(This is a sample cover image for this issue. The actual cover is not yet available at this time.)

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Multi-year black carbon emissions from cropland burning in the RussianFederation

Jessica L. McCarty a,*, Evan A. Ellicott b, Vladimir Romanenkov c, Dmitry Rukhovitch d, Polina Koroleva d

aMichigan Tech Research Institute, Ann Arbor, MI 48105, USAbDepartment of Geography, University of Maryland, College Park, MD 20742, USAcDepartment of Geographic Network of Field Experiments, All-Russian, Institute for Agrochemistry, Moscow, Russian Federationd Soil Dokuchaev Institute, Moscow, Russian Federation

h i g h l i g h t s

< Quantified cropland fire and black carbon emissions for the Russian Federation.< Peaks of BC emissions occurred in the spring, summer, and fall.< Highest BC emissions occurred during spring in European Russia and West Siberia.< The range of average annual BC emissions was 2.49 Gge22.2 Gg.

a r t i c l e i n f o

Article history:Received 30 November 2011Received in revised form24 August 2012Accepted 25 August 2012

Keywords:Cropland firesRemote sensingBlack carbonBiomass burningRussian Federation

a b s t r a c t

Cropland fires are an important source of black carbon (BC) emissions. Previous research has suggestedthat springtime cropland burning in Eastern Europe, more specifically Russia, is a main contributor of BCin the Arctic atmosphere, acting as a short-lived climate forcer strongly influencing snow-ice albedo andradiation transmission. BC emissions from cropland burning were estimated for the Russian Federationfor years 2003 through 2009 using three satellite fire products, the 1 km MODIS Active Fire Product, 0.5�

MODIS Fire Radiative Power monthly climate modeling grid product, and the 500 m MODIS Burned AreaProduct, and a agricultural statistics approach based on a modified method developed and published bythe All-Russian Institute of Organic Peat and Fertilizers to estimate farm- and regional-level residueloading from straw surplus left after grain harvesting, while accounting for agricultural management andagrometeorological inputs. The satellite-based emission calculations utilized several different land coverclassification schemas for defining croplands in Russia for both the 1 km MODIS Land Cover Product andthe 300 m MERIS GlobCover v2.2 data sets. In general, the peaks of BC emissions from cropland burningoccurred during the spring (AprileMay), summer (JulyeAugust), and the fall (October). 2008 had thehighest annual BC emissions. The range of average annual BC emissions from cropland burningcalculated from the different satellite fire products was 2.49 Gge22.2 Gg, with the agriculturalstatistics approach annual average equal to 8.90 Gg. The Global Fire Emissions Database (GFED)version 3 reported an annual average of 11.9 Gg of BC from agricultural burning. The results from thisanalysis showed that the majority of BC emissions originated in European Russia, followed by smallercontributions from West Siberia, Far East Russia, and East Siberia macro-regions. An uncertaintyassessment on data used to calculate the BC emissions found moderate uncertainty in some of the inputdata used in this first attempt to produce spatially and temporally explicit BC emission estimates fromcropland burning in the Russian Federation.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Cropland burning is a common agricultural management prac-tice used globally as an inexpensive, effective, and quick method toremove excess residue to facilitate planting (Chen et al., 2005;Badarinth et al., 2006; McCarty et al., 2007), control pests andweeds (Hari et al., 2003), and produce ash fertilization (Pyne,1993).

* Corresponding author. Tel.: þ1 734 913 6840; fax: þ1 734 913 6880.E-mail addresses: jmccarty@mtu.edu (J.L. McCarty), ellicott@umd.edu

(E.A. Ellicott), viuageoset@gmail.com (V. Romanenkov), soilmap@yandex.ru(D. Rukhovitch), soilmap@yandex.ru (P. Koroleva).

Contents lists available at SciVerse ScienceDirect

Atmospheric Environment

journal homepage: www.elsevier .com/locate/atmosenv

1352-2310/$ e see front matter � 2012 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.atmosenv.2012.08.053

Atmospheric Environment 63 (2012) 223e238

Author's personal copy

Cropland burning is a source of greenhouse gas and carbonaceousemissions, including emissions that negatively impact air quality(Streets et al., 2001; McCarty, 2011). The burning of agriculturalareas within Russia is a well-documented phenomenon (Soja et al.,2004), which began to receive governmental and scientific atten-tion within Russia as early as the 1980s (Derevyagin, 1987). Addi-tionally, cropland fires in Russia are thought to be a cause of forestfires, including the summer 2010 extreme fire season when 10-12million ha was burned as estimated by the Global Fire MonitoringCenter, an activity of the United Nations’ International Strategy forDisaster Reduction (Goldammer, 2010). However, high uncertaintyexists in these estimates, as few ground monitoring data areavailable.

Cropland fires are an important source of black carbon (BC)emissions in the Arctic. Previous research has suggested thatspringtime cropland burning in Eastern Europe, and more specifi-cally Russia, is a main source of BC in the Arctic atmosphere (Stohlet al., 2007; Warneke et al., 2010). Sharma et al. (2004) founda factor of 10 greater BC concentration in the Canadian Arcticduring the winter and spring than during the typical wildfireseason of July through November. Dry deposition of BC onto snowand/or ice cover during the spring can produce a warming effectthat increases snow melt and the snow albedo effect (Quinn et al.,2008). Hegg et al. (2009) found that the source of a large fraction ofBC deposited on the Arctic snow was vegetative biomass burning.Springtime cropland burning in the Arctic region was highlightedby Law and Stohl (2007) as an important source to Arctic BC levelsand strongly influences snow-ice albedo and radiation trans-mission in the atmosphere above the Arctic.

Burned area and active fire products derived from satellitesprovide spatially and temporally explicit estimates of vegetativebiomass burning (Langmann et al., 2009) and are useful tools forquantifying country-level agricultural burning and emissions(McCarty, 2011). Korontzi et al. (2006), using the 1 km ModerateResolution Imaging Spectroradiometer (MODIS) active fire data(Giglio et al., 2003) and 1 km MODIS land cover product (Friedlet al., 2002) to analyze global fire activity by land cover type,found that between 31 and 36% of all agricultural burning globallyannually was detected in European Russia. Similarly, van der Werfet al. (2010) found that agricultural burning was the dominantfire type and main contributor to carbon emissions in southern andEuropean Russia. The purpose of this paper is to provide estimatesof black carbon emissions from cropland burning for the entireRussian Federation for years 2003 through 2009 utilizing threesatellite fire products; 1 kmMODIS Active Fire Product (Giglio et al.,2003, 2006), 0.5� MODIS Fire Radiative Power (FRP) monthlyclimate modeling grid product (CMG) (Vermote et al., 2009), and500 m MODIS Burned Area Product (Roy et al., 2008), andgovernmental statistics of agricultural productivity.

2. Data and methods

2.1. Observations of burning from satellite data

2.1.1. Fire data2.1.1.1. MODIS Active Fire (MOD14/MYD14). The MODIS sensors,onboard the sun-synchronous polar-orbiting satellites Terra andAqua, acquire four near-global observations daily at 1030 and 2230(Terra) and 0130 and 1330 (Aqua), equatorial local time. The firstMODIS sensor was launched aboard the Terra satellite in 1999; thesecond was launched in 2002 aboard Aqua. The MODIS Level 2 fireproduct is collected daily at 1 km resolution and includes the lati-tude, longitude, FRP, and confidence of the fire detection. For thisresearch, each discrete active fire point was considered to representa burned area of 1 km2, a common assumption that burned area is

proportional to simple counts of fire pixels (Giglio et al., 2006).However, this burned area assumption of 1 km2 was assessed usinga contemporary GIS cropland field mask with a scale of 1:25,000developed by the Soil Dokuchaev Institute for Rostovskaya, Mos-kovskaya, and Kostromskaya oblasts (European Russia). If an activefirewas detected inside the boundary of a field, then thewhole fieldwas considered to have burned completely. If there were severalfire points within one field over a short time period, the fire wasregarded as a single burn event and burned area was the same asthe area of that single field. If the fire point distance was less than500 m from a field, it was considered a single burned field. Theresults of this assessment are further discussed in section 3.6.4Uncertainty in fire products.

2.1.1.1.1. MODIS FRP CMG. The radiant component of combus-tion is emitted as electromagnetic waves traveling at the speed oflight in all directions and is proportional to the absolute tempera-ture of the fire (assumed to be a black body) raised to the fourthpower; otherwise known as the StefaneBoltzmann law. It is theinstantaneous radiative component that is estimated from Earthobserving satellite sensors, referred to as FRP, and offers anapproach to quantify the flux of biomass consumed (Wooster et al.,2005) and, with emission factors, the rate of atmospheric load oftrace gases or aerosols.

Based on simulated fires, Kaufman et al. (1998) demonstratedthat a semi-empirical relationship exists between FRP and pixelbrightness temperature measured by MODIS middle infraredchannel (4 mm). This relationship constitutes the current MODISFRP algorithm.

The MODIS FRP CMG product provides monthly mean FRP, aswell as products describing cloud fraction and corrected pixelcounts, at 0.5� spatial resolution (Giglio, 2010). Monthly mean FRPwas multiplied by cloud-and-overpass-corrected fire pixel count,producing the total FRP released within a given grid cell for eachtime period. Additionally, as described later, monthly CMG datafrom both Terra and Aqua was used to offer a long term charac-terization of the temporal variability between the two satelliteobservations. This variability was then used to parameterize theFRP diurnal cycle and ultimately estimate the integral, otherwiseknown as the fire radiative energy (FRE).

2.1.1.2. MODIS Burned Area (MCD45A1) data. The MODIS BurnedArea product is a 500 m daily product derived from both Terra andAqua detections (Roy et al., 2008). The MODIS Burned Area algo-rithm utilizes changes in spectral, temporal, and vegetation struc-tural characteristics to identify and map burned areas (Roy et al.,2008). ‘Burned’ pixels are identified through dramatic drops andrapid changes in 500 m MODIS daily surface reflectance data.Estimated day of burn are determined by identifying the first dateof a detected decrease in surface reflectance. Additionally statisticaltemporal constraints are employed in this algorithm to identifypersistent burned areas and to omit false detections of burned areacaused by clouds, shadow, dark soils, and water (Roy et al., 2005).

