Overview of Urban PM 2.5 Numerical Forecast Models in China

Nianliang CHENG ，Hongxia LI，Fan MENG ，Fahe CHAI

1.College of Water Sciences，Beijing Normal University，Beijing100875，China；2.Beijing Municipal Environmental Monitoring Center，Beijing100048，China；3.Institute of Advanced Technology，Sun Yat-sen University，Guangzhou 510275，China；4.State Key Laboratory for Environmental Criteria and Risk Assessment，Chinese Research Academy of Environmental Sciences，Beijing 100012，China

1 Introduction

In recent years，with rapid development of national economy，acceleration of urbanization，and expansion of industrial scale，the situation of air pollution is increasingly serious［1－2］.Air pollution resulted from fine particulate matter PM2.5greatly threatens public health and its negative influence has been paid more attention to by human beings［3］.PM2.5is named fine particulate matters in air，consisting of primary particulates and secondary particulates with diameters less than or equal to 2.5microns in size，about1／20 of hair diameter［4］.The new Ambient Air Quality Standard（GB3095－2012）included PM2.5into conventional ambient air quality evaluation，which is the first PM2.5standard in China.

In January 2013，10 provinces in central China and east China experienced serious haze pollution，arousing wide attention of government and the public.In the period of Twelfth Five-Year Plan，it is very necessary and urgent to launch urban PM2.5forecast and prediction，develop suitable urban air quality system，and provide accurate，timely，and comprehensive information for administrative department.

Air quality forecast is a complex system work and it is a hot spot and difficult problem of present environmental science［6］.According to methods of forecast，urban air pollution forecast can be divided into potential forecast，statistical forecast，and numerical simulation forecast.With perfect theoretical foundation，reasonable model design，and forecast effect of regular time and quantity，the numerical forecast has become development trend of air pollution forecast.Compared with potential forecast and statistical forecast，the numerical forecast involves multiple criteria（city and weather），many pollutants（chemical reaction of dozens of substances），consideration of many processes（pollutant transfer，diffusion，chemical transport and removal），and high spatiotemporal resolution［7］.

The aerosol model is an essential part of air quality forecast and it was developed in the 1980s［8］.Using models to simulate aerosol size and changes of chemical composition is much more complex than to simulate changes of gaseous pollutants.When processing aerosol，it is necessary to consider how to characterize total particulates，total mass，distribution of particulates，and chemical component of different size of particulates，as well as major physical and chemical processes of particulates［9］.Fig.1 illustrated aerosol simulation process.

Urban air quality numerical model has already had decades of development history in north America and Europe.In foreign countries，common models include American CMAQ and CAMx，Canadian MC2-CALGRID，Germany EURAD，and England Your Air system.China's air quality forecast starts later and fundamental research basically remains passive trace state［5］.In the1980s，Beijing，Shenyang，Lanzhou，Tianjin and Nanjing started to pre-dict urban air quality using statistical method.In the 1990s，numerical forecast model of urban air quality obtained certain development.From 2000，China National Environmental Monitoring Center organized 47 key environment protection cities to launch urban air quality forecast，but only few cities made numerical forecast.The CAPSS urban air pollution numerical forecast system recommended in Environment and Development［2000］No.231 does not rely on list of urban pollution source emission and adds forecast of ozone，but it is still unable to simulate forecast of PM2.5.Since entry to the 21st century，air pollution forecast and prewarning get wide attention of both domestic and foreign experts and government officials.The Action Plan for Air Pollution Prevention and Control（National Development［2013］No.37）issued in 2013 clearly requires environmental protection department to strengthen cooperation with relevant departments and establish heavy pollution monitoring and pre-warning system.

2 Classification of forecast models

Development of numerical model experiences three stages：Gauss mist-plume model，Euler model，and Air Quality Integrated Index model.We mainly introduced 4 common air quality numerical models.

