Andrews 



identifier teams to the reported position. 



Inside of 80 feet EOD/UDT swimmers and scuba divers were used 

 to search visually and simultaneously to identify. This operational 

 group was highly successful in area All, where they were assigned. 

 Approximately four square miles were searched out in six weeks. 

 143 pieces of aircraft debris were sighted and recovered, and the 

 search Commander could report with a degree of confidence approach- 

 ing 100% that the H-bomb was not in All inside of 80 feet. The 

 major problem reported was a difficulty in keeping navigational 

 bouys in place during heavy weather. 



THE FIRST OPERATIONS ANALYSIS (ON-SCENE) 



Two analyses of the acoustic and visual search outside of 80 

 feet were conducted. The first was on scene in support of the 

 operation. Unfortunately the team which carried out this analysis 

 did not arrive in Spain until February 22, 1966, over a month after 

 the air collision. The dispersion of original records throughout 

 the task force, and the lack of DECCA navigation during the early 

 part of the acoustic search by UQS-1, OBSS, and Sea Scanar made the 

 analysis task most difficult. Further, no measure of search effec- 

 tiveness was yet developed. The second analysis was a study of 

 reported acoustic contacts versus reported visual contacts and was 

 conducted in Washington after SALVOPS/MED was concluded. 



The definition of a measure of search effectiveness was 

 provided by Dr. Richardson^, a member of the on-scene analysis 

 group. The chief method by which search was documented prior to the 

 arrival of the analysis team was through preparation of transparent 

 overlays showing, with different cross-hatching, the search vehicles 

 which had been in the several areas. These overlays tended to be 

 misleading since they did not assess the quality of effort which had 

 been expended. The measure chosen to improve this situation was 

 search effectiveness probability (SEP), the probability that if a 

 target were in a particular area, then it would have been found by 

 the amount of search effort considered. The basic inputs to the 

 computation of SEP are initial detection probability (Pj)) and the 

 probability of correct identification (Pq) given a detection. Thus 

 (SEP)n = (PD)n (Pc)n where (SEP)n is the Search Effectiveness 

 probability in the nth area. A value for (PD)n was calculated by 

 considering in the nth area, the sensor detection range from which 

 sweep width (W) was obtained, the length of transit travel (L) in 

 the nth area, and the degradation of search effort due to navigation 

 inaccuracy. For visual search systems (e.g. MIZAR towed camera, 

 ALVIN visual search), it was assumed that a successful identifica- 

 tion always followed a detection. Thus (Pc)n was assigned a value 

 of unity for such systems. For the acoustic systems, a visual/ 

 identifier team had to return to the contact and attempt identifi- 

 cation. Thus (Pc)n for acoustic detection was assumed zero if no 

 revisit had taken place or one if an effective visual revisit of a 



10 



