Grid computing method for atmospheric environmental capacity coupled with ventilation coefficient using CALPUFF simulation and GIS spatial analysis technology

ABSTRACT Atmospheric environmental issues have evolved from point source pollution to regional pollution, leading to controlling specific air pollutant emissions. A-value method has been found suitable for estimating large-scale atmospheric environmental capacity rather than small-scale, resulting in the inaccuracy of developing air pollution control strategy. This study proposed a grid computing method based on the CALPUFF modelling system and GIS spatial analysis tool. The meteorological data from the MM5 model were used to simulate the spatial distribution of air pollutants. The meteorological flow field data was used to simulate the ventilation coefficient. The A value was revised with the simulated to achieve accurate results of atmospheric environment capacity. The credibility was verified by applying this method to Fengtai District, Beijing, China. The research area was divided into small partitions via the ArcGIS spatial analysis tool. The simulation results agreed well with the observation data from actual monitoring stations, even for the PM10 concentration with the most significant error (MRE: 7.05%−13.28%, RMSE: 11.62–17.89, R2: 0.84–0.90). The GIS spatial analysis tools were applied to match the underlying surface types and overcome the restrictions of administrative boundary management. The study proposed four schemes to achieve differentiated air pollutant emission reduction and develop suitable control strategies. Furthermore, this method can be applied on different scales of natural geographic boundaries and realize the precise spatial management of atmospheric environmental capacity. GRAPHICAL ABSTRACT


Introduction
Atmospheric environmental capacity is the maximum allowable quantity of air pollutant emissions when the concentration of air pollutants is no greater than the environmental target in an area [1].The rapid, energyintensive, and coal-fuelled economic growth in China has led to severe air pollution issues [2].Since the 1990s, atmospheric environmental issues have developed from local air pollution caused by bituminous coal combustion to a larger scale of regional air pollution [3].The pollution control strategies have evolved from controlling the emission concentration of specific air pollutants to supervising the overall air quality.Studies on atmospheric environmental capacity are drawing increasing attention in Chinese air quality management, as atmospheric environmental capacity plays a vital role in designing and implementing air pollution control measures [4].
The concentrations of air pollutants need to be calculated first for evaluating atmospheric environmental capacity.The calculation methods include the Atmospheric Dispersion Modelling (ADMS) [5], California Puff Dispersion Modelling (CALPUFF), AMS/EPA Regulatory Model (AERMOD) [6,7], Community Multiscale Air Quality Modeling System (CMAQ) [8], Comprehensive Air Quality Model with Extensions (CAMx) [9] and Weather Research and Forecasting Model (WRF) [10] model coupled with Chemistry (WRF-Chem) [11].Among these methods, CALPUFF is an advanced nonsteady-state meteorological model that has been successfully applied on local and regional scales to conduct air pollutant dispersion simulations of the complex terrain [12].
For atmospheric environmental capacity calculation, the A-value method, the A-P value or box model method [1], the simulation and linear programming method [13], and the system dynamics (SD) method [14] are mainly adopted in China.The A-value method is the Chinese national standard approach for determining atmospheric environmental capacity (GB/T 3840-91: Technical method for making local emission standards of air pollutants) [15].The CALPUFF model and the Avalue method are also highly suitable for estimating atmospheric environmental capacity on a large scale and simulating the distribution of air pollutant emissions in a region [16].However, the differences in the spatial distribution of meteorological conditions are not sufficiently considered when calculating the atmospheric environmental capacity, leading to the inaccuracy of air control strategy formulation in small control areas (e.g.streets).The application of spatial analysis [17] may be significant [18,19] because it reflects spatial variability [20] amongst different areas in the same region [21].
In this study, the CALPUFF model was applied to simulate and analyze the meteorological flow field of different heights and calculate the spatial distribution of four primary pollutants concentrations (i.e.SO 2 , NO X , PM 2.5 , and PM 10 ) in Fengtai District.The Geological Information System (GIS) tool was applied to estimate atmospheric environmental capacity and distribution of air pollutant emissions in every small control area.The simulation results in different control areas will assist in developing comprehensive atmospheric environment control schemes, regional industrial-economic layout plans, and appropriate regional environmental risk prevention and control systems.

