Skip to main content

Full text of "Mercury from wildfires: Global emission inventories and sensitivity to 2000–2050 global change"

See other formats

Atmospheric Environment 173 (2018) 6-15 


Contents lists available at ScienceDirect 

Atmospheric Environment 

journal homepage: 

Mercury from wildfires: Global emission inventories and sensitivity to r® 
2000-2050 global change I 

Aditya Kumar''’’^, Shiliang Yaoxian Huang''’‘^, Hong Liao^ Jed O. Kaplan^ 

^ Department of Geolo^cal and Mining Engineering and Sciences, Michigan Technological University, Hoi^hton, MI, USA 
^ Department of Civil and Envirorunental Engineering Michigan Technological University, Houghton, MI, USA 
School of Environment and Energy, Peking University Shenzhen Graduate School, Shertzhen, China 
Now at Department of Climate and Space Sciences and Engineering University of Michigan, Ann Arbor, MI 48109, USA 
® School of Environmental Science and Engineering Nanjing University of Information Science & Technology, Nanjit^ China 
^ ARVE Research SARL, Fully, Switzerland 



A (climate) 

A (land us^and cover) 

A (mercury anthropogenic 


A (mercury 
wildfire emissions) 



Biomass burning 
Climate change 
Land use 
Land cover 

We estimate the global Hg wildfire emissions for the 2000s and the potential impacts from the 2000-2050 
changes in climate, land use and land cover and Hg anthropogenic emissions by combining statistical analysis 
with global data on vegetation type and coverage as well as fire activities. Global Hg wildfire emissions are 
estimated to be 612 Mg year“^. Africa is the dominant source region (43.8% of global emissions), followed by 
Eurasia (31%) and South America (16.6%). We find significant perturbations to wildfire emissions of Hg in the 
context of global change, driven by the projected changes in climate, land use and land cover and Hg anthro¬ 
pogenic emissions. 2000-2050 climate change could increase Hg emissions by 14% globally and regionally by 
18% for South America, 14% for Africa and 13% for Eurasia. Projected changes in land use by 2050 could 
decrease the global Hg emissions from wildfires by 13% mainly driven by a decline in African emissions due to 
significant agricultural land expansion. Future land cover changes could lead to significant increases in Hg 
emissions over some regions (-1-32% North America, -1-14% Africa, -1-13% Eurasia). Potential enrichment of 
terrestrial ecosystems in 2050 in response to changes in Hg anthropogenic emissions could increase Hg wildfire 
emissions globally (-1-28%) and regionally (-1-19% North America, -1-20% South America, -1-24% Africa, -1-41% 
Eurasia). Our results indicate that the future evolution of climate, land use and land cover and Hg anthropogenic 
emissions are all important factors affecting Hg wildfire emissions in the coming decades. 

* Corresponding author. Department of Geological and Mining Engineering and Sciences, Michigan Technological University, Houghton, MI, USA. 
E-mail address: slwu(a) (S. Wu). 


Received 29 Jime 2017; Received in revised form 11 October 2017; Accepted 28 October 2017 

Available online 31 October 2017 

1352-2310/ © 2017 Elsevier Ltd. All rights reserved. 

A. Kumar et al. 

Atmospheric Envirortment 173 (2018) 6~15 

1. Introduction 

Mercury (Hg) is a toxic and persistent pollutant in the global en¬ 
vironment. Hg emitted to the atmosphere can be transported long dis¬ 
tances before depositing to terrestrial and aquatic systems. The atmo¬ 
spheric emissions of Hg include both anthropogenic sources such as 
fossil fuel combustion, smelting of ores, cement production, waste in¬ 
cineration, and artisanal gold mining (Chen et ah, 2014; Pacyna et ah, 
2010; Pirrone et ah, 2010; Streets et ah, 2011; Veiga et ah, 2006), 
natural emissions from erupting and passively degassing volcanoes, 
geothermal hot springs, topsoil enriched in Hg (Ferrara et ah, 2000; 
Nimick et ah, 2013; Nriagu and Becker, 2003; Pyle and Mather, 2003; 
Varekamp and Buseck, 1986), and biomass burntng/wildfires (Friedli 
et ah, 2003, 2009; Sigler et ah, 2003; Turetsky et ah, 2006; Veiga et ah, 
1994; Weiss-Penzias et ah, 2007; Wiedinmyer and Friedli, 2007). 

Terrestrial vegetation plays an important role in the biogeochemical 
cycling of Hg, and is considered a significant reservoir for atmospheric 
Hg. Obrist (2007) estimated the global uptake of atmospheric Hg by 
vegetation could be more than 1 000 Mg per year. Retention of atmo¬ 
spheric Hg by vegetation delays its transport to soils and aquatic sys¬ 
tems (Graydon et ah, 2012). Hence, vegetation plays an important role 
in Hg sequestration by terrestrial ecosystems. Wildfires, through the 
burning of biomass, can effectively mobilize the Hg stored in terrestrial 
ecosystems and lead to massive emissions of Hg and other pollutants 
into the atmosphere (Andreae and Merlet, 2001; Biswas et ah, 2007; Ito 
and Penner, 2004; Turetsky et ah, 2006; van der Werf et ah, 2006; 
Veiga et ah, 1994). Multiple studies have estimated Hg wildfire emis¬ 
sions at the global and regional scales (Table 2) with global total 
emissions in the range of 104-1 330 Mg Hg year^^ indicating large 
uncertainty in the estimated Hg emissions from wildfires. A number of 
studies have developed wildfire emission inventories for Hg based on 
CO or CO 2 emission inventories by applying fixed emission ratios be¬ 
tween Hg and CO/CO 2 [e.g (Brunke et ah, 2001; Ebinghaus et ah, 2007; 
Sigler et ah, 2003).]. To our knowledge, Friedli et ah (2009) is the only 
study so far that have compiled a multi-year (1997-2006) emission 
inventory for Hg from wildfires at the global scale by accounting for the 
variations in Hg emission factors across geographical regions and eco¬ 
systems. The emission factor (EF) for a given species (Hg in this case) 
can be expressed as the mass of that species emitted per unit dry bio¬ 
mass burned (Andreae and Merlet, 2001). It is affected by both the 
vegetation type and geographical region. 

Significant changes in global vegetation coverage are expected in 
the coming decades driven by either climate change (referred to as land 
cover change hereon) or anthropogenic land use change (referred to as 

land use change hereon) (Bachelet et ah, 2001, 2003; Cox et ah, 2000; 
Falloon et ah, 2012; Notaro et ah, 2007; O'ishi and Abe-Ouchi, 2009; 
Tilman et ah, 2001; Wu et ah, 2012). Wu et ah (2012) predicted sig¬ 
nificant changes in forests and grassland coverage by 2050 with the 
northern mid-latitudes being affected the most. Notaro et ah (2007) 
predicted reduced forest cover over the Amazon, South Africa and 
Australia due to the radiative effect of rising CO 2 and poleward shift of 
the boreal forest due to both radiative and physiological effects. In 
addition, continued anthropogenic conversion of forested land for 
agriculture in the future to support a growing global human population 
and the resulting food and energy demand could result in significant 
degradation of natural vegetation. Tilman et ah (2001) projected an 
18% increase in global agricultural (cropland & pastureland) area in 
2050 relative to 2000. By 2050 (IPCC AIB scenario), significant de¬ 
gradation of natural vegetation for agricultural purposes is predicted in 
Eastern US, Central Africa, Southern and Equatorial Asia and Western 
Europe while agricultural land area could decrease in South America, 
East Asia, Western parts of Australia and Russia (IPCC, 2001; MNP, 

The changes in vegetation type and density associated with future 
land use/land cover change could have a direct influence on future 
wildfire activity. Huang et ah (2015) predicted a 15% increase in global 
wildfire frequency in 2050 due to land cover change alone with major 
increases in Africa and North America. On the contrary, land use 
change could result in significant declines in fire occurrence in Africa 
and Western US whereas increases are predicted in Australia and South 
America (Huang et ah, 2015). Furthermore, Hg emissions from wild¬ 
fires are affected by the Hg content in the vegetation (biomass) (Biswas 
et ah, 2007) making Hg emissions particularly sensitive to future al¬ 
terations in land use and land cover. In addition to land use/land cover 
change, future changes in climate would also influence global fire ac¬ 
tivity. Predicted warmer temperatures in the future together with al¬ 
terations in precipitation patterns driven by climate change could result 
in significantly different fire regimes from the 2000s (Cochrane and 
Barber, 2009; Flannigan et ah, 2009; Yue et ah, 2013). Huang et ah 
(2015) predicted a 19% increase in global fire frequency in 2050 caused 
by changes in meteorology due to climate change. Yue et ah (2013) 
predicted that summertime surface aerosol concentrations resulting 
from wildfire emissions over the Western US could increase by 46-70% 
for organic carbon and 20-27% for black carbon by 2050 relative to the 
2000s driven by climate change. Finally, the changes in anthropogenic 
emissions of Hg can significantly affect Hg emissions from wildfires by 
affecting the atmospheric deposition of Hg. Because atmospheric de¬ 
position is a major source of Hg enrichment in terrestrial vegetation and 

Fig. 1. Definition of global regions used in this 
study. BONA: Boreal North America, TENA: 
Temperate North America, CEAM: Central North 
America, NHSA: Northern Hemisphere South 
America, SHSA: Southern Hemisphere South 
America, EURO: Europe, MIDE: Middle East, 
NHAF: Northern Hemisphere Africa, SHAF: 
Southern Hemisphere Africa, BOAS: Boreal Asia, 
CEAS: Central Asia, SEAS: South East Asia, EQAS: 
Equatorial Asia, AUST: Australia. 


