Polar Sea Ice Concentration Retrieval Based On FY-3C Microwave Radiation Imager Data

Polar sea ice affects atmospheric and ocean circulation and it plays an important role in global climate change.The sea ice concentration is one of the important parameters to characterize the temporal and spatial variation of sea ice.Neural networks have made impressive progress in many areas,but they are less used in the polar sea ice field.The retrieval algorithm of sea ice concentration based on brightness temperature data of FY-3C microwave radiation imager in the polar region was studied.For one thing,After the time-space matching and linear regression,FY-3C microwave radiometers brightness temperature data was corrected.The atmospheric effects were reduced using two weather filters and sea ice mask.A minimum ice concentration array used in the procedure reduced the land-to-ocean spillover effect.The sea ice concentration product was validated by calculating Arctic and Antarctic sea ice extent and area in 2016-2017.The sea ice extent and area trends of this two years were basically consistent with the NSIDC product,with an average difference of 3%.For another,Using the FY-3C brightness temperature data of different channels in 2016 as the training data,the NSIDC products are used as labels,and the optimal model is trained through the combination of different channels data and networks selection.The 8-12-1 neural network trained by the Levenberg-Marquardt algorithm is best suited for retrieval of sea ice concentration.The 23.8GHz channel is effective for reducing weather effects,and location information can reduce the land-to-ocean spillover effect.The sea ice concentration product was validated by calculating Arctic and Antarctic sea ice extent and area in 2017.The Mean Square Error of the Arctic sea ice extent and area is 0.12×10~6km~2 and 0.16×10~6km~2,respectively,and the correlation is good.This research laid the foundation for the release of the polar sea ice concentration business products of China's autonomous satellites,and the products guarantee the continuity of crucial polar sea ice record that might soon be interrupted.The neural network model for sea ice concentration retrieval is given to provide reference for the application of ocean big data...