Exploring the cause of PM2.5 pollution episodes in a cold metropolis in China

Journal Pre-proof Exploring the cause of PM2.5 pollution episodes in a cold metropolis in China Xiazhong Sun, Kun Wang, Bo Li, Zheng Zong, Xiaofei Shi, Lixin Ma, Donglei Fu, Samit Thapa, Hong Qi, Chongguo Tian PII:

S0959-6526(20)30322-X

DOI:

https://doi.org/10.1016/j.jclepro.2020.120275

Reference:

JCLP 120275

To appear in:

Journal of Cleaner Production

Received Date: 23 August 2019 Revised Date:

16 January 2020

Accepted Date: 25 January 2020

Please cite this article as: Sun X, Wang K, Li B, Zong Z, Shi X, Ma L, Fu D, Thapa S, Qi H, Tian C, Exploring the cause of PM2.5 pollution episodes in a cold metropolis in China, Journal of Cleaner Production (2020), doi: https://doi.org/10.1016/j.jclepro.2020.120275. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier Ltd.

Author contributions. Xiazhong Sun: Data curation, Formal analysis, Writing- Original draft preparation, Writing- Reviewing and Editing. Kun Wang: Resources, Methodology, Conceptualization. Bo Li: Resources. Zheng Zong and Xiaofei Shi: Methodology and Formal analysis. Lixin Ma, Donglei Fu and Samit Thapa: Investigation. Hong Qi: Resources, Methodology, Conceptualization, Supervision, Project administration, Funding acquisition, Writing- Reviewing and Editing. Chongguo Tian: Resources, Methodology, Conceptualization, Writing- Reviewing and Editing.

1

Exploring the cause of PM2.5 pollution episodes in a cold metropolis in China

2 3

Xiazhong Suna,b, Kun Wanga,b, Bo Lia,b, Zheng Zongc, Xiaofei Shia,b,d, Lixin Maa,b, Donglei Fua,b, Samit Thapaa,b, Hong Qia,b,*, Chongguo Tiand,*

4 5

a

6

b

7 8

c

9

d

Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology, Harbin 150090,

China School of Environment, Harbin Institute of Technology, Harbin 150090, China

Key Laboratory of Coastal Environmental Processes and Ecological Remediation, Yantai Institute of Coastal

Zone Research, Chinese Academy of Sciences, Yantai 264003, China CASIC Intelligence Industry Development Co., Ltd., Beijing, 100854, China

10 11

Highlights

12

Multifaceted factors were analyzed to explore the cause of PM2.5 pollution episodes.

13

Secondary organic carbon and nitrate were two most abundant chemical components.

14

Biomass burning and secondary aerosol contributes most to fine particle pollution.

15

Regional transportation and adverse weather promote pollution formation.

16 17

Abstract. Harbin, a cold metropolis of China, frequently endures fine particles (PM2.5) pollution

18

during its long heating seasons, which has almost become the main environmental issue. To better

19

understand pollution causes, multifaceted factors including the meteorology conditions, secondary

20

transformation, source apportionment and regional transportation, were analyzed based on the

21

definition of three pollution levels and eight pollution episodes. The results showed that chemical

22

species varied dynamically and followed similar trends to PM2.5, and secondary organic carbon and ∗

Corresponding author.

E-mail addresses: [email protected] (H. Qi), [email protected] (C. Tian)

23

nitrate played more significant driving roles in PM2.5 as the pollution level increased. Five sources

24

were apportioned using the positive matrix factorization model, and the order of importance for the

25

entire study period was as follows: secondary aerosol, traffic emissions, biomass burning, mineral dust,

26

and coal combustion. The PM2.5 in Harbin is mainly affected by transportation in the

27

Harbin-Changchun megalopolis and slightly affected by long-distance transmission from the

28

northwestern region of Harbin. A pollution episode in the fall was quite significant, as it was

29

characterized by long duration time and extremely high pollutant levels, and was most likely attributed

30

to extensive emissions from biomass burning and enhanced secondary transformation. The pollution

31

episodes in winter were characterized by enhanced heterogeneous reactions and weakened

32

photochemical reactions, which were especially sensitive to synoptic variations, and most likely

33

induced by high relative humidity (> 60%), low wind speed (3 m/s), suppressed planetary boundary

34

layer height (< 200 m), and strengthened inversion layer under the control of high-pressure systems.

35

Meanwhile, the episodes in spring were characterized by high proportions of mineral dust, chlorine,

36

and potassium, which was mostly attributed to the prevailing dust events, stronger biomass burning

37

activities, and favorable transportation under low-pressure systems.

38

Keywords: PM2.5 pollution; synoptic weather; secondary transformation; source apportionment;

39

regional transportation.

40

1. Introduction

41

As a cold metropolis in China, Harbin is strongly influenced by the Siberian high pressure and

42

Arctic Vortex. However, it still frequently endures fine particulate matter (PM2.5)-dominated haze

43

pollution during the long heating season. PM2.5 pollution here is not fully controlled as it has received

44

limited attention. The available official data (Fig. S1) indicate that PM2.5 is the primary air pollutant in

45

Harbin. Although the mean annual concentration declined from 2014 to 2018 (2014: 71 µg/m3, 2015:

46

69 µg/m3, 2016: 51µg/m3, 2017: 57µg/m3, 2018: 40 µg/m3), severe pollution still frequently occurs

47

during each heating season, and even triggers the red haze alert. PM2.5 has almost become the main

48

environmental issue during the heating seasons with 45.03% pollution days, while the percentage of

49

pollution days in the on-heating seasons is 6.56%. Furthermore, as a cold metropolis, Harbin has

50

longer heating seasons (last for six months from 20-October to 20-April each year) than most northern

51

cities in China. During the heating seasons, the complex emission systems coupled with adverse

52

meteorological conditions caused severe haze events that threaten public health in the region and may

53

influence climate change across a larger area, such as East Asia and the Arctic, due to its unique

54

geographical location. However, the attention received by haze pollution does not correspond to its

55

severity and complexity. Therefore, haze pollution is being gradually alleviated, but is far from being

56

controlled. Many studies concerning air pollution have been conducted in the tropical and subtropical

57

cities of China, but paid limited attention has been paid to air pollution in the cold metropolises. Air

58

pollution is a global problem that requires local resolutions (Li et al., 2019). Therefore, a deeper

59

exploration of the causes of high-density PM2.5 pollution is urgently required here.

60

The causes of high PM2.5 pollution could be quite complex, as abnormally high PM2.5 can be

61

affected by various factors, including adverse synoptic systems, complex emission systems, secondary

62

transformation, and regional transmission. Haze pollution is the result of the over-accumulation of

63

pollutants, and this process is difficult to reverse once the total amount of pollutants meets the

64

atmospheric capacity threshold, or the atmospheric capacity decreases due to changes in the

65

meteorological conditions (Cai et al., 2017). In the above process, excessive emissions and secondary

66

transformation are the most important causes of haze pollution and have received much attention from

67

domestic and international studies. Multiple primary emissions have been identified (Gao et al., 2018),

68

and various unusual phenomena were also discovered during polluted episodes, such as heterogeneous

69

transformation and hygroscopic growth under high relative humidity, the formation of a two-way

70

feedback mechanism between PM2.5 and the atmospheric boundary layer (Wang et al., 2014; Wang et

71

al., 2014), and the synergistic and competitive conversion of SO2 and NOx (He et al., 2015). Despite

72

this, the unclear phenomena such as the haze evolution process (Seo et al., 2017), heterogeneous

73

catalytic reaction mechanism (Kumar et al., 2018), secondary reaction pathways (Lim et al., 2018), and

74

some other undiscovered phenomena should be further explored. Haze formation is also induced by

75

regional transport and meteorological conditions including synoptic systems, the atmospheric

76

boundary layer, and other meteorological parameters (Ning et al., 2018). It has been reported the

77

meteorological conditions play more important roles in haze formation when the total pollutant

78

emissions are stable. Typical synoptic situations and their roles in severe pollution have been classified

79

and evaluated in numerous studies (Bei et al., 2016). Anticyclonic circulation (Lesniok et al., 2010),

80

synoptic-scale high-pressure ridges (Whiteman et al., 2014), dry low-pressure systems of 700hPa

81

(Ning et al., 2018), shallowed East Asian trough and weakened Siberian high-pressure (Zhang et al.,

82

2016) can all contribute to temperature inversion, and weak horizontal and vertical convection. The

83

above findings suggest that a holistic understanding of the causes of PM2.5 in Harbin is required.

