Abstract
Natural methane (CH4) emissions from wet ecosystems are an important part of today's global warming. Climate affects the exchange of CH4 between ecosystems and the atmosphere by influencing CH4 production and oxidation. The net CH4 exchange depends on ecosystem hydrology and vegetation characteristics. This study focuses on investigating methane emissions specifically from rice paddies in West Bengal, India. The focus is on applying machine learning models to predict methane emissions based on various factors, including wind, temperature, precipitation, and pressure. In this study, we used data from the Copernicus Atmosphere Monitoring Service (CAMS), specifically the CAMS global greenhouse gas reanalysis (EGG4). We applied machine learning models, such as Support Vector Regression, Random Forest, Adaptive Boosting, XGBoost, and Multi-Layer Perceptron, and optimized their hyperparameters using the Optuna framework in Python. To assess the performance of these models, we used 10-fold cross-validation, which showed that the Multi-Layer Perceptron outperformed the others. Furthermore, this study highlights the relevance of hyperparameter adjustment in enhancing model accuracy and finding significant features, which is very useful in environmental monitoring applications.