资源环境科技发展态势分析平台

Using the online-coupled Global/Regional Assimilation and Prediction System coupled with the Chinese Unified Atmospheric Chemistry Environment (GRAPES_CUACE) model, numerical experiments are conducted to compare the impacts of three planetary boundary layer (PBL) schemes—Medium Range Forecast (MRF), Yonsei University (YSU), and Mellor-Yamada-Janjic (MYJ)—on PM_2.5 simulations and their responses to aerosol-radiation feedback (ARF), under different terrain and weather conditions. The baseline experiments without ARF with all three PBL schemes can reproduce the spatial and temporal distribution characteristics of surface PM_2.5 on the whole and show that the model can capture the PM_2.5 variation better in unstable conditions (clean days) than in stable conditions (pollution days). The three PBL schemes show differences in the simulations of meteorological factors, e.g., the MRF scheme produces the best 10 m wind speed, the MYJ scheme produces the best 2m temperature, and the YSU scheme produces the lowest PBL height (PBLH) and vertical diffusivity. Consequently, there is no significant difference among the three PBL schemes in surface PM_2.5 simulations due to the combined action of the surface and the PBL structure. However, all the baseline experiments massively underestimate PM_2.5 during severe pollution days. The online computing of ARF in the three PBL schemes can substantially improve the underestimation of PM_2.5 by reducing sensible heat fluxes, 2m temperature, friction velocity, and the PBL height. More importantly, the response intensity of three PBL schemes to ARF is very different. For PM_2.5 simulations, MYJ has the weakest response, whereas MRF and YSU have similar responses. During the heavy pollution days with more intense aerosol-radiation effect, the increase of PM_2.5 induced by ARF in MRF and YSU is about twice that in MYJ, making the PM_2.5 simulated by MRF and YSU more consistent with the observations. This result shows the ARF dependence on the PBL processes and suggests the importance of PBL-ARF interaction to the PM_2.5 numerical prediction during haze episodes.