Air traffic delays have a major impact on the aviation industry, affecting airlines, passengers, and the broader ecosystem. With increasing regulatory and sustainability pressures, accurate delay predictions are critical as they allow for precise determination of the contingency and discretionary fuel required for flights. This research aims to develop an explainable supervised learning model to improve existing en route delay predictions, focusing on intercontinental flights from North America to Amsterdam Schiphol Airport. While prior studies have explored flight delay prediction, they have not addressed two critical research gaps identified in this research: the inclusion of day-of-operations features, such as passenger information, aircraft weights, and cost index, and the use of transatlantic flight data for predictions 90 minutes before departure. To address these gaps, two Gradient-Boosted models, CatBoost and LightGBM, were trained using internal airline, airport, and METAR data. Both models outperformed the airline’s current in-use statistical model, with CatBoost achieving an MAE of 3.44 minutes and RMSE of 4.61 minutes and LightGBM achieving an MAE of 3.43 minutes and RMSE of 4.56 minutes. The most significant performance increase over the current model was observed under adverse weather conditions. This research advances en route delay prediction by providing more accurate delay forecasts, particularly in critical weather conditions, and proposes practical improvements to support future studies focused on enhancing model adaptability across diverse operational contexts.

EUROCONTROL's Performance Review Commission launched the 2024 PRC Data Challenge in July 2024 with the aim of engaging with data scientists and aviation enthusiasts for the development of an open model to estimate an aircraft's take-off weight. The dataset for the challenge represents a unique instance of otherwise difficult-to-obtain flight information and could be reused for educational purposes or to further improve the outcome of the challenge.

Accurately estimating aircraft fuel flow is critical for evaluating new procedures, designing next-generation aircraft, and monitoring the environmental impact of current aviation practices. This paper investigates the generalization capabilities of deep learning models for fuel flow prediction, focusing on their performance with aircraft types not included in the training data. We propose a novel methodology that combines neural network architectures with domain generalization techniques to improve robustness and reliability across different aircraft types. Using a comprehensive dataset of 101 aircraft types, split into training (64 types) and generalization (37 types) sets with each type represented by 1,000 flights, we introduce a pseudo-distance metric to quantify aircraft type similarity and explore sampling strategies to improve model performance in data-limited regions. Our findings show that for unseen aircraft types, especially with noise regularization, the model outperforms baselines such as corrected proxy estimates. This study demonstrates the potential of blending domain-specific insights with advanced machine learning techniques to develop scalable, accurate, and generalizable fuel flow estimation models.

The OpenSky Network has been collecting and providing crowdsourced air traffic surveillance data since 2013. The network has primarily focused on Automatic Dependent Surveillance-Broadcast (ADS-B) data, which provides highfrequency position updates over terrestrial areas. However, the ADS-B signals are limited over oceans and remote regions, where ground-based receivers are scarce. To address these coverage gaps, the OpenSky Network has begun incorporating data from the Automatic Dependent Surveillance-Contract (ADS-C) system, which uses satellite communication to track aircraft positions over oceanic regions and remote areas. In this paper, we analyze a dataset of over 720,000 ADS-C messages collected in 2024 from around 2,600 unique aircraft via the Alphasat satellite, covering Europe, Africa, and parts of the Atlantic Ocean. We present our approach to combining ADS-B and ADS-C data to construct detailed long-haul flight paths, particularly for transatlantic and African routes. Our findings demonstrate that this integration significantly improves trajectory reconstruction accuracy, allowing for better fuel consumption and emissions estimates. We illustrate how combined data captures flight patterns across previously underrepresented regions across Africa. Despite coverage limitations, this work marks an important advancement in providing open access to global flight trajectory data, enabling new research opportunities in air traffic management, environmental impact assessment, and aviation safety.

Punctuality is a key performance indicator for any airline, especially hub-and-spoke airlines, given their focus on short passenger connections. Flights that are delayed at departure need to compensate for lost time whilst airborne. Because fuelling takes place well before scheduled departure, predicted departure delays determine the planned fuel amounts for en-route speed optimization. To prevent unnecessary fuel burn, airlines benefit from highly accurate departure delay predictions. This study aims to extend previous work on airline departure delay forecasting to a dynamic and probabilistic domain, whilst incorporating novel day-of-operations airline information to further minimize prediction errors. Random Forest, CatBoost, and Deep Neural Network models are proposed for a case study on departure flights of a major hub-and-spoke airline from its hub airport between 1 January 2020 and 1 August 2023. The Random Forest model is selected for its probabilistic performance and high accuracy in predicting delays between 5 and 25 min, for which en-route speed optimization has the largest effect. At the 90 min prediction horizon, the model reaches a Mean Absolute Error of 8.46 min and a Root Mean Square Error of 11.91 min. For 76% of flights, the actual delay is within the predicted probability distribution range. Finally, this study puts a strong emphasis on explainability. Flight dispatchers are therefore provided with the main factors impacting the prediction, explaining the context of the flight. The versatility of the model is demonstrated in two shadow runs within the procedures of an international airline, where delays caused by familiar and unfamiliar factors were successfully predicted.

