3.0 Sampling procedures and data processing

3.1 CTD profiles

CTD data were collected with a Sea-Bird SBE-0911 CTD for the EQ-1 and EQ-2 cruises and a SBE-0911 plus CTD for the EQ-3 cruise, with an internal Digiquartz pressure sensor and external temperature and conductivity sensors. Additional redundant temperature and conductivity sensors were used during the EQ-3 cruise. The Sea-Bird temperature-conductivity duct, which was used to circulate seawater through both the temperature and conductivity sensors, was used on these cruises. The CTD was mounted in a rosette sampler and the package was deployed on a conducting cable, which allowed for real-time data acquisition and data display. The package was first lowered to 10 dbar for priming of the submersible pump, and it was brought up to just below the surface, and then the downcast profile was started. Water samples were taken during the upcasts for calibration of the conductivity sensors.

The station number, latitude, longitude, date and time (GMT), and water depth of each station are provided in Tables 3.1 to 3.3.

Table 3.1
Table 3.2
Table 3.3

Return to top

3.1.1 Data axquisition and processing

CTD data were acquired at the instrument's highest sampling rate of 24 samples per second. Digital signals were stored in real-time on a PC-compatible computer and, for redundancy, the analog CTD signal was recorded on VHS video tapes.

Fig. 4 shows a flowchart of the CTD data processing. The raw CTD data were quality controlled and screened for spikes and missing data. Details of this processing can be found in Winn et al. (1991). The data were aligned, averaged to half-second values and the nominal sensor calibrations were applied. Salinity was then computed. Details of these procedures are described in the following sections. Eddy shed wakes, caused when the rosette entrains water, introduce salinity spikes in the CTD profile data. These contaminated data were handled using an algorithm which eliminated data collected when the CTD's speed was negative or its acceleration was greater than 0.5 m s¯ ². The data were subsequently averaged into 2 dbar pressure bins. For the casts in which the CTD lowering speed was higher than normal, the acceleration cutoff value had to be increased between 0.55 and 0.85 m s¯ ² to allow having enough number of points to average in each bin.

Temperature is reported here in the ITS-90 scale. Salinity was calculated using the UNESCO (1981) routines, reported in the practical salinity scale (PSS-78).

Figure 4.

Return to top

3.1.2 CTD sensor corrections and calibration Pressure

Pressure sensor calibration strategies and procedures are described in detail in Winn et al. (1991). Briefly, this strategy used a high-quality quartz pressure transducer as the laboratory transfer standard and a Russka precision dead-weight pressure tester as a primary standard, which met National Institute of Standards and Technology specifications and was operated under controlled conditions. The transfer standard was a Paroscientific Model 760 pressure gauge equipped with a 10,000 PSI transducer. The transfer standard was calibrated by the Oceanographic Data Facility at Scripps Institution of Oceanography against their primary standard in May of 1991, showing an offset at 0 dbar of 0.3 dbar, and a 0.6 dbar increase in the pressure difference (our standard reading high) over the range 0-4500 dbar. Hysteresis was less than 0.1 dbar throughout the entire range. A recent calibration at the Northwest Regional Calibration Center (27 September, 1994) showed no major change in our standard. The offset at 0 dbar was 0.3 dbar and the pressure difference over the 0-4500 dbar range was 1.2 dbar. Hysteresis was 0.2 dbar. Laboratory calibrations of the CTD pressure sensor were done using our dead weight pressure tester and a manifold to apply pressure simultaneously to the CTD pressure transducer and the transfer standard. Calibrations were performed over a pressure range of 0-4500 dbar with 11 points collected as pressure was increased and decreased. Pressure sensor #43974 was used in the EQ-1 cruise. The results from three in-house pressure calibrations against our standard showed a linear decrease in the CTD-standard difference with respect to pressure of about 6 dbar in the range 0 to 4500 dbar. The sensor was sent to Sea-Bird Electronics Inc. for inspection, where it was found faulty and was replaced by sensor #51412 in February 1993. In order to correct the pressure error to the casts during the EQ-1 cruise, a linear correction as a function of pressure was obtained from the three in-house calibrations. The coefficients of this correction were: offset = - 1.6565 dbar and slope = 1.00121. The results from the in-house calibrations are shown in Table 3.4. These values have been corrected for the shift in the standard.

