A New Instrument Design for Continuous Determination of Oceanic pCO₂

NDP-064 APPENDIX A

by

Christopher L. Sabine and Robert M. Key

Geology Department
Princeton University
Princeton, NJ 08544-1003

• Abstract
• Introduction
• Instrument Design
  • Equilibrator
  • Detector Rack
  • Control and Data Logging
• Instrument Characteristics and Calibration
  • Equilibrator Efficiency
  • Calibration
• Cruise Results
• Conclusions
• Acknowledgements
• References

Abstract

High resolution - high precision measurement of oceanic surface water pCO₂ is critical to the prediction of global climate change due to anthropogenic influences. A new instrument is described which has a response time of approximately one minute and a long term precision and accuracy of approximately 0.4 and 1 ppm, respectively. The equilibrator design is a modification of a counter flow disk stripper which has been used in the past to extract soluble gases from seawater. The detector is a dual beam infrared spectrometer. Calibration and operation of the instrument as well as data logging are computer controlled and require minimal attention. The design is such that other instrumentation can be easily added. Details of the instrument control, calibration, and efficiency tests for this instrument are given to assist others interested in building similar type systems.

Introduction

One of the primary concerns of earth science today is the effect of increasing anthropogenic atmospheric gas levels on future global climate. Of these "greenhouse" gases CO₂ is one of the more important. Many articles have been published in the past decade presenting both data and numerical model results of the concentrations, interactions and fate of atmospheric CO₂. To date, significant differences persist between measured CO₂ concentrations and predictions made by the models. Additional high precision measurements of oceanic CO₂ concentrations are needed to reconcile the differences in net carbon uptake observed between previous direct assessment and the ocean circulation models (Siegenthaler and Sarmiento 1993; Sarmiento et al. 1992; Tans et al. 1990).

Chemical and physical forcing drive the partial pressure (fugacity) of any dissolved gas in a water body toward the partial pressure of that gas in the overlying air. Variations in sea surface pCO₂ are caused by several highly variable physical, chemical and biological factors including temperature, salinity, biological production and respiration, air-sea gas exchange and mixing (Poisson et al. 1993; Weiss et al. 1982; Copin-Montegut 1988; Sarmiento and Siegenthaler 1992; Smith 1985). These processes result in both large scale and small scale temporal and spatial variability of CO₂ in the open ocean. A thorough understanding of the flux rate of CO₂ between the atmosphere and ocean is necessary to predict the effects of increasing atmospheric CO₂ levels on global climate change. It is vital, therefore, to develop instrumentation capable of rapidly and precisely estimating this variability on both large and small scales. To accomplish this, the system must be automated and capable of frequently logging both atmospheric and surface mixed layer CO₂ concentrations. The value actually needed for flux calculations is the CO₂ fugacity. All of the instruments used to estimate fugacity measure the partial pressure (pCO₂) or mole fraction CO₂ from which fugacity is calculated (Weiss 1974 and references cited there).

Various systems have been designed and reported which measure both atmospheric and surface water CO₂ concentrations. Of these techniques the best known to the U.S. oceanographic community was developed by Weiss at Scripps Institution of Oceanography (Butler et al. 1988; Weiss 1981; Weiss et al. 1992). Weiss' system couples a "shower" type equilibrator (Broecker and Takahashi 1966) to a multi-detector gas chromatograph. His system is computer controlled and has been proven reliable and precise over years of at-sea service. The only drawbacks to the Weiss design are (1) the operational mode is such that the spatial resolution is low relative to small scale surface water processes and (2) the level of operator training necessary to repair the system is relatively high. On the plus side, the Weiss design has very few moving parts, the response of the gas chromatographic detector is extremely linear thus simplifying calibration, the detectors can be used to measure multiple gas concentrations (CO₂, CH₄, N₂O), the sample size is very small, and the overall package has been proven accurate so far as is possible with any gas-water equilibrator.

Several new instruments have been built in the past few years which use the shower type equilibrator, but replace the gas chromatographic detector with a dual beam infrared spectrometer (e.g,. Chipman et al. 1993; Wanninkhof and Thoning 1993). This design has the disadvantage that the detector response is somewhat non-linear, however since the sample gas flow is continuous, the spatial and temporal resolution can be better than a discrete injection system if this is the time limiting step in the analytical procedure. Ultimately the spatial and temporal resolution of all these instruments is limited by the equilibration time of the shower equilibrator. In a recent calibration study (Dickson 1994), 13 different seagoing pCO₂ systems were compared. The primary conclusion of this experiment was that different equilibrator designs could give comparable results when supplied with a continuous unchanging stream of water. The degree of complexity in the equipment and the instrumental response times were significantly different.

