Introduction
Emissions from biomass burning (BB) are an important source of mercury (Hg)
to the atmosphere and a major factor in
determining the interannual variations of its tropospheric concentration
. Although the Hg released by BB varies from year to year,
it can amount to up to roughly one third of the anthropogenic emission
estimates . With the eventual
implementation of the Minamata Convention
(http://www.mercuryconvention.org/) and future curbs on industrial
emission, as a by-product of industrial emission abatement measures, its
relative importance will increase in the coming years. A previous modelling
study used the global Hg chemistry model, ECHMERIT, and
three BB inventories to assess the distribution of Hg deposition resulting
from BB. A large part of the Hg released from BB deposits over oceans, where
its re-emission is driven by sea surface temperature, among other factors
, or where it can be converted to toxic
methyl mercury (MeHg) compounds, has important implications for the
food web and, through fish consumption, also for human health (see
, and references therein). The deposition flux of Hg from
BB has been shown to be more sensitive to certain factors, in particular the
chemical mechanism employed in the model and the choice of emission
inventory, than to others such as the vertical profiles of emissions
. In this previous study all Hg emitted from BB was
considered to be Hg(g)0. There is, however, evidence that the
fraction of Hg emitted bound to particulates (HgP) may be sizeable,
up to 30 %, especially when the fuel moisture content (FMC) is high
. These levels, however,
remain uncertain since different methodologies have led to different
conclusions . Little is known about the
mechanisms that control the speciation of Hg in BB emissions, which leads to
uncertainties in the Hg deposition patterns, since the atmospheric lifetime
of HgP is significantly shorter than Hg(g)0, leading
to greater local deposition.
Local Hg deposition due to BB could have important repercussions in regions
such as the South-East Asia, where there is intensive rice cultivation,
which is subject to major BB events, especially during El Niño periods.
Hg deposited to rice paddies can be readily converted to toxic MeHg
that can accumulate in the grains
. Moreover, it has been reported
that HgP from BB deposited to foliage has the ability to enhance
MeHg formation .
The aim of this study is to investigate the effects on simulated deposition
fluxes of Hg resulting from BB when variations in HgP fraction and
production processes are considered. The most recent version of the GFED BB
emission inventory , has been
included in the global online Hg chemical transport model ECHMERIT to
simulate Hg deposition from BB for the year 2013 and to quantify the
influence of variations in model inputs, assumptions and parametrisations.
Methods
The biomass burning inventory
The reference BB inventory in this study, Global Fire Emissions Database
version 4 (GFED4.1s), is based on an updated version of the inventory of
with burned area from , and with the
addition of small fire-burned area .
The standard temporal resolution of the emissions files is monthly, but
data are provided to distribute these daily, and a diurnal cycle based on
is also available. Daily BB emissions from two other global
inventories, GFASv1.2 and FINNv1.5
, were also included in the model for sensitivity runs.
These three inventories are all compiled using the imagery obtained from the
MODIS instruments. However, the way in which the data are filtered or
processed yields substantial differences between the final products; see
and references therein for a detailed description of the
differences among the inventories.
Experimental set-up
The global Hg chemical transport model ECHMERIT uses T42 horizontal resolution (roughly 2.8∘ by
2.8∘ at the Equator) and 19 vertical levels up to 10 hPa.
Hg emissions from BB were included in the model by mapping them to CO
emissions using the global averaged enhancement ratio (ER) of
1.96×10-7, as obtained by , averaging field
measurements from different biomes in various regions around the globe,
including in plume measurements from the CARIBIC project
. Previous modelling studies have used different ERs
, but all these values were well within
the range of uncertainty (0.3–6.0×10-7; see ).
ECHMERIT, in the base configuration, includes the oxidation of
Hg(g)0 to in Hg(g/aq)II oxidation by
O3 / OH in the gas and aqueous phases. OH and O3
concentration fields were imported from MOZART (Model for Ozone and Related
chemical Tracers) . HgP is assumed to be inert,
whether it is emitted from anthropogenic activities or BB, and it is subject
to transport and deposition processes but is not involved in any chemical
reactions. Mechanisms and parametrisations used for calculating the dry and
the wet deposition of the different Hg species are the same as described in
. Beyond this standard configuration a number of alternative
processes and chemical mechanisms have been considered for this study, as
explained in Sect. . Atmospheric reduction of
Hg(g/aq)II to Hg(g)0 has been included in many
models to regulate the residence time of Hg(g)0 in the
atmosphere. However, a number of the proposed mechanisms are unlikely to
occur under most atmospheric conditions or are based on empirical rates to
better match the observations (see for a recent review). Due
to this uncertainty, reduction was not included in this study.
No further HgP particulate matter (PM) dimension distributions
other than the standard log-normal particle size distribution, as described
in detail in , were considered in this study due to large
uncertainties regarding the dynamic size range of PM emitted during BB (see
and references therein). GFED4.1s provides monthly burned
area, fire carbon (C) and dry matter (DM) emissions
(http://www.falw.vu/~gwerf/GFED/GFED4/).
A script is provided to derive gaseous and PM emissions from DM fields making
use of biome-based emission factors based on and
. The resulting emission fields were then interpolated
on to the ECHMERIT T42 grid using the mass conserving remapping function
included in the Climate Data Operators
(https://code.zmaw.de/projects/cdo).
Simulations performed.
Name
Inventory (BB emission (Mg))
Full version
Emiss. time res.