2.1 NAQPMSNAQ PMS（Nested Air Quality Forecast Modeling System）was developed by Institute of Atmospheric Physics in Chinese Academy of Sciences（IAPof CAS）mainly includes Meso-Scale Weather Forecast subsystem and Air Quality Forecast subsystem，suitable for urban and regional air quality forecast.We used aerosol thermodynamic model（ISORROPIA1.7）to calculate component state of inorganic aerosol，adopted an advanced sand dust uplifting mechanism，and added secondary organic aerosol module.This model considers two types of artificially emitted precursors（toluene and other aromatic hydrocarbons）and four types of biologically emitted precursors（isoprene and monoterpenes），and the process of generation of secondary organic matter SOA through photochemical reaction.The heterogeneous chemical reaction model of aerosol surface mainly focuses on influence of sulphides and nitrides.At present，NAQPMS has basically realized parallelization and business of model system，and obtained wide approval.This model is already applied in many provinces and has obtained better effect［10］.

2.2 CMAQCMAQ is amultiple module integrated multi-scale nested grid three-dimensional Eulerian，issued by EPA in 1998，suitable for regional multi-scale environment simulation and evaluation.CMAQ model has perfect chemical mechanism and aerosol module and has the ability of predicting secondary pollution.Aerosol process of CCTM chemical reaction mechanism involves removal of dry deposition，mutual reaction of aerosol and cloud droplet，wet removal，homogeneous nucleation process of sulphuric acid／vapor system，organic aerosol component generated from gaseous precursor，particle condensation and coagulation growth［11］.The aerosol model is evolved from regional particulate model（RPM）established by Binkowski and Shanker（1995），while the RPM is based on Regional Acid Deposition Model（RADM）［12］.Aerosol particulates are divided into fine and coarse particulates.Based on thermodynamic balance principle，we calculated chemical composition of inorganic aerosol，and then calculated gaseous and aerosol composition and balance concentration of components after inorganic matters such as H2SO4，NH3，HNO3，sea salt，and H2O reach thermodynamic balance in gaseous and aerosol phase.Secondary inorganic aerosol mainly considers mixture of NH3-HSO4-HNO3-H2O，secondary organic aerosol（SOA），calculated using Pandis organic aerosol output efficiency method［13］.

CMAQ model is universal，flexible and developable，rapid development in recent years has made it become one of the air quality simulation systems with widest application［14］.At present，NOAA and EPA have developed nationwide air quality forecast（NAQF）system using Eta-12model and CMAQ.David G Streets also simulated the air quality of2008 Beijing Olympic Game stadium using CMAQ model.China has made extensive researches on CMAQ numerical simulation forecast.In national 973 project＂air，water and soil environment pollution and control mechanism for capital Beijing and surrounding areas＂，Liu Yu et al transplanted new generation air quality model CMAQ to Sunway supercomputer of State Meteorological Administration and realized professional air quality simulation of north China.Later，researches transplanted it to IBM supercomputer of National Meteorological Center，and took MM5 numerical forecast products as meteorological field to connect numerical business system of meteorological center，to provide reference for national and provincial level air quality forecast［9］.Feng Yerong et al［15］successfully applied CMAQ／MM5／SMOKE air quality model in south China areas and made analysis on haze weather generated from tropical cyclone in Delta of the Pearl River.Compared with NAQPMS，this model is already widely applied in simulation and forecast evaluation researches of many provinces and cities.

②坡面修整、清理：先将坡面的松石、不稳定的土体清除；磨平坡面锐角岩体部分以免划破生态袋表面；坡面有渗水处做好导水盲沟。每级坡顶、平台修建截水沟防止雨水对坡面的直接冲刷。

2.3 CAMxDeveloped by ENVIRON Company on the basis of UAM-V model，CAMx model is used to make comprehensive evaluation of gaseous and particle state air pollutants in urban and regional scale.The most distinctive benefits of CAMx include two way nesting and elastic nesting，Plume-in-Grid（PiG）module，Ozone Source Apportionment Technology（OSAT），and Particulate Matter Source Apportioning Technology（PSAT），etc.Besides，vertical layer structure is defined externally，so the height of each layer can be defined as function of any space or time.Such flexibility in defining horizontal and vertical grid structure makes CAMx adapt to meteorological model used to provide input field for environmental model，so it is widely applied in China［16］.