This burned area product was used due to its temporal and itsspatial resolution. As a daily product, emission estimates can beproduced at various time steps. The MODIS Burned Area product’sspatial resolution of 500 m, equivalent to 25 ha, is sufficient foragricultural burning as 500 m has been found to be an appropriatespatial resolution to map burned croplands (McCarty et al., 2008).This product is a standard MODIS product, making it a freelyavailable and well-documented source of burned area from 2001 topresent.

2.1.2. Land cover data2.1.2.1. MODIS 1 km Land Cover (MOD12Q1). The MODIS 1 km LandCover dataset (MOD12Q1) (Friedl et al., 2002) was utilized to define

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238224

Author's personal copy

cropland areas for all three fire analyses. Two classification schemaswere applied to the MODIS 1 km Land Cover dataset to producespatial and/or areal definitions of croplands in Russia: the Inter-national GeosphereeBiosphere Programme (IGBP) land coverclasses (Loveland and Belward, 1997) and the University of Mary-land (UMd) land cover classes (Hansen et al., 2000). The maindifference between these IGBP and the UMd land cover classifica-tion schemas is the presence of mixed pixels, i.e., a mosaic of twoland cover classes. The IGBP classification scheme permits landcovermosaics while the UMd classification scheme aims to produce“pure” pixels by requiring >80% content of one land cover class.Table 1 lists the MODIS land cover classification schemas used inthe analysis.

Several definitions of croplands were used in this analysis todefine the spatial and areal extent of croplands in Russia, includingsingle classification schema classes by themselves and a combina-tion of classification schema classes. For this analysis, the combi-nation of both the IGBP Croplands and Croplands/NaturalVegetation Mosaic classes in the MODIS 1 km Land Cover dataset isreferred to as IGBP-Agriculture, the IGBP Croplands class only isreferred to as IGBP-Croplands, and the UMd classification schema isreferred to as UMd-Croplands.

The FRE Analysis utilized both the IGBP-Croplands and theUMd-Croplands definition of croplands extent. The Active Fire andBurned Area Analysis applied the IGBP-Agriculture and IGBP-Croplands definition of croplands extent to clip MODIS activefire and burned area data to all potential agricultural areas inRussia.

2.1.2.2. MERIS 300 m GlobCover v2.2. The Medium ResolutionImaging Spectrometer Instrument (MERIS) 300 m GlobCover LandCover version 2.2 was used in the Active Fire and Burned AreaAnalyses. GlobCover Land Cover v2.2 was developed jointly by theEuropean Space Agency and MEDIAS-France for time period 2005e2006 (ESA, 2011). GlobCover utilizes the United Nations Food andAgriculture Organization Land Cover Classification System (LCCS)as the classification schema (Di Gregorio and Jansen, 2000). TheLCCS is a systematic, a priori, diagnostic, source- and scale-independent classification system developed to address the needfor a standard classification system. Within the LCCS classificationscheme, land cover classes are defined by multi-level inputs, suchas ancillary data of soils, climate, and vegetation data and/or expertknowledge. Like the IGBP classification scheme, the LCCS schemepermits mosaic classes. Table 2 shows the LCCS classificationscheme for agricultural areas used in MERIS 300 m GlobCover LandCover v2.2.

Two definitions of croplands from the MERIS GlobCover data setwere used to provide additional definitions of the spatial and arealextent of croplands in Russia. The LCCS Post-flooding or IrrigatedCropland and Rainfed Croplands classes were combined to derivean LCCS-Croplands classification A combination of all four LCCS

classes from the MERIS 300 m GlobCover product was combined tocreate an LCCS-Agriculture classification.

2.2. Emission calculations

This analysis employed two emission calculation approaches:the bottom-up approach (section 2.3.1) utilized by the Active Fire,Burned Area, and Agricultural Statistics analyses and the conver-sion of fire radiative power to fire radiative energy (section 2.3.2).Fig. 1 provides an overview of the three bottom-up approaches toBC emission estimations and the fire radiative energy approach toBC emission estimations.

2.2.1. Bottom-up approachThe bottom-up methodology developed by Seiler and Crutzen

(1980) was used to estimate BC emissions from agriculturalburning for the Active Fire Analysis and the Burned Area Analysis.The equation for the bottom-up approach is:

Emissions ¼ A�B�CE � ei (1)

where A is cropland burned area, B is the fuel load variable (mass ofbiomass per area), CE is combustion efficiency (fraction of biomassconsumed by fire), and ei is the emission factor for speciesi (mass ofspecies per mass of biomass burned). For this analysis, all agricul-tural burning in Russia was attributed to grain (wheat, barley, rye)production, which accounts for the majority of planted croplandsacreages (USDA FAS, 2003). Variables B, CE, and ei are explainedfurther in section 2.2.1.1.

The analysis based on official agricultural statistics data useda modified approach developed and published by the All-RussianInstitute of Organic Peat and Fertilizers (Derevyagin, 1987;Rusakova, 2006) to estimate field-level residue loading based onstraw surplus left after utilization as fodder and litter in animalhusbandry, compost production, and for household and industrialuses. The amount of crop straw openly burned was deduced afteraccounting for agricultural management and agrometeorologicalinputs. Equation (2) shows the modified bottom-up emissionsapproach used to calculate cropland burning emissions. Thisapproach will be referred to as the Agricultural Statistics Analysis.

Emissions ¼ C�D�CE � ei (2)

where C is annual straw surplus, D is series of coefficients andcorrection coefficients used to account for agricultural manage-ment and agrometeorological conditions (i.e., temperature andprecipitation) that can impact annual crop yield and thus straw

Table 1MODIS land cover classification schemas and corresponding descriptions.

Land coverclassificationscheme

Class Description

IGBP Croplands Temporary crops followed by harvest and abare soil period (single and multiple croppingsystems). Excludes perennial woody crops.

Croplands/NaturalVegetation Mosaic

Mosaic of croplands, forest, shrublands, andgrasslands; no one component comprises>60% of the pixel.

UMd Croplands >80% of pixel covered in crop-producingfields. Excludes perennial woody crops.

Table 2LCCS land cover classification scheme and corresponding descriptions.

Land coverclassificationscheme

Class Description

LCCS Post-flooding orIrrigated Croplands

Post-flooding or irrigated shrubor tree crops; Post-flooding orirrigated herbaceous crops.

Rainfed Croplands Rainfed herbaceous crops; Rainfedshrub or tree crops (cash crops,vineyards, olive tree, orchards.).

Mosaic Cropland (50e70%) /Vegetation (20e50%)

Mosaic cropland (50e70%) /grassland or shrubland (20e50%);Mosaic cropland (50e70%) /forest (20e50%).

Mosaic Vegetation (50e70%) /Cropland (20e50%)

Mosaic grassland or shrubland(50e70%) / cropland (20e50%);Mosaic forest (50e70%) /cropland (20e50%).

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 225

Author's personal copy

production (i.e., crop residues), CE is combustion efficiency, and ei isthe emission factor for speciesi. C is calculated from grain yieldstatistics within individual federal administrative units accordingto a methodology published in Derevyagin (1987). Equation (3)outlines how variable D was derived.

D ¼�12ðd1þ d2Þ=d3

�*d4*d5*d6 (3)

Variables d1 through d3 are oblast-level indicators of agricul-tural enterprise activity in comparison with circa 1990 levels,where d1 is diesel fuel supply, d2 is mineral fertilizers supply, andd3 is arable area dynamics. Variables d4 through d6 representcorrection indicators, where d4 is cropping patterns, d5 is avail-ability and renovation of harvesters (i.e., straw choppers), and d6 isaverage agrometeorological indicators for climatic zones. All indi-cators were derived from agricultural statistics within individualfederal administrative units. In the absence of specific data aboutfield burning of crop residues, di indicators were correcteduniformly for all federal units to reproduce observed dynamics ofstraw burning according to personal communications with localexperts from agricultural institutes and studies of local officialpermission of crop straw openly burned in the field.

2.2.1.1. Input variables: fuel loads, combustion efficiency, and emis-sion factors. Average annual fuel load values for the bottom-upapproach used in the Burned Area Analysis were derived fromgrain yield statistics collected from the Federal State StatisticsService database from each grain producing oblast in Russia andthen applied for the country-level emissions. Fuel loadings forcropland burning in the U.S. are often calculated by multiplyingcrop yield by the yield-to-residue ratio and the dry matter ofresidue statistics (U.S. EPA, 2008), however those statistics aremissing for Russia. Therefore, the Burned Area Analysis in thispaper utilized the grain yield statistics as a proxy for fuel load.Table 3 shows the average annual national fuel loads used in theBurned Area Analysis. The Active Fire Analysis, completed byscientists at the Dokuchaev Soil Institute in Moscow, utilized theoblast-specific fuel loadings from the official statistics and assumeda yield-to-residue coefficient factor of 1.35 to estimate fuel loads.

Without Russia-specific values, the combustion efficiency valuewas compared to wheat combustion efficiency value of 0.85 from

wheat residues in the contiguous U.S. (McCarty, 2011), derived fromexpert knowledge during U.S. fieldwork. The Active Fire Analysisutilized a combustion efficiency value of 0.80 while the BurnedArea Analysis used a combustion efficiency value of 0.85.

Black carbon emission factors were derived from four sources:Andreae and Merlet (2001), Turn et al. (1997), McCarty (2011), andpersonal communication from Dr. Zbignew Klimont at the Inter-national Institute for Applied Systems Analysis (2010). The BCemission factor from McCarty (2011) was derived from a PM2.5

emission factor for wheat in the U.S. of 4.0 g kg�1. An 11.5% BCcontent of PM2.5 was assumed, based on research from Liousse et al.(1996). Table 4 shows the emission factors and the average emis-sion factor utilized in this analysis. The Active Fire and AgriculturalStatistics Analyses utilized the BC emission factor derived from theMcCarty (2011) PM2.5 emission factor and the Burned Area Analysisutilized all emission factors.

