Physical and biogeochemical oceanography data from Conductivity, Temperature, Depth (CTD) rosette deployments during the Antarctic Circumnavigation Expedition (ACE)

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

This data set contains measurements from various sensors mounted on the Conductivity, Temperature, Depth (CTD) rosette that was deployed in the Southern Ocean during the Antarctic Circumnavigation Expedition (ACE). 63 CTD casts were carried out during three legs in the period 21st December 2016 to 16th March 2017, including one test cast and one failed cast, for which no data is available. Data include temperature, salinity, pressure, dissolved oxygen, oxygen saturation, chlorophyll-a concentration, backscatter, and photosynthetically active radiation (PAR) and reported are also the computed variables density, depth, and sound velocity. All data has been quality controlled and post-cruise calibrated, except for the oxygen data. Data is provided at 1 dbar pressure intervals for the up- and down-casts separately and as a merged bottle file when Niskin bottles were closed. This circumpolar data set provides insights into the circumpolar hydrography and biogeochemistry of the Southern Ocean during one austral summer season.

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

A rosette was deployed for 63 casts at 46 stations in the Southern Ocean during the Antarctic Circumnavigation Expedition (ACE). The SBE 23 carousel held 23 10-litre Niskin bottles, one 8-litre Niskin bottle, an SBE 19plus CTD unit and a set of sensors measuring temperature, conductivity, pressure, fluorescence, turbidity, photosynthetically active radiation distance to the seabed (altimeter) and dissolved oxygen (see ctd_sensor_list.pdf for details of when each sensor was deployed on the rosette).

Profile and bottle-firing data were recorded by an SBE 11plus (V5.2) deck unit in the CTD dry laboratory for the downcast and upcast using Seasave software (version unknown; Sea-Bird Scientific) on a computer operating Windows XP. Data were subsequently copied to Network Attached Storage for back-up. Corresponding water samples were taken to measure a number of properties at various depths (see ace_ctd_water_sampling.csv). 

While preliminary data was processed on board, the final post-processed and calibrated data was produced later on shore. For more information about this see the data processing and quality-checking sections below.

Data files are provided for all but two casts (leg 1, casts 0 and 1) where the data files are not available. For another two casts only either the downcast (leg 1, cast 2) or upcast (leg 2, cast 8) data were recorded. Therefore, the extracted bottle data only contains 60 stations since recordings from one upcast (leg 1, cast 2), when bottles were fired, are missing. Communication problems between the CTD and deck unit occurred on three casts. Bottles leaked or failed to close during a number of casts. Full details of these problems can be seen with the deployment information in the file ace_ctd_deployment_summary.csv. 

***** Data processing *****
	 	 	 	
Processed files:
During the data processing, two sets of files were generated: one set of intermediate processing undertaken in SeaBird (see details of processing below and files in the folder Sea-Bird_Processed_Data) and one final set of files for the complete sensor data (ace_ctd_CTD20200406CURRSGCMR). A merged bottle file extracted from the sensor data (ACE_BOTTLE20200406CURRSGCM_hy1.csv) is also provided.

SeaBird processing
A series of processing modules and file manipulation modules were applied within the Sea-Bird Scientific SBE Data Processing Software v7.26.7. A brief description of each processing step is provided here, and more details can be found in the software manual Seasoft V2: SBE Data Processing Software Manual for SBE Data Processing Software v7.26.8 and later (note that the manual was downloaded at time of software download despite version mismatch). To allow users to trace the impact of the processing modules on the data, at the completion of each step a suffix is added to the output filename to indicate the processing modules which have been applied to the data and are noted in the descriptions below. The output files from each processing step can be found in the data folder Sea-Bird Processed Data. 

Data Processing Modules
Data Conversion (no file suffix): Converts raw data (.hex files) to engineering units producing .cnv and .ros files. The process utilises CTD cast specific configuration files (.XMLCON) files. Post-cruise calibration of the temperature sensor (SN 6146) and conductivity sensor (SN 4264) are applied through a sensor offset (for temperature SN 6146) and sensor slope (for conductivity SN4264) in the .XMLCON configuration files which were modified from the original configuration files created during deployments. The modified .XMLCON files are available in the folder Updated_XMLCON_config_files_20190721. See section on post-cruise calibrations and sensor corrections below for more information. Note that for file ACE201601_002 no .ros file was created.