2.4 WRFCHEMWRFCHEM model is the Weather Research and Forecasting（WRF）model coupled with Chemistry recently developed by the United States.It integrates atmospheric chemistry into WRF developed by National Center for Atmospheric Research（NCAR）.The largest benefit of WRFCHEM is complete coupling of meteorological model and chemical transport model in time and space resolution and realizes real online feedback.This model remains preliminary application stage in China and the forecast application is relatively few［17］.

3 Application of models

In provinces and cities already implementing air quality forecast，the technical framework generally consists of statistical model，numerical model，and subjective revision.According to the Notice about Printing Implementation of Air Quality Forecast in Provinces and Cities of China（No.2014.02）issued by China National Environmental Monitoring Center［18］，11 provincial stations launched air quality forecast，including Beijing，Tianjin，Shanghai，Jiangsu，Zhejiang，Fujian，Shandong，Hubei，Hunan，Shaanxi and Qinghai，9 provincial capitals（Taiyuan，Shenyang，Nanjing，Hangzhou，Jinan，Wuhan，Guangzhou，Chengdu，and Xi'an），and 2 cities under direct planning by the state（Qingdao and Shenzhen）.There are 15 provincial stations，13 provincial capitals and 3 cities under direct planning by the state having forecast capability establishment plan.Table 1 listed implementation of part of cities already implementing air quality numerical forecast system.It shows that Beijing，Shanghai，Nanjing and Guangzhou started early and the development level is relatively higher，while some cities and provinces are still at preparatory stage.The models they select are basically CMAQ，CAMx，NAQPMS，and WRFC hem.In this study，we made a brief introduction using data of Beijing，Shanghai，and Nanjing air quality numerical forecast models.

Table 1 Development of air quality forecast models of domestic cities

3.1 Air quality forecast system of BeijingAs early as2001，Beijing launched the air quality numerical forecast.Later，taking 2008 Olympic Game as an opportunity，it realized system upgrade of Beijing air quality forecast technology framework，mainly including development and application of dynamic statistical forecast system，construction and application of comprehensive forecast platform，and setup of numerical forecast models.The numerical forecast system includes NAQPMS model，CMAQ model，and CAMx model.It adopts unified model region setting，unified pollution emission list and SMOKE model to dispose emission sources，and it is driven by unified meteorological model MM5.Besides，the grid region is provided with four layer nesting：the first layer covers all regions of China，the second layer covers north China，the third layer covers Beijing，Tianjin and Hebei，and the fourth layer covers Beijing.Fig.2 illustrated frame structure of this numerical forecast system.

Each air quality model has its own characteristic in different station simulation effect.To optimize air quality forecast results，arithmetic geometric average and weight integration methods were used to issue numerical forecast results，and various statistical pa-rameters were calculated to test the level of this air quality numerical forecast.Application research indicates that the accuracy rate of this forecast model results is higher than 80%.Considering no pollution forecast method has absolute predominance，we made comprehensive integration of various forecast methods，namely potential pollution forecast→ statistical numerical forecast→ heavy pollution day indicator criterion→comprehensive forecast results→ heavy pollution day warning issue.We established technical procedure for heavy pollution warning of Beijing based on Meso-Scale meteorological forecast products，to more accurately recognize and forecast time and cause of heavy pollution process.

3.2 Air quality forecast system of ShanghaiShanghai built air quality forecast system with the aid of Expo2010 Shanghai and made improvement in recent years.This system consists of meteorological model，integrated numerical forecast model，statistical model，and various observation and information tools.

Air quality forecast of Shanghai is jointly implemented by Environmental Atmosphere Monitoring Office and Information Technology Department of Shanghai Environmental Monitoring Center.In 2013，it established Yangtze River Delta Air Quality Forecast and Prediction Center.On August1，2013，it made in-house trial forecast.From September 1，2013，Shanghai Environmental Monitoring Center started to issue AQI forecast to the public.The air quality forecast and warning system of Shanghai is further optimized and improved on the basis of Expo2010 Shanghai.Integrating with technologies newly added in recent years，the system consists of multi-model integrated forecast system，straw burning warning system，automatic alarming system，and joint consultation system.

