Author(s) :

Charlotte Brard.

Volume/Issue :

Abstract :

Turbulence remains one of the most persistent and unpredictable hazards in aviation, yet little research has conducted an integrated comparative analysis of turbulence prediction systems across methodological categories. Most studies have evaluated individual models in isolation, leaving a gap in understanding how different approaches perform relative to one another. This study aimed to examine the strengths and limitations of four representative systems for turbulence forecasting: statistical learning using PCA with support vector classification, the GEKO turbulence model based on generalized k–ω equations, the sensor-driven FUTURA model for real-time anomaly detection, and the ECMWF ensemble-based Integrated Forecasting System. Each system was evaluated on four operational criteria: accuracy, adaptability, efficiency, and performance under adverse weather conditions. Results showed that GEKO achieved the highest accuracy but was computationally too demanding for real-time use. FUTURA excelled in adaptability and speed, although its alerts were limited to short-range predictions. The ECMWF ensemble system demonstrated strong coverage and reliability in adverse weather but suffered from delays. The statistical learning model produced balanced results but remained constrained by data sparsity. The findings suggest that no single model can address all operational needs, but combining complementary approaches into a hybrid framework offers the most effective pathway. This research contributes an evidence-based foundation for developing integrated turbulence forecasting systems to improve aviation safety, efficiency, and reliability.