Cluster Analysis: A New Approach Applied to Lidar Measurements for Atmospheric Boundary Layer Height Estimation

Abstract

Abstract Several procedures are widely applied to estimate the atmospheric boundary layer (ABL) top height by using aerosols as tracers from lidar measurements. These methods represent different mathematical approaches, relying on either the abrupt step of the aerosol concentration between the ABL and the free troposphere (FT) or the statistical analysis of vertical variations of the aerosol concentration. An alternative method—the cluster analysis (CA)—has been applied to lidar measurements for the first time, emerging as a useful and robust approach for calculating the ABL height, taking the advantage of both previous variables: the vertical aerosol distribution as obtained from the lidar range-corrected signal (RCS) and the statistical analysis of the RCS profiles in terms of its variance to determine a region of high aerosol loading variability. CA limitations under real situations are also tested, and the effects in ABL height determination of both noise and cloud contamination in RCS are examined. In particular, CA results are weakly sensitive to the signal noise due to the basic features of this statistical method. In addition, differences in the ABL top height, as estimated under cloudy and clear skies, have been found to be lower than 1.8% for a high RCS signal, while no effect is observed for low RCS cloud conditions. Moreover, the CA performance on the ABL top height determination for real cases is also presented, showing the reliable CA skills in reproducing the ABL evolution.