Contrail Detection and Coverage Estimation across Air Route Traffic Control Centers from Satellite Observations

Contrails are recognised as a net warming contribu- tor to aviation-induced climate impact, and their detection from satellite imagery provides an observational basis for estimating coverage and informing rerouting strategies. While informative, continental-scale estimates are insufficient for operational use, being constrained by air traffic management complexity and costs from CO2/non-CO2 trade-offs. A more feasible approach is to identify contrail-intense air traffic control regions and corridors, where targeted rerouting could deliver greater climate benefit at lower operational cost. We propose a joint spatial–frequency loss that conditions learning to emphasise contrail samples while capturing their line-shaped structures. On the OpenContrails dataset [1], our method consistently improves detection across multiple binarisation thresholds, achieving a global dice score of 70.23%. The proposed method is applied to GOES-16 imagery to estimate contrail coverage over CONUS and at the scale of individual Air Route Traffic Control Centers (ARTCCs). Multi-threshold analysis yields a range of contrail coverage from 0.015% to 0.254%, with a moderate average of 0.115% across 60 days in 2024. The moderate estimate reveals clear seasonal cycles, with greater contrail occurrence in winter and spring, and diurnal peaks between 08:00 and 10:00 local time across most regions. Joint evaluation with meteorological conditions further identifies contrail-intense zones and corridors, implying an uneven potential for region-specific contrail mitigation due to notable anomalies between ARTCCs.