Impact of crop field burning and mountains on heavy haze in the North China Plain : a case study

With the provincial statistical data and crop field burning (CFB) activities captured by Moderate Resolution Imaging Spectroradiometer (MODIS), we extracted a detailed CFB emission inventory in the North China Plain (NCP). The WRF-CHEM model was applied to investigate the impact of CFB on air pollution during the period from 6 to 12 October 2014, corresponding to a heavy haze incident with high concentrations of PM2.5 (particulate matter with aerodynamic diameter less than 2.5 μm). The WRF-CHEM model generally performed well in simulating the surface species concentrations of PM2.5, O3 and NO2 compared to the observations; in addition, it reasonably reproduced the observed temporal variations of wind speed, wind direction and planetary boundary layer height (PBLH). It was found that the CFB that occurred in southern NCP (SNCP) had a significant effect on PM2.5 concentrations locally, causing a maximum of 34 % PM2.5 increase. Under continuous southerly wind conditions, the CFB pollution plume went through a long-range transport to northern NCP (NNCP; with several mega cities, including Beijing, the capital city of China), where few CFBs occurred, resulting in a maximum of 32 % PM2.5 increase. As a result, the heavy haze in Beijing was enhanced by the CFB, which occurred in SNCP. Mountains also play significant roles in enhancing the PM2.5 pollution in NNCP through the blocking effect. The mountains blocked and redirected the airflows, causing the pollutant accumulations along the foothills of mountains. This study suggests that the prohibition of CFB should be strict not only in or around Beijing, but also on the ulterior crop growth areas of SNCP. PM2.5 emissions in SNCP should be significantly limited in order to reduce the occurrences of heavy haze events in the NNCP region.

Abstract. With the provincial statistical data and crop field burning (CFB) activities captured by Moderate Resolution Imaging Spectroradiometer (MODIS), we extracted a detailed CFB emission inventory in the North China Plain (NCP). The WRF-CHEM model was applied to investigate the impact of CFB on air pollution during the period from 6 to 12 October 2014, corresponding to a heavy haze incident with high concentrations of PM 2.5 (particulate matter with aerodynamic diameter less than 2.5 µm). The WRF-CHEM model generally performed well in simulating the surface species concentrations of PM 2.5 , O 3 and NO 2 compared to the observations; in addition, it reasonably reproduced the observed temporal variations of wind speed, wind direction and planetary boundary layer height (PBLH). It was found that the CFB that occurred in southern NCP (SNCP) had a significant effect on PM 2.5 concentrations locally, causing a maximum of 34 % PM 2.5 increase. Under continuous southerly wind conditions, the CFB pollution plume went through a long-range transport to northern NCP (NNCP; with several mega cities, including Beijing, the capital city of China), where few CFBs occurred, resulting in a maximum of 32 % PM 2.5 increase. As a result, the heavy haze in Beijing was enhanced by the CFB, which occurred in SNCP. Moun-tains also play significant roles in enhancing the PM 2.5 pollution in NNCP through the blocking effect. The mountains blocked and redirected the airflows, causing the pollutant accumulations along the foothills of mountains. This study suggests that the prohibition of CFB should be strict not only in or around Beijing, but also on the ulterior crop growth areas of SNCP. PM 2.5 emissions in SNCP should be significantly limited in order to reduce the occurrences of heavy haze events in the NNCP region. 9676 X. Long et al.: Impact of crop field burning and mountains on heavy haze in NCP, China and Liu, 1994;Streets et al., 2003;Wang and Zhang, 2008;Zhao et al., 2010). Large numbers of annual CFB occur in China Yan et al., 2006), especially during the post-harvest seasons Shi et al., 2014;Cao et al., 2008). In addition, most of the CFB occurs on crop growth areas, such as the North China Plain (NCP) (Huang et al., 2012;Li et al., 2008), which have frequently suffered haze events in recent years (Yang et al., 2015;Jiang et al., 2015;Wang et al., 2012).