A. Kumar et al. 

Atmospheric Enviroriment 173 (2018) 6-15 

soils, future alterations in Hg atmospheric deposition to terrestrial en¬ 
vironments could play a key role in determining the emission factors of 
Hg from wildfires. In this study, we estimate the atmospheric Hg 
emissions from global wildfires by accounting for the regional varia¬ 
tions in both fire activities and Hg emission factors and examining the 
impacts from 2000 to 2050 changes in climate, land use/land cover and 
anthropogenic emissions. 

2. Methodology 

Hg emissions from wildfires are calculated based on the classical 
equation for biomass burning emissions (Seiler and Crutzen, 1980) and 
accounting for various factors including vegetation type and density, 
Hg content in biomass, and fire characteristics. Considering the sig¬ 
nificant variation in fire characteristics, land cover and climate across 
geographical regions, the analyses for various regions is carried out 
separately. We use geographical region definitions that have been 
commonly used in the literature (Aldersley et ah, 2011; Friedli et ah, 
2009; Giglio et ah, 2010, 2013, 2006b; hi et ah, 2012; Van der Werf 
et ah, 2010; van der Werf et ah, 2006). 

Fig. 1 shows a map of the geographical regions. The Hg emissions 
model is developed at a spatial resolution of 4° x 5° (latitude x long¬ 
itude). The monthly mean Hg emissions from wildfires at a grid cell (i, 
j), E(ij), are calculated as: 


Eaj) = Yj 


2k=i i) 






EFkgjy Hg emission factors for land cover type k at grid cell (i, j) 
Burned area at (i, j) 

fk(ijy Fractional areal coverage for land cover type k at (i, j) 

Mkjijy Available biomass density for vegetation type k at grid cell (i, 


CF(ij): Combustion fraction at (i, j) 

For Hg emission factors for the 2000s, we follow Friedli et ah (2009) 
to assign region-specific EFs for various regions but update the tropical 
forests EFs with data from Melendez-Perez et ah (2014). The final EF 
(122 ng Hg/g biomass burned) for tropical forests is the average of 
values reported by Friedli et ah (2009) (198 ng Hg/g biomass burned) 
and Melendez-Perez et ah (2014) (47 ng Hg/g biomass burned). Boreal 
regions (BONA and BOAS) and Equatorial Asia (EQAS) have the highest 
EF values (315 ng Hg/g biomass burned) due to high soil mercury pools 
present there (Friedli et ah, 2009) followed by temperate forests 
(242 ng Hg/g biomass burned), tropical forests (122 ng Hg/g biomass 
burned) and grasslands (41 ng Hg/g biomass burned). 

Available biomass density (M) is the amount of dead/live plant 
material available for consumption by fires per unit area. The available 
biomass in a grid cell includes herbaceous vegetation, non-woody and 
woody tree parts and plant litter (decomposable and resistant). 
Available biomass estimates for forest and non-forest ecosystems from 
Jain et ah (2006), generated using the terrestrial component of the 
Integrated Science Assessment Model (ISAM) (Jain and Yang, 2005) are 
used here. These estimates are provided for nine geographical regions 
and four land cover types (tropical, temperate, boreal forests and non- 
forested ecosystems). Combustion fraction (CF) represents the fraction 
of available biomass that burns during a fire. It is a function of vege¬ 
tation type, its spatial arrangement and moisture content (Ito and 
Penner, 2004). We follow the scheme used by Wiedinmyer et ah (2006) 
to estimate the CF. This scheme, based on the percentage of tree cover 
(TC) in a grid cell, classifies the vegetation into different types and 
assigns corresponding CF estimates for trees and herbaceous vegetation 
therein, as: 


030for trees and 0.90 for herbs if %TC > 60% 

_ 0.30 for trees and herbs if 40% < %TC 

< 60% 

0.00 for trees and 0.98 for herbs %TC < 40% 

( 2 ) 

hand cover data used in this work consists of fractional areal ve¬ 
getation coverage output (2000 and 2050) from the Lund-Potsdam Jena 
Dynamic Global Vegetation (LPJ DGVM) model (Gerten et ah, 2004; 
Hickler et ah, 2006; Schaphoff et ah, 2006; Sitch et ah, 2003; Thonicke 
et ah, 2001; Wu et ah, 2012). The LPJ model is a process based global 
model of vegetation dynamics. It simulates the production and loss of 
plant biomass, competition amongst different plant species for re¬ 
sources, vegetation structural properties and soil biogeochemistry 
(Gerten et ah, 2004; Thonicke et ah, 2001) based on the inputs of 
meteorology, soil type and atmospheric CO 2 concentration. Natural 
vegetation in each grid cell is described in terms of fractional coverage 
of 9 plant functional types (PFTs), which include tropical broadleaved 
evergreen tree (TrBE), tropical broadleaved raingreen tree (TrBR), tempe¬ 
rate needleleaved evergreen tree (TeNE), temperate broadleaved evergreen 
tree (TeBE), temperate broadleaved summergreen tree (TeBS), boreal nee- 
dleleaved evergreen tree (BNE), boreal needleleaved summergreen tree 
(BNS), C3 and C4 herbs. Each woody PFT is either evergreen, sum¬ 
mergreen or raingreen depending on water availability and tempera¬ 
ture whereas herbaceous PFTs are C3 or C4 based on the type of pho¬ 
tosynthesis activity associated with them. The fractional natural 
vegetation coverage output used here was obtained by driving the LPJ 
DGVM with meteorology fields (2000 and 2050) generated from the 
GISS Global Climate Model Version 3 (GISS GCM v3) (Rind et ah, 2007; 
Wu et ah, 2007, 2008a, 2008b) following the IPCC AIB scenario for 
future greenhouse gas concentrations. Further simulation details are 
provided in Wu et ah (2012). Here 10-year average vegetation data 
(1991-2000 vs 2041-2050) is used to examine the long-term changes in 

Data for anthropogenic land use consists of cropland areal coverage 
(2000 and 2050) from the IMAGE model following the IPCC AIB sce¬ 
nario (IMAGE Team, 2001; MNP, 2006). The data was regridded from a 
spatial resolution of 1° x l°-4° x 5° in this work. For each grid cell, 
fractional coverage of each of the LPJ PFTs was uniformly reduced in 
proportion to the crop fraction to accommodate for cropland coverage. 
For the 2000s, the LPJ PFT and crop fractions are at the 2000s level. 
The 2000-2050 land use and land cover scenario includes both natural 
vegetation and crop coverage as predicted for 2050. The land cover 
change scenario has natural vegetation changing in response to climate 
change but the cropland coverage is kept fixed at the 2000s level 
whereas in the land use change scenario natural vegetation is kept fixed 
at the 2000s level and cropland coverage following land use trends for 
2050 is used. The climate change scenario involves changes in me¬ 
teorology only (using GISS GCM v3 meteorology fields for 2000 and 

The burned area A(ij) was estimated as: 

A(i,j) = a(ij)(Tf(-,, (3) 

where the proportionality factor aqj) is a function of land cover type 
and coverage (represented by tree cover (Tf^jj), herb cover (Hf^yj) and 
barren land Bf^jj in grid cell (i, j)) and climate (represented by surface 
temperature T^^yj and precipitation Ps^yj in grid cell (i, j)). N,. is the fire 
frequency which is affected by fire ignition sources such as lightning 
strikes and anthropogenic ignitions, meteorological conditions, vege¬ 
tation density and human population density. Huang et al. (2015) ac¬ 
counted for all these factors and estimated the changes in fire frequency 
(at a spatial resolution of 4° x 5° and monthly temporal resolution) in 
response to changes in climate, land cover, land use and population 
density considering a suite of scenarios (land use/land cover change. 


A. Kumar et al. 

Atmospheric Envirortment 173 (2018) 6-15 

climate change, changes in ignition agents and anthropogenic sup¬ 
pression of fires). Their results for the 2050 climate change, land use 
and land cover change scenarios are used here. 

