Content uploaded by Tatiana Loboda
Author content
All content in this area was uploaded by Tatiana Loboda on Jan 24, 2014
Content may be subject to copyright.
Global distribution of agricultural fires in croplands from 3 years of
Moderate Resolution Imaging Spectroradiometer (MODIS) data
Stefania Korontzi,
1
Jessica McCarty,
1
Tatiana Loboda,
1
Suresh Kumar,
1
and Chris Justice
1
Received 7 April 2005; revised 26 October 2005; accepted 20 January 2006; published 24 June 2006.
[1] The Moderate Resolution Imaging Spectroradiometer (MODIS) sensor offers an
improved combination of spectral, temporal, and spatial resolution for global fire
detection compared to previous sensors. The MODIS Terra active fire product was
analyzed to investigate the spatial and temporal occurrence of fires in croplands from
2001 to 2003. Monthly fire counts were analyzed globally, within several regions and
for important crop-producing countries. The annual global total number of fire counts
ranged from 1,472,367 to 1,577,952 during the 3 years. Agricultural fires were found to
account for 8–11% of the annual global fire activity during the 3 years, but the
contribution of agricultural burning was significantly higher on a regional basis. The
Russian Federation was the largest contributor to agricultural burning globally during the
3 years, producing 31–36% of all agricultural fires. The global spatial distribution of
agricultural fires was fairly similar among the 3 years, but a notable interannual change
was observed in the total number of global agricultural fire events. The majority of
regions showed similar magnitude and seasonality in their year-to-year agricultural fire
activity, but in some regions, significant differences were found. At the global scale,
agricultural fire activity showed two peaks, the first occurring during April to May, and
was associated primarily with burning in the croplands of Eastern Europe and European
Russia, and the second in August from burning mainly in the croplands across central
Asia and Asiatic Russia. This timing pattern was observed both in 2001 and 2002. The
August 2003 fire peak was significantly affected by reduced agricultural fire activity in
European Russia. The seasonal and interannual trends in agricultural fire activity are
consistent with known national and regional agricultural practices and reported crop
production estimates.
Citation: Korontzi, S., J. McCarty, T. Loboda, S. Kumar, and C. Justice (2006), Global distribution of agricultural fires in croplands
from 3 years of Moderate Resolution Imaging Spectroradiometer (MODIS) data, Global Biogeochem. Cycles, 20, GB2021,
doi:10.1029/2005GB002529.
1. Introduction
[2] Agriculture is perhaps the most important land use
and since its initial, localized development in the fertile
crescent, has expanded globally replacing both forests and
grasslands [Fischer et al., 2001]. Croplands currently occu-
py about 17 million km
2
, which is more than 10% of the
global land surface [Friedl et al., 2002], and further expan-
sion of agricultural lands is anticipated to meet the growing
demand for a secure world food supply [Fischer et al.,
2001]. Burning in the fields is a common agricultural
practice used during the harvesting, postharvesting or pre-
planting periods. Agricultural burning is undertaken for a
number of reasons including clearing crop residue, fertiliz-
ing the soil, eliminating pests and weed and is often a firmly
entrenched cultural practice [e.g., Chidumayo, 1987; Ekboir,
2002]. An estimated 400 Tg of crop residues are burned in
the developing world with significant regional consequen-
ces on atmospheric air quality [Yevich and Logan, 2003]
and human health [World Health Organization (WHO),
2000]. Agricultural burning is not limited to developing
countries and, although restricted in some countries, con-
tributes to regional air quality and national emissions [U.S.
Environmental Protection Agency, 2003]. The most culti-
vatedcropgloballyiscereals[Food and Agricu lture
Organization (FA O ), 2002a]. Wheat, maize, rice, barley,
millet and sorghum occupy more than two thirds of the
global cultivated area [Leff et al., 2004] and their waste
products provide the main contributors to agricultural bio-
mass burning [Yevich and Logan, 2003]. Sugar cane, soy
and cotton crops also provide co nsiderable amounts of
residues that are burned. Minor crop residues burned in
the field include other various vegetable crops, legumes,
and decorative plants [Kakareka and Kukharchyk, 2003].
With increasing human population and changing land use
GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 20, GB2021, doi:10.1029/2005GB002529, 2006
1
Department of Geography, University of Maryland at College Park,
College Park, Maryland, USA.
Copyright 2006 by the American Geophysical Union.
