GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L09604, doi:10.1029/2008gl033660, 2008 Slope currents around the Kerguelen Islands from demersal longline fishing records Young-Hyang Park, 1 Nicolas Gasco, 1 and Guy Duhamel 1 Received 15 February 2008; revised 19 March 2008; accepted 31 March 2008; published 8 May 2008. [1] The Kerguelen Plateau constitutes a natural obstacle for the eastward progress of the Antarctic Circumpolar Current, especially around the Kerguelen Islands. However, there is little quantitative knowledge of the current field around the islands due to lack of long-term current measurements. We performed a systematic analysis of a total of 28917 points of fishing gear drifts from setting and recovery positions of demersal longlines deployed between 2002 and 2007 for Patagonian Toothfish (Dissostichus eleginoides) fisheries in Kerguelen waters. This enabled the construction of a realistic field of depth-averaged time-mean slope currents along the 1000 m isobath all around the Kerguelen Islands. The strongest depth-mean velocities of 25 cm s 1 are associated with the Polar Front, which rounds the islands from the south and flows northward along the inner continental slope immediately east of the islands, strongly supporting previous hydrographic evidence. These results demonstrate the potential for hitherto unexploited historic longline drift data from demersal fishing grounds to provide valuable quantitative information on the regional circulation. Citation: Park, Y.-H., N. Gasco, and G. Duhamel (2008), Slope currents around the Kerguelen Islands from demersal longline fishing records, Geophys. Res. Lett., 35, L09604, doi:10.1029/2008gl033660. 1. Introduction [2] We present here a novel approach to extracting quantitative information on depth-averaged time-mean currents from available longline fishing records, using longline deployment positions recorded by onboard observers in Patagonian Toothfish (Dissostichus eleginoides) fisheries operating around the Kerguelen Islands. The Patagonian Toothfish is a long-lived (up to 40 50 years) large fish (up to 2 m long and 100 kg in weight) found in deep waters on seamounts and continental slopes around most subantarctic islands of the Southern Ocean. Demersal longlining is the main commercial fishing technique in waters around the Kerguelen Islands, and uses hundreds or thousands of baited hooks hanging from a single line up to tens of kilometers long. During the setting phase, the longline fishing gear drifts with ocean currents while sinking to the ocean floor. Anchors are fixed at both ends of a longline. The Global Positioning System (GPS)-derived geographical position and bottom depth of each anchor are recorded during the setting phase, and the subsequent day when lines are recovered. This enables each longline to provide two 1 Département Milieux et Peuplements Aquatiques, Muséum National d Histoire Naturelle, Paris, France. Copyright 2008 by the American Geophysical Union. 0094-8276/08/2008GL033660 velocity estimates, one from each anchor. The difference in anchor positions when lines are set and when they are recovered represents drift, which is assumed to be linearly proportional to a depth-mean current of the water column. We will show here how these raw drift data accumulated in recent years in Kerguelen waters can be transformed into a reliable current field in the otherwise poorly sampled Southern Ocean area. 2. Materials and Methods [3] The fishing technique used in Kerguelen waters involved setting longlines of variable lengths (range: 1 to 30 km, mean = 11 km, standard deviation (SD) = 3.6 km) on the seabed at depths ranging from 500 m to 2300 m (mean = 1068 m, SD = 326 m). The sink rate of fishing gears was estimated using a miniaturized time-depth recorder attached to a selected fishing gear during a test experiment. It varies only slightly in the vertical around a depth-mean value of 0.28 m s 1 (SD = 0.01 m s 1 ) through the 1000 m water column (N. Gasco, unpublished data, 2007). Here, we examined a total of 28917 points of drift data gathered from 14963 longline sets deployed during six years between 2002 and 2007 over the continental slope around the Kerguelen Islands. [4] We applied the following procedures to estimate a depth-averaged time-mean velocity field from the raw drift data. First, we constructed a network of rectangular grid cells having a size of 30 30 km, with each cell being located along the centroids