Determine the strategy

There are many approaches to habitat mapping, but it is convenient to illustrate the range of applications as lying along a scale spectrum: broad to fine-scale. This spectrum has been adopted throughout the MESH Guide and should now be familiar. Broad versus fine-scale best contrasts the different approaches, but become more difficult to explain for intermediate scales.
GMHM2-1 Broad to fine scale summary
An illustration of the type of maps across the broad to fine-scale spectrum
Benthic habitat mapping will inherently involve some form of modelling. In this context, a model can be seen as any representation of the sea floor based on a systematic investigation and analysis of the evidence for habitat distribution. This is an intentionally broad definition, which can range from expert knowledge interpretation ‘by eye’ and rule-based models through to statistical models based on an extensive analysis of data. Modelling is further described in the section Modelling in habitat mapping. Bear in mind, however, that techniques described under one heading may be applied to other situations. For example, it is possible to use the ‘broad-scale’ modelling approaches for small geographic areas and to use the ‘fine-scale’ detailed approach for very large areas (e.g. national seabed surveys in Ireland (INFOMAR) and Norway (Mareano).
The order of analysis and modelling is important: it is recommended that analysis of ground-truth data is considered before remotely-sensed/physical coverage data because a good understanding of the biological data is required to select and derive physical coverages that are more biologically relevant. This will enable a more meaningful integration with the ground-truth or sample data. This is not to be confused with the recommended order of survey, which is to undertake the remote sensing before the ground-truthing. The accuracy and confidence of the maps are discussed in How Good is my map?.
GMHM4-3 Overview of steps
Scheme of the main methodological steps in habitat mapping

Broad-scale habitat mapping for very large areas

A broad-scale habitat mapping approach, normally applied to whole seas or national waters most commonly uses coverages for the major physical parameters that are combined to obtain the predicted distribution of a wide range of broadly defined habitat types (e.g. EUNIS levels 3 and 4). The data used are derived from multiple sources (see What is habitat mapping?) and the modelling carried out within a geographic information system (GIS). The combined errors and uncertainties of the input data layers will mean that the resulting maps are necessarily rather coarse and generalised in nature.
Flow diagram
A summary of the broad-scale mapping approach
The modelling will be based on applying general rules relying on knowledge of causal relationships, valid across a large geographic area. The area is divided up into habitat categories based on combinations of a limited number of zones for each of the key environmental structuring variables (e.g. depth, substratum and water energy). This may involve expert opinion to reclassify the physical parameters into a small number of biologically relevant categories without recourse to sophisticated statistical methods to justify the upper and lower boundaries of the structuring variables.
The MESH EUNIS modelling approach is an example where the physical variables used to define EUNIS level 3 classes (substratum, depth and incident energy) have been combined to derive the predicted distribution of broad-scale habitats. The approached used by the MESH Partners to derive marine landscape maps adopts a similar approach,but includes a layer for seabed slope making the product easier to understand from a physiographic / topographic viewpoint.
A simple model was used to predict the distribution of EUNIS habitats for the seabed of north-west Europe.

Fine-scale habitat mapping for small areas

At the other end of the modelling continuum, field surveys of small geographic areas usually adopt a data-driven approach. More narrowly defined habitats (e.g. EUNIS levels 4 and 5) can now be predicted and ground-truthing is usually more detailed. Mostly, ‘signatures’ are created from the physical coverages (usually acoustic or electromagnetic spectrum (EMS) reflectance values) using the ground-truth records as training sites. Interpretation of the data from each study is relatively self-contained in terms of its data requirements and no attempt needs to be made to derive rules that can be exported to other studies. This strategy is exemplified by supervised classification of remotely-sensed images and there are many examples of surveys where this approach has been used. It is likely that specific image-processing GIS software will be required. Side-scan sonar images may discriminate a much smaller set of habitats and be more straightforward to interpret than for large areas: intermediate interpretation of side-scan sonar imagery to produce physical habitat maps may be by-passed by interpreting biological habitats directly.
Scheme of the fine-scale mapping approach

Intermediate scale mapping

Intermediate scale modelling lies between the very broad and the fine-scales and it proves to be the most difficult type of modelling on which to offer guidance. It is likely that the input data will come from many surveys over an extended campaign or different campaigns. Some of the inputs may come directly from surveys whilst others will be modelled mathematically (such as bed stress). Modelling strategies might range from:
  1. the use of statistical techniques to investigate relationships between different physical factors and the biology, using spatial statistics to optimise interpolation,
  2. most data-driven and rule-based or knowledge-driven models; or,
  3. mixtures of the two approaches can also be used together.
The first statistical approach is easier to apply to a single habitat (e.g. mussel beds) than to the complete range of habitats found in an area. The second is similar to the approach adopted for very broad-scale mapping and suitable for modelling the range of habitats present. The incorporation of more data-driven statistics into such a model bridges the distinction made between the two approaches. Greater emphasis on statistical models will require the use of specialist statistics and mathematical software requiring specially written routines.
intermediate-scale mapping
A summary of the intermediate-scale mapping approach
An example of an intermediate scale mapping product is presented in the following figure. A probability map (250 m grid resolution) of the macrobenthic communities (EUNIS level 5) of the Belgian part of the North Sea is shown, based on data-driven statistical modelling. Related to the same area, an extract is shown of the EUNIS level 3 MESH EUNIS map (1 nm grid resolution). The differences are mostly observed in the coastal zone where the occurrence of one macrobenthic community is completely missed with the broad-scale approach.
Example of an intermediate (on top) versus a broad-scale (below) mapping product
Links to other chapters:
Links to other sections:
Links to other websites:

All material variously copyrighted by MESH project partners 2004-2010

| List Access Keys | Mapping European Seabed Habitats (MESH) - Home | Overview | News | Work areas in the MESH Project | Partners | Contact Us | Search | Site Map | Partner Extranet |