Analyzing Temporal And Spatial Variation Of Arctic Melt Ponds And Sea Ice Fractions

Sea ice is a hot topic in polar research at present.As an important parameter in sea ice,melt ponds has a significant impact on the change of sea ice,Therefore,this paper uses the Arctic high-resolution measured images of sea ice and melt ponds combined with MODIS reflectance products to retrieve the Arctic melt ponds and sea ice fractions.The spatialtemporal characteristics of the melt ponds and sea ice fraction and their relationship with the sea ice extent in September are analyzed.Using the high-resolution measured images,2786 sets of melt ponds fraction,ice fraction and open water fraction corresponding to the MODIS reflectivity product grid were extracted.Use Stacked Autoencoder to randomly extract 70% of the melt ponds fraction,ice fraction,and open water fraction as output parameters,and train the corresponding surface reflectance as input parameters.The remaining 30% is used to evaluate the model Accuracy.The results show that the correlation coefficient between the melt ponds fraction predicted by the Stacked Autoencoder and the measured data is 0.8,and the root mean square error is 5.4%.The correlation coefficient between ice fraction and measured data is 0.93,and the root mean square error is 9.2%.The evolution of the melt ponds fraction showed relatively asymmetric growth and decline.From May to mid-June,melt ponds fraction showed a strong growth trend.Since then,the growth rate of melt ponds fraction has slowed down and peaked on July 4.During the period from June 26 to July 4,the rate of melt ponds fraction was stable.After July 4,the melt ponds fraction began to decrease.After July 4,the relative melt ponds fraction continued to increase,and a higher relative melt ponds fraction appeared on July 11.After July 11,the relative melt ponds fraction decreased.From May 8 to August 5,2004 to 2017,the interannual change in melt ponds fraction showed a significant downward trend,and it passed the 95%significance test.During the period from May 8 to August 5,the sea ice coverage showed a cyclic decrease from low latitudes to high latitudes.And the interannual change of sea ice area showed a significant downward trend,which passed the 95% significance test,while the downward trend of the sea ice extent did not pass the 95% significance test.By calculating the melt ponds fraction distribution characteristics related to the September sea ice extent,it was found that the melt ponds fraction in areas over 95% significance test had the closest relationship with the September sea ice extent during the period from May 8 to August 5,and the prediction effect of the September sea ice extent was the best.The melt ponds fraction during this period is related to the negative phase of the Arctic Oscillation and the air temperature in the Chukchi Sea and the Eastern Siberian sea,which will affect the September sea ice extent.