Evaluation of Landsat TM and Spot data for shallow water Bathymetry
K. Abduallah, K. Mohd Dimyata, A.P. Cracknell, R.A. Vaughan
Department of Applied Physics and Electronic
& Manufacturing Engineering, University of Dundee
Dundee DD1 4HN, Scotland, United Kingdom
Abstract
The feasibility of using SPOT and Landsat TM data for bathymetri studies on coastal areas near Qatar and the United Arab Emirates has been investigated. We have used one SOPT and two TM sets of data purchased for our use by the Hydrographic Department of the Royal Navy. The applications of single band, two band ratio and three-band hybrid and principal components methods has been investigated. Accurate geometrical rectification of each data set was first performed. Test areas were selected for the calibration of depth-intensity algorithms for each data set using known sounding points. The algorithm was first tested with other known sounding points. The algorithm was first tested with other now sounding points within the test area and then used for two previously unscrewed areas with each data set. The investigation showed that new information about the depth of the water in relatively shallow areas can be obtained from these types of satellite data.
Introduction
Although accurate measurements of water depth can be achieved by using conventional shipboard sounding techniques, such a survey is slow and expensive. To maintain a chart series in an accurate, adequate and up-to-date condition is an immense task to the hydrographers. As a result, investigators have become interested in the application of remotely-sensed data for hydrographic charting. The successful application of passive remote sensing techniques to this problem is restricted to shallow, clear water areas with small changes in bottom types and free from atmospheric contamination.
Algorithms for water depth mapping in coastal areas using satellite data were developed by Lyzenga (1978, 1981, 1985), Paredes and Spero (1983) and Spitzer and Dirks (1987). The principal components approach in bathymetry was developed by Lyzenga (1978).
In the present work the direct use of the algorithms, using single band, 2-band ratio and 3-band hybrid methods, and a principal components analysis approach were investigated in the coastal areas of Qatar and the United Arab Emirates. The potential use of these types of satellite data was tested in two selected previously unsurveyed areas.
Data for the Area Under Investigation
The SPOT scene and two TM scenes that we used were acquired on 2 June 1986, 13 October 1987 and 11 August 1988 respectively. The tidal data related to the scenes and the bathymetric charts were also supplied by the Hydrographic Department of the Royal Navy. Data on suspended sediment concentration, bottom composition materials and other related oceanic parameters were not available for reference.
Algorithm Description
The single band method employed in this work is based on the simple water reflection model (Lyzenga 1978). According to this model the radiance in a single band can be written as
Li = Lsi + Ki RB exp (-KifZi) -----------------------(1)
Where k
i is a constant which includes solar irradiance, the transmittance of the atmosphere and the water surface and the reduction of the radiance due to refraction at the water surface, Ls
i is the radiance observed over deep water, R
B is the bottom reflectance, K
i is the effective attenuation of water, f is a geometrical factor to account for path length through water and Z
i is the water depth.
Rearranging Eq. 1 for a water depth determination algorithm gives
Zi = [1n C1i – 1n (Li – C2i) ] / C3i --------------------------(2)
Where C
1i = K
iR
B, C
2i = L
s1 and C
3i = K
if
By defining X
i = 1n (L
i – L
si), the water depth, Z (determined with data from all the bands), can be written as a linear combination of logarithms of the radiance (Paredes and Spero 1983)
Z = b0 + b1Xi + ….b Xn -------------------------(3)
Where n < 3 in this case, and b
0, b
1…b
n are the regression coefficients.
Eq. 3 is based on the hypothesis that the bottom reflectance and the water composition remain constant within the major part of the image and that the signals in all three bands are influenced by the bottom reflection.
The 3-band version of Eq. 2 (ibrahim 1989) is
Where m
i is the coefficient to be determined, and the C
1i etc. were previously obtained from regression of Eq. 2.
Data analysis and Results for Spot and the first TM data set
For each data set, two test sites were selected for our analysis. The selection of the study areas was based on the availability of good spatial distribution of ground control points with a large number of depth sounding points. Each test area of the imagery was geometrically corrected by rectifying to the corresponding bathymetric chart of the chosen area using a second-degree polynomial transformation.
Depth sounding points were extracted from the chart and their locations were transformed into image coordinates. Pixel values at these points were then obtained for bands 1, 2 and 3. The respective tidal
