| Abstract | Coastal zones are of great ecological and economical importance. Nevertheless, they are threatened by coastal erosion, which forms an increasing global problem. Reliable and timely information on
coastal dynamics is a prerequisite for more sustainable management of coastal zones. Satellite remote sensing has the potential to provide this type of information. The present study aims to develop a procedure for extracting coastlines from a
variety of satellite remote sensing data and to apply this procedure to map the dynamics of the Canadian Beaufort Sea coastline. This coastline has been reported to be particularly sensitive to erosion and hence to the impact of sea level rise.
The Canadian Beaufort Sea coastline forms part of an arctic region. It has a generally low relief of less than 60 m and is composed of unconsolidated sediments, which are bonded by ice. The warm water of the Mackenzie River exerts a great influence
on the area, by inducing thaw settling of the sediments and fastening break-up of the sea ice. During the open water season (July-September) storm winds may generate high storm surges over ice-free fetches. Jointly, these processes destabilise the
coast and induce rapid coastal retreat. The coastline of the active delta is retreating most rapidly. The process of global warming is expected to accelerate coastal erosion in the future. Various methods for coastline extraction are described in
literature. Most of these methods are designed especially for either optical or radar imagery. In this study, a variety of images were used, including a Landsat MSS image of 1973, a Landsat TM image of 1986, four SPOT panchromatic images of 1991 and
a RADARSAT W2 image of 1999. Hence, it was necessary to select a method that could serve to extract coastlines from different image types. The approach adopted may be referred to as local region growing. Image analysis started with the
despeckling of the RADARSAT and then the scaling and geocoding of all images available. Next, coastlines were extracted by means of local region growing. Region growing is an iterative process in which regions are merged based on a homogeneity
criterion. In this study the criterion used was a gray level threshold value. Local region growing applies the algorithm in various subregions. This allows setting different threshold values, and can therefore account for the variable land-water
contrast in the image. The most subregions needed to be defined for SPOT panchromatic images, since they have the least contrast between land and sea. This results from the high reflectance of sediment-rich water in the visible spectrum. Overall,
local region growing provided a good estimate of the coastline position in the individual images. Positional inaccuracies are mainly due to flaws in the process of geocoding, the presence of mixed pixels, and incorrect threshold selection resulting
from misinterpretations of the coastline position. To account for possible inaccuracies, trends in the temporal sequence of the four coastlines were used in the definition of erosional classes. Four classes were distinguished: rapid erosion (>5 m
yr-1), moderate erosion (1-5 m yr-1), no detectable erosion (-1 - 1 m yr-1), and accretion (<-1 m yr-1). The resultant coastal erosion map shows rapid coastline retreat in the active delta in particular. Uncertainties in the positional accuracy
of the extracted coastlines are believed to be too high to make reliable assessments of temporal differences in erosion rates. Therefore, this study can neither confirm nor refute the expected acceleration in coastal retreat due to global warming.
Yet, the dominant coastline dynamics of the Canadian Beaufort Sea could be mapped. This demonstrates the value of satellite remote sensing as a tool in support of coastal erosion studies. In the future, the possibilities for the monitoring of
coastal dynamics by means of satellite remote sensing are expected to improve, because (a) the time span for which data will be available increases, (b) the available image database is expanding as more satellites are being launched and (c)
satellites with enhanced spatial resolution are being introduced. The success of any coastal erosion study depends heavily on the data available. In terms of wavebands, the application of either radar images or infrared images is recommended. In
terms of spatial information content, the user will have to establish a compromise between spatial resolution and spatial coverage. |