Predicting and Explaining for Rice Phenology Across China by Integrating Crop Model and Interpretable Machine Learning

Abstract

This study explores the integration of process-based crop models and machine learning (ML) approaches for predicting rice phenology across China, to gain a deeper understanding of rice growth mechanisms. A multi-model ensemble approach was used to predict flowering and maturity dates at 337 locations across the main rice growing regions of China from 1980 to 2020, combining CERES-rice, ORYZA v3, and RiceGrow. Hybrid models combining standard machine learning algorithms with those three crop models were employed to predict the timing of rice phenological stages. Furthermore, an interpretable machine learning (IML) approach using eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanation (SHAP) was employed to elucidate influence of climatic and varietal factors on phenology. The process-based model ensemble achieved a Root Mean Square Error (RMSE) of 7.64 and 11.55 days and an R-squared (R²) value of 0.82 and 0.70. In contrast, the hybrid model, based on a serial structure and the IML XGBoost, significantly outperformed these metrics, with RMSE of 5.08 and 6.62 days and R² values of 0.92 and 0.89 for flowering and maturity predictions, respectively. SHAP analysis revealed temperature to be the most influential climate variable affecting phenology predictions, particularly under extreme temperature conditions, while rainfall and solar radiation were found to be less influential. The analysis also highlighted the variable importance of climate across different phenological stages, rice cultivation patterns, and geographic regions, underscoring the notable regionality. The study proposed that a hybrid model using an IML approach would not only improve the accuracy of process-based models but also offer a robust framework for leveraging big data in crop modeling, providing a valuable tool for refining and advancing the modeling process in rice.

Keywords: Rice phenology, Crop models, Data-driven, Interpretable machine learning