AR9-WHP work during POLARSTERN ANT-XIV, Leg 4 Status: 26 November 1998 Compiled by Claudia Schmid and Walter Zenk IfM Kiel 0. Preliminary Remark This report summarizes and updates hydrographic work that has been conducted during POLARSTERN cruise ANT-XIV-4 as part of the World Circulation Experiment (WOCE). It first has been described in the cruise report by D.K. Futterer and co-workers (1998). The present summary is designated as accompanying document to the WOCE hydrographic programme. It describes CTD data processing and calibration of the CTD sensors. 1. Cruise Narrative Expedition designation: WOCE-Südatlantik 1997 Kap der Guten Hoffnung Experiment (KAPEX), see also Boebel et al (1998) and http://triton.sea.uct.ac.za Chief Scientist Dieter K. Futterer, AWI, Bremerhaven, Germany Ship FS POLARSTERN, Bremerhaven, Germany Leg ANT-XIV/4: Cape Town - Bremerhaven 21 March - 25 April 1997 Principal Investigator in charge: Walter Zenk Düsternbrooker Weg 20 24105 KIEL, Germany First hydrographer on board Olaf Boebel Ocean Climatology Group Department of Oceanography University of Cape Town Rondebosch 7700 South Africa CTD data processing and validation: Claudia Schmid Düsternbrooker Weg 20 24105 KIEL, Germany For further details see cruise report by Futterer et al (1998). 2. Measurement Techniques, Calibration and Processing 2.1 CTD/Rosette Station numbers are not only related to CTD work; thus they are gappy. CTD profiles are counted consecutively, with gaps occuring only if profiles have been omitted. The only CTD in use was a Neil Brown MKIIIB instrument (IFMK internal identification NB2). This instrument carried a Pt100 Rosemount temperature sensor, a (fast) NTC temperature sensor for analogue time constant compensation, a strain gauge pressure sensor made by Paine Instruments, a standard NBIS 4-electrode conductivity sensor, and a polarographic type Beckman oxygen sensor. The outputs of both temperature sensors are combined in an analogue circuit to a single signal. Pre- and post cruise lab calibration are available for the combined temperature signal and for the pressure sensor. The calibration of conductivity depends on in-situ samples. No oxygen samples are available to calibrate the oxygen sensor. In-situ samples to measure salinity, were drawn from 10 l Niskin bottles mounted on a 24 x 10 l General Oceanics rosette sampler. Bottles were closed on the way up. Samples were drawn immediately after the profile. Salinity samples solely served for CTD calibration. No samples were drawn from bottles that failed to close properly or showed other problems like apparent leaking. These bottles therefore are not included in the bottle file. This also means that all bottles in the file were flagged as 'no problem' (QF2). 2.2 Bottle Salinity Samples to be analyzed for salinity usually were drawn from: - The deepest point of the profile or 20 m above the bottom, for the 1500 m and the bottom stations respectively - The Antarctic Intermediate Water level - The mixed layer where vertical gradients are small All samples were filled to German beer bottles 'Flensburger Pils',a cheap and social method that has been recommended in pre-WOCE daysby Grasshoff et al. (1983) and that keeps samples stable over the typical length of a cruise (4 weeks) better than 0.001 psu. Batch No P122 of IAPSO standard seawater was used to standardize the salinometer. No double samples were considered. The overall accuray of bottle salinities