Two University of East Anglia Seagliders were used in the southern Amundsen Sea polynya in January and February 2022 (Fig. ). They were deployed from the Research Vessel Icebreaker Nathaniel B. Palmer (NBP). This campaign (NBP22-02) was part of the International Thwaites Glacier Collaboration: Thwaites-Amundsen Regional Survey and Network (ITGC: TARSAN) and Accelerating Thwaites Ecosystem Impacts for the Southern Ocean (ARTEMIS) projects. Seaglider 579 (SG579) was deployed approximately 10 km north of DIS on 21 January 2022, completing three circuits of a box (approximately 40 km × 30 km) in front of DIS, before heading north to 73.41° S 114.24° W, where it was recovered on 21 February 2022. SG579 carried out 247 dives to the seabed, or a maximum depth of 1000 m. Seaglider 510 (SG510) was deployed in open water on 17 January 2022 at 73.81° S 112.61° W and headed south, following the eastern branch of the Dotson-Getz Trough (Fig. ), in a straight transect to 74.17° S before following the trough northwest to 73.23° S 114.34° W, then south towards Martin Peninsula, crossing the Dotson-Getz Trough, where data collection ceased on 4 February 2022. Prior to communication loss, SG510 transmitted a total of 143 dives to the seabed (up to 1000 m).

Temperature and conductivity sensors (Sea-Bird CT Sail) were integrated on both Seagliders with sampling intervals of 5 s which corresponded to a measurement approximately every 0.5–1 m (dependent upon glider vertical speed). The raw outputs were processed using the UEA Seaglider Toolbox which incorporates corrections for the thermal lag of the un-pumped conductivity cell that can produce artificial salinity spikes following . The mean difference and root mean squared difference between the processed and raw versions of salinity measurements were -0.0011 and 0.0075, respectively for SG579 and -0.0012 and 0.0123, respectively for SG510. In situ temperature and practical salinity were converted to Θ and SA, using TEOS-10 .

An Aanderaa O₂ optode was integrated on SG510 and was lag-corrected using the method of . The raw measurements were calibrated against an O₂ SBE 43 sensor mounted on the ship's Conductivity-Temperature-Depth (CTD) rosette which itself had been calibrated against discrete water samples for O₂ (Sect. ). To calibrate the glider-measured O₂, six dives were compared with eight CTD casts that were taken within 3 km of the glider profiles. Only measurements at depths greater than 500 m were compared, so that upper ocean influence of biological activity was excluded. The glider O₂ measurements were on average 7.24 µmol L⁻¹ lower than the CTD O₂ sensor measurements, so this offset was applied. This correction is small relative to the magnitude of the O₂ differences observed between the features discussed in this paper and the surrounding waters. The saturation concentration (DO_sat) of O₂ was first calculated following using in situ temperature and salinity. This was then converted to percent saturation: O2sat= DO/(DO_sat) × 100, where DO is the in situ O₂ concentration measured by the optode. For temperature, salinity and dissolved oxygen measurements from the gliders, both ascent and descent profiles are utilised.

A Lab-on-Chip pH sensor developed by the National Oceanography Centre Southampton , was additionally integrated into SG579. It had a sampling interval of approximately 10 min and determined pH on the total hydrogen scale spectrophotometrically . For pH, only measurements taken during the ascent are reported in this study due to the nature of the glider flight which was programmed to travel relatively slowly during the ascent at a vertical speed of 0.09 m s⁻¹, yielding a higher vertical resolution of pH measurements. Final pH values were corrected using in situ temperature, salinity and pressure after calculating the temperature difference between inside the sensor and in situ water following . The pH measurements are reported with an accuracy better than 0.014 and a precision of 0.001 . The pH sensor stopped recording data on 8 February 2022, after 146 dives of successful data collection.

