The following study describes a technique to improve maritime search area prediction by using consensus forecasting to quantify areas of higher probability within a model defined search area. The study included forecasting search areas for 45 five-day drifter tracks, each simulated independently using different ocean models (BLUElink, FOAM, HYCOM and NCOM) throughout 2012 in the eastern Indian Ocean, off the coast of Western Australia.

It was found that zones where all four model search areas overlapped (defined here as a consensus search area) were significantly smaller than those areas generated by any single model forecast. The average consensus search area was quantified to be up to 56.9% smaller at 24 h and 72.5% smaller at 120 h than the average single model search areas at corresponding times. However the average hit rate (the frequency that the drifter was contained within the forecast search area) for the consensus search area was reduced by up to 26.2% at 24 h and 52.8% at 120 h, when compared to average hit rates from single model search areas. This indicated that the four model consensus search area had a higher hit rate per unit of search area than any individual model search area. Hence if search resources were a limiting factor for a particular search effort, then search resources may be most effectively deployed by prioritising the effort initially to the smaller, four model consensus search area.