Estimating Detection Probabilities Fdr Search and Rescue

Abstract

In planning visual search operations, the allocation of Search and Rescue (SAR) resources may be dependent on the statistical probability that a missing aircraft o r crash site can be detected if it is withm given geogra-cal areas. The present study was undertaKen in order t o relate probability of detection (POD) to the relevant enviromntal o r contextual variables. Multidimensional scaling (MDS) and cluster analysis were used to determine the basic stimlus dimensions and characteristics that define different categories of search area. A simulated visual search task generated POD data for representative aerial views that contained synthetic crash sites. POD was a function of target contrast and terrain exposure. Limited field trials have indicated that data from the static visual displays were capable of providing reasonable estimates of detection in dynamic visual search INI'RODUCTIONIn planning visual search operations, the allocation of Search and Rescue (SAR) resources may be based on the statistical probabilities (a) that the aircraft is in one of several geograpucal areas and (b) that the crash site can be detected if it is within any one of the given areas.A model for determining probability areas is currently available (Isbrandt and Hypher, 1979) and the effects of procedural variables, including visual scanning patterns, on visual search have been reported in previous papers (Stager, 1974;Stager and Angus, 1978). Hypher (1985) has recently described the effects of vegetation masKing on airborne search.The present focus was on target/background or contextual variables, such as contrast, that can affect visual detection m e study was undertaKen to develop probability of detection (POD) values that would relate POD to the relevant environmental or contextual variables.The initial problemwas to determine the dimensions that would form the basis for a meaningful classification of these variables. MU^ t iclimensional scaling (MDS) (Carroll and chang, 1970;young, 1984;Young and Lewyckyj, 1980) and cluster analysis (Sarle, 1985) of aerial photographs seemed to provide a method for deriving the most salient environmental characteristics and conceptual dimensions underlying visual sear& A conceptual schema based on data frm operational SAR personnel would reflect (a) their experience with different search environments and (b) their cognitive structure and perception of the search task W s type of approach to, task analysis would ensure that the organization of any environmental classification system was meaningfully based on the SAEWECHs' approach to the performance of the search task.In the present study, various combinat ions of terrain, vegetation, and seasonal conditions that were derived from the MDS analyses were distinguished as clusters or categories of search area. POD data were then generated for these separate types of search area in a simlated visual search task 1. MULTIDIMENSIONAL SCALING OF SEARCH TASK MEXHCD a;u33.i ec t s

Gaze behavior of spotters during an air-to-ground search

“…A downed aircraft must be located quickly so that any survivors can be rescued and treated for injuries. Considerable effort [Christopher and Fournier 1993;Saunders 1987;Stager 1974;Stager 1986;Stager and Angus 1975;Stager and Angus 1978;Taylor and Mack 1988;Vigneault and Young 1989] has been devoted to understanding search effectiveness of spotters, the personnel who sit in the aircraft and look for the target, and to improve spotting success. However, other measures of performance such as eye movement behavior have not been assessed during an actual air-to-ground search.…”