A note on oxygen: During the data conversion process, the ‘tau’ and ‘hysteresis’ corrections were applied to the oxygen data. The ‘tau’ correction improves the response of the measured signal in regions of large oxygen gradients, and also may amplify residual noise in the signal (CalCOFI CTD Data Algorithms). The ‘tau’ correction computes the derivative of the oxygen signal with respect to time with a user-input window size), using a linear regression to determine the slope. The correction was applied using the Data Conversion step where the module uses a window looking backwards in time and a window size of two seconds was defined. Although it is recommended that the ‘tau’ correction be applied during the ‘Derive’ module for more accurate results, a subset of files were processed using that processing order and there was no difference in the derived oxygen values. The hysteresis correction corrects the oxygen voltage values for changes in membrane permeability as pressure varies. Correction coefficients stored in the XMLCON file are applied with calculations based on the current pressure and how long the session spent at previous pressures. Two oxygen variables are provided: (1) dissolved oxygen concentration in micromols/kg and (2) dissolved oxygen saturation, the theoretical saturation limit of the water at the local temperature and salinity value, with local pressure reset to zero (1 atmosphere) (See SBE Data Processing Software Manual for SBE Data Processing Software v7.26.8 & CalCOFI CTD Data Algorithms). Dissolved oxygen saturation is calculated using the equations of Garcia and Gordon (1992) which is based on the equation of Weiss (1970) but reduced the error in cold waters and presents improved performance in waters between -5 to 50 degrees Celsius. 

Wild Edit (file suffix _WildEdit): Marks bad temperature (SN 6146 and SN 5307) and conductivity (SN 4264 and SN 3793) variables with a corresponding flag. Performs two passes over data for each sensor. During the first pass, values are divided into blocks of 100 values and data points which are +- 2 standard deviations from the mean block value are flagged with bad flag. During the second pass, values are again divided into blocks of 100 values, data points which were flagged with bad flag during the first pass are excluded from the calculation of the mean block value in the second pass and data points which are +-10 standard deviations from the mean are flagged with bad flag.

Cell Thermal Mass (file suffix _CellTM): Makes corrections for the thermal mass of the cell, in an attempt to minimise salinity spiking in steep vertical gradients due to temperature/conductivity mismatches. The constants used were: thermal anomaly amplitude α = 0.03; thermal anomaly time constant 1/β = 7.

Filter (file suffix _Filt): Low pass filter applied to pressure sensor data points (filter B: time constant 0.15 sec) and conductivity sensor data points (filter A: time constant 0.03 sec) to smooth high frequency noise. 

Loop Edit (file suffix _Loop): Marks scans (or data points) with bad flag where there is a pressure slow-down or reversal to flag bad values from ship heave or during initial surface soak. If pressure velocity is < 0.25 m/s or pressure value is less than the previous maximum pressure value, it is flagged with bad flag. Surface soak flags are not applied in this step.

Derive EOS-80 Practical Salinity (file suffix _Der): “Derive” uses pressure, temperature, and conductivity from the input .cnv file to compute oceanographic variables, density sigma-theta (kg/m3), salinity (primary, psu), salinity (secondary, psu). Note that this step produces duplicate practical salinity data variables (Sal00 and Sal11; columns 25 and 26 or similar etc.) which are derived using the same formula as for the initial practical salinity Sal00 and Sal11 variables (columns 5 and 8 etc. or similar; See the SBE manual for more information). These derived variables are calculated using the Equation of Seawater 1980 (EOS-80), which is a function of practical salinity PSS-78. See section below on calculation of depth and Appendix V of the Sea-Bird Processing Manual for more information.

Bin Average (file suffix _BA): Averages data for each variable to bin size = 1 using the pressure variable.

Bottle Summary (no file suffix): Reads in .ros file and uses CTD cast specific configuration .XMLCON files to write a bottle summary file containing bottle information (position, datetime), derived variables (density1 (Density00), density2 (Density11), sound velocity (SvCM), sound velocity2 (SvCM1), practical salinity1 (Sal00), practical salinity2 (Sal11) and averaged variables (pressure (PrDM), depth (DepSM), temperature1 (T090C), temperature2 (T190C), conductivity1 (C0S/m) conductivity2 (C1S/m), salinity1 (Sal00) salinity2 (Sal11), oxygen concentration (Sbeox0ML/L and Sbox0Mm/Kg), oxygen saturation (OxsolMm/Kg and OxsatMm/Kg), fluorescence (FlECO-AFL and FlSP), PAR (Par) and backscatter (TurbWETbb0). As in the “Derive” module, this processing step produces duplicate practical salinity variables (Sal00 and Sal11) which are identical.
Note that during this step, .bl files should not be in the same directory as the input files as the position data in the .btl file will be incorrectly derived from the System Upload Time in the .bl files and not the NMEA Time.

