Source Apportionment of PM2.5 in Handan City, China Using a Combined Method of Receptor Model and Chemical Transport Model

  • Zhe Wei
  • Litao Wang
  • Liquan Hou
  • Hongmei Zhang
  • Liang Yue
  • Wei Wei
  • Simeng Ma
  • Chengyu Zhang
  • Xiao Ma
Conference paper
Part of the Environmental Earth Sciences book series (EESCI)

Abstract

Handan is one of the top polluted cities in China, characterized by high concentration of fine particulate matter (PM2.5). In this paper, a receptor model, i.e., the Positive Matrix Factorization (PMF) model, and a chemical transport model, i.e., the Mesoscale Modeling System Generation 5 (MM5) and Models-3/Community Multiscale Air Quality (CMAQ) model, are both applied to apportion the sources of PM2.5 in Handan. It is concluded that regional sources contribute 36.0% of PM2.5, and within local sources, the contributions of major emission sectors are: 22.3% from coal combustion, 10.7% from metal smelting, 7.3% from Zn-OC-Ba, 18.5% from industry, 11.3% from transportation, 10.6% from biomass burning, and 19.2% from dust emissions. It indicates that regional joint air pollution controls should be emphasized in the future control strategy, and local source controls on coal combustion and industries are the key points to mitigate the severe PM2.5 pollution in Handan.

Keywords

PM2.5 Source apportionment PMF MM5-CMAQ Handan 

1 Introduction

Urban air pollution, especially the pollution of fine particulate matter (PM2.5) is one of the most urgent environmental problems to be resolved in China and has raised wide public, governmental, and academic concerns. Hebei province, located in northeastern China, is the top polluted province according to the reports on air quality status publicized by the Ministry of Environmental Protection (MEP) [1, 2]. In 2013 and 2014, seven out of the ten top polluted cities in China are within Hebei province [1, 2]. Handan city, located at the southern edge of Hebei, has an area of 1.2 × 104 km2 and a population of 9.0 million, with about 1.5 million living in the urban area of 419 km2. It is a heavily industrialized city with a large amount of productions of iron and steel, cement, and coke [3]. Air pollution in Handan is very severe and it ranked as the third top polluted city in China in 2013 and the fifth in 2014 [1, 2].

Handan is also a city with complex regional sources surrounding it. It is in the intersection of four provinces, Hebei, Henan, Shandong, and Shanxi (see Fig. 1a). All of them are heavily populated, industrialized, and urbanized and their major air pollutants emissions, e.g., PM2.5, PM10 (atmospheric aerosols with aerodynamic diameter less than or equal to 10 µm), sulfur dioxide (SO2), and nitrogen oxides (NOX) are about as high as one fourth of the total emissions in China [4]. It results in a complex problem of source apportionment of PM2.5 in Handan, which is urgently required by the local government to design an effective control strategy.
Fig. 1

a CMAQ modeling domains at a horizontal grid resolution of 36 km over East Asia (Domain 1 with 164 × 97 cells) and 12 km over an area in northern China (domain 2 with 93 × 111 cells); b location of the PM2.5 sampling site in Handan (the red star, referred to as HEBEU site)

Receptor models and chemical transport models are both effective methods for PM2.5 source apportionment [5, 6]. The Positive Matrix Factorization (PMF) model, as one of the receptor models recommended by US Environmental Protection Agency (US EPA), has been successfully applied in many cities in Asia [7, 8, 9, 10, 11]. Some studies used multiple source apportionment approaches to improve the robustness of the results and understand the spatial characteristics [12, 13]. Contini et al. [13, 14] compared the source contributions of PM10 or PM2.5 at different sites in Venice Lagoon. However, multi-site samplings are not available in Handan before 2015. The Mesoscale Modeling System Generation 5 (MM5) and the Models-3/Community Multiscale Air Quality (CMAQ) model, as one of the source oriented, chemical transport model, has been proved to be an effective tool to identify the sources of air pollution in Chinese cities [15, 16, 17, 18, 19, 20]. PM2.5 source apportionment of PM2.5 in Handan use of those tools was reported in several recent studies [4, 21, 22]. However, neither of them can produce enough detailed and reliable results to fully support the policy making of local government in Handan, because the receptor models cannot easily distinguish the spatial origins of each source sector [5, 6], unless multiple-site sampling data and more advanced analysis methods (e.g., microscopic methods) are employed or trajectory models using detailed meteorological data are employed, which were not available in Handan for this study. As mentioned above, and the reliability of the results from the source-oriented models are highly dependent on the precision of the emission inventory, which is far from developed for the sources in Hebei at this time. Bove et al. [23], applied a hybrid method using the PMF model and the chemical transport model of Comprehensive Air Quality Model with Extensions (CAMx) in the city of Genoa in Italy and found the two methods gave consistent results even the source categories were different. In their study, PAST (Particulate matter Source Apportionment Technology) was used to calculate source contributions [23]. In this paper, the Brute-force method, a source sensitivity method, is applied to calculate source contributions of PM2.5. This method, although computationally expensive, is straightforward and can simulate the non-linear relationship among PM and PM precursors under large emission perturbations. The detailed explanation on choosing this method is described elsewhere [4].

Therefore, in this study a combined method using both a receptor model, the PMF, and a chemical transport model, the MM5-CMAQ, is applied to PM2.5 source apportionment in Handan, e.g., the results by MM5-CMAQ model are used to distinguish the regional or local contributions of PM2.5, and the receptor model is used to apportion the sectoral contributions to limit the uncertainties from the MM5-CMAQ introduced by the uncertainties and incompleteness of the present emission inventory. This study might be the best estimation we can achieve on the source apportionment of PM2.5 in Handan, given the present data and research basis. This paper is organized as below: Sect. 2 describes the method of PM2.5 sampling and analysis, configurations and inputs of the PMF and the MM5-CMAQ model. Section 3 introduces the results of the two methods and the combination analysis. Section 4 summarizes the major findings and conclusions of this paper.

2 Methodology

2.1 PM2.5 Sampling and Analysis

The PM2.5 sampling site is located on the top of a 12 m high building at Hebei University of Engineering (HEBEU) in Handan, Hebei, China (36°34′ N, 114°29′ E) (See Fig. 1b). This location is at the south edge of the urban area of Handan, representing a cultural area, and it is also one of the four official online monitoring sites in Handan. There is no obvious emission source around the site. Four months in 2013, i.e., January, April, July, and October, representing the four seasons, were selected to collect PM2.5 samples continuously from 8:00 to 7:30 a.m. on the next day, using a high volume PM2.5 air sampler (Thermo Scientific Co., model number: HIVOL-CVBLD) with 20.3 × 25.4 cm quartz filters. The flow rate was 1.13 m3 min−1. First, the sampling filters were heated at 350 °C to remove the organic material in the filters. Then they were reserved in constant-temperature and humidity chamber for 24 h before the pre-weighting and sampling. The sampled filters were kept in the chamber for another 24 h before the post-weighting. The differences in the two weights are used to calculate sampled PM2.5 concentrations (the sampling concentrations of 23.5 h were converted to daily concentration). Those sampled filters were stored in a refrigerator under −20 °C until chemical analysis. In total, 120 samples were collected in this study.

Water-Soluble Ions (WSI). Water-soluble cations and anions were measured by ion chromatography (Dionex, DX–600, Dionex, ICS–2100, USA). First, a small section of the quartz filter (1 cm2) was put into 10 mL ultrapure water (18.2 MΩ cm) for extracting water-soluble inorganic ions. The extracted solution was filtered through a microporous membrane filter (0.45 μm of pore size), and stored in a refrigerator at −18 °C until sending into ion chromatography. The measured ions include Cl, NO3, SO42−, Na+, K+, Mg2+, Ca2+, and NH4+. Strictly quality control during all procedures avoid contamination of the samples and ensure the accuracy of chemical analysis. The method detection limit (MDL) of NH4+, Na+, K+, Mg2+, Ca2+, SO42−, Cl, and NO3 were 0.01, 0.001, 0.001, 0.004, 0.005, 0.01, 0.005, 0.01, and 0.01 μg ml−1, respectively.

