A. C. Teodoro, F. Veloso-Gomes, H. Gonçalves
Journal of Coastal Research 24 (sp3), 40-49, (1 July 2008) https://doi.org/10.2112/06-0770.1
KEYWORDS: coastal zone, total suspended matter (TSM) concentration, linear single-band models, linear multiple regression, artificial neural networks
The transport and distribution of sediments driven by mechanisms such as tides and waves, river discharges, wind stress, and turbidity currents and the sediment transport effects can be studied by remote sensing techniques. The study of total suspended matter concentration has ecological importance since it is the main carrier of various inorganic and organic processes.
The main objective of this work was to evaluate the performance of different statistical methodologies in the estimation of total suspended matter concentration, in the breaking zone and in its adjacent area, using multispectral satellite images from TERRA/ASTER, SPOT/HRVIR, and Landsat/TM. These images cover a particular area of the northwest coast of Portugal.
The relationship between the total suspended matter concentration and the spectral response of the seawater, in the visible and near-infrared regions of the electromagnetic spectrum, was quantified through simulations on different beaches of the study area.
Seven images of TERRA/ASTER, SPOT/HRVIR, and Landsat/TM sensors were calibrated and atmospherically and geometrically corrected. Linear single-band models, linear multiple regressions, and artificial neural networks were applied to the visible and near-infrared bands of these sensors in order to estimate the total suspended matter concentration. Statistical analysis using determination coefficients and error estimation was employed, aiming to evaluate the most accurate approach in the estimation of total suspended matter concentration.
The analysis of the root-mean-square error achieved by both linear and nonlinear models supports the hypothesis that the relationship between seawater reflectance and total suspended matter concentration is clearly nonlinear. Artificial neural networks have been shown to be useful in estimating the total suspended matter concentration from reflectance of visible and near-infrared bands with images of ASTER, HRVIR, and TM sensors, with better results for ASTER and HRVIR sensors. The artificial neural network approach was further applied to the seven processed images, and maps of total suspended matter concentration for all satellite images processed were produced.