Based On Remote Sensing Of Suspended Sediment In The Radial Sand Ridges Of The Underwater Terrain

The retrieval of underwater terrain is accepted gradually with the development of remote sensing. It can be achieved to obtain the information of underwater terrain on large scale accurately and effectively by adopting this technique. We take the Xiyang radial sandbank in Jiangsu coastal as the study area. We establish the retrieval model of surface suspended sand with the volume concentration of surface suspended sand and field data of spectral reflectance,moreover, we retrieve the volume concentration of surface suspended sand with regarding the reflectance of each band in Landsat-5 image as the initial data. Based on these work, we explore the contribution of different content of suspended sand to the each image band The contribution of suspended sand content on spectrum is removed for the sake of improving the accuracy of water-depth retrieval. The main results is concluded as follows:1. The data analysis of field spectrum and the concentration of suspended sand show that the correlation between TM3 and concentration is the most significant.2. Comparing the correlation of these models such as single-band model, multi-band linear model, the ratio model and the simulating NDVI model, the result indicates the best retrieval model of the concentration of suspended sand on the surface of study area is the ratio model, the square of correlation coefficient is 0.888 and the testing mean squared error of this model is 18.21 and the mean absolute error is 14.67.3. Furthermore, we analysis the correlation between the volume concentration of suspended sand in this surface area attained by retrieval and the value of each wave band, thus establish the regression model. We find out that the variation of the suspended sand concentration strongly impacts on the spectral reflectance of TM3 and TM4. When the water depth is certain, the contribution of suspended sand to spectral reflectance of TM3 and TM4 is:ΔTM3=0.0002^*C+C²*5.2*10^-7 (1)ΔTM4 = 0.0001^*C + C²*1.3*10^-6 (2)Comparingly the volume concentration of surface suspended sand is zero, where C is the volume concentration of surface suspended sand.4. Without considering the influence of suspended sand on spectral reflectance, we compare these retrieval models where the single-factor linear model, multi-factor linear model, single-factor nonlinear model and multi-factor nonlinear model are included. It is revealed that the retrieval accuracy of non-linear multi-factor model is the best and the mean relative error is 20.18%. When we take the contribution of suspended sand to spectral reflectance into consideration, the mean relative error is 18.78%.