The Study And Application Of The Combined Receptor Models For Ambient Particulate Matter Source Apporitonment

Receptor model is a useful tool for air source apportionment study. Receptor model included two kinds:souce known model and source unknown model. Chemical mass balance model (CMB) is an important model for source known model which needs both of information of source and receptor; principal component analysis/multiple linear regression model (PCA/MLR) and positive matrix factorization model (PMF) belong to source unknown model which only need receptor information. There are defferent strengthen and weakeness for these models. Genarally, collinearity problem is an important problem for these models.Collinearity problem means that there were more than two source categories have similar source profiles. When the collinearity problem is presented, CMB model often obtains negative results; on the other hand, for PCA/MLR and PMF model, the collinear sources usually be extracted in one factor.According to our study, for CMB model, the compatibility between receptor and source is the key reason to resolve the collinearity problem. If the source and receptor are compatible, the acceptable results can be obtained by CMB model even if the collinearity problem was presented.In this study, principal component analysis/multiple linear regression-Chemical mass balance (PCA/MLR-CMB) combined modle and Nonnegative Constrained Principal Component Regression Chemical Mass Balance (NCPCRCMB) model are developed to resolve collienarity problem.In order to access the results of the combined models, the synthetic receptor datasets were developed and studied by combined model.For PCA/MLR-CMB combined model, the actual source profiles which obtained from real world were applied to construct the synthetic datasets. Among the actual source profiles, resuspended dust, soil dust and coal combustion are the collinear sources. So, if the synthetic datasets were studied by CMB model, negative results would obtained. The PCA/MLR-CMB combined model was applied to study the synthetic datasets. The estimated results by combined model were close to the true values.Next, the PCA/MLR-CMB combined model was applied to study the ambient datasets from Chengdu and Taiyuan cities. The difference between the results of PCA/MLR-CMB and CMB models were discussed. The negative results were obtained by CMB model due to the presenting of collinearity problem; while for PCA/MLR-CMB model, acceptable results were obtained. The conclusion indicates that the PCA/MLR-CMB model is feasible.For NCPCRCMB model,100 receptor profiles were developed by actual source profiles. The sources and receptor profiles randomly perturbed in order to make the sources and receptor incompatible. The synthetic receptors were studied by NCPCRCMB model, then the difference between estimated results and true values were discussed. The acceptable results were obtained.Next, the NCPCRCMB model were applied to study the ambient datasets from Xuxi, Yinchuan, Tianjin and Ji’nan. The acceptable results show that the NCPCRCMB model is feasible.