Forecasting Aviation Carbon Emissions with Tree-Based Machine Learning: A Case Study of Turkish Airlines Operational Data

The global aviation industry plays a critical role in economic development and international connectivity. However, it also contributes significantly to environmental challenges, particularly through fuel consumption and carbon emissions. The aviation sector is responsible for approximately 1–2% of global CO₂ emissions, and its environmental impact is expected to increase substantially by 2050. This study proposes a machine learning (ML)-based decision support framework to forecast carbon emissions in the airline sector. Using Turkish Airlines’ quarterly operational data from 2008 to 2023, Decision Trees, Random Forest, XGBoost, and Extra Trees models were applied to predict carbon emissions. Key variables include revenue passenger kilometers (RPK), available seat kilometers (ASK), load factor (LF), fuel cost, number of aircraft, aircraft utilization, and fleet age. Among the tested models, the Extra Trees model achieved the highest accuracy, with an RMSE of 906.98, a MAPE of 6%, and an R² of 0.95. Beyond prediction accuracy, model explainability was enhanced through SHAP (Shapley Additive exPlanations) analysis, which confirmed the dominant influence of RPK and LF and provided interpretable insights for decision-makers. These results demonstrate the effectiveness of machine learning techniques in improving fuel efficiency, minimizing carbon emissions, and supporting sustainable airline operations. Furthermore, this study highlights the importance of integrating predictive analytics into strategic decision-making for emission control and operational optimization. The research contributes to both academic literature and industry practice, particularly in the context of emerging aviation markets. Future studies may incorporate external shocks, hybrid model designs, and adaptive time series approaches.