Comparative analysis of underwater acoustic propagation models: Evaluating the trade-off between accuracy and computational demands for coastal noise predictions

Abstract

The increase of anthropogenic noise in the ocean presents a threat to marine ecosystems. To assess underwater noise pollution effects, acoustic propagation models are commonly used. These models differ in complexity and computational requirements. This study aims to find the optimal trade-off between accuracy and computational efficiency of propagation models in marine coastal environments. We evaluated six models by assessing their accuracy and computational demands to identify the most suitable model. The models' accuracy was determined through the comparison of their estimated received levels with measured received levels. Acoustic measurements were conducted in two areas. Both geometric spreading and Energy flux models were effective across low-frequency bands when compared to field measurements. At 1 kHz, they had high agreement and overlap with the field measurements (geometric spreading: ∼74.1 % & ∼39.7 %, Energy flux: ∼77.3 % & ∼40.3 %), while other models performed poorly. In terms of computational efficiency, geometric spreading model was the fastest, followed by Energy flux model, while Bellhop required significantly more time. Ultimately, geometric spreading and Energy flux models were found to be effective for estimating propagation loss across low-frequency bands in coastal environments. Geometric spreading required minimal computational effort while providing reasonable accuracy in an environment with low variation. However, in highly variable environments, Energy flux model was more accurate but needed substantial computational resources for large datasets. This study emphasises the need to carefully evaluate the complexity of the environment, the quality of the input data and the computational resources available to select the most suitable propagation model.