CMAQ-based response modeling of speciated atmospheric mercury concentrations and depositions for efficient source contribution analysis

Atmospheric mercury (Hg) can undergo long-range transport before depositing into ecosystems, where it may be methylated and bioaccumulated in food webs, creating ecological and health concerns. High-resolution, process-based Hg source-receptor analyses are computationally intensive, limiting systematic evaluation of region-sector emission-control scenarios. To address this, atmospheric Hg response models (RMs) using a ResNetCNN-LSTM architecture trained on emissions, meteorological fields, and Community Multiscale Air Quality (CMAQ)-simulated speciated Hg concentrations and depositions. Coupling the RMs with iterative proportional fitting establishes an RM-IPF framework for efficient multi-region, multi-sector source contribution analysis. Out-of-sample tests showed that the RMs reproduced CMAQ-simulated, rather than directly observed, speciated Hg concentrations and depositions, with mean Pearson correlation coefficients exceeding 0.93 across all targets. Sensitivity analyses yielded response patterns consistent with key CMAQ processes, including dry deposition of gaseous elemental Hg, wet scavenging of reactive Hg, and chemical transformations among Hg species. By integrating RM-derived regional constraints with CMAQ-simulated sectoral constraints, RM-IPF closely reproduced CMAQ brute-force source contribution estimates while avoiding exhaustive zero-out simulations. In Guangdong, China, RM-IPF identified cement production in northern Guangdong and waste incineration in the Pearl River Delta as the largest contributors to anthropogenic Hg concentration (22.96%) and deposition (24.36%), respectively. Distinct dominant contributors highlight that emission control priorities depend on management objectives. This framework enables rapid scenario assessment and source-receptor analysis, supporting targeted Hg emission controls under the Minamata Convention on Mercury.