A Review of Machine Learning and Deep Learning Approaches for Predicting Greenhouse Gas Emissions

Accurate prediction of greenhouse gas (GHG) emissions is critical in guiding informed climatic policymaking but has traditionally suffered from spatial biases, reporting lags, and inconsistent data. Machine learning (ML) and deep learning (DL) have in recent times become viable options for dealing with the complex, non-linear relationships in emissions data. These data-oriented methodologies have been used widely across sectors, showing high accuracy in prediction and flexibility. Here, this paper reviews recent advances in the application of ML and DL for the prediction of GHG emissions through their theoretical underpinnings, empirical performance using public datasets, and the use of explainable artificial intelligence (AI). A contrastive evaluation of exemplary studies identifies patterns in model performance, interpretability, and dependence on data. Special attention is paid to model interpretability and the contribution of explainable AI in increasing the policy applicability of methodologies. This paper contributes to an open, scalable, and actionable approach to data-driven decision support in climatic management, by linking technical advances to policy utility.