However, CFB has adverse impacts on traffic conditions and ecology environments (Shi et al., 2014;Zhang, 2009), and release plenty of pollutants, such as CO, SO 2 , VOC (volatile organic compounds), NO x and PM 2.5 (particulate matter with aerodynamic diameter less than 2.5 µm) (Koppmann et al., 2005;Li et al., 2008). According to Guan et al. (2014) and Lu et al. (2011), annual CFB contributes about 13 % of the total particulate matter (PM) emissions in China . Furthermore, it is more prominent during the harvest periods due to its strong seasonal dependence. Numerous studies have quantified the contribution of biomass burning and CFB to PM pollution in China. According to Yao et al. (2016), Cheng et al. (2013), Wang et al. ( , 2007 and Song et al. (2007), biomass burning has important impacts on the ambient PM 2.5 concentrations (15)(16)(17)(18)(19)(20)(21)(22)(23)(24). Yang et al. (2010) captured a heavy pollution with PM 10 concentrations higher than 350 µg m −3 in some CFB locations. It has been reported that CFB may contribute more than 30 % of the PM 10 increase during CFB incidents Zha et al., 2013;Su et al., 2012). Cheng et al. (2014) reported a summer case that CFB contributed 37 % of PM 2.5 concentrations in the Yangtze River delta.
The impact of CFB to air quality is continental and regional. Air quality in China is influenced by the CFB that occurs in Southeast Asia and on the Indian Peninsula (Qin et al., 2006). Mukai et al. (2014) reported that CFB emissions in Southeast Asia contribute the carbonaceous aerosols in Beijing. Within China, the inter-province transported air pollutants emitted from CFB significantly affect regional PM levels and air quality (Cheng et al., 2014;Zhu et al., 2012). For Beijing, the smoke particles from CFB are expected to be one of the major components Cheng et al., 2013), though the percentage of transported sources are seldom specified . A recent study reported that CFB and regional transport illustrated two of the key processes of haze formation in October 2014, especially on 6 October, but without quantitative estimation in this work (Yang et al., 2015). Related quantification studies are of great importance for the control strategies of CFB in Beijing.
Yanshan and Taihang mountains surround the NCP in the north and west (Fig. 1c). Such topography affects air pollution though the planetary boundary layer (PBL) in complex ways (Miao et al., 2015;Sun et al., 2013;Liu et al., 2009). Hu et al. (2014) reported that the Loess Plateau and NCP result in a mountain-plains solenoid circulation, exacerbat- ing air pollution over NCP. Chen et al. (2009) found that a mountain chimney effect is dominated by mountain-valley breeze, enhancing the surface air pollution in Beijing. The mountain-plain breeze develops frequently in Beijing and may play important roles in modulating the local air quality (Miao et al., 2015;Hu et al., 2014;Chen et al., 2009). Miao et al. (2016) found that the mountains played a significant role in the sea-land aerosol circulation and the pollutants could be transported and accumulated in the NCP areas along the mountains, which is treated as the blocking effect .
In this study, we analyzed a heavy haze episode that occurred in the NCP region from 12:00 to 00:00 local time (LT) on 6 to 12 October in 2014. During the period, the X. Long et al.: Impact of crop field burning and mountains on heavy haze in NCP, China 9677 average PM 2.5 concentrations are much higher than class II standard in both the southern NCP (SNCP) and northern NCP (NNCP). The characteristics of the air pollution were analyzed based on PM 2.5 concentration. Depending on the satellite-based observations of Moderate Resolution Imaging Spectroradiometer (MODIS), a large number of CFB occurred in SNCP, whereas few CFBs occurred in NNCP. A more detailed CFB emission inventory was extracted. Thereafter we analyzed the regional transport of CFB emissions from the SNCP to NNCP driven by prevailing southerly winds. Under continuous southerly wind conditions, the mountains played an important role in northward transport, and caused an accumulation of the aerosol pollutants at the foothills of the mountains. We also analyzed the impact of mountains (especially the Taihang Mountains and the Yanshan Mountains) on the air pollution transport.
2 Description of data

Geographical location
In order to study the effect of CFB on local and regional air pollution, the research domain is located in eastern China, covering a large regional area (more than 10 provinces) (Fig. 1a). The NCP region is in the middle of the research domain, with two mountains in the north and west. The Yanshan Mountains are located in the north of NCP with eastwest directions, and the Taihang Mountains are located in the west of NCP with southwest-northeast directions (Fig. 1b). Figure 1c displays the distribution of online sampling sites and CFB captured by MODIS during the haze episode. We defined two regions of interests according to CFB occurrences, topographic conditions, industrial and agricultural developments. One is the northern NCP (NNCP), including two mega cities (Beijing and Tianjin) and the northern part of Hebei province, where few CFBs occurs. Another is the SNCP, where substantial CFB occurred during the haze episode as shown in Fig. 1c. Because of the severe haze problem in the capital city of China (Beijing), one of the main focuses is to study the long-range transport of CFB pollution from SNCP to NNCP.

