A new machine learning approach to seabed biotope classification

Effective management in the marine environment requires a thorough understanding of the distribution of natural resources, including that of the benthos, the animals living in and on the seabed. Hitherto, it has been difficult to identify broadscale patterns in the benthos as the faunal clusters identified from individual surveys are not directly comparable. As a result, much reliance has been placed on one-off broadscale spatial surveys or matching samples to a common set of biotopes. In this study, new benthic macrofaunal data from discrete surveys are matched to existing broadscale cluster groups identified using unsupervised machine learning (k-means). This objective approach allows for continual improvements in our understanding of macrofaunal distribution patterns, thereby supporting ongoing conservation and marine spatial planning efforts. Other benefits are discussed. Finally, an R shiny web application is presented, allowing users to biotope match their own data.