File Manipulation Modules:
Split (file suffix _Sp): Splits .cnv files into upcast and downcast files. Also creates file prefixes indicating downcast (d) and upcast (u) files. Excludes scans marked as bad during LoopEdit module steps, this is important to ensure that the correct scans for maximum pressure are selected which are used to determine when the downcast ends and upcast begins.

ASCII Out: Applied to downcast data only and outputs only the data portion of the .cnv file created at the last module step (Split).

Note that temperature sensor SN 5307 and conductivity sensor SN 3793 data were not processed further beyond this step, see Sensor Corrections below for more information.

Calculation of Depth:
Depth is a derived variable which is calculated by the Sea-Bird Processing software using the pressure variable and the Equation of Seawater 1980 (EOS-80; Fofonoff & Millard Jnr 1983), which is a function of practical salinity. The variable is depth in salt water and is calculated as per the equation on page 151 of the Sea-Bird Processing manual (from page 28 of Fofonoff & Millard Jnr, 1983), using the file header latitude as an input to calculate the local gravity. 

Sensor Corrections:
- Calibration certificates for the sensors can be found in the file ace_ctd_sensor_calibration_certificates.pdf. Quality-checking and processing of the data has been done utilising these certificates as well as by correcting the sensor data using the collected samples (e.g. salinity and total chlorophyll-a). 
- The post-cruise calibration sheet suggests a slight offset of the temperature sensors,  which is supported by the comparison of Temperature sensor 1 (SN 6146) to temperature sensor 2 (SN 5307) before any correction is applied (Figure 1a). Conductivity sensor 1 (SN 4624) demonstrates a significant drift across voyage when compared to sensor 2 (SN 3793; Figure 1b), World Ocean Atlas 2013 data (WOA2013; Zweng et al., 2013; Figure 2a), and the collected salinity samples (Figures 2b and c) before any correction has been applied. As sensors SN 5307 and SN 3793 failed during leg 3 of the voyage, post-cruise calibration and correction was only performed on SN 6146 and SN 4624 and so sensors SN 5307 and SN 3793 are not included in the final processed dataset provided here.
- The temperature variable CTDTMP (from sensor SN 6146) was corrected using a post-deployment correction procedure as outlined in the Sea-Bird Application Note 31 (Revised June 2016) where sensor drift was corrected based on the pre- and post-cruise calibration coefficients. The temperature offset was calculated as per page 5 using the post-cruise calibration values and applied to the XMLCON files via the ‘offset’ value for the sensor SN 6146 (see the file ace_ctd_sensor_corrections.csv for the list of T offsets applied to each file/cast).
- The post-cruise calibration values for the salinity sensor SN 4624 may have been erroneous as application of the post-cruise calibration through calculation and application of an islope value as per page 2 of Application Note 31 resulted in stronger drift (underestimation) of salinity as compared to bottle values collected by ACE project 18 (DOI: 10.5281/zenodo.1494924). Instead a slope value was calculated as per page 3 of the Application Note 31 and applied to the XLMCON files via the ‘slope value for sensor SN 6146 (see the file ace_ctd_sensor_corrections.csv for the list of S slopes applied to each file/cast). 	 	 	 	
The latter slope value is obtained by first collocating the CTD sensor bottle data with the discrete samples collected throughout the cruise (Haumann et al., 2019). We re-computed the conductivity of the samples from the sample salinity and corrected temperature and pressure values of CTD sensor 1 (SN 4624) using the GSW toolbox for Matlab. We then computed the difference between the re-constructed sample conductivity and the CTD sensor 1 (SN 4624) conductivity (Figure 2c). We removed flagged data from the analysis and additionally remove data points that have a deviation of more than 2 standard deviations from the regression line (red crosses in figure). For each, cast we compute a slope correction according to page 3 of the Application Note 31 (Figure 2d and ace_ctd_sensor_corrections.csv). The slope correction is only computed if 3 or more samples were taken for one cast (blue dots in figure). Slope corrections for casts with less or no samples are linearly interpolated (red dots in figure). The slope correction shows a linear drift throughout the cruise (blue line in figure), and an additional latitudinal dependence for each separate leg with a tendency towards higher values at higher latitudes. After the correction is applied the final and published CTD sensor 1 (SN 4624) conductivity (Figure 2e) and salinity (Figure 2f) agree well with the collected samples. No correction has been applied to sensor 2 (SN 3793) since no post-cruise calibration was performed and the sensor failed during leg 3.
- The oxygen variables have not been corrected further, but should be compared to oxygen bottle data when made available. An initial comparison to World Ocean Atlas 2018 data (WOA18; Figure 3; Garcia et al., 2018) suggests a substantial offset which should be corrected for when using the oxygen data.

