error at 45° North latitude. The two curves 

 reflect a total error between 2 stations of 0.0088 

 DM for gradient sampling and 0.0130 DM for 

 standard depth sampling. For stations 10 miles 

 apart, the standard depth sampHng method can 

 cause an error of up to 6.8 cm/sec or about one- 

 tenth of a knot. In some low current areas, the 

 velocity of the surface current could be of this 

 magnitude resulting in a 100-percent error factor. 

 The gradient samphng error is slightly better, 

 showing a 4.6 cm/gec possible error. These 

 current velocity errors do not reflect any instru- 

 ment measurement error which coidd add more 

 than 0.01 DM of error to dynamic height calcu- 

 lations. 



The close comparison between the dynamic 

 heights determined from standard depth sampling 

 and gradient sampling indicates that the vertical 

 distribvition of temperature and salinity at a 

 number of stations does not have to be com- 

 pletely described in detail in order to do dynamic 

 calculations for the sea surface. A plot of the 

 specific volume anomaly (5) versus depth is, in 

 general, quite uniform with only gradually 

 changing gradients. In many cases only a few 

 data points in the water column are required to 

 draw a curve which describes the mass distri- 

 bution. Due to the regularity of change of the 

 anomaly curve within a stable water mass, the 

 trapezoidal rule employed in the numerical 

 integration of a water column for dynamic 

 heights will give satisfactory answers in many 

 cases from a curve constructed with only a few 

 points. As an example, figure 8D shows large 

 alternating changes of temperature and salinity 

 with depth; however, the specific volume anomaly 

 versus depth curve is extremely regular, indicating 

 that this particular water column could have been 

 sampled by almost any method and with a mini- 

 mum of data points for the determination of the 

 surface dynamics. However, some stable water 

 columns do not have so smooth a curve and 

 many sample points woidd be necessary for an 

 adequate description. 



For Ice Patrol survey work, dynamic height 

 information and a complete description of temper- 

 ature and salinity distribution must be obtained 

 quickly. These two aims require two different 

 sampling techniques: first, only a few data points 

 critically located and rapidly obtained are needed 

 in any given water column for the adeqiuite 

 determination of dynamic heights based on a 

 smooth curve (the exact number of samples 



required needs further study because it depends 

 on the water mass involved) ; second, a detailed 

 description of the properties requiring a great 

 number of point samples which is virtually pro- 

 hibitive from the synoptic standpoint. In this 

 situation the ultimate solution is attainable only 

 with a continuous sampling device. A sampler 

 such as this could be lowered in a water column 

 and provide as many data points as considered 

 necessary to fulfill both requirements of Ice 

 Patrol sm-veys along with adding to the rapidity 

 with which a survey can be conducted. 



The state of the art in the field of electronic 

 continuous samplers leaves much to be desired. 

 Whereas the point sampling technique using 

 Nansen bottles and reversing thermometers can 

 miss maximum-minimum temperature informa- 

 tion and also fail to describe the gradients of 

 density exactly, the inherent inaccuracies of a 

 continuous sampler also pose a problem. 



In using a continuous sampler for the accom- 

 plishment of Ice Patrol survey aims the data 

 abstracted from its output must be sufficiently 

 accurate to perform dynamic height calculations. 

 BasicaUy, two causes of error must be considered 

 when comparing the final results of the point 

 (Nansen bottles) versus continuous sampling 

 systems : 



(o) missed density gradients 

 (6) instrument accuracy 



As discussed above, a good approach to the true 

 dynamic height of a water column can be made 

 from several data points and a smooth drawn 

 anomaly curve. This method could give accu- 

 racies to within ±2 dynamic centimeters or 

 better depending on the water column. However, 

 the possibility of missing the gradients still exists 

 and generally the missed gradient error must be 

 minimized because of additional instrument errors 

 that must be considered in any sampHng system. 



In general, with a fixed number of samples of 

 the vertical distribution of some property, the 

 error in describing the distribution by point sample 

 and using straight fine integration between these 

 points for dynamic computations is shown in 

 figure 9D and given by the relationship : 



_ change of property gradient 



error— ^^^^^ number of samples over given distance 



where the change of gradient of some property (P) 



is -rFT„f and tlie fixed number of samples (n) over 



a given distance is defined by y;. At the limit 



67 