Fraction HgP (%)
Map HgP
Chem. mech.
Vertical profile
Scope
BASE
GFED4.1s (390)
Yes
daily
15
CO
O3 / OH
PBL
Reference
3-hourly
GFED4.1s (390)
3 h
15
CO
O3 / OH
PBL
Emiss. time resol.
Monthly
GFED4.1s (390)
monthly
15
CO
O3 / OH
PBL
Emiss. time resol.
HAM-Profile
GFED4.1s (390)
daily
15
CO
O3 / OH
HAM
Vertical profile
Only first level
GFED4.1s (390)
daily
15
CO
O3 / OH
1st
Vertical profile
Only PBL level
GFED4.1s (390)
daily
15
CO
O3 / OH
level of PBL
Vertical profile
3 h + HAM-prof
GFED4.1s (390)
daily
15
CO
O3 / OH
HAM
V. Pr. & E. T. res.
HgP to PM
GFED4.1s (390)
Yes
daily
15
PM
O3 / OH
PBL
HgP mapping
HgP to OC
GFED4.1s (390)
Yes
daily
15
OC
O3 / OH
PBL
HgP mapping
HgP to FMC
GFED4.1s (390)
Yes
daily
variable
CO
O3 / OH
PBL
HgP mapping
NO HgP
GFED4.1s (390)
Yes
daily
0
NA
O3 / OH
PBL
Fraction HgP
4 % HgP
GFED4.1s (390)
daily
4
CO
O3 / OH
PBL
Fraction HgP
30 % HgP
GFED4.1s (390)
Yes
daily
30
CO
O3 / OH
PBL
Fraction HgP
100 % HgP
GFED4.1s (390)
daily
100
CO
None
PBL
Transport HgP
Partitioning
GFED4.1s (390)
Yes
daily
15
CO
O3 / OH
PBL
Partitioning HgP/II
Partitioning ref.
GFED4.1s (390)
Yes
daily
0
CO
O3 / OH
PBL
Partitioning HgP/II
Reduction
GFED4.1s (390)
Yes
daily
15
CO
O3 / OH + Red.
PBL
Chemistry
Br
GFED4.1s (390)
Yes
daily
15
CO
Br
PBL
Chemistry
Br No HgP
GFED4.1s (390)
daily
0
NA
Br
PBL
Chemistry
Br 30 % HgP
GFED4.1s (390)
daily
30
CO
Br
PBL
Chemistry
Br HgP to OC
GFED4.1s (390)
daily
15
OC
Br
PBL
Chemistry
Br HgP to FMC
GFED4.1s (390)
daily
variable
CO
Br
PBL
Chemistry
GFAS
GFASv1.2 (150; see Sect. 2.3)
daily
15
CO
O3 / OH
PBL
Inventory
GFAS Br
GFASv1.2 (150; see Sect. 2.3)
daily
15
CO
Br
PBL
Chemistry
FINN
FINNv1.5 (550)
Yes
daily
15
CO
O3 / OH
PBL
Inventory
FINN Br
FINNv1.5 (550)
daily
15
CO
Br
PBL
Chemistry
AMAPOH
AMAP2010
NA
NA
NA
O3 / OH
NA
Ratio to anth. emiss.
AMAPBr
AMAP2010
NA
NA
NA
Br
NA
Ratio to anth. emiss.
EDGAROH
EDGAR2008
NA
NA
NA
O3 / OH
NA
Ratio to anth. emiss.
EDGARBr
EDGAR2008
NA
NA
NA
Br
NA
Ratio to anth. emiss.
STREETSOH
STREETS2005
NA
NA
NA
O3 / OH
NA
Ratio to anth. emiss.
STREETSBr
STREETS2005
NA
NA
NA
Br
NA
Ratio to anth. emiss.
Simulations and their scope
The BASE simulation used as the reference case in this study includes
daily BB emissions from GFEDv4.1s, in which a global uniform fraction of
HgP, equal to 15 % of the total Hg emission is assumed. This
value is within the range of observations .
However, since there are uncertainties in the proportion of HgP
emitted from BB , further simulations were carried out with
varying fractions of HgP (0, 4 and 30 %). Simulations were also
conducted mapping the 15 % of the total Hg emitted as HgP to
the geographical distribution of different proxy chemical species (see
Sect. ).
The shorter lifetime of HgP with respect to Hg(g)0
potentially means that the vertical profile of the emissions could have an
impact on the distribution of Hg deposition, as is the case for other
speciated Hg emission sources . Therefore two vertical
profile parametrisations, as well as different emission injection time resolutions,
were also included in the study. The principal vertical profile used
(PBL-Profile) maps the Hg emissions uniformly within the planetary boundary
layer (PBL), whereas in the second the vertical profile of the standard
version of the ECHAM-HAM model was used (HAM-Profile) .
The HAM-Profile is equal to PBL-Profile when the PBL height is greater than
4000 m; otherwise 75 % of the emissions are placed within the PBL and
the remainder in the two layers above the PBL (17 and 8 %). This
threshold value is arbitrary, but it is the standard configuration of
ECHAM6-HAM2 . Biomass burning emissions from
GFASv1.2 and FINNv1.5
were also used in the study to assess uncertainty related to the satellite
imagery processing and inventory compilation. Simulations using GFASv1.2 were excluded from suqsequent analyses since the low Hg
emissions could be due to a technical problem arising from GRIB encoding (see ).