of along-slope drift data points (Figure 1). Note that except for a limited shallow area (< 1000 m) immediately southeast of the islands, the quasi-totality of cells are found close to the 1000 m isobath. We retained only those cells containing at least 80 data points, yielding a total of 75 cells with a data density reaching up to 1259 points per cell (mean = 336, SD = 224). The choice of the cell size is subjective but based on the compromise between the spatial resolution of velocity field and the data density of each cell, although we subsequently found that the results are not very sensitive to the exact cell size between 15 and 30 km (see Section 3). Second, we determined for each cell a common reference depth by averaging the bottom depths corresponding to every data point within the cell. Individual drift distances obtained over various depths were linearly interpolated/ extrapolated to the reference depth. Third, we determined for each cell an area-mean drift by performing a spatial average of all drift vectors. This spatial average is equivalent to a time average, as the data points within each cell are randomly distributed in time over six years of our data acquisition period. Accordingly, we estimated a time-mean drift in each cell. The time-variable parts of drift are L09604 1of6
L09604 PARK ET AL.: SLOPE CURRENTS FROM LONGLINE FISHERIES L09604 Figure 1. Data points (green dots) where longline drifts have been sampled during the past six years (2002 7), superimposed on local bathymetry with isobaths every 500 m. Red rectangles represent grid cells used for computing areaaveraged longline drifts. believed to be largely removed by our averaging process, and may have been caused mostly by dominant tidal currents [Park et al., 2008] as well as by other transient features such as inertial currents. Finally, the corresponding depth-averaged time-mean velocity was obtained by dividing the mean drift by the time taken by an anchor to reach the reference depth. The latter time can be determined by the reference depth divided by the aforementioned depthmean sink rate of the fishing gear of 0.28 m s 1. The direction of the mean current is the same as that of the mean drift. 3. Results and Validation [5] Before presenting a large-scale comparison of our results to previous work later in this section, we will first show a point comparison with the direct current measurements made recently over the northern Kerguelen Plateau southeast off the Kerguelen Islands (49 450S, 73 120E; depth = 1072 m). This consisted of an one-year-long mooring of current-meters with an upward-looking acoustic Doppler current profiler placed at 500 m and two conventional current-meters at 700 m and 900 m, which had been performed as part of the multidisciplinary KEOPS experiment in 2005 [Park et al., 2008]. Our longline-derived timemean velocity vector estimated precisely at this mooring site is 5.3 cm s 1 towards 285. This compares favourably with the direct current measurements showing a vertically averaged time-mean (February 2005 to January 2006) velocity of 6.0 cm s 1 towards 296 [Park et al., 2008]. When a reduced bin size of 15 km is used for spatial averaging, the longline-derived velocity vector is estimated as 6.7 cm s 1 towards 294. This indicates that the method we describe here is quite robust and applies with a difference of about 10% in speed and of 10 in direction relative to the direct current measurements at this KEOPS site. [6] A major source of error in longline-derived velocities is likely to be related to the typical GPS position accuracy of the order of 15 m (http://gps.faa.gov/sitemap/ index.htm), which should yield a velocity error pffiffiof ffi 0.6 cm s 1 for a nominal 1000 m bottom depth (= 2 1500 0.28/1000). Assuming a normal distribution of GPS errors, our velocity estimates may have errors of ±1.2 cm s 1 (= 2 0.6) at the 95% confidence level. Another important source of errors may reside in the mean sink rate, which should be ±0.02 m s 1 (= 2 0.01) or about ±7 % of the nominal mean sink rate (0.28 m s 1) at the 95% confidence level. The same ±7 % errors can be attributed to the depth-mean velocity estimate, as the latter velocity is proportional to the mean sink rate by construction of our method. Therefore, the total root sum square error for a maximum velocity estimate of 25 cm qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1 s (see below) is 2 cm s (= 1:22 þ ð25 0:07Þ2 ). [7] The depth-averaged