In the ground meteorological observation, it generally takes artificial observation to judge whether icing phenomenon occurs. In order to improve the efficiency of icing observation, automation of icing observation has become an inevitable trend. This paper introduces a new automatic observation device for icing through detecting the sudden rise process of temperature that is caused by latent heat release in the icing process, and judges the occurrence of icing phenomenon to realize the automation of icing observation. Combined with temperature observation data, characteristics of each stage in the icing process are analyzed as well. Compared with the artificial icing observation results, it is concluded that the automation observation device for icing can accurately observe icing phenomenon.