Our burned area estimation methodology involves building a sta¬ 
tistical model relating fire activity (independent variable) and burned 
area (dependent variable), which was subsequently used with the fire 
model generated fire frequencies to predict burned area (for 2000 and 
2050). We use available global fire frequencies and burned area data¬ 
sets from satellite observations and consider land cover (% tree cover 
(% TC), % herb cover (% HC) and % barren land (% BL)) and me¬ 
teorology (Ts and P^) as the major factors influencing the relationship 
between fire activity and burned area. In order to account for this de¬ 
pendence, regression tree models are employed. A regression tree is a 
decision tree based statistical model (Breiman et al., 1984; Breiman and 
Meisel, 1976; Loh, 2008, 2011). It recursively partitions the input data 
space {y, Xi, ...Xn}(y is the dependent variable and x's are the in¬ 
dependent variables) consisting of training data for the dependent and 
independent variables into subsets and fits separate linear regression 
models to each subset. The independent variables can be used as 
splitting variables (used to make univariate splits at each node of the 
tree), predietive variables (used to predict the dependent variable) or 
both. The splitting of input data space into subsets allows application of 
regression relations to a homogeneous data space with respect to the 
splitting variables and thus improve the applicability of the relations to 
the data space. The dependenee of burned area on land cover has been 
highlighted by Van Der Werf et al. (2003) and Giglio et al. (2006b) and 
regression trees for predicting global burned area from fire frequencies 
have been previously used [e.g (Giglio et al., 2006b, 2010).]. Separate 
regression trees were developed for each of the 14 geographical regions 
in Fig. 1. This approach allows modeling the effects of fuel type (tree/ 
herbaceous), configuration and availability and favorable/unfavorable 
weather conditions on wildfire spread. 

The splitting variables include %TC, %HC, %BL, Tj and P^. Each 
terminal node of the tree consists of a linear regression model with fire 
frequencies as the independent variable and burned area as the de¬ 
pendent variable. Training data used for constructing the regression 
trees is described below. Observed fire frequencies (2001-2015) are 
from the Collection 5 Terra MODIS Climate Modeling Grid (CMG) fire 
product (data at ftp;// available 
at a 0.5° X 0.5° spatial and monthly temporal resolution (Giglio et al., 
2006a; Justice et al., 2002). For burned area, estimates from the Global 
Fire Emissions Database Version 4 (GFEDv4) (data at ftp://fuoco.geog. ((2001-2015)) (Giglio et al., 2013) are 
used. The spatial resolution of this product is 0.5° x 0.5° (monthly 
temporal resolution). Both datasets were regridded to a 4° x 5° spatial 
resolution for use in this work. LPJ land cover data (accounting for 
cropland coverage as described earlier) was used for the 2000s land 
cover. The PFTs were combined to obtain %TC (sum of all tree PFTs), % 
HC (sum of all herbaceous PFTs) and the remainder after accounting for 
natural vegetation and cropland yielded the %BL (non-vegetated land). 
Surface temperature and precipitation datasets consisted of monthly 
means generated from 3 hourly averaged (A-3) fields from the GEOS-4 
(2001-2003) (Suarez et al., 2005), GEOS-5 (2004-2012) (Rienecker, 
2008) and hourly average (A-1) fields from the GEOS-FP (2013-2015) 
(Lucchesi, 2013; Molod et al., 2012) meteorology products (4° x 5° 
spatial resolution). The datasets for MODIS fire frequency, GFEDv4 
burned area, LPJ land cover and GEOS meteorology at 4° x 5° were 
prepared for each of the 14 geographical regions as defined in Fig. 1. 
Thus, training data for each regional regression tree consisted of a 15- 
year (2001-2015) time series of MODIS Terra fire frequencies, GFEDv4 
burned area estimates, LPJ land cover data for the 2000s and GEOS 
meteorology fields. These data were used to construct regression trees 
for each of the 14 geographical regions. Subsequently, these regression 
tree models were applied to estimate the monthly-burned area for 2000 
and 2050 scenarios based on fire frequencies estimated from the fire 
model along with land cover data from the LPJ model and surface 

temperature and precipitation fields (monthly means from 3 hourly 
averaged (A-3) fields) from the GISS GCM v3 model. With all the input 
data available, the Hg emissions (with monthly and 4° x 5° resolution) 
from wildfires for various regions and scenarios are calculated using the 
fire emissions model (i.e. Eq. (1)). 

To examine the perturbations to wildfire emissions of Hg from 
changes in anthropogenic emissions of Hg, we assume a linear re¬ 
lationship between the Hg emission factors from wildfires and the at¬ 
mospheric deposition of mercury (Dep): EF 2000 /DEP 2000 = EF 2050 / 
DEP 2050 - The actual response of Hg enrichment in vegetation and soils 
to changes in atmospheric deposition could be very complicated and 
non-linear, but without detailed data available, the linear simplification 
allows us to estimate the sensitivity of wildfire emissions of Hg to 
changes in anthropogenic emissions. The atmospheric deposition of 
mercury for the 2000s and 2050 were estimated using the global 
mercury simulation in the GEOS-Chem model (Bey et al., 2001). The 
GEOS-Chem mercury simulation (Corbitt et al., 2011; Giang et al., 
2015; Holmes et al., 2010; Jaegle et al., 2009; Selin and Jacob, 2008; 
Selin et al., 2007, 2008; Smith-Downey et al., 2010; Strode et al., 2007; 
Zhang et al., 2016) includes three Hg species (elemental (Hg (0)), di¬ 
valent (Hg (II)) and particulate bound Hg (Hg (P)). It includes coupled 
land-ocean-atmosphere cycling of Hg. Hg emissions include anthro¬ 
pogenic and natural sources and re-emission of previously deposited 
mercury from terrestrial and aquatic systems. Sinks for Hg inelude dry 
deposition, wet deposition (for Hg (II) and Hg (P)) including sea salt 
uptake (for Hg (II)). This work uses v9-02 of the model and years 
2005-2011 were simulated with the GEOS-5 meteorology. The years 
2005-2007 were used to initialize the model and the results presented 
are averages for 2008-2011. The model was driven by anthropogenic 
emissions for 2050 following the IPCC AIB scenario (Corbitt et al., 
2011; Streets et al., 2009). Hg GDEP was calculated as the sum of de¬ 
position for all Hg species (Hg (0), Hg (II) and Hg (P) dry deposi¬ 
tion -H Hg (II) and Hg (P) wet deposition). 

3. Results 

3.1. Burned area estimates 

Using the regression tree models described above, the calculated 2000s 
(1998-2002 average) annual burned area estimates for various regions, 
are shown in Table la. The regression tree models were able to explain 
most of the variability in burned area (based on GFEE)v4 data) for each 
region (R^ = 0.81 (BONA), 0.72 (TENA), 0.68 (CEAM), 0.75 (NHSA), 0.86 
(SHSA), 0.82 (EURO), 0.81 (MIDE), 0.91 (NHAF), 0.94 (SHAF), 0.81 
(BOAS), 0.85 (CEAS), 0.73 (SEAS), 0.77 (EQAS), 0.96 (AUST)). Global 
annual burned area for the 2000s is estimated at —334 Mha year^^ with 
maximum (51%) contribution from the African continent 
(MIDE + NHAF + SHAF —170 Mha year^^) followed by Eurasia (23%) 
(EURO -H BOAS -H CEAS -H SEAS -H EQAS - 78 Mha year^^), South 
America (14%) (NHSA + SHSA - 47 Mha year^^), Australia (9%) 
(AUST — 29 Mha year^^) and North Tknerica (3%) 
(BONA + TENA + CEAM —10 Mha year^^). Table la contains burned 
area estimates from this work together with available literature estimates 
for comparison. There is considerable variability in the global and con¬ 
tinental burned area estimates in the literature due to different approaches 
used (e.g., process based fire modeling vs. satellite observations) and dif¬ 
ferent year/time period examined. However, our results reproduce the 
major burned area patterns common to most of the studies (e.g. maximum- 
burned area in Africa). Overall, estimates of both global and regional 
wildfire burned area in this work are in reasonable agreement with the 
literature and represent the spatial distribution of burned area well. 

Our calculated changes in wildfire burned area driven by changes in 
climate, land use, land cover are shown in Table lb. We find that the 
2000-2050 climate change could increase the global fire frequencies by 
19% (Huang et al., 2015), which would cause significant increases in 
burned area at both global (1-22%) and regional (Africa (1-28%), 


A. Kumar et ai 

Atmospheric Environment 173 (2018) 6-15 

Table la 

Estimated regional® and global wildfire burned area (in Mha year^^) for the 2000s. 


