Bayesian Inference Approach to Quantify Primary and Secondary Organic Carbon in Fine Particulate Matter Using Major Species Measurements.

Abstract

The determination of primary organic carbon (POC) and secondary organic carbon (SOC) in fine particulate matter using ambient measurements is essential in atmospheric chemistry. A novel Bayesian inference (BI) approach is proposed to achieve such quantification using only major component measurement data and tested in two case studies. One case study composes of filter-based daily compositional data made in the Pearl River Delta region, China, during 2012, while the other uses online measurement data recorded at the Dianshan Lake monitoring site in Shanghai in wintertime 2019. Source-specific organic trace measurement data are available in both the cases so that positive matrix factorization (PMF) analysis is performed, where PMF-resolved POC and SOC are used as the best available reference values for model evaluation. Meanwhile, traditional techniques, i.e., minimum ratio value, minimum R squared, and multiple linear regression, are also employed and evaluated. For both the cases, the BI models have shown significant advantages in accurately estimating POC and SOC amounts over conventional methods. Further analysis suggests that using sulfate as the SOC tracer in BI model gives the best model performance. This methodological advance provides an improved and practical tool to derive POC and SOC levels for addressing PM-related environmental impacts.