Uncertainty Assessment of Gaseous Oxidized Mercury

Subscriber access provided by University of Newcastle, Australia

Article

Uncertainty Assessment of Gaseous Oxidized Mercury Measurements Collected by Atmospheric Mercury Network Irene Cheng, and Leiming Zhang Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b04926 • Publication Date (Web): 06 Dec 2016 Downloaded from http://pubs.acs.org on December 14, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 30

Environmental Science & Technology

Uncertainty Assessment of Gaseous Oxidized Mercury Measurements Collected by Atmospheric Mercury Network Irene Cheng* and Leiming Zhang* Air Quality Research Division, Science and Technology Branch, Environment and Climate Change Canada, Toronto, Ontario M3H 5T4 Canada *Corresponding authors: Irene Cheng Air Quality Research Division Science and Technology Branch Environment and Climate Change Canada 4905 Dufferin Street Toronto, Ontario M3H 5T4 Canada Tel: 1 416-739-4455; Fax: 1 416-739-4281 Email: [email protected] Leiming Zhang Air Quality Research Division Science and Technology Branch Environment and Climate Change Canada 4905 Dufferin Street Toronto, Ontario M3H 5T4 Canada Tel: 1 416-739-5734; Fax: 1 416-739-4281 Email: [email protected]

ACS Paragon Plus Environment


1

ABSTRACT

2

Gaseous oxidized mercury (GOM) measurement uncertainties undoubtedly impact the

3

understanding of mercury biogeochemical cycling; however, there is a lack of consensus on the

4

uncertainty magnitude. The numerical method presented in this study provides an alternative

5

means of estimating the uncertainties of previous GOM measurements. Weekly GOM in

6

ambient air was predicted from measured weekly mercury wet deposition using a scavenging

7

ratio approach, and compared against field measurements of 2-4 hourly GOM to estimate the

8

measurement biases of the Tekran speciation instruments at thirteen Atmospheric Mercury

9

Network (AMNet) sites. Multi-year average GOM measurements were estimated to be biased

10

low by more than a factor of 2 at six sites, between a factor of 1.5 and 1.8 at six other sites, and

11

below a factor of 1.3 at one site. The differences between predicted and observed were

12

significantly larger during summer than other seasons potentially because of higher ozone

13

concentrations that may interfere with GOM sampling. The analysis of multi-year data collected

14

at multiple sites provides a consensus of the systematic bias in GOM measurements, suggesting

15

the need for further development of new measurement technologies and identifying the chemical

16

composition of GOM.

17

2 ACS Paragon Plus Environment

Page 2 of 30

Page 3 of 30


18

1. INTRODUCTION

19

Accurate measurements of gaseous oxidized mercury are indispensable to the study of mercury

20

cycling in the environment, including identifying mercury transformation mechanisms, modeling

21

its dry deposition, and evaluating mercury chemical transport models. The Tekran speciation

22

system (Tekran® Instruments Corporation) is a widely used set of automated instruments for

23

measuring three operationally-defined forms of atmospheric mercury, including gaseous

24

elemental mercury (GEM), gaseous oxidized mercury (GOM) and particle-bound mercury in fine

25

particles (< 2.5µm) (PBM), at high temporal resolution. The instruments have been deployed

26

around the world by researchers and large-scale mercury monitoring networks, such as the

27

Atmospheric Mercury Network (AMNet)1,2 and Global Mercury Observation System3,4. Despite

28

the prevalent use of the Tekran speciation system, the accuracy of GOM measurements remains

29

a very polarizing issue; however, researchers increasingly acknowledge that GOM and PBM

30

have large measurement uncertainties5. This is because the exact chemical compositions of

31

GOM and PBM are unknown and have not been quantified which hinders the development of

32

calibration standards, and recent studies show there are potential sampling artifacts and

33

discrepancies in GOM concentrations among various measurement methods5-9.

34

Studies suggest elevated ozone can cause a positive artifact on PBM measurements10 and loss of

35

GOM11. Water vapor can also interfere with GOM measurements and cause denuders to become

36

passivated10,12,13. There may be chemical reactions between atmospheric mercury and other

37

gases and aerosols during sampling that can result in GOM loss or positive artifacts on PBM6,14.

