New Approach for Soil Moisture Prediction Based on Multiple Influencing Factors

This study explores the intrinsic connection between meteorological characteristics and soil moisture. Based on the measured soil moisture data and meteorological data from 2012 to 2022, this paper screens out 6 key influencing factors among 24 features, and use the Seasonal Auto Regressive Integrated Moving Average (SARIMA) model to predict the regression of the screened key impact factors. The Multi-Layer Perceptron (MLP) model, the Long Short Term Memory (LSTM) model and the Long Short Term Memory optimized by the Sparrow Search Algorithm (SSA-LSTM) model are established to predict the soil moisture. By combining with SARIMA model, the relationship between the impact factors and soil moisture can be obtained. The results show that the LSTM model optimized by Sparrow Search Algorithm has good predictive ability on soil moisture in different soil layers, and R2 values are all above 0.92. Verified by the case, the relative error between the predicted value and the real value is all within 10%, indicating that the method can obtain the predicted value of soil moisture with high accuracy and stability.