Study period 







This work 
















(Giglio et al., 









(Giglio et al., 







(Simon et al., 

GBA 2000'=, 








(Tansey et al., 

(Tansey et al., 




(Tansey et al., 

(Randerson et al., 









Li et al. (2012) 







(Roy et al., 









et al., 2011) 

® Regional definitions: NA = North America (BONA + TENA + CEAM); SA = South 
America (NHSA + SHSA); AFR = Africa (MIDE + NHAF + SHAE); EURAS = Eurasia 
(Europe (EURO) + Asia (BOAS + CEAS + SEAS + EQAS)); AUS = Australia (AUST); 
GLOB = global total. The subcontinental regions such as BONA etc. are as defined in 
Fig. 1. 

Continental estimates calculated from percent distribution in Tansey et al. (2004b) 
[North America: 2%, South America: 3%, Eurasia (Europe + Asia (including Russia)): 
15%, Africa: 64%, Australia + Papua New Guinea: 16%]. 

Table lb 

Projected 2000-2050 changes in wildfire burned area driven by changes in climate, land 
use, and land cover at global and regional® scales. Estimated burned area for the 2000s is 
shown in Mha year^^. 















climate change 

+ 23% 

+ 16% 

+ 28% 

+ 6% 

+ 32% 

+ 22% 

land use change 


+ 12% 



+ 31% 


land cover change 

+ 28% 

+ 2% 

+ 16% 



+ 8% 

land use and land cover 

+ 16% 

+ 18% 

+ 6% 


+ 32% 

+ 6% 


® The regional definitions are the same as in Table la. 

Australia ( + 32%), North America ( + 23%), South America ( + 16%) 
and Eurasia ( + 6%)) scales. An increase of 6% ( + 8% (land cover 
change), —3% (land use change)) in global burned area is predicted 
due to the 2000-2050 changes in land use/land cover. However, at the 
continental scale, more pronounced changes are observed. Burned area 
in North America and Africa is predicted to decline by 13% and 7% 
respectively due to reduction in natural vegetation coverage caused by 
agricultural land expansion. Lesser natural vegetation coverage could 
significantly reduce wildfire activity and limit wildfire spread. On the 
other hand, greater vegetation density and the resulting higher fire 
frequencies due to land cover change could increase burned area in 
both continents ( + 28% North America, +16% Africa). Agricultural 
land expansion is predicted to decline in South America and Australia 
by 2050 resulting in more natural vegetation and greater fire fre¬ 
quencies increasing the burned area by 12% and 31% respectively. In 
Eurasia, burned area could decline by 13% due to increase in agri¬ 
cultural land coverage. The combined effects of 2000-2050 land use/ 

land cover change contribute to greater burned area in all continents 
(North America ( + 16%), South America ( + 18%), Africa ( + 6%), 
Australia ( + 32%)) except Eurasia ( — 11%). 

3.2. Hg emissions from wildfires for the 2000s 

Our calculated Hg emissions from wildfires for the 2000s are shown 
in Table 2. Results from previous literature on Hg emissions from bio¬ 
mass burning are included in the table for comparison. Our best esti¬ 
mated global total Hg wildfire emissions for the 2000s is 612 Mg year^^ 
with 43.8% emissions from the African continent 
(MIDE + NHAF + SHjAF = 268 Mg year^^) followed by Eurasia 
(EURO + BOAS + CEAS + SEAS + EQAS = 31%, 190 Mg year^^). 
South America (NHSA + SHSA = 16.6%, 102 Mg year^^). North 
America (BONA + TENA + CEAM = 7.9%, 48 Mg year^^) and Aus¬ 
tralia (AUST = 0.7%, 4 Mg year^^). Africa and Eurasia are the domi¬ 
nant source regions for Hg emissions from wildfires. High emissions 
from Africa can be attributed to the high fire activity, which results in 
more than half of the global burned area occurring in the continent. The 
significant contribution from Eurasia primarily reflects the high Hg 
pools present over the boreal parts of the continent (Friedli et al., 
2009). Globally, tropical (43.3%) and boreal (33%) forest burning 
contribute the most to wildfire emissions of Hg followed by temperate 
forests (16.4%) and grasslands (7.3%). 

The calculated global total wildfire emissions of Hg in this work 
compare very well with the climatological values reported in the lit¬ 
erature. However, at the continental scale, there are some significant 
differences. Equatorial Asia is not the major source of wildfire emissions 
of Hg, as found by Friedli et al. (2009). In addition, the calculated 
emissions for Africa are much higher, although we follow the same 
regional emission factor assignment methodology used by them. This 
could be due to differences in other inputs to the fire emissions model 
(e.g. predicted burned area from simulated fire frequencies, available 
biomass density and different combustion fraction schemes used). The 
calculated global Hg emissions and the source distribution show much 
better agreement with Streets et al. (2009) with Africa being the most 
important source region followed by Eurasia + Oceania and South 

Our fire emissions model uses inputs from a large number of data 
sources. For example, the values for Hg emission factors are average 
values that have been compiled from several studies in the literature; 
estimates of the burned area involves regression tree models, which 
were developed based on data from MODIS-Terra, GFEDv4, the LPJ 
model as well as the GEOS meteorology. Thus, it is very difficult to 
quantify the uncertainty in our final results associated with all the 
model parameters. Nevertheless, we have carried out some simple 
analyses to examine the sensitivities of our results to various model 
parameters. We first use the bootstrap methods (Efron, 1979; Efron and 
Tibshirani, 1993) to evaluate the uncertainties in our calculated Hg 
emissions from wildfires associated with the inter-annual variability in 
meteorology. Bootstrap methods belong to the class of nonparametric 
Monte Carlo methods. In this work, nonparametric bootstrapping is 
used which regards the data sample as the pseudo-population dis¬ 
tribution with similar characteristics as the true population. It involves 
estimating the sampling distribution of a statistic (e.g. mean of the 
sample) by repeated sampling (with replacement) from the data sample 
and subsequent determination of the properties of the statistic (e.g. 
standard error of the mean). We apply bootstrap methods to the 5-year 
(1998-2002) emissions sample (global and continental) to determine 
the standard error of the mean as a measure of the uncertainty in the 
emissions. In order to provide a range for the mean, the 95% better 
bootstrap confidence intervals (BCa) (Efron and Tibshirani, 1993) are 
reported. Random samples (size n = 5) were selected from the original 
emissions sample and the mean was calculated. This procedure was 
repeated 10,000 times to create a sample of 10,000 means. The stan¬ 
dard deviation of these 10,000 means (standard error (SE) of the mean) 


A. Kumar et ai 

Atmospheric Environment 173 (2018) 6-15 

Table 2 

Model calculated 2000s wildfire emissions of Hg (in Mg year^^) for the global total and various regions.' 


Study Period 







This work 








Brunke et al. (2001) 

510-1 140" 

380-1 330'* 

Chen et al. (2013)^ 



Cinnirella and Pirrone (2006)^ 


Delacerda (1995)^ 


De Simone et al. (2015) 



Ebinghaus et al. (2007) 



Friedli et al. (2009) 







675 (435-915) 

Huang et al. (2011)® 



Michelazzo et al. (2010)^ 



Nelson et al. (2012) 



Roulet et al. (1999)® 

(range for the 1980s) 


Sigler et al. (2003) 


Streets et al. (2009) 

(for 1996 and 2006) 






Veiga et al. (1994)® 



Wiedinmyer and Friedli (2007)* 


44 (20-65) 

Weiss-Penzias et al. (2007) 

Late 1990s 

670 (340-1 000) 

^ The regional definitions are the same as in Table la. 

The specific values shown for this work represent our best estimates; refer to Table 3 and the corresponding discussion in the text for uncertainty analyses. 
" Based on Hg/CO emission ratio. 

Based on Hg/C02 emission ratio. 

® Emission estimates only for China. 

^ Emissions for Europe (1990-2004) and Russian federation (1996-2002) only. 

^ For Amazon only. 

'* For 1996. 

‘ For 2006. 

^ Emissions for lower 48 states of North America & Alaska only. 

Table 3 

Sensitivity of calculated wildfire Hg emissions (in Mg year^^) to various model para¬ 
meters used in this study. 




