Artificial neural network has been proven a useful technique for deriving water and bottom properties from remote
sensing upwelling radiance. Conventionally, a neural network is trained to minimize the overall mean square error of
desired products. The approach does not explicitly take into account the change of spectral shapes of upwelling radiance.
In this study, we have created four groups of training sets, two groups with ratios of Rrs( λi) to Rrs(557), and the others
without. Ratios of Rrs( λi) to Rrs(557) for λi of 409nm, 438nm, 488nm, 507nm, 616nm, 665nm, 683nm, 712nm, 750nm
and 779nm have been used as additional inputs in the training of neural networks. Trained neural networks were then
applied to an independent testing set which was created for optically different coastal waters. The inclusion of 10
spectral ratios in the training significantly improves the accuracy of derived water depth H, backscattering coefficient
bb(438) and the absorption coefficient a(438). The accuracy of the derived coefficients is 86%, 94% and 92%. Our
results clearly show the importance for including spectral ratios in the neural network training process. Remote sensing
upwelling radiance over the identified 11 spectral channels provides adequate information for the retrieval of water
optical property coefficients when an artificial neural network approach is used.
In an earlier ocean-color algorithm, water’s optical properties are classified into two categories. The major properties, such as the absorption and backscattering properties, vary widely and have significant influence on ocean color. The minor properties, such as the spectral slope of the gelbstoff absorption and the spectral power of particle backscattering, affect the ocean color modestly. The main objective of ocean-color remote sensing is to derive the major properties from water color. In model-based inversion algorithms, it is required to know the values of the minor properties. In this study, neural networks (NN) are used to estimate the minor properties. The NN-estimated minor properties are further used in a quasi-analytical algorithm to analytically derive the major properties. Significant improvements are found in the derivation of absorption and backscattering coefficients of coastal waters. The results here indicate an advantage of the neural network approach in inexplicitly linking a water property with water color, especially when there is no apparent relationship that can be explicitly expressed. The results further demonstrate the capability of the quasi-analytical algorithm to analytically derive major water properties from water color.