Seabed sediment textural parameters such as mud, sand and gravel content can be useful surrogates for predicting patterns of benthic biodiversity. Multibeam sonar mapping can provide near-complete spatial coverage of high-resolution bathymetry and backscatter data that are useful in predicting sediment parameters. Multibeam acoustic data collected across a ∼1000 km² area of the Carnarvon Shelf, Western Australia, were used in a predictive modelling approach to map eight seabed sediment parameters. Four machine learning models were used for the predictive modelling: boosted decision tree, random forest decision tree, support vector machine and generalised regression neural network. The results indicate overall satisfactory statistical performance, especially for %Mud, %Sand, Sorting, Skewness and Mean Grain Size. The study also demonstrates that predictive modelling using the combination of machine learning models has provided the ability to generate prediction uncertainty maps. However, the single models were shown to have overall better prediction performance than the combined models. Another important finding was that choosing an appropriate set of explanatory variables, through a manual feature selection process, was a critical step for optimising model performance. In addition, machine learning models were able to identify important explanatory variables, which are useful in identifying underlying environmental processes and checking predictions against the existing knowledge of the study area. The sediment prediction maps obtained in this study provide reliable coverage of key physical variables that will be incorporated into the analysis of covariance of physical and biological data for this area.

The Great Barrier Reef (GBR) World Heritage Area and adjacent Coral Sea Marine Park are under serious threat from global climate change. This study used Daily Optimally Interpolated Sea Surface Temperature (DOISST) data to identify major marine heatwaves (MHWs) that have occurred in this region over the last three decades (1992-2022). We then used Himawari-8 (H-8) SST data to map significant MHW events that occurred between 2015 and 2022. We investigated the mechanisms underlying the MHWs, assessed thier impact on shallow and mesophotic coral reef ecosystems and identified potential coral refugia. MHWs in this region have increased in frequency, intensity and spatial extent. El Niño, especially when it is in phase with positive Indian Ocean Dipole, was the key remote driver leading to intense MHWs. However, the more recent strong MHWs (e.g., 2017 and 2022) occurred in the abscence of these climatic events, signifying the impacts of long-term climate change and local drivers. We also found that reduced wind speed and shoaling mixed layer depth, often together with reduced cloudiness, were the main local drivers pre-conditioning these MHWs. Anomalous air-sea heat flux into the ocean, mainly controlled by shortwave solar radiation (cloudiness) and latent heat flux (wind), was the most constant contributor to the 2015-16 and 2019-20 MHW events. However, local oceanographic dynamics, especially horizontal advection and turbulent mixing, played important roles in MHW heat budgets. This study confirmed that shallow-water coral bleaching severity was positively related to the cumulative MHW intensity in the 2015-16 and 2019-20 MHW events. We identified shallow reefs along the path of the North Queensland Current as potential coral refugia from bleaching because of the cooler waters upwelled from the ocean current. We also found that, except during weather events such as tropical cyclones, mesophotic reefs in the Coral Sea Marine Park may be less susceptible to severe bleaching as the MHWs were more confined within the shallow mixed layer.

Sediment properties are known to influence acoustic backscatter intensity. This sediment‐acoustic relationship has been investigated previously through using physical geoacoustic models and empirical methods and found to be complex and non‐linear. Here we employ a robust machine‐learning statistical model (random forest decision tree) to investigate the most likely non‐linear sediment‐backscatter relationships. The analysis uses co‐located sediment and acoustic backscatter data (collected from a 300 kHz multibeam sonar system) for 564 locations in four different areas on the Australian margin. Seven sediment grain size properties (%gravel, %sand, %mud, mean grain size, sorting, skewness and kurtosis) were used to predict the acoustic backscatter responses at individual incidence angles. The modelling results demonstrate the effectiveness of this multivariate predictive approach for the investigation of sediment‐acoustic relationship. Thus, we find that for incidence angles between 1^o and 41^o, the sediment variables explain around 70% of variance in the backscatter intensity. Sediment mud content was found to be the most important sediment variable in the model, and has a significant negative relationship with backscatter intensity. Mean grain size was the second ranked sediment variable and found to have a positive relationship with backscatter intensity. The results also show that sediment mud content plays a key role in sorting‐backscatter and sand‐backscatter relationships. Using only two sediment properties, mud content and mean grain size, together it was possible to largely explain the sediment‐acoustic relationship. The strongest backscatter return occurred with medium sediment mud content and large mean grain sizes (or muddy coarse sand).

A key requirement for informed marine-zone management is an understanding of the spatial patterns of marine biodiversity, often measured as species richness, total abundance or presence of key taxa. In the present study, we focussed on the diversity of benthic infauna and applied a predictive modelling approach to map biodiversity patterns for three study sites on the tropical Carnarvon shelf of Western Australia. A random forest decision tree model was used to generate spatial predictions of two measures of infaunal diversity, namely, species richness and total abundance. Results explained between 20% and 37% of the variance of each measure. The modelling process also identified potential physical surrogates for species richness and abundance, with sediment physical properties ranked as most important across the study region. Specifically, coarse-grained heterogeneous sediments were associated with higher infaunal species richness and total abundance. Seabed topographic properties were also important at the local scale. The study demonstrated the value of a surrogacy approach to the prediction of biodiversity patterns, particularly when the number of biological samples was limited. Such an approach may facilitate an understanding of ecosystem processes in the region and contribute to integrated marine management.