Unsupervised Classification of Ozone Profiles From Ozonesonde Data
Location
College of Science and Mathematics (COSM)
Session Format
Poster Presentation
Co-Presenters and Faculty Mentors or Advisors
Dr. Dan Jones, Faculty advisor
Dr. Yan Wu, Faculty Advisor
Abstract
The distribution of ozone in Earth’s atmosphere is relevant for both human health and for the operation of the climate system, in part because ozone absorbs powerful ultraviolet radiation from the Sun, changes the radiative balance throughout the atmosphere, and is connected to strong circulation patterns. In this study, we use Gaussian Mixture Modelling (GMM), an unsupervised classification technique, to identify coherent regimes of vertical ozone structure. We applied GMM to observational data derived from ozonesondes, identifying 17 spatially-coherent groups (a.k.a. classes) of ozone structure without using any latitude or longitude information. The classes identified by GMM have a variety of structures; polar and subpolar classes have a lower-altitude tropopause and low concentrations of tropospheric ozone, and the tropical classes have a higher-altitude tropopause. Northern Hemispheric high-latitude classes have higher stratospheric ozone than Southern Hemispheric high-latitude classes. Classes in the region of frequent wildfire/biomass burning have very high tropospheric ozone. This work shows that GMM can be used to identify coherent patterns in an observational dataset as a complement to present classification techniques.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Presentation Type and Release Option
Presentation (Open Access)
Recommended Citation
Fahrin, Fouzia, "Unsupervised Classification of Ozone Profiles From Ozonesonde Data" (2021). GS4 Georgia Southern Student Scholars Symposium. 57.
https://digitalcommons.georgiasouthern.edu/research_symposium/2021/2021/57
Unsupervised Classification of Ozone Profiles From Ozonesonde Data
College of Science and Mathematics (COSM)
The distribution of ozone in Earth’s atmosphere is relevant for both human health and for the operation of the climate system, in part because ozone absorbs powerful ultraviolet radiation from the Sun, changes the radiative balance throughout the atmosphere, and is connected to strong circulation patterns. In this study, we use Gaussian Mixture Modelling (GMM), an unsupervised classification technique, to identify coherent regimes of vertical ozone structure. We applied GMM to observational data derived from ozonesondes, identifying 17 spatially-coherent groups (a.k.a. classes) of ozone structure without using any latitude or longitude information. The classes identified by GMM have a variety of structures; polar and subpolar classes have a lower-altitude tropopause and low concentrations of tropospheric ozone, and the tropical classes have a higher-altitude tropopause. Northern Hemispheric high-latitude classes have higher stratospheric ozone than Southern Hemispheric high-latitude classes. Classes in the region of frequent wildfire/biomass burning have very high tropospheric ozone. This work shows that GMM can be used to identify coherent patterns in an observational dataset as a complement to present classification techniques.