Chapter 1.
1. Underwater passive acoustic monitoring (PAM) is an increasingly popular approach to monitor the health of aquatic environments through the analysis of soundscapes. Standard practices use hydrophones to record ambient sounds. They must either be cabled to surface recording devices or use autonomous instrumentation which comes at a premium cost. However, low-cost consumer-grade action cameras offer an accessible alternative also capable of autonomous underwater acoustic recordings.
2. The performance of two models of GoPro underwater action cameras when used as PAM recorders was evaluated. These were tested against a research-grade hydrophone in field conditions on shallow-water tropical coral reefs.
26
3. Simultaneous recordings of loudspeaker playbacks of known acoustic signals using all three instruments were taken first. Repeated deployments on different coral reef sites in which all three instruments were placed side by to side to record the same natural reef soundscape simultaneously were then undertaken. Eight of the most common eco-acoustic indices used in marine soundscape ecology from these GoPro recordings were calculated. These were used to assess the reliability and accuracy of results from the GoPros compared to the hydrophone.
4. Although not calibrated, GoPros appeared to provide recordings from which select eco-acoustic indices could be calculated reliably, including temporal variability, the acoustic complexity index and acoustic richness. Results from a GoPro can be compared against others of the same model but should not be used interchangeably with a hydrophone or those from another model. We outline the best settings that can be used to collect such soundscape data with GoPros.
5. Underwater action cameras are very popular with marine scientists and potential citizen scientists around the world. Their recordings represent a valuable tool for the global expansion of PAM techniques.

Chapter 2.
Widespread degradation of tropical coral reefs around the world has resulted in them becoming amongst the most threatened habitats globally. This has led to an increased demand for conservation and restoration of these habitats. Adequate monitoring of restored sites is essential to assess their success and identify further areas in need of attention. This investigation builds on previous research that used labour intensive manual listening to explore how PAM can be used to assess the progress of actively restored sites at one of the world’s largest tropical reef restoration projects, in South Sulawesi, Indonesia. The new work presented here applies modern computational approaches to recordings from the same sites to determine whether these could be used to more rapidly assess restoration using PAM data. A set of 12 eco-acoustic indices were calculated for up to three frequency bands; a low (50–800 Hz), medium (2–7 kHz) and full band (0.05–20 kHz), for a total of 33 index-frequency band combindations. Fifteen of these 33 combinations reported a significant difference between healthy and degraded habitats. However, high variability in the distribution of results was observed, offering a limited ability for any one index to discriminate between these two habitats without extensive sampling. This investigation therefore attempted to construct a machine learning model which could better discriminate between these two habitat classes using an optimised set of combined eco-acoustic indices. This used a supervised approach (regularised discriminant analysis) that was trained on labelled one minute recordings from both habitats and then tested blind. The pooled misclassification rate of 1000 cross-validated iterations of the model was 8.27% (± 0.84), demonstrating the first ever successful implementation of PAM and machine learning to determine tropical reef health from acoustic recordings. 1000 repeats of the model were then executed on a set of artificially restored reef recordings from three sites. This reported that a recently restored site (24 months previously that now exhibit an increased coral cover (A: 79.1% ± 3.9; B: 66.5% ± 3.8) received a majority classification of their recordings as healthy (A: 33/39; B: 37/38). Future work should validate this method by investigating trends observed when this tool is applied to additional restored sites. If this method continues to report promising results, this approach could offer a valuable tool that allows marine practitioners to assess habitats rapidly using short snapshot recordings, or effectively monitor habitat recovery over time, with a reduced reliance on frequent labour intensive in-water surveys.

Mesozooplankton (MSP) are the major group of secondary producers in the ocean. They play an important role in both the carbon flux and energy transfer from primary producers to higher trophic levels. The present work explores the significance of how MSP biomass relates to the energy transfer between primary and secondary trophic levels by applying the MSP biomass model developed and validated by Mahesh et al. (2018) in the Bay of Bengal (BoB). The remote sensing products, mean sea surface temperatures (SST) and chlorophyll-a (Chl-a), were acquired from the Aqua (EOS PM) satellite's Moderate Resolution Imaging Spectroradiometer (MODIS) sensor for the pre-monsoon period in 2013 through to summer 2015. These data were incorporated into the model to estimate MSP biomass which ranged from 0.12 to 33.52 mg C m−3 in the summer and post-monsoon periods in 2014, respectively, and was statistically significant between the seasons (F=19. 43; P=0. 008). The MSP productivity rate varied from 0.07 (summer 2014) to 15.35 mg C m−3 d−1(post-monsoon 2015). The rate was low (1 mg C m−3 d−1) across the central bay during the pre-monsoon period, whilst it was slightly higher (1-3 mg C m−3 d−1) along the neritic waters of the Palk Bay and Kakinada coastlines. The mean energy transfer efficiency from primary producers to MSP production was between 8.81 and 62.36 % during the post-monsoon and summer periods, respectively. This varied significantly between seasons (F=16.61; P=0.005) and was higher in the central bay than coastal waters. The complex interactions between water temperature, primary producer abundance, MSP grazing potential, seasonal variability, and the influential role of monsoons were the major controlling factors governing MSP biomass across the BoB.