Meteorological conditions
The reanalysis meteorological data, including wind direction, wind speed and planetary boundary layer height (PBLH) were obtained from the European Centre for Medium-range Weather Forecasts (ECMWF), with a spatial resolution of 0.125 • × 0.125 • . The data are available at http://www.ecmwf.int/products/data/. The average wind direction and wind speed are displayed in Table 1. During the haze episode, the mean wind directions are 174.8 • in NNCP and 165.2 • in SNCP, and the average wind speeds are 2.4 m s −1 in both NNCP and SNCP. The results suggest that the prevailing winds are continuous southerly winds with weak wind speeds, which are favorable to long-range transport of pollution from SNCP to NNCP, which has been reported as one of the major contributors to haze formation in the Beijing City .

PM 2.5 measurements
The hourly PM 2.5 mass concentrations were continuously monitored by the Ministry of Environmental Protection (MEP) of China (http://datacenter.mep.gov.cn), including five sites in NNCP and seven sites in SNCP (indicated by green crosses in Fig. 1c). The data were updated from the website: http://www.pm25.in/. Table 1 summarizes the site information and the observed PM 2.5 concentrations. During the study period, the average PM 2.5 concentrations are 200.0 µg m −3 in NNCP and 184.1 µg m −3 in SNCP. The measured PM 2.5 concentrations are much higher than a class II standard (daily mean of 75 µg m −3 ), indicating an occurrence of a heavy pollution event. We analyzed the characteristics of the air pollution based on the PM 2.5 concentration simulated by WRF-CHEM. Meanwhile, it is worth noting that the highest PM 2.5 concentrations occurred along the foothill sites of the Taihang Mountains. At the foothill sites of Beijing, Baoding, Shijiazhuang and Xingtai, PM 2.5 concentrations are 245.5, 287.7, 257.9 and 320.1 µg m −3 , respectively. The mean PM 2.5 concentration in these four sites is 277.8 µg m −3 , much higher than 147.2 µg m −3 averaged from the other sites. Considering the continuous southerly winds and the topographic conditions, we studied the impact of the mountains on the air pollution transport.

Model description
We used Weather Research and Forecasting Chemical model (WRF-CHEM) (Grell et al., 2005) to simulate the spatial and temporal variability of PM 2.5 concentration. The specific version of the WRF-CHEM model is developed by Li et al. ( , 2011, with a new flexible gas-phase chemical module and the CMAQ (Community Multi-scale Air Quality) (version 4.6) aerosol module developed by US EPA (Binkowski and Roselle, 2003). The wet deposition follows the CMAQ method and the dry deposition is parameterized following Wesely (1989). The photolysis rates are calculated using the FTUV (Fast Radiation Transfer Model) (Li et al., 2005;Tie et al., 2003), in which the impacts of aerosols and clouds on the photochemistry are considered (Li et al., 2011). The gas-phase chemistry was represented in the model by the modified RADM2 (Regional Acid Deposition Model, version 2) gas-phase chemical mechanism (Stockwell et al., 1990;Chang et al., 1987). Meanwhile, the ISORROPIA version 1.7 (http://nenes.eas.gatech.edu/ISORROPIA/) is utilized to simulate the inorganic aerosols, which is primarily used to predict the thermodynamic equilibrium between the Table 1. The average PM 2.5 concentration, wind direction and wind speed of the observations from 12:00 to 00:00 LT on 6 to 12 Oct. The sampling sites located at the foot of mountains are emphasized in bold.
The WRF-CHEM model is configured with resolution of 6 km × 6 km (200 × 300 grid cells) centered in (117 • E, 39 • N). Vertical layers extend from the surface to 50 hPa, with 28 vertical layers, involving seven layers in the bottom of 1 km. The meteorological initial and boundary conditions were gathered from NCEP FNL (National Centers for Environmental Prediction Final) Operational Global Analysis data. The lateral chemical initial conditions were constrained by a global chemical transport model -MOZART4 (Model for Ozone and Related chemical Tracers, version 4) -6 h output (Emmons et al., 2010;. For the episode simulations, the spin-up time of the WRF-CHEM model is 3 days. The surface emission inventory used in this study was obtained from the Multi-resolution Emission Inventory for China (MEIC) , which is an update and improvement for the year 2010 (http://www.meicmodel.org). The emission inventory estimated only anthropogenic emission such as non-residential sources (transportation, agriculture, industry and power) and residential sources related to fuel combustions. The biogenic emissions are calculated online with the WRF-CHEM model using the MEGAN model (Guenther et al., 2006). Additionally, we added emission from CFB in the present study.

