Analysis Of Different Turbid Inland Water Mapping Using Landsat Images

This study focus on different turbid water from upstream to downstream in Xiaolangdi Reservoir in the Yellow River Basin.First,according to the characteristics of spectral reflectance and the water colour between of different turbid water in different bands,clear water,sub-turbid water and turbid water were defined respectively.These three kinds of water bodies were classified by using three kinds of methods.Second,according to the difference of spectral reflectance between different turbid water in different bands,nine water indexes are used to make a comparative study about different turbid water mapping in Xiaolangdi Reservoir.In order to evaluate the ability of these water indexes to enhance water or suppress non-water,the positive value in water and the negative value in non-water were calculated in each NDWI image.To analyse the effect of threshold in water mapping by using different water indexes,Otsu threshold and 0 were used to separate the water and non-water in each NDWI image.Third,using Gram-Schmidt Pan Sharpening(GS)model,the 15m-resulution pan image and multispectral image were fused to new image which was used for accuracy assessment.Finally,three new water indexes were developed to map different turbid water bodies.The results of this study can be concluded as follows:First,although several thresholds need to be used,both NDWI method and classification tree can serve the purpose of classification of water turbidity.The NDWI method can not only classify water color,but also suppress non-water.The result of classification of Fisher'classification tree is better than that of others,while non-water bodies have to be covered.Second,according to the positive value and the accuracy assessment of each NDWI image in different turbid water,we found that the NDWIblue was more suitable to enhance and extract clear water mapping;NDWIgreen was more suitable to enhance sub-turbid water;NDWIred was more suitable to enhance and extract turbid water.In addition,we found that NDWINIR was less capable of extracting turbid water.Third,the best threshold was influenced by three infrared bands in NDWIs.The Otsu threshold was lower than 0 in NDWINIR image and higher than 0 in NDWISWIR1 image.In addition,the Otsu threshold of NDWISWIR1 was near 0.Four,the accuracy of water mapping decreased with the decreasing resolution of image.And in most cases it increased with the increasing positive value.Five,three new water indexes(NDWIclear,NDWIsub-turbid,NDWIturbid)were developed to extract clear water,sub-turbid water and turbid water respectively.