We have developed an automated underway pCO₂ system using an equilibrator design that has a very short equilibration time and is capable of operating on extended cruises with minimal attention. The equilibrator is a modified disk-stripper design that was found to be very efficient at removing radon from seawater (Schink et al. 1970). The system uses a non dispersive dual beam infrared detector. The primary advantages of this system are the rapid response time of the equilibrator and the low level of expertise necessary to maintain the system relative to a gas chromatographic detector design. The gain in equilibrator response time sacrifices simplicity relative to the shower head equilibrator since the disk equilibrator has moving parts in addition to the air circulation pump present on all seagoing pCO₂ instruments. On the other hand, the disk equilibrator design is not sensitive to changes in water pressure from the bowpump.

In the sections that follow the new instrument design is described in detail. This is followed by a discussion of precision and accuracy, both internal and relative to other equipment. Results from the recent WOCE Indian Ocean survey are presented to demonstrate at sea operation.

Instrument Design

In this section the overall system is described. The details are broken into three main components: the equilibrator, the detector, and the instrument control and data logging.

Equilibrator

The disk equilibrator is an acrylic cylinder, 8" in diameter and 24" long (None of the construction materials or parts used in the equilibrator are metric; therefore, actual "American" units are given in this section.), with 60 disks mounted on a stainless steel shaft that runs along the axis of the chamber (Figure 1). The 0.0625" thick disks are 7.5" in diameter and are rotated at 135 RPM by a gear reduction motor connected to the disk shaft with a flexible universal coupler. On the motor end of the equilibrator, the disk shaft passes through the endplate of the cylinder. Gas and water leakage around the shaft is prevented by compressible teflon seals. On the other end of the cylinder the shaft is mounted in a permanently lubricated brass sleeve bearing. The equilibrator is fitted with a high precision absolute pressure transducer (Setra model 270), a calibrated platinum resistance thermometer, a precision mercury thermometer and a water level float switch.

Seawater from the ship's bowpump is run through an acrylic bypass chamber (Figure 2) at the fastest rate possible to minimize the residence time between the equilibrator and the bowpump and to assure that sufficient water is available whenever the system demands. The bypass chamber is fitted with a valve from which discrete water samples can be collected. Water is pumped from the bypass chamber into the equilibrator using a belt driven impeller pump. The pump is powered by a 1/3 HP - 1725 RPM electric motor. Water flows through the equilibrator at 8-18 l/min with the rate being determined by the available supply from the bowpump and regulated by varying pulley diameter ratios on the pump and pump motor. The typical water flow rate for this system (18 l/min) is high compared to shower head equilibrator designs. The pump is automatically turned on and off by a float switch to maintain a constant water level in the equilibrator. The water pump and motor are housed in a wooden container designed to ease any maintenance and to minimize noise. Pump/motor noise is further lessened by using a sectional drive belt rather than the conventional V-belt and by mounting the box on rubber feet. The pump box is fused and equipped with an external power switch. The pump and motor are kept as close to ambient temperature as possible by ventilating the box with a muffin fan. Plumbing connections between the bypass, pump, and equilibrator are made with 3/4" ID flexible nylon-reinforced plastic tubing. The equilibrator along with its disk drive motor, temperature and pressure sensors and electronics, float switch and electronics and bypass are housed in a box suitable for shipping and easy setup. A water meter with plastic internal parts (Omega model FTB6207) is mounted between the exit of the water pump and the inlet side of the equilibrator to monitor water flow rate and pump performance. Water exits the equilibrator via a combination of 1" PVC pipe and flexible plastic tubing. Ultimately the maximum water flow through the equilibrator is controlled by the gravity head between the equilibrator exit and the end of the drain lines. Both the water inlet and drain lines of the equilibrator are fitted with flow control valves mounted adjacent to the equilibrator chamber. By changing the water pump pulleys and adjusting the flow control valves, the system can be adjusted to work with a very wide range of water delivery rates from the bowpump. Once water flow is established through the equilibrator it is simple to balance the flow so that the water pump runs almost all the time. The water inlet to the equilibrator is always set so that the pump does occasionally turn off. That is, the inlet rate when the pump is running is slightly greater than the drain rate. This guarantees that the equilibrator doesn't eventually empty. The pump is wired such that in the event of an electrical component failure or a kinked drain line the pump turns off to prevent flooding. There is no protection built into the system for failure of the ship's bowpump. In that case, the only problem would be the eventual erosion of the water pump impeller.