2.2.2. Fire radiative power to fire radiative energyFire radiative energy (FRE), measured in units of mega joules

(MJ), is the integration of FRP over time and space providinga metric to estimate the total fuel consumed and emissionsreleased. Indeed, an evaluation of traditional methods to estimateemissions reveals that having accurate estimates of FRE one canreplace the variables A, B, and CE in Equation (1) (i.e. mass of fuelconsumed) with the amount of energy emitted during a fire event.However, a current limitation in retrieving FRE from satellites is

Fig. 1. Basic workflow of the four BC emission calculations utilized in this analysis. The calculations are grouped in color by the approach used to calculate the emissions: the fireradiative energy approach is displayed in light grey, the bottom-up emission approach is displayed in white, and the modified bottom-up emission approach is shown in dark grey.

Table 3Average annual fuel load values (kg/ha) for the RussianFederation used in the bottom-up approach emissioncalculations for the Burned Area Analysis; official grain yieldstatistics derived from the Federal State Statistics Servicedatabase of the Russian Federation.

Year Fuel load value (kg/ha)

2003 15402004 18902005 18802006 19002007 20202008 23802009 2270

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238226

Author's personal copy

the resolution (both space and time) and the dynamic nature offires. Ellicott et al. (2009) developed a unique approach toparameterize the temporal trajectory of FRP and calculate theintegral (FRE) using the MODIS CMG product (Giglio, 2010). Theparameterization is based on the long term ratio between Terraand Aqua MODIS FRP and diurnal measurements of FRP made bysatellites with greater temporal resolution. This included thegeostationary sensor Spinning Enhanced Visible and InfraredImager (SEVIRI) and Visible and Infrared Scanner (VIRS) aboardthe Tropical Rainfall Measuring Mission (TRMM). VIRS’s low-inclination orbit (35�) provides observing times which processthrough 24 h of local time every 23e46 days, depending on lati-tude, thus capturing the general diurnal trend of fire activity. Inaddition, high latitude (and thus high overpass frequency) dailyobservations were included. A full description of the approach canbe found in Ellicott et al. (2009) and Vermote et al. (2009). Furtherdescription of this data product, including atmospheric correc-tion, is included in Section S.1 of the Supplemental Materials.

Using the aforementioned approach, FRE estimates were madefrom the MODIS CMG standard product at 0.5� spatial and monthlytemporal resolution for 2003e2009.

2.2.2.1. Emission factors. Emission factor s are defined as theamount (in grams) of gas or particulate matter emitted per unit offuel combusted (in kilograms). However, as presented by Ichoku

and Kaufman (2005), although FRE may serve as a proxy for themass of fuel consumed from biomass burning, any relationshipwith emissions released should be distinguished from traditionalmethods (i.e. Equation (1)); thus, the term emission coefficient (EC)is used to define themass of emissions per unit of energy, or g MJ�1.

In order to establish the EC for black carbon, we used an organicand black carbon aerosol emissions product developed frominversion modeling that accounts for atmospheric processes(transport, deposition, etc.) and derives the aerosol emissions usingsatellite observations. Specifically, the aerosol source informationretrieved by the Dubovik et al. (2008) algorithm from MODIS finemode aerosol AOT measurements was used to estimate biomassburning organic and black carbon. The emission coefficient (ECocbc)was then established by relating the OCBC inversion product esti-mates with FRE for three specific land cover types. The land covertypes (tropical forest, extratropical forest, and non-forest) weretaken from the Global Fire Emissions Database, version 2 (GFEDv2,van der Werf et al., 2006) and chosen to offer continuity andcomparison with the GFEDv2; arguably the most cited biomassburning emissions database currently. Fig. 2 shows the spatialextent of the land cover classes used.

The emission coefficients for non-forest (which includes agri-cultural lands), tropical forest, and extratropical forest were 2.47,7.54, and 11.45 g MJ�1, respectively, and are for organic and blackcarbon combined. The OC-to-BC ratio of 71%e29% from Novakovet al. (2005) was applied to isolate BC emissions only from theFRE emission calculations. Thus, the emission coefficients for BCcan be translated to 0.74, 2.24, and 3.44 g MJ�1 for non-forest,tropical forest, and extratropical forest, respectively.

2.3. Source region and federal subject attribution

Cropland burning and related BC emissions within individualfederal administrative units were identified using a vector GISshapefile of Russian federal subjects, including federal cities,

Table 4Black carbon emission factors utilized in the bottom-up approach; all emissionfactors reported in g/kg.

Source Emission factor (g/kg)

Turn et al. (1997) 0.79Andreae and Merlet (2001) 0.69McCarty (2011) 0.46IIASA (2010) 0.83Average 0.69

Fig. 2. Global land cover types taken from the Global Fire Emissions Database version 2 (GFEDv2; van der Werf et al., 2006).

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 227

Author's personal copy

republics, krais, oblasts, autonomous oblast, and autonomousokrugs, from the Dokuchaev Soil Institute (Moscow). Sourceregions of burning and emissions were also identified. For theburned area and FRE analyses, the Russian Federation was dividedinto four macro-regions: European Russia, West Siberia, EastSiberia, and the Far East (Fig. 3).

3. Results

3.1. Emissions from Active Fire Analysis

3.1.1. Annual and interannual emissionsIn general, the highest annual BC emissions calculated from

the Active Fire Analysis occurred in 2008 and 2009 while thelowest BC emissions occurred in 2003 during the 7-year studyperiod (Table 5). For years 2003 through 2009, LCCS-Croplandschema represented 36%, 37%, and 39% of the average annualBC emissions from the IGBP-Agriculture, IGBP-Croplands, and theLCCS-Agriculture definitions. Excluding the LCCS-Croplandsschema, the annual and average annual BC emissions werewithin less than �1.3 Gg.

3.1.2. Monthly and seasonal emissionsMonthly variability of active fire counts for Rostovskaya oblast

(European Russia) and Omskaya oblast (West Siberia) utilizing theIGBP-Cropland schema illustrates the two distinct patterns ofcropland burning emissions typical for the European and Asianparts of the country, respectively (Fig. 4). The majority of croplandburning BC emissions in Rostovskaya oblast occurred in August andJuly, during or immediately after harvesting of cereals, except foryear 2003which showed highest burning in April. This correspondswell with MODIS monthly time series of agricultural fires fromsouthern European Russia for years of 2001e2003 noted by

Korontzi et al. (2006), including low August burning in 2003. Atypical pattern of cropland burning in West Siberia is illustrated byOmskaya oblast. Peak cropland burning occurred fromApril toMay,with the highest values in 2006, 2008, and 2009. The second peakin October was much lower, with the strongest contributions from2003 and 2007.

The seasonal distribution of cropland burning in for Rostovskayaand Omskaya oblasts also varied. Throughout this analysis andmanuscript, winter is defined as January through March, spring isApril through June, summer is July through September, and fall isOctober through December, respectively. The peak fall burning inRostovskaya oblast occurred in 2005 and 2006 with the leastamount of fall burning in 2003. Springtime burning was a factor of4e9 times less than fall burning, with small peaks observed in2005, 2007, and 2009. For Omskaya, springtime burning wasmost common, with only small contribution from fall burning,specifically in years 2003 and 2007.

Table 6 shows the seasonal BC emissions for the Russian Feder-ation from cropland burning using the four cropland definitions. Onaverage, burning observed within the largest cropland area, theIGBP-Agriculture, had aminorwinter contribution of total annual BCemissions (less than 0.5%), a spring contribution of 55%, a summercontribution of 25%, and a fall contribution of 19%. Interannualvariations of spring, summer, and fall BC emissions were 33e70%,9e38%, and 11e33%, respectively. Average 2003e2009 seasonaldistributions of BC emissions from the IGBP-Croplands and LCCS-Agriculture schemas were similar, with a slightly higher summercontribution (2e3%) at the expense of lower spring BC emissions.The overall contribution of fall burning emissions calculatedfrom the IGBP-Agriculture, IGBP-Croplands, and LCCS-Agricultureschemas were very similar. The smallest definition of croplands,LCCS-Croplands, produced comparatively higher summer andsmaller spring BC emissions estimates (38% and 39%, respectively)

Fig. 3. Macro-regions of the Russian Federation.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238228

Author's personal copy

with 21% of annual BC emissions coming from fall burning, which issimilar to BC emission distributions for the same land cover fromBurned Area Analysis.

3.1.3. Source areas for BC emissions from cropland burning basedon Active Fire Analysis

Federal subjects for each year that account for greater than orequal to 1% of the total annual BC emissions from cropland burning

Table 5Annual, total, and annual average BC emissions from cropland burning in the Russian Federation as calculated by the Active Fire, Agricultural Statistics, FRE, Burned Area, andGFEDv3 approaches; emissions reported by five different definitions of croplands: IGBP-Agriculture¼ IGBP croplands and cropland/natural vegetation mosaic classes from theMODIS 1 km Land Cover data set, IGBP-Croplands ¼ IGBP croplands class from the MODIS 1 km Land Cover data set, UMd-Croplands ¼ UMd cropland class from the MODIS1 km Land Cover data set, LCCS-Agriculture¼ UN LCCS rainfed, irrigated croplands, cropland/natural vegetation mosaics, and natural vegetation/croplandsmosaic classes fromthe MERIS 300 m GlobCover data set, and LCCS-Croplands¼ UN LCCS rainfed and irrigated croplands classes from the MERIS 300 m GlobCover data set; all emissions reportedin Gg.