Organic Carbon (OC) and Elemental Carbon (EC). OC and EC were analyzed by the method of Thermo-optical Reflectance (TOR) which used a Thermal/Optical Carbon Analyzer (Model 2001A) produced by the Desert Research Institute (DRI). First, OC1, OC2, OC3, and OC4 were measured at 140 °C, 280 °C, 480 °C, and 580 °C in an environment with pure He gas without O2, respectively. Second, the sampled filter was further heated to 580 °C, 740 °C, and 840 °C to measure EC1, EC2, and EC3 in an environment with 2% O2 and 98% He, respectively. The precision of OC and EC were <5% and <10% along with the MDL of 0.82 μg cm2− for TOC and 0.20 μg cm2− for TEC, respectively.

Elements. Inductively coupled plasma mass spectrometry (ICP–MS) was used to analyze the concentrations of trace elements, i.e., As, Ba, Cr, Cu, Mn, Pb, Sr, Ti, V and Zn, in PM2.5. A piece of quartz filter (1 cm2) was digested for 25 min at 190 °C with 8 ml concentrated nitric acid (BV–III) and 0.5 ml H2O2 (30%) in a Teflon microwave digestion tank (CEM, MARS). Then, the bulk solution was diluted to 100 ml, of which 10 ml were put into ICP-MS to measure the concentrations of the elements. ICP-MS was calibrated by standard injection each several samples analysis. Relative standard deviation (SD) was controlled below 5%, and internal standard recovery was controlled at the range of 80–120%. All of them aim at ensuring the accuracy of detection.

2.2 The Receptor Model—PMF

The PMF model, as a receptor model, is a data analysis technique of factor analysis based on multivariate statistical methods, of which the fundamental ability is to resolve identities and source contributions of components in an unknown mixture. PMF uses the chemical components of PM2.5 to calculate the source contributions without knowledge of emissions inventory [24]. PMF decomposes a matrix of speciated sample data into two matrices: factor contributions (G) and factor profiles (F). These factor profiles need to be interpreted by the user to identify the source types that may be contributing to the sample using measured source profile information, and emissions or discharge inventories. The key equations of this method are listed as below as Eqs. (1) and (2). A speciated data set can be viewed as a data matrix X of i by j dimensions as in the Eq. (1), where i and j represents the number of samples and chemical species, respectively; gik is the contribution of the pth source to the ith sample; and fkj is the concentration of the jth species in the pth source.
$${{\rm{X}}_{{\rm{ij}}}} = \sum\limits_{{\rm{k}} = 1}^{\rm{p}} {{{\rm{g}}_{{\rm{ik}}}}{{\rm{f}}_{{\rm{kj}}}} + {{\rm{e}}_{{\rm{ij}}}}}$$
(1)
$${\text{Q}} = \sum\limits_{{{\text{i}} = 1}}^{\text{n}} {\sum\limits_{{{\text{j}} = 1}}^{\text{m}} {\left[ {\frac{{{\text{e}}_{\text{ij}} }}{{{\text{u}}_{\text{ij}} }}} \right]} }$$
(2)
where eij is the residual for each sample/species, and uij is an uncertainty term introduced to facilitate a statistical solution of the mass balance as opposed to an analytical mathematical solution. The PMF model is constrained to non-negative species concentrations and source contributions, which makes the results meaningful. Also, PMF applies uncertain matrix (uij) to minimize the Q (2) to optimize the results. The PMF model is described in detail in [25, 26]. A brief description can be found in Wei et al. [22].
PMF has been widely used for both regulatory and research applications [8, 22, 27, 28, 29, 30, 31], and is recommended by US EPA [32]. In this paper, the PMF version 5.0 was selected to calculate the source contributions of PM2.5 in Handan in 2013. Ten percent of the concentrations of each species was used as the uncertainty matrix in the dataset, following the suggestion by Zhang et al. [33], considering that the sampling technique (e.g., high volume sampling with quartz filters may lead to larger uncertainties [34]), analytical method applied, and the number of samples analyzed. If the concentration is less than or equal to the method detection limit (MDL), 1/2 MDL is used to replace the concentration and 5/6 MDL is used as uncertainty [35]. Missing data were substituted with the median value and their uncertainties were replaced by four times the median value. Signal to noise (S/N) was used to review the dataset (Eq. 3). Sij indicates standard deviation of ith sample and jth chemical species. The species were categorized as ‘strong’ when S/N ≥ 2, ‘weak’ (0.2 ≤ signal-to-noise ratio ≤ 2), or ‘bad’ (signal-to-noise ratio < 0.2) species. In this study, all species were categorized as “strong”. A fundamental problem using this method is to identify a reliable number of factors. The number of factors is determined as following: (1) most of the residual matrix is within −3.0 to 3.0; (2) the results tend to be stable when adjusting the number of factors. In this paper, the concentration data of Cl, NO3, SO4−2, Na+, K+, Mg2+, Ca2+, NH4+, OC, EC, V, Cr, Mn, Cu, Zn, As, Sr, Ba, and Pb were used to run the PMF model because of all these species were available for the four months that were studied. As for estimating uncertainty of PMF solutions, this paper selected classical bootstrap (BS) error analysis to estimate the PMF results [36, 37]. Default value was applied to run BS error estimation in Base run and Block size option in PMF model. One hundred BS runs were performed to ensure the robustness of the statistics. R-value was set as 0.6. The first, the second, the third and the fourth factor were mapped with BS in 99, 77, 87, and 97% of the runs, while other factors were mapped 100% of runs. Therefore, the authors were convinced that 8 factors were selected to explain the source contribution of PM2.5.
$$\frac{{\rm{S}}}{{\rm{N}}} = \frac{1}{2}\sqrt {\sum\nolimits_{{\rm{i}} = 1}^{\rm{n}} {{\rm{x}}_{{\rm{ij}}}^2} /\sum\nolimits_{{\rm{i}} = 1}^{\rm{n}} {{\rm{s}}_{{\rm{ij}}}^2} }$$
(3)

2.3 The Chemical Transport Model—MM5-CMAQ

The CMAQ model, a 3-D chemical transport model developed by US EPA [24], has been increasingly used to simulate regional air quality in China in the past decade [15, 16, 17, 18, 19, 20]. The MM5 model, a regional mesoscale meteorological model developed by the National Center for Atmospheric Research (NCAR) and then Penn State University [38], is applied to generate meteorological field to drive the CMAQ model. The source apportionment method used in this study is the so-called Brute-Force method [39], that is, the source contributions of each source region or sector are estimated by calculating the differences between the baseline simulation and the sensitivity simulation in which the emissions from that source region or sector are set as zero. This method has been proven to be an effective way to quantify the source contributions to support the policy making in air pollution controls in China in recent years [15, 16, 17, 18, 19, 20].

