Seawater salinity sample measurements from the Antarctic Circumnavigation Expedition (ACE).

***** Dataset abstract *****

This data set contains salinity measurements from discrete seawater samples that were collected in the Southern Ocean (south of 30deg S) during the Antarctic Circumnavigation Expedition (ACE). 657 samples were collected during the period December 24th, 2016 and March 18th, 2017 in the Southern Ocean from the surface ocean using the ship's underway line (UW; 328 samples) and in vertical profiles using Niskin bottles mounted on the CTD rosette (273 samples). A few additional samples (56) were collected from a parallel cast with a trace-metal rosette, with a bucket, and as duplicates to ensure data quality. All samples were analyzed for their salinity and results are reported on the Practical Salinity Scale 1978 (PSS-78; Sea-Bird Electronics, Inc., 1989). Measurements were performed on a Guildline Autosal Laboratory Salinometer 8400(B) at CSIRO (Hobart, Australia) for samples collected during leg 1, and on a OPTIMARE Precision Salinometer (OPS) at the Alfred Wegener Institute (Bremerhaven, Germany) for samples collected during legs 2 and 3. This circumpolar data set provides insights into the hydrological cycle of the Southern Ocean and the processes (precipitation, evaporation, sea-ice melting and freezing, ice-berg and land-ice melting) that determine the salinity of a certain water mass. It is being used to calibrate the CTD sensor vertical profiles (Henry et al., 2020) and thermosalinograph sensor underway measurements (Haumann et al., 2020) from the ACE cruise.

***** Original data collection *****

Seawater samples for the analysis of their salinity (reported as PSS-78; Sea-Bird Electronics, Inc., 1989) were collected during the Antarctic Circumnavigation Expedition (ACE) with the ship Akademik Tryoshnikov. The cruise consisted of three legs (1 to 3). The first sample was taken on 24-Dec-2016 12:00:00 UTC and the last sample on 18-Mar-2017 09:54:00 UTC. The northern and southern most samples are from -34.8459 deg N and -74.0053 deg N, respectively. Samples were collected circumpolar, spanning a longitudinal range between -179.9517 deg E and 177.2012 deg E.

For the sampling each salinity bottle (200 ml OSIL bottles made of clear type II glass) was filled by 1/10 with seawater, shaken vigorously for 10 seconds and flushed over the plastic seal cap. This procedure was performed three times and subsequently the bottles were filled with the seawater. Samples were filled to just below the shoulder of the bottle, were capped with a plastic seal and with plastic screw caps. Due to a malfunctioning of the salinometer on the ship, samples were processed on shore (see below), which forced us to use additional bottle types during the cruise. Parts of the leg 2 samples were collected in the original salinity bottles and parts of the samples in drinking bottles similar to “coca-cola” bottles and sealed with a normal plastic screw cap only. For leg 3, we borrowed additional salinity bottles from the Scripps Institute of Oceanography that were used throughout leg 3 (except for a few samples in the beginning).

Samples were collected from the underway line in the CTD wet lab and from the Niskin bottles on the CTD rosette (Henry et al., 2020) for most CTD stations at about seven to 12 different depth levels. We also took samples during CTD stations from the underway line, and often these were taken in conjunction with surface samples from the CTD rosette. These duplicates have been used to test the water quality from the underway line (see below). From the underway line, we collected 68 samples during leg 1 (Indian sector), 140 samples during leg 2 (Pacific sector), and 120 samples during leg 3 (Atlantic sector). From the CTD Niskin bottles, we collected 58 samples during leg 1, 135 samples during leg 2, and 80 samples during leg 3. A total of 56 additional samples were collected during legs 2 and 3, of which 38 are duplicates from the CTD Niskin bottles, and 12 are duplicates from the underway line. 5 additional samples were collected from different depth levels of the trace metal rosette (TMR; event/sample number 2653) that was performed right after the CTD cast 6 (event/sample number 2652) at station number 92 on leg 3 for later bottle intercomparison between the two sampling methods. 1 additional sample stems from a bucket taken from the surface ocean in Cumberland Bay (South Georgia) on leg 3.

The depth levels to which the CTD casts were performed were determined to reduce the time spent at each station and to represent the upper ocean water masses. Therefore, casts typically range from shallow/fast casts of a depth of 400 m to deeper casts of about 1000 m south of the Polar Front (where the deep waters are shallower) and 1500 m north of the Polar Front (to capture most of the mode and intermediate water layers). The spatial distribution of the stations was previously selected for a good representation of all water masses on either side of the frontal systems and in between the fronts, as well as to represent an approximate equal spacing. Due to technical issues and weather conditions only 44 of the originally planned 95 CTD stations were carried out of which our project sampled 39. On several occasions one station consisted of multiple casts. During leg 1 (Indian sector), we sampled 10 casts at seven stations. During leg 2 (Pacific sector), we sampled 20 casts at 22 stations, plus 7 casts sampled by another project contained in ‘other’. During leg 3 (Atlantic sector), we sampled 10 casts at 10 stations that were carried out, out of 16 planned CTD stations. Four of the six cancellations of CTD stations during leg 3 were due to weather conditions. The five deep stations that sampled Circumpolar Deep Water (CDW) in the lower most bottles were complemented by five shallow stations that went down to 400 m to capture all surface water characteristics, i.e. summer and winter water layers and sometimes upper CDW. Further information on the cruise and sampling can be found in the cruise report (Walton and Thomas, 2018).