Year Active fire analysis Agricultural statistics FRE Burned area analysis GFEDv3

IGBP-agriculture

IGBP-croplands

LCCS-agriculture

LCCS-croplands

Straw usagefrom officialstatistics

IGBP-croplands

UMd-croplands

IGBP-agriculture

IGBP-croplands

LCCS-agriculture

LCCS-croplands

UMd-croplands

2003 4.82 4.55 3.97 1.43 6.59 21.1 15.7 6.93 6.58 5.45 1.63 18.62004 6.84 6.49 6.55 2.35 7.27 20.7 17.3 6.95 6.64 5.99 1.38 4.42005 7.67 7.52 6.92 3.08 7.51 27.9 23.7 9.13 9.00 8.40 2.32 7.82006 9.22 8.74 8.84 3.11 8.40 23.2 15.3 11.5 11.0 10.6 2.24 9.52007 6.82 6.65 6.04 2.64 9.08 16.2 14.8 11.4 11.2 9.32 2.53 11.22008 13.97 13.51 12.61 5.31 12.11 27.1 22.9 22.7 21.9 18.5 4.61 21.92009 11.25 10.71 10.41 3.81 11.32 19.0 16.6 13.5 12.9 11.2 2.74 9.6Total 66.80 63.99 61.64 23.61 62.27 155 126 82.1 79.1 69.5 17.5 83.0Annual

average8.35 8.00 7.70 2.95 8.90 22.2 18.0 11.7 11.3 9.93 2.49 11.9

a

b

Fig. 4. Monthly variability of BC emissions as detected by the MODIS active fire product for years 2003e2009; emissions variability reported for land cover schema: IGBP-Croplands ¼ IGBP croplands class from the MODIS 1 km Land Cover data set; (a) monthly variability for Rostovskaya oblast, European Russia; (b) monthly variability forOmskaya oblast, West Siberia.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 229

Author's personal copy

calculated from the Active Fire approach, with the macro-region foreach federal subject identified (Fig. 3), are reported inSupplemental Materials section S.2 and Table S1. More than 60% ofall BC emissions, an average of 64% for all years, occurred in Euro-pean Russia. In a given year between 2003 and 2009, approximately50%e79% of all federal subjects with greater than 1% of total BCemissions were within European Russia. On average, West Siberiawas the source of 20% of total cropland burning BC emissions, FarEast Russia accounted for 9%, and East Siberia accounted for 7%.

3.2. Emissions from the Agricultural Statistics Approach

3.2.1. Annual and interannual emissionsThe total straw surplus was calculated at 47.4e85.7 Tg and straw

burning consumption is estimated at18.8e34.4 Tg for 2003 through2009, representing 22e27% of the total output crop straw,respectively. The straw burning consumption range has beenquite consistent based on calculations of Russian Federationagricultural statistics since 1997 and is very similar to regionalestimates in China of straw burning, which were 23% of totalstraw output for years 2000e2003 (Cao et al., 2008). Table 5shows the annual BC emissions calculated from this approach. On

average, 8.90 Gg of BC emissions was emitted from crop residueburning in the Russian Federation. The interannual variation of BCemissions calculated from the Agricultural Statistics Approach wason average � 1.33 Gg - less than 15% variation.

3.2.2. Source areas for BC emissions from cropland burning basedon the Agricultural Statistics Approach

The largest contribution is European Russia (with a range of69e78% of total burning between 2003 and 2009) with less than20% of emissions cumulatively from West and East Siberia. Thefederal subjects with the highest BC emission estimates withgreater than 3% of total BC emissions are listed in theSupplemental Materials as Section S.3 and Table S2. On average,European Russia was the source for 78% of total cropland BCemissions from 2003 to 2009, West Siberia accounted for 21%, EastSiberia accounted for 3%, and Far East Russia accounted for 0%.

3.3. Emissions from FRE

3.3.1. Annual and interannual emissionsAnnual totals of BC emissions from cropland burning classified

calculated from FRE estimations for two land cover classificationschemas are presented in Table 5. In general, the IGBP-Croplandsschema produced higher FRE emissions than the UMd-Croplandsschema. Both of these FRE BC emission estimations peaked inyears 2005 and 2008. The 2005 IGBP-Croplands BC emissions(27.9 Gg) were 1.26 times larger and the 2008 IGBP-Croplands BCemissions (27.1 Gg) were 1.22 times larger than the average BCemissions for all years (22.2 Gg). Similarly, the 2005 UMd-Croplands BC emissions (23.7 Gg) were 1.32 times larger and the2008 UMd-Croplands BC emissions (22.9 Gg) were 1.27 times largerthan the average BC emissions for all years (18.0 Gg) (Table 5).

A time series distribution of fire radiative energy is expressed inFig. 5 which highlights both the extent and intensity of firesannually for Russia regardless of land cover type. Clearly 2003 and2008 stand out as years containing many energetic fire events. TheFRE maps show that southeast Russia is the area that generally seesthe most intense fires.

3.3.2. Monthly and seasonal emissionsThe FRE emission estimates show three peaks of BC emissions

during spring (April), summer (July and August), and fall (October)(Fig. 6). The April peak of emissions was generally higher acrossboth land cover classification schemas, with 2008 being the highestApril BC emissions under the IGBP-Croplands classification schemaand the second highest under the UMd-Croplands classificationschema. When cropland extent was defined by the IGBP-Croplandsclassification, October BC emissions were generally higher thanAugust BC emissions. The UMd-Croplands classification schemaproduced near equivalent BC emissions for both August andOctober. Both classification schemas showed high October BCemissions for year 2005 and low August BC emissions in 2003.

Seasonal BC emissions from cropland burning in Russia weregenerally highest in spring and fall (Table 7). On average, the twocropland definitions for the FRE analysis had similar seasonal BCemissions in terms of percentage of total annual BC emissions. TheIGBP-Croplands schema produced an average seasonal contributionof BC emissions of 7% during winter, 45% during spring, 29% duringsummer, and 20% during fall. The UMd-Croplands schema producedan average seasonal contribution of BC emissions of 9% duringwinter, 39% during spring, 32% during summer, and 20% during fall.The main seasonal emissions difference was springtime burning,which ranged from 45% to 39% of total cropland burning BC emis-sions. Under both schemas, average spring BC emissions were thelargest proportion of BC emissions.. On an annual basis, 2003, 2004,

Table 6Average seasonal BC emissions in the Russian Federation from cropland burning ascalculated from the Active Fire Analysis; emissions reported by four definitions ofcroplands: IGBP-Agriculture ¼ IGBP croplands and cropland/natural vegetationmosaic classes from the MODIS 1 km Land Cover data set, IGBP-Croplands ¼ IGBPcroplands class from the MODIS 1 km Land Cover data set, LCCS-Agriculture ¼ UNLCCS rainfed, irrigated croplands, cropland/natural vegetation mosaics, and naturalvegetation/croplands mosaic classes from the MERIS 300 m GlobCover data set, andLCCS-Croplands ¼ UN LCCS rainfed and irrigated croplands classes from the MERIS300 m GlobCover data set; all emissions reported in Gg.

Year Season IGBP-agriculture

IGBP-croplands

LCCS-agriculture

LCCS-croplands

2003 Winter 0.02 0.02 0.01 0.01Spring 2.95 2.74 2.08 0.67Summer 0.59 0.56 0.57 0.25Fall 1.27 1.23 1.31 0.50

2004 Winter 0.01 0.01 0.01 0.00Spring 3.34 3.01 3.23 0.72Summer 2.43 2.42 2.44 1.23Fall 1.06 1.05 0.88 0.40

2005 Winter 0.11 0.11 0.10 0.05Spring 2.05 1.95 1.57 0.48Summer 3.24 3.23 3.26 1.65Fall 2.27 2.23 1.99 0.90

2006 Winter 0.01 0.01 0.01 0.00Spring 5.22 4.79 4.91 1.24Summer 2.57 2.54 2.65 1.26Fall 1.41 1.39 1.28 0.61

2007 Winter 0.08 0.08 0.07 0.03Spring 2.44 2.32 1.71 0.61Summer 3.09 3.06 3.22 1.53Fall 1.22 1.19 1.04 0.46

2008 Winter 0.08 0.08 0.07 0.03Spring 5.77 5.36 4.74 1.15Summer 5.59 5.57 5.49 2.96Fall 2.53 2.51 2.31 1.17

2009 Winter 0.03 0.03 0.02 0.01Spring 7.45 6.94 6.71 1.91Summer 2.34 2.32 2.34 1.25Fall 1.43 1.42 1.34 0.65

Average Winter 0.05 0.05 0.04 0.02Spring 4.17 3.87 3.56 0.97Summer 2.84 2.81 2.85 1.45Fall 1.60 1.57 1.45 0.67

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238230

Author's personal copy

2006, and 2009 produced higher spring cropland burning BCemissions. Years 2005, 2007, and 2008 produced higher fall crop-land burning BC emissions. Years 2005 and 2008 also produced thehighest and second highest annual BC emissions, respectively, while2007 produced the lowest annual BC emissions.

3.3.3. Source areas for BC emissions from cropland burning basedon FRE results

The BC emissions from the FRE approach were calculated foreach federal subject, i.e. political administrative unit, for each year.Section S.4 and Table S3 in the Supplemental Materials list thefederal subjects for each year that account for greater than or equalto 1% of the total annual BC emissions from cropland burning. Themajority of BC emissions, an average of 70% for all years, from theFRE analysis occurred in European Russia. In each year, approxi-mately 60%e75% of all federal subjects with greater than 1% of totalBC emissions from cropland burning were in European Russia. Onaverage, West Siberia was the source of 19% of total croplandburning BC emissions, Far East Russia accounted for 10%, and EastSiberia accounted for 1%.

3.4. Emissions from Burned Area Analysis

3.4.1. Annual and interannual emissionsAnnual totals of BC emissions from cropland burning classified

by the different land cover data are presented in Table 5, wherebythe emissions were calculated using the average emission factor

(Table 4). Emissions from cropland burning in 2008 were found tobe 52e54% higher than the average emission values from 2003through 2009 depending on the land cover data set and definitionof ‘croplands’ for each land cover data set. Both the seven yeartotal BC emissions and the average annual BC emission werewithin �2 Gg when burned area data was clipped to the IGBP-Agriculture, IGBP-Croplands, and LCCS-Agriculture schemas(Table 5). The seven year total BC emissions and the average annualBC emission derived from burned area data clipped to the LCCS-Croplands schema in Russia was much lower, representing 21% ofthe highest seven year total BC emissions and the average annual BCemissions calculated using the IGBP-Agriculture schema.