These simulations primarily employ a O3 / OH
Hg(g)0 oxidation mechanism. However, since the precise
atmospheric Hg oxidation mechanism remains unclear
, a
number of runs were performed using a Br-based oxidation mechanism.
Some studies suggest that the partitioning of
reactive Hg species the between gas and particulate phases might be driven by
air temperature and on the surface are of the aerosol present in the
atmosphere. Therefore, two other simulations weer conducted including the
temperature-dependent gas-particle partitioning described in
, one assuming BB Hg emissions to be only
Hg(g)0 and another assuming a 15 % of BB Hg emissions to be
HgP.
To estimate the ratio of Hg deposition from BB compared to anthropogenic
sources, six further simulations were conducted including only anthropogenic
emissions using the EDGAR , AMAP2010
and STREETS inventories, employing the
O3 / OH and Br oxidation mechanisms.
This study covers a single year, 2013, chosen due to the availability of
measurements from GMOS network
. All simulations were
performed for a full year, without the rapid re-emission mechanism
, and were continued without further emissions for another
12 months to allow most of the 2013 Hg emissions to be deposited.
Finally, a selection of simulations were re-run including Hg emissions from
all sources, BB, anthropogenic emissions from AMAP2010 ,
dynamic ocean emissions, terrestrial emissions and re-emissions as described
in , to evaluate model performance against measurements
and to evaluate the assumptions made in this study.
A summary of the simulations performed can be found in
Table .
BB emission speciation
The release of Hg from BB occurs prevalently as Hg(g)0. However,
as mentioned previously, a measurable fraction may be emitted as HgP
. No significant amounts
of gaseous oxidised Hg (Hg(g)II) have so far been detected in BB
emissions (, and references therein). The speciation of
Hg emissions is of great importance, since it largely determines the
atmospheric lifetime and hence the distance emitted Hg is transported in the
atmosphere before deposition, as seen for other speciated Hg sources
.
The fraction of HgP released by BB determined in field and
laboratory studies ranges from fractions of a few percent to over 30 %
. The factors determining speciation and whether
HgP is directly emitted or if it is the product of the oxidation of
Hg(g)0 within the plume are not
known. However, foliage, moisture content, fuel type, plant species and
combustion proprieties certainly play a role. HgP emissions were
found to be well correlated with particulate matter (PM) and organic
carbon (OC) emissions . found
that Hg(g)0 is the dominant species in dry fuel combustion,
whereas the fraction of HgP becomes appreciable when FMC reaches
roughly 30 %, above which HgP release appears to increase
linearly with FMC. In the inventory used for the BASE case both
Hg(g)0 and HgP follow the spatial distribution of
CO emissions from BB, and 15 % of the emitted Hg is considered to
be HgP (see Figs. a and a). Hg
emission fields were also compiled in which the HgP fraction of the
total Hg emitted was mapped to OC and PM emissions (see
Fig. b and c). A further emission field was compiled in which
the ratio of Hg(g)0 to HgP is determined by the FMC (Figs. b and d).
A relationship was found to exist between HgP emissions and the fire
burn duration and severity as well as combustion conditions .
In particular high HgP fractions were observed during smouldering phases,
whereas very low or undetectable HgP levels were found during flaming
combustion.
These potential parametrisations were not investigated here due to
the difficulty in finding a suitable proxy data set.
Appendix contains a more detailed description of the methods used
to calculate the different Hg BB emission fields.
Geographical distribution (a–b) and PBL-type vertical
profiles (c–d) of the Hg(g)0 emissions, when mapped to
CO (a, c) and when speciation is determined by
FMC (b, d). For the emissions mapped to CO, only the
speciation (15 : 85 HgP : Hg(g)0) is shown for
clarity.
Geographical distribution (a–d) and PBL-type vertical
profiles (e–h) of the HgP emissions as injected in the
model, when mapped to CO (a, e), PM (b, f)
and OC (c, g) and when speciation is determined by
FMC (d, h). For the emissions mapped to CO, only the
speciation (15 : 85 HgP : Hg(g)0) is shown for
clarity.
Results
Emissions
The total Hg emitted in 2013 based on the GFED inventory is roughly
400 Mg, which is at the lower end of the initial estimates (675±240 Mg)
but is reasonable considering the natural variation of BB
activity and the diminishing trend of the CO emission estimates in the
latest inventory revisions (up to 50 % for some years)
. Considering 15 % of the emissions to be
HgP, in the BASE run this corresponds to approximately 340 Mg
Hg(g)0 and 60 Mg HgP. Interestingly the emissions
of HgP amount to 58 Mg when relating the HgP fraction
to FMC. The exact amount of Hg emitted by BB in the different model runs is
detailed in Table . The spatial distribution and the
vertical profile of the emission injection height, considering the
PBL-Profile for Hg(g)0 and HgP in the different cases
considered are shown in Figs. and . Both the
geographical and vertical distributions of the emissions of the Hg species
reveal notable differences depending on the methodology used, particularly
for HgP. Compared to the cases where HgP emissions are
mapped to CO and PM (Fig. a–b and e–f),
mapping HgP to OC and using the FMC to determine the
speciation (Fig. c–d and g–h) result in enhanced
HgP emissions, above 60∘ N and over some areas the Amazon,
central Africa and East Asia as evident in Fig. . The
timing and location of the enhanced HgP emission at northerly
latitudes could be particularly relevant for Hg deposition to the Arctic.
