The advantages of pan-sharpening in coral reef research

Tony Vanderstraete, Rudi Goossens
Ghent University, Geography Department, Ghent, Belgium
Tony.vanderstraete@Ugent.be, Rudi.goossens@Ugent.be

Tharwat K. Ghabour
National Research Centre, Soils and Water Use Department, Gizah, Cairo, Egypt
thghabour@lycos.com

ABSTRACT

During the last three decades of coral reef research using remote sensing, it became clear that detailed information concerning coral reefs could only be obtained with high spectral and spatial resolution data. However, recently developed high resolution sensors like IKONOS or QuickBird, and hyperspectral sensors, e.g. CASI or HyMap, are not routinely used as these are expensive tools and not practical for monitoring often remote coral reefs located in developing countries. Therefore, moderate resolution sensors, like Landsat or SPOT, are still the most cost-effective option for regional mapping and inventory coral reefs.

Earlier attempts to upgrade the spatial resolution of Landsat datasets by merging them with auxiliary higher resolution satellite data, usually SPOT-Pan, failed as most data fusion algorithms resulted in a significant loss of spectral information. Recent data merging algorithms, like the Adaptive Intensity Matching Filter (AIMF) algorithm developed by de Béthune et al. (1998), guarantee a maximum preservation of the spectral integrity. Besides, with the simultaneously taken high resolution panchromatic scene and multispectral dataset by the Landsat7 ETM+, problems due to different environmental settings during the acquisition of the two datasets are avoided.

The aim of this paper is to investigate the possibilities of pan-sharpening a Landsat7 ETM+ dataset based on the AIMF data fusion algorithm for studying coral reefs. As study area, some coral reef structures offshore Hurghada (Egypt) in the north-western part of the Red Sea, are selected. Different information layers can be derived, these being a bathymetric map, a geomorphological zonation map of the reef structures and an inventory of the main bottom-types occurring on these reef structures. These products are subsequently compared with the results obtained by the processing of the original dataset in order to determine the advantages or disadvantages of pan-sharpening.

It can be concluded that pan-sharpening results in more detailed bathymetric and geomorphological zonation maps. In contrast the benthic habitat classification obtained with the original Landsat7 ETM+ dataset is more accurate. This may be explained by the fact that the classification algorithm greatly depends on the multispectral information available and the data fusion algorithm applied still induces some loss of spectral information. As a great deal of the light reflected by the seabed is already lost by interaction with the overlying water column and atmosphere, any additional loss of reflectance hinders the interpretation of the dataset. The bathymetric method applied and the visual geomorphological zonation mapping, on the other hand, are less dependent on the actual radiance values presented in the dataset.

Last Update: 2005-03-16