Image of a stony coral structures providing habitat for smaller animals including crustaceans, brittle stars, urchins and molluscs

Stony coral structures such as this provide habitat for smaller animals including crustaceans, brittle stars, urchins and molluscs

Using statistics to predict seabed fauna

Environmental statisticians are working to predict the type and distribution of marine seabed fauna on Australia's continental slope.

  • 8 May 2008 | Updated 14 October 2011

Project summary

CSIRO statisticians and marine researchers in the Wealth from Oceans Flagship are working to understand marine ecological processes in a critical range of Australia's continental slope, the depth range of 150 metres to 1 000 metres.

The continental slope is the area of the seabed that occurs after the edge of the continental shelf. It is a steep zone that meets the ocean floor at about 4 000 metres.

This seabed monitoring project will provide an essential tool to enable decision-making about the management of Australia's marine economic zone.

Current activities

This seabed monitoring project uses data obtained from two new sources for this environment:

This model provides a method of predicting (with a measure of uncertainty) seabed fauna, geomorphology and substrate.
  • specially-equipped ships able to bounce multiple sonar beams off the seafloor, which are used to create detailed maps of depth, gradient and seabed roughness ('swath' maps)
  • a video camera, lowered to within a metre or two of the seafloor, that is towed along several kilometres. Marine biologists then score the video for three key traits: substrate type (sandy or rocky, for example), geomorphology (if the rocks are flat or rippled, for example) and fauna type (sponge garden, corals or worms, for example).

The statistical challenge emerges when trying to combine these different data into a model that enables researchers to predict the distribution of fauna over large areas.

The collected video data exhibit a high level of dependence between observations in successive frames. Also, there is dependence amongst the three measurements taken within each video frame.

These dependencies are modelled and the level of dependence is allowed to vary depending on the attributes of the local environment (measured by the swath maps).

This model provides a method of predicting (with a measure of uncertainty) seabed fauna, geomorphology and substrate.

Outcomes

The resulting statistical framework provides researchers and natural resource managers a critical tool to enable prediction of the fauna composition of large areas of the continental slopes.

This information is crucial for assessing proposed exclusion zones, such as marine parks. 

Importantly, the predictions include a measure of uncertainty that allows researchers and resource managers to make more informed decisions.

Find out more about Marine conservation and biodiversity management.