The mass of an aircraft is crucial for performance-related studies, such as predicting flight trajectories and analyzing flight emissions. In these studies, the flight trajectories are often reconstructed using a point-mass aircraft performance model combined with flight profiles from surveillance data and take-off mass information. However, airlines do not usually disclose take-off mass information, considering its sensitive nature. Thus, aircraft masses often need to be assumed or estimated. This paper presents a simple and computationally effective approach for estimating take-off mass using only open data and models. We explore the strong correlation between take-off mass, flight distance, cruise altitude, and partially, the airspeed during the cruise. The main idea is to generate fuel-optimal trajectories with known masses and distances, and then compare them with actual flight data. The optimal trajectories are generated using the open aircraft performance and optimization library. By assuming that actual flights follow quasi-fuel-optimal trajectories, the take-off mass of a flight can be estimated based on simple regression models trained on the optimal trajectory dataset. This open-loop take-off mass estimation approach requires no proprietary information from aircraft manufacturers or airlines. We verified the model with an anonymized dataset containing actual A320 flights with known take-off mass. Our two- and three-feature multi-linear models yield mean absolute percentage errors of 5.95 % and 4.89 %, respectively. This study is another step forward in open science and a contribution to the aircraft trajectory studies.

Condensation trails, or contrails, are line-shaped clouds that are produced by an aircraft engine exhaust. These contrails often impact climate significantly due to their potential warming effect. Identification of contrail formation through satellite images has been an ongoing research challenge. Traditional computer vision techniques struggle with varying imagery conditions, and supervised machine learning approaches often require a large amount of hand-labeled images. This study researches few-shot transfer learning and provides an innovative approach for contrail segmentation with a few labeled images. The methodology leverages backbone segmentation models, which are pretrained on existing image datasets and fine-tuned using an augmented contrail-specific dataset. We also introduce a new loss function, SR loss, which enhances contrail line detection by incorporating Hough transformation in model training. This transformation improves performance over generic image segmentation loss functions. The openly shared few-shot learning library, contrail-seg, has demonstrated that few-shot learning can be effectively applied to contrail segmentation with the new loss function.

Estimating contrail formation and proposing effective mitigation strategies have posed significant challenges for the aviation industry in recent years. This study utilizes the Japanese airspace as a case study to address the challenge of assessing and minimizing the environmental impact of contrails. Initially, we introduce a novel combination of meteorological and flight trajectory data sources, followed by a comparative data quality analysis. Drawing on four months of data during different seasons from 2023, we conduct an in-depth analysis of contrail formation within the Japanese airspace, uniquely quantifying contrails with highresolution data to provide insights into their geographical and seasonal variations. Subsequently, we examine the effectiveness of altitude diversions as a mitigation strategy. Our findings identify clear geographical and seasonal influences on contrail formation in the region. We illustrate that altitude diversions can significantly reduce contrail formation with minimal altitude adjustments of up to 2000 ft. We found that minor altitude diversions can mitigate between 70% and 90% of the persistent contrail formed near the Japan region. Notably, the results also highlight a concerning phenomenon: during warmer months, such as July, a higher quantity and percentage of persistent contrails is observed, and a larger proportion of these contrails cannot be mitigated through altitude diversions. This result could intensify positive radiative forcing during warmer periods, underscoring the need for further research into contrail mitigation strategies.

With the increasing coverage of the OpenSky, flight data gathered from the network of receivers has become a primary source of open data for aviation research. This year marks ten years of OpenSky. In this report, we employ a year's worth of OpenSky data to analyze the formation of contrails and study the potential mitigation with altitude diversions. Persistent contrails, often formed under humid atmospheric conditions, significantly trap outgoing terrestrial radiation. More insights into contrails are essential for studying aviation climate impact. We estimated the potential formation of persistent contrails based on the numerical weather assimilation data. An efficient approach is employed to fuse meteorology data in our analysis, which allows the fast evaluation of these contrail conditions at a very large scale. We designed a simple yet effective algorithm for flight with persistent contrails to study the shortest altitude diversion that would have prevented the contrail formations. We have estimated that between 5 % and 9 % of total flight distances each month could have formed persistent contrails. Furthermore, altitude diversions could have mitigated 70 % of these contrails.

