Jacquomo Monk

Evaluation of Four Supervised Learning Methods for Benthic Habitat Mapping Using Backscatter from Multi-Beam Sonar

An understanding of the distribution and extent of marine habitats is essential for the implementation of ecosystem-based management strategies. Historically this had been difficult in marine environments until the advancement of acoustic sensors. This study demonstrates the applicability of supervised learning techniques for benthic habitat characterization using angular backscatter response data. With the advancement of multibeam echo-sounder (MBES) technology, full coverage datasets of physical structure over vast regions of the seafloor are now achievable. Supervised learning methods typically applied to terrestrial remote sensing provide a cost-effective approach for habitat characterization in marine systems. However the comparison of the relative performance of different classifiers using acoustic data is limited. Characterization of acoustic backscatter data from MBES using four different supervised learning methods to generate benthic habitat maps is presented. Maximum Likelihood Classifier (MLC), Quick, Unbiased, Efficient Statistical Tree (QUEST), Random Forest (RF) and Support Vector Machine (SVM) were evaluated to classify angular backscatter response into habitat classes using training data acquired from underwater video observations. Results for biota classifications indicated that SVM and RF produced the highest accuracies, followed by QUEST and MLC, respectively. The most important backscatter data were from the moderate incidence angles between 30° and 50°. This study presents initial results for understanding how acoustic backscatter from MBES can be optimized for the characterization of marine benthic biological habitats.

Are We Predicting the Actual or Apparent Distribution of Temperate Marine Fishes

Planning for resilience is the focus of many marine conservation programs and initiatives. These efforts aim to inform conservation strategies for marine regions to ensure they have inbuilt capacity to retain biological diversity and ecological function in the face of global environmental change – particularly changes in climate and resource exploitation. In the absence of direct biological and ecological information for many marine species, scientists are increasingly using spatially-explicit, predictive-modeling approaches. Through the improved access to multibeam sonar and underwater video technology these models provide spatial predictions of the most suitable regions for an organism at resolutions previously not possible. However, sensible-looking, well-performing models can provide very different predictions of distribution depending on which occurrence dataset is used. To examine this, we construct species distribution models for nine temperate marine sedentary fishes for a 25.7 km2 study region off the coast of southeastern Australia. We use generalized linear model (GLM), generalized additive model (GAM) and maximum entropy (MAXENT) to build models based on co-located occurrence datasets derived from two underwater video methods (i.e. baited and towed video) and fine-scale multibeam sonar based seafloor habitat variables. Overall, this study found that the choice of modeling approach did not considerably influence the prediction of distributions based on the same occurrence dataset. However, greater dissimilarity between model predictions was observed across the nine fish taxa when the two occurrence datasets were compared (relative to models based on the same dataset). Based on these results it is difficult to draw any general trends in regards to which video method provides more reliable occurrence datasets. Nonetheless, we suggest predictions reflecting the species apparent distribution (i.e. a combination of species distribution and the probability of detecting it). Consequently, we also encourage researchers and marine managers to carefully interpret model predictions.

Habitat classification of temperate marine macroalgal communities using bathymetric LiDAR

Here, we evaluated the potential of using bathymetric Light Detection and Ranging (LiDAR) to characterise shallow water ( 70%), with varying results for the classification of individual habitat classes; for instance, producer accuracy for mixed brown algae and sediment substrata, was 74% and 93%, respectively. LiDAR was also successful for differentiating canopy structure of macroalgae communities (i.e., canopy structure classification), such as canopy forming kelp versus erect fine branching algae. In conclusion, habitat characterisation using bathymetric LiDAR provides a unique potential to collect baseline information about biological assemblages and, hence, potential reef connectivity over large areas beyond the range of direct observation. This research contributes a new perspective for assessing the structure of subtidal coastal ecosystems, providing a novel tool for the research and management of such highly dynamic marine environments.