for calibration purposes of the CTD is estimated by the precision of the overall calibration (0.005 psu) and the accuracy standard seawater (better 0.001 psu) to 0.002 psu. Bottle salinities that differ more than 2.8 and 3.5 times the standard deviation in salinity calibration from the calibrated CTD salinity (see below) were flagged as suspicious (QF3) and bad (QF4), respectively. The bottles may have closed at wrong positions here. However, since no other samples were taken, no corrections for wrong bottle depths have been made. 2.3 Bottle Oxygen No samples drawn, therefore the oxygen values in the data fileshave to be regarded as uncalibrated. 2.4 CTD: Data Processing The CTD used throughout the cruise was a Neil Brown MKIIIB (IFMK identifier NB2). It was mounted below a 24 x 10 l bottle rosette made by General Oceanics and lowered at almost constant speed (about 1 m/s) from 200 m depth on. Data processing is similiar to that described by Millard and Yang (1993). The steps were: - Visually inspect each profile, especially to identify 'strange' effects in the pressure record. - Create a time relative to the start of the profile for each record to well resolve the record interval 1/32 s. - Check that pressure, temperature and conductivity are in reasonable ranges. - Remove spikes in pressure, temperature and conductivity values. - Identify the first 'in water' record and associated pressure offset from the first reasonable conductivity measuremnet. - Remove cycles that were taken at a lowering speed less 0.2 m/s.Monotonize with respect to increasing pressure. For a lowering speed of 1 m/s, the number of remaining cycles then corresponds to the resolution of the pressure sensor. - Correct for different response times of the (combined) temperature and conductivity measurements. Visual inspections in large gradients suggested a 60 ms time constant for a recursive filter to slow down the conductivity response. - Apply a moving average over 29 cyles (corresponding to 3 dbar) - Apply calibrations to pressure, temperature and conductivity (see below). - Interpolate Lagrangian to 2 dbar. - Recalculate salinity and potential temperature. - Identify records as statically unstable if the vertical gradient of potential density (reference level increasing at500 dbar intervals) over a 2 dbar interval is less -0.001 Kg/m^3. Set salinity flag of such cycles to 3. For a 2 dbar output interval after removing spikes etc, the number of basic measurements is 13 on the average. This was transferred as constant to the output files. A special problem showed up in two profiles: At constant lowering rate of the CTD, one expects smooth sensor outputs as a function of time at large depths, say from 1400 m on. However, a problem showed up with the conductivity signal on station 543/profile 3 and on station 579/profile 35. When plotted, temperature is smoothly decreasing and pressure is linearly increasing as expected but conductivity jumps at 1750 dbar at station 543. This jump could not be removed, and therefore the deeper part of this profile was cut off. At station 579, bad conductivity values occurred between 1198 dbar and 1226 dbar. These were interpolated using polynomial of 3rd. order and flagged as such. 2.5 CTD: Sensor Calibration 2.5.1 Temperature Pre- and post- cruise laboratory calibrations are available from July 1992 and April 1993, respectively. They were performed over the whole range at 2 K intervals between -1°C and 28°C. As a secondary standard served a Rosemount Pt25 resistance in a bridge made by SIS, Kiel. The Pt25 was calibrated according to the ITS90. Prior to the CTD calibrations, bias and linear coefficient of the Pt25 basic calibration were adjusted to meet the triple point of water (2 cells independently) and the melting point of Gallium. The adjustments were small (less 1 mK). The quadratic term is believed not to change. A polynomial regression for the CTD's correction to T90 in pre- and post-cruise Calibrations (Tables A1 and A2) shows standard deviation of less than 1 mK with about 10 degrees of freedom. The drift of the sensor output was small (1.5 mK/a at 0°C). High order polynomials are needed to correct for the MKIIIB typical nonlinearity close to 0°C (see Mueller et al., 1995). From these results, temperature outputs TCTD were corrected for both laboratory calibrations and then interpolated in time to the mean cruise date (Tables A1, A2). Figure 2 shows the corrections applied to the CTD temperatures in the bottle file. 2.5.2 Pressure Two aspects are important with the calibration of the Paine strain gauge pressure sensor: (i) nonlinear and temperature dependent static responses to pressure changes (including a hysteresis during up-profiles) and (ii) dynamic response to fast temperature changes. Corrections from, both, the static (PRC) and the dynamic responses (PDYN) are superposed linearly to the sensor output PCTD. The procedure has been described in more detail by Mueller et al. (1994, 1995). PRES = PCTD + PRC + PDYN Static laboratory calibration is performed on a Budenberg dead-weight tester in loading mode up to 6000 dbar in 500 dbar intervals with the pressure sensor being immersed in a water bath of different temperatures, i.e 13 calibration points at fixed temperatures. At the same temperatures, unloading calibrations are achieved in 500 dbar intervals starting at maximum pressures of 2000 dbar, 4000 dbar and 6000 dbar. All calibration points are arranged in a single table. For the loading mode, for each temperature polynomial correction coefficients are calculated (PRC=POLY(PCTD,TEMP). Typical standard deviations in a 3rd to 5th order polynomial regression are less than 1 dbar. The dynamic response model used is written: PDYN = k * (T1l - T2l) where T1l and T2l are lagged from the CTD temperature sensor at record time t(j): Tl1(j)=TCTD(j) + (Tl1(j-1)-TCTD(j))*exp(-(t(j)-t(j-1))/tau1) Tl2(j)=Tl1(j) + (Tl2(j-1)-Tl1(j)) *exp(-(t(j)-t(j-1))/tau2) The three coefficients tau1, tau2 and k are the two time constants representing the temperature response time at the outer (tau1) and the inner (tau2) part of the pressure sensor, respectively, and an amplitude that typically amounts to 0.2 dbar/K. These coefficients are calculated from a laboratory dunck test with the pressure sensor being duncked from a warm (20°C) water pool into a cold (0.5 C) water pool. The sensor is kept there until full response is achieved and duncked back to the warm water pool again. With the dynamic correction applied, the error in the pressure sensor output can be reduced to less than 30% of its amplitude. To process the pressure record in CTD profiles of M28/2, it was assumed that the CTD was in temperature equilibrium before the profile started. Then, for the lowering part pressure measurements were corrected with the polynomial regressions that are valid for the two temperatures that bracket the in-situ temperature with the bias being replaced by the 'in water' offset. The two resulting corrections are linearly interpolated with respect to temperature. If the in-situ temperature was outside a calibration interval the correction was constantly set forth. Finally, the dynamic correction was added. On the way up, hysteresis plays a role, and simple regressions are not possible. Therefore, CTD pressure measurements in the rosette file were corrected by linear interpolation within the calibration table with the offset being replaced by the 'in water' offset. Dynamic correction started with the assumption that the CTD was lowered at a mean speed of 1 m/s to its maximum pressure. For M28/2, laboratory calibrations are available for static effects from July 1992 (pre-cruise, Table A3), for static effects at from April 1993 (post- cruise, Table A4) and for the dynamic response to temperature changes from July 1992 (Table A5). They were applied as described above. The accuracy of corrected pressure values is estimated to be better than 3 dbar at full range (6000 dbar). Figure 3 shows the corrections as applied to the CTD pressure sensor records in the bottle file. 2.5.3 Conductivity and Salinity In the bottle file, bottle salinity and calibrated CTD temperature and pressure are used to calculate in-situ reference conductivity. Then, the CTD cell's output is corrected for a nonlinearity for values CCTD<=32.768 (Mueller et al., 1995) CN = CCTD -0.002 mS/cm. Next, the cell's output CCTD is compensated to temperature and pressure effects (Millard and Yang, 1993). CC = CN*(1+ alpha*(TEMP-T0) + beta*(PRES-P0)) Where alpha=-6.5e-06, T0=2.8 beta=1.5e-08, P0=3000 In-situ calibration coefficients are then estimated for the compensated conductivity measurements applying a linear least square method for a five coefficient correction CRC that includes a drift correction by profile number PROF, i.e. time (Tables A6). COND = CC+CRC where CRC = a1 + (a2+a3*CC)*CC + (a4+a5*PROF )*PROF It was found that the calibration could be done over the whole data set (Table A6, fig. 4). Let a conservative estimate of the number of degrees of freedom in the calibration be either the number of profiles from which samples are used or half of all individual samples (2 samples maximum for each profile), whatever is the minimum. From the statistics below, the precision in CTD salinity then is estimated to 0.001 psu. For stations where bottle salinities were measured, accuracy is the maximum of CTD salinity precision and bottle salinity accuracy, i.e. 0.002 psu. 2.5.4 Oxygen As no oxygen samples were drawn, the CTD oxygen sensor has not been calibrated. The oxygen sensor's current and temperature output are kept as raw data. References Boebel, O., C. Duncombe Rae, S. Garzoli, J.R.E. Lutjeharms, P. Richardson, T. Rossby, C. Schmid and W. Zenk: Float Experiment studies Interocean Exchanges at the Tip of Africa. EOS, Transactions of the American Geophysical Union, 79(1), p.1, 7-8, 1998 Grasshoff, K, M. Ehrhardt and K. Kremling (editors, 1983): Methods of Seawater Analysis. Verlag Chemie, Weinheim. Millard, R.C. and K. Yang (1993): CTD calibration and processing methods used at Woods Hole Oceanographic Institution. Techn. Rep. WHOI-93-44, 96 pp. Mueller, T.J., J. Holfort, F. Delahoyde and R. Williams (1994): Improving NBIS MK IIIB Measurements. In: WOCE Operations Manual, Vol. 3, Sect. 3.1, Part 3.13. WHP Operations and Methods (T.M. Joyce, editor), Rev. 