AI-Driven Identification of Contrail Sources: Integrating Satellite Observation and Air Traffic Data

Abstract

Despite large uncertainties, it is now clear that condensation trails play a major role in aviation contribution to climate change. In order to assess these uncertainties and reduce them, a database of observations needs to be built up to improve prediction models and to enable aircraft trajectories optimization based on climate considerations. In order to build this observations database satellite images are good candidates, but detecting contrails in images is a time-consuming task without automation. In this paper, a dataset from GOES-16 satellite images is used to create a detection algorithm based on segmentation methods. Then, a method is introduced for associating contrails with aircraft trajectories based on ADS-B data. The Hough transform and meteorological forecast reanalysis data are applied to link any contrail with a group of flights that may have contributed to its formation.