3.4.2. Monthly and seasonalA consistent bi-modal distribution of BC emissions during

a late winter-early spring (March through May) and late summer-early fall (July through September) was produced using the fourdefinitions of cropland burning across the four classificationschemas (Fig. 7). Seasonal BC emissions from cropland burningin Russia were generally highest in spring and fall (Table 8). Allfour cropland definitions, from IGBP-Agriculture to LCCS-Croplands, had similar seasonal BC emissions in terms ofpercentage of total annual BC emissions. The strictest (smallest ingeographical extent and area) definition of croplands in Russia,LCCS-Croplands, produced an average winter contribution of BCemissions equal to 2%, spring contribution of 34%, summercontribution of 43%, and fall contribution of 21%, respectively. The

Fig. 5. Time series of intensity (FRE) and extent of vegetative biomass burning for all land cover types in the Russian Federation for years 2003e2009. FRE was estimated asmegajoules (MJ) per 0.5� pixel using the MODIS FRP climate modeling grid (CMG) standard product.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 231

Author's personal copy

LCCS-Agriculture definition had an average winter contribution of2%, spring contribution of 39%, summer contribution of 41%, anda fall contribution of 19%, respectively. IGBP-Croplands definitionproduced an average winter contribution of 4%, spring contribu-tion of 39%, summer contribution of 38%, and a fall contributionof 19%, respectively. The largest cropland definition in terms ofgeographical extent and area, IGBP-Agriculture class, produced anaverage winter contribution of 4% of total annual BC emissions,spring contribution of 41%, summer contribution of 37%, and a fallcontribution of 18%, respectively. The IGBP-Agriculture and IGBP-Croplands definitions showed slightly higher winter contributionof BC emissions (þ2%) and a slightly lower or equal fall contri-bution of BC emissions (�1%). As the cropland definitionincreased the geographical extent and area (from LCCS-Croplandsto IGBP-Agriculture, respectively), the spring emissions increasedby 7% (from 34% to 41%, respectively) and the summer emissionsdecreased by 6% (from 43% to 37%, respectively). Each of thesecropland definitions showed that spring and summer BC emis-sions from cropland burning in Russia, i.e., emissions occurringfrom April to September, accounted for more than 75% of the totalannual BC emissions.

Fig. 8 shows the spatial distribution of seasonal croplandburning in Russia using the IGBP-Croplands cropland definition.Cropland burning duringwinter monthswas focused in East Siberiaand Far East Russia for years 2005, 2008, and 2009 while 2007

showed a lot of winter cropland burning in European Russia. Springburning is spread across all cropland regions in Russia, with strongcontributions from European Russia in 2006 and 2009 and fromWest and East Siberia for years 2003, 2006, 2007 and 2008. In allyears, summer burning was concentrated in European Russia andWest Siberia though in 2005 and 2006 Far East Russia experiencedsome summer cropland burning.

3.4.3. Source areas for BC emissions from cropland burningSection S.5 and Table S4 in Supplemental Materials list the

federal subjects that account for greater than or equal to 1% ofthe total annual BC emissions. Approximately half of all BCemissions occurred annually in European Russia. On average,West Siberia was the source of 22% of total cropland burning BCemissions, Far East Russia accounted for16%, and East Siberiaaccounted for 10%.

3.5. Comparison with the Global Fire Emissions Database, version 3

A comparison of BC emissions with the recently released GlobalFire Emissions Database, version 3 (GFEDv3) developed by van derWerf et al. (2010) was performed. GFEDv3 is a robust and compre-hensive inventory based, bottom-up global emissions database. Theintention of this comparison was to assess the agreement of BCemissions across various methodologies and data inputs.

IGBP-Croplands

0

5

10

15

20

25

30

35

40

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Bla

ck c

arbo

n em

issi

ons

(Gg)

2003

2004

2005

2006

2007

2008

2009

UMd-Croplands

0

5

10

15

20

25

30

35

40

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Bla

ck c

arbo

n em

issi

ons

(Gg)

2003

2004

2005

2006

2007

2008

2009

b

a

Fig. 6. Monthly variability of BC emissions calculated from 0.5� MODIS FRE using the MODIS climate CMG standard product for years 2003e2009; emissions reported by twodefinitions of croplands: IGBP-Croplands ¼ IGBP croplands class from the MODIS 1 km Land Cover data set and UMd-Croplands ¼ UMd croplands class from the MODIS 1 km LandCover data set; all emissions reported in Gg.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238232

Author's personal copy

The GFED mean BC emissions for Russia agriculture wasteresidue burning, one of several partitioned categories within theGFEDv3, averaged 11.9 Gg per annum for 2003-2009. The peak yearwas 2008 at 21.9 Gg and minimum in 2004 at 4.4 Gg. The generaltrend mirrors that of the emission estimates using the officialMODIS burned area product with both the MODIS and MERIS landcover products (Fig. 9), with 2003 being the exception in which theGFEDv3 approaches the FRE-based estimate (18.6 and 21.1 Gg,respectively). Between 2004 and 2006 the GFEDv3 shows thelowest BC emissions of the data presented, but relatively minordifferences with the different land cover schemas used in theBurned Area Analysis. In 2007, the GFEDv3 and MODIS Burned AreaProduct (IGBP-Agriculture) numbers are nearly identical, while theLCCS-Agriculture BC estimate is slightly less, and the FRE-basedIGBP BC emission estimate is about 40% more than the GFEDv3.From 2007 to 2009 the trends for all of the datasets presentedclosely match each other, with the FRE-based numbers again beingthe largest estimates and the LCCS-Agriculture (Fig. 9a) and LCCS-Croplands (Fig. 9b) generally being the lowest.

3.6. Uncertainty analysis of input variables

3.6.1. Emission factorsThe BC emission factors used to calculate the average emission

factor (Table 4) utilized by the Burned Area and Active FireAnalyses varied only by �0.12 g kg�1, or 17% of the average

Table 7Average seasonal BC emissions in the Russian Federation from cropland burningcalculated from 0.5� MODIS FRE using the MODIS climate CMG standard product;emissions reported by two definitions of croplands: (a) IGBP-Croplands ¼ IGBPcroplands class from the MODIS 1 km Land Cover data set and (b) UMd-Croplands ¼ UMd croplands class from the MODIS 1 km Land Cover data set; allemissions reported in Gg.

Year Season IGBP-croplands UMd-croplands

2003 Winter 13.0 13.1Spring 34.0 25.5Summer 9.18 7.12Fall 16.6 8.49

2004 Winter 1.23 1.22Spring 39.3 29.2Summer 17.8 16.8Fall 13.2 12.4

2005 Winter 5.45 5.49Spring 22.0 17.5Summer 28.9 26.3Fall 39.9 32.4

2006 Winter 1.48 0.98Spring 45.9 26.8Summer 21.5 18.3Fall 11.2 6.59

2007 Winter 8.25 8.54Spring 12.5 10.5Summer 21.5 20.4Fall 13.8 11.5

2008 Winter 3.35 6.18Spring 44.7 26.7Summer 37.3 34.9Fall 8.08 11.2

2009 Winter 2.17 2.13Spring 40.2 33.2Summer 16.8 15.8Fall 6.33 6.20

Average Winter 4.98 5.38Spring 34.1 24.2Summer 21.9 19.9Fall 15.6 12.7

0

0.5

1

1.5

2

2.5

3

3.5

4

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Blac

k ca

rbon

em

issi

ons

(Gg)

LCCS-Croplands

2003200420052006200720082009

0

1

2

3

4

5

6

7

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Blac

k ca

rbon

em

issi

ons

(Gg)

LCCS-Agriculture

2003200420052006200720082009

0

1

2

3

4

5

6

7

8

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Blac

k ca

rbon

em

issi

ons

(Gg)

IGBP-Croplands

2003 2004 2005 2006 2007 2008 2009

0

1

2

3

4

5

6

7

8

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Blac

k ca

rbon

em

issi

ons

(Gg)

IGBP-Agriculture

2003 2004 2005 2006 2007 2008 2009

a

b

c

d

Fig. 7. Monthly variability of BC emissions as detected by the MODIS Burned AreaProduct (MCD45A1) and calculated from the average emission factor by different landcover datasets for years 2003e2009; (a) LCCS-Croplands ¼ UN LCCS rainfed and irri-gated croplands classes from the MERIS 300 m GlobCover data set, (b) LCCS-Agriculture ¼ UN LCCS rainfed, irrigated croplands, cropland/natural vegetationmosaics, and natural vegetation/croplands mosaic classes from the MERIS 300 mGlobCover data set, (c) IGBP-Croplands ¼ IGBP croplands class from the MODIS 1 kmLand Cover data set, and (d) IGBP-Agriculture ¼ IGBP croplands and cropland/naturalvegetation mosaic classes from the MODIS 1 km Land Cover data set; all emissionsreported in Gg.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 233

Author's personal copy

emission factor. The lowest BC emission factor - 0.46 g kg�1

(McCarty, 2011) - was 33% lower than the average emission factorwhile the highest BC emission factor - 0.83 g kg�1 (IIASA, 2010) -was 20% higher. Additionally, the assumption of 11.5% BC contentof PM2.5 (Liousse et al., 1996) is lower than the range of BC contentreported by Turn et al. (1997) of 15.6%e16.4%.

The FRE-based emission coefficients can be converted to tradi-tional emission factors (i.e. g kg�1), for comparison sake, using anaverage combustion factor of 0.411 kg MJ�1 taken from research byFreeborn et al. (2008) and Wooster et al. (2005). The resulting BCemission factors are then 0.31, 0.93, and 1.43 g kg�1 for non-forest,tropical forest, and extratropical forest, respectively. Comparedwith Andreae and Merlet (2001) emission factors, the FRE-basedvalues range from a factor of 1.6e5.3 greater and clearlyintroduce a range of uncertainty to the FRE-based emissionresults. The OC to BC ratio used of 7:1 is only applicable to theagriculture residue burning and when comparing the non-forestland cover BC emission factor (1.6 g kg�1) and agriculture residueburning EF (0.69 g kg�1) from Andreae and Merlet (2001), thenumbers are in closer agreement. Applying combustion factorsgenerated by Kaiser et al. (2011, Table 2) for savanna (non-forest),

tropical forest, and extratropical forest to calculate FRE-basedemission factors results in EFs that more closely resembleAndreae and Merlet (2001), ranging from a factor 0.9 to 1.7 differ-ence. When we apply the agriculture conversion factor from Kaiseret al. (2011) to generate an FRE-based emission factor for agricul-ture residue burning, the result is an EF 1.5 times greater than re-ported by Andreae and Merlet (2001). This latter comparison of EFsusing conversion factors from Kaiser et al. (2011) is at best prob-lematic because they used the GFEDv3.1 land cover classes whileour emission coefficients were generated from the relationshipbetween FRE and aerosol observations segmented using land coverclasses defined in the GFEDv2. Using a constant conversion ratio forparsing out BC from an OCBC (i.e. 7:1) estimate may also introduceadditional error.

