CLASP: CLustering of Atmospheric Satellite Products and Its Applications in Feature Detection of Atmospheric Trace Gases.

Satellite instruments have the most potential of capturing trace gas variability as they continually observe the atmosphere and its composition over wide regions. Yet the increasingly large data size of satellite products poses a challenge for their use as traditional data processing methods (e.g., averaging) may not be effective to extract the spatiotemporal variability without prior knowledge of an emission source's spatial and temporal behavior, such as location, time, and plume shape. Here, an agile clustering algorithm entitled CLustering of Atmospheric Satellite Products (CLASP) is presented to identify the spatiotemporal variability of trace gases captured in satellite observations. We find the knowledge discovery method for large data sets, clustering, is suited for identifying the variability of trace gases in satellite observations, as such CLASP is rooted in density‐based clustering methods. CLASP detects features from satellite observations and identifies their spatial, magnitude, and temporal axis leading to a better understanding of the spatiotemporal variability of atmospheric trace gases. To test the applicability of CLASP, the algorithm is applied to TROPOspheric Monitoring Instrument NO2 observations illustrating some of its different capabilities. Implementing CLASP for event identification, capturing plume variability, and source detection, CLASP identified wildfires, observed disruptions from COVID‐19 lockdown restrictions, and detected irregular emissions from oil and gas operations. Plain Language Summary: Satellite instruments capture the location and timing of trace gases as they have daily observations of the atmosphere over a wide spatial coverage. With increasing spatial and temporal resolutions of satellite products, traditional data‐reducing methods such as averaging could be insufficient in analyzing increasingly large satellite observations. The user must have prior knowledge of the location and temporal behavior of emission sources in order to choose the suitable averaging intervals and such information is often lacking or unreliable for non‐urban and point sources. To alleviate this limitation, a clustering algorithm titled CLustering of Atmospheric Satellite Products (CLASP) is presented in this work. CLASP identifies features in satellite observations by their location, amount, and time. CLASP was applied to the satellite observational data set, TROPOspheric Monitoring Instrument NO2, to identify wildfires, observe COVID‐19 lockdown changes, and spot irregular emissions from oil and gas operations. Key Points: Density‐based clustering provides a foundation when built upon can describe the spatiotemporal variability of atmospheric trace gasesA clustering algorithm entitled CLustering of Atmospheric Satellite Products (CLASP) is presented to describe observations of trace gasesCLASP is applied to TROPOspheric Monitoring Instrument NO2 observations for event identification, plume variability, and source identification [ABSTRACT FROM AUTHOR]