PAM Results Field trials demonstrated that the THCs were reliable and capable of detecting harbour porpoise and bottlenose dolphin clicks. Location accuracy was investigated using trials with an artificial porpoise sound and using simulations. Trials also demonstrated that the spherical cluster design had better timing accuracy than the cylindrical design which is likely to be a result of a combination of the different shape of the cowling and also in the spacing of the hydrophones – the spherical clusters had a narrower hydrophone cluster spacing meaning that the signals were less distorted by echoes than the more widely spaced hydrophones in the cylindrical cluster. Changes in timing accuracy affect the accuracy at which sounds can be localised, but do not affect detection range. The simulations for a system consisting of three clusters in a triangular configuration around a turbine structure indicate a localisation accuracy of < 3 m; depth < 0.7 m and angle < 0.5 degrees at 25 m from the hydrophones. While timing accuracy of the VEMCO tag pulses is not as good as it is for porpoise clicks (+/- 7.5 µs), this has little impact on localisation accuracy at short ranges. The PAMGuard software was modified to allow detection of VEMCO acoustic tags. Work has also gone into further developing a data acquisition system in order to make it stable when sharing a network connection with other devices. Further work is required to increase the number of channels from 8 to 12. AAM Results This project has developed and tested a technique to track marine mammals in 3D in a tidally energetic environment using two multi-beam sonars. Two different configurations were tested for this and it was concluded that an overlapping parallel horizontal orientation provided the best results. By measuring the ratio of the sonar intensity of a target imaged simultaneously on two sonars arranged in this way, the depth of the animal was calculated. The error in depth estimated in this way is approximately 1.5 m (although this may be less when the sonars are mounted on a static platform). An efficient algorithm was developed to classify marine mammals in multi-beam sonar data, reducing the high false positive rate reported in previous studies. Cross-validation of the resulting algorithm estimated a cross validation error of 6%. All confirmed seals were correctly classified using the algorithm, while only 8% of non-seal targets were classified as seals. If this result holds with future datasets, the analytical approach will be an effective means of detecting and classifying harbour seals. At present, the effectiveness of these algorithms for classifying other species is unknown; however, it is anticipated that it is likely to be effective for similar sized marine mammals (e.g. grey seals, harbour porpoises, dolphins). The bottom mounted configuration, likely to be used in the turbine site deployment has also been successfully tested in a tidally energetic environment. This has demonstrated that tracking and detection algorithms can still detect marine mammals against a backdrop of additional background noise and surface clutter (in sea states up to Beaufort 2). However, it should be highlighted that the effects on detection and tracking capabilities of sea states above this are largely unknown at present.