4 Case analysis

On October 6-12th，2014，Beijing-Tianjin-Hebei was attacked by a heavy pollution.The air quality level in October 8-10 th was heavy pollution；on October 11 th，it reached hazardous pollution and the average daily valueρ（PM2.5）was up to329μg／m3.Fig.5 depicted spatial distribution of PM2.5concentration in Beijing-Tianjin-Hebeiat20：00 after linear interpolation using CAMx model（the grid resolution is 6 km），which was basically consistent with measured value（http：／／113.108.142.147：20035／emcpublish／）.On October 6 th，the air quality of Beijing-Tianjin-Hebei was excellent；on October 7 th，Beijing-Tianjin-Hebei changed to area of high barometric pressure and southerly air current flowed to north and then to Beijing with high concentration of air pollutants；on October8 th and 9 th，air diffusion condition became worse and regional pollution was further deteriorated；PM2.5concentration was high and evenly distributed，and the level reached hazardous pollution；on October 10 th，in the action of higher southerly wind，pollutants gradually gathered due to influence of terrain of Taihang Mountain；at night of October 11 th，cold air attacked Beijing，the PM2.5concentration of Beijing dropped sharply from 200μg／m3to 10μg／m3，the air quality became excellent，while southeast of Heibei Province was still at moderate pollution level；on October 12 th，influenced by cold high barometric pressure moving eastward and southward，pollutant concentration gradually declined from north to south，and air quality of most areas of Bei-jing-Tianjin-Hebei turned excellent.

Fig.6 illustrated PM2.5regional transport rate of Beijing，Tianjin and Shijiazhuang in this time of heavy pollution caculated by PAST technology，showing the regional transport rate had significant temporal change with pollution degree.Before and after the heavy pollution，the outer source transport rate was below 20%；with deterioration of pollution，the outer source transport rate rose；through calculation，the average outer source transport rate of PM2.5in Beijing，Tianjin and Shijiazhuang during October 7-10 th was57%，48%and 27%respectively，so regional transport plays a great role in PM2.5concentration of Beijing.

Streets et al［22］using CMAQ model simulated the contribution of Hebei Province to PM2.5concentration of Beijing in the action of steady south wind reached 50-70%during 2008 Olympic Games；Wang et al［23］using CAMx model calculated the contribution rate of outer source to PM2.5concentration of Shanghai，the result was close to 50%；Wu et al［24］using CAMx model calculated sources of PM2.5concentration of Guangzhou in spring and found 31.7%came from regional transport；Xue Wenbo et al［25］using CAMx model simulated and calculated average PM2.5concentration of Beijing-Tianjin-Hebei，Yangtze River Delta，and Pearl River Delta regions and the contribution rate reaches22%，37%，and 18%separately，and the contribution rate of outer source to PM2.5concentration of Shandong reached 41%.

Forecast results of this study are comparable with other studies，while there are still differences for following reasons：（i）lack of studies on different regions and long time trans-regional transport rules of PM2.5and major component；（ii）different space resolution，emission source list，parameter setting and calculation methods.

5 Problems and prospects

At present，China's air pollution forecast model system has considerable development，but compared with foreign countries，there are still problems of inadequate pollution source list and low accuracy of forecast.China is still faced with challenges of at mospheric environment，we must further improve atmospheric environment numerical model.

（i）Establishing perfect pollution forecast system.There are problems of large error of pollution sources，few forecast items，limited coverage，and low forecast accuracy in domestic forecast models.Test of model forecast results is also an essential part of urban air quality forecast system.Thus，it is feasible to develop evaluation model to evaluate consistency with spatial distribution and time change of measured values，so as to improve fine forecast effect in the process of heavy pollution.

（ii）Constantly expanding new fields of studies of atmospher-ic environment simulation forecast.Some new research technologies and fields，such asmulti-model integrated forecast technology，pollution source assimilation technology，and aerosol microphysical process studies also bring many new ideas and challenges.

（iii）Researching and developing graphical interface and forecast products.Diversified forecast models and products provide rich reference information and provide timely and accurate forecast warning basis for travel of the public and decision making of government.

