The main objective of this study is long-term monitoring of surface sediment on tidal flats using optical remote sensing. Tidal flat reflectance is sensitive to conditions that constantly change and cause variation of interstitial water contents along with exposure time, remnant water, etc. It is difficult to retrieve sediment grain size from optical reflectance alone without correcting the tidal condition effects. In this study, the tidal flat surface reflectance model according to grain size is proposed by two-step PCA transformation to remove tidal effects. The proposed method showed a potential to classify sediment by grain size regardless of exposure time and tidal conditions. We applied the method to nine scenes of LANDSAT TM images acquired between 1988 and 2009. The preliminary results well demonstrated that the proposed approach is effective to monitor changes of grain size distribution at a large scale independent of tidal conditions. However, this method has a limitation particularly over the vegetated areas and sand shoals mainly because of strong reflection in the NIR channels compared with other areas. For the validation, this study requires further field survey to obtain more in-situ data.