Measuring Underwater noise exposure from shipping

Levels of underwater noise in the open ocean have been increasing since at least the 1960s due to growth in global shipping traffic and the speed and propulsion power of vessels. This rise in noise levels reduces the range over which vocal marine species can communicate, and can induce physiological stress and behavioural responses, which may ultimately have population-level consequences. Although long-term noise trends have been studied at some open-ocean sites, in shallower coastal regions the high spatiotemporal variability of noise levels presents a substantial methodological challenge, and trends in these areas are poorly understood. This thesis addresses this challenge by introducing new techniques which combine multiple data sources for ship noise assessment in coastal waters. These data include Automatic Identification System (AIS) ship-tracking data, shore-based time-lapse footage, meteorological data, and tidal data. Two studies are presented: in the first, AIS data and acoustic recordings from Falmouth Bay in the western English Channel are combined using an adaptive threshold, which separates ship passages from background noise in the acoustic data. These passages are then cross-referenced with AIS vessel tracks, and the noise exposure associated with shipping activity is then determined. The second study, at a site in the Moray Firth, Scotland, expanded the method to include shore-based time-lapse footage, which enables visual corroboration of vessel identifications and the production of videos integrating the various data sources. Two further studies examine and enhance basic analysis techniques for ambient noise monitoring. The first study examines averaging metrics and their applicability to the assessment of noise from shipping. Long-term data from the VENUS observatory are empirically assessed for different averaging times and in the presence of outliers. It is concluded that the mean sound pressure level averaged in linear space is most appropriate, in terms of both standardization and relevance to impacts on marine fauna. In the second study, a new technique for the statistical analysis of long-term passive acoustic datasets, termed spectral probability density (SPD), is introduced. It is shown that the SPD can reveal characteristics such as multimodality, outlier influence, and persistent self-noise, which are not apparent using conventional techniques. This helps to interpret long-term datasets, and can indicate whether an instrument’s dynamic range is appropriate to field conditions. Taken together, the contributions presented in this thesis help to establish a stronger methodological basis for the assessment of shipping noise. These methods can help to inform emerging policy initiatives, efforts to standardise underwater noise measurements, and investigation into the effects of shipping noise on marine life.