Underwater sound propagation modeling and the predictive probability of detection framework

The predictive probability of detection framework (PPD) provides a measure of uncertainty to the binary problem of signal detection in the situation of fluctuating received signal. The fluctuations can be from many causes, including short-term and difficult to predict propagation variability, long-term and possible to predict propagation variability, and noise variability. In PPD, the density functions covering all fluctuating-causing processes are combined and analyzed with respect to a benchmark to give a detection probability. In the traditional PPD analysis, the benchmark propagation loss is a somewhat general function of range, as are other quantities such as array gain. These provide reference levels for signal to noise ratio evaluation. With more detailed data-informed propagation modeling, including 3-D modeling, additional structure can be given to those functions, and the stochastic fluctuation aspects of the method can potentially be reduced, thereby altering computed maps of probability of detection. Example computations will be shown for spatially heterogeneous environments: canyons, slopes, and internal wave areas. [Work supported by the Office of Naval Research.]The predictive probability of detection framework (PPD) provides a measure of uncertainty to the binary problem of signal detection in the situation of fluctuating received signal. The fluctuations can be from many causes, including short-term and difficult to predict propagation variability, long-term and possible to predict propagation variability, and noise variability. In PPD, the density functions covering all fluctuating-causing processes are combined and analyzed with respect to a benchmark to give a detection probability. In the traditional PPD analysis, the benchmark propagation loss is a somewhat general function of range, as are other quantities such as array gain. These provide reference levels for signal to noise ratio evaluation. With more detailed data-informed propagation modeling, including 3-D modeling, additional structure can be given to those functions, and the stochastic fluctuation aspects of the method can potentially be reduced, thereby altering computed maps of probability of det...