Pressure sensor #51412 was used during the EQ-2 and EQ-3 cruises. The results from four in-house calibrations against our standard (Table 3.4) showed a mean offset of -0.25 dbar at 0 dbar, with small variations among calibrations. In the 0-4500 dbar range the offset had a mean increase of 0.6 dbar. Hysteresis had a mean value of about 0.1 dbar. The pre-cruise calibration offset at 0 dbar was used to correct the pressures during EQ-2 and EQ-3 (Note that this offset was only used for real-time data acquisition, as a more accurate offset was determined at the time that the CTD first enters the water on each cast).

Table 3.4 Temperature

CTD temperature calibration strategy relied on the periodic calibration of the sensors at the Northwest Regional Calibration Center (NWRCC) using techniques traceable to the National Bureau of Standards. We used three Sea-Bird SBE-3-02/F temperature transducers, serial numbers 741 (EQ-1, EQ-2, EQ-3), 886 (EQ-3) and 1416 (EQ-3). These transducers were returned to Sea-Bird for calibration by NWRCC approximately once per year prior to 1992, and twice per year thereafter. The calibration coefficients which were determined by NWRCC measurements are given in Table 3.5. These coefficients were used in the following formula that gave the temperature (in °C) as a function of the frequency signal (f):

Table 3.5

We modeled the drift of the sensors as a linear function of time as per the experience of Sea-Bird over many years of working with these sensors. For each sensor, we calculated the 0-30°C average offset for each calibration relative to the oldest one, and applied a linear fit to these offsets. The baseline calibration for each cruise was selected by taking the coefficients from each of the calibration runs and applying the linear drift interpolated to the midpoint of the cruise date. The deviation of each estimate from the ensemble mean was then calculated for the temperature range 0-5°C, where accuracy requirements are greatest. The calibration with the smallest deviation from the mean was selected as baseline calibration. Maximum error from changes in the slope and nonlinear terms in the frequency to temperature conversion is estimated to be less than 1 m°C in the 0-5°C range during the period of these cruises.

Sensor #741 was calibrated on the dates given in Table 3.5. A linear fit to the 0-30°C average offset of each calibration from 5 March 1987 to 26 June 1992 gave a -1.081x10-5 °C day¯ ¹ drift. The 26 June 1992 calibration was used as a baseline calibration for the EQ-1 data. The application of the drift to the mid-point of the cruise dates yielded a correction of 0.0006°C. Applying the drift to the beginning and the end of the cruise dates would change this correction only by ± 0.00015°C. Thus a constant correction of 0.0006°C was applied to all the cruise's temperatures (table 3.6).

Table 3.6

The calibrations for sensor #741 after June 1992 indicated a change in the drift that was confirmed by several tests performed at Sea-Bird on July 1993. The drift obtained from the 18 December 1992 and 13 May 1993 calibrations was 2.698x10-6 °C day¯ ¹. The 18 December 1992 coefficients were used to calculate the temperatures during the EQ-2 cruise. Using the mid-point cruise date, a constant correction of 0.0002°C was obtained and applied. This correction varied only by ± 0.00004°C throughout the cruise.

Dual sensors were used for the EQ-3 cruise. The temperature sensor configuration used during this cruise is shown in Table 3.7. Sensor #741 showed anomalous temperature differences with respect to the secondary sensor and was replaced by sensor #886 after station 25. Secondary sensor #1416 was replaced by #886 in stations 24 and 25 to determine which of the sensors, #741 or #1416, was causing the anomalous differences. Sensor #741 was found faulty after some post-cruise tests at Sea-Bird. Therefore, only data from sensor #1416 are reported here, except in stations 24 and 25 in which the data from sensor #886 are used.

