Different Information Sources And Methods Of Extracting The Comparative Study Of The Snow

Snow is a part of hydrosphere, and also, it has a great impact on global climate. Theremote sensing satellite data, which takes wide application, good continuity and shortinformation acquisition cycle as its features, has made a great convenience for monitoringsnow accumulation continuously. It underlies the basis of studying snow range on a largescale and it has unsubstituted action on reflecting global and regional snow variation.Previously, the research of snow cover is based on field study, topographical map and aerialphotographs, it would take much time to get the figure and the work is inefficiency. Remotesensing is liable to get the spatial data, spectroscopic data and time data in inaccessible areas,it is highly effective to do the snow cover monitoring with this technique. The boundary ofsnow cap is a main factor that causes snowstorm, therefore, it is vitally necessary to surveysnow cover scope and plotting snow cover map, they also have beneficial impact onpreventing snow disaster. With the development of spatial information technology and theincreased demand of monitoring snow cover products by social economy, the efficiency andaccuracy of snow cap mapping have improved rapidly. Based on the geographicaldifferences, researches of various snow cover mapping methods in distinguished areas arebecoming more and more significant.The study takes Manasi River Basin in Middle Tianshan as the research area and thedata is from TM image and MODIS image in2006. It fetches the snow mantle informationof TM image and MODIS image by using90m DEM auxiliary visual interpretation inresearch area. Besides, the snow mantle information of TM image and MODIS image is alsocollected by utilizing ENVI, ArcGIS, MRT and adopting snow cover index, supervisedclassification and unsupervised classification. And on that basis, it compared the accuracy ofsnows which are abstracted by Visual interpreting TM snow method and other methods anddiscussed the threshold range of snow cover.Main conclusions:1. Taking the result of information about the snow which was abstracted from ManansiRiver Basin by TM visual interpreting method, it fingered out that the basin snow area isabout27.26%of the gross area, the lowest altitude of the snow basin is2939m, theaccumulated snow concentrated in southern area of the mountain and there is a littleaccumulated snow in the middle areas.2. Neither the accuracy comparison between supervised data classification and unsupervised data classification on the basis of MODIS, nor the response of NDSI snow capindex is idealized, which shows that the500m resolution data in research area could notanalysis the snow cover information accurately. The maximum likelihood based on TM datasupervised classification and NDSI snow cover index method could reflect the snow coverinformation in research area better, it is a valid supplement of Visual Interpretation.3. Through the analysis of five sample areas in research area, the result shows that thelandform and the vegetation have a great influence on calculating the interpretation accuracyof Snow boundary. In the Manasi River Basin, the misjudgment of mountain shadow due tothe landform accounted the70%portion of all misjudgment. This problem runs through thesupervised classification and unsupervised classification. In addition, the vegetation alsoimpact Snow recognition largely.4. When using high resolution images to definite the snow boundary, the influence ofdifferent methods is greater than using lower resolution images. The result came out byabstracting snow cover by MODIS visual interpreting method is quite different from theresult came out by abstracting snow cover by Markov distance method. By means of otheralgorithms, the results are similar with the result came out by visual interpreting method,differing from2.49%-7.79%. The accuracy of the result came out by unsupervisedclassification method ISODATA、K-Mean to abstract snow cover is slightly lower than theresult came out by supervised classification method, being less than5%of the result cameout by MODIS visual interpreting method. In specific circumstances of the landform andvegetation in research area, it is better to use the supervised and unsupervised methodssupported by ENVI and ArcGIS when it is necessary to use MODIS visual interpretingmethod, otherwise, in the circumstances of being pressed by time and without requiring theaccuracy of the result, to choose unsupervised method to replace the MODIS visualinterpreting method is one of the reasonable way.