The CTD sensor on the RVIB Nathaniel B. Palmer is a Seabird SBE 911+ that takes continuous measurements. The CTD was calibrated post-cruise and the corrections were negligible, within the reported accuracy of the sensors (0.001 °C and 0.0003 S m⁻¹). Also mounted on the CTD rosette was a dissolved O₂ sensor (SBE43) which was post processed in SeaSoft with time lag and hysteresis corrections applied. It was corrected for an offset of 3.13 µmol L⁻¹ (±0.89 µmol L⁻¹) based on comparisons between the sensor and Winkler titrated bottle data at depths greater than 500 m . Water samples for DIC and δ18O were collected from Niskin bottles mounted on the CTD rosette. Samples for DIC were collected in 500 mL borosilicate bottles following SOP1 from . They were analysed at the University of Georgia using a Single Operator Multi-parameter Metabolic Analyser attached to a UIC CM5011 CO₂ Coulometer . Accuracy is based on calibration to two daily certified reference material (CRM) runs, with the average adjustment being 0.15 ± 0.05 %. Precision based on duplicate runs of CRMs was 0.56 µmol kg⁻¹. Precision based on duplicate samples (repeatability) was 0.90 µmol kg⁻¹. Water samples for δ18O were collected from Niskin bottles mounted on the CTD rosette in 20 mL glass scintillation vials, sealed with poly-seal caps and secured with electrical tape to prevent loosening during storage and shipping. Samples were analysed at the British Geological Survey in the Natural Environment Research Council (NERC) Stable Isotope Facility using the CO₂ equilibration method with an Isoprime 100 mass spectrometer plus Aquaprep device. Isotope measurements were calibrated against internal and international standards including VSMOW2 and VSLAP2. Based on duplicate analysis, analytical reproducibility was better than ±0.05 ‰ for all samples.

Measurements for DIC from the ASPIRE campaign are those presented by . The samples were collected between December 2010 and January 2011 and calibrated against CRMs. For further information on collection and analysis see .

The depth-discrete oxygen isotope values (δ18O), along with CTD-observed absolute salinity (SA), were used to estimate vertical profiles of meteoric and sea-ice melt fractions. To do this, the net freshwater balance is partitioned between the total inputs from all meteoric sources (Fmet, precipitation and glacial discharge in the form of melting ice shelves and icebergs) and the input/extraction due to sea ice melt/formation (Fsim). Following , we use a three end member mass balance approach , employing SA and δ18O as tracers. This approach assumes that the observed SA and δ18O values in the water column have resulted from a mixture of sea ice melt, meteoric inputs, and the source water mCDW, all three of which have well-separated SA and δ18O values in pure “end member” form. Following , end members for mCDW and meteoric water were empirically determined based on the mCDW–glacial meltwater mixing line, also known as the Gade line , using all SA and δ18O observations below the winter mixed layer (Winter Water core, identified by the temperature minimum); the extrapolated salinity maximum defined the mCDW end member (34.83 SA, 0.05 ‰ δ18O), while the zero-salinity intercept defined the meteoric end member (0.0 SA, -30.32 ‰ δ18O). The sea ice end member (7.0 SA, 2.1 ‰ δ18O) is based on field observations reported by .

This study makes use of a combined austral wintertime (April to September) temperature and salinity dataset (3457 vertical profiles) obtained from 15 seal-borne conductivity-temperature-depth-satellite relayed data loggers (CTD-SRDLs) deployed in the Amundsen Sea in 2014 and 2022. This combined hydrographic dataset was collected by 9 Southern Elephant seals (Mirounga leonina) and 7 Weddell seals (Leptonychotes weddellii). The 2014 hydrographic dataset (1906 profiles) was collected by CTD-SRDLs deployed as part of the UK’s Ice Sheet Stability Programme (iSTAR) Ocean2Ice JR294 cruise in February 2014 onboard the RRS James Clark Ross . The 2022 hydrographic dataset (9 tagged seals; 1551 profiles) was collected from CTD-SRDLs deployed during the NBP22-02 cruise onboard the RVIB Nathaniel B. Palmer, as part of TARSAN.

