Review of online source apportionment research based on observation for ambient particulate matter

Particulate matter source apportionment (SA) is the basis and premise for preventing and controlling haze pollution scientifically and effectively. Traditional offline SA methods lack the capability of handling the rapid changing pollution sources during heavy air pollution periods. With the development of multiple online observation techniques, online SA of particulate matter can now be realized with high temporal resolution, stable and reliable continuous observation data on particle compositions. Here, we start with a summary of online measuring instruments for monitoring particulate matters that contains both online mass concentration (online MC) measurement, and online mass spectrometric (online MS) techniques. The former technique collects ambient particulate matter onto filter membrane and measures the concentrations of chemical components in the particulate matter subsequently. The latter technique could be further divided into two categories: bulk measurement and single particle measurement. Aerosol Mass Spectrometers (AMS) could provide mass spectral information of chemical components of non-refractory aerosols, especially organic aerosols. While the emergence of single-particle aerosol mass spectrometer (SPAMS) technology can provide large number of high time resolution data for online source resolution. This is closely followed by an overview of the methods and results of SA. However, online instruments are still facing challenges, such as abnormal or missing measurements, that could impact the accuracy of online dataset. Machine leaning algorithm are suited for processing the large amount of online observation data, which could be further considered. In addition, the key research challenges and future directions are presented including the integration of online dataset from different online instruments, the ensemble-trained source apportionment approach, and the quantification of source-category-specific human health risk based on online instrumentation and SA methods.