Tropospheric ozonesonde profiles at long‐term U.S. monitoring sites: 1. A climatology based on self‐organizing maps

Abstract

Sonde-based climatologies of tropospheric ozone (O) are vital for developing satellite retrieval algorithms and evaluating chemical transport model output. Typical O climatologies average measurements by latitude or region, and season. Recent analysis using self-organizing maps (SOM) to cluster ozonesondes from two tropical sites found clusters of O mixing ratio profiles are an excellent way to capture O variability and link meteorological influences to O profiles. Clusters correspond to distinct meteorological conditions, e.g. convection, subsidence, cloud cover, and transported pollution. Here, the SOM technique is extended to four long-term U.S. sites (Boulder, CO; Huntsville, AL; Trinidad Head, CA; Wallops Island, VA) with 4530 total profiles. Sensitivity tests on k-means algorithm and SOM justify use of 3×3 SOM (nine clusters). At each site, SOM clusters together O profiles with similar tropopause height, 500 hPa height/temperature, and amount of tropospheric and total column O. Cluster means are compared to monthly O climatologies. For all four sites, near-tropopause O is double (over +100 parts per billion by volume; ppbv) the monthly climatological O mixing ratio in three clusters that contain 13 - 16% of profiles, mostly in winter and spring. Large mid-tropospheric deviations from monthly means (-6 ppbv, +7 - 10 ppbv O at 6 km) are found in two of the most populated clusters (combined 36 - 39% of profiles). These two clusters contain distinctly polluted (summer) and clean O (fall-winter, high tropopause) profiles, respectively. As for tropical profiles previously analyzed with SOM, O averages are often poor representations of U.S. O profile statistics.

Tropospheric ozonesonde profiles at long‐term U.S. monitoring sites: 2. Links between Trinidad Head, CA, profile clusters and inland surface ozone measurements

“…The surface–6 km amsl monthly averaged O 3MR climatology from Trinidad Head, CA (August 1997–March 2015), is shown in Figure . In terms of O 3MR averages, Trinidad Head exhibits a smaller seasonal cycle than other CONUS ozonesonde sites [ Newchurch et al ., ; Stauffer et al ., ]. A maximum in free‐tropospheric O 3 is observed from April to August, when the combination of transported pollution and STE [e.g., Oltmans et al ., ; Lin et al ., ; Škerlak et al ., ] and photochemical O 3 production has the greatest impact on the site [ Newchurch et al ., ; Cooper et al ., ].…”

“…For more discussion on SOM, additional applications to O 3 profile data, and sensitivity tests comparing SOM and the similar k ‐means clustering algorithm, see Stauffer et al . [] (details in ). In the SOM application to O 3MR data, an array of initial nodes, which are comparable to cluster centroids and resemble O 3MR profiles, is defined.…”

“…This “neighborhood learning,” applied via the neighborhood function, tapers as the algorithm iterates and populates clusters. The neighborhood learning is dependent on a simple user‐chosen function that places similarly shaped clusters in adjacent positions in the SOM output (see Figure ) [ Stauffer et al ., , Figure 4]. After the user‐chosen number of iterations where the O 3MR profile data are input into the algorithm, the SOM converges to its solution as neighborhood learning tapers to zero.…”

Self-organized classification of boundary layer meteorology and associated characteristics of air quality in Beijing

“…Since Kohonen (1982) first proposed the SOM, it has been widely used for data downscaling and visualization in various disciplines (Jensen et al, 2012;Katurji et al, 2015;Dyson, 2015;Stauffer et al, 2016;Pearce et al, 2014;Jiang et al, 2017). In this study, the SOM is introduced to classify the ABL types through detecting topological relationships among the 5-year (2013-2017) radiosonde-based virtual potential temperature profiles.…”

“…Recently, self-organizing maps (SOMs; a feature-extracting technique based on an unsupervised machine learning algorithm) (Kohonen, 2001) were introduced to investigate the ABL thermodynamic structure, indicating the capabilities of SOMs in feature extraction from a large dataset of the ABL measurements (Katurji et al, 2015). In fact, the application of the SOM has increased in atmospheric and environmental sciences during the past several years (Jensen et al, 2012;Jiang et al, 2017;Gibson et al, 2016;Pearce et al, 2014;Stauffer et al, 2016), including a radiosonde-based application in South Africa (Dyson, 2015). However, there has been no SOM application in air-pollution-related ABL structure research thus far.…”