Developing a statistical model to explain the observed decline of atmospheric mercury

Abstract Mercury is a ubiquitous environmental toxicant and a cause for global concern due to its persistence and bioaccumulation in the environment. Evaluating the effectiveness of mercury emission control has become a significant issue after the entry into force of the Minamata Convention on Mercury in 2017. Atmospheric mercury concentration is an important indicator for anthropogenic emission control. Although Eulerian models are generally applied to evaluate emission reduction and policy effectiveness, the uncertainty of mercury reaction mechanisms and the insufficient grid accuracy of simulations limit the applications of this method at particular sites. In this study, we applied a statistical approach (the Generalized Additive Model, GAM) to explain the decline of atmospheric mercury concentration in Beijing, China, which followed a trend (Sen's slope) of −0.37 ng m−3 yr−1 (−8.0% yr−1). The statistical model represented 56.5% of the variance in mercury concentration and the adjusted R2 reached 0.547. Reduction of anthropogenic mercury emission, variation in meteorological condition, and change in global background level explained 51.5%, 47.1%, and 1.4% of the decrease of air mercury concentration, respectively. We validated the results using Hg emission inventories, seasonal Hg/CO value, and meteorological data. Considering the limitations of Eulerian models and the simplicity of statistical models, we suggest the application of GAM as an assessment method for long-term variation of atmospheric mercury.