Accurate seed-yield prediction is essential for optimizing nitrogen (N) management in buffalograss seed production. However, current UAV-based approaches often rely directly on vegetation indices (VIs), which provide limited physiological insight and not transfer well across growing seasons. To address this limitation, we developed a trait-based yield prediction methold that integrates UAV-derived plant height (PH) and canopy nitrogen concentration (CNC), representing crop structural and physiological status, respectively. Field experiments were conducted from 2022 to 2024 under seven N application rates. Using data from 2022 and 2023, we calibrated a quadratic PH-CNC model and then evaluated its predictive performance with an independent 2024 dataset. We also compared this framework with a conventional direct VI-based model. The trait-based model explained 89% of the variation in seed yield during calibration and showed better cross-year predictive performance than the VI-based model (R 2 = 0.70, NRMSE = 17% versus R 2 = 0.52, NRMSE = 22%). In addition, the model captured the decline in seed yield under excessive N input, indicating that it reflected biologically meaningful crop responses. These results demonstrated that combining structural and physiological traits can provide a more robust and interpretable alternative to conventional VI-based methods for UAV-based yield prediction. This framework has practical potential for improving precise and sustainable N management in buffalograss seed production.
Wang et al. (Tue,) studied this question.