Crop field burning emissions
We analyzed the annual and monthly number of open crop fire events captured by MODIS in the research domain from 2008 to 2014. In the NCP region, the CFB activities gradually increase from the minimum fire events of 12 000 times in 2008 to 27 000 times in 2014 (Fig. 2a), suggesting that the CFB is not efficiently controlled in this region. This situation may result from the limitation of local enforcement of regulation despite the fact that CFB has already been banned Shi et al., 2014). The CFB have a seasonal pattern due to the post-harvest activities with two distinct peaks in summer and autumn, especially in June (33-59 %) and October (6-19 %) (Fig. 2b). The strong seasonal dependence character suggests that the CFB emissions during October are much larger than annual averages. In order to have the detailed horizontal distribution of the pollutant emissions of CFB, we elaborated a method to generate emission inventory using the satellite data of "MODIS Thermal Anomalies/Fire product (MOD/MYD14DL)". The MOD/MYD14DL product can detect small opening fires (< 100 m 2 ) (Giglio et al., 2003) and produce the geographic location of open fire activities (van der Werf et al., 2006). In this study, the CFB was defined as MOD/MYD14DL active fires occurred over the cropland, which is classified by the MODIS Combined Land Cover Type product (Friedl et al., 2010).
First, we estimated the CO emission of CFB, utilizing a widely used method (Streets et al., 2003;Cao et al., 2008;Ni et al., 2015) based on the annual provincial statistical data. The provincial emission of crop residues burning can be calculated by Eq. (1): where i stands for each province and k for different crop species of rice, corn and wheat. E i,CO stands for CO emission from CFB of ith province in gigagrams [Gg]. P i,k is the yield of crop in Gg. F i is the proportion of residues burned in the field. D k is the dry fraction of crop residue (dry matter). R k is the residue-to-crop ratio (dry matter). CE k is the combustion efficiency and EF CO is the emission factors of CFB. The P i,k values were taken from an official statistical yearbook (NBS, 2015) ( Table S1 in the Supplement), and the F i on a provincial basis were taken from Wang and Zhang (2008) and Zhang Yisheng (Unpublished doctor thesis-in Chinese) ( Table S1). The parameters of D k , R k and CE k are listed in Table S2. The EF CO from CFB was summarized range from 52 to 141 g kg −1 in China (Table S3). In this study, we used 111 g kg −1 as the average EF CO of crop residue, which was used to estimate the emissions from global open burning . The provincial CO emission was temporally and spatially allocated according to the CFB activities. The detailed daily CO emission of kth grid (E k,CO ) was calculated using Eq. (2): where FC k and FC i are the total CFB fire counts in kth grid and ith province, respectively (Table S1). Thereafter, the emissions of various gaseous and particulate species (E spec1 ) were calculated by the Eq. (3). Furthermore, individual chemical compounds (E spec2 ) were calculated by Eq. (4): where E k,spec1 and E k,spec2 are the kth grid emission of the specify WRF-CHEM species, E spec1 and EF CO are the emission factors of CFB, E k,NMOC is NMOC emission in the kth grid calculated by Eq.
(3) and scale is the value to convert NMOC emissions to WRF-CHEM chemical species. The emission factors for gaseous and particulate species and scale to convert NMOC emissions to WRF-CHEM chemical species from CFB were taken from available data sets Akagi et al., 2011;Andreae and Merlet, 2001), which were summarized by Wiedinmyer et al. (2011) ( Table 2).

Evaluate the crop field burning emission
The provincial CO emissions of CFB were estimated based on Eq.
(1), and there was 8.2 Tg CO emitted from CFB in 2014 (Table S1). This result is comparable to previous studies, which were 4.6-10.1 Tg yr −1 (Cao et al., 2008;Ni et al., 2015;Streets et al., 2003;Yan et al., 2006). According to the MODIS observations, a large number of CFB occurred in SNCP, including the provinces of Henan with 61 % and Shandong with 22 %. Most of CFB occurred on 6 and 7 October, accounting for 75 % (Table 3). Table 4 shows the CFB emissions of gaseous and particulate species on 6 and 7 October, including the mega cities of Beijing and Tianjin, and provinces of Hebei, Henan and Shandong in NCP. Figure 3 displays the CFB activities and related CO emission on 6 and 7 October. Most of the pollutants are emitted from Henan in SNCP, accounting for 73 % on 6 October and 65 % on 7 October. Plenty of pollutants emitted from CFB on 6 October, producing more than 5.1 Gg PM 2.5 and 98.0 Gg CO (1 Gg = 10 9 g).