84

In this study, PM2.5 samples were collected during the heating season of 2017-2018. We aimed to

85

analyze several factors, including physical causes, secondary reactions, source apportionment, and

86

regional transmission. (1) The sampling, chemical analysis and methods are described in Sect. 2. (2)

87

The main results and discussions, including three subsections: the general characteristics of PM2.5 and

88

its chemical species, the influence of synoptic weather, secondary transformation, source

89

apportionment, and regional transportation are presented in Sect. 3. (3) Finally, the main conclusions

90

are summarized in Sect. 4.

91

2. Materials and methods

92

2.1 Site description and Sampling

93

Harbin (125.4-130.1˚ E, 44.0-46.4˚ N) is a well-developed industrial and agricultural mega-city

94

in the Northeast China Plain, with the highest inhabited latitude, lowest temperature (-37.7 ℃), largest

95

land area (10,200 km2), and third largest registered population (5.5 million) among all provincial

96

capitals in China (Chen et al., 2018). The climate of Harbin is mainly continental monsoon, with the

97

lowest temperatures occurring in January. The northern part of Harbin is the XiaoXing'an Mountains,

98

while the southeastern part is the branch hill of Zhangguangcai Ridge, The urban area of Harbin is

99

mainly distributed in the three-ladder formed by the Songhua River, which runs through the central

100

part of the city. The topographic features of this region can be classified as a typical valley landform.

101

The region surrounding Harbin located has fertile land that has been intensively cultivated over the

102

last 50 years. With the increases in the areas of rice paddies, the Northeast China Plain has become one

103

of the most important agricultural regions in China (Wang et al., 2011a; Zhao et al., 2015). However,

104

agricultural waste is commonly burned in the area to reduce straw accumulation, correspondingly

105

causing large amounts of air pollutants (Cao et al., 2016; Qin et al., 2014).

106 107

Fig.1 Study area and sampling site.

108

In this study, PM2.5 samples were collected from the roof of the School of Environment Building

109

(45°45′14″N, 126°40′54″E, approximately 15 m above the ground) in the second campus of the Harbin

110

Institute of Technology, where is located in a mixed zone of commercial and residential buildings, and

111

traffic, without any major fixed pollution sources. Therefore, this area represents the typical urban

112

environment of Harbin. The samples were collected using high-volume air samplers, equipped with

113

PM2.5 fractionating inlets (Laoying., Qingdao, China) every five days typically, or continuously during

114

haze episodes. Ambient air was drawn into the sampler an average flow rate of 1.05 m3/min for 24 h

115

(8:00 AM to 7:59 AM the next day), and particulate matter particles with aerodynamic diameters of ≤

116

2.5 µm (PM2.5) were collected on 8 × 10-in Tissuquartz filters (PALL, NY, U.S.). All filters were

117

sterilized in a Mu℃e furnace at 480 °C for 5 h and then packed in aluminum foil and stored in a sealed

118

bag until they were loaded into the filter cartridge. The PM2.5 concentrations were estimated from the

119

flow-through volume (approximately 1531 m3 in 24h) and the filter net weight recorded at 0.1 mg

120

accuracy. After sampling, the filters were stored at -20 °C until post-chemical analysis were conducted.

121

We also collected gas pollutants data (SO2, NO2) from the local air quality monitoring station for

122

further analysis.

123

2.2 Chemical analysis

124

Two discs with 47 mm diameter were punched from each quartz fiber filter. One was extracted

125

with 20 ml ultrapure water for soluble ions measurement (K+, Ca2+, Na+, Mg2+, NH4+, Cl-, NO3-, F- and

126

SO42-) by ion chromatograph (Dionex, Thermo scientific, Inc.U.S). Another was digested with 15 ml

127

nitric acid (guaranteed reagent) in electric heating plate digestion system for elemental measurement,

128

including V, Cr, Mn, Ni, Cu, As, Cd, Pb, Ba, Ga, Sr, Ti, Co, Se, Rb, Ag, Sn, Sb, Cs, Tl, Fe, Zn, using

129

inductively coupled plasma mass spectrometry (ICP-MS, Thermo Fisher, Inc.US). Two reagents blank

130

and one filter blanks were conducted in each run following the same procedures as samples. The

131

detection limit of ions was 10-30 ng/ml with the uncertainties < 5%, and the detection limits of the

132

elements ranged from 0.01 to 1 ng/ml with error < 5%.

133

A 0.544 cm2 punch from each filter was used to measure organic carbon (OC) and elemental

134

carbon (EC) by a Desert Research Institute (DRI) Model 2001 Thermal/Optical Carbon Analyzer

135

(Atmoslytic Inc., Calabasas CA) following the Interagency Monitoring of Protected Visual

136

Environment thermal /optical reflectance protocol. The measuring range of TOR was from 0.05 to 750

137

µg C/cm2 with the error less than 10%.

138

2.3 Chemical mass closure

139

PM2.5 composed of 10 most abundant categories in this study: organic matter (OM), EC, SO42-,

140

NO3-, NH4+, Cl-, K+, other ions, trace element oxides (TEO), and mineral dust. The factor converting

141

OC to OM has been considered to ranging from 1.4 to 2.2, this study adopted 1.8 as literature

142

suggested in autumn and winter (Tao et al., 2014). The minimum ratios of OC/EC were taken as

143

representative of the primary emission to calculate the secondary organic carbon (SOC) (Li et al., 2016;

144

Zong et al., 2018). OM = 1 . 8 × OC

145

(1)

SOC = OC - POC = OC - EC ×（OC / EC）prim

146

(2)

147

Other ions including soluble ions measured in section 2.2 except SO42-, NO3-, NH4+, Cl- and K+.

148

Mineral dust was calculated by Fe (Zong et al., 2018), TEO was calculated by an equation introduced

149

form other studies based on limited inorganic elements measured in section 2.2 (Tao et al., 2014),

150

calculation formulas were shown as follows: Mineral dust = Fe / 0 . 04

151

(3)

152

TEO=1.3×[0.5×（Sr+ Ba+Mn+Co+ Rb+ Ni+V）+(Cu+Zn+Cd+Sn+Sb+Tl+ Pb+ As+Cs+Ga+ Ag)] (4)

153

2.3 Source analysis

154

Qualitatively and quantitatively sources apportionment were conducted by enrichment factors

155

(EFs) analysis and positive matrix factorization (PMF) analysis in this study. EFs analysis was used to

156

qualitatively assessment elements enrichment by human activities, it could be defined as (Zoller et al.,

157

1974):

158

EFs = (C / C0 ) Aerosol /(C / C0 )Crustal

(5)

159

Where C refers to target elements, and Cₒ is the reference elements. Detailed classification of

160

enrichment levels and degrees are shown in Table S1. (C/Co)Crust in this study referenced from the

161

background value in the soil in China. Generally, Fe, Al and Si can be utilized as reference elements

162

because they were considered less affected by other metal elements or anthropogenic sources, and

163

have relative stable chemical properties and content ratios in the earth's crust (Owens et al., 2005;

164

Kersten and Smedes, 2002; Zhang et al., 2009). In our study, due to the interference from the aluminum

165

packaging and the quartz fiber substrate of the samples, Fe was selected as reference element.

166

PMF 5.0 provided by the US EPA is a multivariate factor analysis tool, which relying on

167

physically significant assumptions, that concentrations at a receptor site are supported by the linear

168

combinations of different source emissions (Gao et al., 2018; Wang et al., 2017; Lee et al., 1999). It can

169

solve the problem of negative factor loadings/scores by integrating non-negativity constrained factor

170

analysis, making factor loadings and scores more interpretable (Paatero et al., 2013) (Text S1). In this

171

study PM2.5 mass concentrations, OC, EC, ions (K+, Ca2+, Na+, Mg2+, NH4+, Cl-, NO3-, F- and SO42-)

172

and elements (V, Cr, Mn, Ni, Cu, As, Cd, Pb, Ba, Ga, Sr, Ti, Co, Se, Rb, Ag, Sn, Sb, Cs, Tl, Fe and Zn)

173

selected as source profiles were put into PMF to quantitatively assess the major source contributions.

174

2.4 HYSPLIT and PSCF modeling

175

Hybrid single-particle Lagrangian integrated trajectory (HYSPLIT version 4.0) model which is

176

available on the National Oceanic and Atmospheric Administration Air Resource Laboratory website

177

(www. arl. noaa. gov/ ready/ hysplit4. html), was used to generate backward trajectories during

178

sampling period. In this study National Center for Environmental Prediction (NCEP) GDAS data

179

(1°×1°) was used to simulated 72-h backward air mass in every sampling day, started at 00:00, 06:00,

180

12:00 and 18:00 of local time. The arrival heights of the trajectories in this study was set as 500 m, this

181

height was considered can effectively reduce the influence of ground surface friction and accurately

182

reflect the characteristic of mean flow field below atmospheric boundary layer (Wang et al., 2009).

