Machine learning based evaluation of airline CO2 efficiency at Istanbul airport

Abstract

The aviation industry’s growing carbon footprint necessitates data-driven evaluation tools.This study assesses the CO2 efficiency of airlines operating at Istanbul Airport by integrating operational flight data with the Atmosfair Airline Index through a machine learning framework. A multiple linear regression model was developed to predict CO2 Efficiency Points (EP) using two primary predictors: total payload and daily landing frequency. Flight observations were collected from FlightRadar24 for passenger aircraft operating on March 28, 2025, while EP values were obtained from the 2024 Atmosfair Index. The model demonstrated a strong explanatory capacity (Adjusted R2 ≈ 0.73) and acceptable predictive accuracy (MAE = 3.82; RMSE = 4.45), indicating that flight frequency and payload are statistically significant determinants of CO2 efficiency.The findings underscore that larger payloads and higher operational intensity are associated with improved efficiency scores, reflecting the critical role of data-informed scheduling and capacity management in sustainable aviation. Despite the limited sample size, the model exhibits robust internal validity and offers a transparent, reproducible approach for airport-level carbon performance assessment. By linking empirical aviation data with environmental performance metrics, this research contributes a lightweight yet scalable analytical framework that aligns with the International Civil Aviation Organization’s (ICAO) net-zero carbon target for 2050. The proposed model provides practical implications for airport operators and policymakers aiming to integrate predictive analytics into emissions monitoring and green airport management systems.