Statistical characteristics of the evaluation
The characteristics of the haze pollution were defined by PM 2.5 concentration, which is significantly affected by the local wind fields and PBLH in the NCP region . In order to evaluate the model performance, the model simulations were compared with the measured results in both species concentrations (PM 2.5 , O 3 and NO 2 ) and meteorological parameters (wind speed, wind direction and PBLH). The normalized mean bias (NMB) and correlation coefficient (R) were used to quantify the performance. Table 2. The gaseous and particulate species emission factors (g kg −1 ) and scales to convert NMOC emissions (kg day −1 ) to WRF-CHEM chemical species (moles-species day −1 ) from crop field burning. The detailed chemical species are described by Stockwell et al. (1990 where P i is the predicted results and O i represents the related observations. N is the total number of the predictions used for comparisons. Meanwhile, P and O are the average prediction and related mean observation, respectively. Figure 4 shows the measured and calculated temporal variations of regional average species concentrations, including PM 2.5 , O 3 and NO 2 . The WRF-CHEM model reproduced the pollution episode well, with a good agreement with observations. The R of simulated and measured PM 2.5 concentrations are 0.88 in both NNCP and SNCP (Fig. 4a). The simulations are overall lower than the observations with NMB of −12 % in NNCP and −7 % in SNCP. Considering the high average PM 2.5 concentration with 200.0 µg m −3 in NNCP and 184.1 µg m −3 in SNCP, obvious underestimates exist with the overall concentrations of 24.0 µg m −3 in NNCP and 12.9 µg m −3 in SNPC. This may be related to the CMAQ (version 4.6) aerosol module, which is likely to underestimated OM (Organic Matter) due to the uncertainty in secondary organic aerosols mechanism (Baek et al., 2011). Meanwhile, the underestimates are also related to the negative bias in S3, which may be related to cloud contamination ( Fig. S1 in the Supplement). Whereas this has only a few impacts on the estimation of CFB contributions since few CFBs occurred during S3. The simulations of O 3 and NO 2 also agree well with observations, with R greater than 0.77 and absolute NMB lower than 17 % (Fig. 4b and c). Figure 5 displays the measured and calculated temporal variations of regional average meteorological parameters, including wind speed, wind direction, and the PBLH. The comparisons between simulated and observed wind fields show good agreements in both NNCP and SNCP ( Fig. 5a and b), with R all higher than 0.64, and the absolute NMB is no more than 15 %. Meanwhile, the R of PBLH is larger than 0.88 and the absolute NMB is no more than 10 % (Fig. 5c).
-S1 (pollution formation): it is dominated by a continuous southerly wind, with a mean wind speed of 2.5 (2.7) m s −1 in NNCP and 3.0 (3.6) m s −1 in SNCP. The backward trajectories, with the HYSPLIT model online version, of Beijing, Tianjin and Baoding during S1 reflected how the CFB influenced the NNCP region (Fig. 6). The air mass mainly came from the south, originating from the SNCP region. The pollutants are continuously transported from SNCP to NNCP, leading to pollutants accumulation in NNCP, which is characterized by the steady rising of PM 2.5 concentration in NNCP from 20.6 (41.0) µg m −3 (at 12:00, 6 October) to 242.7 (217.5) µg m −3 (at 00:00 LT on 8 October) (Fig. 4  a1).
-S2 (pollution outbreak): during S2, the air pollution deteriorates. It is a relative stable period of heavy pollution with average PM 2.5 concentration of 252.0 (241.2) µg m −3 in NNCP and 214.1 (235.0) µg m −3 in SNCP, which are higher than those in other stages. This phenomenon may be related to the relatively lower wind speed and PBLH.
-S3 (pollution clear): during S3, the southerly winds gradually decrease, and turn to northerly winds at the end of S3. Clean airs from the north region of China obviously improve the air quality. Compared with S2, the average PM 2.5 concentrations are decreased in both NNCP and SNCP.
There were several important issues shown in the results, and should be addressed. (1) The PM 2.5 concentrations are extremely high during the S2 period, and the daily average concentrations are greater than the Chinese National Standard (75 µg m −3 ) by 2-3 times. (2) The air pollution is severe in a large region (occurred in both NNCP and SNCP).