183

Potential source contribution function (PSCF) introduced by Ashbaugh et al (Ashbaugh et al.,

184

1985) was used in this study to assess and potential source regions of PM2.5 and main source factors

185

given by PMF result. PSCF is a conditional probability function, and was used to characterize the

186

spatial distribution of possible geophysical source locations of trajectories calculated by HYSPLIT.

187

The probability that any given cell (ijth cell) of PSCF field can be defined by the given equation:

188

PSCFij = mij / n ij

(6)

189

nij represents the total number of endpoints that fall in the ijth cell, and mij represents the number

190

of endpoints-oriented PM2.5 concentration or contributions exceeding given threshold. Here we set the

191

threshold of PM2.5 concentration and contributions as 75 ug/m3 and 0.75 respectively. The grid layer

192

covered by the trajectories was divided into an array of 0.5º×0.5º cells size. In order to remove the

193

uncertainties caused by endpoints (Bressi et al., 2014; Li et al., 2017; Zong et al., 2018), an arbitrary

194

weight function W(nij) shown as following was multiplied with the PSCF value:

 1.00,80 < nij   0.70,20 < nij ≤ 80  w ij =   0.42,10 < nij ≤ 20  0.05, ≤ 10   

195

196

(7)

2.5 WRF simulation

197

Advanced Weather Research and Forecasting model (WRF) was designed as a numerical weather

198

prediction mode with options for various physical processes, but can also be applied as a regional

199

climate model (Chen et al., 2011). This study introduces WRF 3.9, and calculated wind speed (WS),

200

wind direction (WD), temperature (T), relative humidity (RH), and the height of the planetary

201

boundary layer (PBL), also simulated surface (wind field, temperature field, and constant pressure line

202

in horizon profiles) and vertical weather maps (constant temperature line, water vapor concentration

203

and PBL height in vertical profiles) across sampling site. Synoptic data and meteorological field

204

calculated by the National Center for Environmental Prediction (NCEP) data (1º×1º) from the website

205

(https://rda.ucar.edu/). These data are available for 00:00, 06:00, 12:00, and 18:00 UTC, and

206

two-dimensional data assimilation technology was used.

207

3. Result and discussion

208

3.1 Pollution characteristics of PM2.5 and its chemical species

209

The temporal variations in the PM2.5 mass concentration during the heating season are shown in

210

Fig. 2. In this study, we defined haze episodes as days when the PM2.5 mass concentrations exceeded

211

the National Ambient Air Quality (NAAQ) Grade II standard of 75 µg/m3 based on previous studies

212

(Wang et al., 2014; Zheng et al., 2016), and 85% of the samples taken in the heating season were

213

collected on hazy days. We also defined three pollution levels based on the NAAQ standard

214

(http://106.37.208.228:8082/): clean (C, PM2.5 ≤ 75 µg/m3), slightly polluted (SP, 75< PM2.5 ≤150

215

µg/m3), and heavily polluted (HP, PM2.5 > 150 µg/m3). As highly concentrated and long-lasted

216

episodes are more harmful and receive more public attention, this study specially focused on heavily

217

polluted episodes for detailed analysis, and further categorized one special pollution episode (SPE)

218

and seven short-duration polluted episodes (PE1-PE7; Fig. 2). These episodes were characterized by a

219

long duration or abnormally high PM2.5 mass concentrations (exceeding 150 µg/m3), and were selected

220

to explore the causes of haze during the heating season in Harbin. Consecutive heavily polluted

221

samples were merged into one episode. The concentration of PM2.5 in the study period ranged from

222

33.17 to 765.61 µg/m3 and fluctuated significantly. SPE occurred in late October and early November,

223

beginning on October 25. Three peak values occurred on October 27 (411.43 µg/m3), November 1

224

(554.93 µg/m3), and November 6 (765.61 µg/m3), and the event then dissipated on November 7. PEs

225

(EP1-EP7) occurred alternatively with clean days from December to April and were characterized by

226

relatively lower PM2.5 mass concentrations and a shorter duration time than SPE, only two episodes

227

with PM2.5 concentrations exceeding than 250 µg/m3 occurred, which were on January 1 and April 5.

228

The time series of chemical species exhibited a similar temporal trend to PM2.5, with elevated

229

concentrations during polluted episodes and the highest concentration occurring during the SPE (Fig.

230

2). Carbonaceous components were the most abundant species and strongly correlated with PM2.5

231

(rOC=0.878, rEC=0.788, Fig. S4), with average concentrations of 29.18 and 7.89 µg/m3 throughout the

232

observation period, indicating that they predominantly contributed to PM2.5 pollution. Among the

233

water-soluble ions, NO3- was the most abundant, with an average concentration of 8.45 µg/m3,

234

followed by SO42-, Cl-, NH4+, and K+, with average values of 6.19, 5.93, 5.66 and 3.11 µg/m3

235

respectively. The concentrations of other ions (Ca2+, Na+, Mg2+, F-) were low. The fraction of

236

secondary inorganic (SNA, sulfate, nitrate and ammonium) (blow 19.74%) in this study was not

237

predominant, as was observed in previous studies (25.4%-46.49%) (Zong et al., 2018; Li et al., 2017),

238

but was still the second-most predominant component. In this study, NO3- accounted for the largest

239

part of the SNA (Fig. 2-3), and its content sharply increased during SPE and PE1-PE3. The SO42-

240

concentration was relatively stable, and generally exceeded that of NO3- on clean days. The NO3-/SO42-

241

mass ratio has been widely used as an indicator of the relative importance of mobile and stationary

242

sources of nitrogen and sulfur in the atmosphere. Higher (8-13) NO3-/SO42- ratios are typically

243

associated with emissions from gasoline, diesel fuel, or biomass burning (Wang et al., 2006; Cao et al.,

244

2016), while lower ratios (0.27-1.0) are typically associated with coal-burning (Wang et al., 2015;

245

Wang et al., 2015). In this study, the NO3-/SO42- ratio was generally below than 1.0 on clean days, and

246

gradually increased as the pollution level increased, indicating that the drivers of PM2.5 formation

247

shifted from coal combustion on clean days to traffic emissions or biomass burning on polluted days,

248

which was also observed in studies conducted in Shanghai and Chengdu (Li et al., 2017; Wang et al.,

249

2006). The more prominent role of NO3- in haze formation may due to the continuously increasing car

250

ownership and promotion of boiler desulfurization. The properties of Cl- and K+ observed in this study

251

were higher than those observed in previous days (Tao et al., 2014; Zhang et al., 2013; Gao et al., 2018;

252

Zong et al., 2018), and the PM2.5 concentration was more closely correlated with Cl- (r=0.846 ) and K+

253

(r=0.833) than NO3- (r=0.787) and SO42- (r=0.763) (Fig. S4). The contents of potassium and chlorine

254

in aerosols from biomass burning are enriched, and fresh biomass-burning plumes are generally

255

characterized by high potassium chloride contents, while aged biomass burning smoke contains high

256

amounts of potassium sulfate/nitrate (Cao et al., 2016). The stronger relationship (r=0.935, Fig. S4)

257

between potassium and chloride in our study than that of sulfate/nitrate clearly demonstrated that the

258

biomass aerosol likely originated from local emissions. Furthermore, the higher contents of chlorine

259

and potassium observed in SPE, PE4, PE5, and PE6 also corresponded to densely distributed fire

260

points in the Harbin-Changchun megalopolis (Fig. S3).

261

As indicated by the chemical composition of PM2.5 presented in Fig. 3, the content of OM was

262

remarkably high, with a maximum proportion of 35.93% in all categories, followed by mineral dust,

263

NO3-, EC, SO42-, Cl-, NH4+, K+, other ions, and TEO. As the pollution level increased, the fractions of

264

SOC and NO3- increased sharply, those of NH4+, Cl- and K+ fraction increased slightly, while the

265

proportions of primary organic carbon (POC), EC, SO42-, mineral dust, and TEO decreased. The

266

substances with increased proportions may play more important roles in pollution formation (Zheng et

267

al., 2016), and strong positive correlations were observed between them (Fig. S4), indicating that they

268

likely originated from common sources. In SPE, EP1, EP2 and EP3, carbonaceous components and

269

SNA were observed accounting at high proportions (50.82% - 62.12%), these episodes were closely

270

related to the secondary reactions and can be defined as pure-haze. However, EP4-EP7 were driven by

271

mineral dust and can be defined as mixed haze-dust. During SPE, EP5, and EP6, the concentration of

272

K+ and Cl- were extremely enhanced and may have been significantly influenced by biomass burning.