3.6.2. Fuel loads and combustion completenessThe Burned Area and Active Fire Analyses relied on fuel loadings

based on yield statistics from the Russian Federation anda combustion completeness value from research in the contiguousU.S. Without sampled fuel loadings and lab- and/or field-observedcombustion completeness, these variables remain uncertain butrepresent best available estimations for calculation crop residueemissions. The Agricultural Statistics method relied on severalyears of known crop straw usage (including burning) from years ofagricultural management and agrometeorological research. For thisreason, the fuel and combustion completeness variables from theAgricultural Statistics Approach serve as a proxy for in-situmeasurements - which has not been performed to date - andlikely possess less uncertainty than the remote sensing methods.

Although fuel load is unnecessary for FRE-based estimates of BCemissions, it can be calculated for comparison with the bottom-upapproach. The dry biomass (kg) must first be calculated by using anaverage combustion factor (kg MJ�1) multiplied by FRE (MJ) whichyields fuels consumed (kg). This is then divided by the product ofarea burned (ha) and combustion completeness (or efficiency). TheFRE-based fuel loads are on average over a factor of 2 less than theagricultural statistics numbers (Table 9). The difference is due to: 1)uncertainty in the combustion factor which has a relative error of10% and; 2) underestimation of FRP and, correspondingly, FRE. If weuse the combustion factor presented by Kaiser et al. (2011) our fuelload estimates nearly double (factor of 1.9) and are therefore muchcloser to the agricultural statistics. The error in the empirical FRPformula, atmospheric attenuation, and the error in the FREparameterization all contribute to uncertainty in biomass com-busted which is nearly 34% (Vermote et al., 2009). A similar bias(w24%) in FRE estimates produced from MODIS was found byFreeborn et al. (2011).

3.6.3. Uncertainty in land cover productsLand use and land cover in Russia experienced the largest

change during the 20th century for the northern hemisphere whenafter 1990 approximately 43 million ha of agricultural lands(including 30.2 million ha of arable land) were abandonedaccording to official agricultural statistics (Lyuri et al., 2010). Withthe absence of biomass removal and cultivation, any fire occurringon these abandoned lands is generally viewed as open and/orwildland burning but not cropland burning. The large-scale agri-cultural land abandonment combined with the lack of satellite-based land cover product validation and annual updates of theseland cover maps is an important methodological challenge toinvestigate crop burning in Russia. Crop yields in the RussianFederation are not uniform and agricultural practices and economicincentives for farmers vary from the federal to local levels. Toimprove upon the BC emission estimates of the AgriculturalStatistics approach, official data on total amount of crop straw

Table 8Average seasonal BC emissions in the Russian Federation from cropland burning asdetected by the MODIS Burned Area Product (MCD45A1) and calculated from theaverage emission factor; emissions reported by four definitions of croplands: IGBP-Agriculture ¼ IGBP croplands and cropland/natural vegetation mosaic classes fromthe MODIS 1 km Land Cover data set, IGBP-Croplands ¼ IGBP croplands class fromthe MODIS 1 km Land Cover data set, LCCS-Agriculture ¼ UN LCCS rainfed, irrigatedcroplands, cropland/natural vegetation mosaics, and natural vegetation/croplandsmosaic classes from the MERIS 300 m GlobCover data set, and LCCS-Croplands ¼ UNLCCS rainfed and irrigated croplands classes from the MERIS 300 m GlobCover dataset; all emissions reported in Gg.

Year Season IGBP-agriculture

IGBP-croplands

LCCS-agriculture

LCCS-croplands

2003 Winter 1.10 1.04 0.53 0.15Spring 3.50 3.22 2.65 0.90Summer 0.80 0.78 0.78 0.22Fall 1.70 1.69 1.66 0.43

2004 Winter 0.05 0.04 0.00 0.00Spring 3.35 3.12 2.86 0.57Summer 2.21 2.20 2.10 0.52Fall 1.58 1.50 1.30 0.37

2005 Winter 0.19 0.17 0.06 0.03Spring 1.64 1.59 1.23 0.36Summer 4.25 4.24 4.25 1.05Fall 3.36 3.30 3.16 0.96

2006 Winter 0.18 0.17 0.05 0.02Spring 5.14 4.70 5.37 0.84Summer 4.96 4.94 4.56 1.20Fall 1.43 1.41 0.92 0.27

2007 Winter 0.38 0.36 0.22 0.08Spring 3.50 3.34 1.80 0.46Summer 6.32 6.30 6.05 1.66Fall 1.64 1.61 1.75 0.46

2008 Winter 1.41 1.33 0.24 0.10Spring 8.94 8.19 7.31 1.47Summer 8.78 8.78 7.93 2.15Fall 3.89 3.87 3.32 0.97

2009 Winter 0.06 0.06 0.06 0.02Spring 8.20 7.66 6.39 1.47Summer 3.78 3.77 3.53 0.94Fall 1.79 1.72 1.59 0.40

Average Winter 0.48 0.45 0.17 0.06Spring 4.90 4.55 3.94 0.87Summer 4.44 4.43 4.17 1.10Fall 2.20 2.16 1.96 0.55

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238234

Author's personal copy

burned by Russian farmers would need to developed, recorded, andmade available.

The specific satellite-based land cover products used in thisanalysis have reported levels of uncertainty. The croplands andcroplands/natural vegetation mosaic classes have an averageconfidence value of 76.4% and 60.7% (Hodges, 2006), respectively,and were used to develop the IGBP-Agriculture and IGBP-Croplands schemas used in the FRE, Burned Area, and ActiveFire analyses. The classifications used represent the best availableglobal land cover data sets. A country-specific and/or regionalsatellite mapping of cropland extents within Russia wouldimprove the uncertainty.

3.6.4. Uncertainty in fire productsFRE parameterization has been shown to be underestimated by

roughly 20% (Ellicott et al., 2009; Vermote et al., 2009) and is due totemporal and spatial sampling provided by MODIS. Although theintegral of FRP accounts for missed observations within a 24-hcycle, it cannot account for observations which were missed dueto timing (e.g. fire extinguished before overpass or started afteroverpass) and/or location (e.g. fires exist just outside swath ofMODIS overpass). Based on comparisons of MODIS FRE with SEVIRIFRE, as well as FRE-based emission coefficients and traditional (i.e.bottom-up) emission factors (section 3.6.1), we have roughly esti-matedMODIS sampling bias contributes to an underestimate of FREby as much as a factor of 4. Freeborn et al. (2011) also examined the

FRE generated from MODIS and found a similar underestimation.Similarly, Kaiser et al. (2011) suggested a factor of 3.4 increase inaerosol emission estimates is needed from their FRE-based esti-mates to reproduce modeled results.

The FRE emission coefficients are based on 3 broad regionsadopted from the GFEDv2, which includes tropical forest, extra-tropical forest, and grassland/savanna. This limits the separation ofland cover and/or land use heterogeneity and broadly averagesemission coefficients, thus losing resolution and increasing theerror in BC emission estimates. Current work is underway to refinethis methodology.

Our estimates of BC emissions from the GFEDv3 have a degree oferror due to conversion of the database’s estimates from BC m�2 toBC per 0.5� cell. This conversion is performed using the area oflatitude and longitude per cell and slight differences in thisconversion leads to downstream differences in the BC estimatesmade from what the GFEDv3 reports for the Boreal Asia region.

The burned area product utilized in this analysis, MCD45A1, wasfound to have 60% omission error, i.e., omitting 60% of all fires, ina savannah ecosystem in southern Africa (Roy and Boschetti, 2009).Roy et al. (2008) compared the total burned area of the MCD45A1product to quantified burned area from medium and high confi-dence values of the active fire product, MOD14/MYD14. For theNorthern Eurasia region (including the Russian Federation), theactive fire product estimated 126,000 km2 more burned area thanthe burned area product. It was noted in Roy et al. (2008) that the

Fig. 8. Seasonal cropland burning in the Russian Federation from cropland burning as detected by the MODIS Burned Area Product (MCD45A1) for years 2003e2009; definition ofcroplands used: IGBP-Croplands ¼ IGBP croplands class from the MODIS 1 km Land Cover data set.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 235

Author's personal copy

majority of burning detected by the burned area product inNorthern Eurasia occurred in cropland areas. Therefore, it is likelythat small agricultural fires weremissed in the burned area productthat can be detected by the active fire product. During this research,the effectiveness of the MODIS Burned Area Product in mappingcropland fires in the Russian Federation was not assessed. TheMCD45A1 product also provides a pixel QA value ranging from 1(most confidently detected as a burned pixel) to 4 (burned pixelsdetected as burn as they occur spatially close to neighboringburned pixels that passed most stringent temporal and spectralconstraints) (Boschetti et al., 2009). As this product has not beenassessed for accuracy in agricultural landscapes globally or inRussia, this analysis included all QA values, creating uncertainty asto whether cropland burned area estimates and thus the BCemission estimates in this analysis are overestimations orunderestimations.