From Fig. it is evident how the geographical distribution
of the HgP to Hg(g)0 emission ratio differs with the
assumptions considered. However, for OC and FMC there is general
agreement on the areas where the HgP emissions are relatively
higher, especially in the Northern Hemisphere and particularly for areas above
60∘ N. The agreement between OC and FMC is not surprising and
is related to the combustion characteristics that enhance OC
emissions, i.e. lower combustion temperatures and the dominance of the
smouldering phase of combustion , that are likely to occur
where FMC is greatest.
Geographical distribution of the
HgP : Hg(g)0 emissions ratio, when mapped to
PM (a) and OC (b) and when speciation is
determined by FMC (c). In the colour bar the levels
corresponding to the constant speciations (4 : 96, 15 : 85 and 30 : 70
HgP : Hg(g)0) used for the emissions mapped to
CO are indicated.
Emission latitudinal profiles
The latitudinal profiles of Hg(g)0 and HgP emissions,
using the different approaches (Sect. ), are
shown in Fig. a and b. For those emissions mapped to
CO, only the 15 : 85 (HgP : Hg(g)0)
speciation is reported for clarity. The differences in the latitudinal
profiles of the Hg(g)0 emissions (Fig. a) are
sizeable only for the peaks north of 45∘ N, where the FMC-based
speciation has an Hg(g)0 fraction below 85 %. The latitudinal
profiles of HgP emissions mapped to PM and CO look
very similar over the entire domain (Fig. b), apart from a
peak a few degrees north of the Equator. The HgP emissions mapped
to OC and FMC differ from the PM and CO profiles but
are similar to each other between roughly 30∘ S and 60∘ N.
South of 30∘ S HgP emissions mapped to OC are
higher, while peak HgP emissions derived from FMC at 65∘ N
(1.5 g km-2 yr-1) are nearly 30 % greater than those derived
from OC and roughly double those mapped to CO and PM.
Moreover, in the FMC scenario the peak in HgP emissions at
65∘ N are greater than the peak seen at 15∘ S (1.5 vs.
1.4 g km-2 yr-1). As is particularly evident in
Fig. c, the most notable differences among the different
assumptions hypothesised are above 60∘ N, where both the OC
and the FMC cases agree on the location of the greatest HgP
emissions probably due to the linkage between OC emissions and
combustion processes favoured by FMC , and between 30 and
45∘ S, where only OC and PM are greater than BASE.
A previous modelling study focusing on the fate of Hg from BB, where all
emissions were considered as Hg(g)0, showed that the long
atmospheric life of the elemental Hg smoothed the deposition latitudinal
profiles compared to the emission profiles . The four
panels in Fig. compare the normalised latitudinal
deposition profiles obtained for the BASE simulation with those obtained
from the alternative HgP emission scenarios by category.
Figure a demonstrates the very limited impact of the time
resolution used for BB emissions, most likely due to the coarse horizontal
resolution of the model.
The two vertical emission profiles (Fig. b) give deposition
fields that are to all effects indistinguishable, even when considering
varying temporal resolution of the BB emissions, whereas assuming all
emissions to be in the first model level (with an average height of
approximately 35 m) leads to enhanced deposition near emission peaks. In
this instance, the maximum deposition coincides with peak emission, at
approximately 15∘ S, whereas in all other cases maximum deposition
is shifted towards the Equator.
Latitudinal profiles of (a) Hg(g)0 emissions
when mapped to CO and when speciation is determined by FMC;
(b) HgP emissions when mapped to CO, PM and
OC and when speciation is driven by FMC; and
(c) the relevant ratio HgP : Hg(g)0. For
both Hg(g)0 and HgP emissions mapped to CO,
only the speciation (15 : 85 HgP : Hg(g)0) is
reported for clarity, whereas in (c) all the speciations are
reported.
Latitudinal profiles of the normalised Hg total deposition from the
model BASE run, compared with a selection of sensitivity runs, assuming
(a–b) different emission time resolution and vertical profile, as
well as
a combination of both; (c) different HgP emission
geographical distributions, as well as different
Hg(g)0 : HgP ratios. The normalisation was done by
maximum.
The similarities in the latitudinal profiles of HgP emissions when
mapped to CO and PM are reflected in their deposition profiles
(Fig. c). The relatively greater deposition north of
60∘ N seen in Fig. c, obtained when HgP
emissions are mapped to OC and when driven by FMC, reflects the peak in
HgP emissions at this latitude.
The greatest differences in the latitudinal deposition profiles, using the
GFED inventory, are seen when varying the percentage of HgP in the
emissions (Fig. d). Considering emissions to be solely
Hg(g)0 yields a relatively smooth profile extending from pole to
pole, increasing HgP causes enhanced deposition near BB hotspots.
The emission peak at around 50∘ N remains relatively distinct also
in the deposition for all the simulations (although it seen as a shoulder in
the 100 % Hg(g)0 profile). The peak north of 60∘ N
is more dependent on emission speciation, supporting the previous finding
that the location of Hg deposition depends on complex interactions between
emission location and the time of year which influences both atmospheric
transport patterns and oxidant concentration fields .
Geographical distribution of Hg deposition
Due to the uncertainty in the atmospheric oxidation pathway of Hg,
simulations were performed using both O3 / OH and
Br oxidation mechanisms to investigate their impact on Hg deposition
fields.
