Estimation of the atmospheric boundary layer height during different atmospheric conditions: a comparison on reliability of several methods applied to lidar measurements

The performance of six numerical methods usually used to determine the atmospheric boundary layer (ABL) height from lidar measurements was investigated under different atmospheric conditions: results were compared with those obtained from radiosoundings to analyse their reliability for ABL-height retrievals. The selected methods were the gradient method (GM), the logarithm gradient method (LGM), the inflection point method (IPM), the wavelet covariance transform (WCT), the centroid/variance method (VM), and the cluster analysis (CA). Lidar measurements were carried out in the frame of the ‘Atmospheric Minor Species relevant to the Ozone Chemistry’ (AMISOC) project during a multi-instrument campaign conducted at the INTA/Atmospheric Observatory ‘El Arenosillo’ (INTA/ARN) in south-western Spain from 15 May to 20 June 2012. The goal of this work is to analyse the performance and robustness of the different lidar methods in this region, characterized by particular atmospheric conditions. In particular, both events of sea–land breeze regimes and episodes of Saharan dust intrusions were studied. In most days, similar results were obtained by all lidar methods in the events of sea–land breeze regimes, presenting relative absolute differences between lidar and radiosounding retrievals below 12% in average. However, big discrepancies between lidar and radiosounding retrievals are found when residual layers are present in the measurements. In such cases, the vertical extension of lidar and radiosounding profiles must to be limited to the altitude of the residual layer bottom. In a second analysis, focused on diurnal variability in the ABL heights under non-dusty (ND) and dusty (DD) conditions, the methods were tested against intensive radiosoundings launched every 4 h over 2 days. Under ND conditions, the best results were achieved for the LGM, presenting a mean of the relative absolute differences respect to radiosounding measurements of 10%. The rest of methods also provided good results with relative differences below 20% in average. Under DD conditions, however, an increase of the relative differences is found with mean values of up 32%. In this case, best results are given by CA with a mean relative difference of 20%. Despite the limited data set used in this work, results show that unlike the ND conditions for which all lidar methods provide good results respect to radiosounding retrievals, under DD conditions the election of the lidar method is a key factor for ABL estimation. However, we remark the need of extending our analysis to longer periods of time to better characterize the differences observed in this work.