273 274

Fig. 2 Temporal evolution of the PM2.5 concentration, its main components (OC, SOC, EC, POC, NH4+,

275

SO42-, NO3-, K-, Cl-) and component ratios. (The light-yellow bars represent heavily polluted weather

276

conditions)

277 278

Fig. 3 Chemical mass composition of the gravimetric PM2.5, with respect to the pollution level and

279

episodes.

280

3.2 Influence of synoptic weather system

281

The dynamic variations in the PM2.5 mass concentration during the heating season could be partly

282

induced by the unfavorable diffusion conditions, which are reflected by the meteorological parameters,

283

such as PBL, WS, WD, T, and RH (Ning et al., 2018). The PBL was mostly below 500 m from

284

November to March, and below 200 m during heavily polluted episodes (SPE, PE1-PE4), but higher in

285

April. As shown in Fig.4, the influence of PBL on PM2.5 followed two different influence mechanisms.

286

Contrasting trends between the PM2.5 concentration and PBL were observed when the PBL height was

287

below 500 m, and synchronized tends were observed when the PBL height was above 500 m. Westerly

288

winds were predominant throughout the study period; WS fluctuated around an average value of 3.47

289

m/s and was positively correlated with the PBL height, with a correlation coefficient of 0.481 (Fig. S4),

290

The WS in EP1-EP3 were below 3 m/s, but exceeded 3 m/s during the other episodes. The temperature

291

decreased from October to January, and then increased, following an opposite trend to the RH. The

292

Siberian high pressure is the main factor affecting the temperature in Northeast China during winter

293

(Zhang et al., 2016). In this study, temperature was strongly correlated with PBL (r = 0.617) but

294

negatively correlated with RH (r = -0.44). Therefore, the Siberian high pressure was likely the main

295

cause of the low PBL, low temperature and high RH from November to March. The abnormally and

296

frequently elevated PM2.5 mass concentrations from November to March were largely associated with

297

adverse meteorological conditions, such as significantly decreased PBL (< 500 m), WS (< 3 m/s), and

298

temperature, and increased RH (> 60%), which were mainly governed by the Siberian high pressure.

299

Before November 1 and after March 30, the diffusion conditions were favorable, with a high PBL

300

(>500 m) and WS (> 3 m/s), low RH (< 60%), and elevated PM2.5 mass concentrations (PE5-PE7) that

301

may have been caused by the higher primary emissions, regional transportation, and dust storms.

302 303

Fig. 4 Temporal variations in the PM2.5 mass concentration and meteorological parameters.

304

(temperature, relative humidity, wind speed, wind direction and PBL; SPE is indicated by the red bar

305

and PEs are indicated by the yellow bars).

306

To explore the physical vertical mixing and horizontal dispersion mechanisms involved in the

307

formation of SPE and PEs, surface weather maps (Fig. 4) and west-east vertical profiles (Fig. 5) were

308

drawn for each type of episodes. The surface weather maps for polluted episodes were generally

309

categorized into two types: high-pressure system-governed and low-pressure system-governed

310

episodes. During the SPE, a low-pressure system initially formed, accompanied by the first peak value

311

on October 27. In the subsequent days, a high-pressure system developed along with extremely high

312

PM2.5 concentrations on November 1 and 6. The SPE eventually dissipated on November 7, when the

313

region was controlled by a low-pressure system again. In the EPs, the sampling site was successively

314

governed by low-pressure centers (Fig. 5: EP1(1), EP3, EP4), high-pressure ridges (Fig. 5: EP1(2),

315

EP2), and low-pressure troughs (Fig. 5: EP5-EP7). The west–east vertical profiles of temperature and

316

the RH crossing the sampling site were analyzed to investigate the vertical exchange capacity, and

317

adverse vertical weather systems with low PBL, temperature inversion, and high RH were observed

318

during SPE and EP1-EP4 which were mainly governed by a high-pressure system (Fig. 6).

319

Low-pressure system-governed episodes (EP5-EP6) conversely corresponded to high PBL, steep

320

temperature stratification, and low RH in the vertical profiles (Fig. 6).

321 322

Fig. 5 Surface weather maps during the heavily polluted episodes. (sampling site is indicated by the

323

red dots).

324 325

Fig. 6 West-east vertical profiles of temperature and RH crossing the sampling site (45°45′14″N,

326

126°40′54″E; Black shading represents the terrain, and the red dots represent the sampling site).

327

In summary, the synoptic weather system from November to March was generally controlled by

328

the Siberian high-pressure, which decreased the PBL height and surface temperature, increased the RH,

329

and suppressed the diffusion system. When high-pressure systems reached the sampling site, the

330

surface wind speed was reduced while the stability was enhanced, and unfavorable diffusion

331

conditions was further strengthened (PE1-PE4). High-pressure systems can promote strong air

332

mass-sinking motions (Seo et al., 2017; Zhang et al., 2016), suppress the PBL height, induce humid and

333

warm advection, and strengthen the inversion layer. The developed temperature inversion layer acted

334

as a lid to suppress vertical diffusion, while the low surface wind-speed favored the accumulation of

335

pollutants along the horizontal scale. High RH, enhanced stability and transiently rebounded

336

temperature could initiate a series of chemical chain reactions (Wang et al., 2014). However, the

337

weather patterns before November 1 and after March 30 were generally characterized by a

338

low-pressure system, sharp temperature gradient, strong wind, and low RH, which were all conducive

339

to effective regional pollution transport. Previous studies also demonstrated that sand or floating dust

340

caused by cyclonic activity and the thawed land surface could also be a major cause of the deteriorated

341

air quality in Northern China during spring.

342

3.3 Role of secondary aerosol transformation in high density PM2.5 pollution

343

Recent studies reported that secondary aerosols drive haze formation in China (Seo et al., 2017;

344

Zheng et al., 2016). To explore the underlying mechanism, this study conducted further episode-based

345

analysis of SOC, OC/EC, sulfur oxidation ratios (SOR) and Nitrogen oxidation ratios (NOR). OC is

346

generally composed of POC and SOC, while EC is primarily derived from the incomplete combustion

347

of carbon-containing materials (Zong et al., 2018). SOC is catalyzed by crucial oxidants, such as O3

348

and OH, in photochemistry production and aerosol-aging (Jimenez et al., 2009; Hallquist et al., 2009),

349

while the

350

evaluate the contribution of secondary organic chemical reactions (Zheng et al., 2015). In this study,

351

the OC/EC ratio was strongly correlated with SOC (r=0.907), suggesting that it can be used as an

352

indicator of secondary reactions. SOR and NOR (SOR = nSO42-/(nSO42- + nSO2, in mol), NOR =

353

nNO3-/(nNO3- + nNO2, in mol)) are usually calculated to indicate the secondary conversion efficiency

354

of atmospheric sulfur and nitrogen, and it has been reported that the content of SOR in the primary

355

source emissions was below 0.10 (Pearson, 1979; Truex et al., 1980), while it exceeded 0.10 when

OC/EC ratio can eliminate the physical variations of mixing/dilution and be used to better

356

sulfate was produced through the photochemical oxidation of SO2 (Ohta and Okita, 1990; Zheng et al.,

357

2016; Li et al., 2017). Higher values of SOR and NOR indicated extensive secondary oxidation

358

reactions, and lower values were considered to indicate primary source emissions (Sun et al., 2006;

359

Pearson, 1979). The higher OC/EC ratios, estimated SOC fraction, SOR, and NOR, along with higher

360

pollution, suggested that photochemistry and heterogeneous chemistry were enhanced during the

361

heating season (Table. S2).

