STUDY ON THE INVERSION MODEL OF THE SUSPENDED SEDIMENT IN THE MIDDLE YANGTZE RIVER BASED ON REMOTE SENSING TECHNOLOGY
Cheng-da LIN,Bin ZHOU,Qi MA,Meng-xing JIANG
DOI: https://doi.org/10.11870/cjlyzyyhj201408011
2014-01-01
Abstract:Quantitative research on suspended sediments plays an important role in the investigation on the Yangtze River,including water quality,riparian landscapes,and the periphery ecological environment around the river.Meanwhile,it is also of a vital significance for waterway construction and waterway safety.This paper utilizes the quantitative remote sensing inversion of suspended sediment methodology to compute the quantitative concentration of suspended sediment in the Wuhan Section of the Yangtze River.Through a series of experiments,this methodology has been thoroughly studied.The paper was organized as follows.First,the study area dataset in the Wuhan Section of the Yangtze River was introduced.14ETM+images of area acquired at different times were mainly used,combined with the field sampling data of suspended sediment concentration.The ETM+SLC-ON images before 2003 were mainly used for image recovery,and the SCL-OFF images after 2012 were used for spectral reflectance computation.20 filed samples were acquired from 2012,March to 2013 July.12samples were used for model-based inversion,and 8samples were used for model validation.Then,several procedures were implemented on the ETM+images,including image recovery,atmosphere correction,geometric correction,and Gain DV value extraction.Adaptive local regression(ALR)matching algorithm is utilized on SLC-Off images to recover the damaged images.After that,atmospheric and geometric correction was implemented on the images.Then,the DN values of the field sampling area were extracted,used for the spectral reflectance computation,combined with the radiometric calibration parameters recorded in the ETM+images.To choose the most effective band of ETM images that is most sensitive to the suspended sediment concentration,regression analysis was implemented among the suspended sediment concentration and the radiometric reflectance.It was concluded that Band 3is the most appropriate one that reflects the suspended sediment concentration.The key of suspended sediment concentration remote sensing quantitative retrieval is used to build the relations between the spectral reflectance of water and the suspended sediment concentration,which can be modeled as linear,logarithmic,and first-order differential models.Gaussian and logarithmic models were tested in our research,and the Gaussian model is testified as more appropriate for suspended sediment concentration in the research area.To testify the accuracy of the results,8set of measured data were utilized.The retrieval accuracy is 88.91%,the mean error is 0.110 9,and the maximum absolute error is 0.213kg/L.In this paper,suspended sediment field survey data and synchronized landsat7ETM+ remotes sensing images were used for quantitative retrieval of suspended sediment modeling studies in Wuhan section of the Yangtze River.The experimental results demonstrate that,adaptive local regression(ALR)method is effective for recovery of ETM+ SLC-off images.In the process of constructing the relations between the spectral reflectance of the water and the suspended sediment concentration,single band is better,and the retrieval accuracy is higher by using Gaussian model.The correlation coefficient is 0.93 and the accuracy percentage reaches 88.91%.This can meet the actual application requirements.The quantitative inversion model can be used for sediment concentration monitoring in wide area and provide a convenient way for predicting and monitoring of the suspended sediment concentration.