Caicedo et al. (2025) Enhanced Boundary Layer Height Detection Using Ceilometer, Surface Meteorology, and Radiation Products With a Random Forest Ensemble Method

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Journal of Geophysical Research Atmospheres
• Year: 2025
• Date: 2025-11-17
• Authors: Vanessa Caicedo, Joseph Sedlar, Laura Riihimaki, W. M. Angevine, David D. Turner, Damao Zhang, Kathleen Lantz
• DOI: 10.1029/2025jd043385

Research Groups

• Atmospheric Radiation Measurement Southern Great Plains (ARM SGP) user facility
• NOAA Surface Radiation (SURFRAD) Network

Short Summary

This study develops and evaluates a Random Forest (RF) model for estimating planetary boundary layer height (PBLH) using diverse atmospheric measurements, demonstrating its superior accuracy and robustness compared to traditional methods, particularly during daytime and transition periods.

Objective

• To develop and evaluate a Random Forest (RF) model for estimating planetary boundary layer height (PBLH) by integrating ceilometer, surface meteorology, and radiation measurements, trained with radiosonde-derived thermodynamic PBLH estimates.
• To bridge gaps between aerosol-based and thermodynamic-based PBLH estimates.

Study Configuration

• Spatial Scale: Point-based (Atmospheric Radiation Measurement Southern Great Plains (ARM SGP) user facility and a second evaluation site).
• Temporal Scale: 9 years of data.

Methodology and Data

• Models used: Random Forest (RF), Haar Wavelet (HW), Vaisala BL-View software.
• Data sources: Ceilometer measurements, surface meteorological measurements, radiation measurements, radiosonde data (for thermodynamic PBLH estimates used in training).

Main Results

• The Random Forest (RF) model significantly outperformed traditional Haar Wavelet (HW) and Vaisala BL-View software methods for PBLH estimation, especially during daytime and transition periods.
• At the ARM SGP site, the RF model achieved a bias of -4.9 m and a root mean square error (RMSE) of 303.2 m, which was substantially lower than HW (bias 70.9 m, RMSE 404.6 m) and BL-View (bias 124.1 m, RMSE 566.9 m).
• At a second evaluation site, RF achieved the lowest overall RMSE (323.7 m), comparable to HW (326.4 m) and significantly better than BL-View (738.3 m).
• All models showed reduced accuracy under stable nighttime conditions.
• Key predictors for the RF model included the lifting condensation level height (LCLH), aerosol gradients, and month (for seasonal variability).

Contributions

• Development of a robust machine learning (Random Forest) approach for PBLH estimation that integrates multiple observational data sets (ceilometer, surface meteorology, radiation).
• Demonstrated improved accuracy and robustness of the RF model over traditional methods, particularly during daytime and transition periods, with substantial reductions in bias and RMSE.
• Showcased the potential of machine learning to bridge the gap between aerosol-based and thermodynamic-based PBLH estimates.

Funding

• Not specified in the abstract.

Citation

@article{Caicedo2025Enhanced,
  author = {Caicedo, Vanessa and Sedlar, Joseph and Riihimaki, Laura and Angevine, W. M. and Turner, David D. and Zhang, Damao and Lantz, Kathleen},
  title = {Enhanced Boundary Layer Height Detection Using Ceilometer, Surface Meteorology, and Radiation Products With a Random Forest Ensemble Method},
  journal = {Journal of Geophysical Research Atmospheres},
  year = {2025},
  doi = {10.1029/2025jd043385},
  url = {https://doi.org/10.1029/2025jd043385}
}