362

The SOC concentration and OC/EC ratio in HP were higher than those in SP, and extremely

363

enhanced in SPE, EP5, and EP6, which increased their average value in HP and indicated reduced

364

photochemistry and aerosol-aging during the other pollution episodes. However, the SOC fraction was

365

not increased in all PEs, and significantly lower values than those on polluted days and clean days

366

were observed during PE1-PE4 and PE7, suggesting a weakened photochemical reaction. This study

367

attempted to explore the underlying reason for this based on meteorological parameters (Fig. S4), and

368

the results showed that SOC and OC were positively correlated with temperature and PBL, and

369

negatively correlated with RH. As discussed above, the sampling site experienced a high-pressure

370

system during PE1-PE3, and the sinking airflow decreased the PBL height, weakened vertical mixing

371

and aggravated the accumulation of pollutants. Owing to the strong radiation effect of the accumulated

372

particles, the solar radiation intensity and photolysis reaction were weakened, which is important in

373

the production of radicals and near-surface ozone, and eventually inhibited the formation of SOC. The

374

enhanced secondary organic reaction in SPE, EP5 and EP6 corresponded well with the densely

375

distributed fire points as supported by the fire maps (Fig. S3). Furthermore, the PBL in these episodes

376

was relatively high (Fig. 4), and the large amount of aerosol of released organic aerosol together with

377

the strong radiation intensity tremendously promoted the photochemical reactions and SOC formation.

378

Therefore, the photochemical reactions were directly affected by the radiation intensity and oxidant

379

concentration (Ozone (O3) and Hydroxyl radicals (OH)), which was also observed for haze in Beijing

380

(Zheng et al., 2015), and then affected by the synoptic weather processes and primary emission

381

intensity.

382

NOR exceeded SOR throughout the study period, and the concentrations of nitrate exceeded

383

those of sulfate. There was no strong relationship or synchronous variations between SOR and NOR.

384

On normal days, the values of SOR and NOR were low, while they were elevated during episodes of

385

high pollution, reaching average values of 0.16/0.19 during the SPE, when SOR and NOR grew

386

synchronously, although there was a lag in the growth of NOR. During the PEs, the values of SOR

387

were mostly below the threshold of 0.1, and were not strongly correlated with the sulfate concentration

388

(r=0.528), while NOR mostly exceeded the threshold and was strongly correlated with the nitrate

389

concentration (r=0.938). These results suggest extremely enhanced secondary sulfate and nitrogen

390

conversion reactions occurred during the SPE, and sulfate formed faster than nitrate. During the PEs,

391

nitrate was mainly formed by oxidation, while sulfate mainly originated from the primary emission

392

sources. It has been reported that nitrate is generally formed following two pathways: oxidation of

393

nitrogen dioxide by O3 during the day, and oxidation by ·OH and the heterogeneous reaction of

394

nitrogen pentoxide on the surface of the particles during the night (Zheng et al., 2015; Li et al., 2017).

395

During the SPE and PE1-PE3, the high NOR values were most likely linked to the heterogeneous

396

reactions with O2 under the catalysis with a transition metal due to weakened photolysis. Additionally,

397

ammonium was closely related to RH (0.647), sulfate (0.739), nitrate (0.658), and NOR (0.624). As

398

the main alkaline component of SNA, ammonium plays an important role in neutralizing H2SO4 and

399

HNO3, while promoting NOR and SNA formation. With the onset of the aqueous phase due to the high

400

RH (> 60%), the heterogeneous reactions were enhanced in the SPE, and EP1-EP3, resulting in the

401

rapid accumulation of SAN (14.37%-19.74%; Fig. 3, Table S2). The predominant roles of RH,

402

aerosol liquid water content and aerosol acidity in the heterogeneous oxidation of fine particle

403

explosive growth events were also profoundly demonstrated by the online observation in Beijing (Liu

404

et al., 2019; Liu et al., 2017). The secondary reaction was highly complex in EP4-EP7 due to the

405

occasionally enhanced/weakened photochemical and heterogeneous reactions; however, there is no

406

clear cause can be extracted. Multiple factors may have been responsible, including volatile weather

407

systems, stronger biomass burning activities and frequent dust events.

408

3.4 Source apportionment

409

The total concentration of the analyzed inorganic elements was 2.59 µg/m3, which was only 1.58%

410

of the total PM2.5 mass concentration. These elements constitute a minor fraction, but are important

411

tracers in atmospheric environments (Chang et al., 2017). The concentration of Fe was highest among

412

all trace metals, with a proportion of 32.23%, and its concentration was highest in spring, manifesting

413

great contributions from crustal dust. To evaluate the contamination degree of inorganic elements and

414

determine whether they originated anthropogenic or natural sources, the enrichment factors (EFs) were

415

calculated (Fig. 7). The EF values of Ti, Rb, V, Cs, Ba, Co, Mn, and Sr were below 10, and their

416

concentrations were high in spring and autumn but low in winter, This suggests that there was almost

417

no enrichment from anthropogenic sources, and these elements can be used as tracers of mineral dust.

418

The EFs values of Ni, Tl, Cr, Ag, Zn, Sn, As, Pb, Cu, Sb and Cd exceeded 10, indicating that they

419

were strongly enriched by anthropogenic sources.

420 421

Fig. 7 Enrichment factor analysis of the trace metals.

422

This study used PMF version 5.0 to identify the potential sources of PM2.5 (biomass burning,

423

vehicle emission, biomass burning, mineral dust, and secondary aerosols) and estimate their

424

contributions based on above analysis. The PMF model was run 100 times, and most residuals were

425

between -3 and 3. The coefficients of determination for the observed and predicted values (r2) mostly

426

exceeded 0.6, indicating that the noel was reasonable for explaining the information contained in the

427

original data. The modeled source profiles and percentage contributions are shown in Fig. 8 (a). The

428

first factor interpreted low proportions of carbonaceous materials and SNA, but high proportions of

429

crust tracers, such as Sr, Fe, Al, and Ba according to the EFs analysis and Na+, and its normalized

430

source contributions remained near 0 before January, but gradually increased and eventually peaked

431

during March and April (Fig. S3), which agreed well with the prevalence of dust events in northern

432

China during spring. Therefore, this source was identified as mineral dust. The second source was

433

identified as biomass burning due to the relatively high proportions of OC and EC, and extremely high

434

proportions of plant components, such as K+, Cl-, F- and Mg2+. The contribution of the second source

435

during fall and spring were significantly and slightly elevated, respectively, which corresponded well

436

with the burning of agricultural waste as indicated by the fire maps (Fig. S3). As it is one of the most

437

productive agricultural regions of China, agricultural waste is commonly burned in this region, which

438

is the main cause of the occurrence of severe haze after harvesting and before sowing (Wang et al.,

439

2011b; Qin et al., 2014). The third source is traffic emissions, which are strongly enriched with heavy

440

metals, such as Cr, Mn, Cu, Ti and Ga, and crustal tracers, such as V, Ba, Cs, Rb and Fe. V is usually

441

used as an additive in lubricating oil and enters the atmosphere through tailpipe emissions. Cu, Mn, Ga,

442

Ti, and Cr are used in catalytic converters, tire manufacturing, and brake linings, and Ba, Cs, Rb and

443

Fe are mainly contained in fugitive dust (Han et al., 2017; Chang et al., 2017; Lin et al., 2015), The

444

normalized concentration of this factor fluctuated stably and was consistent with the vehicle pollution

445

characteristics. The fourth factor was characterized by high proportions of EC, SO42-, As, Pb, Co, Ga,

446

Ag, Sn, Sb, Tl, and Zn, and was consistent with the emission features of coal combustion. The

447

contribution of this source first increased and then decreased, following an opposite trend to

448

temperature, and the variations in the contribution were consistent with the demand of residential coal

449

burning under low-temperature conditions. The final source is secondary aerosol owning to the

450

significantly high proportions of OC, SO42-, NO3-, and NH4+. The contribution of secondary aerosols

451

was higher in SPE, PE1-PE3, and PE5, which was consistent with the secondary aerosol

452

transformation analysis above. High contributions of secondary aerosol to PM2.5 have been observed in

453

many studies conducted in China, such as Jing-Jin-Ji (9.0%-61.6%) and Chengdu (33%-44%; Table

454

S3).

455

Secondary aerosol (32.30%) played the most important role in PM2.5 pollution during the

456

heating season, followed by traffic emissions (23.90%) and biomass burning (23.30%), while coal

457

combustion (6.2%) and mineral dust (14.4%) were small contributors. The contributions of different

458

sources to the PM2.5 mass concentration under different pollution levels are shown in Fig. 8(b).

459

Secondary aerosol formation contributed the most to the rapid accumulation of PM2.5 during polluted

460

episodes, accounting for 41.40%, and 38.2% in the SPE and PEs respectively. Biomass burning was

461

the fundamental cause of SPE, while traffic emission (24.90%), coal combustion (16.10%) and

462

mineral dust (13.20%) combined with unfavorable weather conditions collectively caused the PEs. As

463

indicated by the analysis above, the emissions from coal combustion and vehicles fluctuated regularly

464

during the heating season, and the PEs occurred irregularly to the unfavorable synoptic weather.