In normal operation, the equilibrator chamber remains approximately half-full resulting in a water residence time in the equilibrator of one minute or less. The rotating disks develop a thin layer of water on both sides thus increasing the exposed surface area of the water 20 fold. Because the disks rotate, the amount of water directly exposed to air per unit time, is increased by a factor much greater than 20. Air is recirculated through the top half of the chamber (approximately 6 liters) in the opposite direction to the water at a rate of approximately 6 l/min resulting in an air residence time in the chamber of approximately 1 min. The high water flow coupled with the very high effective water surface area exposure and the relatively low rate of air flow to the detector (about 0.03 l/min) results in an extremely rapid equilibration rate. The small amount of air removed from the equilibrator for measurement is replaced by a vent line extending from the equilibrator air circulation plumbing to the outside (Figure 2). The vent line is fitted with a baffle volume to minimize air surging in this line caused by the air circulation pump.

Detector Rack

The NDIR detector as well as the associated plumbing and electronics are contained in a thermally insulated rack. The primary components of the analytical system are the detector, a sample selector valve, gas drying equipment and assorted valves and plumbing to control gas flow. Figure 2 is a schematic of the detector rack components and shows the connections between the detector and the equilibrator. The heart of the system is a Li-Cor 6251 infrared detector. The input gas going to the detector is selected with an electronically actuated 6-port valve (Valco model no. ESD6M). Prior to entering the selector valve, sample gases (marine air and equilibrator air) are filtered through stainless steel filter elements with a 0.5 µm pore size. After exiting the selector valve and before entering the detector, both standard and sample gasses are passed through a hygroscopic, ion exchange membrane (Nafion) and a small column with 50% magnesium perchlorate and 50% Aquasorb^TM to remove water vapor.

All gases are adjusted to flow through the detector at a rate of ~0.04 l/min. The standard and reference gas flow rates are adjusted using a combination of pressure, which is set at the tank by high purity two stage regulators, and needle valves located near the detector inlet ports. Solenoid valves are located adjacent to the detector inlets on both the reference and sample gas lines to stop the gas flow prior to taking a measurement. Zero gas flow during actual measurement was found to yield significantly better results than trying to precisely control the gas flow rate. System performance might be further improved by using mass flow controllers to regulate the gas flow between actual measurements (see the next section for an explanation of what is meant by "during actual measurement"). A high precision absolute pressure transducer (Setra model 270) is located between the detector and the vent to record the pressure at the time of measurement. Sample gas temperature at the detector is measured by a platinum resistance thermometer which is part of the Li-Cor. After passing through the detector's reference cell, the reference gas flows around the Nafion drying tube to flush the water vapor extracted from the sample gas.

Laboratory measurements demonstrated that the temperature corrected Li-Cor response was still reacting to the ambient temperature fluctuations. This is the reason that the entire detector rack was thermally insulated. To further ameliorate ambient temperature changes a 60 watt flexible silicone rubber heater was placed inside the detector rack. Air inside the rack is circulated with muffin fans. The air temperature inside the rack is measured and controlled with an Omega (model CN76130) temperature controller connected to a fast response platinum resistance thermometer and the heater. The thermometer monitors the air temperature adjacent to the Li-Cor air vent opening. This temperature controller is rated to ±0.1^oC, however during shipboard operation we have not been able to regulate the detector temperature to better than approximately ±0.15^oC on a daily basis. Figure 3 shows the detector temperature record for one day. Since this data was collected when the ship was in the Antarctic with significant laboratory to outside temperature differences (>25^oC) and ambient laboratory fluctuations >18^oC (shipboard thermostats and pipe-dreams have a lot in common), this should be a near-worst case example. The temperature controller is set to elevate the rack temperature 2-5^oC above the temperature that the insulated rack would reach from heat generated by its various components.

Shipboard electrical power is always more or less noisy. To dampen these effects the Li-Cor is powered by a high precision 12V power supply (Acopian model no. VA12NT200M). To suppress any other electrical noise, all wiring connections inside the detector rack were made with individually shielded cables. Low level DC signals were protected from the possibility of magnetically induced noise by internally wrapping the solenoid coils with Co-Netic magnetic shielding alloy. Analog, digital and AC grounds were carefully segregated by type to minimize the potential of cross talk and ground loops. Both the air circulation fans and the solenoids inside the detector rack are powered by individual modular DC power supplies.

