| Landfast Sea ice refers to the sea ice frozen on land,along the coasts of islands,in front of ice shelves,on shallows or around landing icebergs.Arctic landfast sea ice plays an important role in the Arctic Marine environment and biological system.The Laptev Sea is the most widely distributed sea area of Arctic landfast sea ice,and the change of landfast sea ice in spring is closely related to the local climate,hydrological environment and ecosystem.Therefore,it is very important to accurately monitor the temporal and spatial variations of landfast sea ice distribution in this region.Based on the Moderate Resolution Imaging Spectroradiometer(MODIS)true-color images,an automatic and computationally efficient method for identifying Arctic landfast sea ice is proposed ALFIRM-PP(a fully automatic landfast ice retrieval model-based on a deep learning model Pix2Pix),and on this basis obtained Laptev Sea spring landfast sea ice analysis products 2002-2021.Compared with the artificial visual interpretation results of MODIS true-color images,the average accuracy of landfast sea ice area extracted by ALFIRM-PP model reaches 91.4%,the average recall rate reaches 98.7%,and the average F1score reaches 94.5%.Compared with the traditional method,the model can be largely free from the interference of the floating ice area,waterway and open water area,and has a good identification effect in the thin cloud area and local thick cloud area.In addition,the analytical products produced by this model are better than existing large-scale landfast sea ice products in the Arctic in terms of spatial and temporal resolution(1.25km,7 days).This method has high automation,high computational efficiency and good recognition effect,and can be applied to the identification and monitoring of landfast sea ice in the whole Arctic region.Based on the landfast sea ice products in the Laptev Sea in the spring of 2002-2021obtained in this study,the temporal and spatial variations of the landfast sea ice in the study area were obtained and analyzed with the influencing factors of air temperature and wind vector.The statistics show that:The average area of spring landfast sea ice in Laptev Sea during 2002-2021 was 260.85×10~3km~2,with a trend of-0.67×103km~2/year(R2=0.117,P<0.01),and the maximum area showed a trend of backward movement(-0.34 weeks/year).During 2002-2021,the Laptev Sea landfast sea ice spring experienced the following process:Heavy landfast sea ice years(2002,2004,2006,2009,2010,2012)--transitional years(2005,2008,2011,2014,2016-2019,2021)--light landfast sea ice years(2003,2007,2013,2015,2020),That’s when the ice changes from heavy to light.In addition,the frequency of landfast sea ice occurrence in spring(March to May)in Laptev Sea during 2002-2021 was calculated in this study,and the stability of landfast sea ice was analyzed.The results showed that it was unstable during 2003-2009,relatively stable during 2002,2013 and 2019-2021,and stable during 2010-2012 and2014-2018.It shows that the stability of Laptev sea spring landfast sea ice is changing from unstable to stable.The correlation analysis with temperature and wind vectors that affect the variation of the landfast sea ice range in the Laptev Sea in spring shows that the influence of thermodynamics on the landfast sea ice range is limited(R=-0.02,P>0.05),and its range is mainly affected by dynamics. |
