Lateral boundary conditions are required to drive regional models and are typically prepared from global model results. However, uncertainties arise from the boundary conditions. In this study, the effects of boundary conditions on the performance in the regional model (Community Multiscale Air Quality; CMAQ) for tropospheric ozone (O 3 ), which has a long lifetime, were evaluated over East Asia, mainly Japan, where air quality is affected by local and transboundary air pollution. Here, four lateral boundary conditions were prepared, with a time-independent profile assuming ocean clean air; the Community Atmosphere Model with Chemistry (CAM-Chem) global model; and the Hemispheric CMAQ (H-CMAQ) model over the northern hemisphere with seasonal and daily resolutions. All configurations generally captured the O 3 concentration measured at the surface and by ozonesonde in 2018 because O 3 in East Asia is dominated by local production. The time-independent profile failed to capture the surface seasonal O 3 variation close to the boundary and had difficulty capturing the vertical gradient in the free troposphere, particularly in southern East Asia. CAM-Chem gave an overestimation at the surface level (+16% biases), although this configuration better captured the O 3 vertical profile in the free troposphere compared with ozonesonde. H-CMAQ, seamlessly connected to the regional scale, gave better performance for surface-level (−1% and −3% biases), but underestimated the O 3 profile in the free troposphere. Although the results clarified the effects of boundary conditions on simulated O 3 concentration, other factors (i.e., photochemical production/loss and dry deposition) must be investigated to further improve the regional modeling. • Uncertainties in the regional model arising from boundary conditions were analyzed. • Four types of boundary conditions were prepared. • A time-invariant profile had difficulty near the boundary. • A higher boundary concentration in the global model penetrated the regional domain. • The seamless hemispheric model performed best among the four configurations.
Syuichi Itahashi (Fri,) studied this question.