Control and Data Logging

Normally this system operates automatically with a single microcomputer (80486 CPU) controlling sample selection, valve switching and data logging. Figure 4 shows a typical record of detector voltages recorded for one calibration and measurement cycle. The six-port sample selector valve switches between the reference gas, three CO₂ standard gases, equilibrator air and marine air. Li-Cor detector voltages are constantly monitored and shown on a strip chart display on the computer as gas flows through the detector. When the operator-selected stability criterion is achieved, flow is stopped by closing the solenoid valves, and all data signals are recorded. In addition to the detector voltage, the computer measures and logs voltage signals for detector temperature, detector pressure, detector rack temperature, equilibrator air pressure and equilibrator water temperature. The system can also be linked with the ship's navigational and meteorological computer systems with an RS232 connection so it can log the GPS location, ship's heading and speed, wind direction and speed and sea surface temperature and salinity that corresponds with each CO₂ reading. The system cycles between calibration runs and analysis of equilibrator air and marine air. The frequency of standardization as well as the number of data points collected at each valve setting are user selected and can be easily changed. All of the data is stored on floppy disk, hard disk or both. All of the computer software was written using a commercially available instrument development package (National Instrument's Lab Windows program). The volume of data collected varies depending upon the various parameter settings, but generally falls in the range of 50-100 kilobyte/day in tabular ASCII format. The program calculates and displays approximate (calculated using the most recent standardization data) pCO₂ in real time as well as all of the other measured parameters. The actual measurement of the instrumental voltages and control of the gas selector valve is done with a National Instruments A/D board (model PC+).

Instrument Characteristics and Calibration

The following section describes some of the instrument characteristics and the method used to calibrate the system. A cursory search of the literature and discussions with two noted statisticians have not turned up a rigorous statistical method to analyze the precision of this equipment in spite of the fact that many instruments are operated in similar mode. Work is progressing on this problem, but it turns out to be significantly more difficult than originally imagined. If this research is successful a statistical technique which gives optimal variable settings as well as a rigorous estimate of the instrument precision and accuracy will result. Lacking this method, the settings used have been determined by trial and error and the precision and accuracy are "educated" upper limits. One additional problem that neither this work nor other scientists have been able to address is proof that an equilibrator actually results in true equilibrium between the head space gas and the dissolved gas. The best that can be done at this point is to demonstrate that "steady state" is reached and that the results from one equilibrator are comparable to another (a few additional arguments can be made for multi-detector-multi-gas systems).

Equilibrator Efficiency

The efficiency of the disk equilibrator determines how quickly the system responds to changes in the seawater CO₂ concentration and to what extent the system is influenced by the introduction of air through a leak in the system (e.g. the replacement air vent line). Schink et al. (1970) examined the efficiency of the disk extractor design using an oxygen electrode in the effluent stream, oxygen saturated water as the test solution, and an oxygen free extraction gas. They found that at least 99% of the oxygen was extracted under a wide range of conditions. The extraction was essentially independent of the liquid level from 1/4 to 1/2 full and gas flow rates up to 2.5 l/min. The gas removal was sensitive to liquid flow rates, with complete extraction up to 1.6 l/min dropping to 75% extraction at 2 l/min. Schink and coworkers tested the efficiency of the stripper at removing radon by hooking three strippers in series and treating the gas from each stripper as a separate sample. They found that the first extractor removed 96% of the radon gas from seawater with a water flow rate of 1.6 l/min and a helium flow rate of 1.0 l/min.

For use as an equilibrator, the operation was modified to have liquid flow rates >6 l/min and to recirculate the head space gas. With these conditions, only a small fraction of the dissolved CO₂ is removed from any water parcel. A large seawater volume is thereby equilibrated with a small air volume. Two separate efficiency experiments were performed. Results are presented below. Two hundred liters of artificial seawater were prepared. The alkalinity and pH were adjusted to seawater values using sodium carbonate and HCl.

The first experiment examined the ability of the equilibrator to strip CO₂. Figure 5 shows the results. The letters in parenthesis in this paragraph refer to locations on the figure. First, the equilibrator was run in its normal configuration to establish a steady-state reading. When the detector voltage was steady, water flow was stopped, leaving approximately six liters of water exposed to the disks (A). The detector voltage was allowed to reach a steady- state value then the recirculating air flow was diverted through an Ascarite^TM column (scrubber). This column removes CO₂ from the recirculated air before it is returned to the equilibration chamber (B). The procedure was continued until enough data was collected to calculate the removal rate, then the scrubber was taken out of the recirculation loop (C). Again, the CO₂ readings were allowed to reach steady-state, then water flow was resumed (D) to allow the experiment to return to the original state. The entire experiment was then repeated. During the second run a scrubber was added to the equilibrator air replacement vent line and the extraction time was increased.