- The chlorophyll fluorescence variables CTDFLUOR1 and CTDFLUOR2 were dark corrected and where necessary, corrected for non-photochemical quenching as per Xing et al. (2012) when sun elevation is greater than 5 degrees (calculated using Pysolar python package https://pysolar.readthedocs.io/en/latest/) as per Reda and Andreas (2003) and Boss and Haëntjens (2016). Dark correction was applied to each downcast and upcast by first subtracting the ‘dark’ fluorescence value which for profiles < 200 m was the median of the lowest 10 fluorescence values and for profiles > 200 m was the mean of all fluorescence values (flag = 2) deeper than 200 m (Xing et al. 2012). The correction for non-photochemical quenching was implemented by first applying a rolling mean with a window size of 3 to the fluorescence values within the mixed layer. The median of the maximum 5 values of fluorescence within the mixed layer are calculated and extrapolated to the surface (shallowest observation). The quench corrected fluorescence variables are CTDFLUOR1Q and CTDFLUOR2Q. The methodology requires an estimate of mixed layer depth which was calculated using the temperature threshold method (threshold 0.2 degrees Celsius from a reference depth 10 m) of de Boyer Montegut (2004). For comparison the density threshold (threshold 0.03 kg/m^3) algorithm of de Boyer Montegut (2004) and temperature algorithm of Holte & Talley (2009; https://github.com/rsignell-usgs/mld_demo) were also computed and all three estimates of mixed layer depth for each upcast and downcast file are available in the file ace_ctd_mld_CURRSSRGCMR20200405.csv. Note that where measurements at 10 m were not available, the MLD estimate was set to NaN. For the fluorescence quenching corrections, where the MLD estimate was not available for that upcast or downcast, the alternate cast (upcast or downcast) from the same CTD station was used. The fluorescence variables (CTDFLUOR1, CTDFLUOR2, CTDFLUOR1Q, CTDFLUOR2Q) were calibrated to a concentration of chlorophyll-a using total chlorophyll-a concentration (sum of monovinyl and divinyl chlorophyll-a concentrations derived from high performance liquid chromatography analyses (HPLC; DOI: 10.5281/zenodo.3406983). The slope and intercept of the linear regression of log10 transformed sensor fluorescence (x) and HPLC total chlorophyll-a (y), matched by CTDPRS of the bottle firing depth and the fluorescence sensor observation, were used to correct the fluorescence sensor variables following the equations:
CTDFLUOR1 = 10^(log10(CTDFLUOR1) * 0.4342 -0.7075)
CTDFLUOR2 = 10^(log10(CTDFLUOR2) * 0.4954 -0.5721)
CTDFLUOR1Q = 10^(log10(CTDFLUOR1Q) * 0.4307 -0.7392)
CTDFLUOR2Q = 10^(log10(CTDFLUOR2Q) * 0.4774 -0.6165)
Separate linear regressions were calculated for values above and below the mixed layer as per Xing et al. (2012), but not applied as the result was increased mismatch between HPLC derived and corrected fluorescence values of chlorophyll-a concentration. All variables CTDFLUOR1, CTDFLUOR1Q, CTDFLUOR2, CTDFLUOR2Q were smoothed with a running mean (window size = 5). Finally the decimal places for each variable were set to a maximum of 4, according to the fluorescence sensor sensitivity limits.
- The calibration values entered into the backscatter sensor (FLBBRTD SN 4391) configuration file for leg 3 were corrected as per the calibration sheet. The calibration values were correct for leg 1, although the reported values from the sensor were of bad quality. The sensor was not connected for leg 2. The volume scattering function at 700 nm (Beta700 nm: m^-1/sr), reported by the sensor, was converted to particulate backscattering at 700 nm (Bbp 700 nm: BACKSC m^-1) as per the equation: Bbp = 2 * pi * chi(theta) * Betap(theta, lambda) where Betap(theta, lambda) is the volume scattering function of particles which is equal to the difference between the volume scattering function and the volume scattering function of seawater Betap(theta, lambda) = Beta(theta, lamdba) - Betasw(theta, lamdba, T, S) where theta is the measurement angle of the sensor, or centroid angle (142 degrees) and lambda is the sensor excitation wavelength 700 nm. Beta(theta, lamdba) is reported by the sensor. Betasw is calculated as per Zhang et al. (2009) using the corresponding CTDTMP and CTDSAL values as inputs for T and S respectively. The chi conversion factor is 1.097 (Schechtig et al. 2015). Dark values with the instrument connected to the CTD frame were not collected and it is possible that there is a significant contribution to the overall signal by the dark signal which has negatively impacted the quality of the data. In many cases during leg 3, the calculated volume scattering function of seawater is very high and the resulting values of Bbp are very small. Finally, the decimal places for each variable were set to a maximum of 6, according to the backscattering sensor sensitivity limits Although those values of Bbp are still within the sensors detection limit, caution is advised in using the BACKSC data. The data have been flagged as summarised in the section below.
- For the PAR variable, the decimal places for variable values were set to a maximum of 4, according to the PAR sensor sensitivity limits