The active fire product, MOD14/MYD14, does include a detec-tion confidence value ranging from 0 to 100%. All detection confi-dence values were included in this analysis. Previous work oncropland burning in the contiguous U.S. using high resolutionimagery (15 m) and in situ data has shown that a threshold of 60%

for the active fire detection confidence values removes errors ofcommission from false alarms (McCarty et al., 2008). However, suchan analysis was not completed for Russia due to lack of coincidentalhigh resolution data. Schroeder et al. (2008) found the omissionerror for MODIS active fire detections to be 80% for the Legal Bra-zilian Amazon using high (15 m) to moderate (30 m) resolutionsatellite imagery. The uncertainty associated with the detectionconfidence values was not addressed in this analysis and the rate ofomission is also inherent in the limited overpass times of the activefire product. As discrete active fire points were assigned a burnedarea of 1 km2, there is uncertainty present in the areal estimationsfrom the Active Fire analysis, similar to other studies (Wiedinmyeret al., 2006; Al-Saadi et al., 2008). This analysis did attempt toquantify the uncertainty in using a 1 km2 burned area assumptionfor each active fire detection by using a contemporary GIS fieldmask in scale of 1:25,000 for the European Russian oblasts ofRostovskaya, Moskovskaya, and Kostromskaya. The resultingburned areas for these three European Russian oblasts using the GISfield boundaries was on average � 30% different than the burnedarea estimations derived from the four definitions of cropland:IGBP-Agriculture, IGBP-Croplands, LCCS-Agriculture, and LCCS-Croplands.

4. Discussion and conclusions

Results from this analysis show that average annual BC emis-sions can vary from 2.49 Gg to 22.2 Gg with a median annual BCemission estimate of 8.90 Gg. In general, the highest BC emissionsfrom cropland burning were detected during the spring in Euro-pean Russia and West Siberia, when BC is known contributor toArctic Haze. Additionally, this analysis represents the first attemptto produce a spatially and temporally explicit cropland fire inven-tory and has identified both the source regions and federal subjectswithin Russia.

This fire emission analysis does have inherent uncertaintyassociated with the remote sensing-based fire and land coverproducts as well as the emission factor, fuel loading, and combus-tion completeness variables. Lacking a Russia-specific map ofcropland areas at a moderate resolution (approximately 30 m) alsolimits the accurate mapping of crop residue burning from satellite-based fire products. The results of this analysis show that utilizingcurrent operational fire products will provide a range of BC emis-sion estimates though do not represent a complete inventory of allcrop residue burning. Additionally the emission factors and emis-sion coefficients used are subject to error which propagates into the

a

b

Fig. 9. Annual variability of BC emissions as calculated from the Agricultural Statistics,FRE, Burned Area (BA), Active Fire (AF), and GFEDv3 approaches for years 2003e2009;(a) emissions reported by two different definitions of croplands: IGBP-Agriculture ¼ IGBP croplands and cropland/natural vegetation mosaic classes from theMODIS 1 km Land Cover data set and LCCS-Agriculture ¼ UN LCCS rainfed, irrigatedcroplands, cropland/natural vegetation mosaics and natural vegetation/croplandsmosaic classes from the MERIS 300 m GlobCover data set; and (b) emissions reportedby three different definitions of croplands: IGBP-Croplands ¼ IGBP croplands classfrom the MODIS 1 km Land Cover data set, UMd-Croplands ¼ UMd cropland class fromthe MODIS 1 km Land Cover data set, and LCCS-Croplands ¼ UN LCCS rainfed andirrigated croplands classes from the MERIS 300 m GlobCover data set; the results fromthe Agricultural Statistics approach and GFEDv3 included in both graphs; all emissionsreported in Gg.

Table 9Fuel load estimates made from FRE; assuming a prioriinformation of burned area (BA in units of m2 or ha), biomassconsumed (M in units of g or kg), and combustioncompleteness (CC in %), one can estimate the pre-fire fuelsusing the following formula: Fuel load (kg/ha)¼M/(BA * CC);in the above estimates, biomass (or dry matter) was esti-mated from FRE (MJ) using an average combustion coeffi-cient of 0.41 kg MJ�1, calculated from two publishedcombustion coefficients (Freeborn et al., 2008;Wooster et al.,2005) which expresses the amount of dry matter consumedper MJ of energy released (kg/MJ).

Year Fuel load value (kg/ha)

2003 6932004 11572005 9862006 6242007 6452008 5962009 703

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238236

Author's personal copy

emission estimates. This source of uncertainty may be responsiblefor as much as 50% of the error in the FRE emission estimates(Vermote et al., 2009; van der Werf et al., 2010) and will alsocontribute to error in the emission estimates from the Burned Area,Active Fire, and Agricultural Statistics approaches.

Acknowledgments

This work was funded by support from the Clean Air Task Force(http://catf.us/) and the Bellona Foundation of Russia (http://bellona.org/).

Appendix A. Supplementary material

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.atmosenv.2012.08.053.

References

Al-Saadi, J., Soja, A., Pierce, R.B., Szykman, J., Wiedinmyer, C., Emmons, L.,Kondragunta, S., Zhang, X., Kittaka, C., Schaack, T., Bowman, K., 2008. Inter-comparison of near-real-time biomass burning emissions estimates constrainedby satellite fire data. Journal of Applied Remote Sensing 2, 021504.

Andreae, M.O., Merlet, P., 2001. Emission of trace gases and aerosols from biomassburning. Global Biogeochemical Cycles 15 (4), 955e966.

Badarinth, K.V.S., Chand, T.R.K., Prasad, V.K., 2006. Agriculture crop residue burningin the Indo-Gangetic Plains - a study using IRS-P6 AWiFS satellite data. CurrentScience India 91, 1085e1089.

Boschetti, L., Roy, D., Hoffman, A.A., 2009. MODIS Collection 5 Burned Area Product -MCD45 User’s Guide Version 2.0. Available from: http://modis-fire.umd.edu/Documents/MODIS_Burned_Area_User_Guide_2.0.pdf.

Cao, G.L., Zhang, X.Y., Wang, Y.Q., Zheng, F.C., 2008. Estimation of emissions fromfield burning of crop straw in China. Chinese Science Bulletin 53, 784e790.

Chen, Y., Tessier, S., Cavers, C., Xu, X., Monero, E., 2005. A survey of crop residueburning practices in Manitoba. Applied Engineering in Agriculture 21, 317e323.

Derevyagin, V.A., 1987. Practical Guidance on Straw Utilization as Fertilizer, All-Union Research and Development, pp. 1e11 (in Russian, translated intoEnglish by Bellona Foundation). www.bellona.org. USSR State Agro-IndustrialCommittee on Design and Technological Institute for Organic Fertilizers andPeat, Moscow, Russia. Available from: http://quickr.mtri.org/croplandburning.

Di Gregorio, A., Jansen, L.J.M., 2000. Land Cover Classification System (LCCS):Classification Concepts and User Manual. Available from: http://www.fao.org/docrep/003/x0596e/x0596e00.htm.

Dubovik, O., Lapyonok, T., Kaufman, Y.J., Chin, M., Ginoux, P., Kahn, R.A., Sinyuk, A.,2008. Retrieving global aersol sources from satellites using inverse modeling.Atmospheric Chemistry and Physics 8, 209e250.

Ellicott, E., Vermote, E., Giglio, L., Roberts, G., 2009. Estimating biomass consumedfrom fire using MODIS FRE. Geophysical Research Letters 36, L13401. http://dx.doi.org/10.1029/2009GL038581.

ESA, 2011. European Space Agency Ionia GlobCover Portal. Available from: <http://ionia1.esrin.esa.int/>.

Freeborn, P.H., Wooster, M.J., Hao, W.M., Ryan, C.A., Nordgren, B.L., Baker, S.P.,Ichoku, C., 2008. Relationships between energy release, fuel mass loss, and tracegas and aerosol emissions during laboratory biomass fires. Journal ofGeophysical Research-Atmospheres 113 (D1), D01301. http://dx.doi.org/10.1029/2007jd008679.

Freeborn, P.H., Wooster, M.J., Roberts, G., 2011. Addressing the spatiotemporalsampling design of MODIS to provide estimates of the fire radiative energyemitted from Africa. Remote Sensing of Environment 115 (2), 475e489.

Friedl, M.A., McIver, D.K., Hodges, J.C.F., Zhang, X.Y., Muchoney, D., Strahler, A.H.,Woodcock, C.E., Gopal, S., Schneider, A., Cooper, A., Baccini, A., Gao, F., Schaaf, C.,2002. Global land cover mapping from MODIS: algorithms and early results.Remote Sensing of Environment 83, 287e302.

Giglio, L., 2010. MODIS collection 5 active fire product user’s guide: Version 2.4,edited. Available from: <http://maps.geog.umd.edu/firms/products/MODIS_Fire_Users_Guide_2.4.pdf>.

Giglio, L., Descloitres, J., Justice, C.O., Kaufman, Y., 2003. An enhanced contextualfire detection algorithm for MODIS. Remote Sensing of Environment 87,273e282.

Giglio, L., van der Werf, G.R., Randerson, J.T., Collatz, G.J., Kasibhatla, P., 2006. Globalestimation of burned area using MODIS active fire observations. AtmosphericChemistry and Physics 6, 957e974.

Goldammer, J.G., 2010. Preliminary Assessment of the Fire Situation in WesternRussia. Report to Duma Committee for Natural Resources, Nature Use, andEcology. Available from: <http://www.fire.uni-freiburg.de/intro/about4_2010-Dateien/GFMC-RUS-State-DUMA-18-September-2010-Fire-Report.pdf>.

Hansen, M.C., DeFries, R.S., Townshend, J.R.G., Sohlberg, R., 2000. Global land coverclassification at 1 km spatial resolution using a classification tree approach.International Journal of Remote Sensing 21, 1331e1364.

Hari, O.M., Dhiman, S.D., Hemant, K., Sajjan, K., 2003. Biology and managementof Phalaris minor in wheat under a rice/wheat system. Weed Research 43,59e67.

Hegg, D.A., Warren, S.G., Grenfell, T.C., Doherty, S.J., Larson, T.V., Clarke, A.D., 2009.Source attribution of black carbon in Arctic snow. Environmental Science andTechnology 43, 4016e4021.

Hodges, J.C.F., 2006. Validation of consistent-year V003 MODIS land cover product.Available from: <http://www-modis.bu.edu/landcover/userguidelc/consistent.htm>.