The detailed descriptions on the model configurations, performance evaluation, simulation scenarios and results can be found in [40, 41]. In brief, the modeling domains are seen in Fig. 1a. Two nested domains are applied over East Asia and North China at 36 and 12 km horizontal grid resolutions, respectively. The model configurations and inputs are summarized in Table 1. Besides the base case, twenty-four emission reduction scenarios are simulated, including the emission zero-out simulations of the nine source regions and the five emission sectors in the three cities, Handan, Shijiazhuang and Xingtai (in this study only the results for Handan are used). The nine source regions (see Fig. 1a) are Handan (the local sources), Shijiazhuang, Xingtai, the northern Hebei (other Hebei area besides the above three cities), Beijing and Tianjin (BJTJ), Shanxi (SX), Shandong (SD), Henan (HN), and the other regions (OTH, other regions in Domain 1 besides the above eight). The five emission sectors locally include power plant (PO), industrial (IN), domestic (DO), transportation (TR), and agriculture (AG). Therefore, the local (anthropogenic emissions within Handan) and regional (emissions outside Handan) source contributions to PM2.5 in Handan, and the contributions of each emission sector within Handan are qualified by using those simulations, which is absolutely necessary for policy making in air pollution controls of the local government.
Table 1

Configurations and inputs of MM5-CMAQ modeling in this study

Item

Configuration and input

References and notes

Simulation period

Jan. and Jul., 2013

5-spin up days

Domain

East Asia (36 km), northern China (12 km)

Horizontal resolution

36 and 12 km

Vertical resolution

23 layers (MM5) and 14 layers (CMAQ) from 1000 to 100 mb

Nesting

Two way for MM5; one way for CMAQ

MM5

Version 3.7

CMAQ

Version 4.7.1

Anthropogenic emissions

MEIC

[http://www.meicmodel.org/]

Cumulus scheme

Kain–Fritsch

Kain and Fritsch [42]

PBL scheme

Blackadar

Zhang and Anthes [43]

Moisture scheme

Mixed phase (Reisner 1) explicit

Reisner et al. [44]

Longwave and shortwave radiation

Cloud atmospheric radiation scheme

Dudhia [45]

Surface scheme

Force/restore (Blackadar)

Blackadar [46], Deardorff [47]

FDDA

Wind, temperature, and water vapor mixing ratio in and above PBL

Gas phase chemistry

SAPRC-99 with aqueous and aerosol extensions

Carter [48, 49]

Aerosol phase model

AERO5

Binkowski and Shankar [50]

Aqueous-phase chemistry

Updated regional acid deposition model (RADM)

Chang et al. [51], Walcek and Taylor [52]

Land use data

US geological survey terrain and land use data

Meteorological initial and boundary conditions

National center for environmental prediction (NCEP) final (FNL) operational global analysis data

FDDA input

NCEP automated data processing (ADP) surface and upper air data

Chemical initial and boundary conditions

Clean profile in CMAQ for 36 km domain; nested from 36 to 12 km domain

Photolysis rates processor (JPROC) input

Ozone column data from the Ozone Measurement Instrument (OMI) on the Aura

[http://toms.gsfc.nasa.gov/ozone/ozone_v8.html]

Emission scenarios

Zero-out anthropogenic emissions in Handan, Shijiazhuang, Xingtai, northern Hebei (NHB), Beijing and Tianjin (BJTJ), Shanxi (SX), Shandong (SD), Henan (HN), and other area (OTH)

Zero-out emissions from power plant (PO), industrial (IN), domestic (DO), transportation (TR), and agriculture (AG), one by one, in Handan, Shijiazhuang, and Xingtai

Total 24 scenarios, only results of sectoral zero-out scenarios in Handan are used in this study

3 Results and Discussions

3.1 Chemical Composition of PM2.5 in Handan

Table 2 presents the concentrations of PM2.5 and WSI, OC, and EC in PM2.5 in the four representative months in 2013. It can be seen that the highest monthly-average concentration of PM2.5 was 309.2 μg m−3 in January, followed by 213.3 μg m−3 in October, 103.2 μg m−3 in April, and 85.6 μg m−3 in July. In general, the WSI are the most abundant species in PM2.5, accounted for 43.2% of PM2.5 in January, 48.9% in April, 55.9% in July, and 36.2% in October. The highest fraction of WSI appears in July due to the rapid reaction speed in secondary aerosol formation, e.g., NO3, SO42− and NH4+, led by higher temperature. OC accounted for 15.1% of PM2.5 in January, followed by 13.5% in October, 10.6% in April, and 7.9% in July.
Table 2

The concentrations of PM2.5 and WSI, OC, and EC in PM2.5 in Handan in 2013

 

PM2.5

Cl

NO3

SO42−

Na+

K+

Mg2+

Ca2+

NH4+

OC

EC

(μg m−3)

Jan.

Ave.

309.2

7.3

23.2

36.3

1.3

3.2

0.2

0.5

23.8

33.3

14.8

Min

62.4

1.0

5.0

7.0

0.5

1.1

0.0

0.0

5.6

8.5

4.2

Max

643.0

17.3

43.4

78.2

2.4

6.4

0.4

2.5

43.1

63.7

33.8

S.D.

133.3

3.9

8.4

14.7

0.5

1.2

0.1

0.7

7.8

13.8

7.6

Apr.

Ave.

103.2

3.2

17.2

16.2

0.5

1.1

0.1

2.0

7.3

10.2

3.9

Min

34.8

0.8

3.2

4.9

0.0

0.5

0.0

0.3

2.5

3.5

1.6

Max

224.2

9.2

38.7

29.4

1.3

2.3

0.3

4.4

13.2

18.0

7.9

S.D.

42.3

2.1

9.2

7.0

0.3

0.5

0.1

1.0

3.1

3.2

1.3

Jul.

Ave.

85.6

1.1

12.5

24.8

0.4

0.8

0.1

0.4

9.0

6.1

2.6

Min

36.7

−0.2

2.1

7.7

−0.1

0.2

0.0

0.0

2.1

2.1

1.1

Max

190.5

4.0

37.3

59.7

1.4

1.5

0.2

0.9

19.1

8.8

3.8

S.D.

37.6

1.2

8.0

12.4

0.4

0.3

0.0

0.2

3.7

1.9

0.8

Oct.

Ave.

213.3

5.8

31.6

23.6

0.6

2.2

0.2

1.1

11.5

28.5

4.4

Min

38.6

1.4

4.1

5.3

0.2

0.3

0.0

0.1

3.0

2.3

1.0

Max

402.6

9.9

76.4

51.1

1.4

3.6

0.3

2.4

21.6

67.2

6.4

S.D.

95.0

2.3

14.8

10.3

0.3

0.8

0.1

0.6

4.2

14.1

1.0

S.D. indicates standard deviation

Table 3 presents the concentrations of trace elements in PM2.5 in Handan. They only account for less than 1.0% of mass concentrations in PM2.5. However, the heavy metal elements should be paid more attention to because of their negative effects on human health and the capability of serving as catalyst in the formation of secondary aerosols in the atmosphere. Among those elements, the concentration of Zn is the highest of 314.0 ng m−3, followed by Pb (221.9 ng m−3), Mn (64.4 ng m−3), Ti (49.7 ng m−3), Ba (34.0 ng m−3), As (30.9 ng m−3), Cu (21.4 ng m−3), Sr (8.8 ng m−3), and Cr (6.4 ng m−3). The concentrations of Ca has the highest seasonal variation, with the highest value appearing in April which resulted from dust blowing weather that occurred over northern China.
Table 3

The concentration (ng m−3) of elements in PM2.5 in Handan in 2013

 

January

April

July

October

Ave.

As

30.5

29.3

28.2

35.6

30.9

Ba

11.1

47.7

22.9

54.1

34.0

Cr

10.1

5.5

4.5

5.3

6.4

Cu

28.1

24.0

11.2

22.6

21.4

Mn

80.2

70.9

33.3

73.1

64.4

Pb

350.2

155.3

151.0

231.2

221.9

Sr

5.4

18.6

3.7

7.5

8.8

Ti

13.2

124.5

16.1

45.1

49.7

V

5.9

5.0

0.4

1.7

3.2

Zn

339.0

259.5

246.4

411.2

314.0

3.2 The Results of the PMF Model

Eight factors are identified by PMF model, as shown in Fig. 2. The first factor was recognized as dust/road dust source, because Ca2+ has high loadings in this factor. Ca2+ may also be associated with steel production, cement/concrete, construction activities, and earth crustal material. EC and V emitted from the combustion of diesel oil and deposited into soil. The contaminated soils can be resuspended by mobile transportation or wind blowing. The source of Mn and Ba is related to brake parts on the vehicles. Therefore, this factor was regarded as dust/road dust source, which contributed 12.3% of PM2.5.
Fig. 2

Source profiles in mass concentrations of the factors found by the PMF model

The second factor has high loadings of Cr and Na+. Cr may be from the production process of iron and steel. Because Cr was usually used to produce stainless steel. Iron and steel is a pillar industry in Handan, which produced 51 million tons steel and 45 million tons crude steel in 2013 (http://www.handannews.com.cn/new_epaper/hdrb/html/2014/03/25/content_10718.htm), as well as a large amount of stainless steel (http://www.hgjt.com.cn/index.aspx). Na usually exists in iron ore. As for SO42− and As of the high loadings, they may be from coal combustion because coal was the major energy source for iron and steel industry. Therefore, this factor was identified as metal smelting source, which contributed 11.0% of PM2.5.

