Cluster-based characterization of multi-dimensional tropospheric ozone variability in coastal regions: An analysis of lidar measurements and model results

Abstract

Coastal regions are susceptible to multiple complex dynamic and chemical mechanisms and emission sources that lead to frequently observed large tropospheric ozone variations. These large ozone variations occur on a mesoscale and have proven to be arduous to simulate using chemical transport models (CTMs). We present a clustering analysis of multi-dimensional measurements from ozone lidar in conjunction with both an offline GEOS-Chem chemical-transport model (CTM) simulation and the online GEOS-Chem simulation GEOS-CF, to investigate the vertical and temporal variability of coastal ozone during three recent air quality campaigns: 2017 Ozone Water-Land Environmental Transition Study (OWLETS)-1, 2018 OWLETS-2, and 2018 Long Island Sound Tropospheric Ozone Study (LISTOS). We developed and tested a clustering method that resulted in five ozone profile curtain clusters. The established five clusters all varied significantly in ozone magnitude vertically and temporally, which allowed us to characterize the coastal ozone behavior. The lidar clusters provided a simplified way to evaluate the two CTMs for their performance of diverse coastal ozone cases. An overall evaluation of the models reveals good agreement (R≈0.70) in the low-level altitude range (0 to 2000gm), with a low and unsystematic bias for GEOS-Chem and a high systemic positive bias for GEOS-CF. The mid-level (2000-4000gm) performances show a high systematic negative bias for GEOS-Chem and an overall low unsystematic bias for GEOS-CF and a generally weak agreement to the lidar observations (RCombining double low line0.12 and 0.22, respectively). Evaluating cluster-by-cluster model performance reveals additional model insight that is overlooked in the overall model performance. Utilizing the full vertical and diurnal ozone distribution information specific to lidar measurements, this work provides new insights on model proficiency in complex coastal regions.