Evaluation of Robust Flight Attribution Algorithm for Contrail Avoidance

Condensation trails, or “contrails,” are line-shaped clouds that form when airplanes fly through cold and humid parts of the atmosphere which are ice-supersaturated. Various studies have shown that long-lasting or “persistent” contrails may be responsible for more than half of aviation’s radiative forcing (RF) (Lee, et al., 2021). Efforts to mitigate persistent contrail formation include operational contrail avoidance. Current research suggests that minor (~2000 ft) deviations in altitude of flights during cruise, in conjunction with advancing technologies, have the potential to reduce contrail climate forcing by approximately 90% (Teoh, et al., 2020). Identifying and attributing observed contrails to specific individual flights is critical to demonstrating the success of any contrail avoidance strategy, as it establishes whether a deviated flight created a contrail. Reliable attribution of contrails to individual flights is needed to provide verifiability and accountability before any large-scale implementation of contrail avoidance policies takes place. Flight attribution leverages both Earth-observation methods, such as satellite images and weather data, and flight data. However, temporal and spatial "blindspots" in satellite instruments, coupled with uncertainties in wind fields, have hindered reliable flight attribution. In this work, we consider two approaches: a “Single-instance” probabilistic flight attribution algorithm and a “Multi-frame” probabilistic flight attribution algorithm that accounts for discrepancies in contrail observability and weather data errors. The inputs to both algorithms include contrail detections, ERA5 weather data, and FlightAware ADSB flight data. The Single-instance algorithm computes a probabilistic “match score” for flights and contrails at an individual temporal point. This probability is calculated using distance, heading, and altitude measures. The Multi-frame algorithm computes a probabilistic match score utilizing information from different wind ensembles and previous temporal points in addition to the same baseline measures. To perform this analysis, a dataset of over a hundred manually labeled and attributed contrails was created that captured regional (across the continental United States) and diurnal variation. These were categorized depending on the perceived difficulty of the attribution (due to background cloudiness and flight track density). An initial assessment comparing the outputs of the algorithms to the manually labeled attributions shows that while both our algorithms exhibit strong performance in high contrail observability and low flight density scenes, the Single-instance algorithm demonstrates suboptimal results under conditions of interrupted contrail observability and increased flight density. The Multi-frame algorithm, however, was able to identify flight matches much more accurately in these more challenging scenes. The development of a robust flight-matching algorithm and evaluation dataset is critical to the validation of contrail avoidance efforts. Furthermore, it can provide additional insight into the relationship between meteorology, aircraft parameters, and observable contrails, supporting future efforts to reduce contrail formation through technological or operational means.