Source apportionment for PM2.5 and trace element bound PM2.5 using PMF and PCA at three sites in South Africa

Source apportionment by means of factorization is the traditional means of assigning of sources to factors in air pollution studies. Both PCA and DN-PMF have assumptions, strengths, and limitations. In common these two models have the assigning of sources to factors which can be subjective and dependent on limited information. PCA and DN-PMF is performed on data from three cities which were sampled at the same time, 16 April 2017 to 18 April 2018. The PCA will determine how many factors are assigned and the DN-PMF will determine the assigning of sources to factors. The geographical origin of the air particulates was modelled by means of HYsplit. The DN-PMF was able to give seasonal information to support the source apportionment. Results of the PCA included 6 factors for Thoyohandou and Pretoria and 7 factors for Cape Town. Where PCA produced strong statistical support to the number of factors chosen, correlations between HYsplit and DN-PMF and seasonal output corroborated the assigning of sources to factors. Utilising two models for factorization during source apportionment limits the error due to subjectivity where one method of PM_2.5 sampling was used.