Basin-wide Atlantic marine landscape map

Classifications of the marine environment give a comprehensive and unpretentious overview into regions of similar characteristics and can hence be a stepping stone for sustainable ocean resource handling and protection plans. There have been many efforts to categorise the marine realm into seascapes or hydro-morphologic provinces, using different approaches, applied at a wide range of scales. Some of those categorisations available are based on hierarchical classification schemes with often arbitrary thresholds or use simple algorithms which do not fully account for the high complexity of the data. This study presents a basin-wide classification of the Atlantic seafloor environment, based on nine global datasets: bathymetry, slope, terrain ruggedness index, topographic position index, sediment thickness, POC flux, salinity, dissolved oxygen, temperature, current velocity and phytoplankton. To reduce subjectivity within the analysis, an unsupervised classification was performed on the normalised data using Gaussian finite mixture models. Those models describe a latent distribution structure of the input data set from which the final clusters, here seabed areas (SBAs), are derived. This model-based clustering approach seeks to overcome the shortcomings of other classification techniques by trying to embrace the challenging complexity of the ocean floor environment.

The result is a map of the Atlantic realm subdivided into nine SBAs. Some are clearly defined by geological and geomorphological properties, while others are dominated by hydrographic properties, or by a mixture of both sea floor terrain and water column characteristics. Larger SBAs cover the deep abyssal plain with low hydrographic and seasonal variation – in contrast to smaller SBAs including coastal waters that are subject to high seasonal variability. There are also differences in geographical distributions. The SBAs we found were further compared to other existing classifications (e.g., Global Ocean Seascapes, GOODS, EMU) and supplementary data (e.g., seamount locations) to assess in how far the objectively identified SBAs are represented in former classifications and studies.