Classification of shallow seabeds in high resolution satellite data
The general purpose of this project was to evaluate to what extent high resolution satellite data could be used to increase the information basis regarding seabeds in shallow coastal areas.
Shallow seabeds are often characterized by high biodiversity and variation, which need to be restored or maintained by continued prudent use. National, regional and local authorities need information about the subsurface environment in order to make decisions regarding preservation actions, developments and protected areas in the coastal zone. Sufficient knowledge to make those decisions does not exist for most areas today and it is necessary to develop new methods that can help increase the information basis. Information derived from high resolution satellite data, which covers relatively large areas, could serve as a good basis and a complement to field investigations.The main part of the reported results is based on image data collected over the archipelago of Arkösund within the Municipality of Norrköping.
The possibility to map bathymetry from high resolution satellite data is completely dependent on the transparency of the water and so far, images collected during “clear water” conditions in spring, have generated good results down to 3-3.5 meters. The evaluation of the produced depth maps indicated the possibility to estimate depth with accuracy around 0.3-0.4 meters. Multispectral satellite data is useful for production of maps showing bottoms with vegetation as separated from bare ones, but are limited by the Secchi depth. It is not possible to separate dark vegetation from deep areas based on just one band, which indicates that panchromatic data is insufficient for this purpose. Furthermore, our results indicate that it should be possible divide vegetation covered areas into fucus vesiculosus and other green vegetation, but that a significant amount of field data probably is required. Different vegetation species could be possible to separate in shallow areas if it occurs in larger relatively pure populations. However, any classification on species level requires large amounts of field data collected in the imaged area, and preferably as close as possible to the image collection.
In this study we have used QuickBird data for the analysis and the conclusions are strongly related to this sensor (and similar). There are other sensors, with better spedtral and radiometric properties, available now and the benefit of using these sensors for this application is presently investigated.
The work has been performed as a co-operation between Brockmann Geomatics Sweden AB (SwedPower AB initially) and potential end users, i.e. the Municipality of Norrköping, County of Östergötland, Swedish Environmental Protection Agency and WWF Stockholm. The work was funded by the end users and the Swedish National Space Board.
Petra Philipson, Brockmann Geomatics, email@example.com,
Mobile: 070-699 60 46
25 maj 2009