K+ is a very important marker of biomass burning. The third factor has high loading of K+ and medium loadings of OC and Cl, so that it is attributed to biomass burning source, with the contributions of 10.7% in PM2.5. However, this paper found that Mn, Zn, Pb, and Cu have relative high loadings in this factor, which were usually from the industrial processing that may be influenced by multi-colinearity of PMF model. Therefore, further analyses should be encouraged.

The fourth factor was seen as a mixed source. The reason is that the factor has high loadings of Ba which may be from the brake parts on vehicles or construction activity. At the same time, this factor has close correlations with Zn, Ca2+, Mg2+, OC and As. In general, As is seen as the marker element of coal combustion. OC was also from the coal combustion. Zn was usually made into various product, including brake of vehicles. Ca2+ and Mg2+ may be from steel production, cement/concrete, construction activities, and earth crustal material as mentioned above. This factor was influenced by dust/road dust. Therefore, this factor could be seen as a mixed source with Zn-OC-Ba, which contributes 7.6% of PM2.5.

The fifth factor is coal combustion sources, which has obvious relationship with As, an important maker of coal combustion. Additionally, coal combustion would emit amounts of organic matter and SO2, which also show high loadings in this factor. Zn and Pb may be from the coal combustion. The source contribution of coal combustion to PM2.5 is 13.8%.

Mn, Cu, Zn, As, Sr, Pb and Ba have high loadings in the sixth factor. Zn and Pb is usually from rolling mills and industry processing [9, 35]. In Handan, the iron and steel industries are very important components of its industries. Mn and Cu are in close correlation with metal smelting industry. Sr and Ba were used for the ceramic industry because of their dielectric properties. Therefore, this factor was identified as an industrial source, which contributes 17.5% of PM2.5.

The seventh factor was characterized by high loading of V, EC, and moderate loadings of Mn, Zn, and Pb. V is an auxiliary material used by oil products. EC could be emitted from the vehicle engines due to incomplete combustion of diesel oil. Mn, Zn and Pb are typically related to the brakes of motors. Additionally, Zn is a component of a common fuel detergent and anti-wear additive which also exists in the diesel emissions, as well as in tires, brake linings and pads. This factor is regarded as a transportation source, which contributes 11.2% to PM2.5.

The eighth factor, with high loadings of NH4+, SO42− and NO3, is seen as the secondary sources, because those species are produced in the atmosphere from the primary pollutants of SO2, NOX and NH3. It contributes 16.0% in PM2.5. A secondary source should be identified easily because secondary components are produced in the atmosphere. However, OC, EC, and Cr have high loadings in this factor. There are some limitations for source apportionments of PM2.5 with the PMF model, such as the colinearity of receptors. Further study and analysis should be needed.

The above results on PM2.5 source apportionment in Handan are summarized in Fig. 3.
Fig. 3

The source contributions of PM2.5 in Handan by the PMF model

3.3 The Results of MM5-CMAQ Model

The contributions of each source region and sector to the secondary PM2.5 are discussed in detail as well in [40], because the receptor model can recognize one source as “secondary aerosol”, i.e., NO3, SO42−, and NH4+, as discussed in Sect. 3.2, but cannot distinguish their spatial and sectoral source of origins [5, 6]. The modeling results show that in the total secondary inorganic aerosols (SIA) the contribution of regional sources is 59.1% on annual average, and the local contributions of PO, IN, DO, TR, and AG are 0.4%, 20.6%, 4.8%, 1.5%, and 17.8% [40], respectively. Those results are summarized in Table 4. Those treatments are not pursued for secondary organic aerosols (SOA) because of the insignificance of OC in the “secondary aerosol” recognized by the PMF model (see Fig. 2), and the weakness of the CMAQ model in predicting SOA concentrations at present.
Table 4

The source contributions (%) of regional and local sector emissions to PM2.5 in Handan by the MM5-CMAQ modela

 

% in total PM2.5

% in local contributions

% in SO42−

% in NO3

% in NH4+

% in total SIA

Regional

40.4

 

51.1

65.1

55.1

59.1

PO

0.9

1.5

1.3

−0.6

0.6

0.4

IN

35.9

59.5

34.1

1.5

23.5

20.6

DO

13.5

22.3

6.7

3.1

5.4

4.8

TR

3.7

6.2

1.3

2.9

1.3

1.5

AG

6.2

10.3

4.8

38.5

20.4

17.8

Total local

60.2

99.8

48.2

45.4

51.2

45.2

aThe sum of local and regional contributions is not 100% due to the non-linear response of PM2.5 to its precursors emissions

3.4 Combined Analysis and Results

3.4.1 Regional Contributions

The receptor model apportions the PM2.5 concentrations by the representative trace species of the recognized sources; so, it is not easy to distinguish the spatial origins of those sources unless data from multiple sampling locations can be employed, which was not available in Handan for this study. Therefore, the results of the MM5-CMAQ model are used to quantify the regional contributions in Handan. It indicates 40.4% of PM2.5 in Handan are from regional sources. However, it should be noted that this number may overestimate the regional contributions because of the lack of dust emissions in the MM5-CMAQ simulations in this study (the online dust module has not been involved in CMAQ until version 5.0), as the dust emissions contribute not a small amount in PM2.5 (12.3%), indicated by the PMF model. Here the assumed dust emissions are all from the local area, given the reasons that in 2013 only one dust event occurs on March 6–9th in Handan, which is not within our sampling period; and the urban fugitive dust emissions are relatively large particles and not so active in long-range transport as other aerosols. Therefore, the regional contributions in PM2.5 are revised as 36.0% (40.4%/(1 + 12.3%)). It is still a considerable amount so that the regional-joint air pollution controls need to be fully considered when designing control strategies for Handan.

Although the results from the PMF model are given priority in the estimation of the local sectoral contributions, the results from the CMAQ model are still necessary in two purposes: (1) distinguishing the contributions from the regional sources and the local sources, as discussed in Sect. 3.4.1; and (2) quantifying the contributions from each local sector to the sector of “secondary aerosols” identified by the PMF model. For the purpose (2), the sectors in the CMAQ model need to be transformed to the sectors of the PMF model. Therefore, Eq. (4) is used to map the contributions from the sectors in the CMAQ simulations, i.e., PO, IN, DO, TR, and AG, to the sectors identified by the receptor model, i.e., coal combustion, industry, transportation and biomass burning:
$${\text{S}}_{\text{j}} = \sum\limits_{{{\text{i}} = 1}}^{\text{n}} {\,\sum\limits_{{{\text{k}} = 1}}^{\text{m}} {\left( {{\text{S}}_{{{\text{i}},{\text{k}}}} \times \frac{{{\text{E}}_{{{\text{j}},{\text{i}},{\text{k}}}} }}{{{\text{E}}_{{{\text{i}},{\text{k}}}} }}} \right)} }$$
(4)
where Sj is the source contributions of the local sector j to PM2.5 (the j sectors are those identified by the PMF model). Si,k is the contributions from the local sector i, which are the sectors in the simulations (PO, IN, DO, TR and AG, n = 5), to the chemical species k in PM2.5. Those species include SO42−, NO3, NH4+, OC, EC, and other PM2.5 (m = 6). Ej,i,k are the emissions of species k or the precursors of k (SO2 to SO42−, NOX to NO3, NH3 to NH4+) from sector j in the sector i, and Ei,k are the total emissions of the k from the sector i. For example, the contributions of coal combustion within the IN sector to sulfate in PM2.5 are calculated by multiplying the sulfate contributions from the IN and the fractions of SO2 emissions by coal combustion within the IN sector to the total IN SO2 emissions (note the IN activity includes coal combustion, other fuel combustion and industrial processes). Then the coal combustion contributions from each sector, e.g., PO, IN, DO, to each species in PM2.5 are added together to obtain the total contributions of coal combustion. The Ei,k and Ej,i,k are calculated according to the MEIC inventory and other references on Hebei emissions [40, 54, 55]. Table 5 summarizes the Ej,i,k/Ei,k for the sectors of PO, IN, and DO to be split into the sectors by the PMF model in Handan.
Table 5

