Positive matrix factorization (PMF) is a widely used tool to do the source apportionment, compared with traditional PMF, the organic tracer based PMF has the advantage of providing more chemical information and resolve more specific source type. Although multiple studies incorporating organic tracers in PMF have been performed, very few cover comprehensive data set including other chemical components such as elements and major components.
Work in this thesis aims to improve the understanding of the atmospheric behavior of the organic tracers, and the application of them as source indicators in PMF. Several case studies have been performed using organic tracers-based source apportionment. Besides, the impact of time resolution is also examined using hourly measured aerosol components. The major findings are summarized below:
(1) Source apportionment of PM
2.5 was performed using PMF based on one-year 24-h offline filters collected at two sites in Macau in 2015. The chemical analysis included major inorganic ions, organic carbon and elemental carbon (OC and EC), elements and organic tracers including non-polar organic compounds (NPOCs) and sugar compounds. Trajectory analysis and conditional probability function (CPF)
analysis were performed on the PMF resolved factor contributions to distinguish the local and regional sources. The role of organic tracers was examined by comparing with two traditional PMF runs without organics. The results showed similar contributing sources, while noticeable differences of absolute PM
2.5 contributions from three PMFs were observed. Exclusion of organics overestimated source contributions from biomass burning and vehicle exhaust factor due to absence of source-specific tracers.
(2) Based on a multi-year and multi-site NPOC data set, spatial and seasonal variations, correlation analysis and ratio-ratio plots were used to investigate the source information and degradation of NPOC tracers. In summer, NPOCs showed distinct local emission characteristics, with urban sites having much higher concentrations than suburban sites. In winter, regional transport was an important influence on NPOC levels, driving up concentrations at all sampling sites and diminishing an urban-suburban spatial gradient. A PMF analysis of this large NPOC data set suggests that heavier NPOCs are more suitable source indicators than lighter NPOCs. Incorporating particle-phase light NPOC concentrations in PMF produces a separate factor, which primarily contains those light NPOCs and likely is not a source factor.
(3) We investigate the impact of inclusion of SOA tracers on the source apportionment of OC and PM
2.5 in the Pearl River Delta (PRD) region of China using PMF. In PMF runs incorporating SOA tracers (PMF
w), ten PMF factors were resolved including four secondary factors: (1) SOA_I (α-pinene, β-caryophyllene and naphthalene derived SOA), (2) SOA_II (isoprene derived SOA), (3) a secondary sulfate factor, and (4) a secondary nitrate factor. In PMF tests without SOA tracers (PMF
wo), the SOA_I and SOA_II factors could not be extracted while the remaining eight source factors were resolved. The PMF
w results indicated that SOC from specific precursors may have different formation pathways than secondary sulfate and nitrate formation processes and their source contributions could not be properly resolved without the indicative tracers included in PMF. This study demonstrates the utility of biogenic SOA tracers in resolving isoprene-derived SOA and highlights the need for more SOA tracers, especially those specific to anthropogenic precursors, in improving the source apportionment for those broad OA sources such as industrial emissions.
(4) We demonstrate with actual field data the benefit of using high-time resolution chemical speciation data in achieving more robust source apportionment of PM
2.5 using PMF. Hourly composition data were collected over a month in Shanghai,
including four inorganic ions, thirteen elements, organic and elemental carbon. PMF analysis of the hourly dataset (PMF
1h) resolves eight factors: secondary nitrate/sulfate, vehicular/industrial emissions, coal combustion, secondary sulfate, tire wear, Cr and Ni point source, residual oil combustion, and dust, with the first three being the major sources and each contributing to >20% of PM
2.5 mass. To characterize the benefit gained from time resolution, we carried out separate PMF analyses of 4-h and 6-h averaged data of the same dataset (PMF
6h and PMF
4h). PMF
6h and PMF
4h produce an eight-factor solution sharing similar factors to those by PMF
1h, but show less stability and more mixing in source profiles. Profile mixing was especially noticeable for tire wear, coal combustion and Cr and Ni point source in PMF
6h, as the 6-h averaging significantly decreased between-sample variability and increased rotational ambiguity. While the three sets of PMF solutions were similar in contributions for factors with major species as source markers (e.g., secondary nitrate/sulfate), larger variations existed for factors with trace species as markers due to mixing of major species in the profiles and higher rotational uncertainties in PMF
4h and PMF
6h. Our results indicate that hourly time series of elements and major components could achieve more robust source apportionment through better capturing of diurnal-scale dynamics in source activities.
Post a Comment