Weddell Sea iceberg drift: Five years of observations

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111,, doi: /2004jc002661, 2006 Weddell Sea iceberg drift: Five years of observations M. P. Schodlok, 1 H. H. Hellmer, 1 G. Rohardt, 1 and E. Fahrbach 1 Received 12 August 2004; revised 7 December 2005; accepted 23 February 2006; published 14 June [1] The drift of 52 icebergs tagged with GPS buoys in the Weddell Sea since 1999 has been investigated with respect to prevalent drift tracks, sea ice/iceberg interaction, and freshwater fluxes. Buoys were deployed on small- to medium-sized icebergs (edge lengths 5 km) in the southwestern and eastern Weddell Sea. The basin-scale iceberg drift of this size class was established. In the western Weddell Sea, icebergs followed a northward course with little deviation and mean daily drift rates up to 9.5 ± 7.3 km d 1.To the west of 40 W the drift of iceberg and sea ice was coherent. In the highly consolidated perennial sea ice cover of 95% the sea ice exerted a steering influence on the icebergs and was thus responsible for the coherence of the drift tracks. The northward drift of buoys to the east of 40 W was interrupted by large deviations due to the passage of lowpressure systems. Mean daily drift rates in this area were 11.5 ± 7.2 km d 1. A lower threshold of 86% sea ice concentration for coherent sea ice/iceberg movement was determined by examining the sea ice concentration derived from Special Sensor Microwave Imager (SSM/I) and Advanced Microwave Scanning Radiometer for EOS (AMSR-E) satellite data. The length scale of coherent movement was estimated to be at least 200 km, about half the value found for the Arctic Ocean but twice as large as previously suggested. The freshwater fluxes estimated from three iceberg export scenarios deduced from the iceberg drift pattern were highly variable. Assuming a transit time in the Weddell Sea of 1 year, the iceberg meltwater input of 31 Gt which is about a third of the basal meltwater input from the Filchner Ronne Ice Shelf but spreads across the entire Weddell Sea. Iceberg meltwater export of m 3 s 1, if all icebergs are exported, is in the lower range of freshwater export by sea ice. Citation: Schodlok, M. P., H. H. Hellmer, G. Rohardt, and E. Fahrbach (2006), Weddell Sea iceberg drift: Five years of observations, J. Geophys. Res., 111,, doi: /2004jc As the day wore on the bergs, instead of diminishing in number, actually increased: and wonder we had felt in the morning increased to amazement. Every mile forward brought new columns into sight. They seemed inexhaustible. A grander display of ice power had perhaps never been seen. In a twenty-four hour period ending December 28th Dr. Poulter estimated we sighted 8000 bergs. Byrd [1935, p. 51] 1. Introduction [2] This account from the second Byrd Antarctic Expedition ( ), near 66 S, W, approximately 2000 km north of Marie Byrd Land, illustrates the massive loss from the Antarctic ice sheet due to continuous calving. While sightings on such a scale are rare, they raise questions as to the potential of iceberg impact on hydrology, circulation, and biology which are related to drift and melting. [3] The mean annual iceberg calving rate in the Southern Hemisphere is estimated to be 2016 Gt yr 1 [Jacobs et al., 1992]. This value is associated with uncertainties in size assessment, i.e., volume estimate, event frequency, and 1 Alfred Wegener Institute for Polar and Marine Research, Bremerhaven, Germany. Copyright 2006 by the American Geophysical Union /06/2004JC decay rate between calving and observation. Assuming a mean iceberg density of 850 kg m 3 [Weeks and Mellor, 1978], the calving rate represents an equivalent volume of msv (1 msv = 10 3 m 3 s 1 ) of freshwater which is transferred to the upper Southern Ocean water column. This is approximately 20% of the annual National Centers for Environmental Prediction (NCEP) net precipitation (P-E) value (413 msv) into the Southern Ocean south of 55 S and is far larger than the freshwater flux due to ice shelf basal melting (28 msv) [Hellmer, 2004]. [4] The implications of iceberg meltwater for Southern Ocean hydrography and marine biology depend upon the location and quantity of melting. Numerical studies of the impact of ice shelf basal melt water showed that the weakly stratified continental shelf water column is sensitive to changes in the freshwater fluxes: an addition stabilizes the water column and decreases dense water formation, while a reduction enhances deep convection and leads to sea ice thinning [Hellmer, 2004]. The role of iceberg meltwater in this freshwater balance is not well known. [5] In addition, iceberg meltwater contains meteoric dust (up to kg per 1 m 3 of ice (D. Wagenbach, personal communication, 2003)), accumulated on the ice sheet over several thousands of years. Micronutrients within the dust, such as iron, may contribute to the fertilization of the upper 1of14

2 Figure 1. Dates, positions, and number of buoys deployed in the Weddell Sea during five cruises of R/V Polarstern. The majority of buoys were released off Neumayer Station (inset). ocean when released by melting. Melt-driven upwelling in the vicinity of icebergs may have a similar effect as it can bring nutrient-rich deeper waters to the euphotic zone. [6] Iceberg groundings cause long-term disturbances to the local benthic ecosystem [Gutt and Starmans, 2001] and to hydrographic conditions [Nøst and Østerhus, 1998]. With a decrease in temperature and salinity the microbial community composition can shift in favor of those species which can adapt most rapidly to the new conditions in the water column. In 1986, the giant icebergs A22, A23, and A24 stranded on the shallow Berkner Bank (42 W, 76 S). This grounding altered the sea ice coverage, water mass characteristics and the flow pattern in the Filchner Trough [Grosfeld