Approaches for predicting dairy cattle methane emissions: from traditional methods to machine learning.

Area of Science:

• Agricultural Science
• Environmental Science
• Data Science

Background:

• Traditional methods for measuring dairy cattle methane (CH4) emissions are complex and costly.
• Prediction models offer an accessible alternative by estimating CH4 emissions using proxy data.

Purpose of the Study:

• To review and compare different modeling approaches for predicting dairy cattle CH4 emissions.
• To identify the strengths and limitations of mechanistic, empirical, and machine learning (ML) models.

Main Methods:

• A systematic literature search was conducted in December 2022 using PRISMA guidelines across multiple databases (Scopus, EBSCO, Web of Science, PubMed, PubAg).
• Search terms focused on "Bovine," "Statistical Analysis or Machine Learning," and "Methane Emissions."
• 55 eligible English-language papers investigating CH4 emission prediction in dairy cattle using statistical or ML methods were included.

Main Results:

• Mechanistic models are accurate but require hard-to-obtain data. Empirical models are versatile but limited outside their development range.
• Milk fatty acids (MFA) are a popular trait, but MFA-based models require consolidation for robust accuracy.
• Machine learning (ML) models offer novel approaches, handle diverse data, and can improve prediction accuracy on commercial farms.

Conclusions:

• ML models present a powerful, adaptable methodology for predicting dairy cattle CH4 emissions.
• These models can overcome limitations of traditional methods and enhance predictions in commercial settings.
• Further research is needed to consolidate MFA-based models for reliable CH4 emission prediction.