time-mean currents around the Kerguelen Islands as determined using the method described here are shown in Figure 2, which reveals a detailed picture of slope currents of O (10 cm s 1) rounding the islands. In the upstream area of the northwestern corner of the plateau, the eastward flow impinging on the plateau appears to be divided into northern and western branches north and south of 47 S. From there, the northern branch continues to the east hugging the northern escarpment while the western branch flows to the south to southeast along the western escarpment. Over the inner continental slope immediately east of the Kerguelen Islands near 71 E, there is a 2 of 6
Figure 2. Depth-averaged time-mean currents around the Kerguelen Islands as estimated from longline drift data. Areas shallower than 200 m are shaded grey. The PF location (red) and schematic flows representing the ACC core (black) and a deep western boundary current (blue) are from Figure 4. Green and orange lines are the northern and western branches discussed in the text. remarkable concentration of the strongest northward flow of up to 25 cm s 1 as far north as 48 S. There, it turns abruptly east following the local bathymetry before merging into the southward-bending northern branch. We cannot resolve a continuous flow along the shallower southern inner slope (< 1000 m) due to the absence of longline data. However, the strongest slope currents observed east of the Kerguelen Islands can be traced back to the upstream flow in the area southwest of the islands. This can be easily identified from satellite images by a narrow band of low-chlorophyll concentrations consistently developed during the Austral summer along the southern and eastern inner continental slope, separating two distinct chlorophyll-rich areas north and south of the islands (Figure 3). This chlorophyll-depleted plume advecting unproductive offshore waters has often been used as the clearest surface marker of the Polar Front (PF) of the area [Park et al., 2008, and references therein]. [8] Supporting evidence for this is given by two historical surface buoy trajectories [Daniault, 1984] superimposed in Figure 3. These closely overlap with the PF in the east of the islands, which is also remarkably coherent with our longline-derived velocity vectors shown in Figure 2. Surface velocities at the PF east of the islands from daily buoy positions are 40 cm s 1, as compared to the longlinederived velocities of 25 cm s 1, which is rather expected as the latter velocities are depth-mean values over the 1000 m water column while the strongest velocities are found in the surface layer. This strengthens further the validity of our method. Farther east of the islands, along the SE-NW oriented outer continental slope, a much weaker but significant northwestward flow (5 to 10 cm s 1 ) appears, originating from the south and joining gradually with the aforementioned strongest northward flow associated with the PF. Finally, we note a general eastward flow along the southern escarpment of the Skiff Bank centered at 50 S, 65 E. [9] Our results are consistent with previous work on the regional circulation from conventional hydrography [Park et al., 1993; Park and Gambéroni, 1997]. Recently, Park et al. [2008] established the most comprehensive synthesis of the large-scale circulation over and around the entire northern Kerguelen Plateau. They used a combination of new data from the KEOPS experiment with historical hydrographic data and all available information on mid-depth and near-surface velocities derived from autonomous floats, surface drifting buoys, satellite altimetry, and direct current measurements (Figure 4). Upon comparing Figures 2 and 4, it is immediately clear that the circulation pattern around the Kerguelen Islands is remarkably alike in both maps. Figure 4 shows the Antarctic Circumpolar Current (ACC) core tightly concentrating along the northern escarpment of the Kerguelen Plateau. Our eastward flowing northern branch seen in Figure 2 corresponds to the southern limb of that core. Figure 4 also reveals the PF rounding the Kerguelen Islands anticyclonically from the south to advance northward in the east of the islands before its southeast retroflection at about 48 S, 72 E. Our longlinederived currents faithfully mimic most of these features. Finally, Park et al. [2008] provide clear evidence of a deep western boundary current originating from the Fawn Trough and flowing northwestward along the eastern escarpment southeast of the Kerguelen Islands (see Figure 4b). In 3of6
