                   CTD CALIBRATION: DISCOVERY CRUISE 181                   
                                                                           
1) Instrumentation                                                         
Two  Neil Brown Instrument Systems (NBIS) Mark III CTDs were taken  on  the
cruise, although only one was used.  After an initial problem with the deck
unit  (a loose wire) the system worked well.  An old Beckman oxygen  sensor
was  used  at the beginning of the cruise but by cast 11820 it  had  become
obvious that this was not giving sensible output.  It was replaced by a new
sensor  for cast 11821 and the data collected prior to this were abandoned.
A  SeaTech  1m  transmissometer was incorporated into the  CTD  system  and
worked  well  throughout the cruise.  The multisampler and  rosette  caused
problems  by  consistently  misfiring  at  depth.  A  full  report  of  the
difficulties  encountered  is given in the cruise report (Pollard  et  al.,
1989).                                                                     
                                                                           
2) Data Capture and Reduction                                              
All CTD,  oxygen and transmissometer data were logged on the NBIS deck unit
and displayed on a BBC microcomputer in real time.  Data were also recorded
on digidata tapes as a back up to the main computing system.  The data were
logged via a level A interface to the level B Plessey computer for transfer
to tape and to the level C system: three Sun 3/60 workstations. The level A
despiked and averaged the data from 16 Hz to 1 Hz.  The data were processed
by the procedure outlined by Pollard, Read and Smithers (1987).            
                                                                           
3) CTD Calibration                                                         
The initial calibrations applied by CTDCAL were as follows:                
                                                                           
   Pressure (dbar)       P = 0.0 + P(raw) * (0.1 * 0.99286144)             
   Temperature (deg C)   T = 0.020909 + T(raw) * (0.0005 * 0.998965925)    
   Conductivity          C = 0.0 + C(raw) * (0.001 * 1.00109073)           
                                                                           
The  temperature  calibration  was  obtained in  the  laboratory  with  the
platinum resistance thermometer calibrated in-situ. Deep temperature values
(at  3000  m or deeper) were compared with sample measurements taken  by  a
digital reversing thermometer. Over 23 samples, the mean difference between
temperature  sensor and reversing thermometer was 0.0088 +/- 0.001  deg  C.
Very  few  adequate readings were obtained near the surface because of  the
failure  of the second digital reversing thermometer.  On this evidence  no
change was made to the temperature calibration.                            
                                                                           
Transmittance was also recorded although not shown in this report.  It  was
calibrated with the equation:                                              
                                                                           
   Transmittance (%)  T = 20 * V * 1.0032                                  
                                                                           
where                                                                      
                                                                           
   V (output voltage)   = (4.355 / 4.096) * (T(raw) - 0.0) * 0.0001        
                                                                           
4) Absolute Salinity Calibration                                           
At  the beginning of the cruise six full depth casts were made with  twelve
water  samples  each,  analysed on the Guildline  Autolab  salinometer  for
calibration purposes.  These gave a consistent offset over the entire  cast
which  was comparable to the offset obtained by comparing the 0/S  profiles
to the curve obtained by Saunders (1986) for the deep,  stable water of the
North East Atlantic. A constant offset was therefore applied to salinity to
match the Saunders curve.  After the first few casts the number of salinity
samples  drawn  at each station was reduced to four.  These  were  analysed
during  the  cruise on the Guildline salinometer,  but examination  of  the
results  after  the cruise suggested that the salinometer was  not  working
reliably,  standardisations  on successive sessions showing large  offsets.
All  the  CTD  0/S  profiles  were  compared  to  the  Saunders  curve  and
adjustments  were  made during the cruise as the conductivity cell  drifted
slightly.  The bottle values were compared with the CTD data and gave  poor
statistics  but it is thought that this is most likely to be the result  of
problems  with  the salinometer rather than the  conductivity  sensor.  The
differences  between  bottle  samples and CTD values within 100  m  of  the
surface  are  shown in Fig.  1,  together with the  correction  applied  by
comparison with the Saunders deep curve.                                   
                                                                           
5) Oxygen Calibration                                                      
Calibration  of  the  Beckman  oxygen sensor caused  more  than  the  usual
problems  (Read,  1989).  Data from the casts prior to station  11821  were
discarded  as the old sensor did not give sensible output.  Data  from  the
replacement were given an initial calibration of:                          
                                                                           
   oxygen current       OXYC = 0.0 + OXYC(raw) * (0.001 * 1.05)            
   oxygen temperature   OXYT = 0.0 + OXYT(raw) * (0.128 * 1.0)             
                                                                           
However,  OXYT  was discarded in favour of CTD  temperature  (see  Pollard,
1985).                                                                     
                                                                           
The oxygen sensor was then calibrated with the formula:                    
                                                                           
   oxygen   O = OXYC * rho * exp (alpha* T + beta * P) * OXYSAT            
                                                                           
using a least squares fit to calculate the coefficients rho, alpha and beta
from  bottle samples.  With between ten and twelve samples for  each  cast,
this  should have given a good fit.  However,  it soon became apparent that
many  of  the  bottle sample values were imprecise (Read,  1989)  and  good
values  had  to be carefully selected before the  calibration  coefficients
could be calculated.  This still gave widely varying results for rho, alpha
and  beta for each cast,  so a best fit of the deep gradient  over  several
selected casts was used to fix alpha and beta,  and rho was then calculated
individually for each cast.  This correction improved the fit to the sample
data  but  still  caused the overall value of oxygen to vary  up  and  down
between casts.  The problem was particularly pronounced when casts 11824  -
11845 were contoured.  To minimise the cast to cast variations, the deepest
data  were fitted to the oxygen value given by Saunders (5.67 ml/l) and the
offset added to the whole cast.                                            