The value of Ej,i,k/Ei,k in the Eq. (4) for primary PM2.5 and its major precursors for the sector mapping

Sectors in CMAQ

Sectors in PMF

Primary PM2.5

SO2

NOX

NH3

VOC

POa

Coal combustion

100

100

100

100

100

IN

Coal combustion

12

76

51

100

0

Other fuel and processes

88

24

49

0

100

DO

Coal combustion

49

93

52

95

23

Biomass burning

51

7

48

5

77

aPower plants in Handan are all coal-fueled

3.4.2 Unity the Local Source Sectors

As for estimating the source contributions from the local sectors using the combined results from the receptor model and the chemical transport model, it firstly should be noted that there are consistencies in sector split in the two modeling system, i.e., the sectors recognized by the PMF model are coal combustion, industrial, transportation, dust/road dust, metal-smelting, biomass burning, Zn-OC-Ba, and secondary; however, those in the CMAQ model are PO (power plant), IN (industrial, including industrial coal combustion, other fuel combustion, and industrial process emissions), DO (domestic, including domestic coal combustion and other fuel combustion), TR (transportation), and AG (agriculture). In this study, the results from the PMF model are given priority on the estimation of the local sectoral contributions in Handan, for the three reasons: (1) the sectors identified by the PMF are more detailed than from the CMAQ model; (2) the local sectoral contributions estimated by the CMAQ model are highly dependent on the precision of the emission inventory, which is still far from developed for the sources in Handan; and (3) according to the guidance on the PM2.5 source apportionment released by the MEP [53], the sectors recognized by the PMF model are acceptable.

The model performance evaluation is presented in [4, 40] in detail. In brief, the meteorological predictions are evaluated against the National Climate Data Center (NCDC) integrated surface database. The statistics include the mean bias (MB), the root mean square error (RMSE), the normalized mean bias (NMB), and the normalized mean error (NME), for the parameters of temperature at 2 m (T2), relative humidity at 2 m (RH2), wind speed at 10 m (WS10), wind direction at 10 m (WD10), and daily precipitation. The air quality observations used in model evaluation are the online observations published by the China National Environmental Monitoring Center (CNEMC). The statistics for PM2.5 and PM10 are MB, NMB, NME, the mean fractional bias (MFB), and the mean fractional error (MFE). Those evaluation database, protocols, and results are explained in detailed in [4, 40]. Here only the source apportionment results were presented from the MM5-CMAQ modeling. It is concluded that the regional source contributions to PM2.5 in Handan are 40.6% [40], which is a non-negligible fraction indicating that policy makers should pay more attention to the regional joint controls of air pollution in Handan. The contributions of local emissions of PO, IN, DO, TR, and AG are 0.9%, 35.9%, 13.5%, 3.7%, and 6.2% in total PM2.5 concentrations, respectively, as 1.5%, 59.5%, 22.3%, 6.2%, and 10.3% in local contributions [40], respectively.

The local sectoral contributions from the CMAQ simulations after sector mapping are summarized in Table 6, for each species in PM2.5, as well as in the total local SIA. Those species, i.e., sulfate, nitrate, and ammonia, are formed through the oxidation, condensation, and nucleation of the precursor gaseous pollutants including SO2, NOx, and NH3.
Table 6

The source contributions (%) of local sectors to each species in PM2.5 by MM5-CMAQ after the sector mappinga

 

SO42−

NO3

NH4+

OC

EC

Others

PM2.5

In local SIA

Coal combustion

33.5

1.7

29.3

6.9

13.5

14.6

16.9

79.8

Industry

8.1

0.7

0.0

19.4

21.4

48.2

26.0

13.5

Transportation

0.7

2.9

1.3

5.9

20.1

2.1

3.7

2.4

Domestic biomass burning

0.5

1.5

0.2

23.7

10.8

6.6

7.3

4.4

Sum

42.7

6.8

30.8

55.9

65.9

71.5

54.0

100

aThis table presents the results of Eq. (4) for PM2.5 as well as each species in PM2.5. The Ej,i,k/Ei,k of SO2, NOX, and NH3 are used for the sector mapping for the species of SO42−, NO3, and NH4+, respectively. The Ej,i,k/Ei,k of primary PM2.5 is used for OC, EC and Others in PM2.5

3.4.3 Calculation of Local Sectoral Contributions

Table 7 presents the calculation process using the combined results of the PMF model and the MM5-CMAQ model. First, it is started from the results of the PMF model (see line 1 in Table 7). Then, it is assumed that the dust emissions are all from local sources, for the reasons mentioned in the above Sect. 3.4.1, and the regional contribution estimations were applied from the MM5-CMAQ model. After that, the contributions of the regional and local dust emissions are determined as 36.0 and 12.3%, as other local source contribution is 51.7% in total (step 1 in Table 7).
Table 7

Calculation of source contributions (%) from regional and local sector emissions in Handan using the combined method of PMF and CMAQ

 

Coal combustion

Industry

Metal smelting

Zn-OC-Ba

Motor

Biomass burning

Secondary

Dust

Regional

PMF model

13.8

17.5

11.0

7.6

11.2

10.7

16.0

12.3

Step 1a

51.7

12.3

36.0

Step 2b

8.5

10.9

6.8

4.7

6.9

6.6

7.2

12.3

36.0

% in total PM 2.5 c

14.3

11.8

6.8

4.7

7.3

6.8

12.3

36.0

% in local contributions

22.3

18.5

10.7

7.3

11.3

10.6

19.2

% in local contributions by CMAQ

25.3

38.9

5.5

10.9

19.2d

aThe results after using the regional contribution calculated by CMAQ and assuming all the dusts are from local emissions

bThe results after using the apportionment factor for secondary inorganic PM2.5 calculated by CMAQ and supposing the relative importance of the local sectors is the same as the original PMF model results

cThe results after attributing the local secondary PM2.5 to each original local sector

dThis number is following the above analysis in this table, not from the CMAQ model due to the lack of dust module

To estimate each local sectoral contribution (except for the “secondary aerosols”), it is assumed that their relative importance in the regional contributions and in the local contributions is similar, e.g., the total industrial contribution (17.5%) is higher than those from coal combustion (13.8%); then the local industrial contribution is higher than the local coal combustion for the same ratio, as well as those from regional.

The secondary aerosols are treated differently because they are formed in the atmosphere from gaseous precursors and need to be further identified as their spatial and sectoral origins to support the future controls. In Table 4 the factions to apportion the secondary aerosols to each sector are provided; therefore, firstly Table 4 can be used to estimate the local source contributions in the total secondary aerosols, that about 45.2% can be attribute to local sources, as shown in step 2 in Table 5 (45.2% × 16.0% = 7.2%). Then, the contributions are further apportioned from each sector to secondary aerosols using the data in the last column of Table 6, and obtain the results of final contributions of each sector, e.g., the contribution from coal combustion is 8.5% + 7.2% × 79.8% = 14.3%, as the line of “% in total PM2.5” in Table 7. In addition, the contribution percentages of each sector in the total local contributions are calculated, as 22.3% from coal combustion, 10.7% from metal smelting, 7.3% from a mix source of Zn-OC-Ba, 18.5% from other industry, 11.3% from transportation, 10.6% from biomass burning, and 19.2% from dust emissions.