All sampling locations are taken from the ship’s GPS record for a given sampling time (Thomas and Pina Estany, 2019). Each salinity sample has been paired with a respective sample for analysis of the seawater oxygen isotopic composition (Haumann et al., 2019).

All 126 samples from leg 1 were off-loaded in Hobart for analysis and the empty bottles were re-used for leg 2. 519 samples from legs 2 and 3 stayed on the ship and were processed at AWI in Bremerhaven, Germany. 8 duplicate samples that were processed on the repaired ship’s Autosal during leg 3. All samples that stayed on the ship were shaken up in the middle of leg 3 and again during leg 4 to prevent the salt from precipitating out of the sample. While shaking the samples, we realized that 5 samples were leaking because they were missing the seal caps. We opened the samples carefully, sealed them, and closed them again during leg 3. These samples were LOCEAN 30 GR, LOCEAN 36 GR, and those with the event numbers 1354, 1471, and 1553. Consequently, the results from these samples should be treated with care.

***** Data processing *****

Salinity samples were originally supposed to be processed on board the Akademik Treshnikov using the ship’s Guildline Autosal 8400B Salinometer. However, on leg 1 we found that the salinometer was probably not serviced and aligned since 2010 and in a bad condition. It was partially serviced in Hobart, but major issues were still reported on leg 2. Therefore, salinity samples were only collected but not processed on the ship, increasing our demand in salinity bottles. Samples from leg 1 were processed at CSIRO in Hobart, Australia. Samples from legs 2 and 3 stayed on the ship and were processed at AWI in Bremerhaven, Germany (see above). By the end of leg 3, after considerable troubleshooting, testing, and closely monitoring the room temperature, we could achieve reasonably good measurements with the Autosal on the ship and it seemed reasonable to process salinity samples on the ship’s Autosal. However, samples could not be processed anymore in time. Thus, we only processed 8 duplicate samples. 2 of these duplicate samples were from the underway line (at event 1796) and 2 from the CTD Niskin bottle (15 m, event 1901) at the end of leg 2. All these duplicates were collected in “coca-cola” bottles, as were the original samples. During leg 3, we collected 2 duplicates (1 “coca-cola” bottle and 1 normal bottle) from CTD Niskin bottles at 400 m (event 2965) and another 2 duplicates (1 “coca-cola” bottle and 1 normal bottle) from another cast at 1000 m (event 2968). These 4 duplicates were also processed on the ship. An additional set of each of these salinity duplicates (1 “coca-cola” bottle and 1 normal bottle each) from leg 3 was processed at AWI in Bremerhaven, Germany for comparison (see ‘Quality checking’).

All 126 samples from leg 1 were processed on a Guildline Autosal Laboratory Salinometer 8400B (SN 71611) by Kendall Sherrin at CSIRO in Hobart on January 19th and 20th, 2017. The reference material was OSIL IAPSO Batch P158 (K15: 0.99940, PSU: 34.9880). During the analysis, the lab temperature was held between 21.0 degC and 23.5 degC and the bath temperature was 23.996 degC. Prior to the analysis samples were stored in the Salt Room for 4-6 hours to reach 22 degC.

8 samples were processed on the ship’s Guildline Autosal Salinometer 8400B (SN 70231) by Alexander Haumann on March 16th, 2017. The reference material was OSIL IAPSO Batch P160 (K15: 0.99983). Prior to the analysis all samples were stored in the analysis room for one day. During the analysis the room temperature was between 27.1 and 27.5 degC and the bath temperature was set to 27 degC. A standard was run before and after the analysis. All standard and sample measurements produced 3 stable readings.

All remaining 529 samples from legs 2 and 3 have been evaluated with an OPS (Optimare Precision Salinometer) S/N 003 by Gereon Budéus at the Alfred Wegener Institute in Bremerhaven. Each single measurement session included typically 18 to 20 bottles. At the start of each session, a standard seawater (SSW) control bottle OSIL IAPSO Batch P160 (K15: 0.99983) is used. The reading of this bottle is adjusted to S=34.9930. This additive correction is used throughout the session. At the end of the sessions a second standard seawater check is done. This serves as an indicator for contamination or other problems. A significant lower reading at the end check indicates problems. Evaporation is insignificant over the session period.

