Range versus frequency averaging of underwater propagation loss for soundscape modelinga)

Guidance on efficient methods is needed for the practical application of modeling the sound field from broadband sources such as vessels, seismic surveys, and construction activities. These sound field models are employed for estimating how changes in the soundscape will affect marine life. For efficiency, acoustic propagation modeling is often performed in bands (decidecade or 13-octave), where propagation loss modeled for central frequency is assumed to represent an average propagation loss in the band. This shortcut comes at the expense of accuracy, which can be rectified by averaging the propagation loss across many frequencies in the band. Alternately, the equivalence of range and frequency averaging was shown by Harrison and Harrison [J. Acoust. Soc. Am. 97, 1314-1317 (1995)]. However, when and how to apply range averaging required further investigations. A simple environment with a flat sandy bottom and an isovelocity water-column sound speed profile was considered to test the agreement between the range and frequency averages for decidecade bands typically considered in soundscape modelling (10-1000 Hz). The optimal range smoothing window is a Gaussian window with a width of 10%-16% of the range from the source that switches to a width fixed beyond 20 km distance from the source.