465

Traffic emissions and biomass burning were identified as major factors by the PMF model, which

466

offered new insights to alleviate PM2.5 pollution. Preventing straw burning could significantly reduce

467

the occurrence of pollution episodes during autumn and spring. Additionally, further managing traffic

468

pollution could also be effective. Therefore, clean energy vehicles and improved gasoline should be

469

widely implemented in the public transportation system.

470 471

Fig. 8 Source profile (blue bars) and contribution percentages (red dots) in the PMF model (a); sources

472

contributions to the PM2.5 mass concentration (b).

473

3.5 Regional transportation

474

Regional transportation was found to contribute considerably to the rapid accumulation of PM2.5.

475

Therefore, the trajectories of the air masses and potential geographical origins were analyzed to

476

explore its influence. Fig. 9 shows the 72-h backward trajectories of air masses, trajectory clusters, and

477

the potential geographical origins during the heating season. On the sampling days, the air masses

478

mostly originated over Russian and Mongolia, which experienced fast-moving and long-distance

479

transmissions before reaching the sampling site. Based on the trajectory routes, four clusters were

480

identified which all originated from the northwest. This suggests that the Siberian high-pressure and

481

strong prevailing northwesterly winds were predominant during the heating season. The air parcels

482

during the SPE mostly originated from the northwestern region and passed through the

483

Harbin-Changchun megalopolis (Fig. S2), and some originated from the North China Plain. Severe

484

haze was monitored in other cities of the region at the same time (Table S4), which likely contributed

485

to the rapid accumulation of PM2.5. During the PEs, the air masses mainly originated from the

486

northwestern regions of Harbin, and likely carried pollutants from the Harbin-Changchun megalopolis

487

(Fig. S2), as indicated by the monitoring data (Fig. S4).

488 489

Fig. 9 72-h air mass backward trajectories and trajectory clusters (a-c); Potential geographic origins of

490

PM2.5 (d) and source factors apportioned from the PMF model (e-i).

491

To determine the potential geographical origins of PM2.5 and the source factors given by the PMF

492

model, this study conducted PSCF analysis based on the HYSPLIT model, and the possible

493

geographical origins were indicated in different colors. Regions that were highly likely to be sources

494

of PM2.5 (> 0.3) were mainly located in the Harbin-Changchun megalopolis, and were also

495

sporadically distributed in farther regions, such as the North China Plain and the border of Inner

496

Mongolia and Russian. Therefore, the PM2.5 in Harbin was mainly affected by local emissions and

497

slightly affected by long-distance transmission. The potential geographical origins of the source

498

factors were similar to those of PM2.5, but with different probabilities. The probabilities of biomass

499

burning and secondary aerosol being origins were lower than those of the other factors in lighter colors,

500

reflecting weaker transport of their emissions. The coal combustion and traffic emission sources were

501

stronger than the other factors, and the pollutants received by the sampling site likely had a remote

502

source. Traffic emission sources were densely distributed in the Harbin-Changchun megalopolis,

503

which has heavily used roads, while coal combustion sources were scattered across large regions

504

where coal-fired plants may have existed. The mineral dust sources were distributed in remote areas,

505

but were as the local sources, which was consistent with the distribution of natural sources. In

506

conclusion, the PM2.5 in Harbin was mainly affected by regional transportation from the

507

Harbin-Changchun megalopolis and long-distance transmission from the northwestern region.

508

Biomass burning and secondary aerosol were slightly affected by regional transportation, while traffic

509

emissions, coal combustion, and mineral dust were severely affected. According to Fig. 9, traffic

510

emissions were densely distributed in the Harbin-Changchun megalopolis, while coal combustion and

511

mineral dust were scattered across large scale regions, which was consistent with their distribution

512

features. Based on the above analysis, regional pollutant emissions should be mitigated and prevented

513

by different regions and cities, and at various levels of the government.

514

4. Conclusion

515

Several factors were analyzed in Harbin during heating season to elucidate the causes of the

516

formation of high PM2.5 mass concentrations from four aspects, i.e., synoptic weather, secondary

517

transformation, source apportionment, and regional transportation. The major conclusions are as

518

follows:

519

During the heating season, the level of PM2.5 varied dynamically, and its concentration exceeded

520

75 µg/m3 in 85% of the samples. Three pollution levels (C, SP, and HP) and 8 polluted episodes (SPE,

521

PE1-PE7) were defined based on the NAAQ standard. The time series of the chemical species

522

exhibited a similar trend to the variations in the PM2.5 concentration, and their concentrations were

523

mainly elevated in pollution episodes. Among all species, the concentration of OM was highest,

524

followed by NO3-, EC, SO42-, Cl-, NH4+, and K+. The fraction of sulfate, EC, and POC decreased as the

525

pollution level increased, while those of SOC, NO3-, Cl-, NH4+, and K+ fraction increased and were

526

most likely to be the driving factors. Five source factors were apportioned from the PMF model, and

527

their importance decreased in the following order: secondary aerosol (32.30%) > traffic emission

528

(23.90%) > biomass burning (23.30%) > mineral dust (14.4%) > coal combustion (6.2%). The PM2.5 in

529

Harbin was mainly affected by regional transportation from the Harbin-Changchun megalopolis and

530

slightly affected by long-distance transmission. Biomass burning and secondary aerosols were slightly

531

affected by regional transportation, while traffic emissions, coal combustion, and mineral dust were

532

severely affected. Traffic emissions were densely distributed in the Harbin-Changchun megalopolis,

533

while coal combustion and mineral dust were scattered in wider regions, which was consistent with

534

their distribution features.

535

This study mainly focused on episodes of pollution. The SPE in autumn was the most striking

536

part period studied, and was characterized by longer-lasting and higher pollutant concentrations than

537

those observed in the other periods. Among all chemical species, the concentrations of chlorine,

538

potassium, carbonaceous components, SNA, SOR, and NOR were particularly elevated, and SPE was

539

most likely attributed to extensive biomass burning and enhanced secondary transformation, which

540

contributed to 41.40% and 37.8% of the PM2.5, respectively.

541

The PEs were characterized by slightly elevated chemical species, and two categories were

542

identified according to the different chemical features observed in the different episodes: pure-haze

543

(PE1-PE3) in winter with a high proportion of secondary aerosol, and mixed haze-dust (PE4-PE7)

544

with a high proportion of mineral dust. Enhanced heterogeneous reactions and weakened

545

photochemical reactions were generally observed during pure haze events, which can be attributed to

546

the high RH (> 60%), low WS (3m/s), suppressed PBL height (< 200 m), and strengthened inversion

547

layer under the control of the high-pressure system. In the mixed haze-dust event that occurred in

548

spring, biomass burning activities, mineral dust, favorable diffusion, and regional transportation were

549

observed, which may have caused the pollution episodes. In summary, the PEs were jointly caused by

550

secondary aerosols (38.20%), traffic emissions (23.90%), coal combustion (16.10%), and mineral dust

551

(14.40%).

552 553

Author contributions. Xiazhong Sun: Data curation, Formal analysis, Writing- Original draft preparation,

554

Writing- Reviewing and Editing. Kun Wang: Resources, Methodology, Conceptualization. Bo Li: Resources.

555

Zheng Zong and Xiaofei Shi: Methodology and Formal analysis. Lixin Ma, Donglei Fu and Samit Thapa:

556

Investigation. Hong Qi: Resources, Methodology, Conceptualization, Supervision, Project administration,

557

Funding acquisition, Writing- Reviewing and Editing. Chongguo Tian: Resources, Methodology,

558

Conceptualization, Writing- Reviewing and Editing.

559 560

Declaration of Interest Statement. The authors declare that they have no conflict of interest.

561 562

Acknowledgement. This work was supported financially by the National Science Foundation of China

563

(51979066) and the State Key Laboratory of Urban Water Resource and Environment Funding, Harbin Institute

564

of Technology (No. 2018TS01).

565

Reference

566

Ashbaugh, L.L., Malm, W.C., and Sadeh, W.Z.: A residence time probability analysis of sulfur concentrations at

567

grand Canyon National Park, Atmos Environ, 19, 1263-1270, 10.1016/0004-6981(85)90256-2, 1985.

568

Bei, N., Li, G., Huang, R., Cao, J., Meng, N., Feng, T., Liu, S., Zhang, T., Zhang, Q., and Molina, L.T.: Typical

569

synoptic situations and their impacts on the wintertime air pollution in the Guanzhong basin, China, Atmos

570

Chem Phys, 16, 7373-7387, 10.5194/acp-16-7373-2016, 2016.

