A better understanding of the discrepancies in multi-scale inventories could
give an insight into their approaches and limitations as well as provide
indications for further improvements; international, national, and
plant-by-plant data are primarily obtained to compile those inventories. In
this study we develop a high-resolution inventory of Hg emissions at
0.05
Mercury (Hg), known as a global pollutant, has received increasing attention
for its toxicity and long-range transport. Identified as the most significant
release into the environment (Pirrone and Mason, 2009; AMAP/UNEP, 2013),
atmospheric Hg is analytically defined as gaseous elemental Hg (GEM,
Hg
Due mainly to the fast growth in economy and intensive use of fossil fuels, China has been indicated as the highest ranking nation in anthropogenic Hg emissions (Fu et al., 2012; Pacyna et al., 2010; Pirrone et al., 2010). Emissions of speciated atmospheric Hg of anthropogenic origin in China have been estimated at both global and national scales. For example, AMAP/UNEP (2013) and Muntean et al. (2014) developed global Hg inventories which reported national emissions for China for 2010 and from 1970 to 2008, respectively. At the national scale, Hg emissions have been estimated based on more detailed provincial information on energy consumption and industrial production. Zhang et al. (2015), Zhao et al. (2015a), and Tian et al. (2015) evaluated the inter-annual trends in emissions for 2000–2010, 2005–2012, and 1949–2012, respectively, to explore the benefits of air pollution control polices, particularly for recent years.
There are considerable information gaps between inventories, attributed mainly to the data of different sources and levels of details. For coal-fired power plants (CPP), as an example, the global inventories by AMAP/UNEP (2013) and Muntean et al. (2014) obtained the national coal consumption from the International Energy Agency (IEA), and they acquired the information on control technologies from the “national comments” by selected experts and World Electric Power Plants (WEPP) database, respectively. In the national inventories by Zhang et al. (2015) and Tian et al. (2015), coal consumption of CPP by province was derived from official energy statistics, and the penetrations of flue gas desulfurization (FGD) systems were assumed at provincial level. Zhao et al. (2015a) further analyzed the activity data and emission control levels plant by plant using a “unit-based” database of power sector. Although data of varied sources and levels of details result in discrepancies between inventories, those discrepancies and the underlying reasons have not been thoroughly analyzed in previous studies, leading to large uncertainty in Hg emission estimation.
Existing global and national inventories could hardly provide satisfying
estimates in speciated Hg emissions or well capture the spatial distribution
of emissions at regional/local scales, attributed mainly to relatively weak
investigation of individual sources. When they are used in a chemistry
transport model (CTM), downscaled inventories at global/national scales
would possibly bias the simulation at smaller scales. Improvement in
emission estimation at local scale, particularly for the large point sources,
is thus crucial for better understanding the atmospheric processes of Hg
(Lin et al., 2010; Wang et al., 2014; Zhu et al., 2015). While local
information based on sufficient surveys is proven to have advantages in
improving the emission estimates for given pollutants like NO
In this work, therefore, we select Jiangsu, one of the most developed provinces with serious air pollution in China, as study area. Firstly, we develop a high-resolution Hg emission inventory of anthropogenic origin for 2010, based on a comprehensive review of field measurements and detailed information on emission sources. That provincial inventory is then compared to selected global and national inventories with a thorough analysis on data and methods. Discrepancies in emission levels, speciation, and spatial distributions are evaluated and the underlying sources of the discrepancies are figured out. Finally, the uncertainty in the provincial emission inventory is quantified and the key parameters contributing to the uncertainty are identified. The results provide an insight into the effects of varied approaches and data on development of the Hg emission inventory and indicate the limitations of current studies and the orientations for further improvement on emission estimation at regional/local scales.
