Intelligent terminal airspace management with real-time adaptability to complex meteorological conditions

Abstract

The rapid growth of the aviation industry has precipitated a surge in air traffic density, particularly in terminal areas where meteorological conditions are increasingly affecting airspace management. The Operations Network (OPSNET) delay attribution data indicates that weather-induced disruptions account for 73.8% of total air traffic delays, designating adverse meteorological conditions as the principal risk factor for aviation system resilience. Adverse weather reduces flight safety, disrupts schedules, and leads to system-wide delays, highlighting the urgent need for improved weather-responsive management strategies. Trajectory-based operations (TBO) have emerged as a key concept in Next Generation Air Transportation System (NextGen), aiming to enhance flight predictability and efficiency through data-driven trajectory management. Despite advancements in air traffic management, current TBO frameworks exhibit limitations in incorporating real-time meteorological adaptability. Conventional airspace prediction models rely on deterministic approaches with limited capability to dynamically adjust under evolving weather disruptions. Moreover, while artificial intelligence and data-driven strategies have been explored for en-route traffic management, their application in terminal airspace remains underdeveloped, especially in addressing interactive dependencies between flight safety, traffic flow efficiency, and meteorological uncertainties. The thesis intends to propose a data-driven intelligent airspace management under various meteorological scenarios, and further explore the meteorological impact on the operational performance of terminal airspace. The research consists of three progressive stages: First, it introduces a novel image-based trajectory representation framework for processing 4D flight trajectories. By converting latitude, longitude, flight level, and ground speed into multi-channel image pixels, the proposed approach enables effective trajectory feature extraction through deep convolutional autoencoders (DCAE), demonstrating superior performance in similarity analysis. Second, a Transformer-BiGRU hybrid model is developed to resolve tactical trajectory prediction challenges posed by holding patterns and air traffic control (ATC) interventions, achieving a 13% reduction in horizontal distance deviation. Finally, the research innovatively integrates real-time meteorological impacts via a Spatio-temporal Weather and Airspace Graph Network (SWAG-Net), which synthesises multi-layer weather radar data, ADS-B trajectories, and airspace topology to enhance estimated time of arrival predictions under convective weather conditions. The thesis contributes three innovations: 1) A paradigm transformation in trajectory processing through image processing techniques, enabling DCAE-based analysis of abnormal flight trajectories; 2) Enhanced prediction robustness via online learning architectures that adapt to interactive ATC holding instructions; 3) Quantification of weather-induced operational impacts, with SWAG-Net demonstrating significant accuracy improvements during adverse meteorological events. The methodologies establish a critical pathway for climate-resilient air traffic management, offering implementable solutions that enable NextGen ATM systems to adapt to complex airspace and meteorological scenarios dynamically.