Linkages among U.S. Ozonesonde Profile Variability, Meteorology, and Surface Ozone Measurements based on Self-Organizing Map Clustering

Open Access
- Author:
- Stauffer, Ryan Michael
- Graduate Program:
- Meteorology
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- March 02, 2016
- Committee Members:
- Anne Mee Thompson, Dissertation Advisor/Co-Advisor
William Henry Brune, Committee Chair/Co-Chair
George Spencer Young, Committee Member
John Harlim, Committee Member - Keywords:
- Ozone
Radiosondes
Climatology
Meteorology - Abstract:
- Ozone (O3) profiles have been measured with high accuracy and precision by balloon-borne ozonesonde instruments since the 1960s. Since then, global ozonesonde networks have amassed vast data sets, compiling O3 climatologies with which to develop and validate satellite retrieval algorithms, and evaluate chemical transport model output. A typical way to formulate O3 climatologies is to average ozonesonde profiles on a monthly or seasonal basis, either for specific regions or zonally. Shortcomings of this simplistic averaging are quantified in this dissertation through use of a statistical clustering technique, self-organizing maps (SOM). First, the SOM algorithm and its application to ozonesonde profiles are detailed. Sensitivity tests comparing SOM with the similar k-means algorithm lead to the decision to employ SOM to ozonesonde data sets, and justified the choice of a 3x3 SOM with nine clusters. SOM is then applied to >4500 ozonesonde profiles from the long-term U.S. monitoring sites Boulder, CO; Huntsville, AL; Trinidad Head, CA; and Wallops Island, VA. The clusters of surface – 12 km amsl O3 mixing ratio data are closely linked to large-scale meteorological conditions. At all four sites, profiles in SOM clusters exhibit similar tropopause height, 500 hPa height and temperature, and total and tropospheric column O3. When profiles from each SOM cluster are compared to monthly O3 means, near-tropopause O3 in three of the clusters is double (over +100 ppbv) the climatological O3 mixing ratio. The three clusters include 13–16% of all profiles, mostly from winter and spring. Large mid-tropospheric deviations from monthly means (-6 ppbv, +7 – 10 ppbv O3 at 6 km) are found in two highly-populated clusters with a combined 36–39% of profiles. The latter clusters represent both distinctly polluted (summer) and clean O3 (fall-winter, high tropopause) profiles. Finally, we explore the capability of SOM clusters to identify high surface O3 events. Clusters of O3 profile data over a narrowed altitude range (surface – 6 km amsl) from Trinidad Head, CA, are linked to surface O3 monitor data at inland locations in CA. SOM clusters of O3 profile data from the lower troposphere can discriminate background vs polluted O3 and the conditions linked with each. Two of nine O3 clusters exhibit thin layers (~100s of m thick) of high O3, typically between 1 and 4 km, and residing above a subsidence inversion associated with a northern location of the semi-permanent Pacific subtropical high. Ozone in these clusters is upwind of high-altitude surface O3 monitors at inland locations. The surface O3 monitors at Lassen Volcanic and Yosemite National Parks, and Truckee, CA, display a marked impact of the elevated tropospheric O¬¬3. Days corresponding to the high O3 clusters exhibit hourly surface O3 anomalies at these sites of +5–10 ppbv compared to a climatology; the anomalies can last up to four days.