Plant moisture content (PMC) serves as a crucial indicator of crop water status, directly affecting agricultural productivity, product quality, and the effectiveness of precision irrigation. Conventional methods for PMC assessment predominantly rely on destructive sampling techniques, which are labor-intensive and impede real-time monitoring. This study investigates silage maize cultivated in the Hexi region of China, leveraging multispectral data acquired via an unmanned aerial vehicle (UAV) to estimate PMC across different phenological stages. A stacked ensemble learning framework was developed, integrating Back Propagation Neural Network (BPNN), Random Forest Regression (RFR), and Support Vector Regression (SVR), with Partial Least Squares Regression (PLSR) employed for feature fusion. The findings indicate that incorporating vegetation indices into spectral variables significantly improved prediction performance. The standalone models demonstrated coefficient of determination (R²) values ranging from 0.43 to 0.69, with root mean square error (RMSE) spanning 0.61% to 1.43%. In contrast, the ensemble model exhibited superior accuracy, achieving R² values between 0.61 and 0.87 and RMSE values from 0.54% to 1.38%. This methodology offers a scalable, non-invasive alternative for PMC estimation, facilitating data-driven irrigation optimization in regions facing water scarcity.

Accurately monitoring soil moisture content (SMC) in the field is crucial for achieving precision irrigation management. Currently, the development of UAV platforms provides a cost-effective method for large-scale SMC monitoring. This study investigates silage corn by employing UAV remote sensing technology to obtain multispectral imagery during the seedling, jointing, and tasseling stages. Field experimental data were integrated, and supervised classification was used to remove soil background and image shadows. Canopy reflectance was extracted using masking techniques, while Pearson correlation analysis was conducted to assess the linear relationship strength between spectral indices and SMC. Subsequently, convolutional neural networks (CNNs), back-propagation neural networks (BPNNs), and partial least squares regression (PLSR) models were constructed to evaluate the applicability of these models in monitoring SMC before and after removing the soil background and image shadows. The results indicated that: (1) After removing the soil background and image shadows, the inversion accuracy of SMC for CNN, BPNN, and PLSR models improved at all growth stages. (2) Among the different inversion models, the accuracy from high to low was CNN, PLSR, BPNN. (3) From the perspective of different growth stages, the inversion accuracy from high to low was seedling stage, tasseling stage, jointing stage. The findings provide theoretical and technical support for UAV multispectral remote sensing inversion of SMC in silage corn root zones and offer validation for large-scale soil moisture monitoring using remote sensing.

Accurately monitoring total nitrogen (TN) content in field soils is crucial for precise fertilization management. TN content is one of the core indicators in soil fertility evaluation systems. Rapid and accurate determination of TN in the tillage layer is essential for agricultural production. Although UAV-based multispectral remote sensing technology has shown potential in agricultural monitoring, research on its quantitative assessment of soil TN content remains limited. This study utilized UAV (unmanned aerial vehicle) multispectral imagery and field-measured TN data from four key growth stages of silage corn in 2022 at Huari Ranch, Minle County, Hexi region. The support vector machine–recursive feature elimination (SVM-RFE) algorithm was applied to select vegetation indices as model inputs. A total of 18 models based on machine learning algorithms, including BP neural networks (BPNNs), random forest (RF), and partial least squares regression (PLSR), were constructed to compare the most suitable inversion model for TN in the rhizosphere soil (0–30 cm) of silage corn at different growth stages. The optimal period for TN inversion was determined. The SVM-RFE algorithm outperformed the models built without feature selection in terms of accuracy. Among the nitrogen inversion models based on different machine learning algorithms, the PLSR model showed the best performance, followed by the RF model, while the BPNN model performed the worst. The PLSR model established for the mature growth stage at soil depths demonstrated the highest inversion accuracy, with R and RMSE values of 0.663 and 0.281, respectively. The next best period was the tasseling stage, while the worst inversion accuracy was observed during the seedling stage, indicating that the mature stage is the optimal period for TN inversion in the study area.