The seabed in the Tromsøflaket-Eggakanten area has diverse bottom types and topography created through by a variety of processes including glaciers and submarine landslides. This geological setting, together with the large range in water depth and associated currents gives rise to a range of environmental conditions, within which the seabed dwelling animals find the most favourable area to live. On the shallow part of the continental shelf we have identified three biotopes. The most characteristic biotope here (red) is dominated by sponges (mostly various species of Geodia) and other large sponges which are often associated with sandy muddy sediments. The two other groups seem to be discriminated by the local topography. One (beige) hosts fish species which typically prefer a seabed with more structure, e.g. boulders, where they can form territory. In the remaining shallow-shelf biotope (brown) we find species that generally prefer large areas of uniform seabed. As we move towards the outer continental shelf and upper continental slope we find two further biotopes, and another two biotopes extend deeper on the slope. There are considerable differences in the species composition between these biotopes.
The animals living on the seabed are not evenly distributed but occur in patches related to various environmental factors including bottom type. It is important to understand the patterns of faunal distribution in order to manage the seabed in the most effective manner. For example, where area the vulnerable or species rich habitats that could be damaged by drilling for oil and gas, and which areas are most vulnerable to the impact of trawling? In order to deliver the required information for marine spatial planning and sustainable management it is not enough to provide simple observations from MAREANO video and sampling; we need a full coverage map of the distribution of biotopes. This is where biotope modelling becomes an essential tool. Biotope modelling is the most cost-effective method by which a full coverage map of biotope distribution can be made. MAREANO has shown how data from multibeam echosounder can be used to calculate various terrain variables (e.g. slope) which together give a reasonable indication the environmental variation relevant to the bottom fauna. Together with interpreted maps of geology and geomorphology these full coverage maps can be used to predict the distribution of biotopes to make a full coverage map, using the observation data to train the model.
A summary of the modelling process adopted for the new Tromsøflaket-Eggakanten biotope map is shown below.
Map showing the distribution of dominant biotopes across Tromsøflaket-Eggakanten.
The first stage in biotope mapping is for biologists to review all CAMPOD video data and record all the visible fauna down to the lowest taxonomic level possible. The species observations are then analysed using multivariate statistics (Detrented Correspondence Analysis – DCA) to reveal natural groupings which provide the information for the next stage, biotope classification. Identification of the most appropriate number and best grouping of biotopes can be challenging, and in order to make this part of the work more objective we have introduced the use of statistical clustering techniques (k-means) in analysing data from this area. Classified point observations (representing approximately 200 m of pooled video observations) can be seen in the figure below while the physical characteristics of each biotope are listed in the table, together with typical fauna.
Classified biotopes observed in video shown on shaded relief image (50 m resolution), with inset map zoomed in to see detail of video transect lines.
In order to move from classified points (Figure 2) to a full coverage map (Figure 1) we require some kind of classification or modelling technique making use of available full coverage datasets to provide predictor variables. Multibeam data and ecologically relevant terrain derivatives such as slope provide a good source of such full coverage data, and interpretations of surficial geology and geomorphology are also invaluable predictor variables. Further analysis of the biotope groups and predictor variables provides information which allows selection of the best available predictor variables which are then retained in the final model. To model the biotopes MAREANO has investigated a number of methods, but the present map has been made using Maximum Entropy Modelling (MAXENT) which is a machine learning technique quite widely adopted by scientists modelling both marine and terrestrial habitats, and one which gives valuable information about community-environment relationships. MAXENT allows production of separate maps for each biotope which are then combined into a composite map showing the overall distribution of biotopes. From the analysis it is clear that different predictor variables have different importance to each biotope, but that overall it is bathymetry, landscape, slope and sediment grain size which are most important which is why these variables are retained in the legend.
A summary of the modelling process adopted for the new Tromsøflaket-Eggakanten biotope map is given in Figure 3. Further details on the classification and modelling methods can be found in this report from the Troms II and Nordland VI areas where MAREANO scientists used the same general approach for biotope mapping.
Showing the workflow for production of the new Tromsøflaket-Eggakanten biotope map.