Contrails, formed under specific atmospheric conditions, have a noteworthy role in heat-trapping within the atmosphere. This study bridges the gap between theoretical contrail formation models and real-world data by employing flight information from OpenSky and meteorological data from the European Centre for Medium-Range Weather Forecasts. We introduce a computationally efficient contrail estimation module, leveraging a client-server architecture that allows on-demand weather data interpolation via an API, significantly reducing computational load and enhancing performance locally. The study also benchmarks the entire pipeline, from data acquisition to contrail prediction, offering a robust tool for future air traffic studies requiring interpolated weather data.

The study of the environmental transition of the aviation sector calls for prospective traffic scenarios. Detailed traffic and emissions inventories are often needed to refine the available analyses and to enable the simulation of regionalised scenarios. In the past studies, these are generally based on commercial, proprietary traffic data, making their dissemination problematic and reducing the reproducibility of the science produced. Open-source alternatives do exist but with limited geographical coverage. This study bridges this gap by presenting an innovative open-source dataset detailing 2019's global air traffic flows and associated CO2 emissions. A comprehensive approach that compiles diverse flight data sources is presented. The remaining data gaps are addressed by constructing a route network through systematic Wikipedia parsing and by estimating the related traffic using socio-economic data. Then, an aircraft performance model to estimate CO2 emissions is implemented. This methodology promises reinforced reproducibility and broader data accessibility in aviation environmental research. Several reference datasets are used to evaluate the accuracy of the open-source dataset. Despite various levels of accuracy for individual routes, major traffic flows are well estimated at the country and continental levels, albeit with room for refinement to ensure consistent data reliability. To facilitate the exploration of the dataset, the AeroSCOPE tool has been developed. To initiate research prospects, use cases of this dataset are proposed, concerning the network potential of electric and hydrogen-powered aircraft and inequalities in air transport.

Contrail optimization offers an efficient and cost-effective way for aviation to immediately reduce its climate impact. Open-source optimization, wherein the contrail and emission effects are balanced based on meteorological open data, has been presented in previous work. However, prior research overlooks the importance of using forecasting data, as opposed to post-processed reanalysis data. For contrail optimization to be implementable, forecasting data needs to be available at a sufficient quality in the flight planning stage in order to perform the optimization. In this paper, a fully open non-linear optimal control flight optimization is implemented and applied using both forecasting and reanalysis data. A total of 120 days (175.440 flights) of flight data from OpenSky are used in the analysis. We show that forecasts with larger lookahead times (up to 12 hours) are equally effective when compared to more recent forecasts (1 hour lookahead time) for contrail optimization, with equally high accuracy. However, when compared to more accurate post-processed reanalysis data, there are considerable differences in predicted contrails formed. This research shows there is still a long way to go before we can actually implement contrail optimal flight planning.

Recent developments in language models have created new opportunities in air traffic control studies. The current focus is primarily on text and language-based use cases. However, these language models may offer a higher potential impact in the air traffic control domain, thanks to their ability to interact with air traffic environments in an embodied agent form. They also provide a language-like reasoning capability to explain their decisions, which has been a significant roadblock for the implementation of automatic air traffic control. This paper investigates the application of a language model-based agent with function-calling and learning capabilities to resolve air traffic conflicts without human intervention. The main components of this research are foundational large language models, tools that allow the agent to interact with the simulator, and a new concept, the experience library. An innovative part of this research, the experience library, is a vector database that stores synthesized knowledge that agents have learned from interactions with the simulations and language models. To evaluate the performance of our language model-based agent, both open-source and closed-source models were tested. The results of our study reveal significant differences in performance across various configurations of the language model-based agents. The best-performing configuration was able to solve almost all 120 but one imminent conflict scenarios, including up to four aircraft at the same time. Most importantly, the agents are able to provide human-level text explanations on traffic situations and conflict resolution strategies.

Air traffic sector demand and capacity balancing are necessary for safe and efficient flight execution. Demand and capacity are determined in current operations based on schedules and flight plans. This research aims to improve air traffic demand forecasting by exploring machine learning-based trajectory prediction, specifically the newly emerged transformer-based neural network models. The predicted trajectories are considered to improve demand forecasts for Air Traffic Control in the Netherlands. We successfully built a transformer neural network using available traffic messages from the EuroControl B2B connection and actual trajectories obtained from the OpenSky ADS-B repository. A new lost function is specifically designed to improve this prediction model’s performance. This trajectory predictor could accurately generate trajectories, outperforming the flight plan and other neural network approaches by a good margin. For demand prediction, introducing improved trajectories provided small gains that could lead to more stable predictions.

