Performance evaluation of atmospheric particulate matter modeling for East Asia

Abstract

This study used the Community Multi-scale Air Quality modeling system associated with emission and meteorological datasets to conduct a one-year Particulate Matter (PM) simulation in East Asia. The performance of the PM simulation results was evaluated against observed data using high-spatial resolution monitoring data from Taiwan (e.g., concentration of PM and its compositions and precursors), Air Pollution Index-derived data in 82 cities of China, and satellite-derived surface PM2.5 concentrations. Based on error analysis of two indicators (overall bias and gross error), the validity of model performance is demonstrated. Daily comparison results at the various stations (cities) in 2007 indicate that the simulations at Taiwan's monitoring stations were superior to those in cities in China. Overall, the PM10 and PM2.5 concentrations in Taiwan were underestimated by 11 and 10%, respectively, whereas the simulated PM10 concentrations in six regions of China showed significant differences, with the simulation overall bias ranging between −55 and 52%. There are two primary reasons for the large differences in model performance between Taiwan and China. First, the grid resolution for China was coarser than that for Taiwan because of the limitation of emission data resolution in the Regional Emission Inventory in Asia (REAS). Second, the spatial distribution and magnitude of emissions in the REAS might be incorrect in some regions, particularly in the domain boundary areas of China. In addition to the relatively good emission database for Taiwan, the reasonable emission estimates for the central areas of China may be partially responsible for better model performance of PM in Taiwan.