WRF-Chem modeling of particulate matter in the Yangtze River Delta region: Source apportionment and its sensitivity to emission changes

China has been troubled by high concentrations of fine particulate matter (PM2.5) for many years. Up to now, the pollutant sources are not yet fully understood and the control approach still remains highly uncertain. In this study, four month-long (January, April, July and October in 2015) WRF-Chem simulations with different sensitivity experiments were conducted in the Yangtze River Delta (YRD) region of eastern China. The simulated results were compared with abundant meteorological and air quality observations at 138 stations in 26 YRD cities. Our model well captured magnitudes and variations of the observed PM2.5, with the normal mean biases (NMB) less than ±20% for 19 out of the 26 YRD cities. A series of sensitivity simulations were conducted to quantify the contributions from individual source sectors and from different regions to the PM2.5 in the YRD region. The calculated results show that YRD local source contributed 64% of the regional PM2.5 concentration, while outside transport contributed the rest 36%. Among the local sources, industry activity was the most significant sector in spring (25%), summer (36%) and fall (33%), while residential source was more important in winter (38%). We further conducted scenario simulations to explore the potential impacts of varying degrees of emission controls on PM2.5 reduction. The result demonstrated that regional cooperative control could effectively reduce the PM2.5 level. The proportionate emission controls of 10%, 20%, 30%, 40% and 50% could reduce the regional mean PM2.5 concentrations by 10%, 19%, 28%, 37% and 46%, respectively, and for places with higher ambient concentrations, the mitigation efficiency was more significant. Our study on source apportionment and emission controls can provide useful information on further mitigation actions.

We set two nested domains with horizontal resolutions of 36 km and 12 km, respectively, and the inner domain covered the YRD region (see Fig 1). Vertical layers extended from the surface to 50 hPa (28 layers), with 7 layers in the bottom one km to emphasize boundary layer processes. The meteorological initial and boundary conditions were derived by the National Centers for Environmental Prediction (NCEP) FNL Operational Global Analysis data (https:// rda.ucar.edu/datasets/ds083.2/, last access: 18 November, 2018). The chemical initial and boundary fields were provided by a global chemical model (MOZART, Model for Ozone and Related chemical Tracers) [46]. We conducted four month-long simulations (January, April, July and October) to represent the typical meteorological and air conditions for each of the four seasons in 2015. Each month-long simulation was initialized for the last five days of the previous month.   [21]). MEIC is the most updated available emission inventory for China, which included monthly emissions from industry, power generation, transportation, residential and agricultural activities, and had a native resolution of 0.25˚. We averaged the MEIC emission estimates for the years of 2014 and 2016 to represent that for 2015, because the data for 2015 is not available at present. S1 Fig shows the spatial distributions of anthropogenic SO 2 , NO x , VOCs and PM 2.5 emissions in the YRD region. Anthropogenic actives strongly contributed to emissions in urban areas of central YRD, such as Zhenjiang, Changzhou, Wuxi, Suzhou and Shanghai.
Biogenic emissions were calculated by the Model of Emissions of Gas and Aerosols from Nature (MEGAN, [47]). The MEGAN is coupled into the WRF-Chem model, and provides on-line calculation of biogenic emissions. To drive MEGAN, the leaf area index (LAI), the plant function types (PFTs) and emission factors (EFs) are needed. The LAI and PFTs are based on Moderate Resolution Imaging Spectroradiometer (MODIS) products [48], and the EFs are taken from dominant species from Guenther et al. [47]. Open biomass burning emissions were quantified by the Fire INventory from NCAR (FINN, https://www2.acom.ucar. edu/modeling/finn-fire-inventory-ncar, last access: 18 November, 2018), which is based on satellite observations of active fires and land cover [49,50].

Surface PM 2.5 and meteorological observations
To a thorough evaluation of model performance, the air quality and meteorological observations with sufficient spatiotemporal coverage and resolution are need. We obtained surface hourly PM 2. Even more noteworthy is that all the cities exceeded the national grade II standard of PM 2.5 at 35 μg m -3 , except the island city Zhoushan in Zhejiang province.
We collected meteorological data with temperature, relative humidity and wind involved at 4 surface stations in Shanghai, Nanjing, Hangzhou and Hefei (see Table 1) (http://www. meteomanz.com, last access: 18 November, 2018). Meteorological conditions were generally mild and humid over the YRD region, with the annual mean temperature of 16.8-18.2˚C and the relative humidity of 63-78%.

Mitigation efficiency
In this study, we would explore the potential impacts of emission controls on PM 2.5 reduction later. To better quantify the effects of different emission reduction schemes, we defined mitigation efficiency (ME) as the PM 2.5 reduction ratio divided by the emission reduction ratio, as shown in Eq (1).
where, ME is the mitigation efficiency, E 0 is the initial anthropogenic emission, E 1 is the anthropogenic emission in certain emission control scenario, C 0 and C 1 are the simulated PM 2.5 concentrations using the emission E 0 and E 1 , respectively.