Contrails, formed under specific atmospheric conditions, have a noteworthy role in heat-trapping within the atmosphere. This study bridges the gap between theoretical contrail formation models and real-world data by employing flight information from OpenSky and meteorological data from the European Centre for Medium-Range Weather Forecasts. We introduce a computationally efficient contrail estimation module, leveraging a client-server architecture that allows on-demand weather data interpolation via an API, significantly reducing computational load and enhancing performance locally. The study also benchmarks the entire pipeline, from data acquisition to contrail prediction, offering a robust tool for future air traffic studies requiring interpolated weather data.

Predicting aircraft Take-Off Weight (TOW) has been a long-standing goal for aviation stakeholders, especially for operational and regulatory bodies involved in flight planning. Accurate TOW values would enable better emissions computation, leading to more effective regulation of aviation’s climate impact. However, aircraft operators prefer to keep TOWs confidential because they are sensitive to operational trends and cost indices. Consequently, many works have attempted to circumvent this gap by predicting TOW values. Unfortunately, limited success has been achieved primarily due to the lack of accurate real-world operational data. This study is unique in utilizing operational TOW data provided by airlines. We predict TOW before takeoff based solely on Flight Plan and Terminal Aerodrome Forecast parameters, primarily focusing on flights at Amsterdam Airport Schiphol. The accuracy of several Machine Learning algorithms is directly compared. The best Mean Absolute Percentage Error of 2.17% on the Schiphol testing dataset is achieved. The model is further validated on flights at Paris- Charles de Gaulle Airport and Brussels South Charleroi Airport with errors of 4.07% and 3.41%. We found that the distribution of flights in the training dataset, particularly aircraft and airline types, significantly influenced the model’s applicability. Recommendations are also made on how to improve the model further.

Emissions and contrails are key factors in aviationinduced climate change, often presenting conflicting objectives in flight trajectory optimization. Previous research typically lacks in optimizer efficiency or in addressing these trade-offs, with limited use of extensive meteorological and flight data. In this paper, a fully open non-linear optimal control flight optimization approach is designed for contrail avoidance and emission reduction, with high computational efficiency. This is achieved by leveraging the most recent trajectory optimizer, OpenAP.TOP, and atmospheric data handling tool, fastmeteo. We present a new compound grid-based objective function that considers both contrails and emissions and introduce four different metrics for evaluating the performance. A total of four months’ worth of data, containing around half a million flights, are gathered from OpenSky for analyses. We show that high levels of contrail mitigation can be achieved, without significantly increasing flight time, distance, or emissions.

Current advancements in machine learning have provided new architectures, such as encoder-decoder transformers, for automatic speech recognition. For generic speech recognition, very high accuracies are already achievable. However, in air traffic control, automatic speech recognition models traditionally rely on domain-specific models constructed from limited training data. This study introduces this newly developed transformer model for air traffic control and provides a set of fully open automatic speech recognition models with high accuracies. This paper demonstrates how a large-scale, weakly supervised automatic speech recognition model, Whisper, is fine-tuned with various air traffic control datasets to improve model performance. We also evaluated the performance of different sizes of Whisper models. In the end, it was possible to achieve word error rates of 13.5% on the ATCO2 dataset and 1.17% on the ATCOSIM dataset with a random split (or 3.88% with speaker split). The study also reveals that finetuning with region-specific data can enhance performance by up to 60% in real-world scenarios. Finally, we have open-sourced the code base and the models for future research.

The study of the environmental transition of the aviation sector calls for prospective traffic scenarios. Detailed traffic and emissions inventories are often needed to refine the available analyses and to enable the simulation of regionalised scenarios. In the past studies, these are generally based on commercial, proprietary traffic data, making their dissemination problematic and reducing the reproducibility of the science produced. Open-source alternatives do exist, but with limited geographical coverage. This paper presents a method to aggregate different sources of flight information, in order to obtain an open-source air traffic dataset for 2019. Then, missing flight information is identified and completed using an airline route database built from Wikipedia parsing and related socio-economic data. After that, several reference datasets are used to evaluate the accuracy of the extended open-source dataset. Despite varying accuracy for different routes, major traffic flows are reasonably well estimated at the country and continental levels. Finally, the CO2 emissions are obtained using an existing aircraft performance surrogate model, and the accuracies are examined compared to the results from previous studies.