Constructing Ozone Profile Climatologies with Self-Organizing Maps: Illustrations with CONUS Ozonesonde Data

Abstract:

Ozone (O₃) trends analysis is typically performed with monthly or seasonal averages. Although this approach works well for stratospheric or total O₃, uncertainties in tropospheric O₃ amounts may be large due to rapid meteorological changes near the tropopause and in the lower free troposphere (LFT) where pollution has a days-weeks lifetime. We use self-organizing maps (SOM), a clustering technique, as an alternative for creating tropospheric climatologies from O₃ soundings. In a previous study of 900 tropical ozonesondes, clusters representing >40% of profiles deviated > 1-sigma from mean O^­₃. Here SOM are based on 15 years of data from four sites in the contiguous US (CONUS; Boulder, CO; Huntsville, AL; Trinidad Head, CA; Wallops Island, VA). Ozone profiles from 2 – 12 km are used to evaluate the impact of tropopause variability on climatology; 2 – 6 km O₃ profile segments are used for the LFT. Near-tropopause O^­₃ is twice the mean O^­₃ mixing ratio in three clusters of 2 – 12 km O₃, representing > 15% of profiles at each site. Large mid and lower-tropospheric O₃ deviations from monthly means are found in clusters of both 2 – 12 and 2 – 6 km O₃. Positive offsets result from pollution and stratosphere-to-troposphere exchange. In the LFT the lowest tropospheric O₃ is associated with subtropical air. Some clusters include profiles with common seasonality but other factors, e.g., tropopause height or LFT column amount, characterize other SOM nodes. Thus, as for tropical profiles, CONUS O^­₃ averages can be a poor choice for a climatology.