0886-6236/06/2005GB002529$12.00
GB2021 1of15
practices that foster agricultural expansion and intensifica-
tion, the contribution of agricultural burning as a source of
global biomass burning trace gas and particulate emissions
will likely also rise.
[
3] There is a notable paucity of quantitative data regard-
ing agricultural fires in the scientific literature. Estimates of
agricultural burning have so far relied on national statistics
and simple, generalized assumptions about burned crop
residues as a fraction of the available agricultural waste
[Andreae, 1991; Hao and Liu, 1994; Yevich and Logan,
2003]. A consistent, reliable, large-scale characterization of
the spatial and temporal distribution of agricultural burning
is required to assess its environmental impacts and to
support fire management of natural resources. Satellite
remote sensing of global active fires provides an important
tool to achieve this goal while removing inconsistencies
associated with national fire reporting.
[
4] The present study provides the first global assessment
of fire activity in established agricultural lands and aims at
identifying regional patterns in agricultural burning from
2001 to 2003 using MODIS Terra data. In section 2 the
MODIS active fire and land cover data sets and the
methodology used in this analysis are described. In section
3.1 the regional- to global-scale fire occurrence is analyzed
as a function of land cover. In addition, monthly time series
of acti ve fire detections are analyzed to investigate the
seasonal and interannual variations in agricultural burning.
The agricultural fire patterns for selected major world crop
producing countries are presented in section 3.2. National/
regional reports on agricultural assessment and production
published by the United States Department of Agriculture
Production Estimates and Crop Assessment Division/
Foreign Agricultura l Service (USDA/PECAD/FAS), the
Joint Agricultural Weather Facility (USDA/JAWF) and the
National Agricultural Statistics Service (USDA/NASS) are
used to provide an explanation for the observed seasonal
and interannual variations in agricultural fire activity.
Section 4 contains a summary and conclusions.
2. Data and Methods
[5] MODIS on board the polar orbiting Terra and Aqua
satellites of the National Aeronautics and Space Adminis-
tration (NASA) Earth Observing System (EOS) was the first
instrument designed with fire detection in mind, and it
provide s daily gl oba l fire ob ser vati ons of un par alle led
quality [Kaufman et al., 1998; Justice et al., 2002a]. The
MODIS active fire product (MOD14) [Justice et al., 2002a]
is part of the suite of global land surface products produced
operationally [Justice et al., 2002b]. The MODIS products
are archived and distributed at the NASA Distributed Active
Archive Centers (DAACs).
[
6] The 1-km multiyear (January 2001 to December
2003) active fire data record used for the analysis was
generated with the latest and improved MODIS version 4
active fire detection algorithm [Giglio et al., 2003]. Com-
pared with previous versions of the algorithm, version 4
improves on the detection of small (<100 m
2
) fires, which is
important for monitoring agricultural burning. Validation of
the MODIS active fire product over southern Africa with
coincident Advanced Spaceborne Thermal Emissions and
Reflection Radiometer (ASTER) data has shown that there
is a 95% chance that MODIS will classify a pixel as fire
when the ASTER fire counts within the 2 1 km area,
corresponding to the nominal MODIS pixel, are over 30
[Morisette et al., 2005]. A fire as small as 0.1% of the
ASTER pixel area is the lower size limit for fire identifica-
tion by ASTER (Louis Giglio, MODIS active fire algorithm
developer, personal communication, 2005). Although the
MODIS instrument is flown on two satellites, Terra and
Aqua, there was roughly a 2-year gap between their launch
and their overpass times differ by 4 hours. We included only
MODIS fire detections from Terra to insure comparability
of the data records between 2001 and 2003.
[
7] Corrections were ap plied to the data prior to the
analysis to account for inconsistent data coverage. Morning
and night detections of a single fire event were counted
once. F urthermore, swath overlap in the mid and high
latitudes during adjacent passes could result in multiple
recording of the same fire event. The f ire data records were
also corrected for these multiple detections of the same fire
pixel. Pixels that were detected during the morning and
night overpasses, in a particular day, and had locations
matching to 0.001 degree, were counted only once. A small
number of data losses occurred in the time series (42, 11 and
18 days in 2001, 2002 and 2003, respectively) because of
MODIS being inoperable due to engineering adjustments,
incomplete transmission of instrument and ephemeris data,
and some losses occurring in the production system (David
Roy, M ODIS Land QA lead, personal communication,
2005). Using a moving window before and after the missing
date period, the missing fire counts were adjusted with the
mean value of the fire counts from the moving window.