Burned area 






North America 




46-51 (2.08%) 

South America 




95-111 (4.04%) 





255-281 (2.51%) 





180-199 (2.58%) 





3.4-4.5 (6.75%) 





598-630 (1.32%) 

^ Ranges of Hg emissions represent the 95% confidence intervals. 

was computed to estimate the uncertainty. The standard errors range 
from 1.3% to 7% (SE/mean; Table 3), reflecting the relatively small 
inter-annual variability in the simulated meteorology from the GISS 

We then examine the sensitivities of our calculated Hg emissions 
from wildfires to the uncertainties in model parameters including 
burned area, Hg emission factors and available biomass density. Based 
on literature studies (e.g. Andreae and Merlet, 2001; Brunke et al., 
2001; Friedli et al., 2003, 2009; Ebinghaus et al., 2007; Weiss-Penzias 
et al., 2007), it appears an uncertainty of 20-30% is typical for these 
model parameters. Therefore, we have performed two additional si¬ 
mulations for each model parameter by assuming a 20% uncertainty 
(case 1: —20 to 0% change in the parameter, case 2: 0-20% change in 
the parameter). For each case, a sequence of random numbers between 
the range [-20, 0] or [0, 20]) were generated with a uniform distribu¬ 
tion function. These numbers represent the percentage changes to be 
applied to the model parameters. Therefore, for case 1, all the random 
numbers would be between —20 and 0 and 0 to 20 for case 2. For 
burned area, the random number sequence length equaled to one 
number representing a particular region (1: BONA, 2: TENA ... 14: 
AUST). For emission factors and biomass density, the sequence con¬ 
sisted of different numbers for each vegetation type in a region. Based 

on these perturbation tests, we have summarized the sensitivities of our 
final results to various model parameters, as shown in Table 3. We find 
that the sensitivities of calculated Hg emissions to the three model 
parameters (Hg emission factors, available biomass density, and the 
burned area) are similar and a 20% uncertainty in each of the para¬ 
meter would lead to around 20% uncertainty in our final results. 

3.3. Changes in Hg emissions driven by climate, land use/land cover and 
anthropogenic emissions change 

The perturbations to wildfire emissions of Hg due to 2000-2050 
changes in climate, land use and land cover and anthropogenic emis¬ 
sions are shown in Fig. 2(a), Fig. 2(b) (Supplementary material) and 
Table 4. We find significant increases in wildfire emissions of Hg due to 
2000-2050 changes in climate. Global emissions increase by 14% 
mainly driven by increases in Africa ( + 14%), South America ( + 18%), 
and Eurasia (4-13%). In Africa, emissions increase mainly in the 
southern and northern parts, which could experience significantly 
warmer and drier conditions than the 2000s resulting in greater and 
more severe wildfires. Greater precipitation in Central Africa on the 
other hand causes a decline in wildfire activity and emissions. Eurasian 
emissions increase is primarily due to higher wildfire activity caused by 
warmer conditions in the boreal parts of the region (14% increase in 
BOAS). However, emissions in Equatorial Asia could decline due to 
suppression of wildfire activity caused by greater precipitation. Sig¬ 
nificant increases in wildfires is predicted over the boreal parts of North 
America due to warmer temperatures and over Western US due to 
warmer and drier conditions than the 2000s. These changes in climate 
result in increasing Hg emissions from North America by 8%. Australia 
(34% increase in emissions) and South America could mainly experi¬ 
ence greater wildfire activity in the eastern parts due to higher tem¬ 
peratures and less precipitation in 2050. 

In response to the projected changes in land use and land cover by 
2050, we find that land use change would be a major driving force in 
Africa. Substantial conversion of forests to croplands causes wildfire 
emissions to decline by 36% in the continent, outweighing a 14% 


A. Kumar et al. 

Atmospheric Enviroriment 173 (2018) 6-15 

Hg emissions (2000) 

Hg emissions change CC (2050-2000) 

180* ISO^W i20“W 90“W 60“W 30‘W 

30*E 60°E gp^E t20°E ISO^E 180“ 

0.00 0.29 0.57 0.66 1.14 143 1.71 2.00 > ug/m2V 

Hg emissions change LU (2050-2000) 

180* 150“W 120* W 90°W 60*W 30'W 0* 30‘E 60*E 90*E 1 20*E 150“E 180* 

< -1.00 -O-TI -0.43 -0.14 0.14 0 43 0.71 1.00 > ug/iTi2/y 

Hg emissions change LC (2050-2000) 







W W''' 



if , . : ' 

180* ISOV 120*W 90*W 60*W 30*W 

< <1.00 -0.71 >0.43 4.14 0.14 0.43 0.71 1.00 > 

Hg emissions change LU/LC (2050-2000) 


< -1.00 -0.71 -0.43 4.14 0.14 0.43 0.71 1.00 > ug;in2/y 

Hg emissions change EF (2050-2000) 

180" I50°W ISO“W 90"w 60“W 30“W 

< .1.00 -0.71 ^>.43 .0.14 0.14 0.43 0.71 1.00 > 

t 60°E 00"E 120'E 150°E 180‘ 


ISO" 150"W 120"W 90°W 60‘W 30"W 

30"E 60"E 00"E 1 20*E 150"E ISO" 

< .1.00 .0.71 41.43 41.14 0.14 0.43 0.71 1.00 > llo/nl2V 

Fig. 2. (a): Wildfire emissions of Hg for 2000 C[ig/mVyear) (top left), and projected changes by 2050 Cpg/mVyear) due to climate change (CC) (top right), land use change (LU) (middle 
left), land cover change (LC) (middle right), land use/land cover change (bottom left) and anthropogenic emissions change (EF change) (bottom right). 

Table 4 

Projected 2000-2050 changes in wildfire emissions of Hg driven by changes in climate, 
land use, and land cover. Estimated 2000s Hg emissions are shown in Mg year^^. 















climate change 

+ 8% 

+ 18% 

+ 14% 

+ 13% 

+ 34% 

+ 14% 

land use change 


+ 19% 



+ 58% 


land cover change 

+ 32% 


+ 14% 

+ 13% 


+ 12% 

land use and land cover 

+ 19% 

+ 21% 

- 24% 



+ 0.8% 

anthropogenic emissions 

+ 19% 

+ 20% 

+ 24% 

+ 41% 

+ 18% 

+ 28% 

^ The regional definitions are the same as in Table la. 

increase in emissions due to greater forest coverage and resulting fire 
frequencies caused by land cover change. Overall, the combined effects 
of land use and land cover result in a 24% decline in Hg wildfire 
emissions from Africa. Land use change could also be the dominant 
factor influencing emissions in South America and Australia. Land use is 

projected to decline in both the continents resulting in more forest 
coverage and wildfires than the 2000s. Hence, emissions increase sig¬ 
nificantly in both continents (-1-19% (South America), -1- 58% 
(Australia)). However, land cover change could cause a decline in 
emissions in both continents ( — 5% (South America), —11% 
(Australia)), primarily due to reduction in temperate forest (high Hg 
emission factors) coverage in sub-tropical South America and decrease 
in wildfire frequency in Australia. Overall, 2000-2050 land use/land 
cover change results in increasing Hg wildfire emissions by 21% in 
South America and 182% in Australia. 

On the contrary, future changes in land cover in North America are 
found to have a greater influence on Hg emissions from wildfires (32% 
emissions increase) than the changes in anthropogenic land use (12% 
emissions decrease). Increase in boreal and temperate forest coverage 
in the high and mid-latitudes of the continent and the resulting greater 
fire frequencies lead to increases in Hg emissions over boreal North 
America (1-29%) and the US (1-35%). Land use change would have 
negligible impacts on emissions in boreal North America; however, it 
could result in a significant decrease in emissions from the US (— 23%). 


A. Kumar et al. 

Atmospheric Envirortment 173 (2018) 6-15 

In Eurasia as well, land cover change acts as a major factor causing a 
13% increase in emissions, primarily caused by boreal forest expansion 
and the resulting increase in wildfires. Land use change would have 
negligible overall effects partly reflecting the diverging trends in an¬ 
thropogenic land use in this region (e.g. decreases in Eastern Asia and 
parts of Russia but increases in Equatorial Asia). 

Following the IPCC AIB scenario, the global anthropogenic emis¬ 
sions are predicted to increase with Hg(II) being the dominant emis¬ 
sions constituent in 2050 (Corbitt et al., 2011; Streets et al., 2009). As a 
consequence, we find that the Hg enrichment of terrestrial ecosystems 
driven by changes in atmospheric Hg deposition by 2050 would lead to 
increases in global wildfire emissions of Hg by 28%. The most sig¬ 
nificant increases are calculated over Eurasia (1-41%) and Africa 
(1- 24%) which together account for about 75% of global Hg wildfire 
emissions for the 2000s. Wildfire emissions of Hg from South East, 
Central and East Asian countries increase significantly in response to 
greater Hg anthropogenic emissions in these countries and the resulting 
deposition to terrestrial environments. South America (1-20%), North 
America ( + 19%) and Australia ( + 18%) experience significant in¬ 
creases in emissions as well. Greater Hg deposition to the boreal regions 
results in an increase of 19% in wildfire emissions from boreal North 
America and 35% from boreal Asia. Emissions from tropical peatlands 
in Equatorial Asia also increase due to greater Hg deposition in 2050. It 
should be noted that new developments in technology (such as the 
mercury control technology being used by coal-fired power plants) and 
policy (such as the Minamata Convention) can significantly affect the 
future trends of anthropogenic emissions of Hg, but these factors are not 
accounted for in this study. 

