Tuesday, 29 July 2014 Back to News

Progress in acoustic seabed classification

Cefas and BGS scientists publish comparative study on mapping methods in peer-review journal

Marine conservation and spatial planning require sufficiently detailed maps of the substrates and habitats that make up the seabed, but this information is often missing.

The advent of multibeam echosounder technology has enabled scientists to chart the seabed in high detail. However, such systems do not directly map the parameters necessary to create maps on seabed substrate and habitat.

Seabed substrate and habitat maps are created by a spatial integration of multibeam echosounder and sample data.
Therefore, seabed sampling and observation, which give direct insights on the geology and biology of the seabed, are essential to create seabed maps.

Maps portraying the substrates and habitats of the seabed are then created by a spatial integration of direct observations and multibeam echosounder data.

How such an integration (see Figure) is best achieved remains a hot topic of current seabed mapping research.

MAREMAP therefore hosted a two-day workshop on seabed mapping in October 2012.

The results of this workshop have now been published in the scientific journal Continental Shelf Research.

Cefas Scientist, Dr Markus Diesing said: "This study is a step forward towards validated, repeatable and quantitative seabed mapping, but more research needs to be conducted to increase classification accuracy."

Scientists from Cefas and BGS investigated four approaches in their ability to map seabed substrates. They assessed the accuracy of the different classification results and the similarity of map outputs.

They found differences in classification accuracy, although these were not statistically significant. Pairwise agreement of the resultant maps ranged from 68% to 87%.

The authors concluded that more research needs to be conducted and proposed strategies to increase classification accuracy.

This study is freely accessible until 28th August 2014 by following the link: http://authors.elsevier.com/a/1PKkQ-JmAOHpT
Further questions can be addressed to: markus.diesing@cefas.co.uk

Full reference of the paper:
Diesing, M., Green, S.L., Stephens, D., Lark, R.M., Stewart, H., Dove, D., 2014. Mapping seabed sediments: Comparison of manual, geostatistical, object-based image analysis and machine learning approaches. Continental Shelf Research, 84, 107-119. DOI: 10.1016/j.csr.2014.05.004