island wakes in deep water. J. Phys. Oceanogr. 37 (4):962–981. http://dx.doi.org/10.1175/JPO3047.1. Hasegawa, D., Yamazaki, H., Lueck, R.G., Seuront, L., 2004. How islands stir and fertilize the upper ocean. Geophys. Res. Lett. 31 (16), L16303. http://dx.doi.org/10.1029/ 2004GL020143. Hasegawa, D., Lewis, M.R., Gangopadhyay, A., 2009. How islands cause phytoplankton to bloom in their wakes. Geophys. Res. Lett. 36 (20), L20605. http://dx.doi.org/10.1029/ 2009GL039743. Heywood, K.J., Barton, E.D., Simpson, J.H., 1990. The effects of flow disturbance by an oceanic island. J. Mar. Res. 48 (1):55–73. http://dx.doi.org/10.1357/002224090784984623. Hsu, T.W., Doong, D.J., Hsieh, K.J., Liang, S.J., 2015a. Numerical study of monsoon effect on Green Island wake. J. Coast. Res. 31 (5):1141–1150. http://dx.doi.org/10.2112/ JCOASTRES-D-14-00206. Hsu, T.W., Liau, J.M., Liang, S.J., Tzang, S.Y., Doong, D.J., 2015b. Assessment of Kuroshio current power test site of Green Island, Taiwan. Renew. Energy 81:853–863. http://dx. doi.org/10.1016/j.renene.2015.03.089. Hsu, P.C., Lin, C.C., Huang, S.J., Ho, C.R., 2016. Effects of cold eddy on Kuroshio meander and its surface properties east of Taiwan. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 9:5055–5063. http://dx.doi.org/10.1109/JSTARS.2016.2524698. Huang, S.J., Ho, C.R., Lin, S.L., Liang, S.J., 2014. Spatial-temporal scales of Green Island wake due to passing of the Kuroshio current. Int. J. Remote Sens. 35 (11−12):4484–4495. http://dx.doi.org/10.1080/01431161.2014.916047. Isoguchi, O., Shimada, M., Sakaida, F., Kawamura, H., 2009. Investigation of Kuroshio-induced cold-core eddy trains in the lee of the Izu Islands using high-resolution satellite images and numerical simulations. Remote Sens. Environ. 113 (9):1912–1925. http:// dx.doi.org/10.1016/j.rse.2009.04.017. Li, X.F., Clemente-Colon, P., Pichel, W.G., Vachon, P.W., 2000. Atmospheric vortex streets on a RADARSAT SAR image. Geophys. Res. Lett. 27 (11):1655–1658. http://dx.doi. org/10.1029/1999GL011212.

64

P.-C. Hsu et al. / Remote Sensing of Environment 193 (2017) 54–64

Liang, S.J., Lin, C.Y., Hsu, T.W., Ho, C.R., Chang, M.H., 2013. Numerical study of vortex characteristics near Green Island, Taiwan. J. Coast. Res. 29 (6):1436–1444. http://dx.doi. org/10.2112/JCOASTRES-D-12-00245.1. Lien, R.C., Ma, B., Lee, C.M., Sanford, T.B., Mensah, V., Centurioni, L.R., Cornuelle, B.D., Gopalakrishnan, G., Gordon, A.L., Chang, M.H., Jayne, S.R., Yang, Y.J., 2015. The Kuroshio and Luzon Undercurrent East of Luzon Island. Oceanography 28 (4): 54–63. http://dx.doi.org/10.5670/oceanog.2015.81. Nunalee, C.G., Basu, S., 2014. On the periodicity of atmospheric von Karman vortex streets. Environ. Fluid Mech. 14 (6):1335–1355. http://dx.doi.org/10.1007/s10652-014-9340-9. Pattiaratchi, C., James, A., Collins, M., 1987. Island wakes and headland eddies: a comparison between remotely sensed data and laboratory experiments. J. Geophys. Res. Oceans 92 (C1):783–794. http://dx.doi.org/10.1029/JC092iC01p00783. Peñaflor, E.L., Villanoy, C.L., Liu, C.T., David, L.T., 2007. Detection of monsoonal phytoplankton blooms in Luzon Strait with MODIS data. Remote Sens. Environ. 109 (4): 443–450. http://dx.doi.org/10.1016/j.rse.2007.01.019. Shang, S., Li, L., Li, J., Li, Y., Lin, G., Sun, J., 2012. Phytoplankton bloom during the northeast monsoon in the Luzon Strait bordering the Kuroshio. Remote Sens. Environ. 124: 38–48. http://dx.doi.org/10.1016/j.rse.2012.04.022. SSALTO/DUACS User Handbook, 2016. (M)SLA and (M)ADT Near-real Time and Delayed Time Products, CLS-DOS-NT-06-034, Toulouse, France. Teinturier, S., Stegner, A., Didelle, H., Viboud, S., 2010. Small-scale instabilities of an island wake flow in a rotating shallow-water layer. Dyn. Atmos. Oceans 49 (1):1–24. http:// dx.doi.org/10.1016/j.dynatmoce.2008.10.006.

Topouzelis, K., Kitsiou, D., 2015. Detection and classification of mesoscale atmospheric phenomena above sea in SAR imagery. Remote Sens. Environ. 160:263–272. http:// dx.doi.org/10.1016/j.rse.2015.02.006. Wolanski, E., Imberger, J., Heron, M.L., 1984. Island wakes in shallow coastal waters. J. Geophys. Res. Oceans 89 (C6):10553–10569. http://dx.doi.org/10.1029/ JC089iC06p10553. Yang, Y.J., Jan, S., Chang, M.H., Wang, J., Mensah, V., Kuo, T.H., Tsai, C.J., Lee, C.Y., Andres, M., Centurioni, L.R., Tseng, Y.H., Liang, W.D., Lai, J.W., 2015. Mean structure and fluctuations of the Kuroshio East of Taiwan from in situ and remote observations. Oceanography 28 (4):74–83. http://dx.doi.org/10.5670/oceanog.2015.83. Young, G.S., Zawislak, J., 2006. An observational study of vortex spacing in island wake vortex streets. Mon. Weather Rev. 134 (8):2285–2294. http://dx.doi.org/10.1175/ MWR3186.1. Zheng, Z.W., Zheng, Q., 2014. Variability of island-induced ocean vortex trains, in the Kuroshio region southeast of Taiwan Island. Cont. Shelf Res. 81:1–6. http://dx.doi. org/10.1016/j.csr.2014.02.010. Zheng, Q., Lin, H., Meng, J.M., Hu, X.M., Song, Y.T., Zhang, Y.Z., Li, C.Y., 2008. Sub-mesoscale ocean vortex trains in the Luzon Strait. J. Geophys. Res. Oceans 113 (C4), C04032. http://dx.doi.org/10.1029/2007JC004362. Zheng, Q., Holt, B., Li, X.F., Liu, X.N., Zhao, Q., Yuan, Y.L., Yang, X.F., 2012. Deep-water seamount wakes on SEASAT SAR image in the Gulf stream region. Geophys. Res. Lett. 39 (16), L16604. http://dx.doi.org/10.1029/2012GL052661.