38

However, this may only be an issue during long sampling duration (12 h)14. The mechanisms

39

causing the aforementioned issues are also not well established. More recently, it was revealed

40

that denuders do not capture all forms of GOM15, contradictory to an earlier study demonstrating 3 ACS Paragon Plus Environment


41

high collection efficiency of GOM by denuders16. Other measurement techniques, such as mist

42

chambers, nylon and cation exchange membranes, and Detector for Oxidized Hg, were capable

43

of collecting more GOM than KCl-denuders6,15,17,18. However similar to the Tekran instrument,

44

these alternative methods are not immune to sampling artifacts caused by high water vapor15,18

45

and other gases and aerosols17. Studies suggest the temperature settings of the Tekran instrument

46

favour GOM deposition in the sampling lines and gas-particle partitioning and evaporative losses

47

of PBM within the instrument6,7,11,19,20. Slemr et al.21 also reported that the Tekran Hg analyzer

48

was unable to quantify small signal to noise peaks resulting in very low or non-detectable total

49

gaseous mercury concentrations.

50

With the growing number of studies raising issues regarding GOM and PBM uncertainties and

51

their potential consequences on analysis and modeling of mercury, further studies are needed to

52

determine the magnitude of the GOM and PBM uncertainties, improve the instrumentation

53

detection of Hg species, and develop calibration technologies. Current estimates of GOM

54

measurement uncertainties are a factor of 1.3-5.06,15,18 and up to a factor of 127 based on

55

intercomparisons of various sampling and analysis methods. However, consensus has yet to be

56

reached on the uncertainty magnitude. In this study, an indirect numerical modeling approach

57

was developed to quantify the uncertainties of Tekran measurements of GOM at multiple

58

AMNet sites, taking advantage of the large amount of speciated data that have been made

59

available in the past several years (2009-2014). This dataset is longer than those collected from

60

previous intercomparison studies of different measurement techniques, which typically span over

61

a one month period. The approach predicts GOM concentrations from Hg wet deposition since

62

Hg in precipitation originates from GOM and PBM, and a previous study has shown a strong

63

empirical relationship between atmospheric oxidized mercury and mercury wet deposition22.


Page 4 of 30

Page 5 of 30


64

Predicted GOM are then compared to observed GOM to estimate the potential measurement

65

biases.

66

2. METHODS

67

2.1 Site and data description

68

GOM concentrations were predicted at 13 AMNet sites (co-located instruments at MD98/99)

69

located in the U.S. and Canada (Table 1). Ambient monitoring data from 2009-2014 near the

70

AMNet sites were obtained from various networks (Table S1 of the SI). GEM, GOM and PBM

71

measured using the Tekran speciation system and mercury wet deposition fluxes were obtained

72

from the AMNet and Mercury Deposition Network, respectively 23,24. Trace gases and inorganic

73

ions in air and precipitation were obtained from the Clean Air Status and Trends Network

74

(CASTNET25), National Trends Network26, and the Canadian Atmospheric and Precipitation

75

Monitoring Network (CAPMoN27) for the NS01 site (Sect. 1 of SI). PM2.5 and PM10

76

concentrations were obtained from AirData28 and CAPMoN. Meteorological data and ground-

77

level ozone were obtained from CASTNET and CAPMoN. Due to the different sampling

78

intervals of the networks, the datasets were averaged to a weekly interval. The datasets were

79

also averaged to monthly intervals to examine differences between weekly and monthly data

80

resolution and to estimate the method uncertainties. The data was quality controlled according to

81

the criteria in Tables S2 and S3.

82

2.2 Determination of the HNO3 scavenging ratio

83

The premise behind the method for predicting GOM is based on the knowledge of the mercury

84

wet deposition and assumptions on GOM and PBM wet scavenging efficiencies. The wet

85

scavenging of GOM and PBM were quantified using the scavenging ratios of other pollutants. 5 ACS Paragon Plus Environment


Page 6 of 30

86

Scavenging ratios (equation (1)) are typically determined for particulate pollutants, whereas

87

scavenging ratios of gaseous pollutants are rare because the total pollutant (dissolved +

88

particulate) concentration in precipitation is typically measured like in the case of mercury and

89

ammonium, nitrate and sulfate. In the absence of direct measurements, gaseous HNO3 was used

90

as a surrogate for the scavenging ratio of GOM based on the similarities in their physicochemical

91

properties. Both HNO3 and GOM are water soluble and easily scavenged by precipitation. Both

92

gases have low surface resistance and high dry deposition velocity (1-5 cm s-1)16; thus, GOM is

93

modeled like HNO3 in dry deposition models 29-32.