(3) During the S1 and S2 periods, there is a time lag between SNCP and NNCP for PM 2.5 concentrations. Because it is a continuous southerly wind condition, it shows the important impact of long-range transport of PM 2.5 particles from the SNCP to NNCP.

Contributions of crop field burning
Model sensitivity studies were conducted to separate the individual CFB contribution. Two model simulations were performed, i.e., one with both anthropogenic and CFB emissions while the other with only anthropogenic emission. We calculated PM 2.5 distributions by including CFB emissions (anthropologic and CFB) and excluding CFB emissions (only anthropologic). In this study, the CFB contributions were quantified by regional average contribution in mass concentration (CPM 2.5 ) and daily average contribution proportion (PPM 2.5 ): where TPM 2.5 represents the simulated PM 2.5 concentrations considering total emission; APM 2.5 denotes the simulated PM 2.5 concentrations only considering anthropologic emissions. CPM 2.5 and TPM 2.5 are daily average value for CPM 2.5 and TPM 2.5 , respectively. Figure 7 displays the regional observed and simulated PM 2.5 concentrations considering total emissions (anthropologic and CFB) and only anthropologic emissions. It is clearly shown that the CFB had important contributions to PM 2.5 in both NNCP (Fig. 7a) and SNCP (Fig. 7b). This is  The detailed hourly CFB contributions to PM 2.5 concentrations (CPM 2.5 ) are displayed in Fig. 8. The values of CPM 2.5 in NNCP are generally lag synchronized with that in SNCP, such as P N1 versus P S1 and P N2 versus to P S2 ( Fig. 8a and b). Apparently, the lagged time is not constant and varied with the wind fields. The specific details perform relaxed lag synchronized, especially between the P N 2 and P S2 . This phenomenon further indicates that the CFB contribution in SNCP is mainly due to local emission, whereas the CFB contribution in NNCP is largely a result of long-range transport from SNCP. Indeed, the CFB pollution plume goes through a long-range transport to NNCP can cause an obvious increase to PM 2.5 concentration, with the maximum daily average contribution of 32 % (Table 5)    "sim-total" represents the simulations considering total emissions (anthropologic and crop field burning), whereas "sim-anthro" is the simulations only considering anthropologic emissions. the significant local pollution sources, but also a considerable regional pollution source.
To clearly show the time evolution of the CFB effect on PM 2.5 concentration, four time points were defined in Fig. 8c, such as T1 (23:00, 6 Oct), T2 (05:00, 7 Oct), T3 (20:00, 7 Oct) and T4 (19:00, 8 Oct). At T1, prominent CFB contribution occurred in SNCP with the highest value of 71.9 µg m −3 , but accompanied with unimportant CFB contribution in NNCP with a low value of 7.7 µg m −3 . At T2, the CFB contribution in SNCP decline with a relatively high value of 44.2 µg m −3 , but a rise in NNCP with 51.6 µg m −3 (near the transition between P1 and P2). At T3, the CFB contribution rapidly decreases to a low value of 24.0 µg m −3 in SNCP, but increase to the highest with 47.0 µg m −3 in NNCP. At T4, the CFB contributions largely decrease, becoming lesser in both SCNP (9.1 µg m −3 ) and NNCP (11.4 µg m −3 ). Interestingly, the CFB contribution in SNCP drops faster than that in NNCP (P2 in Fig. 8c), resulting in stronger effects in NNCP than in SNCP, as well as longer effects in NNCP.