Research area
Fengtai District is the third-largest urban area in Beijing with an area of 306.00 km 2 and is located southwest of Beijing.The district lies in the warm temperate zone, featuring a semi-humid monsoon continental climate and four distinct seasons (arid and windy spring, hot and rainy summer, fresh and sunny autumn, and cold and dry winter).The district receives 575.50 mm of precipitation annually, and the annual mean temperature is 11.70°C.The dispersion of air pollutants in this district is greatly affected by meteorological conditions (wind speed, wind direction, and atmospheric stability), especially wind direction.During the Eleventh and Twelfth Five-year Plan, increasing economic activity and energy demand lead to a significant increase in air pollutant emissions in this district.Therefore, Fengtai District was selected as the research area (within the red line of Figure 1).

Data description
Emission data of air pollutants.The emission inventory was obtained from the Multi-resolution Emission Inventory for China (MEIC) (http://meicmodel.org/) of Tsinghua University [22].
Land-use and land cover change (LUCC) data and geological data.The LUCC data and geological data of Fengtai District were obtained from the Geographic Information Systems resource database (www.WebGIS.com) and the United States Geological Survey (USGS) Eros data centre (www.usgs.gov/centers/eros).
Meteorological data.Hourly observation data of meteorological parameters in 2015 (wind direction, wind speed, dry bulb temperature, low cloud amount, air pressure, and relative humidity) were derived from the Information Center of the China Meteorological Administration.The initial meteorological data were the NCEP/NCAR Reanalysis data set from The United States National Centers for Environmental Prediction (www.weather.gov/organization/ncep).The high-altitude meteorological data were derived from mesoscale meteorological simulation in MM5 (https://opensky.ucar.edu/islandora/object/technotes:170).The three monitoring stations in the research area were insufficient for obtaining the detailed meteorological data, and the simulation area was extended to Beijing City.The data are provided in Table 1.

CALPUFF model
California Puff Dispersion Modeling (CALPUFF) is an advanced non-steady-state meteorological and air quality modelling system developed by scientists at Exponent, Inc.The modelling system can determine the mixing height, surface characteristics [23], and dispersion properties of air pollutants [24].The main components of the modelling system are CALMET (a diagnostic three-dimensional meteorological model), CALPUFF (an air quality dispersion model), and CALPOST (a post-processing package).
The basic mathematical representations of the model are given in Equations ( 1) and (2) [25]: where C is the ground-level concentration of air pollutants (g/m 3 ), Q is the pollutant mass (g) in the puff, σ x , σ y , and σ z are the standard deviation (m) of the Gaussian distribution in the along-wind, cross-wind, and vertical direction respectively, d a and d c are the distance (m) from the puff centre to the receptor in the along-wind and cross-wind direction, respectively, g is the vertical term of the Gaussian equation, H e is the effective height (m) above the ground of the puff centre and h is the mixing height (m).

Simulation of meteorological flow field
Objective environmental factors (meteorological, topographical, and human factors) may diminish the accuracy of calculating atmospheric environmental capacity via the A-value method as meteorological factors will impact the atmospheric environment and the distribution of air pollutants [26].Fengtai District is high in the northwest and descends to the southeast with a step-by-step decline in terrain.The western and eastern parts are mountains and account for approximately 3/4 of the region.This district also features many high-rise buildings that could affect the nearsurface diffusion conditions of the region [27].Therefore, the rough terrain conditions suitable for CALPUFF simulation could not reflect the spatial differences in the research area.Furthermore, wind speed is a large-scale meteorological parameter with no apparent geographic boundaries.The research area is too small to get the accurate simulation data of the ventilation coefficient, as the wind speed is too fast.Under these circumstances, the meteorological parameters should be studied.The meteorological flow field was simulated via CALMET model from CALPUFF system, and the results were used to calculate the ventilation coefficient.The spatial distribution of the meteorological flow field was plotted via the CALVIEW drawing module from CALPUFF system.The daily meteorological flow field data of the simulation were categorized into four periods (02:00, 08:00, 14:00, and 20:00) in 4 typical days of spring (March 21, March equinox), summer (June 22, summer solstice), autumn (September 23, September equinox) and winter (December 21, winter solstice) of 2015 (Table 2) (Supplementary Information [SI] Figure S1-S4).The A value was furtherly modified according to the ventilation coefficient to improve the calculation accuracy of the Avalue method.The data and model version were the same with the simulation of the spatial distribution of air pollutant concentration.