（iv）Combining with other technologies，such as3Sand GIS technologies.Remote sensing technology takes space as information acquisition platform and has wide monitoring scope，rapid speed，and low cost.Besides，it is convenient for long-term dynamic monitoring and is able to find pollution source and diffusion state difficult to be revealed by conventional methods.Using GIS technology，it is able to see current air quality condition and forecast information of different time through clicking the corresponding region in website.Combination of forecast models with other technologies can provide better and more detailed references for warning and forecast and generate different forms of warning and forecast products for management department.

（v）Combining numerical forecast and subjective experience.Numerical forecast models automatically calculate forecast results already generated.This is mainly influenced by models themselves and their performance.Subjective experience guides forecast personnel to make forecast according to accumulated knowledge and other auxiliary data.Thus，there are many uncertain factors and may be accidental，but the forecast results are closer to actual situations.Finally，the forecast results are combination of subjective and objective forecast.

（vi）Improving corresponding regulations and requirements.Domestic urban air quality forecast has not established corresponding regulations and requirements and most works are implemented independently.In certain time and period，demands and concern have large fluctuation，and it is difficult to form constant motive force.Besides，meteorological department always strives for air quality forecast business.Demands and pressure rapidly expand，so environmental protection department must bring into play their subjective initiative，use their existing advantages，launch forecast of serving the public and management department.Standardized working methods and well-established assessment system are very necessary and need to be improved.

（vii）Combining forecast and management.The ultimate purpose of model forecast is to provide the public with air quality forecast and warning information services，provide information support for pollution prevention and control decisions of management department，and provide technical guarantee for joint prevention and control of air pollution and emergency response in key areas.This needs fine forecast of heavy pollution through combining forecast and management.

（viii）Strengthening construction of forecast team.It is recommended to strengthen construction of forecast team，consolidate in-house technical personnel to tackle key and difficult problems，and constantly deepen the strategic cooperation with relevant colleges and universities.

From January 1st，2015，Ministry of Environmental Protection officially issued real-time monitoring data of all air quality monitoring stations at the third stage of new standard.In the 338 cities，there are 1436 monitoring stations and all of these stations will monitor air quality using new standard.Distribution of monitoring stations is not simple and it needs full consideration of feeling of the public.China is vast in territory，so it is impossible to establish perfect air quality monitoring network like the United States in short term.As an extremely effective research tool，the air quality forecast model can realize numerical simulation of multiple scale and multiple processes.Also，it is able to forecast PM2.5concentration and change trend in different seasons and different regions，obtain background concentration and forecast concentration of different stations，and obtain simulation results of point sources，non-point pollution and moving sources of different regions through analyzing sources of various pollutants.Through fully using monitoring and simulating forecast results，it is able to provide effective and comprehensive information for management department，so as to take necessary emergency measures to minimize adverse effect of PM2.5pollution.Combining PM2.5concentration of different regions in different meteorological conditions，government may promulgate highly restrictive and mandatory emission standards to improve regional air quality，and continue to do cooperation of emission reduction，rather than simply stressing PM2.5forecast.

［1］REN ZH，WAN BT，SU FQ，et al.Several characteristics of atmospheric environmental quality in China at present［J］.Research of Environmental Sciences，2004，17（1）：1－6.（in Chinese）.

［2］XU XD.Dynamic issues of urban atmospheric pollution models［J］.Quarterly Journal of Applied Meteorology，2002，13（1）：1－12.（in Chinese）.

［3］WEIFS，TENG EJ，WU GP，et al.Concentrations and elemental components of PM2.5，PM10 in ambient air in four large Chinese cities［J］.Environmental Monitoring in China，2001，17（1）：1－6.（in Chinese）.

［4］LILJ，WANG Y，LIJX，et al.The analysis of heavy air pollution in Beijing during2000～2010［J］.China Environmental Science，2012，32（1）：23－30.（in Chinese）.

［5］Ministry of Environmental Protection of the People’s Republic of China.Ambient air quality standards（GB 3095－2012）［S］.Beijing：China Environmental Science Press，2012.（in Chinese）.

