Data and calculations to support the study of the sea-air flux of methane and carbon dioxide on the West Spitsbergen margin in June 2014

The following text in this processing step and the next one describes the acquisition and processing sequence for the data and is only slightly modified from that in the Supporting Information for Pohlman and others, 2017 (see cross-reference). Much of the text here and in the next processing step is used verbatim.
The navigational data, meteorological parameters, water environmental parameters, and gas concentrations were collected by different instruments and correlated based on time stamping or acquisition of the data on a single computer with a program that provides a single, continuous time stamp based on a Network Time Protocol (NTP). Navigational data (latitude and longitude) and meteorological data (true wind speed) were recorded at 60 second intervals in the R/V Helmer Hanssen’s digital log files. The ship’s position is determined using various global positioning system (GPS) receivers. The ship’s positional data were checked against independently-recorded GPS fixes recorded by a U.S. Geological Survey (USGS) receiver on an Airmar PB200 meteorological station and logged using Hypack navigational software or native Hyperterminal logging on laptop computers. In most case, the positional information recorded by the ship and by the Airmar PB200 GPS was found to be within a few meters, except during sharp turns.
The ship’s meteorological sensors are located on top of the bridge, at an estimated 22.4 m above the water line. Although the U.S. Geological Survey independently recorded meteorological parameters using Airmar PB200 Weather Station at several positions on the ship, the true windspeed data obtained from these deployments were not reliable, possibly owing to problems with the directional calibration of the Airmar instrumentation. Because the standard method for determining sea-air flux depends on the square of the windspeed, small oscillations or quality problems in the true windspeed data are magnified by the flux calculation. The ship’s true windspeed record was therefore retained for use in flux calculations.
Near-surface water temperature for this study was measured in two ways. First, a hull-mounted temperature sensor is part of the R/V Hanssen’s standard instrument package, with data recorded at 60 second intervals in the navigation files. Second, the temperature and salinity of the water pumped onto the ship's laboratory via the seawater feed were measured by a YSI EXO2 sonde just before the sample was injected into the equilibrator that forms the front-end to the USGS Gas Analysis System (USGS-GAS), which consisted of a 2201i cavity ringdown spectrometer that received headspace gas from a Weiss-type equilibrator. The nominal difference between these two water temperature measurements is ~0.4 °C for this dataset and represents warming of the water during transport to the YSI EXO2 sonde through the ship’s pipes. Various corrections applied to gas concentrations determined with the CRDS are carried out using the water temperature recorded by the YSI EXO2 sonde. Calculations that refer to the state of near-surface waters (e.g., Schmidt number) employ the hull-mounted water temperature determination.
Concentrations of methane and carbon dioxide were measured with a Picarro G-2201i cavity ring-down spectrometer for near-surface water and a Picarro G-2301f cavity ring-down spectrometer for air. The G-2201i sequentially measures methane and carbon dioxide concentrations for 12C and 13C, in addition to water vapor. Concentrations are determined from the 12C absorption spectrum for those species, and δ13C values are determined from the ratio of the 13C/12C spectra. The measured concentration and δ13C values of methane and carbon dioxide for the G-2201i and the G-2301f, as appropriate, were corrected using a slope and offset correction based on a linear best-fit regression between the measured values and standards of known concentration and isotopic content. The slopes and offsets for the concentration calibration were determined from Air Liquide certified gas standards that contained 1.21, 2.01 and 30 parts per million methane, and 198, 348 and 500 parts per million carbon dioxide (± 5%). Concentrations standards were analyzed at least once per day during the expedition.
The shipboard laboratory component of the surface water analysis element of the USGS Gas Analysis System consists of a seawater feed, a YSI EXO2 sonde, a Weiss-type equilibrator, a single-channel air-handler and a Picarro G-2201i cavity ring-down spectrometer. The intake of the seawater pump was located 3 meters below the waterline of the bow. Based on the characteristics of the shipboard plumbing and a pumping rate of 20 liters per minute seawater to the wet lab, the water samples required 22.5 seconds to traverse the distance between the pump intake at the bow and equilibrator in the shipboard laboratory. A split of the water pumped through the tubing from the seawater intake was directed to the YSI EXO2 sonde, which measured water temperature, salinity, and other parameters (e.g., pH, dissolved oxygen and fDOM). A second split of the flow sprayed into the equilibrator. Gas in the equilibrator was circulated continuously at a rate of 2 liters per minute through the single-channel air-handler, where it was dried and a fraction (80 milliliters per minute) fed continuously through the G-2201i. Gas exiting the cavity ring-down spectrometer was returned to the main circulation loop of the air handler and equilibrator. The 22.5 second lag was applied to the YSI data to correlate them with the correct spatial location. Transport time from the equilibrator to the cavity ring-down spectrometer added another delay of 30 seconds. Thus, the cavity ring-down spectrometer values were time-shifted by a total of 52.5 second to place them at the correct spatial position in near-surface waters. Accurate determination of the time-offset between when a parcel of water was sampled from surface water and when gas from that parcel was analyzed by the cavity ring-down spectrometer was necessary to realistically map the areas of high dissolved gas concentrations.
The Picarro G-2301f cavity ring-down spectrometer measures methane and carbon dioxide concentrations and water vapor pressure. At the beginning of the expedition and prior to initiating the key surveys, the leak rate of the closed-loop gas analysis system that includes the Picarro G-2201i cavity ring-down spectrometer was measured. Tests were conducted by injecting 50 milliliters of a 1000 part per million methane certified gas standard into the 1.5 liter analytical loop with the equilibrator sealed at the base. The change in methane concentration over time (i.e., the leak rate) was then monitored. Measured leak rates of 0.060 and 0.048 part per million per minute for system-diluted concentrations of 32.0 and 29.9 parts per million methane equate to system turnover times of 533 and 623 minutes, respectively. By comparison, the time constant with the equilibrators in operation was much shorter at 598 seconds (see below). The much greater leak rate turnover time means that system leakage did not meaningfully affect the methane concentrations measured in the equilibrator during this expedition. To facilitate point-by-point calculation of sea-air gas fluxes, all time-corrected data were combined onto a common 30 second time grid.
Processing took place over time, 2014-2016.

