Predicting ammonia emissions and global warming potential in composting by machine learning.

Area of Science:

• Environmental Science
• Agricultural Science
• Data Science

Background:

• Accurate prediction of composting gas emissions is crucial for assessing global warming potential (GWP).
• Existing methods for quantifying emissions like ammonia (NH3) are often limited in accuracy.
• Composting processes generate significant greenhouse gases, impacting climate change.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting NH3 emissions and GWP from composting.
• To identify key factors influencing gas emissions and compost maturity.
• To assess the practical applicability of predictive models for environmental management.

Main Methods:

• Utilized three machine learning models: extreme gradient boosting, random forest, and k-nearest neighbors.
• Employed Shapley Additive ExPlanation (SHAP) analysis to determine feature importance.
• Validated model performance using R-squared values and accuracy metrics.

Main Results:

• The extreme gradient boosting model achieved over 90% accuracy in predicting NH3 emissions and GWP.
• K-nearest neighbors model accurately determined compost maturity with 92% accuracy.
• Aeration rate, C:N ratio, and moisture content were key predictors for NH3 emissions.
• Nitrate (NO3-) was the most significant factor influencing GWP predictions.
• Predicted GWP closely matched annual calculations for California.

Conclusions:

• Machine learning, particularly extreme gradient boosting, offers a robust method for predicting composting emissions and GWP.
• Understanding key influencing factors enables targeted strategies for mitigating environmental impact.
• These predictive models have practical applications in environmental policy and waste management.