Research on a Non-linear Forecast Method Based on WRF-CMAQ Model

QIAN Weimiao; CHEN Jing; HAN Juncai; CHENG Xinghong; WANG Xiaomin

doi:10.3969/j.issn.1003-6504.2017.08.017

Abstract：WRF-CMAQ air quality modeling system was applied to forecast six pollutants' (SO2,NO2,CO,O3,PM10 and PM2.5) concentration for the next three days in Shijiazhuang region.Observation data during May to July in 2014 from seven environment monitoring stations in Shijiazhuang were chosen to evaluate the model performance.The results indicate that CMAQ model can well simulate the variation of CO daily concentration,but the other pollutants' concentrations simulated by CMAQ are overestimated more or less.Relatively,the PM10 concentration is simulated better than the other pollutants,the forecast values of SO2,NO2 and PM2.5 are much bigger than observation,and the forecasting result of O3 is the worst.The inaccuracy of CMAQ model could be caused by the uncertainty of emission inventory and large amplitude of daily pollutants' concentration.In order to improve the accuracy of CMAQ model,adaptive partial least squares regression rolling method is used to correct hourly pollutants' concentration from CMAQ model.The correction results obviously improve the forecast values of CMAQ model and have profound guiding significance for fine air quality forecast in counties of Shijiazhuang.Moreover,as the primary pollutants,PM10 and PM2.5 present a better performance of the daily variation characteristics,additionally,the peaks and valleys of the diurnal variation curve is in line with observation.The corrected results can reflect the characteristics of regional distribution of pollutants' concentration in Shijiazhuang,which is useful for analyzing and forecasting the spatial distribution and transport process of pollutants' concentration under different weather background.