Appendix D 



An Evaluation of the International Ice 

 Patrol Drift Model 



D. L. Murphy 



LT I. Anderson, USCG 



Introduction 



Since 1979, International 

 Ice Patrol (IIP) has been using an 

 iceberg drift model as an integral 

 part of its iceberg tracking 

 operations. During the season of 

 maximum iceberg threat, typically 

 March through August, IIP 

 conducts aerial reconnaissance 

 of its operations area (40*^ - 52^, 

 39^ - 57*\V) on alternate weeks. 

 During the week that the IIP Ice 

 Reconnaissance Detachment 

 (ICERECDET) is deptoyed to 

 Gander, Newfoundland (IIP field 

 operations base), daily flights are 

 conducted on five consecutive 

 days, each covering only a small 

 portion of the IIP operations area. 

 As a result of this reconnaissance 

 schedule, IIP must often rely on 

 the nxxiel predictions to set the 

 limits of iceberg danger during 

 periods when no ice 

 reconnaissance is being 

 conducted. In addition, the 

 nnodel drift predictions are used 

 to help recognize icebergs that 

 have been prevtously sighted, 

 either by the ICERECDET or 

 merchant vessels. Lacking this 

 ability to recognize iceberg 

 resights has the effect of inflating 

 the numbers of icebergs south of 

 48°N, the traditional indicator of 

 the severity of an iceberg season. 



Despite the reliance that 

 IIP places on the accuracy of the 

 drift model results, relatively little 

 testing of the nfx>del has been 

 possible, primarily because 



68 



adequate iceberg drift data, with 

 accompanying environmental 

 data, are expensive and often 

 difficult to obtain. Moreover, only 

 in the last few years has 

 navigation in the operations area 

 been accurate and reliable 

 enough to permit the collection of 

 good data. 



Mountain (1980) tested 

 the nfX}del using the tracks of two 

 large tabular icebergs, a large 

 pinnacle iceberg, and a freely- 

 drifting satellite-tracked buoy. 

 The drift durations were from 3 to 

 25 days. The results were quite 

 variable, ranging from a small 9km 

 error for the 3-day drift to a 

 constant 90-1 50km drift error in 

 the 25-day case. Although he 

 recognizes the limitations of this 

 small data set, he suggests that 

 the primary cause of the model 

 error is due to inaccurate inputs, 

 i.e., winds and cun-ents. 



This report describes the 

 results of four case studies in 

 which the perfonnance of the IIP 

 iceberg drift model was examined 

 at four different locations (Figure 

 D-1) in the IIP operations area. 

 The objectives were twofold: 

 first, to test the accuracy of the 

 drift predictions of the operational 

 IIP iceberg drift model, and 

 second, to investigate how the 

 accuracy changes when on- 

 scene measured wind and 

 current data are used to drive the 

 model. 



Model Description 



Mountain (1980) describes 

 the details of the IIP operational 

 drift model; thus, only a brief 

 outline is presented here. The 

 fundamental model balance is 

 between iceberg acceleration, air 

 and water drag, the Coriolis 

 acceleration and a sea surface 

 slope term. The resulting 

 differential equations are solved 

 using a fourth-order Runge-Kutta 

 algorithm. The model is driven by 

 a water cun'ent which combines a 

 depth- and time-independent 

 geostrophic f tow with a depth- 

 and time-dependent current 

 driven by the tocai wind (time- 

 dependent Ekmanftow). 



When used operationally, 

 the IIP drift nnodel emptoys a 

 nnean geostrophic current field 

 based on many years of 

 hydrographic sun/eys (Scobie 

 and Schultz, 1 976). It is on a grid 

 of 20 minutes of latitude by 20 

 minutes of tongitude, except for 

 the Labrador Cun-ent, which is 

 defined on a more detailed grid of 

 10 minutes of tongitude. Wind 

 data, on a 1 degree of latitude by 

 2 degrees of tongitude grid, are 

 provided to the nrx>del every 1 2 

 hours from the surface-wind 

 analysis of the U. S. Navy Fleet 

 Numerical Oceanography Center 

 (FNOC). 



Finally, the nxxJel requires 

 as input the mass and cross- 

 sectional area of the drifting 

 iceberg. Obviously, IIP 

 reconnaissance operations do 

 not pennit precise measurement 

 of each detected iceberg. Often, 

 IIP tocates icebergs using the 

 side-looking airtxime radar 