Due to the limited transmission bandwidth of the Argos satellite system, CTD-SRDLs only send measurements from 17 to 18 pre-defined depths, transmitting only the deepest ascent in every 4 to 6 h window . This study accesses the seal-borne CTD-SRDLs profiles via the Marine Mammals Exploring the Oceans Pole to Pole (MEOP) server which applies standard delayed-mode quality control and vertically interpolates profiles to 1 m (for details see: https://www.meop.net/database/data-processing-and-validation.html, last access: 1 December 2024). Before use in this study, profiles were further calibrated against concurrent summertime ship-based hydrography and historical (1994–2020) profiles from the MIPkit-A dataset . Calibration offsets for temperature and salinity were calculated manually for each CTD-SRDL following comparisons with regional observations of mCDW properties, i.e. Θ greater than 0 °C at depths greater than 700 m. CTD-SRDL GPS coordinates were adjusted using standard repositioning algorithms derived using a Kalman smoother .

Analysis of the water masses in temperature-salinity space reveals two distinct salinity ranges where Θ is less than -1.60 °C (Fig. ). The more saline water mass (with SA greater than 34.2 g kg⁻¹) is denser with temperatures closer to the freezing line (as calculated using TEOS-10 ). The less saline water mass (with SA less than 34.2 g kg⁻¹) is the WW layer that lies immediately below the AASW (Fig. a and b), with an SA between 34.05 and 34.15 g kg⁻¹ (Fig. ). The colder and more saline water appears as isolated patches within the subsurface water column, below the WW layer (Fig. a and b). Such localised subsurface bodies of water with distinct physical and chemical properties are commonly referred to as lenses, for example meddies found in the Northeast Atlantic emanating from the Mediterranean . For the purpose of this study, this terminology will be adopted to describe these isolated, cold features. The lenses are located in the stratified layer at approximate depths of 200 to 500 m between the warmer of the two temperature minima of WW and the warmer deeper mCDW, the deepest water mass (Fig. a and b). The mCDW has Θ above 0.35 °C and SA greater than 34.72 g kg⁻¹ (Fig. ), which agrees well with previous in situ observations in the Dotson region .

In total, ten lenses are identified from the glider transects (Fig. ). The lenses vary in apparent size, with the largest horizontal extent, as captured by the glider, reaching a maximum of 25 km and the smallest being approximately 5 km. All of the lenses are in locations where the total water column depth is greater than 500 m and, with the exception of C and F, all of the lenses are located in water with depths greater than 600 m. Two of the lenses (A and B) are located at approximately 73.9° S, within 25 km from each other and were captured 8 d apart (Fig. a and b). We argue that A and B are the same lens and the differences in size could be due to the glider crossing either the centre of the lens, or an edge. This is supported by the comparison of the core lens properties to the extremities. The Θ of lens A increases towards the edges to -1.60 °C, similar to the core (coldest) value of lens B. Thus, the core Θ of lens A is better representative of the properties of the lenses than that of B, which is also supported by the properties of the other identified lenses shown in Table .

As the gliders only surveyed the eastern branch of the Dotson-Getz Trough (Fig. c), CTD casts covering a wider area (Fig. ) were also investigated for the presence of lenses, allowing further biogeochemical analysis of the lens water masses. The CTD casts with lenses were identified as water masses with Θ less than -1.70 °C at a depth greater than 200 m. These thresholds were chosen as all of the glider lens water masses had temperatures less than -1.70 °C (Table ), apart from lens B which will be explored later. The CTD casts with lenses detected were: 8, 10, 16, 18, 231, 279 and 294 (Fig. ). The casts were located in similar regions to the lenses identified by the gliders, in the proximity of the eastern branch of the Dotson-Getz Trough (Fig. ). No lenses were found either near the Dotson or Getz ice shelf fronts, or along the shallow bank north of Bear Island where there were CTD casts. This suggests that the formation of lenses observed in this campaign was through a mechanism local to the area around the eastern branch of the Dotson-Getz Trough. The CTD casts show that the lenses also appear further north in the trough (between 73.0 and 73.5° S), where the two Seagliders did not survey. Despite being constrained to the trough area, the location of the lenses varies, from the deepest part where the water depth is 900 m to the shallower trough edge with a water depth of 550 m (Figs. c and ).