The results of the two runs were very similar. In both cases the steady-state value decreased by approximately 2 ppm when the water flow was stopped. When the scrubber was inserted into the gas recirculation loop, the CO₂ concentration rapidly decreased. The rate of decrease slowed once the concentration was below 200 ppm. In both experiments the pCO₂ increased significantly and quickly when the scrubber was removed and again when water flow was resumed. The change in extraction rate in this experiment is not fully understood. The pCO₂ decrease was approximately exponential, however, it appears that there was a change in the extraction rate once the CO₂ level fell below ~200 ppm. Another indication of the equilibration efficiency is the very fast recovery when the water flow was resumed. Results from these experiments indicate that the e-folding time for equilibration is approximately one minute. Apparently, there is some other rate limiting step involved when attempting to remove all of the CO₂ from seawater.

The second experiment directly determined the response time of the equilibrator to a step-change in the source water pCO₂. For this experiment two containers, each containing ~100 liters of artificial seawater were prepared. The pCO₂ of one solution was increased by adding a small amount of HCl. The high pCO₂ water was pumped through the equilibrator until a steady detector signal was established, then the intake hose was moved to the low pCO₂ water supply After most of the high pCO₂ water had been flushed through the system, the equilibrator drain line was moved from the high pCO₂ container to the low pCO₂ container. Once the pCO₂ readings stabilized, the intake line was moved back to the high pCO₂ tank. After a short flushing time the equilibrator drain line was returned to the high pCO₂ tank. The experiment was concluded when the measured pCO₂ stabilized.

Figure 6 shows the result of this second experiment. The system responded very quickly to the "instantaneous" change in pCO₂. An initial steady state reading of 859 ppm was established for the high pCO₂ water. Nine minutes after switching the intake tube to the low pCO₂ water a new steady-state value of 313 ppm was established. Similar to the first experiment, the concentration change had an e-folding time of approximately 1 minute. These results indicate that the equilibrator is capable cutting the water/head space difference in half every minute. The e-folding time was the same for increasing and decreasing pCO₂ concentrations in the water. The steady-state value established for the high pCO₂ water at the end of the experiment was lower than the initial steady-state value because some of the low pCO₂ water was mixed into the high pCO₂ container before the equilibrator drain was switched.

The rapid equilibration demonstrated by the experiments described above is irrelevant if all one wants to measure is the large scale open ocean air-sea pCO₂ difference. However, for short time scale changes which can occur near shore and for small spatial scale changes that can occur due to the patchy nature of oceanic biological processes or across sharp frontal regions, rapid equilibration may prove to be critical if one is ever going to understand all of the factors which control both the concentration and flux of CO₂.

Calibration

The primary calibration method for this system is periodic analysis of gas standards having known CO₂ concentrations. The infrared detector response is slightly curvilinear (i.e. not straight) with respect to CO₂ concentration in the sample gas path. Additionally, the detector has been found to have a slow drift over a period of several hours. Steps taken to minimize drift were discussed in the proceeding sections. Frequent calibration against standards can give an estimate of the analytical precision, however this technique has the potential of systematic error with respect to accuracy. In the text that follows both the internal calibration technique used to estimate precision and the results of an external comparison are described. As stated above, there is currently no known rigorous statistical test to determine optimal instrumental settings or for that matter even estimate the expected uncertainty of the results. This is followed by a comparison of results from this instrument to data generated on a cruise where both our system and the Weiss system were operated in parallel.

Internal Calibration

With this instrument the analytical precision primarily depends on three variables:

• The reproducibility obtained when analyzing a gas stream with constant CO₂ concentration over a short time interval (i.e. instantaneous precision),
• The time rate of change when analyzing a gas with constant CO₂ concentration over a long time interval (i.e. instrument drift), and
• The precision of estimating an unknown sample gas concentration based on results derived from standard gas analysis (i.e. the ability to model instrument response as a function of gas concentration and to interpolate those results over time).