Using community-based monitoring with GIS to create habitat maps for a marine protected area in Australia

In recent years there has been an increase in community-based monitoring programmes developed and implemented worldwide. This paper describes how the data collected from such a programme could be integrated into a Geographic Information System (GIS) to create temperate subtidal marine habitat maps. A differential Global Positioning System was utilized to accurately record the location of the trained community-based SCUBA diver data. These georeferenced data sets were then used to classify benthic habitats using an aerial photograph and digitizing techniques. This study demonstrated that trained community-based volunteers can collect data that can be utilized within a GIS to create reliable and cost-effective maps of shallow temperate subtidal rocky reef systems.

Exploring Spatiotemporal Trends in Commercial Fishing Effort of an Abalone Fishing Zone: A GIS-Based Hotspot Model

Assessing patterns of fisheries activity at a scale related to resource exploitation has received particular attention in recent times. However, acquiring data about the distribution and spatiotemporal allocation of catch and fishing effort in small scale benthic fisheries remains challenging. Here, we used GIS-based spatio-statistical models to investigate the footprint of commercial diving events on blacklip abalone (Haliotis rubra) stocks along the south-west coast of Victoria, Australia from 2008 to 2011. Using abalone catch data matched with GPS location we found catch per unit of fishing effort (CPUE) was not uniformly spatially and temporally distributed across the study area. Spatial autocorrelation and hotspot analysis revealed significant spatiotemporal clusters of CPUE (with distance thresholds of 100’s of meters) among years, indicating the presence of CPUE hotspots focused on specific reefs. Cumulative hotspot maps indicated that certain reef complexes were consistently targeted across years but with varying intensity, however often a relatively small proportion of the full reef extent was targeted. Integrating CPUE with remotely-sensed light detection and ranging (LiDAR) derived bathymetry data using generalized additive mixed model corroborated that fishing pressure primarily coincided with shallow, rugose and complex components of reef structures. This study demonstrates that a geospatial approach is efficient in detecting patterns and trends in commercial fishing effort and its association with seafloor characteristics.

Quantification of Spatial and Thematic Uncertainty in the Application of Underwater Video for Benthic Habitat Mapping

This study presents an analysis of the application of underwater video data collected for training and validating benthic habitat distribution models. Specifically, we quantify the two major sources of error pertaining to collection of this type of reference data. A theoretical spatial error budget is developed for a positioning system used to co-register video frames to their corresponding locations at the seafloor. Second, we compare interpretation variability among trained operators assessing the same video frames between times over three hierarchical levels of a benthic classification scheme. Propagated error in the positioning system described was found to be highly correlated with depth of operation and varies from 1.5m near the surface to 5.7m in 100m of water. In order of decreasing classification hierarchy, mean overall observer agreement was found to be 98% (range 6%), 82% (range 12%) and 75% (range 17%) for the 2, 4, and 6 class levels of the scheme, respectively. Patterns in between-observer variation are related to the level of detail imposed by each hierarchical level of the classification scheme, the feature of interest, and to the amount of observer experience.

Use of anthropogenic sea floor structures by Australian fur seals: potential positive ecological impacts of marine industrial development?

Human-induced changes to habitats can have deleterious effects on many species that occupy them. However, some species can adapt and even benefit from such modifications. Artificial reefs have long been used to provide habitat for invertebrate communities and promote local fish populations. With the increasing demand for energy resources within ocean systems, there has been an expansion of infrastructure in near-shore benthic environments which function as de facto artificial reefs. Little is known of their use by marine mammals. In this study, the influence of anthropogenic sea floor structures (pipelines, cable routes, wells and shipwrecks) on the foraging locations of 36 adult female Australian fur seals (Arctocephalus pusillus doriferus) was investigated. For 9 (25%) of the individuals, distance to anthropogenic sea floor structures was the most important factor in determining the location of intensive foraging activity. Whereas the influence of anthropogenic sea floor structures on foraging locations was not related to age and mass, it was positively related to flipper length/standard length (a factor which can affect manoeuvrability). A total of 26 (72%) individuals tracked with GPS were recorded spending time in the vicinity of structures (from <1% to >75% of the foraging trip duration) with pipelines and cable routes being the most frequented. No relationships were found between the amount of time spent frequenting anthropogenic structures and individual characteristics. More than a third (35%) of animals foraging near anthropogenic sea floor structures visited more than one type of structure. These results further highlight potentially beneficial ecological outcomes of marine industrial development.