et al., 2001]. Despite the potentially far-reaching impacts of iceberg melting, little is known about prevalent berg drift tracks, i.e., the location and quantity of freshwater input in the Weddell Sea. [7] Technology for monitoring drift tracks has advanced considerably over the last 30 years. Giant tabular icebergs with edges longer than 18.5 km (satellite sensor detection limit is 10 nm; 1 nm = km) calving along the Antarctic continental coastline have been monitored by active and passive satellite remote sensing since the late 1970s by, for example, the National Ice Center (NIC). The early observations by, e.g., Swithinbank et al. [1977] established that the main northward drift routes in the western Ross and Weddell seas lead icebergs away from the major ice shelves toward and into the Antarctic Circumpolar Current. However, massive calving events, such as the most recent of the Filchner-Ronne (1998/1999) and Ross ice shelves (2001/2002), occur infrequently. Melting and breakup to sizes below the detection minimum resulted in the bergs disappearance from satellite observations. Consequently, the temporal coverage of the drift of massive bergs is incomplete due to the largely unknown drift after breakup, and the spatial coverage is limited to the western boundary current regions where these bergs are found. [8] A larger spatial coverage of drift tracks can be obtained by conventional technology, that is, tagging small to medium-sized icebergs (5 km edge length) which calve continuously. Size distributions observed by Budd et al. [1980] and Orheim [1980] determined edge maxima at 500 m and 1000 m with strong latitudinal dependence. Few icebergs of this size class were tagged in the 1970s [Tchernia and Jeannin, 1984; Vinje, 1980]. The drift of 21 icebergs, the majority tagged in the Indian and Pacific Ocean sectors of the Southern Ocean, was found to follow the path of the coastal current with retroflections corresponding to the main centers of atmospheric lows [Tchernia and Jeannin, 1984]. Vinje [1980] complemented these data with eight iceberg drift observations in the Weddell Sea. His study showed icebergs and sea ice reacting similarly to the effects of wind in areas covered with sea ice and a steering influence of the sea ice movement on the iceberg was suggested. [9] Various numerical studies of iceberg drift have addressed the impact of time- and space-variable ocean drag, wind drag, and Coriolis and other forces on the drift [e.g., Sodhi and El-Tahan, 1980; Gladstone et al., 2001; Lichey and Hellmer, 2001]. However, the latter demonstrated the dominant role of sea ice concentration on iceberg drift: While ocean/wind drag and Coriolis force were found to be the most important driving forces in open water, modeled iceberg trajectories were closer to the observed drift of giant iceberg C7 when sea ice concentrations above 90% controlled the iceberg motion. [10] The objective of this study is to analyze the prevalent iceberg tracks in the Weddell Sea, to relate iceberg to sea ice drift and to estimate the freshwater flux from melting icebergs. For this purpose the Alfred Wegener Institute deployed 52 buoys on icebergs from January 1999 to 2of14

3 January 2003 in the southwestern and southeastern Weddell Sea with the majority of bergs tagged off Neumayer station (70.6 S, 8.25 W, Figure 1). The buoys report daily GPS positions via the ARGOS system and a few buoys also transmit sea level pressure or air temperature data. This 5 year data set provides a detailed view of prevailing drift tracks and their variability in the Weddell Sea. Iceberg displacements were calculated from the position data and three zones determined to illustrate iceberg movements. The results were compared to previous basin-scale sea ice drift studies by, e.g., Wadhams et al. [1989] and Massom, 1992]. The sea ice/iceberg interaction was analyzed using buoy arrays in the eastern and in the southwestern Weddell Sea. Direct information about iceberg behavior during the austral winter in the northward advancing sea ice pack was obtained from an iceberg/sea ice buoy array. Periods of coherent sea ice/iceberg drift were identified as minima in the buoy kinematic time series. The freshwater flux was estimated from a combination of berg volumes (see iceberg dimensions in Table 1), observed drift patterns and modeled melt rates. 2. Data Sampling and Analysis 2.1. Iceberg Buoys [11] Icebergs chosen for tagging were generally of length and width between 100 m and 1500 m, with freeboard heights between 10 m and 70 m, and drafts between 80 m and 550 m. Several larger icebergs were tagged, the largest being A43b, which resulted from the calving events at the Filchner-Ronne Ice Shelf in 1998 and measured 40,000 m 17,000 m 50 m (length width freeboard). The origin of icebergs tagged in the southwestern Weddell Sea was well known. The calving sites of icebergs marked in the eastern Weddell Sea and their prior drift is unknown. [12] Fifty-two icebergs were tagged opportunistically during five R/V Polarstern cruises which are referenced by date of deployment (e.g., Jan/99 for buoys released in January 1999). As R/V Polarstern supplies Neumayer Station once a year between November and March, the majority of buoys were placed on icebergs in the vicinity of Atka Bay (Figure 1, inset). This enables seasonal as well as interannual differences in drift patterns to be determined. To cover a larger area, an inhomogeneous distribution of buoys was sought when possible. This led to considerable variability in oceanographic and atmospheric conditions experienced by the icebergs monitored. [13] A deployment was considered after inspection of the iceberg surface. Where crevasses were present, large connected areas were sought to ensure a position transmission of the buoy/iceberg remains in case of breakups. Usually, this area was