Atmospheric Stability Classification from Floating Doppler Wind Lidar Measurements: A Machine Learning Approach

This work studies the capabilities of the k-nearest-neighbours (KNN) method to estimate the atmospheric stability from floating Doppler wind lidar measurements of the vertical wind profile. A procedure to generate synthetic wind profiles to be used as training data for the KNN is presented. The method is validated with experimental data gathered during IJmuiden campaign by a floating Doppler wind lidar against a metmast as a reference. A global hit rate of 70.2% is obtained when comparing the predicted stability classes to the reference ones.