[
8] To derive the fire distribution over various land
covers, each daily observation was combined with the
MODIS 1 km land cover (LC) data set (MOD12) [Friedl
et al., 2002], illustrated in Figure 1, with the University of
Maryland (UMd) classification s cheme [Hansen et al.,
2000, Table 1]. The MODIS LC was derived using data
for the year 2000. For the analysis, the 14 classes listed in
Table 1 were aggregated into the following general catego-
ries: ‘‘Forests,’’ comprising the Evergreen Needleleaf, Ev-
ergreen Broadleaf, Deciduous Needleleaf, D eciduous
Broadleaf and Mixed Forest LC classes (LC 1–5), ‘‘Shrub-
lands,’’ comprising the Closed and Open Shrubland LC
classes ( LC 6–7), ‘‘Savannas,’’ comprising the Woody
Savannas and Savannas LC classes (LC 8 –9), ‘‘Grassland’’
(LC 10), ‘‘Croplands’’ (LC 12), and ‘‘Other,’’ comprising
LC > 13. The UMd definition of croplands includes lands
with >80% coverage with crops. Subsistence farming,
pastures, rangelands, orchards, cropland/natural vegetation
mosaics in which no one component comprises more than
60% of the landscape (IGBP LC 14), and tropical slash and
burn agriculture (estimated to be 6.5 10
6
ha/yr by
DeFries et al. [2002]) are not included in the UMd ‘‘Crop-
lands’’ LC definition. Therefore we present a conservative
estimate of fires on established agricultural lands. The terms
‘‘agricultural burning’’ and ‘‘cropland burning’’ are used
interchangeably throughout the paper. This study is mainly
focused on global, regional and national monthly and
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
2of15
GB2021
interannual fire patterns in croplands. Csiszar et al. [2005]
present detailed statistics for all land covers derived from
the 2001 and 2002 MODIS active fire data. Validation of the
MODIS LC product over selected areas (denoted as stage 1
validation) deemed an accuracy a 75 –80% globally and
76.4% for the croplands land cover class [Friedl et al.,
2002].
[
9] The following limitations should be noted regarding
the utilization of the MOD IS active fire data for the
purposes of the analysis. As is the case with all active fire
data sets, the active fire count is not representative of the
actual burned area, since substantially smaller fires are
detected than the nominal 1-km pixel size, because the
satellite may not always overpass when burning occurs and
because fire detection may be prevented due to cloud cover
[Robi nson, 1991; Justice et al., 1993]. In addition, the
number of distinct agri cultural fires occurring in small,
discrete land plots within the same 1-km pixel cannot be
inferred from the data. Despite these limitations, the fire
count enables a relative comparison of the proportion of
pixels affected by fire within the various land cover types
and is useful to broadly characterize the intra and interan-
nual spatiotemporal variability of fire phenomena.
[
10] As noted above, we only used MODIS Terra data for
our analysis. The diurnal cycle of fire may be significant in
some regions and still needs to be determined for agricul-
tural fires. However, global analysis of multiyear records of
MODIS Terra and Aqua active fire data by Giglio et al.
[2006] has indicated, that the diurnal fire activity in central
Eurasia, which represents the majority of global croplands,
is insignificant.
3. Results and Discussion
3.1. Global and Regional Agricultural Fire
Distribution
[
11] A total number of 1,577,952, 1,572,884 and
1,472,367 fires were detected globally in 2001, 2002 and
2003, respectively. Figure 2 shows the annual relative
distribution of global fires within predominant land cover
types. The most burning occurred in savannas (48–51%),
Figure 1. Map of the MODIS 1-km Land Cover Types data set [Friedl et al., 2002] and location of the
regional windows used in this study.
Table 1. Global Land Cover Types Based on the MODIS 1-km
Land Cover Data Set, Using the UMd Classification Scheme
a
Class Number Class Name
0 Water
1 Evergreen Needleleaf Forest
2 Evergreen Broadleaf Forest
3 Deciduous Needleleaf Forest
4 Deciduous Broadleaf Forest
5 Mixed Forests
6 Closed Shrubland
7 Open Shrubland
8 Woody Savannas
9 Savannas
10 Grasslands
12 Croplands
13 Urban and Built-Up
16 Barren or Sparsely Vegetated
a
Hansen et al. [2000].