Other file manipulation
After the Sea-Bird processing modules and file manipulation modules were applied some filenames were corrected for consistency and files merged in situations where communication with the CTD was lost during deployment and hence collected data for one cast was divided into two files. The changes applied included:
- addition of ‘0’ to .cnv filenames to maintain consistency. The following filenames were corrected: dACE201602_20_processed.cnv, dACE201602_21_processed.cnv, dACE201602_22_processed.cnv, dACE201602_23_processed.cnv, dACE201602_24_processed.cnv, dACE201602_25_processed.cnv, dACE201602_26_processed.cnv, dACE201602_27_processed.cnv, dACE201602_28_processed.cnv, dACE201602_29_processed.cnv, dACE201602_30_processed.cnv, uACE201602_20_processed.cnv, uACE201602_21_processed.cnv, uACE201602_22_processed.cnv, uACE201602_23_processed.cnv, uACE201602_24_processed.cnv, uACE201602_25_processed.cnv, uACE201602_26_processed.cnv, uACE201602_27_processed.cnv, uACE201602_28_processed.cnv, uACE201602_29_processed.cnv, uACE201602_30_processed.cnv
- Merging of the following .btl files: ACE201602_020_up.btl and ACE201602_020.btl, ACE201602_026_1.btl and ACE201602_026.btl, ACE201603_013.btl and ACE201603_013a.btl. Where necessary, bottle positions were renumbered from 1 to n to retain unique bottle positions and restore the sequential order.
- Merging of .cnv files: uACE201603_013_processed.cnv and uACE201603_013a_processed.cnv. Communication with instruments was lost during casts and data was split across two files.
- Corrected station number in files dACE201601_006_processed.cnv and uACE201601_006_processed.cnv. Correct station number is 19 not 16. 
- uACE201601_002_processed.cnv contained only one data point and so was not processed further and not included in the final processed dataset.
- After quality-checking of the cruise track had taken place, the latitude and longitude of bottle fires in the .btl files were corrected using quality-checked track data and process_ctd_bottle_file_add_latitude_longitude.py. Latitude and longitude were extracted according to the bottle firing date and time, then appended to the rows within the files. Scripts and full details of environment requirements are available (Thomas and Pina Estany, 2019).