94

Since the scavenging ratio of HNO3 (WHNO3) cannot be determined directly from field

95

measurements, the approach for determining WHNO3 required calculating the scavenging ratios

96

(W) of the particulate-phase inorganic ions Ca2+, Mg2+, Na+, and K+ using equation (1).

97

W =

(1)

98

Cprec and Cair are the precipitation and air concentrations, respectively. The scavenging ratio of

99

coarse PM (WcPM) was determined by averaging WCa, WMg, and WNa since these ions dominate

100

the coarse PM fraction33. WK was used as a surrogate for the scavenging ratio of fine PM (WfPM)

101

for inland sites, whereas WK/2 was assumed for coastal sites (NS01, NY06) following the

102

methodology in Cheng et al.33

103

The wet scavenging of particulate nitrate (pNO3-) is then determined using equation (2):

104

[pNO3-] prec = WfPM [pNO3-] air Pf + WcPM [pNO3-] air (1-Pf)

(2)

105

WfPM and WcPM are the scavenging ratios of fine and coarse PM. [pNO3-]air is the NO3-air

106

concentration. Pf is the fine mass fraction of NO3-. Similar to the methodology in Cheng and 6 ACS Paragon Plus Environment

Page 7 of 30


107

Zhang34, a Pf of 0.84 was assumed for the winter months (DJF), while 0.29 was used for all other

108

months since lower temperatures favor ammonium nitrate production in fine particles.

109

Equation (3) is subsequently used to determine WHNO3.

110

WHNO3 =

[ ] [ ]

=

[ ] [ ] [ ]

(3)

111

[Total NO3-]prec is the total NO3- precipitation concentration and [pNO3-]prec is from equation (2).

112

If [HNO3]prec< 0, it is assumed that only pNO3- contributed to total NO3- in precipitation and

113

WHNO3 is not determined.

114

2.3 Determination of GOM wet deposition and prediction of GOM

115

The detailed methodology for determining the wet deposition flux of GOM is described in Cheng

116

et al.33. A synopsis of the methodology is provided here. First, the wet deposition fluxes of fine

117

PBM and coarse PBM are determined using scavenging ratios, air concentrations of PBM, and

118

precipitation amount. The difference between total Hg wet deposition and the wet deposition

119

fluxes of fine and coarse PBM results in the wet deposition flux of GOM. Given the GOM wet

120

deposition flux (FGOM), the assumption that WHNO3 ≈ WGOM from the previous section, and the

121

precipitation amount (P), the GOM concentration (CGOM) can be predicted using equation (4).

122

123

FGOM = WGOMCGOMP (4)

CGOM =

124

2.4 Comparison of predicted and observed GOM

125

The predicted GOM were compared to the observed GOM to assess the bias in the observed

126

GOM measured by the Tekran system. Bias and normalized mean bias (NMB), which are model 7 ACS Paragon Plus Environment


127

evaluation metrics reported in chemical transport modeling studies35, were used to quantify the

128

bias in GOM (equations (5) and (6)). Bias = Pi - Oi (5)

129

∑ !

130

NMB =

131

BiasF = 1.0+NMB

∑ !

(6)

(7)

132

Pi and Oi are the predicted GOM and observed GOM, respectively. The observed GOM is a

133

weekly or monthly average value calculated from the 2-4 h concentrations. N is the number of

134

weekly or monthly cases. Bias is the difference between predicted and observed GOM, while

135

NMB represents an average percent difference between predicted and observed GOM from all

136

the data35. BiasF is defined here as a bias factor, which can be simply used to adjust measured

137

data by multiplying this factor. The scavenging ratio method uncertainties are described in Sect.

138

2 of the SI and shown in Table S4.

139

Initially, the comparison of predicted and observed GOM was performed on all cases. However,

140

this comparison resulted in an unreasonably large NMB (Table S5) and a very low correlation

141

between predicted and observed GOM (r = 0.07, p = 0.071). It was speculated that outliers in

142

certain parameters caused the extremely high predicted GOM. Therefore, data quality control

143

was applied on specific parameters, such as PBM and GOM concentrations and WHNO3. PBM

144

and GOM were constrained to concentrations ≥ 1 pg m-3, which is an estimate of the PBM and

145

GOM detection limits of the Tekran system. The exclusion of highly uncertain PBM and GOM

146

(i.e. low concentrations) would likely result in a more reliable NMB. The range in WHNO3 were

147

constrained between 372 and 6612 (mass basis), which represents the interquartile range (25th-