The effect of the last 2 can be decreased simply by increasing the frequency of calibration. The problem with this is that the measurement platform, a ship, is normally underway so that time spent calibrating results in spatial data gaps. Theoretically, the calibration effort would have to be quadrupled to double the precision. If one assumes equilibrium is reached, then the accuracy of the results at equilibrator conditions are affected by the factors listed plus the accuracy of the standard gases.

Any time the gas selector valve changes, the computer waits a preset minimum "delay time" prior to saving any datum (3 minutes for standards and samples, 5 minutes for reference gas). For samples as well as standards, the delay time is set sufficiently long to allow the "new" gas to completely flush the "old" gas and for the detector response to stabilize. Once data collection begins for any sample, the time separation between points is determined by the time required for the signal to stabilize after closing and opening the solenoid valves. When the solenoid valves are closed - in order to take the actual measurement under no flow conditions at atmospheric pressure - gas pressure builds up behind the solenoid. When the solenoids are opened it takes a finite amount of time for the resulting pressure surge to dissipate. The detector is sensitive to these pressure fluctuations. The time for this pressure surge to pass is significantly less than the initial delay time.

Based on present knowledge of system performance at sea, a complete calibration is run every three hours. Each calibration consists of collecting a preset number, N (generally 5-10), of data points for each of the standard gases and the reference gas (which acts as a fourth standard). The three hour time interval begins at the end of a calibration run. N data points are collected for each standard regardless of the time required. N can be increased for a relatively small time expenditure since, under normal conditions, the initial delay time is a significant fraction of the total time required to collect N data points.

Figure 7a summarizes typical calibration data collected aboard ship over 3 days. At this scale neither the short nor long time scale variability is discernible. Figure 7b shows the subset of the data from the mid-level standard. On this scale both variabilities are obvious. In both A and B, the detector voltage has been normalized to the mean detector temperature for the period and to 1 atmosphere pressure. For this data the time between standardizations was set to 3 hours and N was set to 5 for each standard gas. Figure 7c illustrates the variability with time of the standard deviation of the data in Figure 7b. The mean of these standard deviations is 0.0018 volts which is approximately equivalent to 0.4 ppm CO₂. This can be interpreted as a crude estimate of the analytical precision if one assumes, (1) that the precision obtained when analyzing sample gas is the same as for a standard and (2) that no additional uncertainty is incurred when interpolating calibration runs to the sampling time.

The concentration difference between the reference gas and the highest concentration standard must span the range of expected sample concentrations. If very low sample concentrations are expected, the reference gas should have 0 ppm CO₂. The gases used for the data in Figure 7a had a CO₂ concentration range of 200 to 450 ppm. By using a reference gas with a non-zero concentration, the dynamic range of the instrument is increased. With a non-zero reference gas it is imperative that the reference gas concentration be known accurately. All of the reference and calibration gases currently in use with this instrument were calibrated to an accuracy of 0.3 ppm by R. Weiss against his standards and primary standards measured and prepared by C. D. Keeling.

Given a data set like that shown in Figure 7a , there are numerous techniques that can be used to calculate the concentration of a sample. Two steps are common to all techniques:

• Estimate the functional relationship between instrument response (normalized detector volts for this detector) and gas concentration.
• Define some method to interpolate changes in instrument response for the time interval between calibration runs.

Listed below, in order of calculation, are the steps currently used with this instrument.

1. Estimate the mean (or "best") response for the reference gas and each standard gas for each calibration run.
2. Estimate the response for each gas as a function of time by calculating the set of linear regression lines which connect the estimated responses from the calibration runs. In other words, "connect the dots" generated by step 1 plotted as a function of time. Various smoothing curves could be used here, but this procedure yields the lowest uncertainty of any tried to date (possibly because of the short time scale correlation amongst the 4 results).
3. Based on the four sets (one set for each gas) of regression lines generated by step 2, calculate the response for each standard gas at the time each unknown sample gas was measured.
4. Use the results of step 3 with the detector response for the unknown measurements to calculate the concentration of the unknown samples. Here it is assumed that the relationship between detector response and gas concentration follows a third order polynomial, therefore this step requires finding the real roots of a third order polynomial for each unknown sample measurement.