4. Conclusions 

We investigate the Hg emissions from wildfires in this study. We 
first develop the global and regional emission inventories for the 2000s 
and then examine the perturbations from the projected 2000-2050 
changes in climate, land use, land cover and anthropogenic Hg emis¬ 
sions. Africa (43.8%), Eurasia (31%) and South America (16.6%) are 
found to be the major sources of Hg wildfire emissions in the 2000s. 
Following the IPCC AIB scenario, 2000-2050 climate change would 
lead to more frequent and severe wildfires in most regions around the 
world resulting in significant increases in wildfire emissions of Hg at 
both the global and continental scales. Climate change driven altera¬ 
tions in natural vegetation could also increase global emissions parti¬ 
cularly in the boreal regions, the US and Africa. However, these impacts 
of a future favorable climate for fires and land cover change on global 
emissions are suppressed by continued anthropogenic destruction of 
natural vegetation in order to support agricultural development. As a 
result, emissions in Africa, which is a major source of wildfire emis¬ 
sions, decline in 2050 due to reduced forest cover. In addition, de¬ 
struction of forests in Equatorial Asia and the Western US reduces Hg 
wildfire emissions from these regions. On the other hand, a projected 
rise in anthropogenic emissions in 2050 and the resulting greater Hg 
contamination of terrestrial environments contributes to increasing 
emissions globally and regionally. Wildfire emissions of Hg in the 
boreal regions are predicted to increase in response to the 2000-2050 
changes in climate, land cover and anthropogenic Hg emissions which 
could have significant implications for Hg deposition to the Arctic. 

Conflicts of interest 

The authors declare that they have no conflict of interest. 

This study is supported by NSF (grant #1313755) and U.S. EPA 
(grant # 83518901). J.O. Kaplan was further supported by the 
European Research Council (COEVOLVE 313797). S. Wu acknowledges 

the sabbatical fellowship from Peking University. Superior, a high 
performance computing cluster at Michigan Technological University, 
was used in obtaining results presented in this publication. 

Appendix A. Supplementary data 

Supplementary data related to this article can be found at http://dx. 


Aldersley, A., Murray, S.J., Cornell, S.E., 2011. Global and regional analysis of climate 
and human drivers of wildfire. Sci. Total Environ. 409, 3472-3481. 

Andreae, M.O., Merlet, P., 2001. Emission of trace gases and aerosols from biomass 
burning. Glob. Biogeochem. Cycles 15, 955-966. 

Bachelet, D., Neilson, R.P., Hickler, T., Drapek, R.J., Lenihan, J.M., Sykes, M.T., Smith, B., 
Sitch, S., Thonicke, K., 2003. Simulating past and future dynamics of natural eco¬ 
systems in the United States. Glob. Biogeochem. Cycles 17. 

Bachelet, D., Neilson, R.P., Lenihan, J.M., Drapek, R.J., 2001. Climate change effects on 
vegetation distribution and carbon budget in the United States. Ecosystems 4, 

Bey, I., Jacob, D.J., Yantosca, R.M., Logan, J.A., Field, B.D., Fiore, A.M., Li, Q., Liu, H.Y., 
Mickley, L.J., Schultz, M.G., 2001. Global modeling of tropospheric chemistry with 
assimilated meteorology: model description and evaluation. J. Geophys. Res. Atmos. 
106, 23073-23095. 

Biswas, A., Blum, J.D., Klaue, B., Keeler, G.J., 2007. Release of mercury from Rocky 
Mountain forest fires. Glob. Biogeochem. Cycles 21. 

Breiman, L., Friedman, J., Olshen, R., Stone, C., 1984. Classification and Regression Trees. 
CRC Press, Boca Raton, Florida. 

Breiman, L., Meisel, W., 1976. General estimates of the intrinsic variability of data in 
nonlinear regression models. J. Am. Stat. Assoc. 71, 301-307. 

Brunke, E.G., Labuschagne, C., Slemr, F., 2001. Gaseous mercury emissions from a fire in 
the cape peninsula. South Africa, during January 2000. Geophys. Res. Lett. 28, 

Chen, C., Wang, H., Zhang, W., Hu, D., Chen, L., Wang, X., 2013. High-resolution in¬ 
ventory of mercury emissions from biomass burning in China for 2000-2010 and a 
projection for 2020. J. Geophys. Res. Atmos. 118. 

Chen, Y., Wang, R., Shen, H., Li, W., Chen, H., Huang, Y., Zhang, Y., Chen, Y., Su, S., Lin, 
N., 2014. Global mercury emissions from combustion in light of international fuel 
trading. Environ. Sci. Technol. 48, 1727-1735. 

Cinnirella, S., Pirrone, N., 2006. Spatial and temporal distributions of mercury emissions 
from forest fires in Mediterranean region and Russian Federation. Atmos. Environ. 
40, 7346-7361. 

Cochrane, M.A., Barber, C.P., 2009. Climate change, human land use and future fires in 
the Amazon. Glob. Change Biol. 15, 601-612. 

Corbitt, E.S., Jacob, D.J., Holmes, C.D., Streets, D.G., Sunderland, E.M., 2011. Global 
source-receptor relationships for mercury deposition under present-day and 2050 
emissions scenarios. Environ. Sci. Technol. 45, 10477-10484. 

Cox, P.M., Betts, R.A., Jones, C.D., Spall, S.A., Totterdell, I.J., 2000. Acceleration of global 
warming due to carbon-cycle feedbacks in a coupled climate model. Nature 408, 

De Simone, F., Cinnirella, S., Gencarelli, C.N., Yang, X., Hedgecock, I.M., Pirrone, N., 
2015. Model study of global mercury deposition from biomass burning. Environ. Sci. 
Technol. 49, 6712-6721. 

Delacerda, L.D., 1995. Amazon mercury emissions. Nature 374, 20-21. 

Ebinghaus, R., Slemr, F., Brenninkmeijer, C., Van Velthoven, P., Zahn, A., Hermann, M., 
O'Sullivan, D., Oram, D., 2007. Emissions of gaseous mercury from biomass burning 
in South America in 2005 observed during CARIBIC flights. Geophys. Res. Lett. 34. 

Efron, B., 1979. Bootstrap methods: another look at the Jackknife annals of statistics. 7, 

Efron, B., Tibshirani, R.J., 1993. An Introduction to the Bootstrap: Monographs on 
Statistics and Applied Probability, vol. 57 Chapman and Hall/CRC, New York and 

Falloon, P., Dankers, R., Betts, R., Jones, C., Booth, B., Lambert, F., 2012. Role of vege¬ 
tation change in future climate under the AIB scenario and a climate stabilisation 
scenario, using the HadCM3C Earth system model. Biogeosciences 9, 4739. 

Ferrara, R., Mazzolai, B., Lanzillotta, E., Nucaro, E., Pirrone, N., 2000. Volcanoes as 
emission sources of atmospheric mercury in the Mediterranean basin. Sci. Total 
Environ. 259, 115-121. 

Flannigan, M.D., Krawchuk, M.A., de Groot, W.J., Wotton, B.M., Gowman, L.M., 2009. 
Implications of changing climate for global wildland fire. Int. J. Wildland Fire 18, 

Friedli, H., Arellano, A., Cinnirella, S., Pirrone, N., 2009. Initial estimates of mercury 
emissions to the atmosphere from global biomass burning. Environ. Sci. Technol. 43, 

Friedli, H., Radke, L., Lu, J., Bank, C., Leaitch, W., MacPherson, J., 2003. Mercury 
emissions from burning of biomass from temperate North American forests: labora¬ 
tory and airborne measurements. Atmos. Environ. 37, 253-267. 

Gerten, D., Schaphoff, S., Haberlandt, U., Lucht, W., Sitch, S., 2004. Terrestrial vegetation 
and water balance—hydrological evaluation of a dynamic global vegetation model. J. 
Hydrol. 286, 249-270. 

Giang, A., Stokes, L.C., Streets, D.G., Corbitt, E.S., Selin, N.E., 2015. Impacts of the 


A. Kumar et at 

Atmospheric EnviroTvnent 173 (2018) 6-15 

minamata convention on mercury emissions and global deposition from coal-fired 
power generation in Asia. Environ. Sci. Technol. 49, 5326-5335. 

Giglio, L., Csiszar, I., Justice, C.O., 2006a. Global distribution and seasonality of active 
fires as observed with the Terra and aqua moderate resolution imaging spectro- 
radiometer (MODIS) sensors. J. Geophys. Res. Biogeosci. 111. 

Giglio, L., Randerson, J., Van der Werf, G., Kasibhatla, P., Collatz, G., Morton, D., DeFries, 
R., 2010. Assessing variability and long-term trends in burned area by merging 
multiple satellite fire products. Biogeosciences 7. 

