Effects of Flight Distance Inputs on ICAO-Modeled Aviation CO2 Inventories: Flight-Level ADS-B Evidence from Kunming Airport

Flight distance is consequential but often simplified in aviation carbon inventories. Using 269,408 paired flights at Kunming Changshui International Airport (KMG) in 2023, this study compares distances corrected using the International Civil Aviation Organization (ICAO) Carbon Emissions Calculator (ICEC) with aggregate takeoff-to-landing distances derived from Automatic Dependent Surveillance–Broadcast (ADS-B) data under identical ICAO aircraft-equivalence mapping and fuel curves, thereby isolating the effect of distance input alone. The ADS-B distance inventory was 5.776 Mt CO2, 216,564 t (3.90%) above the 5.560 Mt ICEC distance inventory. An airport-pair resampling analysis produced a 95% composition-sensitivity interval of 133,259–324,634 t (2.77–5.18%). Positive and negative flight-level discrepancies partly offset: the gross absolute discrepancy was 256,312 t, and the cancellation share was 15.5%; the ten largest airport pairs accounted for 48.8% of the gross discrepancy. Historical route and operational calibration reduced modeled CO2 mean absolute error from 857 to 328 kg per flight in a temporal holdout, but the error rose to 743 kg per flight under the whole-airport holdout; adding KMG Meteorological Aerodrome Report (METAR) variables did not improve performance. These findings show that distance input choice materially affects ICAO-modeled inventories and that locally calibrated substitutes can improve within-network reconstruction, but these results should be validated for transferability before application to counterpart airports absent from training.