Table 7 also shows the estimations from the CMAQ model calculated according to Table 6 (considering the dust contribution as 19.2%), in the last line. It can be seen that its estimation on coal combustion is 25.3%, which is very close to a combined result of 22.3%, considering that the Zn-OC-Ba is a mixed source of coal combustion, construction activity, etc. The total contribution from coal combustion and industry is approximate 64.1% (25.3 + 38.9) from the CMAQ model, comparable to the combined results of 58.8% (22.3 + 18.5 + 10.7 + 7.3). The biomass contributions are also quite close from the two estimations, but the transportation contribution is estimated higher by the combined method than that from the CMAQ, which may attribute to the tendency of the CMAQ model of underestimating the transportation contributions.

It should be mentioned that the Primary Component Analysis (PCA) model has also been applied in Handan for the four months of 2013 in the study of [41]. It identified four source factors and concluded that in the local source of Handan, the source contributions from coal and biomass combustion, industrial, dust, and transportation are 29.9%, 26.8%, 18.9%, and 8.8%, respectively [41]. This estimation on dust is quite close to the above conclusion (18.9% vs. 19.2%). The contributions from motor vehicles (8.8%) are between the estimations from the combination method (11.3%) and the CMAQ model (5.5%). It did not distinguish the contributions from coal combustion and biomass burning. Therefore, if the total contributions are compared from coal combustion, industrial, and biomass burning, the PCA result is 56.7% (29.9 + 26.8), lower than the 69.4% (22.3 + 18.5 + 10.7 + 7.3 + 10.6) by the combined method, and the 75.1% (25.3 + 38.9 + 10.9) by the CMAQ model. This may be attributed to the large fraction (around 15.6%) of unidentified sources by the PCA model. Therefore, the PCA results are only presented here as a comparison in this study.

After the above comparisons, more confidence could be placed on the contribution estimations of some sectors, e.g., although the CMAQ model cannot calculate the dust contribution, the PMF model and the PCA model provide quite similar estimations (19.2% vs. 18.9%); the contributions of coal combustion and biomass burning from the combined method (22.3% and 10.6%, respectively) are quite close to those from the CMAQ model (25.3% and 10.9%, respectively); the industrial contribution, if the sum of industrial, metal smelting, and Zn-OC-Ba is taken as a rough estimation of the combined method, it is also consistent to the estimation of the CMAQ model (36.5% vs. 38.5%). At the same time, a rough estimation on the uncertainties could be achieved on the other sectors, e.g., the transportation contribution might be overestimated by the combination method (11.3%) compared to the PCA model (8.8%) and the CMAQ model (5.5%).

4 Conclusions

In this study, a combined method using a receptor model, PMF, and a chemical transport model, MM5-CMAQ, is applied on source apportionment of PM2.5 in Handan, Hebei, China. First, PM2.5 samples were collected daily in Handan in the four months in 2013, i.e., January, April, July, and October, and their chemical compositions were analyzed including water soluble ions, OC and EC, and trace elements. Those composition data were then used in the PMF model to identify the PM2.5 sources and their contributions in Handan. At the same time, regional-scale air quality simulations using the MM5-CMAQ modeling system were pursued over North China to quantify the spatial and sectoral contributions to PM2.5 in Handan using the Brute-Force method. Finally, the two sets of results were combined to better answer the questions on PM2.5 sources in Handan to support the policy making of the local government, given the reasons that the receptor model cannot easily distinguish the spatial origins of each source sector under the present status, and the chemical transport model may not be able to give as detailed sectoral apportionment as the receptor model at present, due to the incompleteness and uncertainties in the emission inventory for Hebei.

It is found that the annual average concentration of PM2.5 in Handan in 2013 is 177.8 μg m−3, as SO42−, NO3, NH4+, OC, and EC are major chemical compositions with the percentages of 18.0%, 14.0%, 8.7%, 11.8%, and 4.1%, respectively. The source apportionment of PM2.5 by the PMF model are: 13.8% from coal combustion, 17.5% from industry, 11.0% from metal smelting, 7.6% from a mixed source of Zn-OC-Ba, 11.2% from motor vehicles, 10.7% from biomass burning, 12.3% from dust, and 16.0% of secondary sources. The MM5-CMAQ model indicates that the regional source contributions to PM2.5 in Handan is 40.6%, and within the local contributions, PO, IN, DO, TR, and AG contribute 1.5%, 59.5%, 22.3%, 6.2%, and 10.3%, respectively (note dust emissions are not included that will result in an overestimation of the regional contributions).

The combination of the two results shows that the regional source contributions are 36.0% in total PM2.5 concentrations, and in the local sources; the contributions from each sector are: 22.3% from coal combustion, 10.7% from metal smelting, 7.3% from a mix source of Zn-OC-Ba, 18.5% from other industry, 11.3% from transportation, 10.6% from biomass burning, and 19.2% from dust emissions. The regional joint air pollution controls should be emphasized in future control strategies, and the local source controls on coal combustion and industries, especially the iron and steel industry, are the key points to mitigate the PM2.5 pollution in Handan. Although some limitations exist in this paper, e.g., the dust emissions are not included in MM5-CMAQ modeling, it still provides a method to do PM2.5 source apportionment in the cities like Handan, where it still lacks of both a bottom-up emission inventory with high precision and long-term PM2.5 chemical observation data so that neither receptor model nor chemical transport model can singly produce source apportionment results reliable and detailed enough to support the policy making of the local government.

