EXPERIMENTAL
Titration system.
	The titration system is similar to the one used in our earlier studies (Thurmond and Millero, 1982) and that developed by Bradshaw and Brewer (1988).  The titration systems used to determined TA consisted of a Metrohm 665 Dosimat titrator and an Orion 720A pH meter that is operated by a personal computer.  Both the acid titrant in a water jacketed burette and the seawater sample in a water jacketed cell were controlled to a constant temperature of 25oC with a Neslab constant temperature bath.  The glass water jacketed cells (volume about 230 cm3) used during the spring cruise were patterned after our earlier design (Thurmond and Millero, 1982).  The cell top was made of Plexiglas with inlets for electrodes, syringe and titrant injection tip.  The cell top can be removed between runs to clean and fill the cell with sample.  It is closed by pushing the top all the way down to the top edge of the cell.  An O-ring on the cell top was used to secure a fixed volume.  For the fall cruise we used a Plexiglas cell.  The cell design was similar to that used by Bradshaw et al. (1988) except a larger volume (about 200 cm3) was used to increase the precision.  This cell had a fill and drain valve and increased the reproducibility of the cell volume.  A GW-BASIC program was used to run the titration, record the volume of the added acid and the emf of the electrodes using RS232 interfaces.  The titration is made by adding HCl to the seawater past the carbonic acid end point.  A typical titration records the emf reading after the readings become stable (ñ 0.05 mV) and adds enough acid to change the voltage to a pre-assigned increment (10 mV).  In contrast to the delivery of a fixed volume increment of acid, this method gives data points in the range of a rapid increase in the emf near the endpoint.  A full titration (25 points) takes about 20 minutes.  Using three systems a 24 bottle station cast can be completed in 3.5 hours. 
Electrodes
	The electrodes used to measure the emf of the sample during a titration consisted of a ROSS glass pH electrode and an Orion double junction Ag, AgCl reference electrode.  A number of electrodes were screened to select those to be used in the titrators.  Electrodes with grossly non-Nernstian behavior (slopes ñ 1.0 mV different from theoretical) were discarded.  A total of 20 electrode pairs were examined before the cruise and six electrodes were selected for use at sea. 
	The reliability of a glass-reference electrode pair was determined by HCl titrations in 0.7 M NaCl solutions, by using seawater buffers (Dickson, 1992) and by determining the TA of standard Na2CO3 solutions and Gulf Stream seawaters.  The HCl titrations in 0.7 M NaCl solutions were used to evaluate the experimental slope in acidic solutions (pH = 2 to 4).  Artificial seawater buffers (Millero, et al., 1993) were used to evaluate the experimental slope near a pH of 8.  The resulting experimental slopes found for the electrodes used in the present study are given in Table 1.  The slopes near a pH of 8 were lower than the theoretical value (59.16 mV); while, the slopes near a pH of 3 were higher than the theoretical value.  The electrodes were also evaluated by determining the TA, TCO2 and pHsw of Gulf Stream seawater to determine the reliability of the electrodes in seawater.  The results (Table 2) indicate that high precision measurements of TA (ñ 2.5 mmol kg-1), TCO2 (ñ 2.8mmol kg-1) and pH (ñ 0.005) can be obtained with weighed samples of seawater in an open cell.  The variations in the experimental slopes for the various electrodes had little or no effect on the values of TA, TCO2 and pHsw.  The precision of the pH measurements for a given electrode (0.003) is better than the average deviation from the mean.  As will be discussed later the accuracies of the values of TCO2 and pH are affected by the reliability of the electrodes.
Standard acids.