Bottle file processing
From the Bottle files (.btl) produced during the Bottle Summary step of Sea-Bird processing (see above) variables CTDPRS, DEPTH, CTDTMP, CTDSAL, CTDDENS, CTDSOUND, CTDOXY, CTDOXYSAT, CTDFLUOR1, CTDFLUOR2, BACKSC, PAR are retained (previously pressure (PrDM), depth (DepSM), temperature1 (T090C), practical salinity1 (Sal00), density1 (Density00), sound velocity (SvCM), oxygen concentration (Sbox0Mm/Kg), oxygen saturation (OxsolMm/Kg), fluorescence (FlECO-AFL and FlSP), backscatter (TurbWETbb0), PAR (Par)). 
Variables CTDTMP, CTDSAL, CTDDENS, CTDSOUND, CTDOXY, and CTDOXYSAT and PAR were visually inspected and no anomalous values were identified. Variables CTDOXY and CTDOXYSAT were flagged with Flag 1 as per the CTD cast data (See section Quality Flags). 
For each bottle, the values for corrected variables CTDFLUOR1, CTDFLUORQ1, CTDFLUOR2, CTDFLUORQ2, BACKSC (and corresponding variable flags; see Sensor Corrections above) were retrieved from the corresponding processed CTD upcast data files, matching the nearest CTDPRS values between the CTDPRS for the bottle and CTD upcast CTDPRS (+- 1 m). Additional metadata including the latitude and longitude at the time of bottle firing was added to the files (see Other File Manipulation), station number (STNNBR), cast number (CASTNO) and bottle status flag (BTLNBR_FLAG using information from the ace_ctd_deployment_summary.csv file) before all .btl files were appended into one file.

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

Quality Flags
Each variable was inspected and quality flags (WHP-Exchange Format, Barna, Swift and Diggs, 2016; https://exchange-format.readthedocs.io/en/latest/) applied using the following conditions:
- Flag 1 Not Calibrated - Oxygen variables (CTD) flagged until comparison with bottle samples or suitable climatology dataset e.g. GLODAP, WOCE or WOA
- Flag 2 Acceptable Measurement 
- Flag 3 Questionable Measurement - CTDTEMP or CTDSAL failed spike or gradient tests (SeaDataNet Data Quality Control Procedures Version 2.0). BACKSC where > 50 % of values flagged as Flag 4, see below.
- Flag 4 Bad Measurement - Bad flag applied during Sea-Bird Data Processing Modules, scan variable value is negative number, variable value identified as being anomalous at beginning or end of upcast or downcast during visual inspection (see ace_ctd_visual_inspection_v2.csv for full list of data points identified), values fall outside acceptable sensor specific ranges (see information on sensor specific flagging below).
- Flag 5 Not Reported - No instances where this was applied
- Flag 6 Interpolated over pressure interval larger than 2dbar
- Flag 7 Despiked - No instances where this was applied.
- Flag 9 Not Sampled - Sensor was not connected
Flags can be identified from the following header format [PARAMETER]_FLAG_W (WHP Exchange Format).

Sensor/variable-specific flagging:
- CTDTMP - SeaDataNet (SeaDataNet Data Quality Control Procedures Version 2.0) spike and gradient test. Flag 3 applied if spike test value > 6 degrees celsius or if gradient test value > 9 degrees celsius. Flag 4 applied if value is < -2.5 degrees celsius or > 40 degrees celsius (Wong et al. 2019).
- CTDSAL - SeaDataNet (SeaDataNet Quality Control Procedures V2.0 2010) spike and gradient test. Flag 3 applied if spike test value > 0.9 PSU or if gradient test value > 1.5 PSU. Flag 4 applied if value is < 2 PSU or > 41 PSU (Wong et al. 2019).
- CTD SOUND and CTDDENS - values are derived from CTDTMP and CTD SAL so if corresponding CTDTMP or CTDSAL value is flagged with Flag 3 or Flag 4, the same flag is applied to CTDSOUND and CTDDENS.
- PAR - Flag 4 applied if value < 0 micromol photons/m^2/s
- CTDFLUOR1 and CTDFLUOR2 - to remove anomalously low values which affect dark correction of sensors, Flag 4 applied if value < 0. Flag 4 applied if value less than mean + 3 * standard deviation for the entire cast.
- BACKSC - to remove anomalously low values which affect dark correction of sensors, Flag 4 applied if value < 0. Flag 4 applied if value less than 3 * standard deviation for the entire cast and after processing to Bbp if value < 1e-10. All remaining values have been flagged with Flag = 3 as most casts contain a high proportion of very small Bbp values which are likely impacted by lack of correct ‘dark’ values.
 	 	 	
Comparison to World Ocean Atlas 2018:
For quality assessment, the CTD temperature, salinity, and oxygen data is compared to World Ocean Atlas 2018 (WOA18; Locarnini et al., 2018; Zweng et al., 2018; Garcia et al., 2018) data (Figure 3). The final and corrected temperature and salinity differences (Figures 3a through f) are small and within the bounds of what one might expect to be caused by natural variability. Oxygen data shows a larger than expected difference (Figures 3g through p), which should be taken into account when using the data. Figure 4 shows that the data set covers most of the Southern Ocean water masses and properties.