Notes

Acknowledgements

This study was sponsored by the National Natural Science Foundation of China (No. 41475131), Hebei Science Fund of Distinguished Young Scholars (No. D2017402086), the Program for the Outstanding Young Scholars of Hebei Province, the Hebei Support Program of Hundred Outstanding Innovative Talents from Universities (SLRC2017025), the Hebei Support Program of Hundred Outstanding Innovative Talents from Universities (SLRC2017025), Hebei Cultivating Project of Talent Development (A2016002022), the Innovation Team Leader Talent Cultivation Fund of Hebei University of Engineering.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. 1.
    MEP: 2013 report on the state of the environment in China. MEP, Beijing, China (2014). http://jcs.mep.gov.cn/hjzl/zkgb/2013zkgb/201406/t20140605_276521.htm. Accessed 20 Aug 2015. (in Chinese)
  2. 2.
    MEP: 2014 report on the state of the environment in China. MEP, Beijing, China (2015). http://jcs.mep.gov.cn/hjzl/zkgb/2014zkgb/201506/t20150608_303142.htm. Accessed 20 Aug 2015. (in Chinese)
  3. 3.
    Handan, B.S.: (Handan Bureau of Statistics): Handan Statistical Yearbook 2013. China Statistical Press, Beijing, China (2014)Google Scholar
  4. 4.
    Wang, L.T., Wei, Z., Yang, J., et al.: The 2013 severe haze over southern Hebei, China: model evaluation, source apportionment, and policy implications. Atmos. Chem. Phys. 14, 3151–3173 (2014).  https://doi.org/10.5194/acp-14-3151-2014CrossRefGoogle Scholar
  5. 5.
    Tang, X.Y., Zhang, Y.H., Shao, M.: Atmospheric Chemistry. Higher Education Press, Beijing, China (2006)Google Scholar
  6. 6.
    Zhang, Y.J., Zheng, M., Cai, J., et al.: Comparison and overview of PM2.5 source apportionment methods. China Sci. Bull. 60, 109–121 (2015).  https://doi.org/10.1360/N972014-00975. (in Chinese)CrossRefGoogle Scholar
  7. 7.
    Chueinta, W., Hopke, P.K., Paatero, P.: Investigation of sources of atmospheric aerosol at urban and suburban residential areas in Thailand by positive matrix factorization. Atmos. Environ. 34, 3319–3329 (2000).  https://doi.org/10.1016/S1352-2310(99)00433-1CrossRefGoogle Scholar
  8. 8.
    Lee, E., Chan, C.K., Paatero, P.: Application of positive matrix factorization in source apportionment of particulate pollutants in Hong Kong. Atmos. Environ. 33, 3201–3212 (1999).  https://doi.org/10.1016/S1352-2310(99)00113-2CrossRefGoogle Scholar
  9. 9.
    Yang, L.X., Cheng, S.H., Wang, X.F., et al.: Source identification and health impact of PM2.5 in a heavily polluted urban atmosphere in China. Atmos. Environ. 75, 265–269 (2013).  https://doi.org/10.1016/j.atmosenv.2013.04.058CrossRefGoogle Scholar
  10. 10.
    Belis, C.A., Karagulian, F., Larsen, B.R., et al.: Critical review and meta-analysis of ambient particulate matter source apportionment using receptor models in Europe. Atmos. Environ. 69, 94–108 (2013).  https://doi.org/10.1016/j.atmosenv.2012.11.009CrossRefGoogle Scholar
  11. 11.
    Viana, M., Pandolfi, M., Minguillόn, M.C., et al.: Inter-comparison of receptor models for PM source apportionment: case study in an industrial area. Atmos. Environ. 42, 3820–3832 (2008).  https://doi.org/10.1016/j.atmosenv.2007.12.056CrossRefGoogle Scholar
  12. 12.
    Hopke, P.K., Ito, K., Mar, T., et al.: PM source apportionment and health effects: 1. intercomparison of source apportionment results. J. Expo. Sci. Environ. Epidemiol. 16, 275–286 (2006).  https://doi.org/10.1038/sj.jea.7500458CrossRefGoogle Scholar
  13. 13.
    Contini, D., Belosi, F., Gambaro, A., et al.: Comparison of PM10 concentrations and metal content in three different sites of the Venice Lagoon: an analysis of possible aerosol sources. J. Environ. Sci. 24(11), 1954–1965 (2012).  https://doi.org/10.1016/S1001-0742(11)61027-9CrossRefGoogle Scholar
  14. 14.
    Contini, D., Cesari, D., Donateo, A., et al.: Characterization of PM10 and PM2.5 and their metals content in different typologies of sites in South-Eastern Italy. Atmosphere 5, 435–453 (2014).  https://doi.org/10.3390/atmos5020435CrossRefGoogle Scholar
  15. 15.
    Chen, D.S., Cheng, S.Y., Liu, L., et al.: An integrated MM5-CMAQ modeling approach for assessing trans-boundary PM10 contribution to the host city of 2008 Olympic summer games-Beijing. China Atmos. Environ. 41, 1237–1250 (2007).  https://doi.org/10.1016/j.atmosenv.2006.09.057CrossRefGoogle Scholar
  16. 16.
    Streets, D.G., Fu, J.S., Jang, C.J., et al.: Air quality during the 2008 Beijing Olympic games. Atmos. Environ. 41, 480–492 (2007).  https://doi.org/10.1016/j.atmosenv.2006.08.046CrossRefGoogle Scholar
  17. 17.
    Fu, J.S., Streets, D.G., Jang, C.J., et al.: Modeling regional/urban ozone and particulate matter in Beijing, China. J. Air Waste Manag. Assoc. 59, 37–44 (2009).  https://doi.org/10.3155/1047-3289.59.1.37CrossRefGoogle Scholar
  18. 18.
    Wang, L.T., Hao, J.M., He, K.B., et al.: A modeling study of coarse particulate matter pollution in Beijing: regional source contributions and control implications for the 2008 Summer Olympics. J. Air Waste Manag. Assoc. 58, 1057–1069 (2008).  https://doi.org/10.3155/1047-3289.58.8.1057CrossRefGoogle Scholar
  19. 19.
    Xing, J., Zhang, Y., Wang, S.X., et al.: Modeling study on the air quality impacts from emission reductions and a typical meteorological conditions during the 2008 Beijing Olympics. Atmos. Environ. 45, 1786–1798 (2011).  https://doi.org/10.1016/j.atmosenv.2011.01.025CrossRefGoogle Scholar
  20. 20.
    Zhou, Y., Fu, J.S., Zhuang, G.S., et al.: Risk-based prioritization among air pollution control strategies in the Yangtze River delta. China. Environ. Health Perspec. 118(9), 1204–1210 (2010).  https://doi.org/10.1289/ehp.1001991CrossRefGoogle Scholar
  21. 21.
    Meng, C.C., Wang, L.T., Su, J., et al.: Chemical compositions and source apportionment of PM2.5 in Handan City, Hebei Province. Environ. Sci. Technol. 39(2), 57–64 (2016).  https://doi.org/10.3969/j.issn.1003-6504.2016.02.010. (in Chinese)CrossRefGoogle Scholar
  22. 22.
    Wei, Z., Wang, L.T., Chen, M.Z., et al.: The 2013 severe haze over the Southern Hebei, China: PM2.5 composition and source apportionment. Atmos. Pollut. Res. 5, 759–768 (2014).  https://doi.org/10.5094/APR.2014.085CrossRefGoogle Scholar
  23. 23.
    Bove, M.C., Brotto, P., Cassola, F., et al.: An integrated PM2.5 source apportionment study: positive matrix factorisation vs. the chemical transport model CAMx. Atmos. Environ. 94, 274–286 (2014).  https://doi.org/10.1016/j.atmosenv.2014.05.039CrossRefGoogle Scholar
  24. 24.
    Pekney, N.J., Davidson, C.I., Robinson, A., et al.: Major source categories for PM2.5 in Pittsburgh using PMF and UNMIX. Aerosol Sci. Technol. 40, 910–924 (2006).  https://doi.org/10.1080/02786820500380271CrossRefGoogle Scholar
  25. 25.
    Paatero, P.: Least squares formulation of robust non-negative factor analysis. Chemometr. Intell. Lab. Syst. 37, 23–35 (1997).  https://doi.org/10.1016/S0169-7439(96)00044-5CrossRefGoogle Scholar
  26. 26.
    Paatero, P., Tapper, U.: Positive matrix factorization: a non-negative factor model with optimal utilization of error estimates of data values. Environments 5, 111–126 (1994).  https://doi.org/10.1002/env.3170050203CrossRefGoogle Scholar
  27. 27.
    Saba, T., Su, S.: Tracking polychlorinated biphenyls (PCBs) congener patterns in Newark Bay surface sediment using principal component analysis (PCA) and positive matrix factorization (PMF). J. Hazard. Mater. 260, 634–643 (2013).  https://doi.org/10.1016/j.jhazmat.2013.05.046CrossRefGoogle Scholar
