Numerical simulations of the effects of regional topography on haze pollution in Beijing

In addition to weather conditions and pollutant emissions, the degree to which topography influences the occurrence and development of haze pollution in downtown Beijing and the mechanisms that may be involved remain open questions. A series of atmospheric chemistry simulations are executed by using the online-coupled Weather Research and Forecasting with Chemistry (WRF-Chem) model for November-December 2015 with different hypothetical topographic height scenarios. The simulation results show that topography exerts an important influence on haze pollution in downtown Beijing, particularly the typical development of haze pollution. A possible mechanism that underlies the response of haze pollution to topography is that the mountains that surround Beijing tend to produce anomalous southerly winds, high relative humidity, low boundary layer heights, and sinking motion over most of Beijing. These conditions favor the formation and development of haze pollution in downtown Beijing. Furthermore, the reduction percentage in PM2.5 concentrations due to reduced terrain height in the southerly wind (S) mode is almost three times larger than that in the northerly wind (N) mode. In the context of the regional topography, the simple S and N modes represent useful indicators for haze prediction in Beijing to some extent, especially over medium to long time scales.

Beijing is located in the northern tip of the North China Plain and surrounded by mountains along the west, north and northeast. Beijing is the capital of China and one of the largest cities in the world; it has with approximately 20 million residents. Given the rapid urbanization and industrial development over the past several decades, Beijing and its adjacent areas are becoming more and more important in China and for the global economy. However, the rapid economic growth and urbanization have increased the level of air pollution in recent decades [1][2][3][4][5][6][7][8] . Beijing and eastern China have suffered from severe haze or smog days frequently in recent years. These events are characterized by high particle mass concentrations and low visibility. Severe haze pollution, exemplified by the persistent haze days in January 2013, represent a substantial threat to human health and traffic safety. These phenomena have stimulated great interest in studying the haze pollution in Beijing or eastern China as a whole [9][10][11][12][13][14][15][16][17][18] . Very serious haze pollution struck again in Beijing in November-December 2015, and heavy pollution red alerts were issued on December 7 th and 18 th , 2015 (although, according to pollutant concentration records, these dates were not the worst cases over that period November-December 2015). That is the first time the capital issued a red alert (the most serious level) for air pollution in Chinese history [19][20][21] .
Haze pollution is generally attributed to the following two factors: the emission of pollutants to the lower atmosphere from fossil fuel combustion, construction and other sources and unfavorable meteorological conditions. Air quality and the occurrence of haze pollution are strongly influenced by meteorology. Meteorological factors have essential impacts on the accumulation or diffusion, spread and regional transport of air pollutants and have important impacts on the formation of secondary aerosols, which are generated by complex physical and chemical reactions [22][23][24][25][26][27][28][29][30][31] . In particular, weather conditions play an essential role in controlling the daily variability of air pollutant concentrations 5,[32][33][34] . Previous studies have suggested that there is a close relationship between the occurrence of winter haze pollution in the Beijing-Tianjin-Hebei region and atmospheric circulation at middle to high latitudes over a long-term perspective [35][36][37] . In addition, on interannual time scales, the air pollution across central and eastern China or even South Asia in the winter and summer is closely related to the East Asian winter and summer monsoons, respectively [38][39][40][41][42][43][44][45] .
Terrain or topography may be another non-negligible factor that influences the haze pollution that occurs in Beijing and eastern China. For example, a previous study has suggested that the "harbor" effect on the westerlies in the eastern lee of the Tibetan Plateau's large topography may be an important factor that influences the regional distribution of haze frequency in eastern China 46 . Beijing is located in the northern part of the North China Plain (Fig. 1a). The Taihang Mountains are located west of Beijing, whereas the Yanshan Mountains lie to the north and northeast of Beijing. Topographical profiles show that the altitude differences in both the west-east and north-south directions exceed 1000 m over very short horizontal distances (Fig. 1c,d). Thus, whether and how the regional topographic variations influence the occurrence and development of haze pollution in Beijing remain open questions. To date, few studies have examined the possible influence of topography on haze in Beijing or other areas quantitatively 47,48 . A close understanding of the effects of topography on air pollution assists in planning the spatial layout of industry areas and emission controls in Beijing and the adjacent areas and is also useful in improving the prediction skill achieved by weather and air quality forecasts produced by numerical models. This study aims to investigate the effects of topography on the occurrence and development of haze pollution using numerical sensitivity experiments.

