In this study, TM bands were cut in the size of 768 lines by 768 pixels after the geometric correction of original data for the areas including three estuarine reservoirs, that is, Yongsan estuary lake, Yongam lake and Kum lake. In the geometric correction, ground control points of TM data are selected on the topographic map of 1:50,000 scale, published in September 1995 in the national geography institute, Korea. The absolute difference between the computed and measured values of the image coordinates was kept down within 1 pixel. The coordinate transformation of which was used Affine transformation equation.

The atmosphere correction in remote sensing that the absorption and the emission of atmospheric energy are regarded as a noise is necessary to correct. The diverse methodologies of atmospheric correction have been proposed. It is classified into three types that consist of the method using radiative transfer code, skylight emission measurements, and regression analysis. The radiative transfer code is used to estimate surface radiance and atmospheric transmittance on the conditions of atmosphere, geometry, and wave number based on radiative transfer equation. The typical codes are MODTRAN (Kneizys et al., 1996) and 6S (Vermote et al., 1994). In addition, the dark object subtract can be also applied when atmospheric affects remove briefly according to study purpose. The method deduct the minimum value of observed DN (Digital Number) values from the image including area that the reflectance is regarded as zero such as wide water area and the shade of clouds. This study uses the radiative transfer code (6S) to execute the atmosphere correction of the image data without a lot of information about atmosphere. When Landsat imagery is used in studies of this nature, there are usually errors associated with the estimation of parameters. The significance of these errors depends on the application of the data. This paper is mainly connected with a synoptic variations in the freshening process of estuarine reservoir, and therefore these errors will not have a significant effect on the objectives of this work (Baban; 1997).

4. Methodology
The study methods of this paper include the water quality observation of field, the simulation of inflow patterns using a numerical analysis, and the processing of Landsat TM images, to analyze the convective mixing patterns of estuarine reservoir.

4.1 Water Quality Observation
The Yongsan river third-stage office of Rural Development Corporation, Korea has been measuring the water quality of lake periodically twice and three times per year since 1991. The salinity has been investigated at the four points of the vicinity of outlet gate (S1), the outlet of link canal (S2), the central part of lake (S3), and the upstream part (S4), along the center line of Yongam lake, as shown in Fig.-1.

4.2 Numerical Simulation of Flow Fields
A simulation related to the mixing patterns of inflows, which is the two-dimensional simulation of flow fields by adopting finite difference method on general curvilinear coordinate system. The various conditions of grid generation in the simulation process were examined, and the temporal variations of horizontal velocity-vector distribution were applied and visualized for the flow fields of final enclosed freshening reservoir.

4.3 Analysis of Satellite Image data
The analysis of TM image data, which is focused and classified for categories such as inflows and stagnant water related to water bodies and convective mixing patterns on the surface of water, after image processing described in the preceding chapter. However, this study was not carried out for a correlation between water quality categories and image data owing to the lack of water quality data.

5. Results And Discussion

5.1 Temporal Variations of Water Quality
The Yongam lake has topographical conditions that is not suitable for freshening because the inflow from watershed is not sufficient to desalt, which is the feature of a standard estuarine reservoir. However, the water depth of lake (6~9m) is shallow relatively except downstream area near outlet gate, maximum depth is approximately 23m. The freshening process is progressing satisfactorily by the introduction of desalted fresh water from adjacent Yongsan lake for desalinization. The salinity that is an important criterion evaluating freshening process has been observed since 1991. Table 1 shows the temporal variations of water quality observed under conditions operated normally by means of desalting facilities as outlet gate, desalt siphons, and suction pumps. These are the results investigated in the downstream (S1) of Yonam lake from Apr. 1994 to Nov. 1996.