entire data window, however. Tlie current pat- 

 terns in the area were complicated by the mooring 

 being on the northern fringe of the Gulf Stream. 

 Changes in Gulf Stream flow caused consequent 

 changes in the currents at the mooring. The cur- 

 rents were further changed in the second half of 

 the record by a warm core ring which was located 

 near the mooring from November 1 through 10. 



Because of the procedure used to obtain the 

 inertial oscillations, I tried to obtain dissipation 

 coefficients which would provide an optimum fit 

 for both the inertial transport solved for by the 

 above method; and for best fit when adding the 

 prediction and the computed net sea current and 

 comparing these to the data set. 



The method I used to test the model and adjust 

 the K and K, coefficients was basically trial and 

 error, and judging the quality of fit from the 

 length of the vectors connecting points of equal 

 time in both sets. One result I discovered while 

 testing the model was that the K ^, or second order 

 dissipation term controlled the velocity of the 

 model, and the K.,, or first order dissipation term 

 controlled the angle the current made with the 

 wind, although each term had an effect on the 

 other's domain. This result may be an artifact of 

 the numerical method, or could be an indicator of 

 part of the physical system. 



The data used to fit the coefficients was three 

 portions of the mooring 280 record. I chose these 

 windows to correspond with certain characteris- 

 tics of the wind record. Window I was a period of 

 relatively constant winds from generally the 

 same direction, from 26 October to 2 November. 

 Window II examined constant velocity winds and 

 varying direction. The winds varied 700 degrees 

 in direction between 3 and 12 October. During 

 Window III from 3 to 13 November, the winds 

 varied both in speed and direction greatly. 



One fundamental principle used to test the 

 computer runs aginst the data must be pointed 

 out. Because of the low amount of dissipation in 

 the wind driven system, the old current motions 

 present in the mixed layer will greatly affect the 

 currents after the start of some new wind event. 

 The results of an applied wind stress will vary 

 greatly depending on the character of the veloc- 

 ity in the mixed layer before the start of that wind 

 event. If the wind comes over an ocean at rest, the 

 velocity and direction of flow will be very differ- 

 ent than if similar conditions acted on a water 

 mass already in motion. For these reasons, the 

 wind history must be applied for some time pre- 



vious to the start of the comparison with the cur- 

 rent data. That time is taken to be four days, as 

 after that time, the results of the initial value 

 problem have been buried in the effects of the 

 later wind history (Pollard and Millard, 1970). 

 The verification of this principle can be seen in 

 later presentations of my data. 



The value for the radiative flux at the mooring 

 location was extrapolated from the data of Han- 

 son (1976). I calculated a radiation flux at 39° 

 lO'N, the latitude of the mooring, of 101.8 cal 

 cm'day' . This value is the average for the value 

 for October and November. 



I will consider the results from each window 

 separately. Data Window I covered the period of 

 26 October to 2 November. The winds throughout 

 the period were blowing towards 110° ± 30° 

 Magnetic, with a velocity of 12-17 meters sec' . 

 The current record shows strong inertial oscilla- 

 tions throughout. Figure 4.6 shows a Lagrangian 

 view of the water trajectory over the entire data 

 window, the water movement being the jagged 

 line. The smoother line is the predictions. The 

 straight lines connecting the two plots show the 

 distance between the prediction and the data 

 every 10 hours. Here the inertial oscillations in 

 the prediction are effectively buried by the addi- 

 tion of the sea current. 



Figure 4.7 shows this same data window, but 

 with the inertial transport separated from the 

 mean flow as described above. Figure 4.7 shows 

 the prediction compared to the inertial period 

 oscillations over the entire length of the data 

 window. Figure 4.8 shows the last four days of the 

 data and prediction. When comparing Fig. 4.7 

 and Fig. 4.8, it is easily seen how the delay in 

 comparing the prediction to the actual current is 

 necessary to eliminate old motions in the mixed 

 layer, as the comparison between the prediction 

 and the current is much closer after that time. 

 The pronounced change in the direction of the 

 data attests to this. 



The 4 day lag hypothesis is also supported by 

 statistical evidence. The hypothesis under study 

 was that the deviation to the left of recombined 

 net sea current and inertial currents was equal to 

 the deviation of the prediction to the right of the 

 separated inertial data. For the runs of the model 

 compared to date after days, such as figures 4.6 

 and 4.7, the hypothesis fails at all significance 

 levels. When the companion is made after 4 days, 

 however, the hypothesis passes at the 85% signifi- 

 cance level. 



14 



