Study On Three-way Source Apportionment Techniques Based On The Physical And Chemical Characteristics Of Particulate Matter

Source apportionment methods can identify the possible source categories and quantify the source contributions to particulate matters(PM)in ambient air.Receptor model,an important tool,plays a vital role on source apportionment works over the world.Factor analysis model,one kind of the receptor models,determines source categories and contributions based on a long-term observation(defied as receptor datasets).Understanding the physical and chemical characteristics of receptor is important for obtaining a good source apportionment result.Traditional source apportionment methods mainly deal with the two-way receptor datasets,containing chemical and temporal dimensions(the time series of species chemical concentrations from one sampling site).With the improvement of the sampling and chemical analysis technology,more three-way receptor datasets are constructed,such as the multi-sites three-way receptor dataset(including chemical,temporal and special dimensions)and the multi-sizes three-way receptor dataset(including chemical,temporal and size dimensions).To deal with the three-way datasets,some three way receptor models like PMF3 have been proposed.If the information of three-way datasets can be made full use of and chemical,temporal,spatial or sizes variability can be considered in the source apportionment,the effectiveness of factors would be improved.The objective of the paper is to develop advanced three-way source apportionment techniques based on the physical and chemical characteristics of three-way datasets,which could make full use of the three-way information and improve the accuracy.First,the physical and chemical characteristics of different three-way datasets were analyzed and the results exhibited that: for the multi-sites three-way receptor dataset,the source profiles are same but the contribution variations are different at diverse sites;for the multi-sizes three-way dataset,the contribution variations are same but the source profiles are different in diverse sizes.Then,the results of synthetic tests showed that traditional two-way and three-way factor analysis models can't make full use of the information among sites or sizes;or they can't deal with three-way datasets with various characteristics.On the basic of the characteristics of three-way datasets and studies on traditional models when dealing with three-way datasets,two kinds of advanced three-way factor analysis models were developed.For the multi-sites three-way datasets,the algorithm defined as the multi-contribution matrixes parallel factor analysis was newly proposed.And the advanced multi-sites three-way factor analysis(AAB three-way)model was developed which can estimate one source profile matrix but different source contributions matrixes for multi sites.For the multi-sizes three-way datasets,the algorithm defined as the multi-profile matrixes parallel factor analysis was newly proposed.And the advanced multi-sizes three-way factor analysis(ABB three-way)model was developed which can estimate one source contribution matrix but different source profile matrixes for multi sizes.The synthetic tests were used to evaluate the advanced models and demonstrated that the advanced three-way models could perform well under different variability of source profiles or source contributions.Additionally,ambient three-way datasets were obtained from Tianjin and Chengdu.The advanced three-way models were applied to study the ambient datasets and acceptable results were obtained.The feasible application and advantages of the models was analysis.This work can provide a new ideal to deal with the three-way receptor datasets,and present basic information for the using of three-way models in source apportionment works in the further.