Table 3.7

Sensor #886 was calibrated on the dates given in Table 3.5 yielding a drift of 6.503x10-6 °C day¯ ¹. The 9 June 1994 calibration was used as a baseline for the EQ-3 casts that used this sensor. A constant correction of -0.0003°C obtained by applying the drift to the mid-point of the dates when the sensor was used, was applied to these casts. This correction varied by only ± 0.00003°C during the sensor utilization period.

Sensor #1416 was calibrated on the dates given in Table 3.5. A drift of 7.609x10-6 °C day¯ ¹ was obtained. Using the coefficients from the calibration on 26 February 1994 and the mid-point cruise date, a drift correction of 0.0004°C was obtained, which only changes by ± 0.0001°C between the beginning and the end of the cruise, thus a constant correction of 0.0004°C was applied to all the casts that used this sensor. Conductivity Nominal calibration

The conductivity cell was calibrated periodically at NWRCC by varying the temperature of a salt-water bath as described by Winn et al. (1991). These nominal calibrations were used for data acquisition, and the final calibration was determined empirically by comparison with the salinities of discrete water samples acquired during each cast (see for a description). Conductivity cell #527 was used for the EQ-1, EQ-2 and EQ-3 cruises. Conductivity cell #1336 was also used for the EQ-3 cruise as dual sensors were used in the cruise. The conductivity values from the two sensors used during the EQ-3 cruise were compared for each cast, and the results showed that both conductivity sensors worked properly. The data obtained from conductivity cell #1336 are reported here for stations 1-25, and from conductivity cell #527 for stations 26-39, as these sensors were physically paired with the temperature sensors that provided reliable data.

Prior to the empirical calibration of conductivity data with water bottle salinities, conductivity was corrected for the thermal inertia of the glass conductivity cell using the recursive filter given by Lueck (1990) and Lueck and Picklo (1990) as described by Chiswell et al. (1990). The value of , which characterizes the initial magnitude of the thermal effect, was calculated separately for each cruise to close the spread between the down- and up-cast T-S curves. Table 3.6 shows the value of used in each cruise. Screening of bottle salinity samples

The nominally calibrated CTD salinity trace was used to identify questionable discrete samples (See Section for a description of the measurement of discrete samples). Potential rosette mistrip problems were resolved, where possible, before data were excluded from use in the calibration of the conductivity cell. Empirical calibration

Calibration of the conductivity cell was performed empirically by comparing its nominally calibrated output against the calculated conductivity values obtained from water sample salinities using the calibrated pressure and temperature of the CTD at the time of bottle closure. An initial estimate of bias (b0) and slope (b1) corrections to the nominal calibration were determined from a linear least squares fit to the ensemble of CTD-bottle conductivity differences as a function of conductivity, from all stations and casts during a particular cruise. This calibration was then used to identify suspect water samples.

The second iteration allowed for the possible addition of a quadratic term (b2) in the correction to conductivity, as well as a revised estimate of slope and bias.The quadratic term was only included when there was a significant reduction of the RMS residuals from the fit. Only the second set of stations (26-39) from the EQ-3 cruise required the addition of the quadratic term. The final conductivity calibration coefficients are given in Table 3.8.

Table 3.8

The final step of the calibration was to perform a profile-dependent bias correction, to allow for a drift of the conductivity cell with time during each cruise, or for sudden offsets due to fouling. This offset was determined by taking the median value of CTD-bottle salinity differences for each profile at temperature below 5°C.

The individual cast offset corrections for the EQ-1 cruise were plotted against time and an obvious pattern was detected. A 7-point running mean was applied to these data (Fig. 5), and the resulting offsets were used to correct each cast. The individual offsets are shown in Table 3.9. Two of the EQ-2 cruise casts and six of the EQ-3 cruise casts required an individual offset to be subtracted.

Table 3.9
Figure 5.

The quality of the CTD calibration is illustrated by Fig. 6, which shows the differences between the corrected CTD salinities and the bottle salinities as a function of pressure for each cruise. Table 3.10 shows the CTD-bottle comparisons after the CTD data were calibrated. These values are comparable to the precision (0.001) and accuracy (0.003) of the water-sample salinities.

