Optimization Study of Low-Altitude Turbulence Intensity Modeling Based On TKE-XGBoost

Abstract

To improve the accuracy of turbulence identification in low-altitude flight safety monitoring, a turbulence intensity modeling and optimization method based on Turbulent Kinetic Energy theory and the XGBoost model is proposed. Firstly, atmospheric stability is determined using the Richardson number. Subsequently, turbulence intensity is calculated by combining different stability conditions and Turbulent Kinetic Energy theory. The data for constructing physical model a originates from observations obtained by wind profile radar and microwave radiometer. Considering that temperature and humidity detection equipment like microwave radiometers are not always available in real-world scenarios, and in observation scenarios relying solely on wind profile radar, a Gradient Boosting Decision Tree algorithm is further introduced to construct a turbulence inversion model based exclusively on wind profile radar data. This model fully exploits the nonlinear relationships between multi-dimensional observational features such as radial velocity, velocity spectrum width, and signal-to-noise ratio of the radar and turbulence intensity. It is trained using the output of the benchmark model. Experimental results indicate that after optimizing the turbulence intensity calculation model b (based solely on wind profile radar) with model a, the model’s MSE decreases by 0.11, MAE decreases by 0.13, and the R² value increases by 0.28. This optimization process reduces reliance on auxiliary temperature and humidity data and effectively addresses the challenge of turbulence identification under limited observational information.