Abstract: Sea fog in polar regions poses a challenge to the research on polar science and sea ice. However, due to the lack of relevant cloud monitoring data in the polar region, the research on sea fog in polar regions is still relatively scare. Based on the CALIOP sensor′s ability to observe cloud information in the vertical direction, the MODIS medium resolution imaging spectrometer with plesiochronous observation is used to analyze cloud information in the Arctic region. Firstly, the deep neural network model is applied to invert the cloud top height. Then, according to the inverted cloud top height, whether it is sea fog can be ascertained. Furthermore, the influence of different wavebands on the inversion results is also analyzed. The results show that the average absolute error of the cloud top height inverted by the deep neural network is 1774.280 m lower than that of the traditional method, indicating that using deep neural network model can invert cloud top height better and more accurately, which can improve the detection accuracy of sea fog.

Key words: cloud-aerosol lidar with orthogonal polarization, moderate resolution imaging spectroradiometer; fog detection, cloud top height, deep learning