Model biases in rice phenology under warmer climates

Area of Science:

• Agricultural Science
• Climate Science
• Computational Biology

Background:

• Climate change significantly impacts crop yields through alterations in plant phenology.
• Accurate simulation of rice phenology is essential for reliable climate-induced crop yield projections.

Purpose of the Study:

• To evaluate the performance of four rice phenology models (GDD, exponential, beta, bilinear) under warmer climates.
• To identify the most accurate models for predicting rice phenological shifts due to climate change.

Main Methods:

• Utilized phenology observations from 775 trials across 5 Asian countries, involving 19 rice cultivars.
• Calibrated and validated four phenology models using varying growing season temperatures (GST) ranging from 2.2 to 8.2°C.
• Assessed model bias and predictability based on cultivar diversity and temperature response patterns.

Main Results:

• In warmer climates, bilinear and beta models exhibited increasing bias in phenology prediction.
• GDD and exponential models demonstrated comparatively constant bias, indicating greater stability.
• Phenology model bias was primarily linked to temperature-dependent patterns, not calibration dataset size.
• Simulations using multiple cultivars improved predictability over single-cultivar models.

Conclusions:

• The GDD and exponential phenology models are recommended for simulating rice phenology under climate change.
• Utilizing multiple rice cultivars in simulations enhances the accuracy of predicting climate change impacts on phenology and yield.