The difference is 15 feet. With a stability index of 13°, the same -winds 

 would give predicted thermocline depths of 102 and 108 feet, respectively - 

 now the difference is only 6 feet. 



Error Due to Stability Index Evaluation 



The second component of random error is introduced by erroneous evalua- 

 tion of the stability index. This error is more likely to occur than that 

 of erroneous wind evaluation, because of (l) the uncertainties and lack of 

 direct means for evaluating At, and (2) the error due to correction for a 

 salinity gradient. 



With a stability index of 5°, 25-knot winds in a normal field, and 

 fully developed sea, predicted thermocline depth would be 165 feet . The 

 predicted depth with a stability index of 7° would be l48 feet . The dif- 

 ference is 17 feet. 



Error Due to Type of Current Field 



Another component of random efror is inexact evaluation of the type 

 of current produced by the wind field, whether it is normal (horizontal 

 flow), convergent, or divergent. This error is largest for low stability 

 indexes and decreases somewhat as stability index increases. 



If a fully developed sea with 25-knot winds is erroneously interpreted 

 as being strongly convergent while it is actually normal, the error in 

 predicted thermocline depth would amount to 52 feet for the At interval 

 of 2° and 1+0 feet for the At interval of 13°. 



Evaluation of the rate of convergence or divergence is a difficult 

 task. Furthermore, methods for determining the existence of convergence 

 or divergence are mainly subjective. For the time being, it is quite sat- 

 isfactory to determine one of the three types of rate O'f convergence or 

 divergence (weak, moderate, and strong) with a good approximation. The 

 solid upper and lower curves in Figures 15 through 22 correspond to strong 

 convergence and divergence. Moderate and weak rates must be interpolated 

 between the normal (central) and the corresponding upper or lower curve. 

 Convergence or divergence produced by pure wind current depends consider- 

 ably on local and seasonal conditions; for example, permanent horizontal 

 flow, vertical boundaries, mean vorticity, permanency of the convergence 

 or divergence, prevailing angle between direction of wind and propagation 

 of the wind field, barometric pressure, upwelling, etc. This error should 

 gradually decrease as experience in a given area increases. 



Components of random error are also produced by factors other than 

 those considered in this prediction technique. Most important of these 

 are mixing processes which may be acting concurrently with mechanical 

 processes to increase or decrease the total mixing effect. Of these, in- 

 stability mixing due to evaporation and surface cooling can be very signifi- 

 cant, especially in late fall. 



72 



