Predicting Acoustic Variability: Pragmatic Considerations for Selecting a Stochastic or Deterministic Approach

Sonar performance prediction models require environmental inputs, including bottom depth and composition, sound-speed profile (SSP), and wave or wind conditions. Temporal and spatial variation and measurement uncertainties in environmental inputs result in uncertain model outputs. Ideally, acoustic fluctuations could be predicted from input parameter variability using a suitable modeling framework. A data set consisting of 460 SSPs and acoustic transmission loss (TL) measurements between 300 and 900 Hz was acquired during GLISTEN15, a 2015 Mediterranean Sea trial led by North Atlantic Treaty Organization's Centre for Maritime Research and Experimentation. The TL was measured in 5-min sampling periods at fixed source-to-receiver ranges of 0.5-10 km over two separate days. A gradual change in TL as large as 15 dB over 5 min was observed at some receivers and frequencies. Predicted TL variability was compared to measured TL variability by using the SSPs as inputs to an acoustic propagation model. Purely stochastic modeling either over- or underpredicted variability, depending on the temporal and spatial separation of the SSPs used as inputs. In this article, an internal wave model derived from the SSP time series was used to demonstrate the possible effect of internal waves propagating perpendicular to the source-receiver direction. The best agreement between modeled and measured variability was obtained for internal waves of 1536-s period and 0.8-1.6-m amplitude. In this way, an approach that included a physically plausible model of internal wave dynamics grounded in a few relevant in situ measurements provided good predictions of observed TL variability and a reasonable explanation of its origin.