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
3of15
GB2021
whereas agricultural fires accounted for 8 – 11% of the
global annu al fires. Interannual change in proportion al
contribution was most pronounced in forest fires with a
doubling to 22% in 2003 compared with 11% in 2001.
[
12] A notable interannual change in the total number of
agricultural fire detections was observed among the 3 years,
including a 14% decrease from 2001 to 2002 and a 21%
decrease from 2002 to 2003. The monthly global total fire
count in croplands during the three year period is presented in
Figure 3. At the global scale the fire occurrences showed two
distinct peaks, the first occurring during April to May and the
second in August. In August 2003 the fire activity was
significantly reduced compared with the other 2 years, by a
factor 5.7 and 3.3 compared with 2001, and 2002, respec-
tively. For the 3-year period, low agricultural fire activity
occurred generally between the months of October to March.
[
13] The monthly spatial distribution of MODIS Terra
detections in croplands globally for the year 2001 is
portrayedinauxiliaryFigureS1
1
. This map shows all
agricultural fire detections with no correction for missing
data but with correction for swath overlap or double
detection in the same pixel within a given day. Fewer fires
during the second half of June, the last week of September
and the first 2 weeks of October are partially the result of
missing data. The highest concentration of agricultural fires
globally extended across Russia and its border with China,
in the latitudinal belt between 45°N–55°N, during the
spring (April–May), as well as in Eastern Europe during
the late summer (August). Auxiliary Figure S2 indicates
similar annual spatial distribution of agricultural fires glob-
ally during the 3-year period, with reduction in fire activity
mainly in Eastern Europe in 2002 compared with 2001 and
further significant reduction in the same region and southern
European Russia in 2003.
[
14] For a more comprehensive analysis of global agri-
cultural fire patterns we distinguished several regions of
significant fire activity for further analysis: North America
(region A) north of 25°N and west of 20°W, Europe and
European Russia (region B) north of 35°N and between
20°W and 60°E, Asiatic Russia and central and northeast
Asia (region C) north of 30°N and east of 60°E, Central
America and the Caribbean (region D) between 0°N and
25°N and west of 20°W, northern Africa (region E) between
0°N and 35°N and 20°W and 60°E, Southeast Asia and
India (region F) between 10°S and 30°N and east of 60°E,
South America (region G) south of 0°N and west of 20°W,
southern Africa (region H) south of 0 °N and between 20°W
and 60°E and Australia and New Zealand (region I) south of
10°S and east of 60°E. The six regional-scale windows are
outlined in Figure 1. Figure 4 shows the fire distribution
within predominant land cover types in each region. Figure 5
presents the monthly time series of agricultural fire detec-
tion in each region.
[
15] The Northern Hemisphere produced the majority of
annual agricultural fire counts glob ally (94% in all
3 years). In North America (region A) agricultural burning
contributed 9 –16% of all fires detected over the various
land covers (Figure 4a) during the 3-year period. The
majority of the burning (33–49%) in region A occurred in
forests. Broadly described, North America is composed of
two wheat belts and one maize belt in the north which
transitions gradually into a soybean belt in the south [Leff et
al., 2004]. In North America there were two distinct peaks
in the timing of agricultural fire activity, the first during
March to April and the second during September to October
(Figure 5a). In Canada the fires occurred in the wheat-barley
belt in the provinces of Alberta and Saskatchewan during
the spring but also during the fall harvesting period. In the
United States the spr ing fires w ere concentrated in the
wheat belt between Canada and the Great Plain states and
transitioned progressively southward to occur largely in the
southeastern United States during September and October.
Besides the southeastern United States, fires also occurred
in the Pacific Northwest during the fall. A number of other
major crops, for which fire is used besides wheat, are
cultivated in the United States, including maize, barley,
canola, sorghum, rice, soybeans, cotton and sugar cane. In
2001 and 2002 the timing and magnitude of the agricultural
fires were similar, and the spring and fall fire activities were
comparable (Figure 5a). In 2003, more spring burning took
place relat ive to the fall burning.
[
16] Cropland burning accounted for a remarkable 48–
73% of all fires in the various land cover types in Europe
(most countries south of Scandinavia) and European Russia
Figure 2. Distribution of global fire occurrence within
various land covers in 2001–2003.
