Monitoring of Soybean Soil Moisture Content Based on Hyperspectral and Machine Learning Models

Soil moisture content (SMC), as a key crop physiological index, is of great significance in intelligent agriculture. In this study, the hyperspectral data of soybean canopy and the soil moisture content of soybean root zone in shallow (0–20 cm), middle (20–40 cm) and deep (40–60cm) layers were measured and obtained. 20 empirical vegetation indices and 20 ‘trilateral’ parameters were selected as spectral features, and correlation analysis was carried out with soil moisture content in each layer. Among the ‘trilateral’ parameters, empirical vegetation index and all spectral parameters, five indexes with the highest correlation with SMC were selected to form combination 1, combination 2 and combination 3. Subsequently, these three combinations were combined with random forest (RF), extreme learning machine (ELM), extreme gradient boosting (XGBoost), gradient boosting decision tree (GBDT) and partial least squares regression (PLSR) algorithms to establish SMC inversion models for each layer of soybean root zone. The results show that under the same model and soil depth, the inversion model with ‘trilateral’ parameters as input variables has the highest accuracy. In the case of the same model and combination, the established shallow (0–20 cm) soybean root zone SMC inversion model has the highest accuracy. In the case of the same combination and soil depth, the inversion model established by the random forest (RF) algorithm has the highest accuracy. The SMC inversion model of shallow (0–20 cm) soybean root zone established by using pure ‘trilateral’ parameters as input variables combined with random forest (RF) model obtained the highest accuracy. The R2 values of the training set and the validation set were 0.854 and 0.872, the RMSE values were 0.013 and 0.022, and the MRE values were 6.686% and 6.405%, respectively. The model can be used to predict SMC at different soil depths during soybean flowering, providing technical support for the realization of smart agriculture. © (2024), (Taiwan Joint Irrigation Associations). All rights reserved.