Assessment of Different Bathymetry Statistical Models Using Landsat-8 Multispectral Images

Abstract

Remote sensing has many applications including bathymetry mapping in shallow water areas from multispectral images. It is considered a useful reconnaissance tool to save time and cost to be used in the preliminary survey. This paper aims at assessing the performance of three techniques used for bathymetry modeling to obtain water depth. In this research, three models are applied to obtain water depth using optical bands of Landsat-8 multispectral images. The selected algorithms (models) are principle component analysis (PCA), ratio model (RM), and 3rd polynomial model. The images are first corrected for the effects caused by atmospheric conditions and sun specular. The bands are then calibrated using ground truth measurements obtained using GPS and single-beam echo sounder in the area of the eastern harbor of Alexandria, Egypt. The study area represents a low turbid, mud bottom, and shallow water area. The three algorithms are implemented, and the results are compared with echo sounder data. The root mean square error (RMSE) is computed for the three techniques for different spectral bands combinations. RM resulted in RMSE 1.04 m using the coastal and green bands. The RMSE of PCA was 1.02 m resulted from the blue and green bands. The 3rd polynomial yielded 0.91 m using from the green band. The capability of using Landsat-8 images in bathymetry was approved. In addition, the outperformance of green band usage in bathymetry was also argued.