Study Of Antarctic Snowmelt Detection Methods Using Sentinel-1 Data

The freezing and thawing of polar ice caps is closely related to global climate change,sea level change,and ice shelf collapse.Accurate observation of polar ice cover changes has important scientific significance for understanding global change research.Sentinel-1 SAR data combined with the Google Earth Engine(GEE)remote sensing big data cloud computing platform provides an important support for polar observation.Based on the GEE platform,this paper uses the Sentinel-1 EW SAR data to study the method of detecting the freezing and thawing of the Antarctic ice sheet by taking the Larsen C ice shelf and its surrounding area of the Antarctic Peninsula as examples.The main findings obtained are as follows:(1)For the Sentinel-1 EW data provided by the GEE platform,there are unequal widths and black bars of invalid values of more than ten kilometers.This paper proposes a method based on information entropy + buffer combination,which can effectively Remove black edges and reduce their effects on subsequent image fusion and analysis.The ice formation process and distribution characteristics of different glaciers in the Antarctic ice sheet were analyzed.The variation of melting and freezing backscattering coefficient and the characteristics and distribution of the backscattering coefficient of the glacier belt were analyzed in detail by using Sentinel-1.The results were consistent with the results of the analysis of Radarsat-1 SAR data by Liu et al.(2)This paper comprehensively applies the random forest,classification regression tree and support vector machine classification method of GEE platform,and selects the massive Sentinel-1 data to carry out the experiment of the Antarctic freeze-thaw information detection method.The experimental results show that the detection accuracy based on the classification regression tree and the support vector machine is high,and the accuracy of the random forest is low.The three methods can detect the freeze-thaw information of the ice cover to some extent,but the dry snow and wet snow with a certain degree of confusion.is low.The three methods can detect the freeze-thaw information of the ice cover to some extent,but the dry snow and wet snow with a certain degree of confusion.(3)This paper firstly proposes an Antarctic ice sheet freezing and thawing detection method based on change detection + decision tree.On the GEE platform,the median value of the Sentinel-1 data in the winter months of June,July and August is the base image,and the difference between the summer and the base image of the same orbit is detected,and the threshold and elevation of the snow zone backscattering coefficient are determined.The threshold value is divided into thresholds to realize the detection of the freezing and thawing information of the Antarctic ice sheet.Use this method to detect information on the freezing and thawing of the Antarctic ice sheet from December 2015 to March 2016,October 2016 to March 2017,and from October 2017 to March 2018.For the detection results,the accuracy of the self-selected samples and automatic weather station data is used for verification.The average overall accuracy of the self-selected sample verification is 93%,the average Kappa coefficient is 0.90,and the average accuracy of the automatic weather station verification is 83.86%.The method is applied to the ice cover freezing and thawing detection in Greenland,and the verification accuracy is about 90%.Therefore,the Antarctic ice sheet freezing and thawing detection method based on change detection + decision tree has certain application prospects in ice sheet freezing and thawing detection.