Table 3.10
Figure 6.

The downcast and upcast CTD salinity data were compared with the near surface salinity measured by thermosalinograph. The comparison indicated that the upcast CTD salinity matches with the thermosalinograph salinity. But the downcast CTD data are saltier than the upcast data on average by 5.25 mpsu (see section Water sample analysis

Salinity samples were collected in 250-ml polyethylene bottles. Samples collected during the EQ-1 and EQ-2 cruises were sealed with parafilm and stored for later analysis on land. Samples collected during the EQ-3 cruise were analyzed onboard. All water samples were run on a Guildline Autosal #8400A. For the EQ-1 cruise, IAPSO standard seawater batch P115 was used. For the EQ-2 cruise, batch P118 was used, and for the EQ-3 cruise, batch P121 was used. Previous analyses show that typical precision (one standard deviation of triplicate samples from the same Niskin bottle) is about 0.001 (Tupas et al., 1993).

The bottle salinities collected during the EQ-1 cruise were checked against bottles from a previous COARE 156E-1 cruise to the same region (August-September 1991) obtained below 3000 dbar, and from bottles obtained at the same location on both cruises, showing good agreement between them. All salinities were within an envelope of 0.005 wide.

The bottle salinities collected during the EQ-2 cruise were compared against bottles from the previous COARE 156E-1 and the EQ-1 cruises obtained below 3000 dbar and from bottles obtained at the same location on the three cruises. The salinities from the EQ-2 cruise showed higher values than the others, probably owing to evaporation from the bottles that were in storage in a container for about 100 days in Pohnpei. A salinity correction because of evaporation of 0.05x10¯ ³ psu day¯ ¹ was applied based on the results of some laboratory tests designed to find the rate of salinity change due to evaporation (Appendix). Similarly, bottle salinities from the EQ-3 cruise were also compared against bottles from previous cruises (EQ-1, EQ-2, and COARE 156E-1), showing good agreement. All salinities were within a 0.005 wide envelope.

Return to top

3.2 Thermosalinograph

3.2.1 Data acquisition

During the EQ-1 and EQ-2 cruises, a custom made thermosalinograph built around a Sea-Bird CTD SBE-3 temperature and SBE-4 conductivity sensor pair was installed in the pumped intake line for cooling of one of the winches. This intake is in the ship's hull in the engine room at about 4 m below the sea surface. Data were acquired at a rate of 1 Hz using a Sun Microsystems workstation. During the EQ-1 cruise, the conductivity cell failed after two days of operation. As no spare was available, no further salinity data were acquired. During the EQ-3 cruise, a SBE-21 Seacat thermosalinograph was used. A remote temperature sensor was installed in a sea chest located at the bow of the ship to acquire temperature from a depth of about 3 m. This location allowed for undisturbed water to enter the thermosalinograph. The SBE-21 uses an internal temperature sensor along with a conductivity sensor to calculate salinity. The data were obtained every 10 sec. The correlation between the temperature from the remote sensor and the temperature from the internal sensor indicated that the time lag was about 10 sec because of spacial differences. Thus the temperature data from the internal sensor were advanced by 10 sec to match the temperature data from the remote sensor. Bottle salinity samples were taken periodically from the thermosalinograph water intake to calibrate the conductivity sensor. To calculate salinity, a pressure of 20 dbar was assumed which included the pressure from the pump (30 psi).