not located at the center of the iceberg. Helicopter altimeter records gave height, and GPS readings at the iceberg edges size estimates (Table 1). [14] The iceberg buoys transmitted their position daily (12:00 h) using the ARGOS satellite system. GPS positions were given with an accuracy of ±15 m. In case of GPS instrument failure the ARGOS transmitter positions with lower accuracy were recorded (200 m to 300 m, depending on satellite availability). Transmission and/or functionality problems sometimes resulted in position reports at varying times. Where necessary, interpolation routines were applied for a time interval between positions of 24 hours. Mean daily drift velocities were derived from the displacement between successive positions giving u and v, the eastward and northward velocity components, respectively. An error of ± ms 1 resulted from the accuracy of the sensor. [15] Buoys failing to transmit for more than two weeks were regarded as malfunctioning. Some of the newer generation of instruments recorded air pressure and/or air temperature with a sampling interval of 3 hours, thus enhancing the sampling interval for iceberg position. Sea level pressure and/or air temperature records were entered into the GTS system and used in the analysis model of the European Centre for Medium-range Weather Forecast (ECMWF). ECMWF data (pressure, air temperature, wind velocities) were used to interpret the drift trajectories. [16] With a 24 hour sampling interval, the data do not resolve high frequency signals like the M 2 tide, the main component of tidal energy in the Weddell Sea [Robertson et al., 1998]. However, they represent the overall drift pattern, as well as mesoscale and large-scale variability Sea Ice Concentration [17] From March to September a roughly 1-m thick sea ice cover is formed in the Weddell Sea which can extend to about 60 S, and to the west of the South Sandwich Islands farther north. During maximal sea ice extent an area of km 2 is covered. In October/November the ice edge begins to retreat near Maud Rise (3 E, 65 S). The area of minimal sea ice cover in February/March is km 2 with perennial sea ice in the western Weddell Sea. The seasonal cycle of sea ice cover lags the cycle of air temperature by 2 months. The interannual variability in sea ice extent in the 1980s and 1990s was seen as a second mode wave pattern of the Antarctic Circumpolar Wave (ACW) with maxima every 4 years [White and Peterson, 1996]. [18] To obtain a relationship between iceberg and sea ice drift, sea ice concentrations were derived from Special Sensor Microwave Imager (SSM/I) data (before June 2003) using the 85 GHz channel using the Arctic Radiation and Turbulence Interaction Study (ARTIST) sea ice (ASI) algorithm [Kaleschke et al., 2001]. The SSM/I data were available on a grid of 12 km, and from mid-2003, the higher resolution Advanced Microwave Scanning Radiometer for EOS (AMSR-E) data were available on a 6 km grid. AMSR sea ice concentrations were also calculated with ASI, using the 89 GHz channel. The ASI AMSR sea ice concentration is slightly overestimated compared to the BOOTSTRAP algorithm, whereas the sea ice edge is much better represented (G. Spreen, personal communication, 2004). The pixel containing the iceberg buoy was taken to give the sea ice concentration Differential Kinematic Parameters [19] The behavior of iceberg movement with and within the sea ice cover was studied by calculating the timedependent, two-dimensional horizontal flow field of several buoy arrays. Four triangular buoy arrays were available from the 52 buoys deployed for analysis. Two arrays (sections 3.2 and 3.3) were assumed to represent the regional differences of iceberg movement within sea ice in 3of14

4 Table 1. Iceberg Buoy Information a No. ID Date of Deployment Iceberg Dimension L W F Latitude Longitude Last Date of Transmission Latitude Longitude Jan S E 25 Mar S E Jan S W 8 Feb S E Jan S W 21 Dec S W Jan S W 17 Jan S W Jan S W 7 Feb S W Jan S W 18 Feb S W Jan S W 5 May S W Jan S W 25 Dec S W Feb S W 10 Mar S W Feb S W 10 Nov S W Mar S W 04 Mar S W Jan S W 18 Mar S W Jan S W 30 May S E Jan S W 15 Apr S W Jan S W 21 Feb S W Feb S W 25 Feb S W Feb S W 10 Jun S W Feb S E 25 Feb S E Feb S E 2 Apr S W Feb S E 12 Mar S E Dec S W 3 Feb S W Dec S W 7 Jan S W Dec S W 16 May S W Dec S W 28 Feb S E Dec S W 28 Feb S W Dec S W 27 Feb S E Dec S W 20 Feb S W Dec S W 3 Mar S W Dec S W 31 Aug S W Dec S W 2 Mar S W Dec S W 28 Mar S W Dec S W 20 Mar S W Dec S E 1 Sep S W Jan S W 1 Sep S W Jan S W 1 Sep S W Jan S W 1 Sep S W Jan S W 12 Feb S W Jan S W 1 Sep S W Jan S W 20 Mar S W Jan S W 1 Feb S W Jan S W 1 Sep S W Dec S W 17 Jan S W Dec S W 5 Mar S W Dec S W 5 Mar S W Dec S W 6 Mar S W Dec S E 7 Mar S W Dec S E 17 Feb S W Dec S E 22 Dec S E Dec S E 10 Jan S E Dec S E 7 Mar S W Dec S E 7 Mar S E a Iceberg dimensions L, length; W, width; and F, freeboard, are given in meters. the eastern and western Weddell Sea, and are presented here. [20] The deformation of the arrays was examined following [Massom, 1992] by using the change-of-area method. This provides four components of movement, i.e., divergence, vorticity, shear and stretch deformation referred to as the differential kinematic parameter (DKP). The method was developed by Saucier [1955] and used in the Southern Ocean for sea ice kinematics by, e.g., Wadhams et al. [1989]. The DKP components were compared with sea ice concentrations and available atmospheric observational and/ or model data. The differential motion of a sea ice field is related to changes in ice concentration. This in turn is related to atmospheric forcing events. Hence icebergs trapped in large-scale sea ice fields should be seen as minima in DKP components. We assume that divergent and convergent movement is small when icebergs are in an area of high sea ice concentration but large in open water conditions. The horizontal divergence D is expressed as the rate of change over time of a triangular area A which encloses the iceberg buoys: @y 1 da A dt ¼ 1 A 1 A 0 A 0 t 1 t 0 ð1þ 4of14

