Research On Forecast Model Of Field Soil Temperature And Moisture

Water is the source of life,and people cannot live without water.The lack of water resources is a serious problem facing today’s society.Hebei Province is a large agricultural province with a large amount of agricultural water,and the serious groundwater exploitation has led to a serious shortage of water resources in Hebei Province.Soil temperature affects many physical,chemical,and biological processes in the soil.It is important to study soil temperature changes and predict soil temperature changes for the growth and development of crops.This paper studied the real-time monitoring of soil temperature changes in 2-3 years of wheat-corn annual planting patterns in 6 typical areas of Hebei Province,Tangshan City,Baoding City,Shijiazhuang City,Hengshui City,Cangzhou City and Xingtai City.The differences in the temperature changes of different soil layers of 10-30 cm,the daily average soil temperature changes in large periods of the year and the time average soil temperature changes in small periods are studied respectively.And take Gaocheng District,Shijiazhuang as an example,to predict soil temperature and soil moisture.(1)Through the study of soil temperature changes in different soil layers in 6 typical wheat-corn annual rotation planting pattern regions in Hebei Province,it is found that the laws are basically similar,and the change laws are similar to a sinusoidal curve.The overall change trend of soil temperature in the six regions showed a gradual decrease in soil temperature from south to north.The highest soil temperature throughout the year occurs from June to July,and the lowest soil temperature occurs around January.The soil temperature of 10cm and 20cm varies greatly,and the temperature of 30cm soil is relatively stable.(2)The 10cm,20cm,and 30cm large-period daily average soil temperature prediction models established based on LSTM neural network respectively have an average determination coefficient of R^2=0.992 and an average mean square error(MSE)of 0.874.The training accuracy is better and can meet the daily forecast requirements of soil temperature.The mean square errors of 10cm,20cm,and 30cm are 1.554,0.711,and 0.359,respectively.The 30cm model fits the best results,while the mean square error of 10cm is larger.The reason is that the surface soil temperature is more severely affected by weather conditions,and the wind speed,Air temperature and solar radiation intensity will affect it,while the soil at 30cm is less affected by weather conditions,and the prediction accuracy is more accurate.(3)Based on the LSTM neural network,the average coefficient of determination of the time-averaged soil temperature prediction models for 10cm,20cm and 30cm small periods is R^2=0.988,and the average mean square error(MSE)is 0.126.The time-averaged soil temperature simulation effect is ideal,and the mean square error is very small,indicating that this model is very practical as a time-averaged temperature prediction model,but there is a certain deviation between the predicted value and the true value in a local period,such as the highest temperature period of the day,The simulated value of this model is often 0.1～0.2℃ lower than the measured value.(4)A soil moisture prediction model was established based on the principle of farmland water balance in Gaocheng District,Shijiazhuang.Using the three-year real-time monitoring of soil moisture data and meteorological data,in the wheat-corn planting mode,calculate the day-to-day crop coefficients during the whole growth period of wheat and com,and use this as the local guiding crop coefficient,combined with local records for many years Establish a soil moisture prediction model with reference to crop evapotranspiration(ET0)to predict soil moisture.In the prediction calculation result with 5 days as the step length,the average relative error of wheat is 1.75mm,and the average absolute error is 0.74%.The average error of corn is 3.48mm,and the average relative error is 1.28%.(5)Apply the soil temperature prediction model and the soil moisture prediction model to provide calculation models and data support for the water-saving irrigation forecast management system in the irrigation forecast system.