As shown in Fig. S1 in the Supplement, Jiangsu province (30
In global/national inventories, Hg emissions were first calculated by sector based on activity data and emission factors that were obtained or assumed at global, national, or provincial level, and were then downscaled to regional domain with finer spatial resolution. Various methods and data were adopted in multi-scale inventories to estimate Hg emissions for different sectors, as summarized briefly in Table S1 in the Supplement. Three national inventories were developed by Nanjing University (NJU; Zhao et al., 2015a), Beijing Normal University (BNU; Tian et al., 2015), and Tsinghua University (THU; Zhang et al., 2015), with major activity data at provincial level obtained from Chinese national official statistics. Compared to the NJU and BNU inventories, which applied deterministic parameters relevant to emission factors, THU developed a model with probabilistic technology-based emission factors to calculate the emissions. Based on international activity statistics at national level, two global inventories for 2010 were developed by the joint expert group of the Arctic Monitoring and Assessment Programme and United Nations Environment Programme (AMAP/UNEP, 2013) and the Emission Database for Global Atmospheric Research (EDGARv4.tox2, unpublished). The AMAP/UNEP inventory developed a new system for estimating emissions from main sectors based on a mass-balance approach with data on unabated emission factors and emission reduction technology employed in different countries. The EDGARv4.tox2 inventory calculated the emissions for all the countries by primarily applying emission factors from EEA (2009) and USEPA (2012), combined with regional technology-specific information on emission abatement measures.
In contrast to the downscaling approach, the emissions are calculated plant by plant based on information on individual sources and then aggregated to provincial level in a bottom-up method. We refer to the inventory as the bottom-up or provincial inventory hereinafter. Information for individual emission sources are thoroughly obtained from the Pollution Source Census (PSC, internal data from the Environmental Protection Agency of Jiangsu Province). The PSC was conducted by local environmental protection agencies, in which the data for individual sources were collected and compiled through on-site investigation, including manufacturing technology, production level, energy consumption, fuel quality, and emission control devices. Differences in total energy consumption and industrial production levels exist between the PSC data and the energy/economic statistics. For example, the coal consumption by power plants in the PSC was 6 % larger than the provincial statistics for Jiangsu in 2010. We believe that, compared to the energy and economic statistics that were commonly used in global/national inventories, the plant-by-plant PSC data could provide more detailed and accurate information on specific emitters, particularly for power and industrial plants.
According to the availability of data, anthropogenic sources are classified into three main categories. Category 1 includes coal-fired power plants (CPP), iron and steel plants (ISP), cement production (CEM), and other industrial coal combustion (OIB). Note that the emissions from coal combustion in cement production are not included in CEM but in OIB, following most other inventories included in this paper for easier comparison. The information on geographic location, activity levels (consumption of energy or raw materials), and penetration of air pollution control devices (APCDs) is compiled plant by plant from the PSC, with an exception that the technology employed in CEM is obtained from CCA (2011). Category 2 includes nonferrous metal smelting (NMS), aluminum production (AP), municipal solid waste incineration (MSWI), and intentional use sector (IUS: thermometer, fluorescent lamp, battery, and polyvinyl chloride polymer production). Geographic location information for those sources is obtained from the PSC, while other activity data come from official statistics at provincial level. Category 3 includes emission sources that are not contained in the PSC: residential and commercial coal combustion (RCC), oil and gas combustion (O&G), biofuel use/biomass open burning (BIO), rural solid waste incineration (RSWI), and human cremation (HC). They are defined as area sources, and the data sources for them are discussed later in this section.
In general, annual emissions of total and speciated Hg are calculated using
Eqs. (1) and (2), respectively:
For CPP/OIB and CEM, Eq. (1) can be revised to Eqs. (3) and (4), respectively,
with detailed fuel and technology information on individual sources
incorporated:
For ISP, Eq. (1) could be revised to Eq. (5):
Activity data for NMS, AP, MSWI, RCC, and O&G are derived from national
statistics (NMIA, 2011; NSB, 2011a, b), while Hg consumption in IUS is
estimated based on the internal association commercial reports that provide
national market and economy information collected at
Regarding the spatial pattern of emissions, the study domain is divided into
4212 grid cells with a resolution at 0.05
For better understanding the sources of discrepancies between inventories, a
comprehensive sensitivity analysis is conducted to quantify the differences
between selected parameters used in multi-scale inventories and the
subsequent changes in emission estimation for Category 1 sources. The
relative change (RC) in a given parameter (
In particular, a new parameter, total abatement rate (TA), is defined for the
sensitivity analysis, combining the effect of the penetrations of APCDs and
their removal efficiencies on emission abatement:
The uncertainties in speciated Hg emissions at provincial level are
quantified using a Monte Carlo framework (Zhao et al., 2011). Given the
relatively accurate data reported in the PSC, the probability distributions of
activity levels for individual plants of CPP, OIB, ISP, and CEM are defined
as normal distributions with the relative standard deviations set at
10, 20, 20, and 20 % respectively. As summarized in Tables S6
and S7 in the Supplement, a database for Hg emission factors/related
parameters by sector and speciation is established for China, with the
uncertainties analyzed and indicated by probability distribution function
(PDF). The PDFs of Hg contents in coal mines by province are obtained from
Zhang et al. (2015). For Hg content in limestone (HgC
Table 1 provides the Hg emissions by sector and species for Jiangsu in 2010
estimated from the bottom-up approach. The provincial total Hg emissions of
anthropogenic origin are calculated to be 39 105 kg, of which 51 % is
released as Hg
Emission estimates for Jiangsu in 2010 and species from multi-scale inventories by sector. Abbreviations for emission sources, as in Sect. 2 – CPP: coal-fired power plants; RCC: residential coal combustion; O&G: oil and gas combustion; OIB: other industrial coal combustion; CEM: cement production; ISP: iron and steel plants; NMS: nonferrous metal smelting; AP: aluminum production; LGM: large-scale gold mining; MM: mercury mining; HC: human cremation; MSWI: municipal solid waste incineration; RSWI: rural solid waste incineration; BFLP: battery/fluorescent lamp production; BIO: biofuel use/biomass open burning; and PVC: PVC production.
