GAW-qc: A data science-based dashboard for quality control of atmospheric composition measurements

The Global Atmosphere Watch (GAW) Programme of the World Meteorological Organization coordinates a worldwide network of hundreds of ground-based in-situ monitoring stations that provide reliable scientific data on the chemical composition of the atmosphere. In the framework of the GAW Programme, the Quality Assurance/Scientific Activity Centre at Empa has developed an interactive dashboard based on data science to support station operators in timely detecting issues in their in-situ measurements of various trace gases.

The application (GAW-qc), currently in beta testing, makes use of a mixture of purely data-driven and hybrid anomaly detection techniques. It exploits historical measurements made at the target station as well as the archive of gridded numerical forecasts by the Copernicus Atmosphere Monitoring Service (CAMS). The accuracy of the latter for the specific site is improved through machine learning using various predictors, including meteorological parameters and aerosol concentrations.

GAW-qc allows station operators to upload their latest measurements, visualize the data with different temporal aggregations, and easily detect anomalous values using just their internet browser. By combining the information gathered from the dashboard with logbook entries and local expertise, they can effectively flag problematic measurements and even detect instrumental issues that would remain unnoticed otherwise. First case studies indicate that this process can indeed facilitate the detection of malfunctionings in the analytical setup and reduce the ingestion of erroneous data into the international data repositories. Moreover, it has the potential to shorten data gaps if applied timely. Therefore, it may become a game-changer towards reliable, comparable and traceable world-wide datasets in the field of air quality and greenhouse gases. The software is freely available through a GitHub repository and can be adapted to analyze other atmospheric variables.