Based on the observations of three different kinds of underlying surface temperatures and the automatic meteorological sensor measurements during January and February of 2009 winter in Enshi and Jinsha, Hubei Province, this paper studies the relationship of the three kinds of underlying surface temperatures and the air temperature by establishing multi〖HJ*3/7〗ple regressions accordingly with the statistic analysis method as well as the meteorological conditions of icy underlying surface. The results show that the range of soil and concrete road surface temperature is less than asphalt, and the former two variation trends are highly consistent. Air temperature is one of the most important factors influencing the underlying surface temperature. The underlying surface temperature in different weather conditions predicted by multiple regression are well validated, with 0.9 above correlated with observations except for 0.88 of asphalt underlying surface in Jinsha. Besides, the revealed close relationship between the low temperature around 0 ℃ on the underlying surface and the wet state for the road ice development will benefit the meteorological departments for their practical operation to some extent.