Exploiting multi-wavelength aerosol absorption coefficients in a multi-time resolution source apportionment study to retrieve source-dependent absorption parameters
In this paper, a new methodology coupling aerosol optical and chemical parameters in the same source apportionment study is reported. In addition to results on source contributions, this approach provides information such as estimates for the atmospheric absorption Ångström exponent (α) of the sources and mass absorption cross sections (MACs) for fossil fuel emissions at different wavelengths. A multi-time resolution source apportionment study using the Multilinear Engine (ME-2) was performed on a PM10 dataset with different time resolutions (24, 12, and 1 h) collected during two different seasons in Milan (Italy) in 2016. Samples were optically analysed by an in-house polar photometer to retrieve the aerosol absorption coefficient bap (in Mm−1) at four wavelengths (λ=405, 532, 635, and 780 nm) and were chemically characterized for elements, ions, levoglucosan, and carbonaceous components. The dataset joining chemically speciated and optical data was the input for the multi-time resolution receptor model; this approach was proven to strengthen the identification of sources, thus being particularly useful when important chemical markers (e.g. levoglucosan, elemental carbon) are not available. The final solution consisted of eight factors (nitrate, sulfate, resuspended dust, biomass burning, construction works, traffic, industry, aged sea salt); the implemented constraints led to a better physical description of factors and the bootstrap analysis supported the goodness of the solution. As for bap apportionment, consistent with what was expected, biomass burning and traffic were the main contributors to aerosol absorption in the atmosphere. A relevant feature of the approach proposed in this work is the possibility of retrieving a lot of other information about optical parameters; for example, in contrast to the more traditional approach used by optical source apportionment models, here we obtained source-dependent α values without any a priori assumption (α biomass burning =1.83 and α fossil fuels =0.80). In addition, the MACs estimated for fossil fuel emissions were consistent with literature values. It is worth noting that the approach presented here can also be applied using more common receptor models (e.g. EPA PMF instead of multi-time resolution ME-2) if the dataset comprises variables with the same time resolution as well as optical data retrieved by widespread instrumentation (e.g. an Aethalometer instead of in-house instrumentation).