Regularized log-linear models for predicting underwater radiated noise spectra of small vessels

Underwater radiated noise (URN) from vessels is an anthropogenic stressor that negatively affects marine ecosystems. Accordingly, the ability to estimate URN levels is a critical component for environmental assessment. Predictive models are widely used to estimate URN emissions from different vessel types and operating conditions; however, comparable models for small recreational vessels are largely lacking, even though these vessels are arguably a dominant source of noise pollution in coastal areas. Here, we use a large dataset of acoustic measurements of recreational vessels to extend the empirical framework to small motorized vessels (motorized yachts and sailboats) by constructing regularized log-linear regression models that provide length and speed dependent source spectra over 23–24,000 Hz.