A machine learning approach for contrail detection and persistence prediction using airborne measurements from the ECLIF II/ND-MAX and ecoDemonstrator flight campaigns

Abstract. Contrails or condensation trails are a major contributor to aviation-induced cloudiness, which represents a significant, yet highly uncertain, component of aviation's environmental impact. The reliable detection and characterization of contrails has become increasingly important for quantifying their radiative forcing and developing mitigation strategies. However, contrail detection and prediction remains challenging due to their variable optical properties, lack of accuracy in humidity, temperature and pressure sensor measurements, as well as in weather prediction. This paper investigates the use of machine vision on board aircraft to inform contrail formation and classification in real time. The focus of this work is the comparison of two major airborne measurement campaigns conducted by NASA, industry, and international partners: the 2018 NASA – DLR ECLIF II/ND-MAX and the 2023 NASA – Boeing ecoDemonstrator flight tests. The atmospheric data from the campaigns are used with the Schmidt-Appleman criterion to identify periods of contrail formation. For the ND-MAX dataset classification accuracies of 94.0 %, 87.1 %, and 81.5 % were obtained for contrail absence, short-lived and persistent categories, respectively. For the ecoDemonstrator dataset the corresponding accuracies were 90.4 %, 89.5 %, and 93.3 %. Comparison between the two campaigns reveals that camera placement affects the classification performance; longer visible contrail segments improve detection for the absence and short-lived categories while reduced airframe intrusion in the camera's field of view improves persistent contrail classification. Based on these findings, some recommendations for camera placement on flights are provided. The approach requires only an onboard camera as additional instrumentation, making it a cost-effective and scalable tool that may complement existing contrail monitoring and mitigation strategies.