Salinity (and therefore effectively Density)

In polar waters, with temperatures below approximately 4°C, density profiles largely follow the shape of the salinity profile. This means that salinity checks can also include density profile checks and the dynamical unlikeliness of density overturns. There are a limited number of casts with significant density overturns. As these would make the profiles unstable (dense water above less dense water) then is almost all cases they can be ascribed to sensor problems. They can happen throughout a profile but are more common at the surface of bottom of the profile. They were filtered by looking for an overturn of >0.05 kg m3 and also by looking for unusually large deviations between different mixed layer depth calculations (including using the 10m depth as the reference value). Spikes are then identified and removed manually in salinity in the initial processing (rats_cnv2mat). This is usually between 1 and 7 metres of data, though some profiles are completely removed (including events 1495 and 1999), where pump problems make all data invalid.

The precision of the salinity data is ensured by salinity samples being collected and by joint casts between the RATS CTD(s) and that on the R/V Laurence M Gould, with adjustments applied in initial processing.

Temperature

Temperature has little effect on density in the range encountered and is therefore free to vary both up and down with depth such that there is no way to ascribe a profile to be physically implausible. The temperature data has been very robust, with no suspicious profiles and very tight matches in all joint casts, it is therefore presented as recorded, except for profiles with pump profiles, where the temperature looks less wrong than salinity but the depth the data is recorded at could be significantly different to the depth the water actually was when it entered the CTD.

PAR

From 2017 there have been repeating problems with the PAR sensors, despite servicing and changing sensors. Some values at depth are easily filtered as impossible but other times the values are within bounds, but the shape of the profile is unlikely. There are standard sampling issues, caused by the shade of the boat, ice and clouds, that means light can increase rapidly with time and/or depth. This makes filtering the problem profiles harder, without removing data where the sensor is working well. Often the shape of the profile is more important than the absolute values so these profiles that increase with depth are of reduced value.

The first filtering is to use a mask created from the first 700 events and also remove values <-1. This removes wild numbers, accounting for the variability driven by weather and attenuation (which can vary considerably with phytoplankton concentration). Away from changing shading/cloud conditions the expectation is for an exponential dropoff of PAR with depth, and so significant deviations from this can flag up potentially problematic profiles. Estimations of attenuation from PAR profiles are calculated by fitting a regression line to log(PAR) in overlapping 5m depth intervals down the profile. Checking profiles with negative ''attenuation'' catches further profiles that are judged to be problematic due to the sensor (rather than natural effects) and these PAR profiles are removed after individual checking.

Two profiles in 2015 (1667, 1673) show very unlikely increases at depth. Given the similarities and closeness of the profiles these have been deemed a sensor problem and also removed, despite being before the period of regular problems.

Mixed layer depth (derived variable)

Mixed layer depth is sensitive to the definition used, which is inevitably a processing choice, with no outright correct answer. Profile-by-profile checking has shown that the use of a 0.05 kg m3 density difference criterion gives a good match to the reduction of chlorophyll with depth, for the period of the year where the mixed layer exceeds the photic zone. This is a good indication of the mixed layer depth that is calculated describing the depth that is, or has very recently been, connected to the surface through vertical mixing.

A time where this has been found to be too tight a criterion is during winter or early spring where spring melt can produce a very shallow layer of fresher, less dense water, above a homogenous zone that has clearly mixed recently. This is exacerbated by our inability to take profiles in the windy conditions that drive the mixing, due to boating safety reasons. To counter this a mixed layer depth relative to 10m is also calculated, which should catch the deep mixing events that happen between sampling events.

Variable names: mld, mld10m

Stratification (derived variable)

There are frequent times through summer where there is strong stratification through the surface depths (to within 5 metres, or even 10cm of the surface) due to the large input of meltwater from sea ice, icebergs and the glaciers. In these circumstances, a mixed layer depth of 1m or 2m does little to describe the conditions for mixing or phytoplankton. Due to this, stratification is calculated from the surface to 10m, 20m, 30m through to 100m in 10m increments. This is calculated as the additional potential energy that is required to homogenise the depth interval, which gives a physically based metric, with units of joules/m2. This is considered more relevant that a profile of Brunt Vaisala frequency.

Variable name: strat10s

Chlorophyll

There is a step change to higher sample values at water sample event 1930 (matches CTD event 1931). Awaiting HPLC samples to compare with these, data after this point should be blanked from the series for now, hopefully it is recoverable. Chlorophyll from the CTD is also blanked from this point, with fluorescence (flsc) retained as these are raw values recorded from the fluorometer on the CTD.

Ammonium

Detection limit is 0.01 µM. Any negative values were reset at 0.001 µM.

Macronutrients from 1998 to 2017

Detection limits are 0.3 µM for nitrate, 0.1 µM for nitrite, 0.2 µM for orthophosphate, and 1.2 µM for silicic acid (OSIL/NOCS). From 2017 to 2018, detection limits are nitrate and orthophosphate were 0.02 µM, and 0.01 µM for nitrite, and 0.02-0.03 µM for silicic acid (PML). The typical uncertainty of the analytical results was between 2-3%. Clean sampling and handling techniques were employed during the defrosting, sampling and manipulations within the laboratory, and where possible carried out according to the International GO-SHIP nutrient manual recommendations of Hydes et al. (2010). Seawater nutrient reference materials (KANSO Ltd. Japan) were analysed to assess analyser performance and for quality control purposes.

Spikes were removed from the macronutrient dataset as follows:

- Nitrite - Events 502 and 504 (APR 2003)

- Nitrate - Event 328 (15 JUN 2001)

- Orthophosphate - Event 486 (21 FEB 2003)

- Silicic acid - Event 314 (18 APR 2001)

Oxygen isotopes

From 1998 to 2012, isotope measurements used internal standards calibrated against the international standards VSMOW and VSLAP. Errors during this period were typically +/- 0.08 per mille for delta-O-18. From 2012 to 2017, isotope measurements used internal standards calibrated against the international standards VSMOW2 and VSLAP2. Errors during this period are typically < 0.05 per mille for delta-O-18.