Return to top

3.2.2 Data processing and sensor calibration Nominal calibration


We used Sea-Bird temperature sensors serial number #621 (EQ-1, EQ-2) and #1392 (EQ-3). The calibration coefficients obtained at the NWRCC are given in Table 3.11. As these sensors are the same type as those used for the CTD measurements, the same procedures for drift estimation were followed. The 20 February 1992 calibration was used to calculate the temperature from the sensor frequency for the data obtained during the EQ-1 cruise. Assuming that the temperature sensor drifted linearly, a correction of 0.0033°C was calculated from a drift of -1.105x10-5 °C/day, obtained from the 18 March 1985, 14 August 1986, 12 October 1990 and 20 February 1992 calibrations. The 19 June 1993 calibration indicated a change in the drift compared to that from the previous calibrations and was not included in the calculations for the temperature correction of EQ-1. The last two calibrations of this sensor (20 February 1992 and 19 June 1993), indicated a decrease in the drift to about 3x10-7 °C/day, which implied a correction of 0.000025°C to the EQ-2 cruise temperatures using the 19 June 1993 coefficients. This correction was inconsequential, therefore, it was not applied to the data. The EQ-2 temperatures were calculated using the 19 June 1993 coefficients. The EQ-3 sensor #1392 temperatures were calibrated with the 3 June 1993 coefficients and corrected by 0.0014 C. This correction resulted from a drift of 4.4x10-6 °C/day obtained from the 3 June 1993 and 22 September 1994 calibrations. The 29 September 1994 calibration was obtained after the sensor's electronics was reworked, which usually causes a change in the sensor drift. This calibration was not used in the analysis.

Table 3.11


Sea Bird sensors serial number #375 (EQ-2) and #1392 (EQ-3) were used. The EQ-2 data were nominally calibrated with coefficients from the NWRCC 14 January 1993 calibration, and the EQ-3 data were nominally calibrated with coefficients from the NWRCC 3 June 1993 calibration. Processing

The data were screened for gross errors, setting upper and lower bounds of 35°C and 18°C for temperature and 6 Siemens m¯ ¹ and 3 Siemens m¯ ¹ for conductivity. No gross errors were detected in the data.

A 5-point running median filter was used to detect temperature and conductivity glitches. Glitches in temperature and conductivity were immediately replaced by the median. Threshold values of 0.3°C for temperature and 0.1 Siemens m¯ ¹ for conductivity were used for the median filter. Seven points were replaced with median values for the EQ-2 data. For the EQ-3 data, 373 data points were replaced owing to conductivity glitches, 8 data points were replaced owing to temperature glitches of the internal sensor, and 13 points were replaced because of temperature glitches of the remote sensor. A 3-point triangular running mean filter was then finally used to smooth the edited temperature and conductivity. Empirical calibration

The thermosalinograph salinity was calibrated empirically by comparing it to bottle salinity samples drawn from the plumbing near the thermosalinograph. Bottle salinity samples were analyzed as described in Section


In order to compare the thermosalinograph data with the bottle data in conductivity units, the conductivity of the bottle sample was computed using the salinity from the bottle, thermosalinograph temperature and a pressure of 20 dbar (4 dbar at depth of intake in addition to the pump's pressure).

A cubic spline was fit to the time series of the differences between the bottle conductivity and the thermosalinograph conductivity averaged over 2 minutes around the sampling time, and the spline was used to correct the thermosalinograph conductivities (Fig. 7a). Thermosalinograph salinity was calculated using these corrected conductivities, thermosalinograph temperatures and a pressure of 20 dbar.


The time delay between the water passing through the thermosalinograph and it reaching the bottle sampling area was determined to be about 50 seconds using autocorrelation between bottle and thermosalinograph samples. The thermosalinograph data were extracted within ± 15 seconds around the sample time minus the 50 second delay for the comparison with the bottle data.

A cubic spline was fit to the time series of the differences between the bottle conductivity and the thermosalinograph conductivity separately for leg 1 and leg 2 (Fig. 7b) because there is an apparent shift of calibration while the ship was in port. The correction of the thermosalinograph conductivities was obtained from the cubic spline fit. Salinity was calculated using these corrected conductivities, thermosalinograph temperatures and pressure of 20 dbar.

Figure 7.
Figure 8. Comparison with CTD data

The corrected thermosalinograph salinity was compared with the downcast CTD salinity at 4 bar for the purpose of checking the calibration.


The average thermosalinograph salinities within 10 sec of the acquisition time of the CTD data were used for comparison. A linear fit was applied to the time series of the difference between the CTD salinity and the thermosalinograph salinity. This comparison showed that the CTD data were saltier by 0.012 at the beginning of the cruise and by 0.005 at the end of the cruise (Fig. 8a).

