The impact of multi-species surface chemical observation assimilation on air quality forecasts in China

Zhen Peng,Xingxia Kou,Junmei Ban,Dan Chen,Jianning Sun,Lili Lei,Aijun Ding,Zhiquan Liu,Kekuan Chu

doi:10.5194/acp-18-17387-2018

Abstract

Abstract. An ensemble Kalman filter data assimilation (DA) system has been developed to improve air quality forecasts using surface measurements of PM10, PM2.5, SO2, NO2, O3, and CO together with an online regional chemical transport model, WRF-Chem (Weather Research and Forecasting with Chemistry). This DA system was applied to simultaneously adjust the chemical initial conditions (ICs) and emission inputs of the species affecting PM10, PM2.5, SO2, NO2, O3, and CO concentrations during an extreme haze episode that occurred in early October 2014 over East Asia. Numerical experimental results indicate that ICs played key roles in PM2.5, PM10 and CO forecasts during the severe haze episode over the North China Plain. The 72 h verification forecasts with the optimized ICs and emissions performed very similarly to the verification forecasts with only optimized ICs and the prescribed emissions. For the first-day forecast, near-perfect verification forecasts results were achieved. However, with longer-range forecasts, the DA impacts decayed quickly. For the SO2 verification forecasts, it was efficient to improve the SO2 forecast via the joint adjustment of SO2 ICs and emissions. Large improvements were achieved for SO2 forecasts with both the optimized ICs and emissions for the whole 72 h forecast range. Similar improvements were achieved for SO2 forecasts with optimized ICs only for the first 3 h, and then the impact of the ICs decayed quickly. For the NO2 verification forecasts, both forecasts performed much worse than the control run without DA. Plus, the 72 h O3 verification forecasts performed worse than the control run during the daytime, due to the worse performance of the NO2 forecasts, even though they performed better at night. However, relatively favorable NO2 and O3 forecast results were achieved for the Yangtze River delta and Pearl River delta regions.

Highlights

Predicting and simulating air quality remains a challenge in heavily polluted regions (Wang et al, 2014; Ding et al, 2016)
It shows that the magnitudes of the total spreads were close to the root-mean-square errors (RMSEs), indicating that the data assimilation (DA) system was well calibrated (Houtekamer et al, 2005)
It shows that the ensemble spread of all the scaling factors were very stable throughout the ∼ 10-day experiment period, which indicates that MSF can generate stable artificial data to generate the ensemble emissions

Summary

Introduction

Predicting and simulating air quality remains a challenge in heavily polluted regions (Wang et al, 2014; Ding et al, 2016). It has been widely used to assimilate aerosol measurements from both ground-based and spaceborne platforms, including surface PM10 observations (Jiang et al, 2013; Pagowski et al, 2014), surface PM2.5 observations (Li et al, 2013; Zhang et al, 2016), lidar observations (Yumimoto et al, 2007, 2008), aerosol optical depth products from AERONET (the AErosol RObotic NETwork) (Schutgens et al, 2010a, b, 2012), and various satellites (Sekiyama et al, 2010; Liu et al, 2011; Dai et al, 2014) These studies indicate that assimilating observations can substantially improve the spatiotemporal variations of aerosol in the simulation and forecasts. Barbu et al (2009) applied an EnKF to optimize the emissions and conversion rates using surface measurements of SO2 and sulfate. McLinden et al (2016) constrained SO2 emissions using space-based observations

Results

Discussion

Conclusion