Dataset: CO2 - Surface Underway pCO2 Reduced

See: SO-GasEx cruise report, Section 5.4.3 ppgs 21-23

System configuration:
Water was supplied from the ship's surface intake supply, tapped in the main lab.
Flow went through CStar, WetStar, SBE43, membrane-contactor equilibrator, SBE45,
and then to over-board-draining sink. Analog sensors were logged via a NI USB6009
14-bit A/D card.

CO2 measurement approach:
Atmospheric air was introduced to the equilibrator (LiquiCel 2.5x8) in counter-flow
arrangement at a controlled 300 ml/min. Wet equilibrator effluent was delivered to
a custom pCO2 analyzer unit centered on a LICOR LI840 NDIR unit. The NDIR unit
vented directly to the room atmosphere.

The linearized xCO2 LI840 output data was calibrated with three dry gases of known
xCO2 (tied to primary Takahashi standards), and a linear calibration function.
Calibration gas mixtures were 197.8, 405, and 616 ppm, each known to about ± 2ppm.

Calibrated xCO2 data was converted to pCO2 by multiplying by the total gas-side
equilibrator pressure. Equilibrator pressure was determined by adding the gauge
pressure in the equilibrator to the LI840 cell pressure. Gauge pressure was determined
with a Honeywell ASCX05DN compensated pressure sensor with a 31.7V/atmosphere pressure
sensitivity. Equilibrator differential pressure was typically about 0.005 atm.

Data reduction:
Raw data was collected at 1 Hz. Data was reduced to 21.6 second (0.00025 day) intervals
using a running cubic polynomial smoothing function originating on each time interval.
Data from ± 0.0025 day from each interval were allowed in the smoothing function.
Smoothing was performed if there were raw data bracketing each time interval, and if
the polynomial was more than 5x over-constrained with raw observations. All entries
were reduced in this way except xCO2_atm, which was only sampled about once every 4 hours
during standard sequences. This data was interpolated to the full 1-Hz resolution of
the raw samples, and subsequently smoothed as the other entries.

Data uncertainties:
Data are believed to be accurate to within about ±4 µatm. About half of this is due to
calibration uncertainties, and a little less due to temperature correction uncertainties
(assuming a 0.1°C total temperature correction uncertainty). The pressure correction
term is estimated to contribute less than 0.3 µatm to the uncertainty.