5 with the areas A 0 and A 1 calculated at times t 0 and t 1, respectively. [21] The vorticity, shear and stretch deformations can be calculated similarly using the rotation of the u and v vectors [Molinari and Kirwan, 1975; Wadhams et al., 1989]. The physical interpretation of these parameters has been provided by Kirwan [1975]; divergence is the rate of area change without change in orientation or shape; vorticity is the rate of area change without change in shape or size; shear deformation is the rate of area change in shape due to forces acting parallel to the array sides (an elongated SE to NW axis is associated with negative, a shortened SW to NE axis with positive values), and stretch or normal deformation is the rate of elongation of the array, that is, compression or dilation along its N-S and/or E-W axis [Massom, 1992]. 3. Iceberg Drift 3.1. General Features [22] The basin-wide iceberg drift (Figure 2), showed two distinctive patterns to the east and west of approximately 40 W which is in accordance with previous sea ice drift observations [e.g., Vihma et al., 1996]. Iceberg drift west of 40 W was more uniformly northward with little meandering while to the east of 40 W the drift was more complex with larger deviations from the main course north. [23] Icebergs tagged west of 40 W on the western continental shelf moved northward with mean speeds between 2.9 ± 4.3 km d 1 (westernmost track) and 9.5 ± 7.3 km d 1 (easternmost track). The drifts divide into a northerly course (Figure 2d) with iceberg export out of the Weddell Sea and an easterly course following the Weddell Gyre circulation (Figure 2a) depending on the time of year the icebergs arrived off Joinville Island in the northwestern Weddell Sea. It is still unclear what physical mechanisms, for example, seasonality of sea ice concentration, prevalent atmospheric conditions or ocean circulation associated with rough bottom topography, determined whether icebergs remained within the Weddell Sea, followed the ACC fronts in the Scotia Sea or headed due north across the ACC fronts (for the latter, see Grosfeld et al. [1998]). [24] Buoys deployed east of 40 W, in particular along the Greenwich Meridian (GM) north of 67 S in open water conditions, transmitted the shortest position records (less than 6 months). Observations made during deployment suggest that these icebergs were already eroded due to a prolonged drift and melt history. This supports the statement of Orheim [1980], who suggested that icebergs do not survive 2 months in the open sea. In contrast to this finding, icebergs in the central Weddell Sea experiencing a second season drifted in waters with low ice concentrations or open water for up to 90 days (Figure 2c) before being entrained in sea ice again. [25] Although Orheim [1980] reported very low numbers of icebergs in the central Weddell Sea during the 1980s, the majority of icebergs tagged in the CC area off Neumayer in this study (Figure 1, inset) drifted into the central Weddell Sea. The observed drift patterns varied from year to year, depending on the location of deployment in the eastern Weddell Sea, and the mean speed varied from 10.5 ± 7.3 km d 1 to 12.2 ± 8.3 km d 1 before reaching the ACC. [26] Satellite observations of icebergs off Dronning Maud Land drifting in the CC have indicated a westward intensification in speed [Aoki, 2003]. This westward intensification west of 10 W is not evident in our data. The mean speed of the buoys following the CC on the narrow continental shelf was 13.7 ± 8.8 km d 1 with maxima up to 50 km d 1 off Neumayer. The interannual variability is apparent in the different southernmost positions reached before the turn north into the central Weddell Sea. This occurred between March and May with sometimes a dispersed (Figure 2b) and sometimes a narrow spreading from the CC (Figure 2c). Iceberg groundings (15% of the buoys) occurred on the continental shelf between 12 W and 21 W, lasting for 30 to 450 days. The disturbance of the local ecosystem due to this iceberg size class is more prominent in the eastern than in the western Weddell Sea. [27] The existence of two distinctive drift patterns east and west of 40 W can be substantiated by examining the meander coefficients of this large drift data set. The meander coefficient is the ratio of total distance covered to the overall displacement and thus a measure of directness of drift from point A to point B [Limbert et al., 1989; Massom, 1992]. A minimum coefficient of one represents a straight path. The circulation of the Weddell Gyre was considered in three sections; the Coastal Current (CC) with westward transport, the inner Weddell Sea (IW) with mainly northward transport (divided at 40 W), and the northern part of the Weddell Gyre and/or Scotia Sea (NW) with mainly eastward transport. The meander coefficients are presented in Table 2. Low coefficients of 1.2 to 1.9 were found in the fast flowing CC which is directed west-southwest parallel to the coast and represents the southern branch of the Weddell Gyre. The very low interannual variability in meander coefficients may have resulted from the similar deployment time in austral summer. With little sea ice cover the bergs continued to drift with the CC until being forced northward, presumably by catabatic winds from the continent. [28] The western boundary current of the Weddell Gyre (part of the IW section) was also associated with low meander coefficients. From close to the Antarctic Peninsula up to approximately 40 W the values ranged from 1.6 to 2.4 which is slightly higher than the values provided by Vihma et al. [1996] for sea ice buoys. Persistent southerly and southwesterly