Establishment of a high-resolution emission inventory and its impact assessment on air quality modeling in Jiangsu Province, China

Abstract:

A high-resolution emission inventory of Jiangsu Province was developed for 2012, using the bottom-up method with the best available domestic emission factors and activity data incorporated. Information of over 6,000 point sources including geographical location, fuel type, burner type and removal efficiency were investigated from various available data sources. The point sources were estimated to account for 83.9%, 71.2%, 63.7% and 54.5% of the total SO₂, NOx, PM_2.5 and VOCs emissions respectively. Improvement of this provincial emission inventory was assessed by comparisons with emission estimation at national level. For SO₂ from power plants, NOx from transportation and PM_2.5 from industry, correlation coefficients were 0.703, 0.814 and 0.335, indicated other than power plants and transportation, there was an improvement in estimation of small industrial pollution sources which were usually estimated as area sources in national emission inventory. Correlation analysis of NOx emission and tropospheric NO₂ vertical column density measured by Ozone Monitoring Instrument (OMI) were also conducted. The correlation coefficient rose from 0.52 to 0.57 after revisions on geographical locations of 20 large point sources. Such result indicated the local source information from Environmental Statistics should be carefully examined before it can be applied for emission inventory development. In order to assess the improvement in spatial distribution and emission estimation on air quality modeling, the provincial and national emission inventory were input to Community Multi-scale Air Quality Model (CMAQ) simulations. Simulations performed better when emissions were updated from Multi-resolution Emission Inventory for China (MEIC) to provincial inventory, indicating the necessity of improved spatial and temporal distribution of emissions on air quality modeling, especially for gaseous pollutants. For SO₂, the normalized mean bias (NMB) and normalized mean error (NME) decreased from 53.9% to -29.3% and from 81.3% to 52.3% respectively. NMB of NO₂ decreased from 22.0% to -4.6%. For PM_2.5 and O₃, the predicted concentrations from both of the two simulations were lower than observations in heavy polluted days. This might be attributed to the deficiency of secondary pollution mechanisms in air quality model.