To better understand the discrepancies and their sources between various
studies, the emissions from multi-scale inventories are summarized in Table 1 for comparison. Among all the inventories, the total emissions in the
provincial inventory are the largest, i.e., 28, 7, 19, 22,
and 70 % higher than NJU, THU, BNU, AMAP/UNEP, and EDGARv4.tox2,
respectively. The elevated Hg emissions compared to previous studies could
be supported by modeling and observation work to some extent. Based on the
chemistry transport modeling using GEOS-Chem (Wang et al., 2014), or
correlation slopes with certain tracers (CO, CO
Direct comparison between inventories is unavailable for every sector, as the definition of source categories is not fully consistent with each other. Therefore, necessary assumption and modification are made on source classification for global inventories. In Table 1, CPP, OIB, and RCC for EDGARv4.tox2 represent the emissions for all of the fossil fuel types, and they are 1316 and 5342 kg lower and 986 kg higher than our estimation from coal combustion, respectively. For AMAP/UNEP, the emissions from regrouped stationary combustion (industrial sources excluded), industry, and intentional use and product-waste-associated sources (see Table 1 for the detailed definition) are respectively 3382 and 2032 kg higher and 3118 kg lower than our estimation with the bottom-up method. Figure 1 shows the ratios of the estimated Hg emissions in national/global inventories to those in the provincial inventory by source. The CPP emissions are relatively close to each other, but larger differences exist in some other sources. The estimates for CEM and ISP in the provincial inventory are much higher than the NJU, BNU, and EDGARv4.tox2 inventories, while those for NMS are considerably smaller. The reasons for those differences are analyzed in detail in Sect. 3.1.2–3.1.4.
Figure 2a and b represent the relative changes in given parameters between the provincial and other inventories, and the subsequent differences in Hg emissions for Category 1 sources, using Eqs. (6) and (7), respectively. For CPP, the differences between provincial and national/global inventories are mainly determined by AL, HgC, TA, and IEF (integrated emission factors), as indicated by the calculation methods summarized in Table S1. (Instead of analyzing HgC and RR separately, integrated input emission factors were applied in AMAP/UNEP and EDGARv4.tox2.) For activity level (AL), the coal consumption data are collected and compiled plant by plant in the provincial inventory, while they were obtained from Chinese official statistics (NSB, 2011b) in national inventories. As a result, the coal consumptions in the NJU and THU inventories are 17 and 6 % smaller than our provincial inventory, resulting in 1968 and 760 kg reduction in Hg emission estimate, respectively.
The ratios of estimated Hg emissions for Jiangsu and 2010 in global/national inventories to that in the provincial inventory for selected sources and anthropogenic total.
Sensitivity analysis of selected parameters in Hg emission
estimation for Category 1 sources.
