Bathymetry in clear waters from Landsat-5 Satellite Data
2. Case Study – Pulau Tioman
The study was carried out on the coastal waters of Pulau Tioman, Malaysia Using the Landsat-5, Thematic Mapper data. The waters in this areas is very clear with Secchi disk readings of about 40 m. The study area is as shown in Figure 1. The image was acquired on 1 April 1990 at about 02.50 GMT by the Landsat-5, satellite when the height of the tide was 1.5 m above lowest astronomical tide. The satellite data on abn 1 ( 0.45 – 0.52mm) was used to compute water depths since it has the best depth penetration capability.
Figure 1. Location map of study area
In the following sections, geometrical rectification of satellite data, depth algorithm used and some results that were obtained are presented.
2.1 Geometrical Rectification of Satellite Data
The satellite data used in the study was geometrically rectified to enable quantitative comparison to be made between the remotely – sensed image and existing maps and charts. This was achieved by carrying out an image-to-image registration on the Dipix ARIES-III Image Analysis System between a scanned hydrographic chart of Pulau Tioman and the corresponding Landsat-5 TM image at the Centre for Remote Sensing, UTM. The rectification was carried out to subpixel accuracy.
2.2 Method of Depth Determination and Depth Algorithms
For the remotely-sensed data used in the study, pixel intensities ( digital numbers ) were extracted at some points of known depth ( calibration depths ) in order to fit algorithms relating pixel intensities to the depth. The depths at the calibration points were determined on-site by lead-line measurements at the time of satellite pass while the positions of these points were obtained by sextant resection to some ground control points. These points were plotted on the hydrographic chart prior to scanning on the Dipix ARIES-III System so that upon registration to the Landsat-5 image, it was possible to extract the pixel intensities at these points.
A simple algorithm based on the model of Polcyn and Lyzenga ( 1975) was used , i.e.
I – A
1 + A
2 Exp ) A
3Z) (1)
Equation ( 1) can usefully be expressed in the alternative form as follows.
Z = [ In(A
2) – In ( I – A
1)] A
3 (2)
These equations express the expected exponential relationship between pixel intensity I and depth Z. The coefficients A
1, A
2 and A
3 describe the parameters outlined by the model of Polcyn and Lyzenga ( 1975). The coefficient A1 in particular, describes the effects due to the atmosphere which in most cases contribute about 50-80% of the total signal received by the satellite sensor. Therefore, a program was written specifically to calculate and correct the satellite data from the effects. The coefficient of Markham and Barker ( 1986).
Equation (2) can now be written as follows,
A=[In (A
2) – In ( I
corr )] / A
3 (3)
Where I
corr are the pixel intensities corrected for atmospheric effects. A least-squares minimization was carried out relating the corrected pixel intensities Icorr to depth at nine calibration points in order to obtain the best values of the coefficients A
2, and A
3 in equation (3) Table 1 lists the values of depths and their corresponding corrected pixel intensities at the nine calibration points used in the least squares minimization. Having determined the coefficients, the depth at any point on the image was obtained by using the values of these coefficients and the corrected pixel intensity values at this point from the relevant spectral band, i.e. band 1 of the satellite data.
