Deployment 2018 to 2020
temperature and conductivity calibration


The calibration adjustments were based on the drift determined from the laboratory
calibrations at SeaBird before and after the deployment. The SeaBird sheets, while
very helpful, can be confusing at first sight (they have confused me in the past).
Below, the adjustment process is described for the example of microcat #16932.
It is followed by a summary table with the parameters for all seven microcats,
and the SeaBird calibration figures.


Temperature:
The plot below shows the lower portion of the SeaBird calibration sheet for the January 2021
post-cruise calibration, a graphical display of the residuals from the current and the most
recent previous calibration. When comparing calibrations from different times, the SeaBird
sheets use the current (2021) coefficients to convert current instrument and bath readings
(the triangles close to the zero line) as well as the earlier instrument and bath readings
(circles). This general guide was described by SeaBird's Mark Baumann in an email from
June 2000:

      "In general all comparisons on the cal sheet, be it temperature or
      conductivity are done using the coefficients for the calibration date that
      the coefficients are calculated for (the date at the upper left corner of
      the cal sheet) and the sensor outputs from the other calibrations (the
      dates to the right of the residual plot)."

Residuals are defined as instrument temperature minus bath temperature.



The figure indicates a small drift of about 0.002 degrees C over the range of values
encountered during deployment. The cell has drifted [very slightly] "low"
: using the 2018 instrument readings with the 2021 cal, instrument temperature was larger
(residuals positive) than bath temperature. This implies that converting sensor frequencies
collected in 2021 into temperatures using the 2018 cal generates instrument readings that
are too low.


A simple example:
Lets say: temperature = cal * sensor frequency, a simplified conversion relation;
- the 2021 sensor frequency = 5 Hz; bath temp = 10 degC; => 2021 cal = 2 degC/Hz;
- we know: 2018 frequencies and 2021 cal generated positive residuals,
      i.e., higher temps; lets say 15 degrees; => 2018 frequency was 7.5;
- back in 2018, a frequency of 7.5 gave the correct bath temp of 10, so the
      2018 cal was 1.33.
- in 2021, when the sensor frequency was 5, applying the 2018 cal gives only 6.67 degrees
      => we have to increase the temps calculated by converting frequencies measured
      with the 2018 cal.

Since we are using the 2018 cal (i.e., the T values downloaded from the instrument
after recovery), we need to add 0.002 deg C at the end of the deployment.
Lacking additional information, we assume the drift occurred linearly in time over
the course of the deployment, from no correction at the start to full correction at
the end of the record.




Conductivity:
The conductivity correction followed a similar scheme. Again, the sensor had drifted "low",
leading to positive residuals on the cal sheet. So conductivity of the downloaded data was
increased in time, again going linearly from no to full correction over the course of the
record. In contrast to temperature, however, the correction itself was a linear funtion of
conductivity ("slope correction"). The reason is that the origin (zero output at zero
conductivity) has been found to be very stable.




Pressure:
The pressure sensor tends to be quite stable. This was confirmed by the
SeaBird post-deployment calibration, and no adjustment was needed for
any of the microcats.



Summary info for all microcats

All calibration adjustments were small. The list below summarizes conductivity slope
and temperature bias corrections. Both varied linearly from no to full correction over
the course of the deployment. At the time of recovery,

      T_corrected = T_recovered + T_offset;
      C_corrected = C_recovered * C_slope;

with the following values :

serial # depth T offset C slope
16758 35m -0.001 deg 1/0.9988215
16937 51m flooded no cal
16936 67m 0.0005 deg 1/0.9996203
16935 81m 0.0008 deg 1/0.9995243
16934 95m 0.0 deg 1/1.0003291
16932 109m 0.0025 deg 1/0.9995392
16930 123m -0.0005 deg no postcal; use typical 1/0.9995
16800 135m 0.0 deg 1/0.9993320


Two of the conductivity cals were further investigated using TS diagrams:

For 16930, conductivity had clearly failed during the last 29 days of
deployment, and no postcal was obtained at SeaBird. However, the time
series until that point looked fine, including it's TS "signature"
when compared to the remaining MCs. Lacking better information, I assumed
a typical conductivity drift, and applied a slope correction of 1/0.9995.
This small correction was also not distinguishable in the overall TS diagram.

For 16934, the conductivity cell had drifted to "saltier". In addition,
the "baseline" from the recent cal on Jan 2021 did not show residuals near zero,
as for all other cals. I will try to understand this better by checking with
Brian Hogue, and with SeaBird. In the meantime, I applied the correction as
stated with my usual procedure, and checked the resulting TS footprint against
the overall record. It is not obvious how any further adjustments would
improve this comparison.


Below, the SeaBird T and C calibration sheets as well as the adjustment plots
are shown in one row of plots per microcat.