China University of Geosciences (Beijing)
Suspended Sediment; Quantitative Inversion; Scaling Effect ; Multi-Observations Analysis Algorithm; Multi-Models Information Assimilation Algorithm
As the development of economy and society, the water environmental problems become more and more serious. As an essential water quality monitoring approach, remote sensing techniques can be used to monitor the water quality of lakes quickly with large scale, dynamically and lowly costing, and has sigificance meaning for water environmental protection and water resources management. The paper is a case study on the case II waters in Taihu Lake, and carries on systematic research. The mainly research contents as follows:① The study on the optical properties of the turbid waters in Taihu Lake. In the Taihu Lake, the max peak of spectral curves was moving to the direction of short-wave as the increasing of suspended sediment concentration, namely the blue shift of wavelength. The area enclosed by spectral curve and wavelength coordinate axis in the range of sensitive bands has preferably linear relationship with the suspended sediment concentration (Curve Srea Model). The trapezoidal area model which was an approximation of curve area model could also excellently reflect those relationships, and be greatly suitable for multispectral satellite imagery retrieval such as Landsat/TM.②The scaling effect of quantitative models for water quality remote sensing was studied. The factors which cause the scaling effects of quantitative models, are the inhomogeneous distribution of suspended sediment concentration in waters and the nonlinear of inversion model. The scaling effect results in the suspended sediment concentration of inversion results are lower than the in situ measurements. Accordingly, we could define two different conceptions, apparent average concentration and average concentration, for suspended sediment concentration before and behind the scaling effects errors correcting.③The Influencing factors from the errors of imagery on the quantitative result were studied. The data error would be reformed by the quantitative model, and then melted into the retrieval errors. The performance is correlation with the first derivative of inversion model, and could be grouped into two classes, convergence and divergence. According to the data errors reforming mechanisms, the method based the highest of correlation coefficient or the least of root mean square error would not be reasonable for quality comparing between the quantitative models.④The mul-observation analysis algorithm was developed. The algorithm could be used to accuracy evaluating for retrieval results. Three-observation analysis algorithm is a special case of the mul-observation analysis algorithm, and the research on the three-observation analysis algorithm would help to understand the mechanisms of multi-observation analysis algorithm.⑤The multi-model information assimilation algorithm was developed, which basing on the kalman filter algorithm. This approach could be used to information extracting from multi-model. The error data should be input when uses this approach, and the resolutions of three-observation analysis algorithm could provide such parameter.