Relevant data columns from the navigation and CRDS files and from the serial feed log files were extracted using Unix command line processing, yielding separate files for geographic position and some meteorological parameters, the cavity ring-down spectrometer near-surface water and atmospheric boundary layer data, the YSI EXO2 sonde data, and other meteorological parameters. The data were read in with Matlab (version R2015B) routines, with a Savitsky-Golay filter applied to smooth some of the datasets where rapid oscillations could propagate through calculations and create nonphysical results, particularly in the determination of sea-air flux. The Savitsky-Golay filter is fast and retains more of the recorded signal than other filtering approaches. The Matlab version of the Savitsky-Golay filter was implemented here using a third-order polynomial and smoothed datasets on their original, uneven time grids over different-duration time windows, depending on the degree of oscillatory behavior in the original data. For example, air concentration data were smoothed with a 51-point filter (generally less than 2 minutes), water concentration data with a 31-point filter (31 minutes), hull-mounted temperatures with an 11-point filter (11 minutes), and δ13C values for CH4 and CO2 in near-surface waters, which constituted the noisiest dataset, with a 501-point filter (~20 minutes).
Once the data were read in and smoothed, they were interpolated for each day of the cruise at common 30 second intervals using Matlab. Only navigation data and ship-supplied parameters (e.g., temperature of water at the hull) were recorded less frequently than 30 seconds. After gridding at 30 seconds, the interpolated data were written into spreadsheets where they were edited for quality control.
Prior to determining the sea-air flux, concentrations measured by the cavity ring-down spectrometers were corrected to account for the delay related to time required for the headspace of the equilibrator to reach equilibrium with the incoming water. Four laboratory-based tests were conducted in 2014 and one shipboard-based test was carried out in 2015 to derive the time constant representing the amount of time required for the system to increase to e (exponential number = 2.71828) times its initial value when the ship enters an area of elevated methane. From these analyses, the estimated delay time is 598 seconds for the rising methane case, and this was doubled to 1196 seconds for times when concentrations are decreasing. The decreasing concentration scenario represents the time required for re-equilibration to 1/e of the highest value measured in an area of elevated methane.
Following the method of Kodovska et al. (2016; see cross-reference), the lag correction was applied to the already-corrected methane values from the near-surface water measurements. To remove spikes and negative values caused by overcorrection of the data using this approach, the results were smoothed by 51 points (25 minutes) for the data acquisition on most days and by 71 points (35 minutes) for data acquisition on June 25, 2014. Given the greater solubility of carbon dioxide (and the faster response of the instrumentation), it was not necessary to lag-correct those data. The measured carbon dioxide values are assumed to be in equilibrium.
Processing took place over time, 2014-2016.

Added keywords section with USGS persistent identifier as theme keyword.