|OCEANS 2000 - Students Poster Session|
Comparison of Methods to Derive a Digital Elevation Model for the Wadden Sea from SAR data
by Danielle Hoja
One major interest of geoscientific applications like natural hazard management
support is the investigation and visualization of the earth and its changes.
Especially for the investigation of the temporal movement of geomorphological
structures in coastal regions, a great number of potential users in the area of
public administration, like government agencies monitoring waterways, have
shown interest on an operational analysis of remote sensing data for their
tasks. In this paper we analyze methods to monitore the wadden sea topography
by satellite and airborne radar data. Due to the all weather capability and the
high resolution of radar data, they have been shown to be optimal for
continuous investigations of coastline changes.
Here a comparison of several methods to map the continuously changing mudflat topography of the German Bight is presented. The results compared are an airborne interferometric digital elevation model (DEM), an interpolation of extracted coast lines from Synthetic Aperture Radar (SAR) images of the European Remote Sensing Satellites (ERS-1/2) and coherency maps (from ERS-1/2 tandem SAR images). Two investigation areas (50 km by 20 km and 10 km by 10 km) situated in the mud flats of the estuaries of the rivers Elbe and Weser in the German Bight are investigated.
Properties of different sensor data and evaluation algorithms used are analyzed. The airborne digital elevation model is a result of the interferometric processing of cross track interferometry data of the AeS1 sensor (Aerosensing Radarsysteme GmbH). Result is an high precision topography model of the wadden sea (resolution 2.5 km). Coherency data of the ERS Tandem Mission in 1995 and 1996 are used for coastline extraction by analysis of the much larger coherency of land as compared to sea surface. The coherency data show unexpectedly areas of high coherency in the tidal flats, which have been flooded between the two acquisitions. With coastline extraction from coherency data, only the line of the highest water level between the two images can be achieved. To get the actual land-water-boundary in ERS SAR scenes (resolution 12.5 km) at the acquisition time a coastline extraction method was used for the comparison, which combines edge detection by wavelets, a blocktracing algorithm and an active contour algorithm. To convert extracted coastlines derived from several scenes into a DEM additional knowledge about the water level from tide gauge and on the form of the coast lines is used. Each land-water-boundary is attributed with the actual water level and then all lines are interpolated into a DEM. Another possibility shown is the updating of an existing DEM with this data.
In a second step the algorithms to derive digital elevation models have been assessed in terms of the transferability of the methods to other and larger areas, e. g. the whole German Bight. Therefore acquisition parameters of the data, like the acquisition capability and frequency of each sensor, the applicability for operational use of each method and the height accuracy of the generated DEM was taken into account.
It is shown that the coastline extraction and interpolation method offers a good possibility for operational use and the interferometric algorithm leads to very good height accuracy (5 cm). Because the coastline extracted from coherency data provides only the line of the highest water level between the two images, this method is not suitable to achieve a DEM of the wadden sea, but yields additional information, e. g. on sediment type current speeds during flooding.
The final comparison uses statistical methods and additional knowledge about the water level and geomorphologic forms in this region. The coastline extracted from ERS SAR images is overlaid on the interferometric DEM deduced from the airborne data. The result is analyzed visually and statistically by comparing the coastline pixels to the DEM heights.
In general the comparison of the methods show a very good overall agreement for DEM generation in spite of the different sensor resolutions and algorithm accuracy. The high morphodynamics in this area can be mapped and monitored.
So for further use of the digital elevation model of the wadden see, e.g. as input into a mathematical tide model, ERS SAR data of coarse resolution can be used as an overview. Where high quality topography data are needed because of intense variation in a small area, a DEM generated by the airborne system yields the necessary fine resolution.