Direct mapping
 

Direct interpretation of images/reclassification of images

When dealing with intertidal habitats, the interpreter can go to the field and obtain direct observation of ground facies against how they appear on the imagery. The observer can quite easily forge their own idea of a signature. (Note that this use of the term ‘signature’ differs from that used in supervised classification. In the present context signatures indicate a relationship between habitat characteristic as observed directly and patterns as observed in the coverage image. (For supervised classification, signatures have a more statistically rigorous meaning.) Generally such signature data are rather abundant as ground-truthing the intertidal zone is not so heavily constrained, and signatures can be refined to a high level of quality. In the case of subtidal habitats, direct observation is more difficult to achieve. Therefore getting reliable signatures may be more difficult in subtidal mapping.
 
Sidescan images may be directly interpreted in terms of physical habitat features (sediment, bedform, topography) with a minimum of direct observation and then re-interpreted knowing which biological communities they support. For example, side-scan images may have been interpreted in terms of bedform and sediment types prior to integration and might need a reclassification. Ground-truth biological records can be overlain and matched by eye. It is then assumed that similar patterns support the same biota. If other layers, such as depth contours and slope, are also incorporated, then the substrata can be divided by depth zone etc and the extrapolation of the pattern/biota links will be further constrained by these additional factors, improving the final map. This may be relatively successful for small, simple areas but as the survey area increases, confusion over the distribution of biota over similar patterns in the backscatter is more likely to occur.
 
When interpreting remote sensing imagery of seabed habitats, the interpreter needs to relate the expression of ground units given by the imagery to their reality in the field, in other words he looks at signatures of his facies or habitats.
 
Signatures can vary for a given facies or ground unit. They depend on natural factors, which will make them different to the observer even though they may be clearly assigned to a given class. Slight colour nuances in the composition of the facies itself can be of considerable influence on the way they are remotely-sensed. For example muddy sand can be covered with some microflora, which modifies its aspect on aerial photographs. The mappers use their knowledge and expertise to allow for this variability.
 
An example is given in the Bay of Concarneau (France) with a “pockmarked texture” (figures), a spectacular facies never highlighted in previous studies in this bay. These bedforms are usually seen in deeper water. The pockmarks form a highly unusual chain pattern here. The numerous samples taken from these bottoms were made up of a compact mud with highly concentrated Haploops communities on their surface.
 
GMHM4-42_Still_camera_view_small-sized_pockmark.jpg
Still camera view showing a small-sized pockmark
 
GMHM4-43_Side-scan_sonar_imagery_of_pockmarks.jpg
Side-scan sonar imagery of pockmarks in the Bay of Concarneau. The EUNIS class is “Sandy mud”. The facies is described as “clear grey facies with craters”.
 
These signatures can then either be used in automated classification or in direct manual interpretation, depending on the methodological choice of the interpreter. The quality of the interpretation will depend directly on the quality of these signatures.
 
Links to other sections:

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 |