Simulation of the spatial distribution of air pollutant concentration
The simulation area was extended from the research area (with only three air quality monitoring stations) to the Beijing City (with 34 air quality monitoring stations) to obtain precise results of air pollutant concentration.During the simulation process, grids were developed to cover the simulation area (180 km × 180 km) at a cell size of 1 km.Time parameters (year, month, day, hour), meteorological parameters (wind direction, wind speed, dry bulb temperature, low cloud amount, atmospheric pressure, and relative humidity), and terrain parameters (elevation, land use, surface water identification, and vegetation structure and composition) were fully considered based on the characteristics of meteorology and terrain in the research area.The CALPUFF model (version 6.42) was applied to simulate the spatial distribution of four primary pollutants (i.e.SO 2 , NO X , PM 2.5 , and PM 10 ) in every partitioned grid.The Briggs plume rise algorithm was applied to calculate the rise of the pollutant plume caused by effluent buoyancy and exit momentum.The plume penetration into the stable atmospheric stratification and the transitional plume rise were also considered.The simulated meteorological flow field data from CALMET and the simulated annual average concentration data of four pollutants from CALPUFF were combined to deduct the average annual concentration distribution of four pollutants in Fengtai District.

Calculation method of atmospheric environmental capacity
The A-value method was based on the box model and the control area was regarded as a box [28].The bottom and top of the box represent the underlying surface and top of the atmospheric mixed layer, respectively.The circumference was determined by the range of the area [29].
The atmospheric environmental capacity was calculated using Equation (3): where A is the atmospheric environmental coefficient (Avalue, 10 4 km 2 /a), representing the physical self-purification of the atmosphere, C si is the standard annual average pollutant concentration limit (mg/m 3 ), S is the total research area and S I is the functional area (S = n i=1 S I , km 2 , the research area is divided into n functional areas and S i is one of them).Considering the impact of meteorological conditions, the A value can be further modified using the ventilation coefficient in the research area.The mathematical representation is given in Equation ( 4): where V C is the ventilation coefficient, calculated as where m is the regional average wind speed and H F is the height of the atmospheric mixing layer).
According to the Guidelines for Environmental Impact Assessment Atmospheric Environment (HJ/T2.2-93),when the atmospheric stability is at the level of A, B, C, or D (Tables 3 and 4): And when the regional atmospheric stability is E or F (Tables 3 and 4): where H is the mixing height (m), U 10 is the average wind speed at 10 m height (m/s), μ is the ground rotational speed (calculated value, 7.29 × 10 −5 rad/s), φ is the geographic latitude of the research area (39.8°N in this study), and a and b are atmospheric mixing layer coefficients.
The above method was used to calculate H under different atmospheric stability conditions in the research area.
Considering the background concentration of air pollutants in the environment, the mathematical representation for estimating the environmental capacity in the research area is: where C b is the background concentration of pollutants in each cell of the grid applied to the research area with a resolution of 1 km × 1 km.In this study, C b adopted the simulation results of the spatial distribution of air pollutant concentration.

Spatial analysis of atmospheric environmental capacity
The vector data of Beijing City was segmented into fishnet grids using the Data Management Tool in ArcGIS [18], and a basic grid map with 17,568 grids was generated with a grid resolution of 1 km × 1 km (477 of which covered the research area).All the simulations were carried out based on the generated grid map.Each grid contains several simulation values of air pollutant concentration and ventilation coefficient.The arithmetic mean value was used to represent the spatial distribution of air pollutant concentration and ventilation coefficient to simplify the calculation process.The overlay analysis of the grid map and CALPUFF simulation results were carried out to get the arithmetic mean value in each grid using the Zonal Statistics function of ArcGIS.

Simulation results of ventilation coefficient
According to the simulation results, the vertical direction of the CALMET model contains ten layers, with heights of 20, 40, 80, 160, 320, 640, 1200, 2000, 3000, and 4000 m, and the research area was divided into grids with a horizontal grid resolution of 1 km × 1 km.The mixing height (H) and wind speed were considered as crucial parameters for evaluating the ventilation coefficient of an urban area [30].In this study, H in each cell was 700 m (Table 4), and the average wind speed was 2.30 m/s (Table 2).The ventilation coefficients in all grids were calculated and illustrated in Figure 2.This result proved that the ventilation coefficient variated closely to the meteorological changes even in a small research area.Therefore, the atmospheric environmental capacity calculation accuracy will be improved as the traditional A-valued method did not modify the A value with ventilation variation.

