This paper seeks to develop an effective model for forecasting aviation carbon emissions by leveraging comprehensive flight-level data for the year 2019. To achieve this objective, we employed traditional time series models – Autoregressive Integrated Moving Average (ARIMA) and Vector Autoregression (VAR) – alongside machine learning approaches: Random Forest (RF), Multilayer Perceptron (MLP), and Long Short-Term Memory (LSTM) networks. These methodologies were utilized to predict carbon emissions for individual flights as point estimates. In addition, we applied the Variational Mode Decomposition (VMD) method to decompose the data into its constituent components, leading to the development of the models VMD RF, VMD MLP, and VMD LSTM for comparative analysis. The findings reveal several key insights: firstly, the VMD LSTM model demonstrated superior performance in prediction accuracy, closely followed by the VAR model; secondly, the prediction errors associated with VMD RF, VMD MLP, and VMD LSTM were consistently lower than those of their non-decomposed counterparts (RF, MLP, and LSTM), which highlights the effectiveness of data decomposition in enhancing predictive outcomes. This paper contributes innovative methodologies for forecasting aviation carbon emissions, providing critical insights to support initiatives aimed at reducing carbon emissions within the aviation sector.