Since the Action Plan for Air Pollution Prevention and Control (the Action Plan) was implemented at the end of 2013, the ambient air quality in China is significantly improved. However, PM2.5 (particles with diameter <= 2.5 mu m) levels in some cities still exhibit clear interannual oscillations. For example, the annual mean PM2.5 levels in Shanghai decreased by 16.1% in 2014, while increased by 2.2% in 2015 according to year-on-year comparisons. To better understand the corresponding causes, the obliquely rotated T-mode principal component analysis (PCA) method and WRF-Chem model are jointly employed in this study. Results show that the west wind frequency and the accumulative wind (<1.8 m/s) duration are the key indicators affecting local PM2.5 transport and dispersion significantly. Moreover, four typical synoptic patterns conductive to PM2.5 pollution are illustrated as mid-ward path cold (CM), eastward path cold (CE), L-share high (GL) and near high center (GC), in which GL is the most adverse circulation pattern. The year-on-year changes of meteorology have positive effects on PM2.5 year-on-year variations. The significant decline of PM2.5 levels in 2014 compared with those in 2013, as well as the obvious increase in 2015 compared with 2014, both well correspond to year-on-year variations of meteorological indicators. Model results present that PM2.5 interannual variations result from the changes of meteorology during 2013 to 2015, are consistent with measured oscillations. By comparing measured and modeled PM2.5 year-on-year variabilities, a greater PM2.5 decreasing at 9.4% is estimated under favorable meteorological conditions, while a less increasing at 6% under unfavorable meteorological conditions due to emission reductions, indicating the initial improvement has been achieved by the Action Plan. Otherwise, since the current Action Plan has difficulties in completely offsetting the PM2.5 rise attributed to the adverse weather, more stringent program should be drawn up for unfavorable meteorological conditions. (C) 2016 Elsevier B.V. All rights reserved.