Figure 3. MODIS data-derived chlorophyll concentrations in January 2005, showing a typical narrow band of chlorophyll-depleted waters extending along the Polar Front south and east of the Kerguelen Islands. Trajectories of two historical surface buoys launched respectively west (red) and east (white) of the islands [Daniault, 1984] are superimposed. Figure 2 we also see this feature clearly by a weak northwestward branch converging gradually into the northward flowing much stronger PF. 4. Discussion and Conclusions [10] We have demonstrated how longline fishing records can be used to identify and quantify the depth-averaged time-mean slope currents around the Kerguelen Islands. Knowing precise depth-mean currents has an important implication for directly diagnosing the transport as long as one has some knowledge of the current width. For example, assuming an effective width of 30 km (a reasonable value when judged from local hydrographic data) for the PF having a typical depth-mean velocity of 20 cm s 1 over 1000 m waters in the east of the Kerguelen Islands (see Figure 2), we can estimate a transport of 6 Sv (1 Sv = 10 6 m 3 s 1 ). This value is close to the bottom-referenced geostrophic transport of 5 Sv previously attributed to the PF in a nearby site from SUZIL data [Park et al., 1993]. [11] Also noteworthy is that the Kerguelen Plateau effectively controls impinging eastward streams to pass round the shallow topography, tending to run parallel with bathymetry, instead of crossing it. This somewhat expected fact, in view of the potential vorticity conservation constraint (f/h = const, where f is the Coriolis factor and h is the layer thickness), has practical implications for critically evaluating the appropriateness of several conflicting positions of the PF suggested in the literature. For a surfaceintensified flow with a typical upper-layer thickness of the order of 500 m [e.g., Park et al., 2008], there are only two viable zonal passages satisfying the potential vorticity conservation principle: one along the northern escarpment north of 47 S and the other through a trough (sill depth = 650 m) developed immediately south of the islands between 50 S and 51 S [see Park et al., 2008, Figure 1]. The PF location of Belkin and Gordon [1996], which has been defined by a subsurface temperature minimum of 2.5 C, may correspond to the northern passage. However, this temperature criterion of 2.5 C is in conflict with the more traditional criterion of the PF, i.e., the northernmost limit of a subsurface temperature minimum less than 2 C [e.g., Park et al., 1993; Orsi et al., 1995]. Sparrow et al. [1996] suggested a double-pf pattern, with the surface PF passing through the Fawn Trough and the subsurface PF hugging the Kerguelen Islands from the north. This latter front and that of Orsi et al. [1995] cross a shoal much shallower than 200 m that is widely developed around the islands between 47 S and 50 S. It can be easily verified that these frontal locations over the shallow shoal (< 200 m) hardly satisfy the vorticity conservation for the 500 m thick upper-layer flow, as f varies only a little (< 10%) over the area and the relative vorticity is negligible compared to f. Moreover, the subsurface temperature minimum does not exist over the entire plateau north of the Kerguelen Islands [Park and 4of6
Gambéroni, 1997]. In contrast to this, our longline-derived velocity vectors are in close proximity of such a PF location that satisfies not only the vorticity conservation principle but also the aforementioned traditional criterion of the PF. [12] Previously, two Lagrangian methods employing satellite tracking of surface drifters and mid-depth floats have been used for constructing the velocity field at a specific depth. These methods yield a mean velocity field averaged within 1 latitude 1 longitude boxes for surface drifters [Niiler et al., 2003] or objectively interpolated using a 300 km radius of influence for autonomous floats [Davis, 2005]. While limited in the spatial extent, our method provides a far finer velocity field along the continental slope, permitting a detailed description of local branches of the ACC that are invisible by the above large-scale Lagrangian methods. [13] Can our method yield a further finer velocity field by reducing the bin size? A velocity field constructed using a 15 km grid (not