Results
Analysis of observational data. The daily mean PM 2.5 concentrations in downtown Beijing in November-December 2015 are presented in Fig. 2. The standard deviations determined from the 24-hour records from each day are indicated by error bars. Days with a high PM 2.5 concentrations and a small error bars, such as December 1 st and December 25 th , indicate heavy pollution throughout the day. In contrast, a large error bar generally indicates intense fluctuations in the PM 2.5 concentrations, which may reflect either the accumulation or elimination of pollutants or both. It is clear that most (36 days, approximately 59%) of the daily PM 2.5 concentrations exceed 75 μg/m 3 , which suggest polluted conditions, according to the ambient air quality standards of China. The mean concentration over this period was 140.6 μg/m 3 , which is far larger than the annual mean for 2015 (80.   Fig. 3a,b, respectively. As seen in the observations, most of Beijing, Tianjin, the south-central portion of Hebei province and the northwest part of Shandong province experience moderate (≥115 μg/m 3 ) to high levels (≥150 μg/m 3 ) of pollution during these two months. The spatial distribution of the simulated mean PM 2.5 concentrations is generally consistent with that of the observations. Heavy pollution extends from the south-central portion of Hebei province along the foot of the Taihang Mountains to the Beijing areas. The spatial correlation coefficient between the observed mean PM 2.5 concentrations at all of the stations (770 in total) located within the simulation domain and the corresponding PM 2.5 concentrations derived from the simulations is 0.62 (p < 0.01). The temporal correlation coefficients between the observed and simulated PM 2.5 concentrations are also calculated; on the daily and hourly time scales, the percentages of positive correlations are 93.8% and 94.2%, and the mean correlation coefficients are 0.34 and 0.29, respectively. These significant correlations suggest that the simulations generally capture the spatial distribution patterns and temporal fluctuations of the PM 2.5 concentrations found in the observations over the entire areas. In downtown Beijing, the correlation coefficients between the observed and simulated PM 2.5 concentrations are 0.48 and 0.46 on the daily and hourly time scales, respectively. Both of these correlations are significant at the 0.01 level. The comparison of the temporal variations in the observed and simulated hourly PM 2.5 concentrations in  downtown Beijing (Fig. 3c) indicates that the simulated PM 2.5 variability is generally consistent with the observed, although the model does not capture the peak concentrations. Note that the skill of the simulation in representing the typical development of haze events (featured by gradual increases in PM 2.5 concentrations) is better than its ability to represent the intense fluctuations in haze (featured by rapid increases or decreases in PM 2.5 concentrations over very short time periods). Generally, the simulation results are reliable and useful in examining the effects of topography on the occurrence of haze pollution in Beijing.

Sensitivity analysis.
We first examine the general characteristics of the response of PM 2.5 concentrations to changes in topography. As the terrain heights drop, the PM 2.5 concentrations decrease over most of the North China Plain, especially from south of Beijing to the south-central portion of Hebei province (Fig. 4a-c). In contrast, the PM 2.5 concentrations clearly increase over most of the North China Plain as the terrain heights is increased (Fig. 4d). Moreover, the responses of other important pollutants, such as PM 10 , SO 2 , NO 2 , and CO, to topographic height changes are roughly consistent with those of PM 2.5 (not shown). The daily variations in the concentrations of PM 2.5 , PM 10 , SO 2 , NO 2 , and CO in downtown Beijing from November to December 2015 derived from the simulations performed using different topographical scenarios are shown in Fig. 5. The pollutant concentrations generally increase as the terrain heights increases, especially during the typical development of haze, such as during the period of the 7 th to the 16 th of November (Fig. 5).
To quantitatively understand the influence of the regional topography on haze pollution in downtown Beijing in the different scenarios, the mean concentrations of PM 2.5 , PM 10 , SO 2 , NO 2 , and CO derived from all of the scenario simulations over the entire period and the typical haze events are calculated. When the topographic heights are scaled by 0.75 (S2), 0.50 (S3) and 0.25 (S4), all of the pollutant concentrations in downtown Beijing are clearly less than those in scenario S1 over both the entire period and the typical haze period. In contrast, when the height of the topography is scaled by 1.25 (S5), all of the pollutant concentrations in downtown Beijing are larger than these of S1 in both periods (Table 1). We further calculated the percentage changes in the pollutant concentrations in the scenarios with changed terrain heights (S2, S3, S4 and S5) relative to the scenario S1 (Table 2). During November to December 2015, the PM 2.5 concentrations in downtown Beijing decrease by 9.2%, 18.3%, and 26.1% as the topographic heights are lowered by 25%, 50%, and 75%, respectively, whereas the concentration increases by 12.9% when the topographic height increases by 25%. During the typical development of haze, the PM 2.5 concentrations appear to decrease by 15.3%, 36.5%, and 59.2% as the topographic heights decrease by 25%, 50%, and 75%, respectively; however, they increase by 10.4% when the topographic height increases by 25%. The percentage changes in the PM 10 concentrations are very similar to the percentage changes in the concentrations of PM 2.5 , SO 2 , NO 2 , and CO. However, there is an exception; namely, no distinct changes in the NO 2 concentrations occur in scenarios S5 relative to S1 during the period of November 7 th to 16 th (0.1%).

