Enhancing global methane emissions forecasting using hybrid time series models

Global warming is a major environmental issue that raises the average air temperature on Earth's surface. Human activities have played a key role in increasing greenhouse gas emissions, which contribute to higher temperatures and climate change. Methane is the second most significant greenhouse gas driving global warming. This study focuses on predicting global methane emissions using the SARIMA (Seasonal Autoregressive Integrated Moving Average) statistical model and three machine learning models: MLP (Multilayer Perceptron), LSTM (Long Short-Term Memory), and GRU (Gated Recurrent Unit). Two hybrid models, SARIMA-MLP and SARIMA-GRU, were also applied. The models’ accuracy was assessed using statistical metrics, including Mean Square Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The findings show that the SARIMA model outperformed the standalone machine learning models. However, the hybrid models demonstrated better forecasting performance, with SARIMA-GRU emerging as the most effective model for predicting global methane emissions. The forecast results indicate a continuous rise in methane emissions over time.