Giglio, L., Randerson, J.T., Werf, G.R., 2013. Analysis of daily, monthly, and annual 
burned area using the fourth-generation global fire emissions database (GFED4). J. 
Geophys. Res. Biogeosci. 118, 317-328. 

Giglio, L., Van der Werf, G., Randerson, J., Collatz, G., Kasibhatla, P., 2006b. Global 
estimation of burned area using MODIS active fire observations. Atmos. Chem. Phys. 
6, 957-974. 

Graydon, J.A., St Louis, V.L., Lindberg, S.E., Sandilands, K.A., Rudd, J.W., Kelly, C.A., 
Harris, R., Tate, M.T., Krabbenhoft, D.P., Emmerton, C.A., 2012. The role of terres¬ 
trial vegetation in atmospheric Hg deposition: pools and fluxes of spike and ambient 
Hg from the METAALICUS experiment. Glob. Biogeochem. Cycles 26. 

Hickler, T., Prentice, I.C., Smith, B., Sykes, M.T., Zaehle, S., 2006. Implementing plant 
hydraulic architecture within the LPJ d 5 mamic global vegetation model. Glob. Ecol. 
Biogeogr. 15, 567-577. 

Holmes, C.D., Jacob, D.J., Corbitt, E.S., Mao, J., Yang, X., Talbot, R., Slemr, F., 2010. 
Global atmospheric model for mercury including oxidation by bromine atoms. Atmos. 
Chem. Phys. 10, 12037-12057. 

Huang, X., Li, M., Friedli, H.R., Song, Y., Chang, D., Zhu, L., 2011. Mercury emissions 
from biomass burning in China. Environ. Sci. Technol. 45, 9442-9448. 

Huang, Y., Wu, S., Kaplan, J.O., 2015. Sensitivity of global wildfire occurrences to various 
factors in the context of global change. Atmos. Environ. 121, 86-92. 

IMAGE Team, H., 2001. The IMAGE 2.2 Implementation of the SRES Scenarios: a 
Comprehensive Analysis of Emissions, Climate Change and Impacts in the 21st 
Century. RIVM CD-ROM Publication 481508018. 

IPCC, 2001. Climate change 2001: the scientific basis, contribution of working group I to 
the third assessment report of the intergovernmental panel on climate change. In: 
Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., Linden, van der, P.J. (Eds.), 
Cambridge Lfnited Kingdom and New York. Cambridge University Press, NY, USA. 

Ito, A., Penner, J.E., 2004. Global estimates of biomass burning emissions based on sa¬ 
tellite imagery for the year 2000. J. Geophys. Res. Atmos. 109. 

Jaegle, L., Strode, S.A., Selin, N.E., Jacob, D.J., 2009. The Geos-chem Model, Mercury 
Fate and Transport in the Global Atmosphere. Springer, pp. 533-545. 

Jain, A.K., Tao, Z., Yang, X., Gillespie, C., 2006. Estimates of global biomass burning 
emissions for reactive greenhouse gases (CO, NMHCs, and NOx) and C02. J. Geophys. 
Res. Atmos. 111. 

Jain, A.K., Yang, X., 2005. Modeling the effects of two different land cover change data 
sets on the carbon stocks of plants and soils in concert with C02 and climate change. 
Glob. Biogeochem. Cycles 19. 

Justice, C., Giglio, L., Korontzi, S., Owens, J., Morisette, J., Roy, D., Descloitres, J., 
Alleaume, S., Petitcolin, F., Kaufman, Y., 2002. The MODIS fire products. Remote 
Sens. Environ. 83, 244-262. 

Li, F., Zeng, X., Levis, S., 2012. A process-based fire parameterization of intermediate 
complexity in a Dynamic Global Vegetation Model. Biogeosciences 9, 2761-2780. 

Loh, W.Y., 2008. Classification and Regression Tree Methods. Encyclopedia of Statistics in 
Quality and Rehability. 

Loh, W.Y., 2011. Classification and regression trees. Wiley Interdiscip. Rev. Data Min. 
Knowl. Discov. 1, 14-23. 

Lucchesi, R., 2013. File Specification for GEOS-5 FP (Forward Processing). 

Melendez-Perez, J.J., Fostier, A.H., Carvalho, J.A., Windmoller, C.C., Santos, J.C., Carpi, 
A., 2014. Soil and biomass mercury emissions during a prescribed fire in the 
Amazonian rain forest. Atmos. Environ. 96, 415-422. 

Michelazzo, P.A.M., Fostier, A.H., Magarelli, G., Santos, J.C., Carvalho, J.A., 2010. 

Mercury emissions from forest burning in southern Amazon. Geophys. Res. Lett. 37. 

MNP, 2006. Integrated modelling of global environmental change, an overview of IMAGE 
2.4. In: Bouwman, A.F., Kram, T., Klein Goldewijk, K. (Eds.), Netherlands 
Environmental Assessment Agency (MNP), Bilthoven, The Netherlands. 

Molod, A., Takacs, L., Suarez, M., Bacmeister, J., Song, I.-S., Eichmann, A., 2012. The 
GEOS-5 Atmospheric General Circulation Model: Mean Chmate and Development 
from MERRA to Fortuna. 

Nelson, P.F., Morrison, A.L., Malfroy, H.J., Cope, M., Lee, S., Hibberd, M.L., Meyer, C.M., 
McGregor, J., 2012. Atmospheric mercury emissions in Australia from anthropogenic, 
natural and recycled sources. Atmos. Environ. 62, 291-302. 

Nimick, D.A., Caldwell, R.R., Skaar, D.R., Selch, T.M., 2013. Fate of geothermal mercury 
from yellowstone national park in the madison and Missouri rivers. USA. Sci. Total 
Environ. 443, 40-54. 

Notaro, M., Vavrus, S., Liu, Z., 2007. Global vegetation and climate change due to future 
increases in C02 as projected by a fully coupled model with dynamic vegetation. J. 
dim. 20, 70-90. 

Nriagu, J., Becker, C., 2003. Volcanic emissions of mercury to the atmosphere: global and 
regional inventories. Sci. Total Environ. 304, 3-12. 

O'ishi, R., Abe-Ouchi, A., 2009. Influence of dynamic vegetation on climate change 
arising from increasing C02. Clim. Djm. 33, 645-663. 

Obrist, D., 2007. Atmospheric mercury pollution due to losses of terrestrial carbon pools? 
Biogeochemistry 85, 119-123. 

Pacyna, E.G., Pacyna, J., Sundseth, K., Munthe, J., Kindbom, K., Wilson, S., Steenhuisen, 
F., Maxson, P., 2010. Global emission of mercury to the atmosphere from anthro¬ 
pogenic sources in 2005 and projections to 2020. Atmos. Environ. 44, 2487-2499. 

Pirrone, N., Cinnirella, S., Feng, X., Finkehnan, R., Friedli, H., Leaner, J., Mason, R., 
Mukherjee, A., Stracher, G., Streets, D., 2010. Global mercury emissions to the 

atmosphere from anthropogenic and natural sources. Atmos. Chem. Phys. 10, 

Pyle, D.M., Mather, T.A., 2003. The importance of volcanic emissions for the global at¬ 
mospheric mercury cycle. Atmos. Environ. 37, 5115-5124. 

Randerson, J., Chen, Y., Werf, G., Rogers, B., Morton, D., 2012. Global burned area and 
biomass burning emissions from small fires. J. Geophys. Res. Biogeosci. 117. 

Rienecker, M., 2008. File Specification for GEOS-5 DAS Gridded Output. NASA Goddard 
Space Flight Center. 

Rind, D., Lerner, J., Jonas, J., McLinden, C., 2007. Effects of resolution and model physics 
on tracer transports in the NASA Goddard Institute for Space Studies general circu¬ 
lation models. J. Geophys. Res. Atmos. 112. 

Roulet, M., Lucotte, M., Farella, N., Serique, G., Coelho, H., Sousa Passos, C., De Jesus da 
Silva, E., Scavone de Andrade, P., Mergler, D., Guimaraes, J.-R., 1999. Effects of 
recent human colonization on the presence of mercury in Amazonian ecosystems. 
Water. Air,&Soil Pollut. 112, 297-313. 

Roy, D.P., Boschetti, L., Justice, C.O., Ju, J., 2008. The collection 5 MODIS burned area 
product—global evaluation by comparison with the MODIS active fire product. 
Remote Sens. Environ. 112, 3690-3707. 

Schaphoff, S., Lucht, W., Gerten, D., Sitch, S., Cramer, W., Prentice, I.C., 2006. Terrestrial 
biosphere carbon storage under alternative climate projections. Clim. Change 74, 

Seiler, W., Crutzen, P.J., 1980. Estimates of gross and net fluxes of carbon between the 
biosphere and the atmosphere from biomass burning. Clim. Change 2, 207-247. 

