EARSeL eProceedings Vol. 3, No. 1, 34-42, 2004

Abstract
The 1992 European Habitats Directive (92/43/EEC) requires reporting of the status of a variety of habitats on a six-year cycle. One potential method of carrying out this monitoring is to use remote sensing. Improvements in remote sensing techniques would allow more accurate monitoring to be carried out, thereby providing more accurate indications of the extent and status of these habitats. This paper describes a study carried out to increase the potential of remote sensing for monitoring saltmarsh and sand dune Special Areas of Conservation (SACs). Methods of increasing coastal habitat classification accuracy by adding elevation derived data to multispectral data are examined.
Data were gathered using the ITRES Compact Airborne Spectrographic Imager (CASI) to provide multispectral data and the Optech Airborne Laser Terrain Mapper (ALTM) to provide digital surface models (DSMs) of two UK test sites. Multispectral remote sensing has previously been used for mapping the extent of coastal vegetation classes. However, there are ecological basis for including additional data in classifications, particularly slope and in the case of intertidal vegetation, elevation. This study used data derived from the ALTM DSMs to provide additional data layers in classifications. Statistical and neural network classifiers were used to assess increases in saltmarsh and sand dune vegetation classification accuracy when ALTM data were used in addition to multispectral data. Results are presented that show an increase in discrimination between intertidal land cover types when ALTM data are used in conjunction with fine spatial resolution multispectral imagery and that neural network classifiers can be more accurate for classifying coastal habitats particularly when multisource data are used.

ISSN 1729-3782