Rolling positive matrix factorisation for aerosol optical apportionment

This is the authors' abstract. We don't add key points for this paper.

Traditional aerosol optical apportionment methods, like the Aethalometer model and Multi-Wavelength Absorption Analyzer (MWAA) model, only rely on light absorption coefficients and work through stringent hypotheses on the number of acting sources and on the wavelength dependence of their absorption properties. In this paper, an improvement to optical apportionment is presented by adapting Positive Matrix Factorisation (PMF) to accept both aerosol absorption and scattering coefficients as input data. This made it possible to abandon the two main assumptions of current widespread aerosol optical apportionment models, namely the fixed number of absorbing sources and their absorption wavelength dependence, thus giving way to a more comprehensive approach where not only absorption, but scattering coefficients too, are contemplated and apportioned. The novel approach was tested on a 2-year long dataset of multi-wavelength absorption and scattering coefficients measured in Lecce, Italy. Four aerosol optical profiles were clearly identified and were interpreted as related to Black Carbon (BC), Brown Carbon (BrC), secondary aerosol, and coarse aerosol through validation against independently measured chemical composition variables.