winds dominated the northward sea ice drift and, in turn, that of icebergs enclosed in the sea ice cover. The motion was interrupted by southward deviations during the passage of low-pressure systems. Low values could indicate less atmospheric influence on the perennial sea ice cover or, in the westernmost tens of kilometers, the closer proximity to more compact land-fast ice with higher viscosity, which dampened the atmospherically induced variability. [29] Higher values east of 40 W are in accordance with sea ice buoy coefficients found in 1980 [Massom, 1992] and 1992 [Vihma et al., 1996]. Here, the sea ice cover is unconsolidated owing to the frequent passage of atmospheric low-pressure systems across the area. Additionally, icebergs surviving more than one winter experience similar sea ice conditions again in the northern Weddell Sea, although the conditions are more variable and the effect less marked. [30] Most of the icebergs tagged in the eastern Weddell Sea needed more than one seasonal cycle to reach the 5of14

6 Figure 2. Drift tracks of icebergs in the Weddell Sea tagged with buoys deployed in (a) January 1999, (b) January 2000, (c) December 2000, (d) January 2002, and (e) December The labels, e.g., Jan/99, refer to month and year of deployment. northern fringe of the Weddell Sea (60 S). After turning north into the central Weddell Sea between March and May, and being entrained into the sea ice, they generally followed a northeastern course in a large-scale cyclonic feature with a diameter of more than 400 km. At about 66 S the start of a second cyclonic feature of similar size (e.g., Figures 2c, 2d, and 2e) indicated the onset of the second winter season. The freezing and advance of the sea ice edge determined whether the iceberg was incorporated into the sea ice and this, together with the large-scale circulation of ocean and atmosphere, determined the drift pattern Western Weddell Sea (January September 2002) [31] In this section, the main focus will be on a sea ice/ iceberg buoy triangle ( ) deployed at the summer sea ice edge (Figure 3). The sea ice buoys (see Kottmeier et al. [1997] for buoy specifications) were positioned between 62 km (8060) and 110 km (9781) away from the iceberg (9367). The combination of such a set of buoys enables a direct comparison of iceberg and sea ice drift. The kinematics of the triangle together with AMSR-E-derived sea ice concentrations provide information about the sea ice/ iceberg interaction. [32] The buoys, released just north of the continental shelf break, closely followed the continental slope on the way north (Figure 3). Their northward progression can be separated into three sections; the southern, with speeds of 6±4kmd 1 and sea ice concentrations of 68 ± 16% from January to March (day 90); the middle, with dominant northward drift, speeds of 12.2 ± 7.5 km d 1 and ice concentrations of 97 ± 3% from April to July; and the northern, with prevailing eastward velocity components, speeds of 17.5 ± 10.7 km d 1 and ice concentrations of 82 ± 14% from July to the end of the record. [33] The northward advance of the sea ice edge was faster than the northward progression of the sea ice/iceberg buoy 6of14

7 Table 2. Meander Coefficient for Iceberg Buoys With a Drift History of More Than 6 Months a ID CC IW NW Jan Jan Dec Jan Dec a The drift region is partitioned into the Coastal Current (CC), the inner Weddell Sea (IW) and northern Weddell Sea (NW) areas, where the IW area represents the main northward flow and the NW area represents the main eastward flow including the Scotia Sea. array. As sea ice buoys deployed in 1980 showed a similar behavior with respect to the ice edge advance, we consider the thermodynamic process described by Massom [1992] to be the responsible driving mechanism. The array s position remained near the sea ice edge for only a few weeks before it was entrained in the pack ice. The eastward turn of the buoy ensemble at around 64 S is in accordance with the large-scale ocean circulation and is caused by the steering effect of bottom topography (Joinville Ridge, South Orkney Plateau) on the Weddell Gyre. Deviations from the mean northward course, that is anticyclonic loops, superimposed on the drift before the ensemble turned east were associated with atmospheric forcing. [34] The amplitude of the meridional and zonal excursion around latitudes 68 S and 66.5 S was represented by the difference between two anticyclonic circulation features (April and May 2002) which completed cycles in 14 and 7 days, respectively (Figure 3b). During the first loop in April 2002, the eastern buoys (47 W) covered a N-S excursion of 48 km, which was lower, at 36 km, to the west. The westernmost buoy (9728, 52 W, Figure 3) which was not part of the triangle, covered a N-S excursion of only 16 km. The decrease in excursion amplitude with increasing proximity to the Antarctic Peninsula induced by a passing lowpressure system was also observed in sea ice motion by Vihma et al. [1996] in In April 1992, sea ice buoys performed a 10-day cyclonic loop at a similar latitude (68.5 S) with north-south and east-west scales of 40 km at 40 W and 20 km at 50 W. While the N-S excursion scales are similar for the respective longitudes, the E-W diameter in 2002 of 11 km was smaller and varied by just 1 km between the two pairs. An air temperature rise to just below freezing (day 120) followed by a pressure decrease to 975 hpa, and cold southerly winds afterward occurred at the time of the April 2002 anticyclonic feature (Figure 4). Northerly winds are associated with warmer temperatures and force the sea ice cover (100%) together with the iceberg, southward. [35] In May 2002 (from day 140) the anticyclonic movement was ellipsoid in shape with the major E-W axis decreasing westward from 21 km to 10 km. It became more circular in shape toward the