Retrieval of Planetary Boundary Layer Height from Calipso Satellite: A Big Data and Machine Learning Approach

This study presents a Random Forest (RF) model trained on an extensive dataset to estimate the planetary boundary layer height (PBLH) retrieved from Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) on board of Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) Level 1 measurements during both daytime and nighttime. The novelty of the method lies in the RF’s capability to automatically classify the observed scenarios, enabling P BLH estimation in cloudy, multi-layer, and dust-laden conditions across diverse global regions. The RF model was trained using 10 years of CALIOP data, with radiosonde-derived P BLH retrievals coinciding with CALIPSO overpasses serving as ground truth. The results demonstrate that the RF model surpasses existing state-of-the-art methods in spatial and temporal coverage, achieving a correlation coefficient (R2 = 0.6) and root-mean-square error (RMSE) of 333.60 m, without requiring atmospheric typing, data filtering, or additional ancillary information.