Figure a–d compare the geographical distribution of the
modelled Hg deposition field using emission fields with 0 % and of
15 % HgP, for each of the oxidation mechanisms. The
O3 / OH mechanism leads to enhanced deposition in the
tropics, whereas the Br mechanism leads to relatively higher
deposition over the South Atlantic and Indian oceans.
Assuming a fraction of HgP in the emissions subtracts some Hg(g)0
from the global pool, and this fraction is deposited nearer to emission sources
in central Africa, South-East Asia, the Amazon and near the wildfires which occur
in North America and in North Asia in the northern hemispheric summer.
From Fig. , it appears that assuming a fraction of the BB
emissions to be HgP causes the deposition field simulated using the
Br oxidation mechanism to more closely resemble that using the
O3 / OH mechanism.
To better understand the combined effect of Hg speciation and oxidation
pathway on the deposition distribution, agreement maps were created to
highlight the similarities and differences in the distribution of
high-deposition (⩾μ+1σ, the average plus 1 standard
deviation) model cells in the different simulations as described in
. Figure a and b show the agreement
maps of the deposition for three different HgP fractions using the
two oxidation mechanisms.
Using the O3 / OH mechanism, the number of model cells
in which the model predicts high deposition in all three emission speciation
scenarios is higher than when using the Br mechanism (631 vs. 248).
This is due to the combination of high emissions and high oxidant
concentrations in the tropics when using the O3 / OH
mechanism, constraining Hg deposition to a relatively narrow latitude band.
Using the Br mechanism, Hg has a greater possibility of being transported
to mid- and high latitudes before being oxidised and deposited.
In both the oxidation scenarios the higher deposition over the remote areas of
North America and North Asia occurs only when the fraction of HgP in the
emissions is greater than zero.
High local contributions to Hg deposition from BB using the Br
mechanism occur more frequently when the fraction of HgP is
non-zero (purple in Fig. b), unlike the
O3 / OH simulations.
Figure contrasts the results from the two oxidation
mechanisms with varying percentages of HgP and a simulation in
which the HgP fraction was assumed to be 100 %, so that it
behaves as an inert tracer. The agreement maps show clearly that the
similarity in the deposition fields increases with increasing HgP
fraction, reflected in the number of cells where all three simulations agree
(grey in the figure) and the decrease in the number of cells where only one
simulation predicts deposition higher than μ+σ (red, blue and
yellow).
Geographical distribution of the Hg total deposition from model runs
including only BB emission sources and assuming two different HgP
emission fractions, 15 % (a, c) and 0 % (b, d), for
the two oxidation mechanisms considered,
O3 / OH (a–b) and Br (c–d).
Agreement maps of high Hg deposition model cells obtained
considering only BB emissions and assuming 0, 15 and 30 % to be
HgP under both the oxidation mechanisms considered,
O3 / OH (a) and Br (b). The maps
show the areas where deposition is greater than μ+σ.
Agreement maps, under three different speciation scenarios,
0 % (a), 15 % (b) and 30 % (c)
HgP, of high Hg deposition model cells obtained considering only BB
and using the O3 / OH, the Br oxidation
mechanisms, and a sensitivity run where all Hg BB emissions were considered
inert (i.e. all HgP). The deposition field from for this “inert”
run was retained under the three different speciation scenarios. The maps
show the areas where deposition is greater than μ+σ.
Characteristics of ground-based sites measuring HgP.
Long name
Short name
Lat
Long
Elev. (m)
Amsterdam Island
AMD
-37.8
77.58
70
Cape Hedo
CHE
26.86
128.25
60
Longobucco
LON
39.39
16.61
1379
Manaus
MAN
-2.89
-59.97
110
Mauna Loa
MAU
19.54
-155.58
3399
Mt. Changbai
MCH
42.4
128.11
741
Mt. Waliguan
MWA
36.29
100.9
3816
Rao
RAO
57.39
11.91
5
Constraints from global measurements networks
The output from the simulations including all emissions (as indicated in
Table ) for the year 2013 were compared to measurement
data available from GMOS and other monitoring networks. The sites are the
same as those used in , the measurements from which have
been reviewed .
Table summarises a selection of metrics
from the comparison for total gaseous mercury (TGM; Hg(g)0 +
Hg(g)0) and for Hg in wet deposition.
The results are in line with those obtained from previous studies
focusing on a different time period,
and they indicate a generally good agreement between measured and simulated TGM, especially
for the run with the Br-driven oxidation mechanism.
For the Hg wet deposition fluxes, the results show poorer performance due
to the difficulties for coarse-resolution global models to simulate precipitation
events correctly .
Since the different sensitivity runs considering HgP from BB differ by a
only a small perturbation in the speciation of total Hg emitted from the
BASE (or the relevant reference) case, the results are actually indistinguishable
from BASE (or the relevant reference) case.
Therefore the table reports the comparison only from runs which yield different results.
Also, this means that neither wet deposition nor TGM is the most appropriate variable
to assess the validity of any of the assumptions concerning HgP emitted
during BB.
During 2013, within the GMOS and other Hg monitoring initiatives, a number of
measurement sites collected samples of atmospheric HgP. These
stations and their precise locations are reported in the
Table .
The result of the comparison with the measurements from these sites is
summarised in Fig. . Figure a
shows the annually averaged surface concentrations of HgP as
simulated by the BASE run for 2013. As is evident, surface HgP hotspots are close to the industrial areas of eastern Europe, India, East Asia
and South Africa and to areas characterised by significant BB activity,
including Indonesia, central Africa and boreal areas of Canada and Asia.
