Soil Moisture Prediction Based on the ARIMA Time-Series Model

In order to solve the long-time prediction problem of soil moisture under different soil depth conditions, this paper proposes a soil moisture prediction model based on ARIMA time-series model. Soil moisture data at three different soil depths of 10cm, 40cm and 100cm are smoothed and noise-reduced, and then the processed data are input into the constructed ARIMA model and LSTM model for prediction, and the prediction results show that the combined model has better prediction results compared with LSTM, and finally, based on the Pearson correlation coefficient to filter out the precipitation (mm), average The four independent variables, precipitation (mm), mean wind speed (knots), mean temperature (°C) and mean visibility (km), were selected based on Pearson correlation coefficients. A least squares-based linear regression model was constructed for these four groups of variables to fit the moisture data from April to July 2022, and the results corroborated the correctness of the prediction results of the ARIMA model and verified that the algorithm outperformed the prediction of the LSTM soil moisture model.