Discussion
Analysis of the relevant mechanisms. The simulation results suggest that topography indeed exerts an important influence on air pollutant concentrations in downtown Beijing. We thus wonder what the possible mechanisms underlie the relationships and whether the modulation depends on the favorable weather conditions. First, we investigate the response of surface winds and relative humidity to the topographic changes. Fig. 6 shows the composite anomaly fields of surface wind vectors and relative humidity during November 7 th to 16 th . The anomalous northerly winds and low relative humidity become dominant in Beijing and the surrounding areas as the terrain is lowered, especially when the terrain drops by 50% (Fig. 6b) and 75% (Fig. 6c). In contrast, most of the study area is dominated by anomalous southerly winds and high relative humidity when the terrain height increases (Fig. 6d). The anomalous southerly winds are likely accompanied by static winds, due to the blocking produced by the high topography to the north and west. A similar response is also found in the planetary boundary layer height (PBLH), which increases (decreases) over Beijing and its adjacent areas when the terrain height drops (increases). Furthermore, the profiles of zonal and vertical winds over central Beijing in the simulations that evaluate different scenarios are examined. The results show that increases in terrain height may lead to a sinking vertical vortex in the lower troposphere on the eastern lee of the regional topography, which promotes the development of sinking motion over downtown Beijing (Fig. 7). This sinking motion typically depresses the vertical exchange in pollutants and is often accompanied by or leads to a lower PBLH and weak winds. This phenomenon is similar to the "harbor" effect on the westerlies in the eastern lee of the Tibetan Plateau's large topography, which may be an important factor that influences the regional distribution of haze frequency in eastern China 47 .
Thus a simple conceptual model can be described as that increase in terrain height lead to anomalous southerly winds, high relative humidity, low boundary layer heights, and sinking motion, which favor the formation and development of haze pollution in Beijing, and vice versa.

Implications of the S and N modes. The observed data show that the PM 2.5 concentrations in downtown
Beijing are distinctly related to the wind direction. The southerly and northerly winds correspond to high and low (or no) pollutions, respectively (Fig. 8a). The relationship between the winds and haze pollution in downtown Beijing is well reproduced in the atmospheric chemistry simulation (Fig. 8b). This result suggests that the simple southerly (S) and northerly (N) wind modes are very useful predictors of haze in downtown Beijing to some extent. Thus, it is of interest to understand the influence of topography on haze pollution in downtown Beijing under the different wind modes. The percentage changes in the pollutant concentrations in the different scenarios and the different wind modes are calculated (Table 3). For example, in the context of the N mode, the PM 2.5 concentrations decreases by 5.0%, 9.6%, and 14.5% as the terrain height drops by 25%, 50%, 75%, respectively. However, in the context of the S mode, the PM 2.5 concentrations decrease by 13.5%, 27.0%, and 37.6%, respectively. The reduction percentage in PM 2.5 concentrations due to terrain height decreases in the S mode is almost Figure 6. The composite anomaly fields of surface wind vectors and relative humidity for S2-S1 (a), S3-S1 (b), S4-S1 (c) and S5-S1 (d), respectively. three times larger than that in the N mode. This opposite relationship is noted when the terrain height increases by 25%. Similar characteristics are also found in the concentrations of other pollutants, such as PM 10 , SO 2 , NO 2 , and CO (Table 3). The response of pollutants concentrations to terrain height changes are closely related to the responses of meteorological variables, such as wind speed, relative humidity, and the PBLH. When the terrain height drops, the increases (decreases) in wind speed and the PBLH (RH) in the S mode are larger than those in the N mode, and vice versa (Table 4). Thus, characterizing the different modulating effects of topography on haze pollution in downtown Beijing under the S and N modes is important. This information would be very useful in producing medium-range to long-term (monthly to seasonal) predictions or evaluations of air pollution.
Uncertainties. Moreover, note that some uncertainties exist in the qualitative response of haze pollution to terrain height changes, due to the different spatial scales involved. In principle, differences are expected between the influence of regional topography on haze pollution in downtown Beijing and the influence of national-or continental-scale topography. Thus, further study is needed to explore this issue.

