Extracting Algal Blooms Information Of Taihu Lake Using Chinese Satellite Imagery

Taihu Lake,located in the economic zone of Yangtze River Delta,is important to the local development.The eutrophication and water pollution problem of Taihu lake is becoming incresingly serious due to the rapid economic development and industrialization improvement.The remote sensing technology can efficiently monitor macroscopic land cover information with low-cost,which is of great significance to the monitoring and management of blue-green algal blooms in lake.With rapid development of the resources and environment satellites in China,plentiful free data provides a wider support for the remote sensing monitoring of blue-green algal blooms in the lake.According to previous researches,the main problems in remote sensing monitoring of algal blooms are:(1)The automatic masking of the lake boundary;(2)The distinguishment between blue-green algal blooms and aquatic plants;(3)The selection of the extraction threshold.Taking the algal blooms extraction in Taihu Lake as the research target,and based on the domestic satellite images such as HJ-1-CCD,HJ-1-HSI,CBERS-2/02b-CCD,etc.,this study built a new decision tree model for algal blooms extraction with one image,and validate the method with 15 images.The images dated from April to September,and November to December in the year of 2003 to 2013,including 14 images of Taihu lake and 1 image of Chaohu lake as to the coverage area,10 HJ-1-CCD images,2 HJ-1-HSI images,and 3 CBERS-2/02b-CCD images as to the sensors,among which 14 images have clear blue-green algal blooms and one has none.The extraction algorithm of blue-green algal blooms was built using HJ-1-CCD image on August 3rd,2010 in Taihu lake,and validated and analyzed using other images.The main results and conclusions are as follows:(1)The distinction between blue-green algal blooms and backgrounds.The band ratio of SWIR and Green band(R25)was used to identify the lake body from the land.By comparing the performance of typical vegetation indices in algal blooms extraction,including RVI,DVI and NDVI,NDVI was selected as the best index to distinct between blue-green algal blooms from clear water.The transparency index NSDD was built to distinct the algal blooms from aquatic plants.(2)Construction and validation of the determine tree extraction model of blue-green algal bloomsCombined with the index between algal blooms with land,clear water and aquatic plants,the decision tree extraction model of algal blooms was built based on the junction of R25,NDVI,NSDD,which initially solved the confusion problem between the algal blooms and aquatic plants.Compared with the NDVI threshold method,Chla inversion model and the FAI method,the decision tree extraction model proposed in this study can eliminate the influence of land and aquatic plants.The precision was calculated and evualted based on the result of visual interpretation,the accuracy of extraction is up to 94.12%,the missing rate is 5.88%,and the error rate is 0.53%,showing that its accuracy is much higher than other methods.(3)The rapid extraction of blue-green algal blooms with artificial auxiliaryBased on the extraction results of algal blooms from images,this study initially established statistical threshold suitble to the extraction model of HJ-1-CCD,with the initial threshold of R25,NDVI,NSDD being 2.75,-0.22,0.78,and the range of threshold being 2.5?2.7,-0.25?-0.13,0.57?0.87.Designed the prototype program extracting of blue-green algal blooms with artificial auxiliary,which can quickly extract the algal blooms information by making small adjustments within the threshold range,thus increased the extraction efficiency of blooms.