Unveiling PM2.5 sources: Double and tracer conjugate PMF approaches for high-resolution organic, BC, and inorganic PM2.5 data

Abstract

A number of recent source apportionment studies have explored high-time resolution organic particulate matter, elemental particulate matter (PM), and Black Carbon (BC) datasets and attributed them independently to specific sources. However, proper unmixing of the actual sources operational on the ground most of the time cannot be achieved based on such an independent source apportionment approach, especially owing to the presence of secondary aerosol factors. Therefore, a combined analysis of all major PM_2.5 species is needed to better recognize the actual physical sources. Accordingly, in this study, using a combined dataset consisting of non-refractory PM_2.5 organic factors/major m/z signals from organics, elements, and BC, we evaluated two disparate factor analytic methodologies – namely, double-PMF (D-PMF) and Tracer-conjugate PMF (TC-PMF), to apportion PM_2.5 sources in Delhi winter (from December 15, 2020 to February 28, 2021) through real-time instrumentation (ACSM, Xact, and Aethalometer(AXA)). During the study period, the average PM_2.5 concentration was 182 μg/m³ (C-PM_2.5 = sum of NR-PM_2.5 (Organics, NO₃⁻, SO₄⁻², NH₄⁺), BC, and elements). For D-PMF, organic aerosols (OA) were initially deconvolved with positive matrix factorization (PMF) into hydrocarbon-like OA (HOA), biomass burning OA (BBOA), low volatile oxidized OA (LVOOA 1 and 2) and semi-volatile oxidized organic aerosols (SVOOA) before being coupled with elemental species and BC for a second PMF. The TC-PMF combined the major m/z signals from organics with the elemental species and BC. Both D-PMF and TC-PMF identified biomass burning, industrial, waste incineration, dust-related, traffic, secondary chloride, Pb-rich, power plant, and LVOOA dominated as the sources. Both solutions (D-PMF, TC-PMF) were found to be dominated by biomass burning (33.3% and 26.5%), followed by the power plant (27.4% and 18.4%) and the LVOOA dominant (14.2% and 18.6%) factors. The D-PMF and TC-PMF improved the interpretation of organic factor sources, such as apportioning considerable contributions of LVOOA2 (85%) to the power plant factor, which is often linked with regionally carried aged OA in the Organics PMF (O-PMF). Lastly, the D-PMF results significantly agreed with TC-PMF, indicating that either of the two techniques could be used to unmix the complex variety of PM_2.5 sources in the Delhi-NCR (National Capital Region) region and, arguably, the larger Indo-Gangetic Plains.