shown) reveals an overall circulation pattern quite similar to the 30 km grid solution, except for several directionally inconsistent vectors appearing mostly in regions of weak currents. Note that each longline data point represents about 1 h of drift through the nominal 1000 m water column, considering the mean sink rate of 0.28 m s 1. As already mentioned, the minimum data points validated for the 30 km grid are 80, which is equivalent to a cumulative drift time of 80 h or 6.5 semidiurnal tidal cycles. The reduction in grid size to 15 km (minimum data points = 25) shortens considerably this drift time by a factor of 3, thus seriously affecting the efficiency of averaging out tidal and other transient currents. Therefore, a finer spatial resolution would better await the accumulation of more data in the future. Finally, investigation of the interannual variability of slope currents is beyond the scope of this study but warrants a systematic analysis of accumulating data, as part of a climate-related research. [14] In conclusion, we consider that our quantitative results of slope currents obtained as a by-product of demersal longline fishing activities in Kerguelen waters will serve as a useful reference to future frontal studies of the area. The technique we developed here can also be deployed in other sectors of the Southern Ocean easily, e.g., around the subantarctic islands where there have been similar demersal longline fishing activities such as the Crozet Islands (4426 lines deployed since 2002) or South Georgia Islands (16500 lines deployed since 2001). Around the Heard-MacDonald Islands, a systematic data collection has been made since 2006 only, so additional years of data collection seem to be necessary before any scientific exploitation. The present work highlights the potential for hitherto unexploited longline drift data to reveal a realistic regional circulation in such scientifically and economically important areas of the Southern Ocean. Figure 4. (a) Satellite altimetry-derived mean surface dynamic topography. (b) Schematic of the large-scale circulation around the northern Kerguelen Plateau. Thin arrows with Arabic numerals stand for directly measured upper-layer velocities from the KEOPS experiment. Adapted from Park et al. [2008]. [15] Acknowledgments. We are grateful to all fishing masters and onboard fisheries observers who collected since 2002 records of longline deployment positions in fisheries off the Kerguelen Islands. We have much benefited from a stimulating discussion with Igor Belkin. We thank him and an anonymous reviewer for their constructive comments on the manuscript. We acknowledge Isabelle Durand for her contribution in data analysis and graphics. References Belkin, I. M., and A. L. Gordon (1996), Southern Ocean fronts from the Greenwich Meridian to Tasmania, J. Geophys. Res., 101, 3675 3696. Daniault, N. (1984), Apport des techniques spatiales à la connaissance des courants de surface. Application à l océan Antarctique, Ph.D. thesis, 68 pp., Univ. de Bretagne Occidentale, Brest, France. Davis, R. E. (2005), Intermediate-depth circulation of the Indian and South Pacific oceans measured by autonomous floats, J. Phys. Oceanogr., 35, 683 707. Niiler, P. P., N. A. Maximenko, and J. C. McWilliams (2003), Dynamically balanced absolute sea level of the global ocean derived from near-surface velocity observations, Geophys. Res. Lett., 30(22), 2164, doi:10.1029/ 2003GL018628. Orsi, A. H., T. Whitworth III, and W. D. Nowlin (1995), On the meridional extent and fronts of the Antarctic Circumpolar Current, Deep Sea Res., Part I, 42, 641 673. Park, Y.-H., and L. Gambéroni (1997), Cross frontal injections of Antarctic Intermediate Water and Antarctic Bottom Water in the Crozet Basin, Deep Sea Res., Part II, 44, 963 986. 5of6
Park, Y.-H., L. Gambéroni, and E. Charriaud (1993), Frontal structure, water masses and circulation in the Crozet Basin, J. Geophys. Res., 98, 12,361 12,385. Park, Y. H., F. Roquet, I. Durand, and J. L. Fuda (2008), Large-scale circulation over and around the Northern Kerguelen Plateau, Deep Sea Res., Part II, doi:10.1016/j.dsr2.2007.12.030, in press. Sparrow, M. D., K. J. Heywood, J. Brown, and D. P. Stevens (1996), Current structure of the south Indian Ocean, J. Geophys. Res., 101, 6377 6391. G. Duhamel, N. Gasco, and Y.-H. Park, Département Milieux et Peuplements Aquatiques, Muséum National d Histoire Naturelle, F-75005 Paris, France. (yhpark@mnhn.fr) 6of6