Figure 3. Seasonal and interannual variability of MODIS
fire detections in croplands globally in 2001 –2003.
1
Auxiliary material is available at ftp://ftp.agu.org/apend/gb/
2005gb002529.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
4of15
GB2021
(region B) du ring the 3 years (Figure 4b). Europe is
composed of agriculturally homogeneous regions planted
predominantly with wheat, barley, maize, potatoes and
sunflower [Leff et al., 2004]. Even though many countries
in Western Europe have banned open field burning [Jenkins
et al., 1992], some agricultural burning was detected
by MODIS in several western European countries (see
auxiliary Figure S2). The majority of crop burning in region
B though occurred in the cropland-dominated areas of
Eastern Europe and southern European Russia. The maxi-
mum fire activity in 2001 and 2002 in this region occurred
in August, during the harvesting of the winter and spring
grains. Fire activity in this region was the determinant for
the August peak observed at the global scale, indicating that
this is the major center of agricultural fire activity globally
(Figures 3 and 4). The significantly lower fire activity in
August 2003 compared with the previous years (10 times
lower than in August 2001) can be mainly attributed to
unfavorable weather conditions. Inclement winter and un-
seasonably cold spring weather resulted in a reduction of
planted ar ea and unusually high winterkill to grai ns in
European Russia [USDA/PECAD/FAS, 2003a].
[
17] Extensive agricultural burning occurred also in the
region of Asiatic Russia, central and northeast Asia
(region C) and accounted for 18–29% of all fires detected
within the various land covers (Figure 4c). The relative
contribution of forest fires more than doubled from 2001 to
the following 2 years. The most dominant crops cultivated
Figure 4. Annual fire distribution within predominant land cover types in the study regions in 2001 –
2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
5of15
GB2021
in region C are wheat, potatoes, maize, barley, groundnuts
and sugar cane [Leff et al., 2004]. The spring (April–May)
agricultural fire maximum shown in Figure 5c was mainly
from burning in central and southeast Russia, followed by
burning in northern and southern Kazakhstan and the
Russia-China border (see auxiliary Figure S1). The smaller
fall (September – October) fire peak resulted from agricul-
tural burning mainly in the Russia-China border. The entire
Russian Federation accounted for 31–36% of the global
agricultural fires during the 3 years.
[
18] In Central America and the Caribbean (region D),
savanna fires prevailed (35–42%), followed by burning in
forests (24–36%) (Figure 4d). Agricultural burning
accounted for 9% of all fires detected over the various
land covers during the 3 years. The main period for
agricultural burning occurred from March to May during
harvesting in the maize belt throughout Central America
(Figure 5d). Other major crops grown in Mexico include
sorghum, pulses, rice, sugar cane and wheat. A mixture of
maize, pulses and rice is cultivated in Costa Rica, whereas
the Carib bean is dominated by sugar cane [Leff et al., 2004].
The seasonal distribution and magnitude of agricultural fires
were fairly similar among the 3 years (Figure 5d).
[
19] Fire activity in northern Africa (regi on E) was
dominated by savanna burning (75%), whereas agricul-
tural fires made a small contribution (2%) (Figure 4e).
Agricultural burning heightened from November through
February (Figure 5e), which is also general ly the main
period for savanna burning in this region [Csiszar et al.,
2005]. The main crops grown in Sahelian Africa are the
drought resistant millet and sorghum, whereas in the more
humid coastal regions of West Africa maize and rice are also
cultivated [Leff et al., 2004]. No signific ant interannual
variations in the amount and seasonality of agricultural fire
activity were found in region E (Figure 5e).
[
20] Forest fires produced the majority of fire counts (38–
50%) in Southeast Asia and India (region F) followed by
savanna fires (28–30%) (Figure 4f). A major reason for
Figure 5. Monthly time series of MODIS Terra agricultural fire detections in the study regions in
2001–2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
6of15
GB2021
setting forest fires in Southeast Asia is to clear land for
shifting agriculture, plantations and transmigration-program
settlers [Sastry, 2002]. Cropland burning was also substan-
tial in this region contributing 16 –20% of the total fire
counts. Rice and wheat form the dominant crop belts, but a
number of other crops, including maize, cassava, pulses, oil
palms and groundnuts, are cultivated throughout this region.
