The Temporary Water Bodies Extraction And Monitoring In Northeast China Through Spectral Mixture Analysis

Temporary water,usually classified as one of wetland types,arises phase-state transformation between water and wetland cycle.The regional and global variation law can be reflected by the formation,expansion,contraction and triggered eco-environment change of temporary water.Simultaneously,the temporary water,which provides favorable habitat for a large number of biological resource,exerts such a profound effect on ecological system.In addition,the temporary water,the component of global hydrologic cycle,carbon cycle,and nitrogen cycle,performs a favorable role in ecology.There is vast territory,various climate types and differential landform in Northeast China.Many kinds of waters here,like saline lakes,marshes,shallow lakes and other waters sensitive to climate change,appear in plain area on account of developed water,dense waterway and more lakes.Under the background of global climate change and the effect of human activities,some serious ecological problems have been produced by reduced wetland area,overdraft groundwater,land desertization and salinization,which leads to vulnerable eco-environment in the Northeast China.Using GOCI remote-sensing images of high time resolution,a comparative study,selected in the concentrated-distribution plain of lakes and bogs,is carried out between 2013(high flow year)and 2016(low flow year)on non-ice season in the Northeast China.Based on BP neural network method,mixed-pixels decomposition are used to extract water component information in monitoring time images and the high frequency change monitoring of temporary water component is conducted.Besides,water component datasets of time series are created to monitor every pixel score of water component,and then to extract the water pixels of changeable phase-state(temporary water body).Furthermore,this thesis compares and analyzes the spatial distribution characteristic,frequency and spatio-temporal change law of different climatic temporary water in 2013 and 2016.The results indicate that:(1)BP neural network extraction method,constructed by the mixed-pixel decomposition,has higher extraction precision than traditional methods such as band ratio method,NDWI,MNDWI,spectrum-photometric method,single-band threshold method and so on.More specifically,the extraction results of fine canal,small water body and water-land boundary are better than those used traditional methods.(2)According to calculation,the total area of temporary water is 5819.5km2 during 2013,while the total area of temporary water is 4398.55km2 in 2016.(3)Most varying phase-state temporary water has taken place 1to 3 times during May to November,and the percentage of varying temporary water pixel of total temporary water pixel is 89.13% in 2013,whereas the percentage of varying temporary water pixel of total temporary water pixel is 88.41% in 2016.(4)The number of changeable water phase-state pixels(the water appears or disappears)are summed up in every monitoring time.According to statistical results and other hydro-meteorological data like temperature and precipitation,the temporary water is sensitive to the change of climate conditions,which means the changes of climate conditions can be correspond to the changeable phase-state of temporary water.(5)The greater the proportion of pure water pixels and non-water pixels are,the more stable the changeable state is in time series,and these pure water pixels are mainly distributed in the water and land boundary.While the greater the proportion of the mixed pixels is,the more changeable the fluctuation is in time series,and these non-water pixels are mainly distributed in depression and dense waterway area.