Three lenses (D, J and CTD 279) are located at approximately 73.43° S, within 5 km from each other, within a 19 d period (Figs. a and ). This suggests that it is the same lens, although there is a significant difference in size between D and J (approximately 20 km). Lens D was surveyed by SG579 just before recovery and therefore the coverage of lens D was cut short. Additionally, the two gliders were not following the same path, which means different parts of the lens were likely captured. The lens identified in CTD 18 is in approximately the same location as lens E, three hours later. Since the CTD is a single profile, it is not possible to determine the size or shape of the lens captured. The occurrence of these repeatedly surveyed lenses indicates that lenses are not short-lived features on the timescale of hours or days. The comparison between lenses in similar locations suggests that the size variation between surveys of the same lenses is an artefact of how the lens was surveyed.

The end members of WW and mCDW and the core properties of each identified lens are shown in Table . These have been determined using both the glider and CTD measurements, where measurements of a variable exist for each. For mCDW and WW, the associated biogeochemical properties are those that align with the Θ and SA end members. For the lenses the core is defined as the temperature minimum. This was checked using the standard deviations for the properties in all of the lenses (Table ) which were all small. With the exception of lens B, all of the identified lenses have a core Θ less than -1.70 °C. The mean core Θ of all of the lenses (excluding B) is -1.80 °C which is 0.05 °C lower than the end member of the overlying WW and 2.51 °C lower than the mCDW end member. The core of the lenses are also more saline and denser than WW, with a maximum difference of 0.23 g kg⁻¹ and 0.16 kg m⁻³, respectively. Conversely, mCDW is up to 0.51 g kg⁻¹ saltier and 0.30 kg m⁻³ denser than the core of the lenses.

Measurements of pH show that the core of lens A had a pH 0.01 less than overlying WW (Table ) and greater than deeper mCDW by 0.08. The lowest pH is found in mCDW as it is an old water mass that has remained isolated from the surface for an extended period of time, so DIC has accumulated in it through remineralisation. In WW, pH is higher than in mCDW as there has been recent surface exchange with the atmosphere, during a recent winter. As the lenses exhibit a pH more closely resembling that of WW it could be assumed that they have been at the ocean surface at a similar time. Concentrations of O₂ in the core of the lenses range between 278–293 µmol L⁻¹ (Table ) with a mean of 291 ± 2. They are similar to, or slightly higher than, O₂ concentrations in WW (281 µmol L⁻¹) and much higher than in mCDW (195 µmol L⁻¹) as O₂ has been consumed in mCDW through remineralisation and there has been no exchange with the surface to replenish depleted O₂. As with pH, the similarity between O₂ in the lenses and WW suggests recent interaction with the atmosphere. The slight increase from WW to the lenses could be a result of colder water being able to absorb more O₂ from the atmosphere during ventilation, however, the O₂ saturation is only 2 %–3 % higher than WW in more than half of the lenses that have O₂ measurements.

The differences between the lenses captured by both CTD casts and gliders (i.e. D, J and 279 as well as 18 and E) could be due to the fact the CTD casts are a single profile and do not extend horizontally. The temperature minimum in each single CTD profile is considered the core temperature of the lens, whereas the glider profiles reveal that the lenses can be as large as 25 km horizontally. The glider profiles also show that the lens properties can vary between the core and extremities. This highlights a limitation of lower resolution CTD sampling.

