Analysis of elaziğ airport aviation emissions using machine learning

The increasing share of aviation in the transportation sector has led to a growing interest in examining its impact on global warming and environmental pollution. In light of these factors, assessing the environmental impact of the aviation sector has become a significant research topic due to rising global awareness. Particularly considering the sector's contribution to greenhouse gas emissions, accurate emission inventories and forward-looking estimation studies are essential for shaping the future of aviation and implementing preventive measures against emissions. This thesis aims to create a detailed inventory of the gases (HC, CO, NOx, CO₂) released into the atmosphere during aircraft ground operations (taxiing) at Elazığ Airport and to evaluate the predictability of these emissions using machine learning methods. Based on calculations involving 11,042 commercial flight operations at Elazığ Airport in 2023 and 2024, it was estimated that approximately 1,265 kg of HC, 20,942 kg of CO, 6,093 kg of NOx, and 3.98 million kg of CO₂ were emitted during taxiing operations. Following the two-year emission inventory, short-term prediction performance was tested using Linear Regression, Support Vector Regression (SVR), Random Forest (RF), and Gradient Boosting Machine (GBM) models. Model performance was evaluated using RMSE, MAE, and MAPE metrics. The results indicated that ensemble learning methods provided higher accuracy compared to other models, with the best model achieving a MAPE of approximately 12%. This thesis presents a detailed analysis demonstrating that high-accuracy emission estimates can be achieved for Elazığ Airport using actual taxi times, and that these emissions can be reasonably modeled using machine learning approaches. The study is expected to offer valuable insights and methods for policymakers, urban planners, and decision-makers in the future.