Figure D-6. Hourly wind vectors for Case III 



The Case III system/system 

 run (Figure D-1 1 ) produced errors 

 that increased persistently, 

 exceeding the IIP error estimate 

 after about 36 hours of drift. At 

 the end of the drift period, the 

 errorwas overSOnm {--56km), 

 which is 73% of the total 

 predicted drift. In the early part of 

 the drift perbd (<48hrs.), the use 

 of the observed current and wind 

 data produced no improvement in 

 the results; indeed, at one point, 

 the results were less accurate 

 than the system/system case. 

 This result is not surprising 

 because the iceberg moved 

 rapidly to the north while both 

 buoys remained close to the 

 deployment area. When the 

 buoys were retrieved and 

 redeployed at the iceberg (-60 

 hrs.), the mode! results computed 

 using observed data improved 

 somewhat. 



Using the observed data to 

 drive the model in Case IV (Figure 

 D-1 2) made an enormous 

 improvement in the results. The 

 system/system mn produced 

 errors between 30-45nm (-56- 

 83km) while, for the observed 

 data, the en-ors were 

 approximately half those values. 

 For most of the drift period, the 

 currents measured at 58m 

 provided more accurate model 

 results than those measured at 

 8m. At 84 hours this situation 

 reversed, and the 8m data 

 produced better results. This is 

 an expected result because as 

 this small iceberg deteriorated, its 

 rrxjtion should have been nxjre 

 consistent with the 8m currents 



74 



<2> 



Dote 

 116 



V/I // 117 



/^. ; / / ^ 



.<::^:^^.\\\ 



/ y 



llUl /.x , 



118 



119 



120 



I I 1 1 I I 1 1_ 



OOz 



I2z 



—I ■ ■ ■ i_ 



10 KTS 



23z 



W 85 



than the 58m currents. However, 

 the data are few and the 

 difference between the results 

 (8m vs. 58m) is small so there is 

 no certainty that the reversal is 

 meaningful. 



Conclusions 



No f inn conclusions can be 

 drawn from this small data set, but 

 there is some consistency in the 

 results that is worthy of note. 



In all four cases, using on- 

 scene measured data improved 

 the model accuracy over the runs 

 made using geostrophic cuaents 

 and FNOC winds. The accuracy 

 improvement was substantial in 

 two cases: Case I and Case IV. 



Thus, the results of this study 

 support the IIP practice of using 

 TOD drift data to rrxxJify the 

 geostrophic current. The more 

 widespread the use of TOD's, the 

 nrwre we can rely on the model 

 results. No attempt was made to 

 separate the improvements due 

 to on-scene current data and on- 

 scene wind data because of the 

 small anxjunt of data. However, 

 Case I, for which there were no on- 

 scene wind data, showed 

 considerable improvement when 

 on-scene current data were used 

 in the model predictions. 



In three of the four cases 

 (Case II excepted), the observed 

 drift error was larger than the IIP 

 estimated error when the system 



