High‐Resolution Simulation of CO₂ Using WRF‐GHG Over the Kanto Region in Japan

Abstract The Kanto region in Japan is one of the largest economic zones in the world, an area of high anthropogenic emissions, surrounded by heterogeneous orography and complex coastlines. It hosts ground and aircraft observations critical for monitoring greenhouse gas (GHG) emissions from the region that are yet to be fully utilized. To analyze these observations, we developed a high‐resolution CO 2 model at smallest grid size of 1 × 1 km 2 using the Weather Research and Forecasting model coupled with GHGs fluxes (WRF‐GHG). The model used three different anthropogenic emission inventories: EAGrid (Japan region, 1 × 1 km 2 ), EDGAR (0.1° × 0.1°), and EDGAR‐downscaled (0.01° × 0.1°) and was run for February and May 2018 and February 2020. At 1 × 1 km 2 resolution, simulations resolve point‐source emission, which gradually attenuate as the model resolution is coarsened to 3 × 3, 9 × 9 or 27 × 27 km 2 ; however, at surface sites, the 1 × 1 km 2 model introduces transport errors weakening the correlation with observations compared to the 27 × 27 km 2 model. CO 2 simulations are found to be sensitive to planetary boundary layer dynamics on event basis. The 1 × 1 km 2 model better reproduced the CO 2 aircraft observations between 0 and 3 km compared to 27 × 27 km 2 model, suggesting critical roles of improved representation of emission patterns and transport processes. We demonstrate that a systematic bias is introduced in the CO 2 vertical distribution in the mid‐upper troposphere (altitude >3 km) when lateral and top boundary conditions are neglected in WRF‐GHG. These systematic biases in vertical distribution can have profound impact when comparing with remote sensing observations of total columns.