Simulation results and spatial distribution of annual average concentrations of air pollutants
The results indicated that the annual average concentrations of PM 10 , SO 2 , and NO X were below the concentration limits, while PM 2.5 exceeded.The percentage exceeding the concentration limits of NO X and PM 10 were 0.41% and 20.00%, respectively.Detailed information was provided in Table 5.
As illustrated in Figure 3, areas around Wangzuo Town and Yungang Street in the northwest part of the research area had the highest SO 2 concentration, and areas around Yungang Street also had the highest concentration of NO X .Areas with high concentrations of PM 2.5 were mainly located around the northwest border between Mentougou and Shijingshan Districts, while the north-western and western parts of the research area had remarkably high PM 10 concentrations.
The results revealed that the concentrations of PM 2.5 [31] and PM 10 [32] in Beijing are pretty high, and the spatial distributions determined were consistent with air pollution sources in Fengtai District.

Verification of the simulation results
The observation values of three monitoring stations were selected to verify the accuracy of the simulation results.Differences between the simulation results and observation values of the annual average air pollutant concentration were 1-5 μg/m 3 for SO 2 , NO X , and PM 2.5 and 10 μg/m 3 for PM 10 .Therefore, the results of PM 10 were adapted for evaluating the credibility of the simulation, as the differences were the largest.As shown in Figure 4, all the R 2 values exceeded 0.84, all the MRE values were within ±13.28%, and all RMSE values were within ±17.89.According to these results, the simulated data showed considerable agreement with the observed data, which provided a strong foundation for calculating atmospheric environmental capacity.

Spatial analysis of atmospheric environmental capacity
The integration of the ventilation coefficient with the spatial distribution of the annual average air pollutant concentration revealed differences in the atmospheric environmental capacity at a local scale (Figure 5).
As shown in Figure 5, the atmospheric environmental capacity of PM 10 ranged from −200 to −50, with a difference of 150 in the whole research area.The atmospheric environmental capacity of SO 2 ranged from 150 to 350 with a gradient difference    of 200.For PM 2.5 and NO X , the atmospheric environmental capacity ranged from −40 to 125 with a difference of 165 and ranged from −100 to 275 with a difference of 375, respectively.According to these results, the critical control regions for SO 2 and NO X should be the northwest and north of Wangzuo Town.Same with NO X , the critical control region for PM 2.5 should be the north of Wangzuo Town.For PM 10 , the critical control regions should be the south-eastern and north-eastern parts of the research area.
The atmospheric environmental capacity of different air pollutants at street level in Fengtai District was illustrated in Figure 6.The total environmental capacities for SO 2 , NO X , PM 2.5 , and PM 10 were 54547.46,29594.82,−21001.38,and −20139.57t, respectively.The remaining atmospheric environmental capacities of SO 2 and NO X at all streets are high, while the remaining atmospheric environmental capacities of PM 2.5 and PM 10 exceeded regional standards.

Spatial analysis of air pollution and pollution risk
According to simulation results of air pollutant concentration and atmospheric environmental capacity, all the grids in the research area were categorized into high-risk grids for NO X and high-risk grids for PM 10 and PM 2.5 .Statistical analysis of the spatial distribution of air pollutant concentration in the whole area was conducted by considering 60 point-source and several non-point-source emissions.As shown in Figure 7, the results revealed that the point-source emissions of SO 2 were very high in Changxindian and Donggaodi Streets, and the point-source emissions of NO X were high in Lugouqiao Street and west of Nanyuan Street.
Based on the analysis of emission intensity of air pollutants, the numerical segmentation was applied to the raster datasets of air pollutant concentration using the Raster Reclass tools of GIS spatial analysis function.The segmentation results were furtherly used to analyze the spatial distribution characteristic of atmospheric environmental quality and the pollution risk of different air pollutants.
As shown in Figure 8, the research area was categorized into three risk-level control areas (i.e.low-risk area, medium-risk area, and high-risk area) to develop air pollution prevention and control strategies.The risk level distribution of SO 2 and NO X were roughly the same, with high-risk areas in the western part of the research area.For PM 2.5 , most parts were moderately polluted, and the heavily polluted parts were located north of Wangzuo Town, Yungang Street, and the northern part of Taipingqiao Street.For PM 10 , most parts were also heavily polluted, and a few parts were lightly polluted.Most streets in the eastern part of the research area were heavily polluted, and the moderately polluted parts were located at the west of Fengtai Street.Only the middle of Changxindian Street was slightly polluted.These results generally reflected high pollution risks in the west and low pollution risks in the east.Therefore, to control the regional atmospheric environmental capacity and the total amount of emissions, it is necessary to comprehensively consider the impact of meteorological differences in the entire area rather than control the emission intensity of air pollutants.