(a) Annual averaged surface HgP concentrations as
simulated by BASE run including all emission sources.
(b) Differences in annual averaged surface HgP
concentrations as simulated by BASE and by NO HgP runs, both
including emissions from all sources. Black dots indicate that differences
are not significant based on a Student t test at a 95 % confidence
interval. Blue bigger points indicate the locations of measurements sites
reported in Table . Short names are depicted for sites
where the differences between BASE and NO HgP runs are significant.
(c) Scatter plot of annual averaged HgP concentrations
measured at sites of Table compared with those obtained
by different sensitivity runs. The blue circles in the figure indicate values
relative to the sites further investigated at an higher temporal resolution (see Fig. ),
whereas the red circles indicate values relative to high-altitude sites affected by processes other than BB.
A first analysis to find those areas where the model run, assuming a fraction
HgP from BB (i.e. BASE), gives results that are statistically distinguishable from
the model run assuming Hg from BB to be only Hg(g)0 was performed
to identify the measurements sites best suited for further analysis.
The geographical distribution of these differences is reported in panel b of
Fig. . The areas were the anthropogenic input is the
greatest differ little between the simulations (based on a Student t test
at 95 % level of confidence), as indicated by dot points in the panel.
Most of the stations, depicted by the blue solid points in the same panel,
are within these regions and therefore unsuitable for the analysis. Only
three stations are in areas where the model results are significantly
different. These, the short names of which are reported in the panel, are
Amsterdam Island (AMD), Manaus (MAN) and Mauna Loa (MAU). However, MAU and Mt.
Waliguan (MWA) are high-altitude sites and affected by processes other than
BB. For both the remaining stations (AMD and MAN), the fraction of
HgP that is assumed to be emitted by anthropogenic activities, as
estimated by AMAP2010 inventory , is not sufficient
alone to explain the averaged HgP concentrations collected over the
year, as is evident from Fig. c. The inclusion of
30 % HgP from BB emissions at MAN and AMD and also the
inclusion of 15 % HgP from BB as using the FINN inventory at
MAN significantly improve the model performances, in terms of the annual
average HgP concentrations.
The result of the comparison between the HgP concentrations
collected at these two stations with the same modelled at the same points by
a selection of sensitivity runs at an finer temporal resolution (daily
averages) is reported in the two panels of Fig. .
The same comparisons for all the stations, among with the box and whisker
plot of distributions of the HgP concentrations measured and
modelled, are reported in Fig. . Although the
measurement coverage of the year at MAN is sporadic, it is an important
station because it is situated in a remote area where the local Hg emissions
are due only to ASGM (only Hg(g)0) and BB
. The consistent reduction of the error between
measured and modelled HgP concentrations when considering a fraction
of particulate bound Hg emitted from BB (NRMSE from 48 to 34 % and 27 for
30 % HgP and FINN, respectively) clearly indicates the role of
BB on the observed HgP values. At AMD
(Fig. b), the inclusion of the fraction of
HgP from BB results only in a slightly better agreement with the
measurements (NRMSE from 16 to 14 %). However, the HgP event
matching grows from 25 to 32 %, especially in the last part of the year. These HgP events have been associated with BB events in the central Africa in . Peaks was evaluated using the “findpeak” function
in MATLAB, available from
https://it.mathworks.com/help/signal/ref/findpeaks.html.
To summarise, it seems that the emissions of a fraction HgP from BB is plausible and
supported by the measures of atmospheric HgP, at least for the period investigated
and for the location of the two remote stations AMD and MAN.
However, it has to be noted that the uncertainties related to the precise
nature of atmospheric HgP and to the processes it undergoes in the
atmosphere could have an appreciable impact on the model results. For
example, the assumption of a temperature-dependent gas-particle
HgII partitioning proposed by (i.e. the
“Partitioning” and “Partitioning ref” runs) yield overall better model
agreement with annually average HgP concentrations (stars in
Fig. c). However, comparing the modelled daily average
time series with measurements results in clearly poorer performance at both
the AMD and MAN stations (see Fig. b and c). More
importantly, this assumption tends to render statistically indistinguishable
(Student t test at 95 % level of confidence) the contribution of any
eventual HgP from BB, as evident from
Fig. a.
Temporal evolution of daily averaged surface HgP
concentrations measured at Manaus (MAN) and Amsterdam Island (AMD) for the
entire 2013, compared with a selection of sensitivity runs.
Left column: temporal evolution of the daily averaged surface
HgP concentrations measured at all sites from
Table for the entire 2013, compared with the modelled
values as simulated by BASE and by NO HgP runs, including emissions
from all sources. Right column: box plots of the distribution of the of the
daily averaged surface HgP concentrations, for the entire 2013, as
measured and simulated by the different sensitivity runs. Note the
logarithmic for both MAU and MWA subplot.
(a) Differences in annual averaged surface HgP
concentrations as
simulated by Partitioning and by Partitioning ref. runs, both including emissions
from all sources and the temperature-dependent HgII gas-particle partitioning
as implemented in . Black dots indicate that differences are
not significant based on a Student t test at
a 95 % confidence interval. Bigger blue points indicate the locations of
measurements sites reported in Table .
Temporal evolution of daily averaged surface HgP concentrations
are measured at Manaus (MAN) and Amsterdam Island (AMD) for the entire 2013, compared
with the modelled values from the same sensitivity runs.
