James J. Daniell

Application of machine learning methods to spatial interpolation of environmental variables

Machine learning methods, like random forest (RF), have shown their superior performance in various disciplines, but have not been previously applied to the spatial interpolation of environmental variables. In this study, we compared the performance of 23 methods, including RF, support vector machine (SVM), ordinary kriging (OK), inverse distance squared (IDS), and their combinations (i.e., RFOK, RFIDS, SVMOK and SVMIDS), using mud content samples in the southwest Australian margin. We also tested the sensitivity of the combined methods to input variables and the accuracy of averaging predictions of the most accurate methods. The accuracy of the methods was assessed using a 10-fold cross-validation. The spatial patterns of the predictions of the most accurate methods were also visually examined for their validity. This study confirmed the effectiveness of RF, in particular its combination with OK or IDS, and also confirmed the sensitivity of RF and its combined methods to the input variables. Averaging the predictions of the most accurate methods showed no significant improvement in the predictive accuracy. Visual examination proved to be an essential step in assessing the spatial predictions. This study has opened an alternative source of methods for spatial interpolation of environmental properties.

Geology–benthos relationships on a temperate rocky bank, eastern Bass Strait, Australia

To better understand the possible relationships between the geology of the seabed and the associated biological communities, a multibeam sonar survey over New Zealand Star Bank in the eastern Bass Strait was conducted. A hierarchical method of benthic habitat mapping was applied to the secondary biotope and biological facies levels at the site (<10 km) scale. Four secondary biotopes and four biological facies have been defined on the basis of geomorphology revealed by the bathymetry model and the results of statistical analysis of the sediment and underwater video transect data over the bank. The major differences that control the distribution of biological communities in the New Zealand Star Bank area appear to be related to variations in substrate. (1) Hard-ground features related to high-relief granite outcrops are associated with diverse and abundant sessile and motile fauna. These faunal communities may be biologically modified to patchy barrens habitat by grazing urchins. (2) Unconsolidated sediment on a flat seabed is associated with sparse small sponges on the inner shelf. On the middle shelf and seaward of bank, the flat and muddy seabed supports a community dominated by infauna. (3) Unconsolidated sediment on a low-relief seabed is associated with an increase in the density and sizes of sponges concentrated on any low-relief feature raised above the surrounding flat seabed.

Neogene evolution of the mixed carbonate-siliciclastic system in the Gulf of Papua, Papua New Guinea

This paper outlines the evolution of the late Cenozoic mixed carbonate-siliciclastic depositional system in the Gulf of Papua (GoP), using seismic, gravity, multibeam bathymetry, well data sets, and Landsat imagery. The deposition of the mixed sedimentary sequences was influenced by dynamic interplay of tectonics, eustasy, in situ carbonate production, and siliciclastic sediment supply. The roles of these major factors are estimated during different periods of the GoP margin evolution. The Cenozoic mixed system in the GoP formed in distinct phases. The first phase ( Late Cretaceous-Paleocene) was mostly driven by tectonics. Rifting created grabens and uplifted structural blocks which served later as pedestals for carbonate edifices. Active neritic carbonate accumulation characterized the second phase (Eocene-middle Miocene). During this phase, mostly eustatic fluctuations controlled the large-scale sedimentary geometries of the carbonate system. The third phase ( late Miocene-early Pliocene) was characterized by extensive demise of the carbonate platforms in the central part of the study area, which can be triggered by one or combination of several factors, such as eustatic sea level fluctuations, increased tectonic subsidence, uplift, sudden influx of siliciclastics, or dramatic changes in environmental conditions and climate. The fourth phase ( late Pliocene-Holocene) was dominated by siliciclastics, which resulted in the burial of drowned and/or active carbonate platforms, although some platforms still remain alive until present-day.

Predictive modelling of seabed sediment parameters using multibeam acoustic data: a case study on the Carnarvon Shelf, Western Australia

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.

Deep water geomorphology of the mixed siliciclastic‐carbonate system, Gulf of Papua

