关键词: 雪面温度预报模型, 神经网络, 小海坨

Abstract:
Abstract: Based on the hourly meteorological observations at Erhaituo station with high altitude and Changchonggou station with low altitude in the Xiaohaituo mountain area of Yanqing, Beijing from October 2019 to March 2020, the characteristics of snow surface temperature and its correlation with meteorological factors were analyzed. The forecast models of snow surface temperature of two stations were established and tested by using neural networks and stepwise regression methods. The results are as follows: (1) The hourly variation of snow surface temperature during snow cover period in the Xiaohaituo mountain area was obviously stronger than that of air temperature, air temperature and total solar radiation were positively correlated with snow surface temperature and were the main factors causing the change of snow surface temperature. (2) The performance of snow surface temperature forecast model based on neural network method was superior to the one based on stepwise regression method, the model forecast effect at low altitude station was better than that at high altitude station, and the model forecast effect in the daytime was better than that during nighttime. (3) The model established by distinguishing daytime and nighttime was more suitable for the low altitude station.

Key words: Key words: snow surface temperature forecast model, neural network, Xiaohaituo