west as the N-S axis changed only slightly in extent from 14 km to 9 km. This westward change in amplitude and/or shape of circulation may be due to the transition to sea ice regimes with more compact conditions and thus higher viscosities. Vihma et al. [1996] suggested, while observing a similar second loop at 66 S in May 1992, that the radius of loops increases linearly with distance from the Antarctic Peninsula. However, in 1992 and 2002 the loop size decreased with distance from the southern Weddell Sea Ice Shelves despite the fact that atmospheric pressure systems move from the northwest to the east across the Antarctic Peninsula. [36] During the middle drift section iceberg and sea ice tracks were very similar. To evaluate the coherence of sea ice and iceberg drift, we correlated the buoy velocity components as well as changes in differential kinematic parameters (DKP) compared to sea ice concentrations. [37] The correlation between u and v velocities of iceberg and sea ice buoys (April, May, and June) was strong with proportion of variance explained (r 2 ) above 0.94 (n 90). For the v component of the entire track r 2 ranged from 0.84 to 0.89 while for the u component r 2 was 0.92 (n 212). The coherence was most pronounced during this period, during which the iceberg and sea ice buoy (8060) were separated by 34 ± 6 km (Figure 3b, bottom). [38] If the sea ice concentration around the ensemble is 100% a mechanical link exists and the drift can be expected to be coherent. Ice concentrations in the vicinity of the buoys ranged between 50% and 80% at the beginning of the drift and increased from day 87 to more than 90% until the ensemble reached the northern Weddell Sea off Joinville Island where it entered the west wind zone. From April to mid-june, when the buoy velocity correlations were high, the sea ice concentration was above 97%. This indicates that a high sea ice concentration is necessary for a coherent iceberg/sea ice motion (Figure 5), which in this case, is higher then the 90% proposed by Lichey and Hellmer [2001]. 7of14

8 Figure 3. (a) Drift patterns of iceberg (9367, white solid) and sea ice buoys 8060, white dotted; 9781, black solid; 9728, black dotted) from point of deployment in January 2002 to September Iceberg data are available until the beginning of September, and sea ice buoy data are available until the end of July. The bottom topography is derived from the GEBCO 1-min [GEBCO, 2003] data set with isopleths shown at 1000-m intervals. Abbreviations are EI, Elephant Island; JI, Joinville Island; SOI, South Orkney Islands; SOP, South Orkney Plateau; and SSI, South Shetland Islands. (b) (top) Iceberg and sea ice buoy drift patterns from April to July (bottom) Comparison between iceberg path (9367) with 24-hour sampling interval and sea ice buoy (8060) with 3-hour sampling interval for the circled area in Figure 3b (top). The black tracks represent the daily buoy positions at noon; the gray track on the right connects the 3 hourly sea ice buoy position. [39] Low and fluctuating sea ice concentrations at the beginning of the record reflected summer conditions at the ice edge while the small fluctuations in sea ice between April and June indicated the passage of synoptic lowpressure systems. The largest DKP changes, in generally low values (Figures 5b 5e), occurred during low sea ice concentrations (day 40 to 60) and were associated with the shift of buoy position within the array. After the relative positions were fixed by a more compact sea ice cover (97%), the divergence was small ( ± s 1 ). Deviations in DKP minima occurred at the end of the anticyclonic motion (day 120) as a change of array shape due to negative shear. The change, not present in divergent or rotational (vorticity) DKP moments, reflected the elongation of the array s SE to NW axis (Figure 5d). [40] Although the second anticyclonic movement (day 153) did not disturb the ensemble s progress, the entire array turned sharply toward the west. The DKP changes are seen particularly in the vorticity and shear deformation (Figures 5c and 5d), associated with a low-pressure system (960 hpa, Figure 4a), i.e., strong winds influencing the sea ice cover and, in turn, the buoy array. After the ensemble turned east south of Powell Basin, sea ice cover began to decrease below 90% (from day 175). The following cyclonic motion (day 183) was associated with a strong shear in DKP causing a new arrangement of the array with the SE-NW axis elongated and the SW-NE axis shortened (Figure 5d). [41] Our buoy array in the western Weddell Sea, with distances between individual buoys of 30 to 150 km, occupied a smaller area than, for example, the sea ice buoy arrays used by Massom [1992] and Wadhams et al. [1989]. However, the rather small DKPs associated with a uniform northward drift in the Western Weddell Sea, also found by 8of14

9 Figure 4. (a) Pressure versus time and (b) temperature versus time recorded at sea ice buoy (9781) representing atmospheric conditions during the drift through the western Weddell Sea. The solid curve in Figures 4a and 4b represents the monthly mean value. Figure 5. (a) SSMI/AMSR derived sea ice concentrations at sea ice (8060, 9781) and iceberg (9367) buoys. Associated DKP moments of the sea ice/iceberg buoy array for days 4 to 216, (b) divergence, (c) vorticity, (d) shear deformation, and (e) stretch deformation of the ensemble. 9of14