Ichoku, C., Kaufman, Y.J., 2005. A method to derive smoke emission rates fromMODIS fire radiative energy measurements. IEEE Transactions of Geoscienceand Remote Sensing 43, 2636e2649.

IIASA, 2010. Personal Communication with Zbigniew Klimont, Ph.D. of Interna-tional Institute for Applied Systems Analysis (IIASA), 26 October 2010. BCemissions factor from the GHG-Air pollution Interaction and Synergies(GAINS) model.

Kaiser, J.W., Heil, A., Andreae, M.O., Benedetti, A., Chubarova, N., Jones, L.,Morcrette, J.-J., Razinger, M., Schultz, M.G., Suttie, M., van der Werf, G.R.,2011. Biomass burning emissions estimated with a global fire assimilationsystem based on observed fire radiative power. Biogeosciences Discussion 8,7339e7398.

Kaufman, Y.J., Justice, C.O., Flynn, L.P., Kendall, J.D., Prins, E.M., Giglio, L., Ward, D.E.,Menzel, W.P., Setzer, A.W., 1998. Potential global fire monitoring from EOS-MODIS. Journal of Geophysical Research 103, 32215e32238.

Korontzi, S., McCarty, J., Loboda, T., Kumar, S., Justice, C.O., 2006. Global distributionof agricultural fires from three years of MODIS data. Global BiogeochemicalCycles 20 (6), 2021. http://dx.doi.org/10.1029/2005GB002529.

Langmann, B., Duncan, B., Textor, C., Trentmann, J., van der Werf, G.R., 2009.Vegetation fire emissions and their impact on air pollution and climate.Atmospheric Environment 43, 107e116.

Law, K.S., Stohl, A., 2007. Arctic air pollution: Origins and impacts. Science 315,1537e1540.

Liousse, C., Penner, J.E., Chuang, C., Walton, J.J., Eddleman, H., Cachier, H., 1996.A global three-dimensional model study of carbonaceous aerosols. Journal ofGeophysical Research 101, 19,411e19,432.

Loveland, T.R., Belward, A.S., 1997. The IGBP-DIS global 1-km land cover dataset, DIScover: first results. International Journal of Remote Sensing 65,1021e1031.

Lyuri, D.I., Goryachkin, S.V., Karavaeva, N.A., Denisenko, E.A., Nefedova, T.G., 2010.Dynamics of Agricultural Lands of Russia in XX Century and PostagrogenicRestoration of Vegetation and Soils, Geos, Moscow. 420 pp. (in Russian).

McCarty, J.L., 2011. Remote sensing-based estimates of annual and seasonal emis-sions from crop residue burning in the contiguous United States. Journal of Airand Waste Management Association 61, 22e34.

McCarty, J.L., Justice, C.O., Korontzi, S., 2007. Agricultural burning in the south-eastern United States detected by MODIS. Remote Sensing of Environment 108,151e162.

McCarty, J.L., Loboda, T., Trigg, S., 2008. A hybrid approach to quantifying cropresidue burning in the US based on burned area and active fire data. AppliedEngineering in Agriculture 24, 515e527.

Novakov, T., Menon, S., Kirchstetter, T.W., Koch, D., Hansen, J.E., 2005. Aerosolorganic carbon to black carbon ratios: analysis of published data and implica-tions for climate forcing. Journal of Geophysical Research 110, D21205. http://dx.doi.org/10.1029/2005JD005977.

Pyne, S.J., 1993. Keeper of the flame: a survey of anthropogenic fire. In: Crutzen, P.J.,Goldammer, J.G. (Eds.), Fire in the Environment: The Ecological, Atmospheric,and Climatic Importance of Vegetation Fires. John Wiley & Sons, New York,pp. 245e266.

Quinn, P.K., Bates, T.S., Baum, E., Doubleday, N., Fiore, A.M., Flanner, M., Fridlind, A.,Garrett, T.J., Koch, D., Menon, S., Shindell, D., Stohl, A., Warren, S.G., 2008. Short-lived pollutants in the Arctic: their climate impact and possible mitigationstrategies. Atmospheric Chemistry and Physics 8, 1723e1735.

Roy, D.P., Boschetti, L., 2009. Southern Africa validation of the MODIS, L3JRC andGLOBCARBON burned area products. IEEE Transactions of Geoscience andRemote Sensing 47, 1032e1044.

Roy, D.P., Jin, Y., Lewis, P.E., Justice, C.O., 2005. Prototyping a global algorithm forsystematic fire affected area mapping using MODIS time series data. RemoteSensing of Environment 97, 137e162.

Roy, D.P., Boschetti, L., Justice, C.O., Ju, J., 2008. The Collection 5 MODIS burned areaproduct e global evaluation by comparison with the MODIS active fire product.Remote Sensing of Environment 112, 3690e3707.

Rusakova, I.V., 2006. Straw. In: Yes’kov, A.I. (Ed.), Organic Fertilizer Resources inRussian Agriculture: Information Analysis Reference Book, All-Russian Instituteof Organic Fertilizers and Peat. Russian Federation, Moscow, pp. 113e119 (inRussian, translated into English by Bellona Foundation). www.bellona.org.Available from: <http://quickr.mtri.org/croplandburning>.

Schroeder, W., Prins, E., Giglio, L., Csiszar, I., Schimdt, C., Morisette, J., Morton, D.,2008. Validation of GOES and MODIS active fire detection products using ASTERand ETMþ data. Remote Sensing of Environment 112, 2711e2726.

Seiler, W., Crutzen, P.J., 1980. Estimates of gross and net fluxes of carbon betweenthe biosphere and the atmosphere from biomass burning. Climatic Change 2,207e247.

Sharma, S., Lavoué, D., Cachier, H., Barrie, L.A., Gong, S.L., 2004. Long-term trends ofthe black carbon concentrations in the Canadian Arctic. Journal of GeophysicalResearch 109, D15203. http://dx.doi.org/10.1029/2003JD004331.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238 237

Author's personal copy

Soja, A.J., Cofer, W.R., Shugart, H.H., Sukhinin, A.I., Stackhouse Jr., P.W., McRae, D.J.,Conard, S.G., 2004. Estimating fire emissions and disparities in boreal Siberia(1998e2002). Journal of Geophysical Research 109, D14S06. http://dx.doi.org/10.1029/2004JD004570.

Stohl, A., Berg, T., Burkhart, J.F., Fjraa, A.M., Forster, C., Herber, A., Hov, Ø.,Lunder, C., McMillan, W.W., Oltmans, S., Shiobara, M., Simpson, D., Solberg, S.,Stebel, K., Ström, J., Tørseth, K., Treffeisen, R., Virkkunen, K., Yttri, K.E., 2007.Arctic smoke e record high air pollution levels in the European Arctic due toagricultural fires in Eastern Europe in spring 2006. Atmospheric Chemistryand Physics 7, 511e534.

Streets, D.G., Gupta, S., Waldhoff, S.T., Wang, M.Q., Bond, T.C., Yiyun, B., 2001. Blackcarbon emissions in China. Atmospheric Environment 35, 4281e4296.

Turn, S.Q., Jenkins, B.M., Chow, J.C., Pritchett, L.C., Campbell, D., Cahill, T.,Whalen, S.A., 1997. Elemental characterization of particulate matter emittedfrom biomass burning: wind tunnel derived source profiles for herbaceous andwood fuels. Journal of Geophysical Research 102, 3863e3869.

U.S. EPA, 2008. United States Environmental Protection Agency: Inventory of U.S.Greenhouse Gas Emissions and Sinks: 1990-2006, (USEPA 430-R-08-005),Research Triangle Park, North Carolina. Available from: <http://www.epa.gov/climatechange/emissions/downloads/08_CR.pdf>.

USDA FAS, 2003. Russia: Agricultural Overview. United States Department ofAgriculture Foreign Agricultural Service Production Estimate and Crop Assess-ment Division. Available from: <http://www.fas.usda.gov/pecad2/highlights/2005/03/Russia_Ag/index.htm>.

van der Werf, G.R., Randerson, J.T., Giglio, L., Collatz, G.J., Kasibhatla, P.S., 2006.Interannual variability in global biomass burning emission from 1997 to 2004.Atmospheric Chemistry and Physics 6, 3423e3441.

van der Werf, G.R., Randerson, J.T., Giglio, L., Collatz, G.J., Mu, M., Kasibhatla, P.S.,Morton, D.C., DeFries, R.S., Jin, Y., van Leeuwen, T.T., 2010. Global fire emissionsand the contribution of deforestation, savanna, forest, agricultural, and peatfires (1997e2009). Atmospheric Chemistry and Physics 10, 11707e11735.

Vermote, E., Ellicott, E., Dubovik, O., Laypyonok, T., Chin, M., Giglio, L., Roberts, G.,2009. An approach to estimate global biomass burning emissions of Organicand Black Carbon from MODIS Fire Radiative Power. Journal of GeophysicalResearch 114, D18205. http://dx.doi.org/10.1029/2008JD011188.

Warneke, C., Floyd, K.D., Brioude, J., Bahreini, R., Brock, C.A., Cozic, J., de Gouw, J.A.,Fahey, D.W., Ferrare, R., Holloway, J.S., Middlebrook, A.M., Miller, L., Montska, S.,Shwarz, J.P., Sodemann, H., Spackman, J.R., Stohl, A., 2010. An important contri-bution to springtime Arctic aerosol from biomass burning in Russia. GeophysicalResearch Letters 37, L01801. http://dx.doi.org/10.1029/2009GL041816.

Wiedinmyer, C., Quayle, B., Geron, C., Belote, A., McKenzie, D., Zhang, X., O’Neill, S.,Wynne, K.K., 2006. Estimating emission from fire in North America for airquality modeling. Atmospheric Environment 40, 3419e3432.

Wooster, M.J., Roberts, G., Perry, G.L.W., Kaufman, Y.J., 2005. Retrieval of biomasscombustion rates and totals from fire radiative power observations: FRP deri-vation and calibration relationships between biomass consumption and fireradiative energy release. Journal of Geophysical Research-Atmospheres 110,D24311. http://dx.doi.org/10.1029/2005JD006318.

J.L. McCarty et al. / Atmospheric Environment 63 (2012) 223e238238