KPH TSG and fCO2#

Note

Based on meeting with Ylva Ericson on 30.09.23 (notes by ØF, revised by YE). The scripts and materials prepared by YE are available in an NPIOcean internal Gitlab repository.

Overview#

  • fCO2 calculated by Ylva based on measurements from ship water intake

  • There are three sources of data going into this

    1. Thermosalinograph data (SBE files). Include a feed from SBE38 temp sensor at the intake which is taken as “SST”.

    2. CO2 sensor (also connected to the intake). Measures xCO2 in air and water, which are calibrated against standard gas measurements.

    3. Ship weather station (use atmospheric pressure in the fCO2 calculation, use relative wind direction for flagging ship exhaust).

(The ship has its own GPS feed, can be found in ship log files. GPS data at high resolution is found in the data files from the weather station. Not sure if this is from a separate GPS.)

(Note that these sample at different time intervals.)

Intakes#

  • There are 3 intakes at various depths:

    • Intakes at 2, 4, 7 m

    • Normally, the 4m one is used.

    • The intake depth is given in the name of the file.

  • Note: There has been a recent rebuild of the system (?) - could have changed something.

  • Ceslav deals with the actual instruments (CO2 only?) - replacing parts, switching out stadard gases etc.

  • IMR deal with the intake and with calibration (annually?) of the TSG sensors.

  • Also connected to the intake:

    • Oxygen sensor

    • CDOM

    • Fluorometer

In sea ice, there seems to be a persistent issue with clogging. So in practice, no data from sea ice.

Common issues#

Some problems with the data that are accounted for in the procedure:

  • Repeating time stamps

  • Time lag between intake and TSG [1]

    • Found it hard to determine dT based on correlation, so doing this manually. [2]

    • Could possibly depend on flow speed (flux), but applying a fixed value (140 s) to all 2018-2019 data.

      • YE: I did it very rough using 140s for the period aug 2018 to dec 2019. But definitiely better to do it per cruise.

  • Flushing of the cell results in junk near-null values. These need to be edited out.

  • Sensors take some time to equillibrate after the intake is turned on; after flushing but in particular after clogging or if ice has formed in the pipes. So we need to remove bad data for some time after turning on (on the order of 30 min?).

    • When clogging/icing, the waterflow drops low or to 0. After this, it takes some time for the S and T to measure properly. The time is not fixed - it depends on the amount of clogging/icing.

  • Occasional data gaps in the various data sources, e.g. the weather station data.

  • Air pCO2 goes into the procedure -> can be contaminated by ship influence during certain wind directions (particularly on station, less when steaming).

Calibrations#

YE has applied a correction for drift between pre- and post calibration values. This is the only step where data outside teh cruise disk is needed. Resulting changes to end variables tend to be very small - YE thinks it would be fine to just use pre-cal values.

It can occasionally be difficult to find the actual calibration values; there is no system to keep track of these things for external users.

YE’s scripts#

Series of Matlab scripts. Look nicely laid out and well-commented.

Stephen van Heuven has written the original CO2 scripts based on the paper by Pierrot et al. (2009).

The scripts do the following:

  • Read data from TSG, CO2 sensor and weather station.

  • QC / clean these time series using various flagging procedures.

  • Calibrates Co2 (xCO2) data vs standard gases

  • Calculates xCO2 -> pCO2 -> fCO2.

Ylva’s scripts and documentation are available a the NPIOcean internal Gitlab repository (repo moved from earlier location on CodeRefinery).

Shoudl be possible to reproduce data from the Q4 2019 cruise if we want to redo this procedure (both raw and processed data are available).

Misc#

  • Ultimately, these data are syupposed to be handled and processed by the QuinCe project (PI Steve D. Jones)

    • Important to figure out whether there is actually any point of developing an NPI procedure!

    • It would probably be good to at least have a procedure for TSG data, though..

  • Olaf Schneider from the data section will be working with KPH data.

  • There is supposed to be a cross-intitution working group for KPH data which will convene in September.

  • Idea: Develop a SIC time series (or at least binary ice/no ice) fromo ship cameras etc.

Footnotes:#