Remote Sensing Identification Of Ulva Prolifra And Sargassum And Evolution Of Green Tide In The Yellow Sea And The East China Sea

During 2007 and 2013, large scale of Ulva prolifra green tide had been burst out in the Yellow Sea for 7 successive years, and it becomes one of the main marine ecological disasters in this sea area. Recent years, floating Sargassum has expanded in the East China Sea and the Yellow Sea. The Sargassum appeared in the sea area of Ulva prolifra green tide, and grew mixed with it, which affected the remote sensing monitoring of green tide. This study researched on the spectrum measuring and analysis of seawater, Ulva prolifra and Sargassum both in field and laboratory, to detect the difference of their spectral characteristics, and applied to the identification method research of Ulva prolifra and Sargassum from hyperspectral and multispectral images. The research analyzed the temporal and spatial characteristics of green tide from the time series between 2007 and 2013, divided the growth and disappearing cycle of green tide into 4 phases, and researched the environmental characteristics and discriminant method in each phase. The main researches and conclusions can be summarized as:(1) The seawater, Ulva prolifra and Sargassum spectrum measuring in field and laboratory showed that they have obvious difference in spectral characteristics. In the visible band, seawater reflectance was generally lower than that of Ulva prolifra and Sargassum. Ulva prolifra had a reflectance peak in the band of 510-580nm, while Sargassum’s reflectance peak exists in the band of 580-650nm, and there were differences between the spectral characteristics of Ulva prolifra and Sargassum in 510-650nm. In the near infrared band, reflectance of Ulva prolifra and Sargassum were high, while Seawater reflectance was generally very low. The spectral characteristics of seawater, Ulva prolifra and Sargassum had the greatest difference in near-infrared band.With the increase of algae suspended depth, surface spectral quantity value decreased, the red to near infrared spectral value decreased most obviously.(2) Two HSI hyperspectral satellite images were processed, such as atmospheric correction, and analyzed the spectrum characteristics difference of Ulva prolifra and Sargassum on the images. With the remote sensing Supervised classification, identification and information extraction of macro algaes can achieved using unique and continuous reflection spectra of macro algaes.(3) For the identification of Ulva prolifra and Sargassum of multi spectral satellite images, this paper adopted two steps:first separate seawater and macro algaes, then identifies Ulva prolifra and Sargassum. As to the second step, this research adopted an algorithm, i.e. Green Algae Index (GAI), whose formula was:. . GAI method performs well in Ulva prolifra and Sargassum identification on HJ1A/1B and Landsat 8 satellite images, especially identified the two algaes in mixed-growing condition.(4) Many HJ1A/1B and MODIS satellite images containing green tide from 2007 to 2013 were collected to conduct green tide satellite remote sensing information extraction, and researched the temporal and spatial characteristics of green tide. The results of remote sensing show that, during 2007 and 2013, green tide took place between May and September, the lasting period had some difference in each year. Green tide had a whole emergence, development and disappearing cycle each year. The annual and interannual coverage and distribution areas of green tide had a large amplitude of variation. Green tide distributed mainly at the sea area from Yancheng Jiangsu to Rushan Shandong, and it had one complete drift route from Yancheng Jiangsu to the south coast of Shandong peninsula at least. The green tide drift routed incline to the west during 2007,2008,2010,2011 and 2013.(5) The growth and disappearing of green tide in the Yellow Sea was divided into 4 phases, i.e. emerge, outbreak, maintain, and disappear. This study compared 12 environmental elements in each phase of the green tide from 2008 to 2013.(6) From the maritime environment monitoring datum of green tide,208 training samples and 88 testing samples were selected, and discriminant analysis method and support vector machine method (S VM) were applied in each phase of green tide based on environmental elements. The results showed that discriminant analysis and SVM perform well in each phase of green tide based on environmental elements, and they were optional methods. Compared with discriminant analysis, SVM method had more advantages in the discrimination in each phase of green tide, the discrimination accuracy of back substitution test was improved from 92.8% to 99.5%, and the discrimination accuracy of direct test was improved from 92.0% to 96.6%.This study drew some meaningful conclusions, provided a suitable method for the identification of Ulva prolifra and Sargassum on moderate and high resolution multispectral satellite images, deepened on the understanding of Chinese green tide disaster regularity, and laid the foundation for the following green tide forecasting.