571

Bressi, M., Sciare, J., Ghersi, V., Mihalopoulos, N., Petit, J.E., Nicolas, J.B., Moukhtar, S., Rosso, A., Féron, A.,

572

Bonnaire, N., Poulakis, E., and Theodosi, C.: Sources and geographical origins of fine aerosols in Paris

573

(France), Atmos Chem Phys, 14, 8813-8839, 10.5194/acp-14-8813-2014, 2014.

574 575

Cai, W., Li, K., Liao, H., Wang, H., and Wu, L.: Weather conditions conducive to Beijing severe haze more frequent under climate change, Nat Clim Change, 7, 257-262, 10.1038/nclimate3249, 2017.

576

Cao, F., Zhang, S., Kawamura, K., and Zhang, Y.: Inorganic markers, carbonaceous components and stable carbon

577

isotope from biomass burning aerosols in Northeast China, Sci Total Environ, 572, 1244-1251,

578

10.1016/j.scitotenv.2015.09.099, 2016.

579

Chang, Y., Huang, K., Deng, C., Zou, Z., Liu, S., and Zhang, Y.: First long-term and near real-time measurement of

580

atmospheric trace elements in Shanghai, China, Atmospheric Chemistry and Physics Discussions, 1-49,

581

10.5194/acp-2017-613, 2017.

582

Chen, F., Kusaka, H., Bornstein, R., Ching, J., Grimmond, C.S.B., Grossman-Clarke, S., Loridan, T., Manning,

583

K.W., Martilli, A., Miao, S., Sailor, D., Salamanca, F.P., Taha, H., Tewari, M., Wang, X., Wyszogrodzki, A.A.,

584

and Zhang, C.: The integrated WRF/urban modelling system: development, evaluation, and applications to

585

urban environmental problems, Int J Climatol, 31, 273-288, 10.1002/joc.2158, 2011.

586

Chen, Q., Yuan, Y., Huang, X., He, Z., and Tan, H.: Assessment of column aerosol optical properties using

587

ground-based sun-photometer at urban Harbin, Northeast China, J Environ Sci-China, 74, 50-57,

588

10.1016/j.jes.2018.02.003, 2018.

589

Gao, J., Wang, K., Wang, Y., Liu, S., Zhu, C., Hao, J., Liu, H., Hua, S., and Tian, H.: Temporal-spatial

590

characteristics and source apportionment of PM2.5 as well as its associated chemical species in the

591

Beijing-Tianjin-Hebei region of China, Environ Pollut, 233, 714-724, 10.1016/j.envpol.2017.10.123, 2018.

592

Hallquist, M., Wenger, J.C., Baltensperger, U., Rudich, Y., Simpson, D., Claeys, M., Dommen, J., Donahue, N.M.,

593

George, C., Goldstein, A.H., Hamilton, J.F., Herrmann, H., Hoffmann, T., Iinuma, Y., Jang, M., Jenkin, M.,

594

Jimenez, J.L., Kiendler-Scharr, A., Maenhaut, W., McFiggans, G., Mentel, T.F., Monod, A., Prévôt, A.S.H.,

595

Seinfeld, J.H., Surratt, J.D., Szmigielski, R., and Wildt, J.: The formation, properties and impact of secondary

596

organic aerosol: current and emerging issues, Atmospheric Chemistry and Physics Discussions, 9, 3555-3762,

597

10.5194/acpd-9-3555-2009, 2009.

598 599

Han, F., Kota, S.H., Wang, Y., and Zhang, H.: Source apportionment of PM2.5 in Baton Rouge, Louisiana during 2009–2014, Sci Total Environ, 586, 115-126, 10.1016/j.scitotenv.2017.01.189, 2017.

600

He, H., Wang, Y., Ma, Q., Ma, J., Chu, B., Ji, D., Tang, G., Liu, C., Zhang, H., and Hao, J.: Mineral dust and NOx

601

promote the conversion of SO2 to sulfate in heavy pollution days, Sci Rep-Uk, 4, 10.1038/srep04172, 2015.

602

Jimenez, J.L., Canagaratna, and M.R., Donahue, N.M, et al.: Evolution of Organic Aerosols in the Atmosphere,

603 604 605 606 607

Science, 326, 1525-1529, 10.1126/science.1180353, 2009. Kersten, M., and Smedes, F.: Normalization procedures for sediment contaminants in spatial and temporal trend monitoring, J Environ Monit, 4, 109-115, 10.1039/b108102k, 2002. Kumar, M., Li, H., Zhang, X., Zeng, X.C., and Francisco, J.S.: Nitric Acid–Amine Chemistry in the Gas Phase and at the Air–Water Interface, J Am Chem Soc, 140, 6456-6466, 10.1021/jacs.8b03300, 2018.

608

Lee, E., Chan, C.K., and Paatero, P.: Application of positive matrix factorization in source apportionment of

609

particulate pollutants in Hong Kong, Atmos Environ, 33, 3201-3212, 10.1016/S1352-2310(99)00113-2,

610

1999.

611

Lesniok, M., Malarzewski, L., and Niedzwiedz, T.: Classification of circulation types for Southern Poland with an

612

application to air pollution concentration in Upper Silesia, Phys Chem Earth, 35, 516-522,

613

10.1016/j.pce.2009.11.006, 2010.

614

Li, H., Wang, Q., Yang, M., Li, F., Wang, J., Sun, Y., Wang, C., Wu, H., and Qian, X.: Chemical characterization

615

and source apportionment of PM2.5 aerosols in a megacity of Southeast China, Atmos Res, 181, 288-299,

616

10.1016/j.atmosres.2016.07.005, 2016.

617

Li, L., Tan, Q., Zhang, Y., Feng, M., Qu, Y., An, J., and Liu, X.: Characteristics and source apportionment of PM2.5

618

during persistent extreme haze events in Chengdu, southwest China, Environ Pollut, 230, 718-729,

619

10.1016/j.envpol.2017.07.029, 2017.

620

Li, X., Jin, L., and Kan, H.: Air pollution: a global problem needs local fixes, Nature, 570, 2019.

621

Lim, Y.B., Seo, J., Kim, J.Y., and Turpin, B.J.: Multiday haze in the East Asia: Transport and chemical

622

aging

623

10.5194/acp-2018-364, 2018.

of

hygroscopic

particles,

Atmospheric

Chemistry

and

Physics

Discussions,

1-28,

624

Lin, Y.C., Tsai, C.J., Wu, Y.C., Zhang, R., Chi, K.H., Huang, Y.T., Lin, S.H., and Hsu, S.C.: Characteristics of trace

625

metals in traffic-derived particles in Hsuehshan Tunnel, Taiwan: size distribution, potential source, and

626

fingerprinting metal ratio, Atmos Chem Phys, 15, 4117-4130, 10.5194/acp-15-4117-2015, 2015.

627

Liu, Z., Hu, B., Ji, D., Cheng, M., Gao, W., Shi, S., Xie, Y., Yang, S., Gao, M., Fu, H., Chen, J., and Wang, Y.:

628

Characteristics of fine particle explosive growth events in Beijing, China: Seasonal variation, chemical

629

evolution

630

10.1016/j.scitotenv.2019.06.068, 2019.

pattern

and

formation

mechanism,

Sci

Total

Environ,

687,

1073-1086,

631

Liu, Z., Xie, Y., Hu, B., Wen, T., Xin, J., Li, X., and Wang, Y.: Size-resolved aerosol water-soluble ions during the

632

summer and winter seasons in Beijing: Formation mechanisms of secondary inorganic aerosols,

633

Chemosphere, 183, 119-131, 10.1016/j.chemosphere.2017.05.095, 2017.

634

Ning, G., Wang, S., Yim, S.H.L., Li, J., Hu, Y., Shang, Z., Wang, J., and Wang, J.: Impact of low-pressure systems

635

on winter heavy air pollution in the northwest Sichuan Basin, China, Atmos Chem Phys, 18, 13601-13615,

636

10.5194/acp-18-13601-2018, 2018.

637 638

Ohta, S., and Okita, T.: A chemical characterization of atmospheric aerosol in Sapporo, Atmospheric Environment Part A General Topics, 24, 815-822, 1990.

639

Owens, P.N., Batalla, R.J., Collins, A.J., Gomez, B., Hicks, D.M., Horowitz, A.J., Kondolf, G.M., Marden, M.,

640

Page, M.J., Peacock, D.H., Petticrew, E.L., Salomons, W., and Trustrum, N.A.: Fine-grained sediment in river

641

systems: Environmental significance and management issues, River Res Appl, 21, 693-717, 10.1002/rra.878,

642

2005.