The result obtained from these four steps is the mole fraction (xCO₂) of the measured dry gas. This value can be corrected to pCO₂ or fCO₂ at in situ conditions. These adjustments have been described in great detail (DOE 1994) and are not repeated here. In order to calculate ΔpCO₂ between surface water and the atmosphere, the atmospheric results are interpolated to the times surface water measurements were made, then the difference is taken. If 0.4 ppm is used as the uncertainty for each standard determination, a rough error propagation results in an uncertainty of just less than 0.6 ppm for the equilibrator and marine air measurements and somewhat less than 1 ppm for the difference. This estimate assumes that no significant error is incurred in converting to in situ conditions. This assumption is reasonable with respect to precision, but not necessarily for accuracy. The primary correction to get pCO₂ at in situ conditions is for the warming of the water as it passes through the ship and equilibrator. A one degree increase in temperature results in roughly a 4% or ~14ppm change in pCO₂ (Takahashi et al. 1993; Copin-Montegut 1988; Weiss et al. 1982); therefore, both sea surface and equilibrator water temperatures must be known to an accuracy of ~.05^oC to calculate pCO₂ values accurate to ±1ppm.

Another method to estimate the system precision is to repeatedly measure the same sample under real conditions. This is extremely difficult with this sort of equipment; however, the experiment can be approximated by examining data collected while the ship is stationary. Figure 8 shows two days (7 stations) of data collected on station during WOCE cruise I3 in the southern Indian Ocean. For this data "on station" is defined as a ship speed of less than 1 knot. The data shown are all equilibrator gas measurements corrected to sea surface conditions. We assume that "on station" the ship is stationary, the water is stationary and that the ship has no influence on the surrounding water, none of which is true. Nevertheless, the standard deviation of data collected for each station ranged from 0.17 to 0.42 ppm with an average of 0.26 ppm. This is comparable to the precision estimated for the determination of the standard gases. Based on these two estimates it is reasonable to assign a precision of 0.4 ppm to results obtained with this instrument under at-sea conditions.

Cruise Results

Several successful cruises in both equatorial and polar waters with earlier versions of the system have shown that the disk equilibrator-IR package is reliable and operates well in both warm and cold waters. Improvements made in the system after examination of the results of these cruises (i.e., thermal control of the instrument package) led to the design described in this manuscript. This version of the system was run for the first time as part of the WOCE Indian Ocean one-time survey aboard the R/V Knorr. Weiss' shower head-GC system was also run in parallel with our instrument. Both systems shared the same marine air supply and took water from the uncontaminated bow pump plumbing at essentially the same point. The sampling frequency of the two systems was very different. Approximately 25,000 water measurements and 8,000 air measurements were automatically logged by the Princeton instrument along the 10,000 km cruise track of WOCE leg I9N (from Fremantle, Australia to Colombo, Sri Lanka). By contrast, the Weiss system made approximately 2,000 water and air measurements (two samples per hour) on the same cruise. The high sampling frequency for the Princeton system (average water sample interval was 2.5 minutes) was designed to allow examination of the small scale spatial variability in surface pCO₂ values. Changes of 10 to 20 ppm over a distance of 10 km are not uncommon in open ocean surface waters. These gradients can be an order of magnitude greater in frontal regions or in coastal waters. Despite the different designs of the two systems (e.g. GC vs. NDIR and shower vs. disk equilibrator) the Princeton and Weiss CO₂ systems gave nearly identical results. Figure 9 is a plot of ΔxCO₂ (Princeton - SIO) for surface waters versus time for WOCE leg I9N. To make a fair comparison, given the very different sampling rates, CO₂ values were interpolated from each data set to 24 evenly distributed times per day (the top of every hour) for the entire cruise. The range of surface water CO₂ concentrations covered in this comparison was approximately 300 to 420 ppm. The mean difference between the two systems (0.86 ± 2.7 ppm) was not statistically different from zero. The standard deviation of the difference not only reflects the potential variability introduced from the interpolations, but also any real variability that may have been sampled by one system and missed by the other.

Conclusions

Results from 14 months of continuous operation during the WOCE Indian Ocean survey indicate that the disk equilibrator - infrared detector pCO₂ system described herein has a long term precision and accuracy of 0.4 and 1 ppm or better, respectively. With the settings normally used in continuous mode, 15% of the time is dedicated to standard analysis and 85% to sample analysis. The fast response time of the equilibrator will allow study of the surface seawater distribution of pCO₂ on finer spatial and temporal scales than previously reported. The instrument design is such that ancillary instrumentation can be easily added. A cursory error analysis emphasizes the extreme importance of highly accurate temperature measurement and calibration of standard gases.

Acknowledgements

We would like to thank all of the members of the DOE CO₂ survey team for helpful advice while we were building the system and for helping to run the system during the Indian Ocean survey. In particular we thank R. Weiss and R. Van Woy for helpful advice and calibration of standard gases and the captain and crew of the R/V Knorr for assistance throughout the survey. This work was supported by a grant from the US Department of Energy's Office of Health and Environmental Research (DE-FG02-93ER61540) and the Princeton University Department of Geosciences.

