Automated identification of local contamination in remote atmospheric composition time series
A common challenge of atmospheric measurements in remote environments is to identify pollution from nearby activities that interfere with the purpose of the observations. Pollution, particularly from combustion, typically reveals itself in enhanced particle- , CO2 or CO concentrations and affects many atmospheric variables. It can vary in time scales from a few seconds to several hours. Here, we present an automated algorithm used to clean the year-long continuous (10s-time resolution) dataset of particle concentration measurements collected during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition onboard RV Polarstern. We identify pollution in our dataset based on the gradient, i.e., time derivative, of the particle number concentration. If this gradient exceeds a certain threshold, the data is flagged as polluted. We describe the performance of the algorithm and compare it to other commonly-used techniques. This method has two main advantages: It allows the detection of pollution from both stationary and non-stationary sources, and polluted periods can be identified without a need for other datasets (e.g., wind direction or CO2 concentration). This algorithm will be made open-source and user-friendly to allow wide use in the MOSAiC and larger atmospheric chemistry community.
2022-04-25
Python Algorithm Code
Event name | Event place | Event date |
Potsdam, Germany | April 25-29, 2022 | |
Relation | URL/DOI |
Is Supplemented By | |
IsSupplementedBy | |