Figure D-7. Iceberg and TOD trajectories for Case IV 



winds and currents were used. 

 For these three cases, the drift 

 errors were 52-73% of the total 

 predicted drift; for Case II the drift 

 error was 22% of the total 

 predicted drift. While it is 

 tempting to suggest that the 

 estimated position error be linked 

 to the total length of the 

 predicted drift (distance along the 

 predicted path), no clear 

 guidance can be given based on 

 these results. The limited data 

 show that if there is a TOD 

 providing current information in 

 the vicinity of a drifting iceberg, 

 the model will probably produce 

 positions that are wrthin the IIP 

 error limits. If only geostrophic 

 data are available, the errors can 



be substantially larger, even for 

 drifts of short duration. This issue 

 is particulariy important when an 

 iceberg is being used to set the 

 limits of iceberg threat. 



The importance of 

 collecting current data as close as 

 possible to the tracked iceberg 

 cannot be overemphasized. 

 Early in Case III, when the TOD's 

 and the iceberg separated 

 rapidly, there was no 

 improvement in the model errors 

 when observed inputs were 

 entered. Later in the drift period 

 (afterthe buoys were 

 redeployed), the model errors 

 were smaller when the observed 

 data were used. 



Finally, the results of this 

 study provide some guidance on 

 the deployment of IIP operational 

 TOD's. Although TOD drift data 

 directly north of Flemish Cap are 

 useful, the results of Case II 

 showed that the model 

 performed within the error 

 estimates using the geostrophic 

 currents. The TOD's deployed in 

 the Labrador Cuaent (Case I) and 

 south of Flemish Pass (Cases III 

 and IV), on the other hand, 

 provided bigger payoffs. 



75 



