Abstract: Wind erosion has been a serious problem in the northern and northwestern China. Wind speed is one of the driving factors for the wind erosion and plays an important role. The wind erosion models require high resolution wind data, such as hourly data, to estimate the event wind erosion amount. The wind speed data from the metrological stations is difficult to meet the requirements from wind erosion models on temporal and spatial coverage. The stochastic simulation of wind speed is one of the alternatives to solve the problem. Harmonic analysis of hourly surface wind speed was performed to study the temporal and spatial characteristics of diurnal surface wind speed change based on the data from 239 meteorological stations in the north and northwest of China. F-test was used to detect significant harmonics which have significant contribution to diurnal wind speed change. The main conclusions are as follows:For 204 stations(about 85.3% of the total), only the first harmonic had significant contribution to the diurnal wind speed change, which indicated the wind speed cycle of diurnal variation is 24 h. For 34 stations(about 14.2% of the total), the first two harmonics had significant contribution. The explained variations of the first harmonic is larger than the second harmonic, which suggested the main-cycle of diurnal variation is 24 h and the sub-cycle is 12 h for these stations. The daily mean wind speed ranged from 0.96 to 8.36 m·s^-1 with an average value of 2.42 m·s^-1. The stations with recorded wind speed greater than 3 m·s^-1 were located in the northern Mongolian Plateau, the Hexi Corridor in Gansu Province, flat areas in the Tibetan Plateau and the northeastern Xinjiang Province. The seasonal mean wind speed had a descending order from spring to winter to summer and to autumn. The characteristics of the amplitude and phase for the winds recorded by the 204 stations with only the first harmonic being significant were analyzed. The amplitude ranged from 0.28 to 3.28 m·s^-1. The phase ranged from -1.56 to 4.67(with a concentration on the range between 3.21 and 4.67), which indicated the daily maximum wind speed appearing during the time period from 13:00 PM to 18:00 PM. This study on diurnal variation can provide theory basis for the stochastic simulation of hourly wind speed, which in turn can provide hourly wind speed data to improve the estimation of wind erosion.

Key words: wind erosion, hourly wind speed, diurnal variation, harmonic analysis, temporal and spatialvariation