Predicting the exposure of diving grey seals to shipping noise.

There is high spatial overlap between grey seals and shipping traffic, and the functional hearing range of grey seals indicates sensitivity to underwater noise emitted by ships. However, there is still very little data regarding the exposure of grey seals to shipping noise, constraining effective policy decisions. Particularly, there are few predictions that consider the at-sea movement of seals. Consequently, this study aimed to predict the exposure of adult grey seals and pups to shipping noise along a three-dimensional movement track, and assess the influence of shipping characteristics on sound exposure levels. Using ship location data, a ship source model, and the acoustic propagation model, RAMSurf, this study estimated weighted 24-h sound exposure levels (10-1000 Hz) (SELw). Median predicted 24-h SELw was 128 and 142 dB re 1 µPa2s for the pups and adults, respectively. The predicted exposure of seals to shipping noise did not exceed best evidence thresholds for temporary threshold shift. Exposure was mediated by the number of ships, ship source level, the distance between seals and ships, and the at-sea behaviour of the seals. The results can inform regulatory planning related to anthropogenic pressures on seal populations.

An adaptive grid to improve the efficiency and accuracy of modelling underwater noise from shipping.

Underwater noise pollution from shipping is a significant ecological concern. Acoustic propagation models are essential to predict noise levels and inform management activities to safeguard ecosystems. However, these models can be computationally expensive to execute. To increase computational efficiency, ships are spatially partitioned using grids but the cell size is often arbitrary. This work presents an adaptive grid where cell size varies with distance from the receiver to increase computational efficiency and accuracy. For a case study in the Celtic Sea, the adaptive grid represented a 2 to 5 fold increase in computational efficiency in August and December respectively, compared to a high resolution 1 km grid. A 5 km grid increased computational efficiency 5 fold again. However, over the first 25 km, the 5 km grid produced errors up to 13.8 dB compared to the 1 km grid, whereas, the adaptive grid generated errors of less than 0.5 dB.