Abstract: There are some problems of missing data or discontinuity in long-time series of soil relative humidity data obtained from the observational stations during the spring sowing in Liaoning province. Using the soil relative humidity records from Haicheng observational station as a case study, the relationship between soil relative humidity (0-20 cm) and meteorological factors, relative humidity from the surrounding weather stations was analyzed. A statistical regression model for the construction of monthly soil relative humidity during the spring sowing period was established, and missing soil relative humidity data were simulated using the above method. Meanwhile, the monthly soil relative humidity data during the spring sowing period from 1981 to 2012 in 20 observational stations in Liaoning province were rebuilt. The results indicate that soil relative humidity in Haicheng observational station is significantly correlated with local rainfall and the last precipitation before freezing up in autumn, and the multiple correlation coefficients are over 0.60 and 0.30, respectively. The correlation coefficient of soil relative humidity between Haicheng and Benxi observational stations is over 0.40. A regression model on soil relative humidity in April and May is established by the above three elements, and its multiple correlation coefficients are 0.79 and 0.77, respectively. The mean relative error is 2.6 % between simulated and observational values. The multiple correlation coefficient of the model established by the other loss data in Liaoning province is over 0.50, and the fitting precision is higher than 85 %. The mean relative error is within 15% between simulated and observed values. The rebuilding of monthly soil relative humidity during spring sowing for 20 stations in Liaoning province during 1981 to 2012 is reasonable.

Key words: Soil relative humidity, Spring sowing period, Regression model, Data rebuilding