Methodological guidelines: Cow milk mid-infrared spectra to predict reference enteric methane data collected by an...

Area of Science:

• Animal Science
• Spectroscopy
• Environmental Science

Background:

• Methane (CH4) emissions from cattle contribute to greenhouse gases.
• Accurate prediction of CH4 emissions is crucial for mitigation strategies.
• Milk mid-infrared (MIR) spectroscopy offers a non-invasive method for estimating CH4 production.

Purpose of the Study:

• To identify optimal methodological protocols for predicting CH4 emissions using milk MIR spectroscopy.
• To evaluate the impact of different measurement durations and data processing techniques.
• To assess model performance on datasets including cows treated with 3-nitrooxypropanol (3NOP).

Main Methods:

• Collected milk MIR spectra and GreenFeed system (GF) measured CH4 emissions from 129 Holstein cows.
• Developed prediction models using partial least squares regression, splitting data into calibration and validation sets.
• Investigated effects of averaging spectra, measurement duration (1-4 basic measurement units - BMU), days in milk (DIM) correction, and inclusion of phenotypic data (parity, milk yield, FPCM).

Main Results:

• Models using average day spectra and longer GF measurement durations (4 BMU) improved CH4 prediction accuracy (R2V = 0.60 for g/d) compared to shorter durations (1 BMU, R2V = 0.52).
• Coupling GF data with average MIR spectra over the measurement period yielded better predictions (R2V = 0.70) than using single-day spectra.
• Correcting spectra by DIM enhanced model performance (R2V = 0.67), and including milk yield or FPCM further improved predictions for average spectra models (R2V = 0.73).
• Models failed to accurately predict CH4 emissions in cows treated with 3NOP, indicating a need for specific models for such cases.

Conclusions:

• Milk MIR spectroscopy, particularly when using average spectra over extended periods and incorporating DIM corrections, is a promising tool for predicting enteric CH4 emissions.
• Longer methane measurement durations and incorporating phenotypic data like milk yield can enhance prediction accuracy.
• The efficacy of these models is compromised in cows treated with 3NOP, necessitating the development of distinct predictive models for this specific scenario.