In national and provincial inventories, as mentioned in Sect. 2, the Hg
contents in the raw coal (HgC
Total abatement rate (TA) of APCDs installed for CPP is calculated at
57 % in the provincial inventory, 6.7 and 8.2 % smaller than that in
the THU and AMAP/UNEP inventories, respectively, and 12 % larger than
that in the NJU inventory. The differences result mainly from the varied
removal efficiencies (RE) and application ratios (AR), as shown in
Table S5. For RE, local tests on FF, ESP
For OIB, the comparison of HgC is similar to that for CPP. AL from the PSC in the provincial inventory is very close to that in the THU inventory obtained from NSB (2011b), while AL in the NJU inventory was much lower as the coal consumption of CEM and ISP was excluded. The RR from industrial boilers in this work is estimated at 82 % based on domestic measurements (Wang et al., 2000; Tang et al., 2004), much lower than the result in the THU inventory measured by Zhang et al. (2012), i.e., 95 % for stoker-fired boilers. Given the limited samples in both inventories, large uncertainty exists in RR of industrial boilers. Compared to the provincial inventory, ARs of ESP and FGD were clearly underestimated in the NJU and THU inventories (Table S5); hence, the TA in NJU was calculated 23 % smaller than that in the provincial inventory, leading to a 747 kg increase in the Hg emission estimate. In the THU inventory, however, the much higher RE of WET reduced the difference between national and provincial inventories, and TA in the THU inventory was only 2 % smaller than the provincial one.
For CEM, both the provincial and THU inventories adopted the data from Yang (2014), who measured provincial Hg contents in raw materials (limestone and
other raw materials) and Hg removal efficiency of DPT
For ISP, difficulty exists in emission estimation due to various Hg
input sources and complex production processes, and there is no consistent
method in multi-scale inventories so far. It was found that raw material
production (limestone and dolomite), coking, sintering, and pig iron smelting
with blast furnace account for most Hg emissions in typical ISPs in China
(Wang et al., 2016). In our study, 11 factories containing those processes
are collected in the PSC, and the emissions factors of 0.043 and
0.068 g t
In general, the detailed activity and technology information including manufacturing procedures and APCDs was investigated for individual plants in our provincial inventory to improve the emission estimation, in contrast to previous inventories that applied simplified or regional-average data. However, some crucial parameters (e.g., Hg contents in coal and limestone, and Hg removal efficiencies of APCDs) are still unavailable at plant level due to a lack of measurements. Such a limitation indicates the necessity of more efforts on plant-specific emission factors and also motivates the uncertainty analysis for the provincial inventory, as presented in Sect. 3.4.
For categories 2 and 3, differences also exist in EF and AL between
inventories. For example, an emission factor of 0.22 g t
Hg speciation profiles by sector and the mass fractions to total emissions in multi-scale inventories (%).
Besides the total emissions, Hg speciation has a significant impact on the distance of Hg transport and chemical behaviors. Table 2 summarizes the mass fractions of Hg species in emissions by sector for multi-scale inventories.
In general, as shown in Table 2, reduced Hg
Hg speciation profiles used in provincial and national inventories for typical APCDs (%).
Spatial distribution of Hg emissions for Jiangsu 2010 at a
resolution of 0.05
As mentioned above, the “universal” profiles were applied for many sectors
in the AMAP/UNEP inventory, ignoring the effects of various types of APCDs on Hg
speciation, particularly for coal combustion. However, the fate of Hg
released to atmosphere can primarily be affected by the removal mechanisms of
APCDs. As shown in Table 3, for example, Hg
Figure 3 presents the spatial distributions of total and speciated Hg
emissions in Jiangsu province at 0.05
Differences in gridded Hg
In order to compare the spatial distribution of the provincial inventory to that
of the NJU, THU, AMAP/UNEP, and EDGARv4.tox2 inventories, we upscale the gridded
provincial emissions from 0.05
As illustrated in Fig. 4, differences in gridded emissions between provincial
and other inventories (NJU, THU, AMAP/UNEP, and
EDGARv4.tox2) are respectively in the ranges of
With much fewer large emitters, discrepancies in gridded emissions for another part of Jiangsu resulted largely from the allocation of emissions as area sources in national and global inventories. For example, in spite of an estimation of 8496 kg smaller than the provincial inventory in total emissions, the NJU inventory applied proxies (e.g., population and GDP) to allocate the emissions except those from CPP, resulting in higher emissions in central and most part of northern Jiangsu (Fig. 4). Similar patterns are also found for THU (Fig. 4b) and AMAP/UNEP (Fig. 4c) compared to the provincial inventory.
