The Prospective Application of Machine Learning in Turbulence Forecasting over China

Abstract

Abstract This study analyzes the spatiotemporal distribution of turbulence in China from 2020 to 2022 using pilot reports. Results reveal a higher frequency of moderate-to-severe turbulence during spring and winter, particularly in January. Spatially, the primary regions of turbulence occurrence are eastern China, Xinjiang Province, Sichuan Province, and the Qiongzhou Strait, with a focus on altitudes at or above 6000 m. Machine learning models, especially random forest and extreme gradient boosting (XGBoost), demonstrate high accuracy in turbulence prediction, notably for high-altitude events. The random forest model shows optimal performance in winter, achieving an area under the curve of 0.92. The study highlights the importance of thermally related diagnostics, indicating a significant presence of convectively induced turbulence in high-altitude turbulence events. This research not only deepens the understanding of turbulence dynamics in the China region but also underscores the potential of machine learning in enhancing turbulence forecasting.