Selin, N.E., Jacob, D.J., 2008. Seasonal and spatial patterns of mercury wet deposition in 
the United States: constraints on the contribution from North American anthro¬ 
pogenic sources. Atmos. Environ. 42, 5193-5204. 

Selin, N.E., Jacob, D.J., Park, R.J., Yantosca, R.M., Strode, S., Jaegle, L., Jaffe, D., 2007. 
Chemical cycling and deposition of atmospheric mercury: global constraints from 
observations. J. Geophys. Res. Atmos. 112. 

Selin, N.E., Jacob, D.J., Yantosca, R.M., Strode, S., Jaegle, L., Sunderland, E.M., 2008. 
Global 3-D land-ocean-atmosphere model for mercury: present-day versus pre¬ 
industrial cycles and anthropogenic enrichment factors for deposition. Glob. 
Biogeochem. Cycles 22. 

Sigler, J., Lee, X., Munger, W., 2003. Emission and long-range transport of gaseous 
mercury from a large-scale Canadian boreal forest fire. Environ. Sci. Technol. 37, 


Simon, M., Plummer, S., Fierens, F., Hoelzemann, J.J., Arino, O., 2004. Burnt area de¬ 
tection at global scale using ATSR-2: the GLOBSCAR products and their qualification. 
J. Geophys. Res. Atmos. 109. 

Sitch, S., Smith, B., Prentice, I.C., Arneth, A., Bondeau, A., Cramer, W., Kaplan, J., Levis, 
S., Lucht, W., Sykes, M.T., 2003. Evaluation of ecosystem dynamics, plant geography 
and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Glob. 
Change Biol. 9, 161-185. 

Smith-Downey, N.V., Sunderland, E.M., Jacob, D.J., 2010. Anthropogenic impacts on 
global storage and emissions of mercury from terrestrial soils: insights from a new 
global model. J. Geophys. Res. Biogeosci. 115. 

Streets, D.G., Devane, M.K., Lu, Z., Bond, T.C., Sunderland, E.M., Jacob, D.J., 2011. All- 
time releases of mercury to the atmosphere from human activities. Environ. Sci. 
Technol. 45, 10485-10491. 

Streets, D.G., Zhang, Q., Wu, Y., 2009. Projections of global mercury emissions in 2050. 
Environ. Sci. Technol. 43, 2983-2988. 

Strode, S.A., Jaegle, L., Selin, N.E., Jacob, D.J., Park, R.J., Yantosca, R.M., Mason, R.P., 
Slemr, F., 2007. Air-sea exchange in the global mercury cycle. Glob. Biogeochem. 
Cycles 21. 

Suarez, M.J., daSilva, A., Dee, D., Bloom, S., Bosilovich, M., Pawson, S., Schubert, S., Wu, 
M.-L., Sienkiewicz, M., Stajner, L, 2005. Documentation and Validation of the 
Goddard Earth Observing System (GEOS) Data Assimilation System, Version 4. 

Tansey, K., GrEgoire, J.-M., Binaghi, E., Boschetti, L., Brivio, P.A., Ershov, D., Flasse, S., 
Fraser, R., Graetz, D., Maggi, M., 2004a. A global inventory of burned areas at 1 km 
resolution for the year 2000 derived from SPOT VEGETATION data. Chm. Change 67, 

Tansey, K., Gregoire, J.M., Defourny, P., Leigh, R., Pekel, J.F., van Bogaert, E., 

Bartholome, E., 2008. A new, global, multi-annual (2000-2007) burnt area product 
at 1 km resolution. Geophys. Res. Lett. 35. 

Tansey, K., Gregoire, J.M., Stroppiana, D., Sousa, A., Silva, J., Pereira, J., Boschetti, L., 
Maggi, M., Brivio, P.A., Fraser, R., 2004b. Vegetation burning in the year 2000: 
global burned area estimates from SPOT VEGETATION data. J. Geophys. Res. Atmos. 

Thonicke, K., Venevsky, S., Sitch, S., Cramer, W., 2001. The role of fire disturbance for 
global vegetation dynamics: coupling fire into a Dynamic Global Vegetation Model. 
Glob. Ecol. Biogeogr. 10, 661-677. 

Tilman, D., Fargione, J., Wolff, B., D'Antonio, C., Dobson, A., Howarth, R., Schindler, D., 
Schlesinger, W.H., Simberloff, D., Swackhamer, D., 2001. Forecasting agriculturally 
driven global environmental change. Science 292, 281-284. 

Turetsky, M.R., Harden, J.W., Friedli, H.R., Flannigan, M., Payne, N., Crock, J., Radke, L., 
2006. Wildfires threaten mercury stocks in northern soils. Geophys. Res. Lett. 33. 

Van Der Werf, G.R., Randerson, J.T., Collatz, G. J., Giglio, L., 2003. Carbon emissions from 
fires in tropical and subtropical ecosystems. Glob. Change Biol. 9, 547-562. 

Van der Werf, G.R., Randerson, J.T., Giglio, L., Collatz, G., Mu, M., Kasibhatla, P.S., 
Morton, D.C., DeFries, R., Jin, Y.v., van Leeuwen, T.T., 2010. Global fire emissions 
and the contribution of deforestation, savanna, forest, agricultural, and peat fires 
(1997-2009). Atmos. Chem. Phys. 10, 11707-11735. 

van der Werf, G.R., Randerson, J.T., Giglio, L., Collatz, G.J., Kasibhatla, P.S., Arellano Jr., 
A.F., 2006. Interannual variability in global biomass burning emissions from 1997 to 
2004. Atmos. Chem. Phys. 6, 3423-3441. 

Varekamp, J.C., Buseck, P.R., 1986. Global mercury flux from volcanic and geothermal 


A. Kumar et at 

Atmospheric EnviroTvnent 173 (2018) 6-15 

sources. Appl. Geochem. 1, 65-73. 

Veiga, M.M., Maxson, P.A., Hylander, L.D., 2006. Origin and consumption of mercury in 
small-scale gold mining. J. Clean. Prod. 14, 436-447. 

Veiga, M.M., Meech, J.A., Onate, N., 1994. Mercury pollution from deforestation. Nature 
368, 816-817. 

Weiss-Penzias, P., Jaffe, D., Swartzendruber, P., Hafner, W., Chand, D., Prestbo, E., 2007. 
Quantifying Asian and biomass burning sources of mercury using the Hg/CO ratio in 
pollution plumes observed at the Mount Bachelor Observatory. Atmos. Environ. 41, 

Wiedinmyer, C., Akagi, S., Yokelson, R.J., Emmons, L., Al-Saadi, J., Orlando, J., Soja, A., 
2011. The Fire INventory from NCAR (FINN): a high resolution global model to es¬ 
timate the emissions from open burning. Geosci. Model Dev. 4, 625. 

Wiedinmyer, C., Friedli, H., 2007. Mercury emission estimates from fires: an initial in¬ 
ventory for the United States. Environ. Sci. Technol. 41, 8092-8098. 

Wiedinmyer, C., Quayle, B., Geron, C., Belote, A., McKenzie, D., Zhang, X., O'Neill, S., 
Wynne, K.K., 2006. Estimating emissions from fires in North America for air quality 
modeling. Atmos. Environ. 40, 3419-3432. 

Wu, S., Mickley, L.J., Jacob, D.J., Logan, J.A., Yantosca, R.M., Rind, D., 2007. Why are 
there large differences between models in global budgets of tropospheric ozone? J. 
Geophys. Res. Atmos. 112. 

Wu, S., Mickley, L.J., Jacob, D.J., Rind, D., Streets, D.G., 2008a. Effects of 2000-2050 
changes in climate and emissions on global tropospheric ozone and the policy-re¬ 
levant backgroimd surface ozone in the United States. J. Geophys. Res. Atmos. 113. 
Wu, S., Mickley, L.J., Kaplan, J., Jacob, D.J., 2012. Impacts of changes in land use and 
land cover on atmospheric chemistry and air quality over the 21st century. Atmos. 
Chem. Phys. 12, 1597-1609. 

Wu, S., Mickley, L.J., Leibensperger, E.M., Jacob, D.J., Rind, D., Streets, D.G., 2008b. 
Effects of 2000-2050 global change on ozone air quality in the United States. J. 
Geophys. Res. Atmos. 113. 

Yue, X., Mickley, L.J., Logan, J.A., Kaplan, J.O., 2013. Ensemble projections of wildfire 
activity and carbonaceous aerosol concentrations over the western United States in 
the mid-21st century. Atmos. Environ. 77, 767-780. 

Zhang, H., Holmes, C., Wu, S., 2016. Impacts of changes in climate, land use and land 
cover on atmospheric mercury. Atmos. Environ. 141, 230-244.