To further understand the evolution of CFB to heavy haze pollution, we analyzed the horizontal distributions of PM 2.5 concentration (TPM 2.5 ) and related CFB contribution (CPM 2.5 ) at T1, T2, T3 and T4 (Fig. 9). The pattern comparisons between simulated and observed near-surface PM 2.5 concentrations (TPM 2.5 ) perform well (Fig. 9, left panels). Meanwhile, the regional average CFB contributions are shown in Table 6, including mass concentration and related percentage as well as the related time lag of NNCP corresponding to SNCP. At T1, massive local pollutants are Figure 9. The distributions of TPM 2.5 and CPM 2.5 of the key point-in-local-times of T1, T2, T3 and T4, which represent different pollution phases of emission from CFB to PM 2.5 . Left panels also show the pattern comparisons of simulated vs. observed near-surface PM 2.5 concentrations (TPM 2.5 ), with PM 2.5 observations of colored circles. Black arrows denote simulated surface winds. emitted from CFB in SNCP and the CFB plume had not yet been largely transported to NNCP (see CPM 2.5 of Fig. 9 T1). The CFB contribution is high in SNCP with 72.6 µg m −3 , accounting for 71 % of the total PM 2.5 , whereas the CFB contribution is low with 8.1 µg m −3 in NNCP, only accounting for 21 %. At T2, high CFB contribution occurred in both SNCP and NNCP with 37 µg m −3 , suggesting that plenty of CFB pollutants were emitted from SNCP and were transported to NNCP (see CPM 2.5 of Fig. 9 T2). At T3, CFB contribution rapidly reduced in SNCP with 20.2 µg m −3 (13 %). It is worth noting that the high CFB contribution with 50.4 µg m −3 (58 %) still remained in NNCP (see CPM 2.5 of Fig. 9 T3). At T4, the CFB contribution largely decreased in both SNCP and NNCP (no more than 6 %) (see CPM 2.5 of Fig. 9 T4). The time lag of NNCP to SNCP is 7-12 h, and gradually increases from T1 to T4, implicating that the effect of CFB remains longer in NNCP than in SNCP. The highest PM 2.5 concentrations are along the foothills of the Taihang Mountains (Left panels of Fig. 9), which may be related to the mountain effects.

Impact of mountains
Sensitivity experiments were conducted to quantify the impacts of the Taihang Mountains (referred as R-T), the Yanshan Mountains (R-Y) and both of them (R-TY) on heavy pollution. The mountains were removed from the model calculation, in which the altitude of mountains were reduced to the average altitude of NCP (30 m). With the reduction of altitudes of the topography, the dynamical conditions calculated from WRF-CHEM changed, which affect pollution transport, especially along the foothills of mountains. In this study, we utilized the differences between the simulations with or without mountains to represent the effect of the topography on PM 2.5 concentration, which were calculated based on Eq. (9). As an online dynamical model, the topography changes in WRF-CHEM can lead to dynamical changes, such as the wind speeds at the foothills of the mountains. This is a useful and traditional sensitivity analysis method for numerical model to quantify the mountains effects, but with some shortcomings, which are to bring uncertainties to the sensitivity experiment. First, the impact of topography is too complicated to be completely quantified only by the altitude remove behavior. Second, the initial NCEP FNL data with mountains are treated as "real" in scenarios without mountains. The sensitive configuration and related enclosing scope are displayed in Fig. S2.
where IPM 2.5 is the net impacts of mountains on PM 2.5 ; RPM 2.5 denotes the simulated PM 2.5 concentration with removal behaviors, involving R-TY, R-T and R-Y; TPM 2.5 rep- resents the simulated PM 2.5 concentration considering emission of anthropologic and CFB, which corresponds to the case of R0 (Fig. S2a). The sensitivity study period was selected from 12:00 to 00:00 LT on 7 to 10 October. Figure 10 displays the elevation contours and the horizontal distributions of PM 2.5 concentration with the effect of mountains, exhibiting a good performance of the pattern comparisons between simulated and observed near-surface PM 2.5 concentrations. The results illustrate that the mountains had important impacts on regional PM 2.5 concentration, especially for the region along the foothills of mountains with a heavy pollution area, covering sampling sites of BJ, BD, SJZ and XT. Here, it is attributed to the mountain blocking effect, which has two categories of influences. First, the mountains block the airflows, causing pollutant accumulation and resulting in high PM 2.5 loading at the foothills of mountains (influence-1, block). Second, the mountains redirect the airflows, causing the pollutants to move toward the downwind foothill areas (influence-2, redirect). Both influences act to prevent the pollutant plume to disperse toward western mountains, causing accumulations of the air pollutants along the foothills of mountains. These two influences of mountain blocking effects are illustrated as schematic pictures in Fig. S3. Figure 11 displays the simulated PM 2.5 concentration due to the mountain effects (RPM 2.5 ), with the three cases (R-TY, R-T and R-Y). The heavy pollution accumulation (Fig. 10) along the foothills of mountains is significantly reduced, especially with the removal of Taihang Mountains (R-T and R-TY) (Fig. 11 a1 and a2). In these two cases, the pollution plumes dispersed westerly (Fig. 11 b1 and b2). The Figure 11. The average spatial distribution of PM 2.5 concentration and horizontal winds during 12:00 to 00:00 LT on 7 to 10 Oct. (a) Simulated PM 2.5 loading with remove behaviors RPM 2.5 , involving R-TY, R-T and R-Y. (b) The related impacts of mountains on PM 2.5 (IPM 2.5 ), which represent the effect of related mountains. The bold black lines were used to stress enclosing scope of each remove behavior. PM 2.5 concentrations increase 40-120 µg m −3 in the western part of Taihang Mountains, and reduce 20-60 µg m −3 in NCP. The distribution of the reduced pollution plume shows a northeast band plume, indicating the mountain blocking effect. With the removal of the Yanshan Mountains (R-Y), the high PM 2.5 concentrations still remained along the foothills of the Taihang Mountains (Fig. 11 a3), but more pollutants are pushed forward along the foothill, toward the northeastern NCP. Without the blocking effect of the Yanshan Mountains, the PM 2.5 concentrations increased 20-80 µg m −3 in the northern part of the Yanshan Mountains, and decreased 10-60 µg m −3 in the southern part of the Yanshan Mountains (Fig. 11 b3).
