To overcome the limitations of traditional yield estimation methods—which are often subjective, costly, and difficult to implement at scale—this study developed a high-precision, interpretable model for predicting walnut yield by integrating multi-source remote sensing technology with interpretable machine learning. To provide a theoretical foundation for precise water and fertilizer management as well as intelligent production in walnut orchards. By employing interpretable machine learning and a multi-stage integration strategy, the model achieves not only high-precision yield estimation but also elucidates the influence pathways of water–fertilizer coupling on yield formation at a mechanistic level. This advancement offers reliable technical support and a decision-making framework for the precise management of orchards. This study focused on the Xinjiang ‘Wen 185’ walnut, employing field experiments with varying water and fertilizer gradients. A UAV equipped with a multispectral sensor was utilized to capture canopy images, from which vegetation indices and texture features were extracted. This process resulted in a comprehensive dataset that integrated remotely sensed features with management practices. Various machine learning algorithms, including random forest, support vector machine, partial least squares regression, and ridge regression, were applied. An innovative stacked integration model for growth stages was proposed, and the SHAP framework was incorporated to analyze feature contributions and enhance model interpretability. In this study, texture features—particularly those derived from the red-edge band—showed higher predictive importance than traditional vegetation indices. This suggests that they may be more sensitive to canopy structural heterogeneity under the tested conditions. Among the models, random forest showed numerically higher values in terms of R² and RPD compared to the other individual models under the present dataset, achieving a validation R² of 0.670 and an RPD of 1.836. The proposed growth stage stacking ensemble (GSSE) model further enhanced prediction accuracy, achieving validation R² of 0.789, an RMSE of 0.494, and an RPD of 2.296. Additionally, the results revealed that texture may have a potential ability to captured canopy heterogeneity as the primary mechanism underlying yield variation, and the integration of multi-stage spectral information was associated with higher estimation accuracy in this dataset in improving estimation accuracy, with the oil conversion stage contributing up to 60% to the final prediction.

Accurate monitoring of the leaf area index (LAI) throughout the entire growth cycle of walnut trees using UAV multispectral imagery is essential for digital orchard management. In this study, focusing on the ‘Wen 185’ walnut variety in Xinjiang, we simultaneously acquired UAV multispectral images and ground-measured LAI data during four critical growth stages: expansion, hard shell, oil conversion, and maturity. A total of 25 vegetation indices and 48 texture features derived from the gray-level co-occurrence matrix were extracted. Hybrid feature selection combining linear (Pearson correlation), nonlinear (maximum information coefficient and random forest importance), and multiple consensus strategies was employed to reduce redundancy. LAI prediction models were constructed using four algorithms: Random Forest (RF), Support Vector Machine (SVM), LASSO, and Ridge Regression (RR), with model interpretability enhanced by SHAP analysis. Results showed that the multiple consensus screening reduced feature redundancy by an average of 69.6%. SHAP identified five core features: Redge_750_Mean, NDVI, B_Mean, RENDVI, and G_Homogeneity. Importantly, predictor importance shifted significantly with phenology: texture features dominated during the expansion stage, while red-edge indices (RENDVI and Redge_750_Mean) became predominant during the hard shell and oil conversion stages, effectively mitigating the saturation problem commonly observed in traditional indices such as NDVI within the LAI range of 1.5–5.8 in this study. The hybrid feature subset combining “red-edge spectrum + spatial texture” with the Random Forest algorithm achieved superior performance across all stages, with the RPD value exceeding 2.0 during the oil conversion stage, indicating excellent estimation capability. This study demonstrates that a “quality over quantity” feature selection strategy not only reduces model complexity but also enables high-precision, dynamic LAI monitoring throughout the entire walnut growth cycle, providing a scientific basis for intelligent management of large-scale orchards in arid regions.

Rapid, non-destructive estimation of leaf chlorophyll content (SPAD) is crucial for assessing plant photosynthetic health and nutrient status. However, conventional methods rely on specialized instruments (e.g., SPAD meters and hyperspectral sensors) which are costly, cumbersome, or unsuitable for large-scale field deployment. While RGB image analysis offers a low-cost alternative, most existing approaches depend solely on color features, which are susceptible to environmental interference and lack robustness across growth stages. To address these limitations, this study proposes a novel machine learning framework that fuses both color and texture features from smartphone-captured RGB images for accurate SPAD estimation in walnut seedlings and explores its linkage with potassium nutrition. ‘Wen 185’ walnut seedlings were subjected to seven potassium concentration treatments to induce a chlorophyll gradient. From the leaf images, 22 color indices and 8 texture features based on the Gray-Level Co-occurrence Matrix (GLCM) were extracted. Prediction models were built and compared using Random Forest (RF), XGBoost, and a Support Vector Machine (SVM), with two fusion strategies: data-level and feature-level fusion. Results demonstrated that the RF model with feature-level fusion achieved optimal performance (validation set: R² = 0.939, RMSE = 0.014, and RPD = 4.539), significantly outperforming models using single-feature types. SHAP analysis identified normalized red, normalized blue, and green-band correlation as the most influential features. This work fills a critical gap by establishing a robust, cost-effective, and interpretable method for SPAD monitoring using ubiquitous RGB imagery. Furthermore, the strong correlation between image-predicted SPAD and potassium levels confirms the method’s high potential for early and non-destructive diagnosis of potassium deficiency in orchard management.