An accurate estimation of anthropogenic emission is primarily important in improving air quality forecast skills using an atmospheric chemistry-transport model. An inventory-based bottom-up estimation method is applied based on the CAPSS inventory for a national anthropogenic emission, which is used in an operational air quality modeling system. However, accuracy of the estimated emission has not been quantitatively verified. This study applied a Lagrangian inverse modeling technique based on the Bayes’ probabilistic theory to quantitatively verify the inventory-based carbon monoxide (CO) emission over the Seoul metropolitan area. The anthropogenic emission of MICS-ASIA 2010 (Model Inter-Comparison Study for ASIA 2010) was verified using surface CO concentrations measured from routine air quality monitoring sites. The coupled WRF-VUCM (Weather Research and Forecasting-Vegetated Urban Canopy Model) was used to simulate meteorology during a month period of April 2010 which is provided to the Lagrangian dispersion modeling system of WRF-FLEXPART (WRF-FLEXible PARTicle dispersion model). Sensitivity experiments were conducted to investigate quantitative uncertainty in calculating atmospheric concentration from Lagrangian modeling. Simulated meteorology was evaluated statistically against surface and upper observed meteorological data, and it was found that the northwesterly wind relevant to high-pressure system in China and the local sea breeze are important meteorological factors that significantly influence on the atmospheric concentration over the Seoul metropolitan area. The top-down verification results from application of the Lagrangian inverse modeling technique show that the anthropogenic emissions are underestimated by 85.8%±18% and by 52.6%±11% for Seoul and the Gyeonggi province, respectively. It is also found that the forecast skill of the Lagrangian model can be improved over the Seoul metropolitan area by applying the optimized anthropogenic emission.