Deriving boundary layer height from aerosol lidar using machine learning: KABL and ADABL algorithms

Abstract. The atmospheric boundary layer height (BLH) is a key parameter for many meteorological applications, including air quality forecasts. Several algorithms have been proposed to automatically estimate BLH from lidar backscatter profiles. However recent advances in computing enabled new approaches using machine learning that are seemingly well suited this problem. Machine can handle complex classification problems and be trained by human expert. This paper describes compares two machine-learning methods, the K-means unsupervised algorithm AdaBoost supervised algorithm, derive Atmospheric Boundary Layer (KABL) (ADABL) codes used study free open source. Both methods were compared reference BLHs derived colocated radiosonde data over 2-year period (2017–2018) at Météo-France operational network sites (Trappes Brest). A large discrepancy between root-mean-square error (RMSE) correlation with radiosondes was observed sites. At Trappes site, KABL ADABL outperformed manufacturer's while performance clearly reversed Brest site. We conclude promising (RMSE of 550 m Trappes, 800 manufacturer) but has training issues need resolved; lower Trappes) than much more versatile.