Despite their low concentration levels in airborne fine particulate matter (PM_2.5), transition metals (e.g., Cu, Mn, and Fe) have received significant attention due to their toxicity and catalytical capability. Laboratory studies have demonstrated that Cu and Fe, once inhaled into lungs, can induce the generation of harmful reactive oxygen species. Furthermore, these catalytic metals are also recognized to enhance the formation of secondary sulfate in the atmosphere. To mitigate the harmful effects, it is crucial to gain knowledge about their sources and spatial distributions.

Traditional air quality models rely on emission inventories of pollutants, while the current inventories of trace metals are compromised with large uncertainties, resulting in model predictions that are incongruent with field observations. Efforts have been made to develop “hybrid models” that incorporate ambient measurements into air quality models to reconcile the disparities. Building on previous studies, we construct a new observation-constrained hybrid model that enables a more accurate and spatially informative simulation of trace metals and other primary components in PM_2.5.

The observation data required for the hybrid model are the mass concentrations of PM_2.5, major species such as OC, EC, sulfate, nitrate, and ammonium, as well as trace metals. These data are generally available for air quality monitoring sites in PM_2.5 speciation networks. Additional inputs to the model are the source apportionment results of primary PM_2.5 (PPM_2.5) from regional air quality models, such as The Community Multiscale Air Quality Modeling System (CMAQ). The hybrid model adjusts PPM2.5 results to improve their alignment with the observations. After carefully reviewing past methods, we also implemented in the hybrid model two new techniques, which were validated and published elsewhere. The hybrid model and its evaluation are described schematically in Figure 1.

Using observations from ten air quality monitoring sites, we analyze eight trace metals and other primary species in PM_2.5 in the Pearl River Delta (PRD) region of China in 2015. During the cross-validation stage, our new model demonstrated better prediction accuracy compared to three existing methods, and model errors were acceptable for most of the analyzed species.

As examples, we provide a brief summary of the modeling results for Cu and Mn. Our model reveals that area source sector (31%), industrial emissions (27%), and power generation (20%) contribute to over 75% of the ambient concentration of Cu in the PRD region. In some districts of Guangzhou, Foshan, Dongguan, and Zhongshan, Cu concentration can exceed 50 ng m^-3 (Fiure 2). Meanwhile, area sources (40%), power generation (17%), and marine vessel emission (13%) are the primary sources of Mn. In heavily polluted areas, ambient concentration of Mn can exceed 30 ng m^-3. These high concentrations of Cu and Mn raise concerns about the potential toxicity of aerosols. Similar analyses were conducted for other trace metals and primary species.

With the establishment of increasing number of PM_2.5 speciation monitoring stations in China and other countries, we expect to see a wider application of this novel hybrid model in analyzing the source contributions and spatial distribution of trace metals and other primary species in PM_2.5.