depth will be adequately described. Past Ice 

 Patrol data indicate stable water columns below 

 100 meters are the rule. 



Ice Patrol requirements are better fulfilled by 

 the gradient sampling method. First, better 

 temperature-salinity information is obtained for 

 the dynamic height calculations required for sur- 

 face current detemiinations and, second, a more 

 accurate picture of the vertical variations of tem- 

 perature is described for tracing and analysis of 

 the Labrador Current. As an added feature to 

 this method, it is noted from figure 4D that many 

 areas of the Grand Banks are quite unifonn in 

 their property distribution allowing fewer Nansen 

 bottles than usual to be used on a cast, thus de- 

 creasing time on station, reducing data processing 

 time, and risking less equipment. This is illus- 

 trated by Table ID, where each station on two of 

 the surveys of 1965 is compared for the number of 

 bottles used with the standard depth and the 

 gradient sampling methods. 



The advantage of the gradient sampling method 

 lies in the obtaining of better definition of the 

 property distribution. With a more accurate de- 

 scription of the temperature and salinity struc- 

 ture, the accuracy of the dynamic height calcula- 

 tions will also improve. To show this, a standard 

 model of the temperature and salinity structure 

 for each station taken during the first and third 

 surveys was prepared. All observed data points 

 obtained from the gradient sampling method were 

 used to provide a graphic plot of temperature and 

 salinity versus depth. Supplemental tempera- 

 tures were gleaned from the bathythennograph 

 records and along with the observed data were 

 plotted on H.O. 17325 temperature-salinity- 

 density paper. Corresponding supplemental sa- 

 linities were then obtained for an assumed stable 

 water column. Tlie resulting composite tempera- 

 ture-depth and salinity-depth plots are certainly 

 a close approach to describing the properties of 

 each water column. Figure 5D shows an example 

 of a property distribution plot based on observed 

 values from Nansen bottles and the bathythenno- 

 graph supplemental points. 



The dynamic heights based on observed gradient 

 sampling values w ere computed immediately dur- 

 ing the survey. Later, a second set of points 

 representing the temperature and salinity values 

 at the standard Ice Patrol depths were abstracted 

 from the composite distribution curves to 1,000 

 meters and processed on tlie computer for dynamic 

 heights. Figure 6D shows what the property 



distribution would look like if this water mass had 

 been sampled using the standard Ice Patrol depths 

 only. Although the density distribution is ade- 

 quately described, in figure 6D, important tem- 

 perature and salinity distribution detail is lacking. 

 A third set of points representing any significant 

 change in temperature and salinity witli depth 

 were abstracted from the curves and processed 

 by the computer. This third set of points was 

 sufficient to adequately describe the distribution 

 of temperature and salinity and the resulting 

 dynamic height is considered the closest approach 

 to the true value. Figure 5D, shows an example 

 of the vertical distribution of these 3 sets of 

 computer processed points: the observed Nansen 

 bottle points, abstracted standard depth points, 

 and abstracted true points. 



Table IID shows the comparison between the 

 surface dynamic heights of the two sampling 

 methods and the true value. It can be seen that 

 both methods give residts quite close to the 

 true dynamic height value of the station with the 

 gradient method giving slightly better results; 

 standard deviations are (r=± 0.0039 DM for 

 gradient sampling and (7= ±0.0052 DM for stand- 

 ard sampling. These values are for all 119 sta- 

 tions taken on the first and third surveys of Ice 

 Patrol. These stations were taken both on and 

 off the Grand Banks and include shallow stations 

 where the cumulative errors are minimal in many 

 cases. A total of 62 stations were taken in water 

 greater than 1,000 meters, the assumed depth of 

 no motion, and a more realistic comparison is 

 shown from these; (r=± 0.0044 DM for gradient 

 sampling, and (r=± 0.0065 DM for standard 

 sampling. The gradient method is, however, head 

 and shoulders above standard sampling because 

 the extreme temperature values obtained more 

 accurately describe the existing distribution. 

 These values are essential for the proper analysis 

 and interpretation of the water masses of the 

 Ice Patrol area. Table IID also presents the 

 maximum differences from the true value. In 

 the case of gradient sampling, a large difference 

 occurred due to the missing of gradients caused 

 by encountering a greater than anticipated wire 

 angle. 



The errors listed above for the gradient sampling 

 method and the standard depth method can sig- 

 nificantly influence the calculated current. The 

 influence is dependent upon the station spacing 

 and the current velocity. In figure 7D, station 

 interval is plotted against calculated current 



66 



