COMPUTER PROCESSING OF LANDSAT IMAGERY OF THE SEA

Abstract

Many satellites are orbiting the Earth gathering data that can be used for a wide variety of studies. In particular data gathered by the Landsat series of satellites have simulated a great deal of interest. The thesis describes a study of data gathered by Landsat II for an area covering Singapore and its surrounding regions; an area designated as path 134 row 59 in Landsat terminology. Analysis is made of data of this area acquired at two different dates, namely 21 March 1977 and 1 June 1977. A general purpose computer at the University Computer Centre is used to process the multi­ spectral scanner data of the sea on computer compatible tapes. In order to analyse the data correctly an accurate knowledge of the Physics of how they have been acquired is essential, The Landsat data acquisition system and the Physics of atmospheric transmission of electromagnetic waves are described in detail in the thesis. Landsat was not primarily optimised to study the sea. On the contrary it was designed to study vegetation on land. The pixel radiance range of the sea portion of the imagery is therefore small and hence imposing a limitation on what could be achieved. The high incidence of cloud cover in the tropics, 40% and 30% in the two imagery under analysis respectively, complicates matters even more. Quite apart from just obstructing the view the cloud creates radiometric errors in the readings if cloud and sea surface are scanned in the same line. Scanning the highly reflecting cloud saturates the sensor and if the sea with a very low reflectance is then scanned the value of the sea radiance obtained is higher than it should be. These radiometric errors which do not bother investigators of land areas become an important factor and the errors have to be processed to minimise them before an analysis can be carried out. Information of sea depth is obtained by devising an algorithm with radiometric correction and elimination of atmospheric flunctuation incorporated into it. The model is developed taking refinements such as light scattering in the water and internal reflections at water surface into account. This is compared to models using either the green band or the red band in single channel analysis and using both the green and red bands simultaneously in a ratio analysis. Some very interesting results have been obtained. In the clear shallow water depth variation shows up prominently, The pixel radiance value varies gradually with depth in coastal water but the variation is more abrupt in deeper sea. It indicated that there exists a definite correlation betw0en the intensity contours of Landsat data and the depth contours of the sea. This result was presented at the 13th International Symposium on Remote Sensing of the Environment (1979) and the paper was published in its proceedings (page 1807). This study has also provided precise information on the capabilities and limitations of remote sensing techniques. Such information could prove useful in evaluating the appropriateness of employing remote sensing techniques in studying any particular aspect of the sea.