1. November 1994. Woods Hole, MA, U.S.A Mueller, T.J., J. Holfort, F. Delahoyde and R. Williams (1995): MKIIIB-CTD: Improving ist system output. Deep-Sea Res. 42, 2113-2126. WOCE (1991): WOCE Operation Manual, Vol. 3, Sect. 3.1, Part 3.1.3. WHP Operations and Methods. WHP Office Report WHPO 91-1. Woods Hole, MA, USA, 1991. Futterer, D.K. And Cruise Participants (1998): The expedition ANTARKTIS - XIV /4 of RV "POLARSTERN" IN 1997.Berichte zur Polarforschung, 259, 39pp. Table A1: ANTXIV/4 pre-cruise temperature calibration of MKIIIB CTD, IFMK NB2, NOV 1993. TCTD and T90 are the CTD's temperature signal and the reference temperature (secondary standard), respectively. Polynomial correction of TCTD with coefficients c (values below) gives TEMP.TDIF=T90-TLAB is the residuum. TEMP = c(0) + (1 + c(1))*TCTD + c(2)*TCTD^2 + c(3)*TCTD^3 + ... Temperature calibration in ITS90 with CALTRC.M. IFMK NB2 FEB96 TCTD T90 TLAB TDIF 62181.0000 30.9871 30.9871 -0.0000 62182.0000 30.9876 30.9877 -0.0001 56430.0000 27.9313 27.9312 0.0001 50753.0000 24.9159 24.9159 0.0000 41471.0000 19.9888 19.9884 0.0004 32035.0000 14.9815 14.9822 -0.0007 22628.0000 9.9942 9.9942 -0.0000 13210.0000 5.0032 5.0031 0.0001 11294.0000 3.9881 3.9880 0.0001 9422.0000 2.9968 2.9963 0.0005 7534.0000 1.9960 1.9962 -0.0002 5651.0000 0.9990 0.9989 0.0001 4716.0000 0.5036 0.5037 -0.0001 3753.0000 -0.0069 -0.0063 -0.0006 3761.0000 -0.0021 -0.0021 0.0000 2799.0000 -0.5116 -0.5116 -0.0000 1864.0000 -1.0068 -1.0067 -0.0001 720.0000 -1.6123 -1.6125 0.0002 714.0000 -1.6155 -1.6156 0.0001 Polynomial degree is M=3 Number of data pairs is N=19 Coefficients, starting at lowest order: co(0)= -1.993698e+00 co(1)= 5.294941e-04 co(2)= 1.168238e-11 co(3)= 4.656400e-17 Statistics: Range: minimum is -1.615500e+00 maximum is 3.098760e+01 Number of data points is 19 Degree of fit is 3 Degree of freedoms is 15 Test sigq=rms/(N-M) is 1.740346e-05 Mean error is 1.940402e-15 66 perc error, rms is 2.784554e-04 95 perc error, 2*rms is 5.569108e-04 99 perc error, 3*rms is 8.353661e-04 Minimum of error is -6.637854e-04 Maximum of error is 5.292969e-04 Table A2: ANTXIV/4 post-cruise laboratory pressure sensor calibration of MKIIIB CTD, IFMK NB2, NOV 1993. Calibration with the sensor immersed into a bath at two temperatures (1°C and 10°C). Unloading modes starting at different maximum pressures. Pressure calibration with CALPRC.M. IFMK NB2 MAY96 Input data with PCTD at reference pressure and temperatures: N O T E : If spikes were removed do not use the last table in the output. Repeat calculation then with spikes removed from start on: TEMP 0.2 0.5 0.7 0.3 10.9 11.2 11.3 11.0 24.9 24.8 24.8 24.9 PRES 0.0 1.5 2.6 2.5 2.9 2.5 2.6 2.6 3.3 1.9 2.7 2.9 3.2 500.0 500.3 505.6 505.1 505.4 501.0 506.1 505.7 506.3 500.4 505.9 505.5 505.8 1000.0 1002.1 1008.4 1007.1 1008.2 1002.7 1008.7 1007.7 1008.7 1002.1 1008.4 1007.3 1008.2 1500.0 1503.9 1509.2 1506.9 1509.1 1504.3 1509.7 1507.4 1509.7 1503.6 1509.2 1506.9 1508.9 2000.0 2005.0 2009.0 2004.9 2009.0 2005.3 2009.3 2005.3 2009.3 2004.5 2008.6 2004.8 2008.5 2500.0 2505.3 2507.9 -9999.0 2507.8 2505.5 2508.1 -9999.0 2508.2 2504.4 2507.2 -9999.0 2507.2 3000.0 3005.0 3006.4 -9999.0 3006.7 3005.0 3006.6 -9999.0 3006.8 3003.9 3005.6 -9999.0 3005.8 3500.0 3504.4 3504.9 -9999.0 3505.4 3504.5 3505.1 -9999.0 3505.7 3503.1 3504.0 -9999.0 3504.3 4000.0 4003.9 4003.6 -9999.0 4004.6 4003.7 4003.9 -9999.0 4004.5 4002.4 4002.5 -9999.0 4003.3 4500.0 4503.5 -9999.0 -9999.0 4504.0 4497.1 -9999.0 -9999.0 4500.9 4501.9 -9999.0 -9999.0 4502.4 5000.0 5003.5 -9999.0 -9999.0 5003.7 5003.4 -9999.0 -9999.0 5003.6 5001.8 -9999.0 -9999.0 5001.8 5500.0 5503.7 -9999.0 -9999.0 5503.9 5503.4 -9999.0 -9999.0 5503.9 5501.6 -9999.0 -9999.0 5502.0 6000.0 6004.3 -9999.0 -9999.0 6004.6 6004.3 -9999.0 -9999.0 6004.4 6002.4 -9999.0 -9999.0 6002.2 Loading curve at temperature T0= 0.5 PCTD PREF PPOL PDIF 1.5 0.0 1.5 -1.5 500.3 500.0 499.5 0.5 1002.1 1000.0 999.8 0.2 1503.9 1500.0 1500.1 -0.1 2005.0 2000.0 2000.2 -0.2 2505.3 2500.0 2500.1 -0.1 3005.0 3000.0 2999.9 0.1 3504.4 3500.0 3499.9 0.1 4003.9 4000.0 4000.0 0.0 4503.5 4500.0 4500.0 -0.0 5003.5 5000.0 5000.1 -0.1 5503.7 5500.0 5499.9 0.1 6004.3 6000.0 6000.0 -0.0 Coefficients for static correction at temperature T0=0.5°C PRES(T0)=PCTD(T0)+Pol(PCTD(T0)) Polynomial degree is M=5 Number of data pairs is N=13 Coefficients, starting at lowest order: co(0)= 0.000000e+00 co(1)= 0.000000e+00 co(2)= -3.963898e-06 co(3)= 1.985301e-09 co(4)= -3.434035e-13 co(5)= 1.988652e-17 Statistics: Range: minimum is 0.000000e+00 maximum is 6.000000e+03 Number of data points is 13 Degree of fit is 5 Degree of freedoms is 7 Test sigq=rms/(N-M) is 5.699552e-02 Mean error is -7.872762e-02 66 perc error, rms is 4.559641e-01 95 perc error, 2*rms is 9.119283e-01 99 perc error, 3*rms is 1.367892e+00 Minimum of error is -1.499991e+00 Maximum of error is 4.644453e-01 CTD pressure output first order corrected with respect to loading at T0= 0.5 TEMP 0.2 0.5 0.7 0.3 10.9 11.2 11.3 11.0 24.9 24.8 24.8 24.9 PRES 0.0 1.5 2.6 2.5 2.9 2.5 2.6 2.6 3.3 1.9 2.7 2.9 3.2 500.0 499.5 504.8 504.3 504.6 500.2 505.3 504.9 505.5 499.6 505.1 504.7 505.0 1000.0 999.8 1006.1 1004.8 1005.9 1000.4 1006.4 1005.4 1006.4 999.8 1006.1 1005.0 1005.9 1500.0 1500.1 1505.4 1503.1 1505.3 1500.5 1505.9 1503.6 1505.9 1499.8 1505.4 1503.1 1505.1 2000.0 2000.2 2004.2 2000.1 2004.2 2000.5 2004.5 2000.5 2004.5 1999.7 2003.8 2000.0 2003.7 2500.0 2500.1 2502.7 -9999.0 2502.6 2500.3 2502.9 -9999.0 2503.0 2499.2 2502.0 -9999.0 2502.0 3000.0 2999.9 3001.3 -9999.0 3001.7 2999.9 3001.6 -9999.0 3001.8 2998.8 3000.5 -9999.0 3000.7 3500.0 3499.9 3500.4 -9999.0 3500.9 3500.0 3500.6 -9999.0 3501.2 3498.6 3499.5 -9999.0 3499.8 4000.0 4000.0 3999.7 -9999.0 4000.7 3999.8 4000.0 -9999.0 4000.6 3998.5 3998.6 -9999.0 3999.4 4500.0 4500.0 -9999.0 -9999.0 4500.5 4493.6 -9999.0 -9999.0 4497.4 4498.4 -9999.0 -9999.0 4498.9 5000.0 5000.1 -9999.0 -9999.0 5000.3 5000.0 -9999.0 -9999.0 5000.2 4998.4 -9999.0 -9999.0 4998.4 5500.0 5499.9 -9999.0 -9999.0 5500.1 5499.6 -9999.0 -9999.0 5500.1 5497.8 -9999.0 -9999.0 5498.2 6000.0 6000.0 -9999.0 -9999.0 6000.3 6000.0 -9999.0 -9999.0 6000.1 5998.1 -9999.0 -9999.0 5997.9 Table A3: ANTXIV/4, MKIIIB CTD, IFMK NB2, APR 1993, pressure senor's dynamic response to temperature changes. Coefficients are outerand inner sensor time constants tau1 and tau2 and the amplitude k (Mueller et al., 1995; see text). Coefficients for dynamic pressure correction tau1/s tau2/s ishift/s k/(dbar/K) 52.0518 1530.1525 345.7106 0.1193 Table A6: ANTXIV/4, MKIIIB CTD, IFMK NB2: Calibration of conductivity cell. Model CRC=a1 + (a2+a3*C)*C + (a4+a5*PROF )*PROF Vector of coefficients: 1 -0.0015 2 3.4870e-04 3 0 4 1.0001e-04 5 -1.2447e-06 Final statistics of residuals: Number of cycles N=60 Cond. Salinity mS/cm psu Min -0.0026 -0.0031 Max 0.0028 0.0032 Mean 0.0001 0.0000 Median -0.0000 -0.0000 Std. 0.0014 0.0016