A Study On The Method Of Wetland Information Extraction And Its Dynamic Changes Of Minjiang River Estuary

The Minjiang River is the mother river of people in the Fujian Province. The wetland in Minjiang Estuary covers six districts of Fuzhou City, Changle City, Minhou County and Lianjiang County, which is the political, economic and cultural center of Fujian Province, and it is great significant for the development in Fujian.The protection for the wetland in Minjiang estuary is a prerequisite and guarantee for economic development in the area. It is instructive to obtain the resources of status and dynamic changes of the wetland timely and accurately for the development and protection of wetland resources in this area, and its sustainable economic development. Due to the characterize with observation of large range, access to information fast, repeatable observations, the amount of information obtained by less restrictions, the remote sensing technology could provide large amount of data for the investigation of wetland resources, which becomes a new approach of wetland investigation.Combined with the previous classification system and the actual situation of the Wetland in Minjiang River Estuary, it was classified into river, beaches, paddy fields, ponds, lakes and reservoirs by using of Landsat-TM images. Taking the region that ranges Tingjiang River to the outside coastal of the Minjiang River estuary as the sample areas and with the image acquired in October31,2010, respectively, the wetland information were extracted by means of the maximum likelihood supervised classification, decision tree classification and object-oriented classification. With the comparison between the result of the three methods and visual interpretation respectively, the overall classification accuracy of the result by object-oriented classification reaches86.30%, and the Kappa Coefficient was0.8181, which is better than by the supervised classification (the overall accuracy is77.28%and the Kappa Coefficient is0.6983) and decision tree classification (the overall accuracy is83.20%and the Kappa Coefficient is0.7744). Consequently, the object-oriented classification method is used to extract the wetland information in this study.The study used a new method of water extraction, namelySRWI.And discussed the water body extraction effect of SRWI using TM imagery with atmospheric correction based on the existing research results.It showed that it could improve the water extraction accuracy of SRWI by using TM imagery with atmospheric correction. So the study used SRWI and TM imagery with atmospheric correction to extract the water body information of Minjiang River Estuary,and combined with area、shape index to realize the classification of the water.The wetland information in Minjiang River Estuary was extracted in support of ENVI EX software with the Landsat TM images imaging at low tide, and four periods images were taken into the study, which were1996,2001,2007and2010.By superimposing the results of water extraction and non-water kind of wetland, it complete the wetland information extraction of Minjiang River Estuary.Under the analysis of the wetland distribution in the four periods, it could be found that the total area of the wetlands in Minjiang River estuary was decreasing from1996to2010, and it was mainly for the paddy fields reduction. The reason of the increasing of river area and the reduction of beaches area was that tide level of images were not exactly the same. Meanwhile, there is the most dramatic change of ponds, because of a strong seasonal effect on the presence state of irrigation ponds Additionally, it was stable for the distribution of reservoir and lakes, the edges might vary little by human influence, while the total area remained stable.From the view of the changing pattern of wetlands, the stable types of wetlands were mainly river, beaches and paddy fields, which were in larger area. The regressed type of wetlands was paddy field, which were mainly in the Nantai Island, and some areas of Minhou County. The fluctuating wetland area also accounted for a large proportion that was mainly due to conversion between the paddy fields and bare land for the seasonal variations.