The Gulf of Papua (GoP) has become a focal point for understanding the deposition and accumulation of siliciclastic and carbonate material along and across a low-latitude continental margin. Although studies have addressed submarine geomorphological features on the inner and middle shelves, as well as processes that may have led to their formation, the seafloor of adjacent slope regions remains poorly documented. This study presents and interprets results from approximately 13,000 line-km of multibeam bathymetry, 9500 line-km of 3.5 kHz seismic, and 122 sediment cores that were collected from the GoP shelf edge and slope, primarily on two cruises (PANASH and PECTEN). Bathymetric maps, in conjunction with the seismic profiles and cores, were used to make extensive observations, descriptions, and interpretations of seafloor geomorphology and begin to address several key issues regarding the delivery and accumulation of sediment. This study divided the GoP slope region into physiographic regions including intraslope basins: Ashmore Trough, southern Pandora Trough, northern Pandora Trough, Moresby Trough and intraslope plateaus/platforms: carbonate platforms and atolls and Eastern Plateau. Ashmore Trough contains a very linear northern margin capped by a drowned barrier reef system. This shelf edge is also defined by a broad promontory with channels extending from its apex, interpreted as a relict shelf-edge delta. Southern Pandora Trough is characterized by pervasive slope channels and slump scars extending down slope to a thick depocenter and an extensive mass-transport complex. In contrast, northern Pandora Trough has few visible slope channels. Seismic observations reveal a wedge of sediment extending down slope from northern Pandora Trough shelf edge and filling preexisting bathymetry. Large fold-and-thrust-belt ridges are also present on the seafloor in this region and may act to divert and/or catch sediment, depending on sediment transport direction. Moresby Trough contains a large axial submarine channel that extends almost the entire length of the intraslope basin. In addition, an extensive system of canyons lines the NE margin of Moresby Trough. Mass-transport deposits have been fed from the canyons and in one case deposited a large (similar to 2000 km(2)) mass-transport complex. Fold-and-thrust-belt ridges also extend into Moresby Trough. Here they trend perpendicular to slope and catch gravity flow deposits on their updip side. GoP carbonate platforms/atolls all display very pronounced scalloped-margin morphology, which may indicate pervasive mass-wasting processes on carbonate margins. Northwest Eastern Plateau is dominantly carbonate and displays the characteristic scalloped margin morphology; however, most of the plateau is characterized by parallel seismic reflectors. These seismic observations in conjunction with core data indicate that accumulation on Eastern Plateau is primarily mixed pelagic and hemipelagic sediment. Observations and interpretations of the bathymetry have revealed the deep water GoP to contain very diverse geomorphology and suggest it is a dynamic system influenced by a variety of sediment transport processes, particularly mass wasting and other gravity flow processes.

Evidence of extensive reef development and high coral cover in nearshore environments: implications for understanding coral adaptation in turbid settings

Mean coral cover has reportedly declined by over 15% during the last 30 years across the central Great Barrier Reef (GBR). Here, we present new data that documents widespread reef development within the more poorly studied turbid nearshore areas (<10 m depth), and show that coral cover on these reefs averages 38% (twice that reported on mid- and outer-shelf reefs). Of the surveyed seafloor area, 11% had distinct reef or coral community cover. Although the survey area represents a small subset of the nearshore zone (15.5 km2), this reef density is comparable to that measured across the wider GBR shelf (9%). We also show that cross-shelf coral cover declines with distance from the coast (R2 = 0.596). Identified coral taxa (21 genera) exhibited clear depth-stratification, corresponding closely to light attenuation and seafloor topography, with reefal development restricted to submarine antecedent bedforms. Data from this first assessment of nearshore reef occurrence and ecology measured across meaningful spatial scales suggests that these coral communities may exhibit an unexpected capacity to tolerate documented declines in water quality. Indeed, these shallow-water nearshore reefs may share many characteristics with their deep-water (>30 m) mesophotic equivalents and may have similar potential as refugia from large-scale disturbances.

Insights into environmental drivers of acoustic angular response using a self-organising map and hierarchical clustering

Acoustic backscatter from the seafloor is a complex function of signal frequency, seabed roughness, grain size distribution, benthos, bioturbation, volume reverberation, and other factors. Angular response is the variation in acoustic backscatter with incident angle and is considered be an intrinsic property of the seabed. An unsupervised classification technique combining a self-organising map (SOM) and hierarchical clustering was used to create an angular response facies map and explore the relationships between acoustic facies and ground truth data. Cluster validation routines indicated that a two cluster solution was optimal and separated sediment dominated environments from mixtures of sediment and hard ground. Low cluster separation limited cluster validation routines from identifying fine cluster structure visible with an AR density plot. Cluster validation, aided by a visual comparison with an AR density plot, indicated that a 14 cluster solution was also a suitable representation of the input dataset. Clusters that were a mixture of hard and unconsolidated substrates displayed an increase in backscatter with an increase in the occurrence of hard ground and highlighted the sensitivity of AR curves to the presence of even modest amounts of hard ground. Remapping video observations and sediment data onto the SOM matrix is innovative and depicts the relationship between ground truth data and cluster structure. Mapping environmental variables onto the SOM matrix can show broad trends and localised peaks and troughs and display the variability of ground truth data within designated clusters. These variables, when linked to AR curves via clusters, can indicate how environmental factors influence the shape of the curves. Once these links are established they can be incorporated into improved geoacoustic models that replicate field observations.