Expert-eye analysis of remote-sensing data

When remote sensing has been used to draw up a draft physical map of an area, the interpretation of that map in terms of the substratum classification and bedforms should undergo some form of validation procedure using observational and/or sampling techniques. Failure to do so will reduce the confidence in the final mapped output.
Ideally, each of the mapped polygons should have been sampled during the ground-truth survey, but this is often impractical. It is more likely that each substratum class has been sampled using a variety of techniques. The data and/or information from these samples should now be used to validate the draft physical map. If the provisional classifications prove to be incorrect, they should be modified accordingly.
Where soft sediments predominate, a full geophysical analysis of samples will provide the highest confidence in validating the sediment classes. This may involve a particle size analysis and/or testing the geotechnical properties of the sediments (penetrometry, shear strength etc). If this is not available, moderate confidence can be gained from sediment descriptions resulting from direct inspection of samples, such as in a walk-over survey of a beach or examining sediments from grabs or core samples. Failing this, the validation may have to rely on the results of remote observation, where video and/or photographs have been used to image the seabed.
Where hard substrata predominate (i.e. rock outcrops, boulders, cobbles), validation relies heavily on observational techniques. In-situ, direct human observation provides the greatest confidence, using results from walk over surveys of the shore or diver surveys in shallow waters. For deeper waters, observation may be restricted to the use of video or photographic techniques, and confidence can be less on account of their limited field of view. If observational techniques have not been used, some degree of validation can be gained from samples collected by rock dredges or very large grabs.
The draft physical maps may include some interpretation of bedforms, such as ‘sand waves’ or ‘mega-ripples’. In many cases, the remote survey techniques are the best method of identifying these bedforms, and opportunities for validation are extremely limited. Here, cross-validation may be appropriate, comparing the outputs of two remote sensing techniques to see if they both detect the same bedform (e.g. mega-ripples detected by both multibeam and side-scan sonar). Sampling techniques can rarely be used to validate bedform interpretations, as the scale at which they sample is usually far smaller than the scale at which bedforms exist (grabs may only sample an area of 0.1 m2, but sandwaves may have a wavelength of several metres). Observational techniques (video and diver surveys) are probably the only affective ground-truth method for verifying sub-tidal bedforms, though they may only be applicable to bedforms on the metre rather than decimetre scale (due to restricted visibility).
The validation process should also consider verifying the boundaries between different sediment types and/or bedforms detected by remote survey techniques. The placement of borders may be the most critical part of a habitat map, so their location and characteristics are important to the overall confidence of the map. Are the borders in the right place, and do they represent distinct or gradual changes in sediment type? Here again, cross validation between the outputs of two or more remote survey techniques can be informative. Direct observation or video transects can also be used. In shore surveys, a portable GPS system can be used to track a detailed outline of habitat boundaries.
Examples of interpretation of sediments, bedforms and borders from side-scan sonar surveys. The interpretations should be validated using data from ground-truth surveys

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 |