Modeling rice leaf area index and canopy height in the US Mid-South region

Crop growth modeling plays a critical role in addressing the global challenges of food scarcity, carbon cycling, and water management. By simulating crop development from environmental factors, these models help predict harvest yield and carbon or water cycle terms and thus can inform policy and investment decisions. However, for some agricultural regions, such as the US Mid-South, there is a lack of comprehensive data specific to rice (Oryza sativa) cultivars and their growing conditions. Here, we use 30 field seasons of observational data to predict leaf area index (LAI) and canopy height (H_can), key inputs for crop growth models, for numerous rice cultivars grown under different conditions in east-central Arkansas. Our results show that a peaked response to the accumulated growing degree day (LAI: R² = 0.83, root mean square error (RMSE) = 0.97 m² m⁻²; H_can: R² = 0.90, RMSE = 10.9 cm) is a better functional form than one without temperature information, that is, driven only by days after planting (LAI: R² = 0.73, RMSE = 1.22 m² m⁻²; H_can: R² = 0.83, RMSE = 14.5 cm). Additionally, predictions from a cultivar-agnostic model are comparable to the predictions from cultivar-specific models, suggesting that the broader cultivar-agnostic model is sufficient and can be widely applied to rice production systems in this region. Such a generalized model can support yield prediction efforts, disentangle carbon cycle terms, or be used for crop stress detection.