[AAS03-P13] Modeling the source-receptor relationships in highly polluted regions
キーワード:air pollution, source-receptor , emission
Efficient prediction of air pollution responses to precursor emissions is a prerequisite for an integrated assessment system in developing effective control policies. Representing the nonlinear response of air pollutants to emissions control with accuracy remains a major challenge for policymakers and modelers. Here, we propose a novel method by applying the deep-learning approach in atmospheric chemical transport model (CTM), and employ the Extended Response Surface Modeling (ERSM) technique to develop response surfaces of PM2.5 and O3 concentrations as a function of pollutant emissions. The proposed method significantly enhances the computational efficiency and practicality by requiring CTM simulations with mean and 95th maximal normalized errors within 5% and 10% across all grid cells. With the newly developed method, we systematically investigate the nonlinear response of PM2.5 and O3 to emissions of multiple pollutants from different sources in three highly polluted regions in China including the Northern China Plain (NCP), Fen-Wei Plain (FWP), and Chuan Yu region (CYR). We found that the effectiveness of NOx controls for reducing PM2.5 and O3 are largely influenced by the ambient levels of NH3 and VOC, exhibiting strong nonlinearities characterized as NH3-limited/-poor and NOx-/VOC-limited conditions, respectively. The contribution from regional sources becomes larger during more polluted episodes, suggesting the importance of joint controls on regional sources for reducing the heavy air pollution.