Improving Planetary Boundary Layer Height Estimation From Airborne Lidar Instruments

Abstract The height of the planetary boundary layer (PBLH) influences processes such as pollutant distributions, convection, and cloud formation within the troposphere. Aerosol observables play a critical role in deriving the mixed layer height (MLH) using retrieval techniques like the Haar wavelet covariance transform (WCT), which employs gradients in aerosol backscatter to estimate MLH. Currently, backscatter‐only approaches struggle with identifying very shallow stable boundary layers, distinguishing PBL from lofted residual or other aerosol layers, and profiles with very low aerosol loading. Here, we reflect on the WCT method's performance and evaluate different approaches to improve PBLH estimations. We aggregate lidar observables from recent NASA airborne field campaigns and compute MLHs based on the WCT method. Machine learning (ML) approaches are explored to produce PBLH estimates by training lidar information on thermodynamically derived PBLHs over marine and land settings. A linear model is found suitable for producing PBLH estimates in marine settings (improving mean bias by 71 m), while an ensemble tree method proves more suitable for PBLH types over land, as indicated by improved biases (13 m mean bias), errors (179 m mean error and 391 m RMSE), and correlations (+0.3) for the models explored. The algorithms are additionally tested on “unseen” data to gauge differences between MLH and PBLH estimates produced from each of the models. The PBLH estimates, composed of information from lidar and thermodynamic profiles, further support the use of ML for an automated method of PBLH prediction. Overall, these improved predictions will help evaluate models and deepen our understanding of PBL‐aerosol interactions.