Discriminating sediment and clear water over coastal water using GD technique

Abstract

Abstract Currently two algorithms are being used routinely by the MODIS Atmosphere and Ocean Team in order to distinguish sediment influence and clear water pixels over turbid water area. These two algorithms require complicated computational analyses. In this paper, a simple algorithm based on empirical technique to detect the sediment-influenced pixels over coastal waters is proposed as an alternative to these two algorithms. This study used apparent reflectance acquired from MODIS L1B product. This algorithm is based on the gradient difference of the line connecting the 0.47- and 1.24-μm channels and 0.47- and 0.66-μm channels of a log-log graph of the apparent reflectance values against MODIS wavelengths. Over clear-water areas (deep blue sea), the 0.47-, 0.66- and 1.24-μm channels fitted very well in line with correlation R > 0.99. Over turbid waters, a substantial increase of 0.66 μm in the reflectance leads to a low correlation value. By computing the difference between the gradient of the line connecting 0.47 and 0.66 μm and the gradient of the line connecting 0.47 and 1.24 μm, the threshold to discriminate turbid and shallow coastal waters from clear-water pixels can be obtained. If the gradient difference is greater than 0, the pixels were then marked as sediment-influenced pixels. This proposed algorithm works well for MODIS Terra and Aqua sensor. The comparison of this algorithm with an established algorithm also showed a good agreement.