Model evaluation
Hourly observations of meteorology at 4 sites (Table 1) and PM 2.5 at 134 sites (Table 2) were compared with the simulated results for validation. Fig 1 compares the simulated and observed annual mean PM 2.5 concentrations at each of the 26 YRD cities. We averaged the grid concentrations matching at individual observation station level in each city to represent the city's value. The simulated annual mean PM 2.5 concentration was 61.8 μg m −3 averaged for the YRD region, 7% higher than the observed 57.6 μg m −3 (see Fig  1 and Table 2). Fig 2 shows the simulated and observed hourly PM 2.5 concentrations in each of the 26 YRD cities. The seasonal mean simulated PM 2.5 was the highest in winter and the lowest in summer, consistent with the observations. On the scale of individual cities, our model well reproduced the observed PM 2.5 concentrations in most cities, with the normal mean biases (NMB) less than ±20% for 19 out of the 26 cities (Table 2). However, the model systematically overestimated the observed PM 2.5 in Shanghai (NMB = 13%), Jiangsu (NMB = 11%) and Anhui (NMB = 20%). The overestimation has also been found in other simulation studies in China, but our difference was relatively smaller compared with the previous results [51], which is partly because our study used the latest emission inventory. In addition, our model incidentally overestimated the summertime PM 2.5 in Shanghai, Jiangsu and Zhejiang (see Fig 2), mainly because of the omission of precipitation prediction.    North China by monsoon in winter. Summer was the least polluted season in the year, mainly because of high mixing height and rich precipitation. We also analyzed the spatial distributions of seasonal mean PM 2.5 components, including sulfate, nitrate, ammonium, element carbon (EC), primary organic aerosol (POA), secondary organic aerosol (SOA) and other components (i.e. primary PM 2.5 mostly from dust) (see Fig  5). Secondary inorganic components (sulfate, nitrate and ammonium) and primary anthropogenic components (EC and POA) had similar seasonal variation to the total PM 2.5 , which was the highest in winter and the lowest in summer. As mentioned above, higher concentrations of secondary inorganic components and primary anthropogenic components in winter were mainly due to the lower mixing height and transport effects from North China. In contrast, SOA showed a different pattern, and the highest SOA occurred in summer. As major precursors of SOA, biogenic VOC emissions abundantly enhanced during summertime, causing the SOA concentration in summer 80% higher than the annual mean value. Nevertheless, we should note that uncertainties exist in sulfate and SOA simulations. Current models frequently underestimate sulfate especially in haze events. This may in part due to the omission of heterogeneous formation of sulfate [52,53] in chemical models. Li et al. [54] evaluated this process might contribute 20% of the sulfate in non-polluted days and the contribution would greatly enhanced as ambient PM 2.5 increasing. In addition, current chemical models also tend to underestimate SOA due to imperfect understanding of its sources and formation pathways [55].

PM 2.5 concentrations from different sources
We conducted intensive sensitivity simulations to quantify the contributions of individual sources to PM 2.5 concentration in the YRD region, by turning off the emissions from different sectors (industry, power generation, residential, transportation, and open biomass burning in the YRD region), in turn and all at once. Other simulation settings were the same as the base case.
On an annual scale, industry was the largest local source, contributing 27% to the PM 2.5 concentration averaged for the YRD region, followed by residential (19%) and transportation activities (15%). However, the PM 2.5 source characters exhibited remarkable variations in subregions in different seasons. Pie charts in Fig 1 show the corresponding contributions of each source (5 sources) to seasonal mean (4 seasons) PM 2.5 concentrations in Shanghai, Nanjing, Hangzhou and Hefei (the capital cities of each province). Industry was the most important local source to the PM 2.5 in all these four cities in summer (26-46%), fall (33-41%) and spring (24-30%), which was mainly due to the relative large emissions from coal-based industries in the region. In winter, the residential activities turned to be the dominant local contributor (28-45%), mostly caused by fossil-fuel and biofuel usage for domestic heating in the cold season. Transportation was also a considerable source of PM 2.5 in the YRD region, which was the second largest source in summer (17-26%), fall (13-24%) and spring (14-18%), and even contributed 4-14% in winter. Other sources, power generation and open biomass burning, jointly contributed 3% year-round. At a city-level, contributions from industry and transportation were more obvious in Hangzhou (35% and 20%, respectively), compared with those in the other three cities (27-29% and 13-18%, respectively). Residential source played a more important role in Hefei, which was~2 times as high as that in other cities.
In addition to the local sources, transport from outside of the YRD region (turned off the emissions from all sectors in the YRD region) also considerably contributed to the PM 2.5 concentrations in spring (40-46%) and winter (32-44%), mainly because of the spring dust storm and winter monsoon transport, respectively. On an annual mean basis, outside sources contributed 36% to the PM 2.5 concentration averaged for the YRD region, nearly half of the local sources' contribution. It indicated the importance of both local emissions and external transport to the PM 2.5 pollution in the YRD region and emphasized the necessity of regional cooperative pollution control.

