利用观测数据和人工智能探测凝结尾迹的形成.pdf

1、?Observational Data and AI to Detect Contrail FormationEUROCONTROL Aviation Sustainability Unit CO2 and non-CO2 effects of aviation For any type of emissions or cloud induced effects,radiative forcing results from the imbalance between the reflection of the incoming solar radiation an outgoing terre

2、strial radiationThe climate impact of non-CO2 could be of the same order of magnitude as CO2 emissionsObservational Data and AI to Detect Contrail FormationPhysical models(+observations)Contrail formation&developmentRadiative forcingGround-camera observationSatellite observationTrafficWeather observ

3、ationsDetection,tracking,and flight attributionDetection and trackingColocationObjectives:Scientific:modelling contrails Operational:monitoring their impactContrail research in the Aviation Sustainability Unit Labelled datasets of contrails on remote sensors Models and algorithms for contrail detect

4、ion,tracking and flight attribution ContrailNet,a data portal to support contrail research Participation to NEATS,a system to monitor,report and verify non-CO2 effects of aviation Other research topics(altitude estimation,climate surrogate models,uncertainty quantification)A labelled dataset for con

5、trail detection,tracking and flight attribution 90 sequences of 1 or 2 hours,with an image every 30 sec 18K images Contrails are instances(sets of polygons)consistently labelled throughout the sequence Forming vs.legacy contrails For forming contrails,the corresponding flight is providedA panoptic s

6、egmentation approach to contrail detection and tracking Fine-tunned a cutting-edge transformer universal segmentation model with Things:Contrails Stuff:Sky Trivial extension to learn from sequences(videos)All-in-one contrail detection and t