［6］WANG ZF，XIE FY，WANG XQ，et al.Development and application of nested air quality prediction modeling system［J］.Chinese Journal of Atmospheric Sciences，2006，30（5）：778－790.（in Chinese）.

［7］FANGXY，JIANGWM，WU J，etal.Study on the development of numerical model system to predict urban air quality［J］.Acta Scientiae Circumstantiae，2004，24（1）：111－115.（in Chinese）.

［8］PILINISC，SEINFELD JH，SEIGNEUR C.Mathematical modeling of the dynamics of multicomponent atmospheric aerosols［J］.Atmospheric Environment，1980，（21）：943－955.

［9］YAN P，LIWL，QIN Y.A review of the current development of atmospheric aerosol models［J］.Quarterly Journal of Applied Meteorology，2004，15（5）：629－640.（in Chinese）.

［10］WANG ZF，XIE FY，WANG XQ，et al.Development and application of nested air quality prediction modeling system［J］.Chinese Journal of Atmospheric Sciences，2006，30（5）：778－790.（in Chinese）.

［11］CEMPD，UNC.Operational guidance for the community multiscale air quality（CMAQ）modeling system［EB／OL］.Chapel Hill：Commuunity Modeling and analysis System Center，2010［2014－09－01］.http：／／www.cmascenter.org／help／model＿docs／cmaq／4.7.1／CMAQ＿4.7.1＿OGD＿28june10.pdf.

［12］BINKOWSKIF S，SHANKAR U.The regional particulate matter model 1：Model description and preliminary results［EB／OL］.JGR，1995，100（D12）：26191－26209.

［13］PANDIS SN，HARLEY RA，CASS GR，et al.Secondary organic aerosol formation and transport［EB／OL］.Atmos Environ，1992，26A：2269－2282.

［14］FAN Y，MENGWG，WANGXM，et al.Amulti-selective air quality model system［J］.Chongqing Environmental Science，2003，25（11）：134－137.（in Chinese）.

［15］FENG YR.Mechanical analyses and numerical simulations on the aerosol pollution over the Pearl River Delta［D］.Guangzhou：Sun Yat－sen University，2006.（in Chinese）.

［16］TONG YC.Air quality forecast and development in major cities of China［J］.Environmental Monitoring in China，2006，22（2）：69－71.（in Chinese）.

［17］FANGXY，JIANGWM，WU J，et al.Study on the development of numerical model system to predict urban air quality［J］.Acta Scientiae Circumstantiae，2004，24（1）：111－115.（in Chinese）.

［18］CNEMC.The notification about issuing the survey results of local environmental air quality forecast（No.2014］02）［Z］.（in Chinese）.

［19］China Research Academy of Environmental Sciences.Study on key technologies of ambient air quality standard management of heavy pollution cities［R］.Beijing：China Research Academy of Environmental Sciences，2011.（in Chinese）.

［20］China Research Academy of Environmental Sciences.Pre－research on atmospheric haze characteristics and control ways in typical regions［R］.Beijing：China Research Academy of Environmental Sciences，2013.（in Chinese）.

［21］HE KB，YANG FM，DUAN FK，et al.Atmospheric particulate matter and regional compound pollution［M］.Beijing：Science Press，2011：165－185.（in Chinese）.

［22］STREETSDG，FU JS，JANGCJ，et al.Air quality during the 2008 Beijing Olympic Games［J］.Atmospheric Environment，2007，41：480－492.

［23］WANG YJ，LIL，CHEN CH，et al.Source appointment of fine particulate matter during autumn haze episodes in Shanghai，China［J］.Journal of Geophysical Research，2014，119（4）：1903－1914.

［24］WU DW，FUNG JC，YAO T，et al.A study of control policy in the Pearl River Delta region by using the particulate matter source apportionment method［J］.Atmospheric Environment，2013，76：147－161.

［25］XUEWB，FU F，WANG JN，et al.Numerical study on the characteristics of regional transport of PM2.5 in China［J］.China Environmental Science，2014，34（6）：1361－1368.（in Chinese）.