Page 8 of 30

Page 9 of 30


148

75th percentile) of the WHNO3 from our previous study of Canadian rural sites34. Limited field

149

data on scavenging ratios of HNO3 and other gases are available; therefore, the WHNO3 predicted

150

from our previous long-term multi-site study was used to constrain WHNO3 in this study. The

151

constraints on PBM and GOM concentrations had little effect on the NMB and correlation

152

between predicted and observed GOM. However, narrowing the WHNO3 range led to a more

153

reasonable NMB (Table S5) and correlation coefficient (r = 0.2, p

202

predicted GOM. The positive and negative biases suggest that the low GOM bias associated

203

with the Tekran instrument cannot be generalized to all locations and measurements. More field

204

intercomparison studies of different GOM measurement methods are needed to assess the spatial

205

variability of the GOM bias and confirm whether there are positive and/or negative biases.

206

3.2 Seasonal variation in bias

207

Aggregating all the sites together, BiasF of weekly GOM was 1.9 in winter, 1.8 in spring, 2.9 in

208

summer, and 2.0 in fall, and the corresponding method uncertainties were a factor of 1.7, 1.5,

209

1.01, and 1.3, respectively. These results indicate that the biases in GOM measurements were

210

the largest in summer (> factor of 2 after considering the method uncertainties) (Fig. 3). The

211

seasonal trend of the predicted GOM was similar to that of the observed GOM with the

212

exception of the higher predicted GOM during summer (Fig. 3). The larger bias in GOM during

213

summer may be attributed to ozone interferences on GOM measurements as reported in



214

experimental studies11, since ozone concentrations are frequently elevated in the summer (Fig.

215

3). However, the mechanisms causing the loss of GOM when KCl-coated denuders are exposed

216

to elevated ozone during sampling are not yet known.

217

218

3.3 The case of GA40 in southeastern U.S.

219

It is noted that 11 of the 13 sites have BiasF values smaller than 3.0 and one site (OK99) has a

220

value of 5.4. However, an extremely high BiasF value of 14.3 was found at GA40, a

221

southeastern U.S. site. This is because AMNet sites in the southeastern U.S observed much

222

higher mercury wet deposition, although similar ambient GOM concentration, than sites

223

elsewhere in the U.S.36-38. Elevated Hg wet deposition in southeastern U.S. and Puerto Rico has

224

been attributed to the efficient scavenging of GOM from the free troposphere by deep convective

225

storms38-41. Elevated GOM was indeed observed in the free troposphere by aircraft

226

measurements. 42-44 About 60% of the total Hg wet deposition was estimated to be from upper-

227

level (or in-cloud) scavenging in southeastern U.S. based on model simulations40,41. In this case,

228

the predicted GOM represents more of the upper altitude than surface concentration. For GA40,

229

if the total wet deposition was reduced by 60%, then the revised BiasF would be 5.7, which is

230

more in line with values at other sites. Thus, the discrepancy between the predicted and

231

observed GOM at GA40 should be a combination of the enhanced wet deposition due to deep

232

convection and the instrument measurement uncertainties, the latter was the focus of the present

233

study. Note that the exact contributions from these two sources to the total BiasF cannot be

234

quantified in this study.


Page 12 of 30

Page 13 of 30


235

The higher BiasF during summer than other seasons can also be explained by the higher

236

predicted wet deposition of GOM aside from elevated ozone during summer. The higher wet

237

scavenging of GOM during summer is not due to the higher precipitation amount, since it was

238

lower than the precipitation amount in spring (Fig. S2 of SI). It was due to the higher

239

precipitation Hg concentration during summer. Figure 4 illustrates the strong relationship

240

between the predicted- observed GOM difference and the precipitation Hg concentration. This

241

relationship can be used to predict the discrepancy between predicted and observed GOM given

242

the precipitation Hg concentration. With the precipitation Hg concentration < 20 ng l-1 at most

243

Mercury Deposition Network (MDN) locations in the U.S. and Canada36, the expected difference

244

between the predicted and observed GOM concentrations would be up to ~15 pg m-3. Based on

245

the geographical distribution of precipitation Hg concentrations36, this difference is expected to

246

be larger in southeastern U.S. which further supports the high BiasF at GA40.

247

3.4 Uncertainties of PBM measurements on GOM bias - a sensitivity analysis

248

The method of predicting GOM in this study relies on PBM (

Uncertainty Assessment of Gaseous Oxidized Mercury

Recommend Documents