10 Figure 6. Excerpt of drift trajectories of iceberg buoys 25886, 25718, and released in Dec/00 off Neumayer Station. Array positions every 50 days are overlaid as joined triangles. Vihma et al. [1996], indicate that sea ice/iceberg drift in the western Weddell Sea occurs in a synchronous manner due to the pack serving as the mechanical glue Eastern Weddell Sea (December 2000 to February 2002) [42] Although 11 iceberg markers were deployed in the eastern Weddell Sea in December 2000, the focus in this section will be on a triangle of three iceberg buoys ( ) released in December Since in situ sea ice drift observations were not available for the area during this period, sea ice/iceberg interaction was studied using iceberg velocity correlations and DKP moments associated with SSM/I-derived sea ice concentrations. [43] The buoys were released within 100 km of the coast off Neumayer station (40 to 90 km distance between deployments) and lasted as an ensemble for more than 430 days with individual buoys surviving two winter periods. Initially, the array followed the gyre rotation westward, with a maximum speed of 34 km d 1 in the CC, before turning north (at around day 420) into the central Weddell Sea (Figure 6). [44] The drift in the CC was marked by a large variability in DKP values which were associated with an incoherent ensemble movement. At the end of summer and autumn the sea ice concentrations were low and highly variable (60% ± 30%). Air temperatures well below zero from March (Figure 7), fuelling sea ice growth, and the northward advance of ice from the southwestern Weddell Sea both contributed to the high sea ice concentration once the bergs left the CC. Shortly after the turn, the ensemble had settled into a common northwest drift locked in a sea ice cover of more than 90% (April to October). [45] Minimum DKP values of to s 1 suggest a coherent drift from April to October. Sea ice concentrations were high, and thus the common forcing field acted via the sea ice cover on the icebergs. Atmospherically forced deviations and/or reversals occurred regularly during the mean northward drift and were associated with perturbations in DKP values (Figure 8). The largest departure of the ensemble from the mean track was a 20-day NE excursion of 145 km due to a change from southeasterly to southwesterly winds with a 90 turn in the track. The rearrangement of the array (Figure 8, day 470) was caused by a passing low-pressure system (965 hpa) with warmer temperatures. At the maximum excursion a passing high pressure system (above 1000 hpa) led to the reversal of the course to a position within 50 km of the point of origin (Figure 7, day 490) and a new rearrangement of the ensemble. In both cases DKP values showed maxima with positive shear deformation, negative stretch deformation and variable vorticity. This deviation from the main Figure 7. (a) ECMWF sea level pressure versus time and b) ECMWF air temperature versus time at iceberg buoys 25886, 25718, and resembling atmospheric conditions during the drift in the eastern Weddell Sea. The solid curves in Figures 7a and 7b represent the monthly mean value. 10 of 14

11 Figure 8. (a) SSMI/AMSR derived sea ice concentrations for the drift period depicted in Figure 6 at iceberg (25886, 25718, and 25887) buoys. Associated DKP moments of iceberg buoy array for days 349 (2000) to 780 (2002), (b) divergence, (c) vorticity, (d) shear deformation, and (e) stretch deformation of the ensemble. track was far larger than the those observed in the western Weddell Sea. Departures from the main northwestern course following this event were more circular, 10 km(20 km) to 40 km(60 km) in N-S(E-W) extent with loop cycles lasting from 10 to 16 days. These scales are more in agreement with the loop scales farther west [Vihma et al., 1996]. The subsequent vorticity and negative stretch deformation was associated with a cyclonic loop in the aftermath of a lowpressure system, again associated with high temperatures (day 540). The largest DKP variation occurred just before the major 90 turn to the east (day 574) and subsequent mean eastern ensemble drift. However, while the previous DKP variations corresponded to atmospheric conditions and lower sea ice concentrations this event could not be attributed to a large pressure disturbance or lower sea ice concentration. A possible cause might be sea ice age or sea ice strength, the influences of which have yet to be analyzed. [46] Using the correlation between u and/or v buoy velocities as a measure of coherence, the relationship between percentage sea ice cover and coherent drift of icebergs (and sea ice) was not discernible from December to March. The proportion of variance in coherence explained by sea ice cover exceeded 0.8 (n > 200) for sea ice concentration above 86% from April to October, the period with minimum DKP. The correlation between u and v velocities of the icebergs was strong with r 2 above 0.95 (n 215). Although not all DKP variations could be explained by passing atmospheric pressure systems, horizontal ECMWF wind and iceberg buoy velocities were correlated with r 2 above 0.85 (n = 215), indicating that the wind acting on the compact sea ice cover indirectly controlled the iceberg motion. [47] At the end of winter, with no recent ice formation, high sea ice concentrations close to the icebergs (Figure 8a) were due to advection from the southeast. Leads within the ice pack had formed and account for the open water portion of the satellite pixel and suggest the loss of the mechanical link between adjacent bergs. The DKP variability increased (from day 660) as sea ice concentration decreased to well below 85% marking the end of the coherent iceberg ensemble motion. While the distance between buoys slowly 11 of 14