Uncertainty and biomass burning versus anthropogenic impact
Besides the uncertainty related to the atmospheric Hg oxidation mechanism
there are a
number of other factors that lead to uncertainty in ascertaining the fate of
Hg released by BB. Some of the model assumptions and parametrisations, in
particular emission height, made little difference to the eventual deposition
fields in the case where emissions from BB were considered to be 100 %
Hg(g)0 . Other sensitivity studies of the
speciation of anthropogenic emissions reveal that varying the fractions of
Hg(g)II and HgP can result in quite different Hg
deposition patterns due to their shorter residence time compared to
Hg(g)0 .
Horizontal pattern correlation (R) and probabilities that the Hg deposition fields of
the different runs belong to the same distribution as the BASE run (PKS).
The checks in the ensemble column indicate the inclusion of the respective run in the ensemble
in Fig. .
Sim.
R
PKS
Ensemble
Time resolution
3-hourly
1
1
and
Monthly
1
0.99
vertical profile
HAM-Profile
1
1
3 h + HAM-Profile
1
1
HgP mapping
HgP to PM
1
1
HgP to OC
1
0.42
✓
HgP to FMC
0.99
0.45
✓
HgP fraction
NO HgP
0.94
0.38
✓
4 % HgP
0.97
0.72
✓
30 % HgP
0.97
0.5
✓
Inventory
GFAS
0.98
0
✓
FINN
0.96
0
✓
Oxidation mech.
Br
0.96
0
✓
and
Br No HgP
0.81
0
✓
combination
Br 30 % HgP
0.91
0
✓
Br HgP to OC
0.95
0
✓
Br HgP to FMC
0.94
0
✓
GFAS Br
0.94
0
✓
FINN Br
0.92
0
✓
Hg deposition (Mg) coming from BB to the oceans as obtained by the different runs for the
2013. The last two columns reports the percentage of the total Hg that deposits over sea and land.
Run
Total deposition/Mg
%
N. Atlantic
S. Atlantic
N. Pacific
S. Pacific
Indian Ocean
Med. Sea
Arctic
S. Ocean
Sea
Land
BASE
31.7
32.5
75.3
67.4
45.9
1.1
5.0
2.3
66
34
NO HgP
32.1
32.4
82.0
74.4
48.9
1.2
4.7
2.6
71
29
30 % HgP
31.3
32.5
69.3
61.0
43.2
1.0
5.2
2.0
62
38
HgP to FMC
31.4
32.1
74.3
66.6
44.7
1.1
5.8
2.3
66
34
Br No HgP
26.6
39.4
75.8
83.0
55.3
1.1
3.7
7.6
74
26
Br 30 % HgP
28.0
36.4
61.7
61.1
44.9
0.9
4.8
4.6
62
38
Br HgP to FMC
27.3
36.8
66.6
68.8
47.1
1.0
5.6
5.8
66
34
Mercury deposition (Mg) to the oceans for 2013 from BB and comparison (ratio)
with deposition from anthropogenic activities for both oxidation mechanisms.
O3 / OH
N. Atlantic
S. Atlantic
N. Pacific
S. Pacific
Indian Ocean
Med. Sea
Arctic
S. Ocean
Only BB
29.8
29.9
72.1
63.0
43.0
1.1
4.7
2.1
Only anthropogenic
144.0
80.0
417.7
206.7
151.3
10.0
34.3
11.0
Ratio
0.21
0.37
0.17
0.31
0.28
0.11
0.14
0.19
Br
N. Atlantic
S. Atlantic
N. Pacific
S. Pacific
Indian Ocean
Med. Sea
Arctic
S. Ocean
Only BB
25.7
34.7
65.1
66.2
46.2
0.9
4.2
5.1
Only anthropogenic
153
85.33
457.3
188.3
140
12.33
34
27.3
Ratio
0.17
0.41
0.14
0.35
0.33
0.08
0.12
0.19
Geographical distribution of the total Hg deposition from BB
emissions obtained from an ensemble of simulations for the year
2013 (a) in terms of the average (μ) and standard deviation
σ of the ensemble. The comparison of the BB simulation with an
ensemble of runs including only anthropogenic emissions
shows (b) the geographic distribution of the fraction of the BB
contribution to the Hg deposition from the anthropogenic sources.
Ratio of the Hg deposition due to biomass burning with respect to Hg
deposition due to anthropogenic emissions for three anthropogenic emissions
scenarios for 2035: (a) current policy (CP), (b) new policy
(NP) and (c) maximum feasible reduction (MFR).
However, the choice of the two main vertical profile of the BB emissions used
in this study, also when combined with the temporal resolution of the
emissions, actually has little influence on the final Hg deposition fields.
Emitting all of the Hg in a single model layer does have an
impact. However, these cases are a little speculative, and therefore not
included in the final analysis.
The factor which has the greatest influence on the Hg deposition pattern is
the choice of emission inventory, whereas for a given inventory the most
important factors are the fraction of HgP and the oxidation
mechanism, although as seen in Sect. the impact of
the oxidation mechanism decreases with increasing HgP fraction. The
method of calculating the HgP fraction has a limited impact on
deposition on a global scale, with 66 % of Hg deposited over the oceans,
but the regional impact does change. Using FMC to determine the HgP
fraction increases deposition to the Arctic by 16 and 13 %
(O3 / OH and Br) and to the Southern Ocean by 30
and 25 % (O3 / OH and Br); see
Table .