Besides the total emissions, differences in spatial distribution of speciated
Hg emissions between multi-scale inventories are presented in Fig. S3 in the
Supplement. The various patterns for species are largely influenced by the
distribution of different types of large point sources, as the speciation
profiles vary significantly between source types in the national and
provincial inventories (Table 2). Compared to other inventories, larger
Hg
The vertical distribution of Hg releases, which is crucial for the transport
range of atmospheric Hg, is also analyzed. Four groups of release height are
defined: 0–58, 58–141, 141–250, and > 250 m. Based on the
detailed information on emission sources, the fractions of Hg releases into
the four groups for CPP are 2, 66, 31, and 1 %, respectively,
and the analogue numbers for OIB, ISP, and CEM are 85, 13, 2,
and 0; 4, 44, 12, and 4 %; and 6, 94, 0, and
0 %, respectively. The release heights for rest sources are uniformly
assumed to be in the range of 0–58 m. As a result, the fractions of total Hg
emissions in the four groups are estimated as 35, 53, 11, and
1 %. In the AMAP/UNEP inventory, as a comparison, the fractions at the height
of 0–50, 50–150, and > 150 m were estimated at 23, 53, and
24 % respectively, with a larger share in Hg emitted over 150 m than that in
our provincial inventory. The smaller fraction of Hg emissions under 150 m
and larger fraction of Hg
Uncertainties in Hg emissions in Jiangsu in provincial and national (NJU) inventories by source, expressed as the 95 % confidence intervals of central estimates.
As summarized in Table 4, the uncertainties in speciated Hg emissions in the
provincial inventory are estimated at
The parameters contributing most to uncertainties and their contributions to
the variance of corresponding emission estimates are summarized by sector in
Table S8 in the Supplement. For CPP and OIB, parameters related to emission
factors contribute most to the uncertainties in Hg
In most cases, parameters with largest contribution to uncertainty in Hg
Taking Jiangsu province in China as an example, we have thoroughly analyzed the discrepancies and their
sources of atmospheric Hg emission estimations in multi-scale inventories,
applying various methods and data. Using a bottom-up
approach that integrates the best available information on individual plants and
most recent field measurements, we find that the total Hg emissions in Jiangsu in 2010 are
estimated larger than any other national/global inventories. CPP, ISP, CEM,
and OIB collectively accounted for 90 % of the total emissions.
Comparisons between available studies demonstrate that the information gaps
of multi-scale inventories lead to large differences in Hg emission
estimation. Discrepancies in emissions between inventories for the
above-mentioned major sources come primarily from various data sources for
activity levels, Hg contents in coals, and total abatement effects of APCDs.
A notable increase in Hg
The method developed and demonstrated for Jiangsu could potentially be applied to other provinces, particularly for those with intensive industrial plants. As estimated in this work, for example, cement and iron and steel industries were the two most important sectors in which the Hg emissions were significantly underestimated by previous inventories for Jiangsu. The underestimations came mainly from ignoring the high Hg release ratio of precalciner technology with dust recycling, as well as from the application of relatively low emission factors for steel production. We could thus cautiously infer that Hg emissions might be underestimated for China's other regions with intensive cement and steel industries in previous inventories. For power plants and industrial boilers, however, the Hg emissions were influenced largely by Hg contents in coal and APCD application. Whether the emissions of those sources were underestimated or not for other parts of the country could hardly be judged unless detailed information becomes available for the regions. Extensive and dedicated measurements are urgently required for Hg contents in coal/limestone and removal efficiency of dominant APCDs to further improve the emission estimation at regional/local scales and, eventually, for the whole country.
The gridded Hg emissions for Jiangsu province 2010 at a horizontal resolution
of 0.05
This work was sponsored by the Natural Science Foundation of China (91644220 and 41575142), the Natural Science Foundation of Jiangsu (BK20140020), the Ministry of Science and Technology of China (2016YFC0201507), the Jiangsu Science and Technology Support Program (SBE2014070918), and the Special Research Program of Environmental Protection for Commonweal (201509004). We would like to acknowledge Hezhong Tian from Beijing Normal University and Simon Wilson from the UNEP/AMAP Expert Group for the detailed information on national/global Hg emission inventories. Thanks also go to the two anonymous reviewers for their very valuable comments to improve this work. Edited by: L. Zhang Reviewed by: two anonymous referees