Control strategy for total air pollutants
According to the spatial analysis results of air pollution and pollution risk, the atmospheric environment control schemes for Fengtai District are listed below: Scheme 1: Reducing the total emission of smoke and dust (20768.51t) by coordinating with external regions.
Scheme 2: Conducting the total control of air pollutants in different parts of Fengtai District by allocating control units.The detailed index allocation of control units is shown in Figure 9. Scheme 3: . Strengthening the dust management system in urban areas. .Promoting the implementation of 'coal-to-electricity' and 'coal-to-gas' policies.  .Proposing clean energy reduce the air pollution caused by coal consumption.
Scheme 4: Optimizing the spatial layout of the research area and stopping the construction of new residential areas within 1 km of the existing pollution sources.For existing residential areas near the pollution sources, creating a green gridded network zone and setting up isolation belts to reduce air pollution.Besides, the set-up of ventilation corridors can also improve the diffusion conditions of polluted areas.
The spatial distribution results of the atmospheric environmental capacity of different pollutants in the research area may be effectively promote targeted comprehensive management and develop industrialeconomic layout plans from different regional perspectives.

Conclusions
This work proposed a novel method for calculating atmospheric environmental capacity at a small scale (i.e.street level) by combining CALPUFF simulation system with GIS spatial analysis technology.The CALPUFF system was applied for ventilation coefficient and air pollutant concentration simulation, and the atmospheric environmental capacity was calculated by combining the simulation results with the Zonal Statistics tool in ArcGIS.The calculation results of this method  well with the observation data from the monitoring stations in the research area and the shortage of monitoring stations could also be made up.The atmospheric environmental capacity calculation accuracy was improved by modifying the A value with the ventilation coefficient.The calculated annual average concentration of SO 2 was lower than the concentration limit, and the concentration of NO X , PM 10 , and PM 2.5 were 0.41%, 20.00%, and 100.00% exceed the corresponding concentration limits.The variation of atmospheric environmental capacity in different parts of the research area indicated that this method could be applied to small-scale areas to provide accurate calculations.Based on the spatial analysis results, the air pollution control strategy of the research area could be refined to street level, and the targeted, comprehensive management and industrial-economic layout plans could be formulated.For future research, the modification of the A-value could be extended to other factors such as meteorology, topography, and other influencing factors to obtain more accurate results.Furthermore, as Jiang et al., proposed [33], the concentration of PM2.5 varies widely in the vertical direction, which should be considered when using CALPUFF model for atmospheric environmental capacity calculation at an hourly scale.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Figure 1 .
Figure 1.Location of Fengtai District in China, Beijing (yellow area with red border) and location of the monitoring stations in Fengtai District (purple dots).

Figure 3 .
Figure 3. Spatial distribution of annual average concentration (μg/m 3 ) of SO 2 , NO X , PM 2.5 and PM 10 in Fengtai District.

Figure 4 .
Figure 4. Comparison of monthly simulation results with the observation values of PM 10 at three monitoring stations.

Figure 5 .
Figure 5. Spatial distribution of the atmospheric environmental capacities (t) of SO 2 , NO X , PM 2.5 and PM 10 in Fengtai District in 2015.

Figure 6 .
Figure 6.Atmospheric environmental capacity of SO 2 , NO X , PM 2.5 and PM 10 at street level in Fengtai District.

Figure 7 .
Figure 7. Spatial distribution of air pollutant emissions in Fengtai District in 2015.

Figure 8 .
Figure 8. Pollution risk level analysis of SO 2 , NO X , PM 2.5 and PM 10 in Fengtai District in 2015.

Figure 9 .
Figure 9. Spatial distribution of atmospheric pollution sources and environmental control areas in Fengtai District.

Table 1 .
Information of data sources.

Table 2 .
Flow fields in spring, summer, autumn, and winter.

Table 4 .
Mixing heights of the research area under different stability.

Table 5 .
Statistical results of air pollutant concentrations in Fengtai District.