Analysis of Corn Yield Prediction Potential at Various Growth Phases Using a Process-Based Model and Deep Learning

Area of Science:

• Agricultural Science
• Crop Modeling
• Machine Learning

Background:

• Accurate grain yield prediction is vital for food security and policy.
• Identifying key crop growth phases and features enhances yield prediction efficiency.

Purpose of the Study:

• To develop a hybrid approach using the WOFOST model and deep learning for corn yield forecasting.
• To analyze yield prediction potential at different growth stages and identify key predictive features.

Main Methods:

• Utilized the World Food Studies (WOFOST) model to generate a simulated dataset with meteorological, soil, crop, and management data.
• Employed machine learning and deep learning methods with various feature combinations across different growth phases to forecast yield.
• Evaluated prediction performance at flowering, milk maturity, and maturity stages.

Main Results:

• Key features for vegetative and reproductive stages were identified as growth state and water-related features, respectively.
• Yield prediction accuracy significantly improved after entering the reproductive growth stage.
• Optimal model performance achieved R² values of 0.53 (flowering), 0.89 (milk maturity), and 0.98 (maturity).

Conclusions:

• The hybrid WOFOST and deep learning method enhances corn yield prediction accuracy across various growth stages.
• The findings provide reliable analysis for agricultural decision-making and can be applied to other crops.
• Improved yield prediction supports farmers and governments in agricultural production and policy formulation.