12 increased from 120 to 200 km over 200 days, it subsequently increased to 300 km within a few days. This suggests that the length scale for coherent iceberg and sea ice motion is at least 200, km which is higher than the value of 150 km found for the western Weddell Sea Freshwater Export [48] Iceberg volume and iceberg drift define the mass flux of freshwater contained in icebergs in the Southern Ocean. Iceberg meltwater is a vital component of the freshwater budget in the Weddell Sea. The fraction of the Weddell Sea s freshwater budget can be estimated by balancing the calving rate, the import from the east, and the export to the north. [49] Gladstone et al. [2001] stated a mean iceberg calving rate in the Weddell Sea of 410 Gt yr 1, and our tagged icebergs are equivalent to less than 3% of this flux. As calvings of huge tabular icebergs are infrequent, they are not taken into consideration for this estimate. For example, iceberg A43b alone, tagged in 2002, represents 88% of this rate and thus a higher calving rate can be assumed in this year. [50] Although one has to consider that only a minor portion of icebergs entering the Weddell Sea from the east were tagged with buoys, and that a significant number of bergs calving from Weddell Sea ice fronts move out of the region, we assume that the tagged bergs give a representative picture of the export patterns. The latter can be described in three scenarios: (1) complete export of tagged icebergs (Figures 2a and 2d), (2) partial export to the north with some bergs remaining in the Weddell Gyre for more than 1 year (Figure 2c), and (3) all tagged icebergs remain in the Weddell Gyre (Figure 2b). [51] In order to give a first estimate of the freshwater budget for each scenario, we assume that import from the east into the eastern Weddell Sea is compensated by export to the north, and write the equation C ¼ F þ E þ R with C the calved iceberg volume, F the volume melted in the Weddell Sea, E the volume exported from the Weddell Sea, and R the solid ice remaining in the Weddell Sea. F can be approximated as F = Cm, the product of calved volume (C) and the proportion (m) of melted to initial volume of an average sized iceberg with a melt rate r. The three scenarios correspond to the assumptions: (1) R = 0, (2) E, R > 0 and (3) E = 0. For the simplest scenario (3), the 410 Gt yr 1 are equivalent to a freshwater flux of 15.3 msv (r ib = 850 kg m 3 ) into the Weddell Sea. This corresponds to about 55% of the ice shelf basal melting and 40% of the NCEP-Reanalysis P-E excess on the continental shelf in the Weddell Sea. [52] For an estimate of the proportion of an iceberg that melts as it transits the Weddell Sea size, thickness, and traveltime were provided by the tagged icebergs. The median of observed iceberg area was A mdn = 0.26 km 2, the mean draft D = 330 m (Table 1), and the mean transit time t ws = 1 year. The iceberg melting rate r was taken from the drift/melt model based on the work by Lichey and Hellmer [2001]. The latter was improved through implementing the wave radiation force as a supplemental driving ð2þ force, and iceberg decay parameterized through basal and lateral melting [Hellmer et al., 1998] and wave erosion [Gladstone et al., 2001]. The freshwater input increased from a minimum of 4 m yr 1 at the ice shelf front to a maximum of 90 m yr 1 in the Scotia Sea near the South Orkney Islands [Schodlok et al., 2005]. Similar maximum values were estimated for the Polar Front by Morgan and Budd [1978]. The mean r in the central Weddell Sea was 25 m yr 1. This value was chosen to be representative for the inner Weddell Sea, including most of the observed tracks, and results in a melt proportion m = With a calved iceberg volume (C) of 410 Gt yr 1, the calculated freshwater input F was equal to 31 Gt yr 1. Substituting these values into equation (2), the freshwater budgets for the different scenarios were (1) E = 379 Gt yr 1, (2) E = R = Gt yr 1 (assuming a nominal 50:50 split between export and remainder), and (3) R = 379 Gt yr 1. [53] The export dominated melting in scenarios 1 and 2, and the large values for fraction R in scenarios 2 and 3 confirm that the Weddell Sea freshwater budget is profoundly influenced by the prevailing export scenario in a given year. In addition, the drift tracks obtained from this study suggest that the freshwater released is not only confined to the Coastal Current as suggested by Gladstone et al. [2001] but spread across the entire Weddell Sea. 4. Discussion and Conclusions [54] One prominent feature of the observed iceberg drift patterns is the episodic wintertime coherent movement. While the observed iceberg drift patterns showed a considerable interannual variation in speed and direction the eye is drawn to regions of coherent motion. The drift in ice-free waters mainly represents wind forcing and forcing effects of the upper m of the water column acting upon the large keel, whereas in consolidated ice the sea ice cover was dominant indicating wind forcing of sea ice indirectly steering the icebergs as suggested by Lichey and Hellmer [2001]. [55] Early observations by Vinje [1980] showed an example of three nearly parallel, northward drifting icebergs between 10 W and 20 W within 6/8 to 8/8 sea ice cover. However, with a large uncertainty in sea ice concentration surrounding the icebergs, due to the lower resolution in early satellite technology, the interpretation of sea ice/ iceberg interaction remained problematic. [56] Our results using minima in drift kinematics indicate that the sea ice concentration threshold at which coherent sea ice/iceberg motion occurs depends on the location. In the western Weddell Sea an iceberg/sea ice buoy array in combination with satellite-derived sea ice concentration gave a threshold of 93%. In the eastern Weddell Sea the threshold was found to be lower at 86% for a combination of tagged icebergs and satellite-derived sea ice concentrations. These results are in broad agreement with the threshold of 90% obtained by modeling iceberg drift [Lichey and Hellmer, 2001] but indicate that the threshold itself is sensitive to the physical properties, such as sea ice strength, sea ice age, which is greater in the western Weddell Sea, and dominant driving mechanism in a given area. [57] The general iceberg drift is in accordance with basinwide sea ice drift observations, with two distinctive patterns 12 of 14