An Operational Early-Warning Model for Low-Level Turbulence Based on Radar Wind Profiler Observations

Low-level atmospheric turbulence impacts urban air quality and poses substantial risks to aviation safety, underscoring a critical demand for reliable short-term early warning. Radar wind profilers (RWPs) provide continuous vertical observations of wind fields, emerging as a key tool for low-level turbulence monitoring. However, existing turbulence forecasting relies mostly on diagnostic or retrieval approaches, which are often prone to systematic biases and not tailored for short-term operational early warning. In this study, we develop an operational early-warning model for low-level turbulence using a data-driven machine learning approach, leveraging RWP observations. Drawing on long-term observations at Shiyan Station in Shenzhen of Guangdong Province, we integrate RWP measurements with estimates of in situ turbulence dissipation rate (ε), as derived from ultrasonic anemometers installed at 160 and 320 m on a meteorological tower. A random forest model is trained to predict the mean ε in the subsequent 30 minutes, with predictors consisting of preceding half-hourly averaged RWP variables. The results demonstrate that the proposed model effectively captures the nonlinear relationship between RWP observations and ε, delivering robust predictive performance. The reliability of model exhibits a clear dependence on atmospheric and seasonal conditions. Under clear-sky scenarios and during autumn and winter, the model achieves enhanced stability, with correlation coefficients exceeding 0.7. In contrast, prediction uncertainty increases under complex conditions characterized by low wind speeds and weak turbulence signals. Further error analysis reveals systematic associations between prediction errors and observed spectral width, wind speed, and true turbulence intensity magnitude. Notably, applying threshold constraints to observed spectral width further improves the reliability of model predictions. This study provides a promising pathway to advancing turbulence monitoring and short-term forecasting in complex urban environments, with direct implications for low-level aviation safety.