Suitability Evaluation of Crop Variety via Graph Neural Network

Area of Science:

• Agricultural Science
• Artificial Intelligence
• Genomics

Background:

• Global population growth necessitates increased food production.
• Current AI models for crop yield prediction suffer from limited data and performance issues.
• Optimizing land and crop variety suitability is key to boosting agricultural output.

Purpose of the Study:

• To develop an advanced AI model for crop suitability evaluation using maize as a case study.
• To address limitations in existing AI models, including data scarcity and performance.
• To guide future crop breeding experiments through improved suitability assessments.

Main Methods:

• Collected extensive environmental climate and crop phenotypic traits data across multiple experimental sites for maize.
• Constructed a comprehensive dataset integrating diverse agricultural variables.
• Introduced and implemented a graph neural network (GNN) model for crop suitability evaluation.

Main Results:

• The graph neural network model demonstrated effective crop suitability evaluation.
• The model achieved a good evaluation effect, outperforming existing methods.
• Generated evaluation results provide a reliable reference for expert assessments.

Conclusions:

• The developed GNN model offers a robust approach to crop suitability assessment.
• The model's findings can guide expert evaluations and inform future breeding strategies.
• This AI-driven method aids in optimizing crop selection for diverse trial sites and enhancing food security.