Water depth retrieval models with remote sensing sediment parameter

Abstract

针对悬浮泥沙影响水体遥感测深精度的问题,选择长江口南港至南槽为研究区,通过对遥感测深方法研究,结合悬浮泥沙光谱特性分析,把泥沙因子引入到水体遥感测深反演模型中,研究表明:1)单因子非线性模型中,指数模型对0-2m的水深反演效果较好,对数模型对2-7m的水深反演较好,二次回归模型对7-14m的水深反演效果较好;2)建立的BP人工神经网络水深反演模型综合了多个波段具有的水深信息,模型的反演效果好于单因子非线性模型;3)实验构建的泥沙遥感参数综合了不同波段具有的悬沙信息,削弱了叶绿素和外界环境条件对泥沙信息的干扰,可较好地反映悬沙浓度变化特征;4)建立的BP人工神经网络泥沙因子水深反演模型削弱了悬浮泥沙对遥感测深的影响,模型实际反演能力明显优于单因子非线性模型和多因子BP人工神经网络水深反演模型.;In order to solve the problem of suspended sediment influencing water depth measurements by remote sensing technology, the remote sensing sediment parameter was constructed by experimental methods and introduced into the research of water depth retrieval methods to weaken the effect of suspended sediment to remote sensing bathymetry. The study area is the South Channel to the South Passage in the Yangtze River Estuary. Results showed that: 1) the exponential models can estimate the water depth of less than 2m well, and the logarithm models can retrieve water depths from 2m to 7m accurately and the quadratic regression models can calculate the water depths from 7m to 14m better in the single factor nonlinear inversion models of water depth; 2) the multiple factor water depth inversion model of BP artificial neural network (BPANN) integrate water depth information of multiple wavelengths, and it performs the stronger capability to retrieve the water depths than single factor nonlinear models; 3)remote sensing sediment parameter constructed by experimental methods integrate suspended sediment information of multiple wavelengths, and weakens the effect of chlorophyll and environmental factors, and it can characterize the variations of suspended sediment concentration (SSC) better; 4) the sediment parameter water depth retrieval model of BP artificial neural network (SPBPANN) weakens the influence of SSC to remote sensing bathymetry, it performs the strongest capability than single factor nonlinear models and BPANN model.