The OPS is an automated system. Rinsing, blind measurements and valid measurements are performed automatically from one bottle. A single measurement cycle consists of 50s pumping and 10s pausing. Rinsing is set to 1 cycle, 3 blind measurements follow, and then 3 or more single measurements follow. A measurement set for one bottle is regarded as valid if 3 consecutive single measurements show readings within a range of deltaS = 0.0003. This is a restrictive requirement which is relevant for related work on sea water density. If readings vary by an amount which is larger but still acceptable for salt accuracy, the instrument has been stopped manually. The need to do so depends mainly on the bottle type. All measurements have been made at 20.000 degC +/- 0.3 mK. The sample bottles have been carefully cleaned from salt residuals before further treatment. All samples have been slightly degassed to allow for precision measurements. Warming to 30 degC for 1 hour, overpressure release and subsequent cooling to room temperature (20 to 24 degC) preceded measurements on the following day. This procedure prevents the formation of microbubbles and is mandatory for high precision work.

The quality control shows that the additive correction reveals a trend during the first 6 sessions. This is not usual and suggests contamination either already present in the instrument before the sessions or introduced by early samples. While results within each session are sufficiently consistent, the situation is not ideal. The SSW end check values, being lower than the start check, also indicate dirt of some kind. It was therefore decided to do a complete cleaning and recalibration before evaluations were continued. The succeeding sessions show insignificant variation of the SSW correction and also insignificant differences between SSW start and end check.

There are three types of bottles: Cola-style screw cap bottles, Meplat (OSIL) bottles with screw cap and inner plug, and square (Scripps) bottles with screw cap and inner plug, each type with specific issues. The bottleneck of the Cola-Style bottles could not be rinsed before opening them. The screw cap includes an insert for tightness. Screw caps were solidly torqued, so there seems little danger of evaporation. But there is no idea whether or not salt crystals formed under the screw caps and whether these could possibly fall into the bottles. With respect to degassing and homogenizing, the Cola-style bottles were the easiest to handle: their content had inevitably to be migrated into our standard borosilicate infusion bottles, because the bottles are simply too large to fit in any of the instruments. Thus, degassing and filling level were well controlled, as is usual for our standard bottles. The three consecutive measurements from each bottle therefore cover typically a negligible range (0.0000 or 0.0001 in salinity).

The majority of the Meplat bottles were new, and cleaning of the inner plug as well as degassing was possible. The inner plugs were tight, so rinsing the bottle before opening was no problem. All of them had salt crystals underneath the screw cap, so cleaning was definitely necessary. An issue is the form of the inner plugs, as these were not of the smooth type but had three lamellae. Water is trapped between these lamellae – and exactly which kind of water is unclear. The bottles were unplugged and measured though some drops from the inner plug went back into the bottle. As a consequence, the range covered by the three measurements per bottle is mostly larger. Measurements were accepted if the range did not exceed 0.001 or 0.002 in S. Most results were better.

The square bottles were the most difficult ones to measure. When opening the screw cap, many of them lost their inner plug immediately, many only a few seconds later. These inner plugs are not tight and it must have been already difficult to rinse them after filling in sea water samples. It was often impossible to control the salt deposits beneath the screw cap because the inner plug just went off. The size tolerances of bottlenecks and/or inner plugs are obviously too large. The shape of the bottlenecks is also adverse: the bottlenecks are slightly V-shaped towards the upper end, giving rise to an outward pressure on the inner plug. Therefore plugs may also jump off after some time. Unfortunately, these bottles were also difficult to homogenize. Whether this is due to their angled design or to an unfavorable height/base area ratio is unclear. In the end, their contents were refilled into standard borosilicate infusion bottles. Measurement ranges of the latter were negligible as usual, while those from the square bottles had simply to be accepted after four or five measurements. The ‘deviation cloud scatter’ from these samples will certainly be largest.

***** Quality checking *****

Any issue associated that occurred either during sample collection, storage, or analysis is reported through the sample flag ‘SALNTY_FLAG_W’ (see https://exchange-format.readthedocs.io/en/latest/quality.html#water-quality-codes). These flags also include outliers that were identified by comparing the salinity to the underway thermosalinograph (Haumann et al., 2020) or CTD sensor salinity (Henry et al., 2020). 

Accuracy of Autosal at CSIRO (leg 1 samples) and the OPS (leg 2 and 3 samples) are reported to be better than +/-0.001 PSU. The estimated measurement uncertainty is reported in the data set as the variable ‘SALNTYERR’. Uncertainties from sampling, bottle type, caps, and bottle handling will be the major source of errors. In order to estimate these, we performed a number of comparisons.

