Study On Muddy Coastal Water Body Index Based On Modis Data

With the economic development and urbanization of the estuary delta.Water bodies in coastal areas are facing threats such as continued area reduction and water pollution.The change in water area and water quality caused by the pursuit of economic development has become a problem that local land planning departments need to solve.Remote sensing technology provides a convenient and rapid technical means for monitoring water changes in the estuary delta,and water index method is widely used in water information extraction and area change monitoring.As an important type of estuary delta,the muddy coast has a complex and unique water environment,and most of the existing water index objects are inland water.Therefore,it is of great scientific significance to deeply study the application of water index in the muddy coastal area.This paper took the Yellow River Delta coast as the study area,and used the MODIS and Landsat remote sensing data from 2000 to 2015.Studied the spectral characteristics of the main land cover types in the tidal flat area of the muddy coast.Compared the water extraction accuracy of six water indexes(NDWI,MNDWI,AWEI_nsh,AWEI_sh,TCW,WI₂₀₁₅)by ROC model.Analyzed the effect of main land cover types on the extraction of water index in muddy coastal areas.On this basis,the advantages and disadvantages of empirical value method,trial number method and automated water index construction method are analyzed.For the MODIS data dimension is higher,and it have redundancy information between the bands.Three data dimensionality reduction methods（PCA,BTBC,SDA）were used to study the optimal band combination for extracting muddy coastal waters,and the coefficients of each band were determined by linear Support Vector Machine（SVM）.Proposed a Muddy Coastal Water Index（MCWI）that based on data dimensionality reduction and automation build method.The water separation performance of pure pixels for MCWI and other water indices(NDWI,MNDWI_B6,MNDWI_B7,AWEI_nsh,AWEI_sh,TCW1,TCW2,TCW3,WI₂₀₁₅,MBWI)was compared and analyzed by t-test method,and calculated the water extraction accuracy and error of pure pixels and mixed pixels by ROC model.The results show that the MCWI has a slight advantage over the other water indexes in the construction method,water separation performance and extraction accuracy,and improves the accuracy of water extraction.It can provide reference for the study of water distribution and variation characteristics in the muddy coastal area.The main conclusions are as follows:（1）The spectral characteristics of the main land cover types（tidal flat water,tidal flat soil and tidal flat vegetation）in the muddy coastal tidal flat area are significantly different from those in the inland areas.The reflectance of tidal flat water,soil and vegetation in the short-wave infrared is very significant.The reflectance difference in the near-infrared band is significant,and the difference in reflectance between the blue and red bands is significantly higher than that of the green band.The green band distinguishes between muddy coastal water bodies and non-water bodies.It is more difficult for the green band to distinguish between muddy coastal water and non water.（2）Due to the unique geographical environment and water features of the muddy coastal area,the water extraction accuracy of the six water indices is slightly different.The water extraction result of AWEI_nsh is better than other water indices.WI₂₀₁₅ is second,and the water extraction results of NDWI,AWEI_sh and TCW are relatively close.The water extraction effect of MNDWI is poor.The mapping accuracy of seawater extracted by each water body index is higher than 90%.The extraction resluts of land water is general,and the drawing accuracy is lower than80%.NDWI has poor recognition ability for tidal flat water,and the drawing accuracy is lower than other water body indexes.The different water indexes have high omission error of land water,and the omission errors of seawater and tidal flat water is low.The MNDWI have the highest omission error of seawater.The influence of the tidal flat soil on the water extraction is the greatest,and the second is the cultivated soil.The sparse vegetation,luxuriant vegetation,and built-up area have the least impact.（3）According to the results of three kinds of data dimensionality reduction methods,the optimal bands for extracting muddy coastal water by MODIS data are red band（B1）and blue band（B3）in visible band,B6and B7 in short wave infrared band,among which B6 has the highest identification capacity for muddy coastal water.（4）The Muddy Coastal Water Index（MCWI）based on data dimensionality reduction has good separation performance,and the t value of water and other land cover types are higher than 8.8,and the highest is 17.86.The extraction accuracy of pure pixels and mixed pixels have slightly advantages compared with other water body indexes.The water body extraction accuracy of pure pixels is higher than 97.5%,and the omission error,commission error and total error are 2.28%.,2.48%and 4.76%.The water body extraction accuracy of the three types of mix pixels is higher than 85%,and the highest is 95.6%.The the total error of the three types of mixed pixels is lower than 25.8%,and the lowest is9.38%.