References

• Butler, J.H., J.W. Elkins, C.M. Brunson, K.B. Egan, T.M. Thompson, T.J. Conway and B.D. Hall, Trace gases in and over the West Pacific and East Indian Oceans during the El Niño - Southern Oscillation event of 1987, NOAA Data Report, ERL-ARL-16, 104pp, Air Resources Laboratory, Silver Spring, MD, 1988.
• Broecker, W.S., and T. Takahashi, Calcium carbonate precipitation on the Bahama Banks, J. Geophys. Res., 71, 1575-1602, 1966.
• Chipman, D.W., J. Marra and T. Takahashi, Primary production at 47° N and 20° W in the North Atlantic Ocean: A comparison between the¹⁴C incubation method and mixed layer carbon budget, Deep-Sea Res., 40, 151-169, 1993.
• Copin-Montegut, C., A new formula for the effect of temperature on the partial pressure of CO₂ in seawater, Mar. Chem., 25, 29-37, 1988.
• Dickson, A.G., The plastic menagerie: CO₂ teams compare seagoing systems, U.S. JGOFS News, 5(4), 5, 1994.
• DOE, Handbook of Methods for the Analysis of the Various Parameters of the Carbon Dioxide System in Sea Water, Version 2, A.G. Dickson and C. Goyet, eds., ORNL/CDIAC-74, 1994.
• Poisson, A., N. Metzl, C. Brunet, B. Schauer, B. Brex, D. Ruiz-Pino, and F. Louanchi, Variability of sources and sinks of CO₂ in the western Indian and Southern Oceans during the year 1991, J. Geophys. Res., 98, 22,759-22,778, 1993.
• Sarmiento, J.L., and U. Siegenthaler, New production and the global carbon cycle, In: Primary Productivity and Biogeochemical Cycles in the Sea, P. Falkowski, ed., Plenum Press, New York, pp. 317-332, 1992.
• Sarmiento, J.L., J.C. Orr, and U. Siegenthaler, A perturbation simulation of CO₂ uptake in an ocean general circulation model, J. Geophys. Res., 97, 3621-3646, 1992.
• Schink D.R., J.J. Sigalove, R.L. Charnel, and N.L. Guinasso, Jr., Use of Rn/Ra ratios to determine air-sea gas exchange and vertical mixing in the ocean, Office of Naval Research, Tech. Rep., Contract # N00014-69-C-0254, Contract Authority #NR083-245, 1970.
• Siegenthaler, U., and J.L. Sarmiento, Atmospheric carbon dioxide and the ocean, Nature, 365, 119-125, 1993.
• Smith, S.V., Physical, chemical, and biological characteristics of CO₂ gas flux across the air- water interface, Plant, Cell and Envoron., 8, 387-398, 1985.
• Takahashi, T., J. Olafsson, J.G. Goddard, D.W. Chipman, and S.C. Sutherland, Seasonal variation of CO₂ and nutrients in the high-latitude surface oceans: A comparative study, Global Biogeochem. Cycles, 7, 843-878, 1993.
• Tans, P.P., I. Fung, and T. Takahashi, Observational constraints on the global atmospheric CO₂ budget, Science, 247, 1431-1438, 1990.
• Wanninkhof, R. and K. Thoning, Measurement of fugacity of CO₂ in surface water using continuous and discrete methods, Mar. Chem., 44, 198-204, 1993.
• Weiss, R., Carbon dioxide in water and seawater: The solubility of a non-ideal gas, Mar. Chem., 2, 203-215, 1974.
• Weiss, R.F., Determination of carbon dioxide and methane by dual catalyst flame ionization chromatography and nitrous oxide by electron capture chromatography, J. Chromatographic Sci., 19, 611-616, 1981.
• Weiss, R.F., R.A. Jahnke, and C.D. Keeling, Seasonal effects of temperature and salinity on the partial pressure of CO₂ in seawater, Nature, 300, 511-513, 1982.
• Weiss, R.F., R.A. Van Woy, and P.K. Salameh, Surface water and atmospheric carbon dioxide and nitrous oxide observations by shipboard automated gas chromatography: Results from expeditions between 1977 and 1990, Scripps Institution of Oceanography Reference 92-11, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge, TN, ORNL/CDIAC-59 NDP-044, 144pp, 1992.