The sampling error is estimated by comparing the sample salinity to calibrated senor salinity (Henry et al., 2020; Haumann et al., 2020) from the thermosalinograph for underway samples (Figure 1) and the CTD for the Niskin bottle samples (Figure 2). The RMSE from the difference of the underway samples is 0.005 PSU and the one from the Niskin bottles is 0.001 PSU (when excluding samples with a flag larger than 2). The larger uncertainty in the surface samples most likely arises from a larger variability of salinity in the surface ocean and larger uncertainty in the exact timing when the sample was taken and the measurement by the thermosalinograph. Comparing samples taken from the underway line within one hour of the sample taken from the upper most Niskin bottle (shallower than 20 m) from the CTD cast (Figure 3) results in an RMSE of 0.008 PSU (22 comparison points for sample flags smaller than 4). The larger RMSE for the latter comparison most likely results from the larger spatial and temporal distance of the water sampled by the CTD and underway system compared to the comparison between the samples and the corresponding sensor. Figure 3 also shows no systematic bias between samples from the surface Niskin bottles and the underway line, showing that there is little or no effect of potential remaining rust particles in the underway system (Walton and Thomas, 2018).

Duplicate samples that were taken from the same source (Figure 4) show an RMSE of 0.004 PSU (19 samples with flags smaller than 4), of which 12 pairs are from the underway line with an RMSE of 0.006 PSU and 7 pairs are from the Niskin bottles with and RMSE of 0.003 PSU. 10 pairs of duplicates that were measured on both the ship’s Autosal and the OPS at AWI resulted in an RMSE of 0.002 PSU (Figure 5), indicating that storage time was not an issue and that the ship’s Autosal was providing reasonable results by the end of the cruise. Similarly, a comparison of 8 pairs of samples from different types of bottles reveals an RMSE of 0.002 PSU (Figure 6), illustrating that sampling with different bottle types did not influence the results.

***** Standards *****

All samples are analyzed using standard seawater from OSIL IAPSO of either Batch P158 or Batch P160 and are reported on the Practical Salinity Scale 1978 (PSS-78; Sea-Bird Electronics, Inc., 1989).

***** Dataset contents *****

Processed Data
- ace_18_data_salinity_ctd_20200812.csv, text format; contains salinity measurements of seawater samples collected from the Niskin bottles mounted on the CTD rosette
- ace_18_data_salinity_uw_20200812.csv, text format; contains salinity measurements of seawater samples collected from the underway line
- ace_18_data_salinity_other_20200812.csv, text format; contains salinity measurements of miscellaneous samples: Duplicate seawater samples; seawater bucket sample from Cumberland Bay, South Georgia; seawater samples from Niskin bottles mounted on the trace-metal rosette.

Metadata
- data_file_header.txt, metadata, text format
- README.txt, metadata, text format
- figure*.pdf, metadata, portable document format

***** Dataset contacts *****

F. Alexander Haumann, Princeton University, USA; British Antarctic Survey, UK; ETH Zurich, Switzerland. ORCID: 0000-0002-8218-977X. Email: alexander.haumann@gmail.com
Jenny Thomas, Swiss Polar Institute, Switzerland. ORCID: 0000-0002-5986-7026. Email: jenny.thomas@epfl.ch, jen@falciot.net

***** Acknowledgements*****

This research was supported by Swiss National Science Foundation (SNSF) grant numbers PZ00P2_142684, P2EZP2_175162, and P400P2_186681, and a grant from the BNP Paribas Foundation. I.G. thanks FCT/ MCTES for the financial support to CESAM (UIDP/50017/2020+UIDB/50017/2020) through national funds. We thank Kendall Sherrin (CSIRO) for processing the leg 1 samples, and Susan Becker and Lynne Talley (Scripps) for lending us their sampling bottles. We thank Stefan Vogel for the Autosal cleaning/checking and logistics in Hobart and Gwenael Renard for assistance with samples/cargo handling in Bremerhaven. We thank Yvonne Firing and Rob Craft (NOC) for providing Autosal training. It would have not been possible to assemble this data set without their help. The Antarctic Circumnavigation Expedition was made possible by funding from the Swiss Polar Institute and Ferring Pharmaceuticals.

***** Dataset license*****

This physical and biogeochemical oceanography dataset is made available under the Creative Commons Attribution 4.0 License (CC BY 4.0) whose full text can be found at https://creativecommons.org/licenses/by/4.0/

***** Dataset citation *****

Haumann, F. A., Leonard, K., Budéus, G., Meredith, M. P., Gorodetskaya, I. V., Hutchings, J., Stammerjohn, S., Tsukernik, M. and Thomas, J. (2020). Seawater salinity sample measurements from the Antarctic Circumnavigation Expedition (ACE). (Version 1.0) [Data set]. Zenodo. doi:10.5281/zenodo.1494924