Study On Arctic Sea Ice Concentration Obtained By Classification Of SAR Imagery

With global warming, the Arctic sea ice is also undergoing an unprecedented rapid change. The sea ice extent of summer Arctic continuously breaks the recorded low since the satellite remote sensing became available. And the recorded low of 3.41x106 km2 in September 16 of 2012 is even the lowest Arctic sea ice extent in September in past 1450 years. Under the background like this, the important impact of Arctic sea ice change on the climate and ultimate weather events in China has gradually emerged, and the open of new Arctic passages also provides opportunities for China to extricate ourselves from a single way of energy transportation and to improve the energy safety of the nation. Therefore, studies on Arctic sea ice have an important economic value and strategic significance to China.Observations on Arctic sea ice using the Synthetic Aperture Radar (SAR) have advantages of full daytime, full weather conditions and high spatial resolution, and thus become the most important method of sea ice remote sensing currently. In this study, four SAR images with single-polarization from the satellite Radarsat-2 of Canada are employed to analyze the sea ice concentration (SIC) on the key region of Northeast Passage of Arctic where the East Siberian Sea and the Chukchi Sea are joined together. The SIC retrieved by the brightness temperature data of the microwave radiometer onboard HY-2 satellite, and the SIC product of the Special Sensor Microwave Imager/Sounder (SSMIS) released by the National Snow and Ice Data Center (NSIDC) of USA are also used for comparisons.The results reveals that the Refined Lee filtering method during pre-process of SAR image can effectively suppress the speckle noise, and not break the complete outlines of the ice edge on image. The K-means clustering algorithm can realize the automatic classification of SAR image, and both the automatic degree and the classification result are better than that of the ISODATA algorithm. Comparisons among SIC of SAR, HY-2 and SSMIS show that the standard deviations of the difference between the different data are similar. The mean of difference between HY-2 and SSMIS is smallest and that between SAR and SSMIS is largest. The correlation coefficient between HY-2 and SSMIS is highest, and the correlation coefficient between SAR and SSMIS agrees with that between SAR and HY-2. These differences mainly come from the different spatial resolution of these satellite sensors, and the impact of open ocean waves and melt ponds on summer sea ice on the back-scattering features of SAR image, and the influence of large amount of water vapor in the marginal ice zone on the retrieval algorithm of HY-2. However, from the overall trend of view, the distributions of SIC from SAR, HY-2 and SSMIS are similar in trend, and the sea ice extent also agrees well with each other, implying the feasibility of the present steps and methods to process the SAR images and finally obtain the SIC, and the reliability of the SIC retrieval algorithm from the brightness temperature data of the microwave radiometer onboard HY-2. This study provides important bases for monitoring the sea ice conditions on the Arctic passages using high-resolution satellite remote sensing data, to direct the safety of navigation during the future Arctic shipping of China, and also has positive meanings for releasing routine Arctic sea ice product using Chinese satellite resource, to improve the initiative of China in international Arctic sea ice research.