The up and down cast CTD data at 4 dbar differ on average by 0.00525 which explains the CTD-thermosalinograph difference in the latter half of the cruise. The down cast is more salty. This difference is equivalent to the difference between the down cast CTD salinity and the thermosalinograph salinity at the end of the cruise; however, the difference during the first half of the cruise is not explained. Also, there is no obvious explanation for the difference between the upcast and downcast CTD data.


The thermosalinograph data were averaged using data during one minute after the acquisition time of the CTD sample. The linear fit of the time series of the difference between the CTD salinity and the thermosalinograph salinity showed that the CTD salinity was higher than the thermosalinograph salinity most of the time, and the average offset was 0.0059 (Fig. 8b), which could be largely explained by the CTD up-down cast differences previously mentioned.

Return to top

3.3 ADCP measurements

An RDI model VM-150 shipboard ADCP was used during the three cruises. Navigation information on cruises was obtained using the Global Positioning System (GPS). Details of data acquisition and processing are found in Firing (1991).

One of the four ADCP transducers failed prior to the EQ-1 cruise, and extensive efforts to repair it in port at Guam before the cruise were unsuccessful. Thus, the redundancy in computing currents, which normally provides some independent measure of system performance, was absent. There were no problems with the instrument during the EQ-2 and EQ-3 cruises.

The system operated nearly continuously during the entire cruise periods, with a few brief interruptions. Most of the time, the penetration depth for good quality returns was from 20 m to about 200 m, though the system sometimes reached 350 m.

Return to top

3.4 Meteorology

Meteorological data were collected at four-hour intervals during the EQ-1 cruise and at six-hour intervals during the EQ-2 and EQ-3 cruises by the ship's officers on the bridge, using standard procedures (Ship's Weather Observations, 1968). During the EQ-1 and EQ-3 cruises, meteorological observations were also made by the science group every 6 hours.

The meteorological data collected during these cruises include wind speed and direction, atmospheric pressure, sea surface temperature, wet and dry bulb air temperature, cloud coverage, and swell height and direction. A shaded non-aspirated psychrometer was used for the measurements by the bridge officers during the EQ-1, EQ-2, and EQ-3 cruises. An aspirated psychrometer was used for the measurements by the science group during the EQ-3 cruise. The smallest scale division of the thermometers determines the resolution of the wet and dry bulb air temperature and of the sea surface temperature (SST). These were 1.0°C for the wet and dry bulb air temperature of the thermometer used by the bridge officers, 0.5°C for the wet and dry bulb air temperature of the thermometer used by the scientists, and 0.05°C for the SST.

The time series of the data were plotted and obvious outliers were identified and flagged. The SST-dry air temperature and wet-dry air temperature plots helped to identify further outliers. No extra correction was applied to any of the data, except in the case of wind direction data obtained by the science group during the EQ-3 cruise. The anemometer used to measure the winds had a 37 degree misalignment because the lock screw holding the aerovane monitor pole in a north direction had fallen out. This situation was corrected on April 23 at 0142Z. All the data before this date were corrected by this offset. A calibration of the instrument obtained after the cruise revealed that the instrument was performing correctly, thus no further correction was applied. Although the wind directions obtained by the science group seem to match those from the bridge after the constant offset correction was applied (Fig. 42(l)), there is the possibility that the aerovane misalignment had fluctuated during the cruise. Thus the winds measured by the science group before the correction should be used with caution.

During the EQ-1 cruise, a meteorological tower and instrumentation were installed on the bow of the R/V Moana Wave at 10 m-height. The sensor suite was comprised of the following:

Data were acquired by a Campbell CD-7 data logger and transferred to the shipboard Sun network. Sampling rate was 1 Hz. These observations from the meteorological tower are not included in this report.

Return to top

3.5 XBT

During the EQ-1 cruise, a total of 25 T-4 XBTs were launched in support of TOGA thermal structure mapping. Data from 18 profiles were sent via GOES satellite to provide real-time ocean thermal structure information to the Global Telecommunication System (GTS).

Return to top

Back to Table of Contents

Previous Next