Table B-3. Iceberg Size Distribution (SLAR) 1984 - 1985 



1985 IIP experiment range from 

 growler through medium. The 

 targets selected for this study 

 were small and medium icebergs 

 (ground truthed by on-scene 

 vessels) and ship targets of similar 

 radar retum. Small and medium 

 icebergs represent 59% of the 

 icebergs recorded by IIP SLAR in 

 1 984 and 68% in 1 985, as shown 

 in Table B-3. These percentages 

 are comparable to the pre-SLAR 

 value of 64% for the period 1 960 

 through 1982. 



The second mitigating 

 factor is the ambiguity of the 

 targets used, i.e., the absence of 

 cues. Since the methods of this 

 study eliminated these cues, the 

 targets used represented the 

 most ambiguous available. 



In order to assess the 

 impact of these results on IIP 

 iceberg reconnaissance, it is 

 necessary to estimate the 

 proportion of IIP SLAR targets 

 that are cueless. It can be 

 conservatively assumed that 40% 

 of SLAR targets are cueless, 

 based on IIP operational 

 experience. Indications that this 

 is a reasonably conservative 

 assumption are that the data set 

 used for this study in which 74 of 

 1 60 targets (46%) were cueless, 

 and the fact that 65% of IIP 

 search flight mileage offers 200% 

 search coverage, which is 

 assumed to greatly increase the 

 probability of cues being present. 

 A further assumptbn is that the 

 presence of cues results in 1 00% 

 correct identifcation 



Applying a worst-case error 

 rate of 45% to the (estimated) 

 cueless 40% of the small and 

 medium icebergs detected by 

 SLAR, yields an estimated SLAR 

 error of 161 and 68 misidentified 

 icebergs in 1 984 and 1 985, in the 

 small and medium size range. 



Conclusions 



The probability of correctly 

 identifying ambiguous (cueless) 

 iceberg and ship SLAR targets is 

 just above chance (55-60%). 

 Therefore, the International Ice 

 Patrol uses search tactics to 

 maximize cues and visual 

 confirmation during SLAR 

 reconnaissance. 



Based on this limited study 

 of cueless SLAR targets, the 

 SLAR error rate and iceberg bias 

 of SLAR operators could inflate 

 the number of icebergs that IIP 

 reports. This inflation is 

 insignificant when compared with 

 the increased efficiency that 

 SLAR provides iceberg 

 reconnaissance. Even though 

 visual searches provide 

 unquestionable identification, 

 they were historically flown only 

 on 50% of the deployment time 

 and each visual flight covered 

 one-third less area than a SLAR 

 flight does. 



An important issue not 

 addressed by this study is the 

 SLAR identification error rate for 

 unambiguous targets, i.e., targets 

 with cues. If this error is 

 quantified by further study, a 

 better estimate of the overall error 

 rate would be possible. 



References 



Lapin, L. 1975. Statistics: 

 Meaning and Methods. Harcourt, 

 Brace and Janovich. New Yort<. 



Robe, R.Q., N. C. Edwards, Jr., D. 

 L. Murphy, N. B. Thayer, G. L. 

 Hover, M.E.Kop. 1985. 

 Evaluation of Surface Craft and 

 Ice Target Detection Performance 

 by the AN/APS- 135 Side- 

 Lool<ing Airt)orne Radar (SLAR), 

 CG-D-02-86. U. S. Coast Guard, 

 Washington, DC 



Rossiter, J. R., L. D. Arsenault, A. 

 L. Gray, E. V. Guy, D. J. Lapp, R. 

 O. Ramseier, E. Wedler. 1984. 

 Detection of Icebergs byAirtxirne 

 Imaging Radars^ Proceedings of 

 the 9th Canadian Symposium on 

 Remote Sensing, St. John's, 

 Newfoundland, Canada 



55 