643

Paatero, P., Eberly, S., Brown, S.G., and Norris, G.A.: Methods for estimating uncertainty in factor analytic

644

solutions, Atmospheric Measurement Techniques Discussions, 6, 7593-7631, 10.5194/amtd-6-7593-2013,

645

2013.

646 647 648 649

Pearson, W.R.B.W.: Sulfate Emissions from Catalyst-Equipped Automobiles on the Highway, Air Repair, 3, 255-257, 1979. Qin, X., Yan, H., Zhan, Z., and Li, Z.: Characterising vegetative biomass burning in China using MODIS data, Int J Wildland Fire, 23, 69, 10.1071/WF12163, 2014.

650

Seo, J., Kim, J.Y., Youn, D., Lee, J.Y., Kim, H., Lim, Y.B., Kim, Y., and Jin, H.C.: On the multiday haze in the

651

Asian continental outflow: the important role of synoptic conditions combined with regional and local

652

sources, Atmos Chem Phys, 17, 9311-9332, 10.5194/acp-17-9311-2017, 2017.

653 654

Sun, Y., Zhuang, G., Tang, A., Wang, Y., and An, Z.: Chemical Characteristics of PM2.5 and PM10 in Haze−Fog Episodes in Beijing, Environ Sci Technol, 40, 3148-3155, 10.1021/es051533g, 2006.

655

Tao, J., Gao, J., Zhang, L., Zhang, R., Che, H., Zhang, Z., Lin, Z., Jing, J., Cao, J., and Hsu, S.C.: PM2.5 pollution in

656

a megacity of southwest China: source apportionment and implication, Atmospheric Chemistry and Physics

657

Discussions, 14, 5147-5196, 10.5194/acpd-14-5147-2014, 2014.

658

Truex, T.J., Pierson, W.R., and Mckee, D.E.: Sulfate in Diesel Exhaust, pp 1118-1121 Environ. Sci. Technol, 1980.

659

Wang, G., Cheng, S., Li, J., Lang, J., Wen, W., Yang, X., and Tian, L.: Source apportionment and seasonal variation

660

of PM2.5 carbonaceous aerosol in the Beijing-Tianjin-Hebei Region of China, Environ Monit Assess, 187,

661

10.1007/s10661-015-4288-x, 2015.

662

Wang, J., Wang, S., Jiang, J., Ding, A., Zheng, M., Zhao, B., Wong, D.C., Zhou, W., Zheng, G., Wang, L., Pleim,

663

J.E., and Hao, J.: Impact of aerosol-meteorology interactions on fine particle pollution during China's severe

664

haze episode in January 2013, Environ Res Lett, 9, 10.1088/1748-9326/9/9/094002, 2014.

665

Wang, Q., Zhuang, G., Huang, K., Liu, T., Deng, C., Xu, J., Lin, Y., Guo, Z., Chen, Y., Fu, Q., Fu, J.S., and Chen, J.:

666

Probing the severe haze pollution in three typical regions of China: Characteristics, sources and regional

667

impacts, Atmos Environ, 120, 76-88, 10.1016/j.atmosenv.2015.08.076, 2015.

668

Wang, X., Zong, Z., Tian, C., Chen, Y., Luo, C., Li, J., Zhang, G., and Luo, Y.: Combining Positive Matrix

669

Factorization and Radiocarbon Measurements for Source Apportionment of PM2.5 from a National

670

Background Site in North China, Sci Rep-Uk, 7, 10.1038/s41598-017-10762-8, 2017.

671

Wang, Y., Yao, L., Wang, L., Liu, Z., Ji, D., Tang, G., Zhang, J., Sun, Y., Hu, B., and Xin, J.: Mechanism for the

672

formation of the January 2013 heavy haze pollution episode over central and eastern China, Science China

673

Earth Sciences, 57, 14-25, 10.1007/s11430-013-4773-4, 2014.

674

Wang, Y., Zhang, Q., Jiang, J., Zhou, W., Wang, B., He, K., Duan, F., Zhang, Q., Philip, S., and Xie, Y.: Enhanced

675

sulfate formation during China's severe winter haze episode in January 2013 missing from current models, J

676

Geophys Res-Atmos, 119, 10.1002/2013JD021426, 2014.

677

Wang, Y., Zhuang, G., Zhang, X., Huang, K., Xu, C., Tang, A., Chen, J., and An, Z.: The ion chemistry, seasonal

678

cycle, and sources of PM2.5 and TSP aerosol in Shanghai, Atmos Environ, 40, 2935-2952,

679

10.1016/j.atmosenv.2005.12.051, 2006.

680

Wang, Y.Q., Zhang, X.Y., and Draxler, R.R.: TrajStat: GIS-based software that uses various trajectory statistical

681

analysis methods to identify potential sources from long-term air pollution measurement data, Environ

682

Modell Softw, 24, 938-939, 10.1016/j.envsoft.2009.01.004, 2009.

683

Wang, Z., Song, K., Ma, W., Ren, C., Zhang, B., Liu, D., Chen, J.M., and Song, C.: Loss and Fragmentation of

684

Marshes

685

10.1007/s13157-011-0209-0, 2011a.

686

in

the

Sanjiang

Plain,

Northeast

China,

1954

-

2005,

Wetlands,

31,

945-954,

Whiteman, C.D., Hoch, S.W., Horel, J.D., and Charland, A.: Relationship between particulate air pollution and

687

meteorological

688

10.1016/j.atmosenv.2014.06.012, 2014.

variables

in

Utah's

Salt

Lake

Valley,

Atmos

Environ,

94,

742-753,

689

Zhang, R., Jing, J., Tao, J., Hsu, S.C., Wang, G., Cao, J., Lee, C.S.L., Zhu, L., Chen, Z., and Zhao, Y.: Chemical

690

characterization and source apportionment of PM2.5 in Beijing: seasonal perspective, Atmospheric Chemistry

691

and Physics Discussions, 13, 9953-10007, 10.5194/acpd-13-9953-2013, 2013.

692

Zhang, W., Feng, H., Chang, J., Qu, J., Xie, H., and Yu, L.: Heavy metal contamination in surface sediments of

693

Yangtze River intertidal zone: An assessment from different indexes, Environ Pollut, 157, 1533-1543,

694

10.1016/j.envpol.2009.01.007, 2009.

695

Zhang, Z., Zhang, X., Gong, D., Kim, S.J., Mao, R., and Zhao, X.: Possible influence of atmospheric circulations

696

on winter haze pollution in the Beijing–Tianjin–Hebei region, northern China, Atmos Chem Phys, 16,

697

561-571, 10.5194/acp-16-561-2016, 2016.

698

Zhao, H., Tong, D.Q., Gao, C., and Wang, G.: Effect of dramatic land use change on gaseous pollutant emissions

699

from biomass burning in Northeastern China, Atmos Res, 153, 429-436, 10.1016/j.atmosres.2014.10.008,

700

2015.

701

Zheng, G., Duan, F., Ma, Y., Zhang, Q., Huang, T., Kimoto, T., Cheng, Y., Su, H., and He, K.: Episode-Based

702

Evolution Pattern Analysis of Haze Pollution: Method Development and Results from Beijing, China,

703

Environ Sci Technol, 50, 4632-4641, 10.1021/acs.est.5b05593, 2016.

704

Zheng, G.J., Duan, F.K., Su, H., Ma, Y.L., Cheng, Y., Zheng, B., Zhang, Q., Huang, T., Kimoto, T., Chang, D.,

705

Pöschl, U., Cheng, Y.F., and He, K.B.: Exploring the severe winter haze in Beijing: the impact of synoptic

706

weather, regional transport and heterogeneous reactions, Atmos Chem Phys, 15, 2969-2983,

707

10.5194/acp-15-2969-2015, 2015.

708 709

Zoller, W.H., Gladney, E.S., and Duce, R.A.: Atmospheric concentrations and sources of trace metals at the South pole., Science (New York, N.Y.), 183, 198-200, 10.1126/science.183.4121.198, 1974.

710

Zong, Z., Wang, X., Tian, C., Chen, Y., Fu, S., Qu, L., Ji, L., Li, J., and Zhang, G.: PMF and PSCF based source

711

apportionment of PM2.5 at a regional background site in North China, Atmos Res, 203, 207-215,

712

10.1016/j.atmosres.2017.12.013, 2018.

Declaration of Interest Statement. The authors declare that they have no conflict of interest.