Abstract: Real-time and accurate prediction of moisture content is to carry out irrigation forecasts, and to achieve precise management of farmland water, which is an important measure to improve water efficiency. Based on the principle of water balance in the root zone (0−60cm soil layer), the crop transpiration and water flux at the lower interface of the root zone are linearized by using the Taylor series. On this basis, a dynamic soil moisture prediction model was constructed with the real-time average soil moisture content of the root zone as an independent variable. The real-time monitoring data (soil moisture content at 30cm and 60cm below the ground surface) of the wireless soil moisture monitoring system (including three monitoring points) in Xilv Village, Wuqing District, Tianjin City are used, and 5 days, 10 days, 15 days and 20 days are selected as the modeling series length respectively, and regression analysis is performed to determine the model parameters. The prediction accuracy of soil moisture was analyzed, using the two forecast periods of 10 days and 15 days. The results showed that: (1)the real-time prediction model fits well, and the deterministic coefficients under the condition of the three modeling series length can above 0.80 (the number of samples are all greater than 550).(2) The relative error of 15 days modeling series is the smallest.(3) Under the conditions of 15 days modeling series length, 15 days prediction period, and 10% relative error limit value, the moisture prediction pass rates of the three monitoring points reached 98%, 100% and 89%, respectively. It can be seen that the real-time moisture prediction model proposed by the research has high prediction accuracy, which is convenient for modeling and analysis, and provides a new method for soil moisture prediction.

Key words: 土壤墒情, 动态建模, 传感器, 水量平衡, 墒情预测