In the foothill sampling sites (BJ, BD, SJZ and XT), the average PM 2.5 concentrations are reduced 54.2 µg m −3 for the case of R-T, which is much higher than the case of R-Y (28.4 µg m −3 ). For the other non-foothill sites, the average reduction is 34.7 µg m −3 for the case of R-T, which is also much higher than the case of R-Y (2.4 µg m −3 ), sug-gesting that the Taihang Mountains have stronger effects than the Yanshan Mountains. Meanwhile, the higher impacts in the foothill sampling sites than non-foothill sites are further demonstrated.

Conclusions
In recent years, the NCP region, including the capital city of Beijing, has suffered serious haze pollution problem, especially in winter and summer. Most studies concerned about the intense secondary formation, huge regional transport of pollutants, stationary meteorological conditions and large local emission. In autumn, CFB and movement of wind based on large-scale topography are important in NCP, whereas the percentage of transported CFB emission sources are seldom specified. This probably resulted from the contingency of CFB activities during harvest period and the limitation of temporal resolution of CFB emission inventories. In this study, we extracted a more detailed CFB emission inventory based on the provincial statistical data and CFB activities captured by MODIS. The WRF-CHEM mode was applied to study the effect of CFB on the PM 2.5 concentrations in NCP, especially the evaluation of CFB plums pollution, such as local influence and long-range transportation. We get some insights of how could CFB affect the air quality in NNCP and Beijing under heavy haze condition, though more and longer studies are needed to get more representative conclusions. The results are summarized: 1. A more detailed CFB emission inventory was generated in NCP. The daily CFB emissions were estimated depending on CFB activities captured by MODIS. Plenty of pollutants emitted from SNCP on 6 and 7 October, producing plenty of PM 2.5 pollution, but few in NNCP during the entire haze period.
2. The WRF-CHEM model reproduced the pollution episode with a good agreement with observations. The correlation coefficients (R) of simulated and measured PM 2.5 concentration are 0.88 in both NNCP and SNCP, and the related NMB are −12 % in NNCP and −7 % in SNCP. The simulated winds and PBLH are also in good agreement with observations in both NNCP and SNCP.
3. The WRF-CHEM model was used to investigate the impacts of CFB contribution and its evaluation on PM 2.5 concentration. The SNCP region is mainly influenced by the local CFB emissions, causing a maximum of 34 % PM 2.5 increase. Whereas the NNCP region is mainly affected by the long-range transport of pollution plume emitted from CFB in SNCP, causing a maximum of 32 % PM 2.5 increase in NNCP.

The research domain includes two regions of interests.
One is the NNCP, including two mega cities (Beijing and Tianjin), where few CFBs occurred. Another is the SNCP, where substantial CFB occurred. This study shows that there is a substantially long-range transport of CFB plume from SNCP to NNCP. More importantly, the effect of CFB remains longer in NNCP than in SNCP along the foothill areas of the Taihang Mountains, causing significant enhancement in Beijing in both time and magnitude.
5. Another major finding is that the mountains, surrounding the NCP in the north and west, play significant roles in enhancing the PM 2.5 pollution in NNCP through the blocking effect. Mountains block and redirect the airflows, causing the pollution accumulation along the foothills of mountains. The Taihang Mountains had greater impacts on PM 2.5 concentration than the Yanshan Mountains.
On account of various factors, such as pollutant long-range transport and pollutant accumulation caused by mountain effects, the prohibition of CFB should be strict not only in or around Beijing, but also on the ulterior crop growth areas of SNCP. Other PM 2.5 emissions in the SNCP should be significantly limited in order to reduce the occurrences of heavy haze events in NNCP region, including the Beijing City.