Lenses observed in each CTD cast are located at different depths, with the Θ minimum found within the range of 240 and 405 m (Fig. ). The varying profile shapes within the lenses possibly reflect the same feature observed by the glider transects, where the difference in lens shape and size depends on which part of the lens was captured by the single profile. Three of the CTDs with lenses (Fig. ) were additionally sampled for δ18O and DIC as shown in Fig. d and e. For CTD 8 the δ18O decreases from -0.45 ‰ at 200 m in WW to -0.50 ‰ at 300 m within the lens. For CTD 18 the decrease in δ18O is from -0.40 ‰ at 200 m (WW) to -0.48 ‰ within the lens at 400 m. For CTD 279 the decrease is from -0.47 ‰ at 245 m (WW) to -0.55 ‰ at 345 m, just below the lens. In the WW (depths between 100–200 m) δ18O is approximately 0.5 ‰ more depleted than in mCDW (depths greater than 450 m). Although there are only seven data points of δ18O within the lenses and sample distribution with depth is sparse, these results suggest that the δ18O values of the lens cores are more depleted than those of WW (Table ).

The effect of sea ice formation and melt on δ18O is small compared with meteoric inputs, whereas the effect on salinity is large. The derived freshwater contributions of meteoric input and sea ice formation, calculated from δ18O, allow us to identify the processes affecting the lenses. The freshwater fractions (Fig. ) reveal a negative sea ice contribution to the lenses which indicates net sea ice formation, greatest for CTD 18 and least for CTD 8. There is no δ18O measurement at the depth within the lens from CTD 279, but the measurement just below the lens also shows a negative sea ice melt contribution. Water above the lenses, especially close to the surface, has a positive sea ice meltwater contribution, indicating net sea ice melt which has occurred locally in the spring.

DIC concentrations in the lenses are slightly higher than in WW, by up to 10 µmol kg⁻¹ (Table ). Similar to pH and O₂, mCDW has the highest DIC concentration (2261 µmol kg⁻¹) as it has spent the longest time isolated from the atmosphere. Additionally, DIC accumulates in the water column due to the breakdown of organic matter which produces DIC.

The CTD casts with a lens are compared with a CTD profile where a lens was not observed (Fig. ). For temperature, salinity, O₂ and δ18O, the comparison CTD cast (CTD 194) was in the same location as the lens CTD cast (CTD 18), 20 d later (Fig. ). No additional CTD casts were sampled for DIC in the region of the Dotson Trough during the TARSAN-ARTEMIS campaign. Instead, two CTD casts were sourced for comparison from the ASPIRE campaign in December 2010–January 2011 , within 25 km of CTD 18 (Fig. ). Three consecutive SG579 dives immediately after the glider exited lens A were selected for a comparison of pH profiles in the lens with pH profiles in surrounding water.

Temperature and salinity are lower in the lenses than in nearby water at the same depth (Fig. ) by up to 1.50 °C and 0.16 g kg⁻¹, respectively. The dissolved O₂ concentration and O₂ saturation is higher in the lenses than in surrounding water (by approximately 50 µmol L⁻¹ and 11 %, respectively) indicating the ability of cold water to dissolve more O₂. The δ18O in the lenses is approximately 0.18 ‰ lower than surrounding water. The lens has DIC concentrations very similar to one of the ASPIRE CTD casts (CTD 9) but approximately 18 µmol kg⁻¹ lower than the other (CTD 72, collected about 2 weeks later in the season and over a deeper part of the trough). The lower DIC concentration is consistent with pH which is 0.02 higher in the lens than in surrounding water at the same depth.

Evidence of deep convection from the surface to depths less than 400 m or to the seabed (if shallower than 400 m), as indicated by seal profiles, was observed on the shallow shelf around Martin Peninsula and to the east of the Dotson-Getz Trough (Fig. a), plus very occasionally in deeper water across the trough. Temperatures within these profiles reach a minimum of -1.85 °C at depths of approximately 300–400 m (Fig. b). These temperatures are associated with SA greater than 34.25 g kg⁻¹, consistent with the properties of the lenses.