***** Standards *****
 
Data files have been created following the WHP-Exchange Format (Barna, Swift and Diggs, 2016; https://exchange-format.readthedocs.io/en/latest/). 

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

Raw Data
- ace_ctd_raw_files, folder
-- Original_XMLCON_config_files, folder
--- ACETTTLL_XXX.XMLCON, configuration files, ASCII
-- Updated_XMLCON_config_files_20190721, folder
--- ACETTTLL_XXX.XMLCON, configuration files, ASCII
-- Raw_Data, folder
--- ACETTTLL_XXX.hdr, header information, ASCII
--- ACETTTLL_XXX.bl, raw bottle data file, ASCII
--- ACETTLL_XXX.hex, raw data file, hexadecimal

Intermediate Data
- Sea-Bird_Processed_Data, folder - see SeaBird processing section above for information on file naming convention

Processed Data
- ace_ctd_CTD20200406CURRSGCMR, folder - Complete CTD sensor files (CTD) from upcasts and downcasts
-- cACETTTTLL_XXX_ct1.csv, processed data files, comma-separated values
- ace_bottle_BOTTLE20200406CURRSGCMR, folder
-- ace_ctd_BOTTLE20200406CURRSGCMR_hy1.csv, processed data file, comma-separated values

Metadata
- figure1.pdf, metadata, portable document format
- figure2.pdf, metadata, portable document format
- figure3.pdf, metadata, portable document format
- figure4.pdf, metadata, portable document format
- ace_ctd_sensor_list.pdf, metadata, portable document format
- ace_ctd_sensor_corrections.csv, metadata, comma-separated values
- ace_ctd_sensor_calibration_certificates.pdf, metadata, portable document format
- ace_ctd_deployment_summary.csv, metadata, comma-separated values
- ace_ctd_water_sampling.csv, metadata, comma-separated values
- ace_ctd_mld_CURRSSRGCMR20200405.csv, metadata, comma-separated values
- ace_ctd_visual_inspection_v2.csv, metadata, comma-separated values
- ace_physical_biogeochemical_oceanography_ctd_change_log.txt, metadata, text format
- data_file_header.txt, metadata, text format
- README.txt, metadata, text format

CTD cast data files are named in the format cACEYYYYLL_XXX.yyy or ACEYYYYLL_XXX.yyy where c is the type of cast, either upcast (u) or downcast (d), ACE is the name of the expedition (Antarctic Circumnavigation Expedition), YYYY is the start year of the expedition (2016), LL is the leg number of the cruise (01 - 03) and XXX is the CTD cast number within the leg (this starts again from 001 at the beginning of each leg). Finally, yyy is the file extension. If c is missing, then the file contains both upcast and downcast. NMEA location in the .btl, .ros and .cnv SeaBird Processed files are not from the corrected cruise track data. 

Within the files, SECT_ID is equivalent to leg described elsewhere in this documentation. 

Null values: In the processed files null values are filled as ‘999’.

Data flags: See section above about Quality Flags

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

Jenny Thomas, Swiss Polar Institute, Switzerland. ORCID: 0000-0002-5986-7026. Email: jenny.thomas@epfl.ch, jt.sciencedata@gmail.com 
Charlotte M. Robinson, Curtin University, Australia. ORCID: 0000-0001-8519-5641. Email: charlotte.robinson@curtin.edu.au, charlotte.mary.robinson@gmail.com 
Tahlia Henry, University of Cape Town, South Africa. ORCID: 0000-0002-7370-2957. Email: tahliahenry@gmail.com
F. Alexander Haumann, Princeton University, USA; British Antarctic Survey, UK; ETH Zurich, Switzerland. ORCID: 0000-0002-8218-977X. Email: alexander.haumann@gmail.com

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

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

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

Henry, T., Robinson, C., Haumann, F.A., Thomas, J., Hutchings, J., Schuback, N., Tsukernik, M. and Leonard, K. (2020) Physical and biogeochemical oceanography from Conductivity, Temperature, Depth (CTD) rosette deployments during the Antarctic Circumnavigation Expedition (ACE). (Version 1.1). [Data set]. Zenodo. https://doi.org/10.5281/zenodo.3813646