Deep Learning-Based Prediction of Short-Term CO₂ Emissions in Ireland

Accurate forecasting of carbon dioxide (C0<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions is crucial for developing effective environmental policies and emission reduction strategies in response to global climate change. This study explores the use of Long Short-Term Memory (LSTM) networks, a type of deep learning model specifically designed for time series data, to predict short-term CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions in Ireland. Daily CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> data from a monitoring station in Galway were used to train the LSTM model. Experimental results demonstrate that LSTM outperforms traditional machine learning methods, especially for medium to long-term forecasts. The application of regularization techniques, including Dropout and Early Stopping, further enhanced the model's performance by mitigating overfitting. Future work will involve the use of Bidirectional LSTM (BiLSTM) models and the comparative analysis of other greenhouse gases to improve forecasting accuracy and deepen the understanding of emission dynamics.