Regional CO 2 and CH 4 inversion system using WRF-Chem (v4.4)/DART (v9.8.0) and continuous high-precision observations over the Korean Peninsula

Abstract

Abstract. Quantifying greenhouse gas (GHG) emissions over complex terrain remains a significant challenge for conventional inversion systems due to the high sensitivity of tracer transport to surface heterogeneity. We develop a high-resolution dual-species GHG top-down inversion framework by integrating the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem v4.4) and the Data Assimilation Research Testbed (DART v9.8.0) in a cycling ensemble Kalman filter system. Built on community tools, the framework is designed to be portable and configurable, enabling applications to other regions, resolutions, and observing-system configurations (e.g., expanded surface networks and additional data streams). This framework jointly assimilates near-surface CO2 and CH4 concentrations to produce dynamically consistent updates of emissions. By employing a unified Eulerian framework that simultaneously updates meteorology and 3-D tracer fields, the system is designed to maintain consistency between transport and concentrations and to reduce the risk that transport-related concentration errors are aliased into flux adjustments over the complex landscapes of the Korean Peninsula. To improve the simulation of turbulent GHG dispersion in the atmospheric boundary layer over complex terrain, we incorporate surface heterogeneity parameterizations (roughness sublayer and canopy height) into the model physics in the inversion system. The system assimilates high-precision continuous in situ observations from three World Meteorological Organization/Global Atmosphere Watch (WMO/GAW) stations to constrain CO2 and CH4 emissions. Prior flux estimates include anthropogenic emissions from the Emissions Database for Global Atmospheric Research (EDGAR v8.0), biogenic exchanges (the region-optimized Vegetation Photosynthesis and Respiration Model), biomass burning (Fire Inventory from the National Center for Atmospheric Research v2.5), and oceanic CO2 exchanges (SeaFlux). In a 2020 case study, the top-down estimates improve the agreement with ground observations, reducing root-mean-square errors by 30 %–60 % and lowering posterior mean bias to 1–2 ppm for CO2 and 20–30 ppb for CH4 at the high-precision surface observation sites. Independent aircraft profiles provide external evaluation and indicate residual CH4 discrepancies consistent with prior emissions and boundary condition uncertainties. Controlled observing-system simulation experiments show that, under prescribed perturbations, the system produces bounded and interpretable emission responses to transport-model, boundary-condition, and observation-error perturbations. They also indicate that recovery is strongest within station footprints and remains coverage-limited under the current three-station network, while a dense-network known-truth experiment highlights the value of expanded observational coverage for improving domain-wide emission constraints. Posterior adjustments suggest reduced CO2 emissions over the Seoul Metropolitan Area and parts of the western coastal region and increased CH4 emissions over inland agricultural source regions relative to the priors, highlighting priorities for follow-on evaluation of inventory components. Our posterior CO2 total (620 ± 45 Mt yr−1) is consistent with the Republic of Korea Biennial Transparency Report (ROK-BTR) estimate (624 Mt yr−1) at the national scale, while the posterior CH4 total (54.7 ± 5.2 Mt CO2eq yr−1) exceeds the ROK-BTR estimate (35.5 Mt CO2eq yr−1) by 19.2 Mt CO2eq yr−1, consistent with the larger structural uncertainty in CH4 source characterization and spatial allocation noted in previous studies.