Apart from the polar oceans the oceanic basins, most influenced by the
fraction of HgP in the BB emissions are the North and South Pacific
and the Indian ocean. The total deposition to individual basins from the
limiting 0 and 30 % HgP cases is included in
Table .
The horizontal pattern correlation method and
the non-parametric Kolmogorov–Smirnov two-sample test were used to assess the
differences in the deposition fields obtained from the simulations summarised
in Table , as in . The results of
the comparison of the simulations with the BASE run are presented in
Table .
The results of the Kolmogorov–Smirnov two-sample test were exploited to construct an
inspected ensemble, following the approach of and previously
employed in . The ensemble includes only those simulations with
realistic assumptions and deposition fields with little or no probability of
belonging to the same distribution.
Hg deposition from the resulting ensemble is shown in
Fig. a. The figure shows how the inclusion of
HgP in the BB emissions causes greater deposition near the hotspots of central Africa, Brazil, South-East Asia, North America and North
Asia. Nonetheless approximately 70 % of Hg deposition occurs over the
oceans, with the Tropical Atlantic, Tropical Pacific and Indian oceans most
impacted (see Table ).
Figure b compares the BB ensemble results with an
ensemble constructed using only anthropogenic emissions, using the EDGAR
, AMAP2010 and STREETS
inventories (considering both oxidation mechanisms; see
Table ). It can be seen that the contribution of BB to
Hg deposition is close to or greater than that from anthropogenic activities
in the areas near the locations of wildfires, central Africa, the Amazon,
part of the Southern Atlantic and North Asia. The contribution to Hg
deposition from BB relative to anthropogenic emissions is greater than
25 % everywhere in the Southern Hemisphere and exceeds 30 % in the
South Pacific and South Atlantic (Table ).
As anthropogenic Hg emissions decline the relative impact of BB Hg will rise,
as shown in Fig. , where the Hg deposition due to BB is
compared with Hg deposition from anthropogenic sources in three different
emission scenarios for 2035 (see , for details of the
emission scenarios).
Comparison of the results of BASE and Br simulations including all emissions sources with
observations from measurement networks for 2013.
Total gaseous mercury
Wet deposition
Regression
Stats
Regression
Stats
Intercept
Slope
r
NRMSE %
Intercept
Slope
r
NRMSE %
BASE
0.36
0.62
0.72
10.54
5.84
0.04
0.12
6.89
Partitioning
0.34
0.7
0.73
11.9
3.71
0.03
0.14
4.76
Br
-0.08
0.96
0.74
15.68
7.1
0.08
0.18
9.12
Conclusions
That a fraction of HgP is present in BB Hg emissions has been
confirmed by several field measurements , and
this fact has been suggested as an explanation of high HgP
observations at a remote site , but this is the first time
it has been included in a model study to assess its effects on a global
scale.
A previous modelling study assuming emissions from BB to be 100 %
Hg(g)0 suggested that as much as 75 %
of the Hg emitted by BB was deposited to ocean basins, with global
implications for food webs and human health.
Including a fraction of HgP in the BB Hg emissions has an impact on
the geographical distribution of the deposition fluxes for the year analysed,
reducing input to the global oceans and some high-latitude regions, while
enhancing potentially negative effects on ecosystems close to areas where
significant BB occurs.
The presence of HgP in the emissions decreases the differences seen in
Hg deposition patterns produced by employing different oxidation mechanisms.
In the remote areas of North Asia and North America, BB has a strong local
impact if the HgP fraction is non-zero. This latter result is independent
of the atmospheric oxidation pathway.
In simulations with 30 % HgP in the BB emissions, deposition
over the Arctic increases by 11 % with respect to 0 % HgP
(30 % in the Br simulations) and by 16 % when the
HgP fraction is determined by FMC (37 % in the Br
simulation).
The fraction of HgP released from BB while having an impact on the
land–sea distribution of global Hg deposition, has a more significant impact
in particular regions including the polar regions, the South Atlantic and
Pacific and Indian oceans. These results apply for the investigated year
(2013) and may differ for other years due to the complex interaction of the
numerous factors determining the final fate of Hg. However, few alternatives
of analysis period exist due the limited time coverage of global measurement
network(s). Indeed the year selected for the analysis allowed for the
hypotheses tested in this study to be supported by observations at a number
of sites from GMOS, which has extended the observational network in the
tropics and the Southern Hemisphere .
The eventual emissions of a fraction of HgP from BB cannot be
evaluated by comparison with observed gaseous atmospheric Hg concentrations or Hg in
wet precipitation samples due to the very small impact of HgP from
BB on both the atmospheric burden and wet deposition relative to all other
emissions sources (≈ 1–2 %). Conversely, its contribution to
atmospheric HgP is comparable to that of anthropogenic activities
and therefore may be investigated. The inclusion in the model run of a
fraction of HgP from BB contributes to better model performances at
two remote sites, Manaus and Amsterdam Island. However results are not
definitive due to the large uncertainty related to HgP emissions
and transformation processes. Further modelling and more measurement sites,
particularly in remote areas, would help reduce some of the uncertainties
associated with Hg emissions from BB and constrain these processes.
Biomass burning has and will continue to play a significant role in the
cycling of legacy Hg, and its relative importance is likely to increase as
anthropogenic emissions are reduced and global temperatures rise.