Conclusion
The possible effects of different topographic scenarios on haze pollution in downtown Beijing are quantitatively examined using the online-coupled WRF-Chem model, output from the FNL reanalysis obtained from the National Centers for Environmental Prediction (NCEP), high-resolution anthropogenic emissions data derived from the Multi-resolution Emission Inventory for China (MEIC) and the hourly PM 2.5 records generated at surface stations. The results suggest that the terrain height indeed plays an important role in controlling the haze pollution in downtown Beijing, especially during the typical development of haze, which featured by a gradual increase in PM 2.5 concentrations. Concretely, the PM 2.5 concentrations in downtown Beijing decrease by approximately 9.2% (15.3%), 18.3% (36.5%), and 26.1% (59.2%) as the topographic height in most of northern China is lowered by 25%, 50% and 75%, and it increases by 12.9% (10.4%) as the topographic height is increased by 25%  during November-December 2015 (during a typical haze event), respectively. The statistical relationships are consistent with the simple mechanisms demonstrated by these simulations. For example, increases in terrain height lead to anomalous southerly winds, high relative humidity, low boundary layer heights, and sinking motion in the lower troposphere, which favor to the formation and development of haze pollution in downtown Beijing, and vice versa. Moreover, the reduction percentage in PM 2.5 concentrations due to reductions in terrain height in the southerly wind mode (S mode) is almost three times larger than that in the northerly wind mode (N mode), suggesting that the S and N modes represent good indicators of haze pollution in downtown Beijing in relevant operations or studies, especially over medium to long time scales.

Methods
Data and study area. The hourly concentrations of PM 2.5 (particulate matter with aerodynamic diameters less than 2.5 μm) in downtown Beijing used in this study are derived from eight stations operated by the Ministry of Environmental Protection of the People's Republic of China during November-December 2015 (Fig. 1b). Moreover, the PM 2.5 concentration data measured at the American Embassy station are also used for comparison and to fill in missing data using regression. From November 1 st to December 31 st in 2015, the percentages of missing hourly PM 2.5 concentration records at the American Embassy station and the average of the eight operational stations are 0.68% (10 hours missing) and 3.62% (53 hours missing), respectively. In fact, the hourly PM 2.5 records from the American Embassy station are highly consistent with the average values for the eight stations. The correlation coefficient is 0.96, which is significant at the 0.01 level (i.e., p < 0.01). The difference in their mean values is only 0.81 μg/m 3 . To better represent the data across the city, we take the average records from the eight stations as the actual PM 2.5 concentration for downtown Beijing in this study. The meteorological records from five stations in downtown Beijing are shown in Fig. 1b (solid blue circles). The spatial scope in Fig. 1a also represents the model domain of the WRF-Chem simulations. For the daily mean values calculated here, a meteorological day is defined as the 20:00-20:00 (local time) throughout the paper.
Model and simulation design. To investigate the influence of topography on haze pollution in Beijing, a series of sensitivity simulations involving five scenarios of topographic changes are evaluated using version 3.7 of the online-coupled Weather Research and Forecasting with Chemistry (WRF-Chem) model [49][50][51][52][53] . The first simulation involves the real (unchanged) terrain heights (labeled as S1). In addition, the real terrain heights are then scaled by 0.75, 0.50, 0.25, and 1.25 in the second (S2), third (S3), fourth (S4), and fifth (S5) scenario simulations, respectively. The output from the final (FNL) reanalysis produced by the Global Forecast System (GFS) of the National Centers for Environmental Prediction (NCEP) is used as the lateral meteorological boundary conditions in the simulations. The temporal and spatial resolutions of the FNL data used here are 6 hours and 1.0° × 1.0°, respectively. The primary settings and schemes used in the WRF-Chem simulations are shown in Table 5. The
To better capture the process of haze development near the surface, 13 levels are specified under 1,500 m. A monthly dataset that describes anthropogenic emissions in 2012 at high spatial resolution (0.1° × 0.1°) is used in the simulations and was provided by the Multi-resolution Emission Inventory for China (MEIC) 56,57 .
All of the simulations cover November-December 2015 because a very serious haze pollution event struck Beijing in November-December 2015, and red alerts were issued for the first time in Chinese history, as mentioned in the Introduction. Both the typical development of haze events (which feature a gradual increase in PM 2.5 concentrations, i.e., November 7 th to 16 th ) and the fast-changing haze events (which feature explosive growth and/ or dramatic reductions in PM 2.5 concentrations, i.e., November 29 th to December 2 nd and December 22 nd to 26 th ) occurred during this haze season. Thus, this study increase our understanding of the possible effects of topography on haze pollution in downtown Beijing under different weather and pollution conditions.  Table 5. List of the experimental settings and schemes used in WRF-Chem.