Aspects of the biology of Galaxias maculatus.

The biology of three landlocked and a riverine population of Galaxias maculatus were examined in western Victoria, Australia. All systems supported reproducing populations of these fish, including Lake Corangamite which had salinities that on occasion reached 82. Spawning sites in Lake Corangamite were located in adjacent tributaries and not in the main lake as was the case for other populations. The smallest fish were found in the fresh water Lake Purrumbete and the largest in the hypersaline Lake Corangamite. The size at which 50% of the population attained sexual maturity varied across sites, with fish maturing at a smaller size in Lake Purrumbete, followed by the Merri River, Lake Bullen Merri and Lake Corangamite. Condition was higher in the freshwater Lake Purrumbete and there was no relationship between condition and temperature, conductivity, turbidity and pH; but there was a positive relationship between condition and dissolved oxygen. Length frequency analysis suggested that the majority of fishes live for a year.

Combining pixel and object based image analysis of ultra-high resolution multibeam bathymetry and backscatter for habitat mapping in shallow marine waters

Habitat mapping data are increasingly being recognised for their importance in underpinning marine spatial planning. The ability to collect ultra-high resolution (cm) multibeam echosounder (MBES) data in shallow waters has facilitated understanding of the fine-scale distribution of benthic habitats in these areas that are often prone to human disturbance. Developing quantitative and objective approaches to integrate MBES data with ground observations for predictive modelling is essential for ensuring repeatability and providing confidence measures for habitat mapping products. Whilst supervised classification approaches are becoming more common, users are often faced with a decision whether to implement a pixel based (PB) or an object based (OB) image analysis approach, with often limited understanding of the potential influence of that decision on final map products and relative importance of data inputs to patterns observed. In this study, we apply an ensemble learning approach capable of integrating PB and OB Image Analysis from ultra-high resolution MBES bathymetry and backscatter data for mapping benthic habitats in Refuge Cove, a temperate coastal embayment in south-east Australia. We demonstrate the relative importance of PB and OB seafloor derivatives for the five broad benthic habitats that dominate the site. We found that OB and PB approaches performed well with differences in classification accuracy but not discernible statistically. However, a model incorporating elements of both approaches proved to be significantly more accurate than OB or PB methods alone and demonstrate the benefits of using MBES bathymetry and backscatter combined for class discrimination.

An evaluation of the error and uncertainty in epibenthos cover estimates from AUV images collected with an efficient, spatially-balanced design

Efficient monitoring of organisms is at the foundation of protected area and biodiversity management. Such monitoring programs are based on a systematically selected set of survey locations that, while able to track trends at those locations through time, lack inference for the overall region being “monitored”. Advances in spatially-balanced sampling approaches offer alternatives but remain largely untested in marine ecosystems. This study evaluated the merit of using a two-stage, spatially-balanced survey framework, in conjunction with generalized additive models, to estimate epifauna cover at a reef-wide scale for mesophotic reefs within a large, cross-shelf marine park. Imagery acquired by an autonomous underwater vehicle was classified using a hierarchical scheme developed under the Collaborative and Automated Tools for Analysis of Marine Imagery (CATAMI). At a realistic image subsampling intensity, the two-stage, spatially-balanced framework provided accurate and precise estimates of reef-wide cover for a select number of epifaunal classes at the coarsest CATAMI levels, in particular bryozoan and porifera classes. However, at finer hierarchical levels, accuracy and/or precision of cover estimates declined, primarily because of the natural rarity of even the most common of these classes/morphospecies. Ranked predictor importance suggested that bathymetry, backscatter and derivative terrain variables calculated at their smallest analysis window scales (i.e. 81 m2) were generally the most important variables in the modeling of reef-wide cover. This study makes an important step in identifying the constraints and limitations that can be identified through a robust statistical approach to design and analysis. The two-stage, spatially-balanced framework has great potential for effective quantification of epifaunal cover in cross-shelf mesophotic reefs. However, greater image subsampling intensity than traditionally applied is required to ensure adequate observations for finer-level CATAMI classes and associated morphospecies.