  28. 28.
    Baumann, K., Jayanty, R.K.M., Flanagan, J.B.: Fine particulate matter source apportionment for the chemical speciation trends network site at Birmingham, Alabama, using positive matrix factorization. J. Air Waste Manag. Assoc. 58, 27–44 (2008).  https://doi.org/10.3155/1047-3289.58.1.27CrossRefGoogle Scholar
  29. 29.
    Brown, S.G., Frankel, A., Raffuse, S.M.: Source apportionment of fine particulate matter in phoenix AZ, using positive matrix factorization. J. Air Waste Manag. Assoc. 57, 741–752 (2007).  https://doi.org/10.3155/1047-3289.57.6.741CrossRefGoogle Scholar
  30. 30.
    Huang, S., Rahn, K.A., Arimoto, R.: Testing and optimizing two factor-analysis techniques on aerosol at Narragansett. Rhode Island. Atmos. Environ. 33, 2169–2185 (1999).  https://doi.org/10.1016/S1352-2310(98)00324-0CrossRefGoogle Scholar
  31. 31.
    Pekey, H., Pekey, B., Arslanbas, D.: Source apportionment of personal exposure to fine particulate matter and volatile organic compounds using positive matrix factorization. Water Air Soil Pollut. 224, 1403 (2013).  https://doi.org/10.1007/s11270-012-1403-2CrossRefGoogle Scholar
  32. 32.
    U.S. EPA (U.S. Environmental Protection Agency): EPA positive matrix factorization (PMF) 3.0 fundamentals & user Guide. U.S. EPA, Research Triangle Park, NC, USA (2008)Google Scholar
  33. 33.
    Zhang, J.G., Wang, Y.S., Wang, S., et al.: Source apportionment of non-methane hydrocarbon in Beijing using positive matrix factorization. Environ. Sci. Technol. 32(5), 35–39 (2009). (in Chinese)Google Scholar
  34. 34.
    Viana, M., Chi, X., Maenhaut, W., et al.: Influence of sampling artefacts on measured PM, OC, and EC Levels in carbonaceous aerosols in an urban area. Aerosol Sci. Technol. 40(2), 107–117 (2007).  https://doi.org/10.1080/02786820500484388CrossRefGoogle Scholar
  35. 35.
    Yu, L.D., Wang, G.F., Zhang, R.J. et al.: Characterization and source apportionment of PM2.5 in an urban environment in Beijing. Aerosol Air Qual. Res. 13, 574–583 (2013).  https://doi.org/10.4209/aaqr.2012.07.0192
  36. 36.
    Brown, S.G., Eberly, S., Paatero, P., et al.: Methods for estimating uncertainty in PMF solutions: examples with ambient air and water quality data and guidance on reporting PMF results. Sci. Total Environ. 518–519, 626–635 (2015).  https://doi.org/10.1016/j.scitotenv.2015.01.022CrossRefGoogle Scholar
  37. 37.
    U.S. EPA: EPA positive matrix factorization (PMF) 5.0 fundamentals and user guide (2014)Google Scholar
  38. 38.
    NCAR: PSU/NCAR mesoscale modeling system tutorial class notes and user’s guide: MM5 modeling system version 3. National Center for Atmospheric Research, USA (2003)Google Scholar
  39. 39.
    Dunker, A.M., Morris, R.E., Pollack, A.K., et al.: Photochemical modeling of the impact of fuels and vehicles on urban ozone using auto oil program data. Environ. Sci. Technol. 30, 787–801 (1996).  https://doi.org/10.1021/es950175mCrossRefGoogle Scholar
  40. 40.
    Wang, L.T., Wei, Z., Wei, W., et al.: Source apportionment of PM2.5 in top polluted cities in Hebei, China using the CMAQ model. Atmos. Environ. 122, 723–736 (2015).  https://doi.org/10.1016/j.atmosenv.2015.10.041CrossRefGoogle Scholar
  41. 41.
    HEBEU: Report on PM2.5 source apportionment in Handan. HEBEU, Handan, Hebei, China (2015)Google Scholar
  42. 42.
    Kain, J.S., Fritsch, J.M.: Convective parameterization for mesoscale models: the Kain-Fritsch scheme. In: Emanuel, K.A., Raymond, D.J. (eds.) The Representation of Cumulus Convection in Numerical Models, p. 246. American Meteorological Society (1993).  https://doi.org/10.1007/978-1-935704-13-3_16CrossRefGoogle Scholar
  43. 43.
    Zhang, D.L., Anthes, R.A.: A high-resolution model of the planetary boundary layer-sensitive tests and comparisons with SESAME-79 data. J. Appl. Meteorology 21, 1594–1609 (1982).  https://doi.org/10.1175/1520-0450(1982)021<1594:ahrmot>2.0.co;2CrossRefGoogle Scholar
  44. 44.
    Reisner, J., Rasmussen, R.J., Bruintjes, R.T.: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Q. J. Royal Meteorological Soc. 124B, 1071–1107 (1998).  https://doi.org/10.1002/qj.49712454804CrossRefGoogle Scholar
  45. 45.
    Dudhia, J.A.: A non-hydrostatic version of the Penn State/NCAR mesoscale model: validation tests and simulation of an Atlantic cyclone and cold front. Monthly Weather Rev. 121, 1493–1513 (1993).  https://doi.org/10.1175/1520-0493(1993)121<1493:anvotp>2.0.co;2CrossRefGoogle Scholar
  46. 46.
    Blackadar, A.K.: Modeling the Nocturnal Boundary Layer. In: Proceedings of the Third Symposium on Atmospheric Turbulence, Diffusion and Air Quality, American Meteorological Society, Boston, MA, pp. 46–49 (1976)Google Scholar
  47. 47.
    Deardorff, J.W.: Efficient prediction of ground surface temperature and moisture, with inclusion of a layer of vegetation. J. Geophys. Res. 83, 1889–1903 (1978).  https://doi.org/10.1029/JC083iC04p01889CrossRefGoogle Scholar
  48. 48.
    Carter, W.P.L.: A detailed mechanism for the gas-phase atmospheric reactions of organic compounds. Atmos. Environ. 24, 481–518 (1990).  https://doi.org/10.1016/0960-1686(90)90005-8CrossRefGoogle Scholar
  49. 49.
    Carter, W.P.L.: Implementation of the SAPRC-99 chemical mechanism into the models-3 framework. Report to the U.S. EPA, prepared by Cater, W.P.L., Statewide Air Pollution Research Center, University of California, Riverside, CA (2000)Google Scholar
  50. 50.
    Binkowski, F.S., Shankar, U.: The regional particulate model. 1. Model description and preliminary results. J. Geophys. Res. 100, 26191–26209 (1995).  https://doi.org/10.1029/95JD02093CrossRefGoogle Scholar
  51. 51.
    Chang, J.S., Brost, R.A., Isaksen, I.S.A., et. al.: A three-dimensional acid deposition model: physical concepts and formulation. J. Geophys. Res. 92, 14681–14700 (1987).  https://doi.org/10.1029/JD092iD12p14681CrossRefGoogle Scholar
  52. 52.
    Walcek, C.J., Taylor, G.R.: A theoretical method for computing vertical distributions of acidity and sulfate production within cumulus clouds. J. Atmos. Sci. 43, 339–355 (1986).  https://doi.org/10.1175/1520-0469(1986)043<0339:atmfcv>2.0.co;2CrossRefGoogle Scholar
  53. 53.
    MEP: Guidance on the report of source apportionment of atmospheric aerosols. MEP, Beijing, China (2015)Google Scholar
  54. 54.
    Cao, G.L., Zhang, X.Y., Gong, S.L., et al.: Emission inventories of primary particles and pollutant gases for China. Chin. Sci. Bull. 56 (2011).  https://doi.org/10.1007/s11434-011-4373-7. (in Chinese)CrossRefGoogle Scholar
  55. 55.
    Guan, D.B., Liu, Z.: The truth of haze: PM2.5 pollution source analysis and reduction strategy research. China Environmental Press, Beijing, China (2014)Google Scholar

Copyright information

© Springer International Publishing AG, part of Springer Nature 2019

Authors and Affiliations

  • Zhe Wei
    • 1
  • Litao Wang
    • 1
  • Liquan Hou
    • 1
  • Hongmei Zhang
    • 2
  • Liang Yue
    • 3
  • Wei Wei
    • 4
  • Simeng Ma
    • 1
  • Chengyu Zhang
    • 1
  • Xiao Ma
    • 1
  1. 1.Department of Environmental Engineering, School of City ConstructionHebei University of EngineeringHandanChina
  2. 2.School of Economics and ManagementHebei University of EngineeringHandanChina
  3. 3.Environmental Monitoring Center of HandanHandan Environmental Protection BureauHandanChina
  4. 4.Department of Environmental ScienceBeijing University of TechnologyBeijingChina

Personalised recommendations