Impacts of different emission controls on PM 2.5 reduction
We further conducted 5 scenario simulations to explore the potential impacts of emission controls on PM 2.5 reduction. Based on the standpoint of regional cooperative control, we assumed all anthropogenic emissions in the YRD region as well as its circumjacent regions were proportionately reduced by 10%, 20%, 30%, 40% and 50%, respectively. Other simulation settings were the same as the base case. S2 Fig compares spatial distributions of the simulated annual mean PM 2.5 concentrations in the YRD region under different emission control scenarios, as well as the corresponding concentration changes. The PM 2.5 concentration would effectively be reduced by mitigating anthropogenic emissions, and larger reductions tend to occur at the place of higher ambient concentrations (Shanghai, Anhui and southern Jiangsu). Table 3 shows the variations of mitigation efficiency of different schemes averaged for all YRD cities, and demonstrates an approximate linear correlation between anthropogenic emission reduction and the PM 2.5 concentration decrease. To be specific, a 10% reduction of anthropogenic emission caused a 10% reduction of PM 2.5 concentration. If the initial emission was reduced by 20%, 30%, 40% or 50%, the corresponding PM 2.5 concentration would decreased by 19%, 28%, 37% or 46%, respectively. The calculated ME (as defined in Method and Data) was in the range of 0.92 to 1.00, slightly decreased as the emission reduction enhanced. The results suggested that the anthropogenic emission controls at varying degrees between 10-50% could all effectively reduce the PM 2.5 concentration. The ME would slow down along with the declining initial anthropogenic emission, mainly because of the background PM 2.5 concentrations and the contributions from non-anthropogenic (biogenic and open biomass burning) sources.

Discussion
The YRD region has been arousing more and more attention in recent year, and the heave air pollution issue manifested along with the fast economic development. We summarized the main findings of this study and proposed corresponding mitigation suggestions as follows. 1) We quantified the contribution to PM 2.5 pollution in the YRD region was 64% from the local sources, and the rest 36% was from circumjacent regions, which emphasized the importance of both local emissions and external transport. The result highlights the necessity of regional cooperative control in environmental policy making. 2) Turning our attention to the YRD pollution, we found the dominant contributor was different from other high-polluted regions in China. In the Beijing-Tianjin-Hebei region, residential sources dominated [21], while in the YRD region we indicated industry contributed about half of the local PM 2.5 concentration. Thus, to pointedly reduce air pollution in the YRD region, the effective method might be controlling pollutant emission from industry activities in priority order, potentially via industrial structure innovation, emission control techniques development and fossil fuels replacement [56]. Compared with residential source, industrial pollutant control is relatively more implementable, since the former includes massive emissions from necessity for human life and sometimes detrimentally from fugitive sources. 3) Our scenario simulations give us the confidence that the PM 2.5 pollution could be effectively reduced as long as the mitigation actions bring into operation. We will always receive positive feedback from different degrees of emission controls, which is quite different with the O 3 governance. However, this study presented a big rough picture about the pollutant gross-control in the YRD region, and the emission control scenarios were supposed to proportionately reduce the pollutant emissions from all sectors. Based on the effects of gross-control as the first step, further studies are still needed to formulate and optimize implementable mitigation actions down to sector-specific level for individual pollutants.

Conclusion
In this study, the WRF-Chem model well captured the magnitudes and variations of observed PM 2.5 , with the normal mean biases less than ±20% for 19 out of the 26 cities from 134 ground stations. Based on the confidence built by the model evaluation, a series of sensitivity simulations were conducted to quantify the contributions from individual source sectors and from different regions to PM 2.5 in the YRD region. The calculated results show that local sources and surrounding transport contributed 64% and 36% to the annual mean PM 2.5 concentration, respectively, which indicated the importance of both local emissions and external transport to the PM 2.5 pollution. Among the local sources, industry activity was the most significant sector in spring (25%), summer (36%) and fall (33%), while residential source was more important in winter (38%). We further conducted 5 scenario simulations to explore the potential impacts of varying degrees of emission controls on PM 2.5 reduction, in which all anthropogenic emissions in the YRD region as well as its circumjacent regions were proportionately reduced by 10%, 20%, 30%, 40% and 50% respectively. The result demonstrated that regional cooperative control could effectively reduce the PM 2.5 level. A preliminary emission control of 10% could reduce the PM 2.5 concentration by 10%. Further emission controls would continue to reduce the concentration approximate linearly, and for places with higher ambient concentrations, the mitigation efficiency was more significant.

Author Contributions
Conceptualization: Nan Li.