	The HCl acid solutions (20 L) used throughout the cruise were made, standardized, and stored in 500 cm3 glass bottles in the laboratory for use at sea.  The 0.25 M HCl solutions were made with 1 M Mallinckhodt standard solutions in 0.45 M NaCl to yield an ionic strength equivalent to that of average seawater (0.7 M).  The acid was standardized by titrating weighed amounts of Na2CO3 and TRIS dissolved in 0.7 M NaCl solutions.  The blanks in the 0.7 M NaCl solutions were determined by using coulometry and by titrations of the NaCl solutions with and without added Na2CO3 and TRIS.  The blanks of the titrations of TRIS were determined by extrapolation to zero added salt (Goyet and Hacker, 1992).  The alkalinity blanks in the NaCl were generally about (14 ñ 1 mM).  The concentration of the standard acids obtained from Na2CO3 and TRIS were in good agreement (Table 3).  Recently we have determined the concentration of HCl by using a coulometric technique.  The system we used is similar to the earlier design of Taylor et al. (Taylor and Smith, 1959; Marinenko and Taylor, 1968) and constructed by Dickson (personal communication).  The protons are titrated coulometrically and released as H2(g) on the Pt anode.  The current is generated from a Sargent constant current supply and recorded as a function of time by measuring the voltage across a standard one ohm resistor with a Keithley 617 electrometer.  An ORION combination electrode was used to measure the pH of the solution during the titration.  The endpoint is determined from the plot of the concentration of H+ and OH- in the solution as a function of the coulombs delivered (Figure 1).  The precision of the coulometric titration is very good (0.01%) and the results agree with titrations made in Na2CO3 and TRIS solutions (see Table 3).
Volume of the cells.
	The volumes of the cells used at sea were determined in the laboratory by weighing the cells filled with degassed Millipore water.  The density of water at the temperature of the measurements (25oC) was calculated from the international equation of state of seawater (Millero and Poisson, 1981).  The nominal volumes of all the cells were about 230 cm3 and the values were determined to ñ 0.03 cm3.  The reliability of the volumes was assessed by comparing the values of TA obtained for standard solutions with open (weighed sample) and closed cells.  If the volume is correct the TA from the open and closed cells should be the same (provided the same acid, titrator, and electrodes are used).  The open cell titration was found to affect the TCO2 derived from the titration, due to the loss of CO2, but not the TA.  If an accurate weight of the cell cannot be made, the open and closed cell titration offers a precise way of determining the volume of the cell.  The volumes of the plastic cells determined in this manner were more reproducible than the glass cells.  The valves in the plastic cells close off a more reproducible volume than the O-ring seal used on the glass cells.  For this reason we replaced the glass cells with the plastic cells for the Fall cruise.  If the cells were modified during the cruise, adjustments were made to the volumes using the daily titrations on low nutrient surface seawater (collected before the first station and poisoned with HgCl2) and the CRM.
Solutions.
	The NaCl, Na2CO3 and NaHCO3 salts used to make up the solutions were Baker reagent grade.  The details on the preparation and calibration of the seawater buffers (Dickson, 1993) are given elsewhere (Millero, 1993).  Approximately 20 liters of standard carbonate solutions in 0.7 M NaCl were prepared for the calibrations of the acids.  The solutions were equilibrated with air to provide an alkalinity and nearly constant TCO2 standard.  The TCO2 in the blanks and carbonate solutions was measured daily by coulometry (UIC Inc.).  The coulometer was calibrated using CO2 gas loops and monitored with the CRM.  The Gulf Stream seawater used in the laboratory was collected off the coast of Miami.  The seawater was filtered through a 0.45 mm Millipore filter and poisoned with HgCl2 (250 ml of saturated HgCl2 solution per liter of seawater).  The salinity was determined with a Guildline salinometer on the practical salinity scale.  The densities of all the solutions were determined with a Mettler densimeter.  The densimeter was calibrated with nitrogen, water and standard seawater (Millero and Poisson, 1981).
Volume of Titrant.
	The volume of HCl delivered to the cell is traditionally assumed to have small uncertainties (Dickson, 1981) and equated to the digital output of the titrator.  Calibrations of the burettes of the Dosimats with water at 25oC indicate that the systems deliver 3.000 cm3 (the value for a titration of seawater) to a precision of ñ 0.0004 cm3.  This uncertainty results in an error of ñ 0.4 mmol kg-1 in TA and TCO2.  The accuracy of volume of acid delivered by the Dosimats, however, is ten times lower than that of precision.  Since the titration systems are calibrated using standard solutions, this error in the acuracy of volume delivery will be partially cancelled and included in the value assigned to the concentration of HCl and the volume of the cell.  When TA CRM becomes available the calibration of the burettes must be incorporated into the calibration of the system.