Method Of Wetland Mapping Using MODIS Time Series

Wetlands,oceans,and forests constitute the World's three major ecosystems.Wetland mapping is of great significance to wetland research,protection and management.However,deriving accurate wetland classification maps from remote sensing imagery is a big challenge due to the strong dynamics of wetlands.Although there had been a lot of researches on wetland mapping,the current wetland-related products are commonly based on classification of single-date remote sensing image by using traditional supervised/unsupervised classification methods,and intelligent classification methods.Therefore,the resultant wetland-related products cannot fully reflect the dynamic characteristics of wetlands.In addition,the shortage of human-made wetlands in existing China wetland maps have limited its usages in wetland management and decision-making policies.Remote Sensing Time Series has shown potentials in monitoring and mapping land use/cover.(1)Reconstruction of time series data.The time series vegetation index(EVI)of Dongting Lake International Wetland was reconstructed by using the pixel reliability data provided by MODIS13Q1 as the weighted coefficient in S-G filtering process.The results show that the improved S-G filtering algorithm can effectively repair the outliers in different situations.In the process of reconstructing time series data,it is necessary to effectively remove the outliers,while preserving the original information as much as possible.According to the fast changing characteristics of rice field vegetation index,the MODIS09A1 time series data of China were reconstructed by time interpolation method.The results show that the time series interpolation algorithm based on time order can effectively repair the outliers while preserving the originally unaffected value in the sequence data.(2)In the wetland classification method,the spectral time characteristic curves of wetland types such as permanent water,flooded wetlands(mudflats),permanent swamps,seasonal swamps and paddy rice(single season / double season paddy rice)was constructed.And theses reference curves were applied to the Dongting Lake wetland classification.Long time series monitoring of wetland in Dongting Lake area was carried out(from 2001 to 2014)using the spectral matching of minimum distance(SMMD).The results showed that the overall classification accuracy of the study area was 88% in 2014,and the accuracy of seasonal swamp classification was 85%.The good classification accuracy indicated that SMMD could meet the requirement of wetland dynamic characteristic monitoring.Based on the long-term sequence,the results show that the natural wetland of Dongting Lake is fluctuate and shows a decreasing trend,and the area of permanent water is reduced.In addition,the change of spatial location of seasonal wetland and the large area of tree plantation in the marsh area have changed the original structure of Dongting Lake ecological System.And these factors has a potential impact on the functions of ecological services in Dongting Lake.According to the phenological characteristics of paddy field,the paddy field was further divided into single season paddy rice and double season paddy rice.The classification strategy of China paddy rice was constructed based on the relationship between EVI and land surface water index(LSWI)during paddy rice growth,as well as the characteristics of time series EVI curve.And the classification strategy was performed to classify and map paddy rice among China.The results showed that the overall classification accuracy of paddy rice was 91%,which indicated that the classification method based on time series data could meet the precision requirement of paddy field extraction.The results of the area show that the national single season rice field area is 20,163 thousand hectares in 2015,and the double season paddy field area is 10447 thousand hectares.The provinces among China are distributed in paddy rice(except Qinghai province)with a total area of 36610 thousand hectares.(3)The results of our study show that MODIS time series data can effectively characterize the dynamic characteristics of wetland,and the classification of seasonal wetland has derived acceptable accuracy.In the future research,more attention should be focused on the reconstruction technology of remote sensing data and classification method of high spatial and temporal resolution data and its application in remote sensing monitoring of large scale wetland.