The agricultural fire activity was concentrated in the months
of December through April (Figure 5f) during the main fire
period for this region [Csiszar et al., 2005]. The magnitude
of agricultural fire events was fairly similar among the
3 years. In 2002 the peak fire activity lagged a month
behind the 2001 February peak, whereas the 2003 peak fire
activity occurred 2 months earlier compared with 2001.
These temporal variations are likely related to different
lengths in the growing seasons.
[
21] In the Southern Hemisphere the majority of agricul-
tural fires occurred in the late dry season (July–October) in
the subequatorial tropics of South America (region G)
(Figure 5g) and in southern Africa (region H) (Figure 5h)
and coincided with the main period of savanna burning
[Csiszar et al., 2005]. Both of these regions were clearly
dominated by savanna fires. In South America cropland
burning accounted for a small 2% of all fires in various land
cov er types, whereas savanna fires compr ised 42 – 55%
(Figure 5g). The decrease in relative contribution of savanna
fires from 2001 to 2003 was marked by a concurrent
doubling in forest fires. In South America, maize and
soybean cultivation dominate in the northern parts of the
continent as well as in southeastern Brazil, whereas wheat
cultivation dominates in the southern portions of Argentina
and Chile [Leff et al., 2004]. The magnitude and seasonality
of agricultural burning in South America was fairly similar
among the 3 years (Figure 5g). South America is a region of
intense agricultural fire activity, but the majority occurs as
conversion of forest to pasture or pasture to cropland
[Morton et al., 2005], rather than in established agricultural
lands examined in the present analysis. In Mato Grosso
alone, about 1 10
6
ha of cerrado and forest have been
cleared for cropland conversion between 2001 and 2004
[Marris, 2005].
[
22] In southern Africa, savanna fires contributed about
80% and agricultural fires only 1% of all fires during the
3-year period (Figure 4h). Through out southern Africa,
maize is a winter crop generally harvested after August,
which is the month of maximum agricultural fire activity as
shown in Figure 5h. Other important crops include millet,
cassava, wheat, sorghum, cotton and sugar cane [Leff et al.,
2004]. The smaller peak observed during March is due to
harvesting of summer crops (wheat, maize, rice) in South
Africa, Madagascar and southeastern Mozambique (see
auxiliary Figure S1). The drop in the 2002 and 2003 late
dry season fire activity was possibly related to drought
conditions that affected several countries in southern Africa
and likely also to low crop prices [USDA/PECAD/FAS,
2002a, 2002b, 2003b, 2003c].
[
23] The majority of annual fire occurrences in Australia
and New Zealand (region I) were associated with burning in
shrublands (36–62%) and savannas (25–39%), whereas
agricu ltural fi res were comparatively insignificant (1%)
(Figure 4i). The timing of agricultural fires in southern
Australia is counterseasonal to the tropical savanna burning
in northern Australia [Russell-Smith et al., 2003; Csiszar et
al., 2005]. Agricultural burning in Australia occurred mostly
during March to April (Figure 5i) in the wheat producing
regions of southwestern Western Australia and the eastern
parts of the wheat belt that stretches from southern South
Australia to central Queensland (see auxili ary Figure S1). An
El Nin˜o drought resulted in significant decrease in the 2002/
2003 wheat production [USDA/PECAD/FAS, 2003d] and
this was also reflected in the 2003 fire activity (Figure 5i).
The April 2003 agricultural fires were 3.4 times lower than
in the two previous years. Even though sugar cane is also
grown in Australia, most of the harvesting is done mechan-
ically without burning [Vallis et al., 1996].
3.2. Case Studies of Agricultural Fire Use in
Important World-Crop Producing Countries
3.2.1. Southeastern United States
[
24] MODIS detected a total of 9587 fires in 2001, 8096
fires in 2002 and 7758 fires in 2003 in the southeastern
United States. Agricultural burning accounted for an aver-
age of 15% of all burning, and it was the fourth largest
contributor to burning (Figure 6). Forest fires dominated
this part of the United States with an average of 38.6% of all
fire detections. This is not surprising owing to the influence
of industrial forestry on the South’s economy. Savanna fires
accounted for 22% of burning.
[
25] Many different crops that are managed with fire are
cultivated in the southeastern United States [USDA/NASS,
1997]. A strong relationship exists in this region between
location and timing of burning, associated with the crop
type processing and harvesting. Most of the agricultural
burning in the southeastern United States was found in three
areas: the corners of Alabama, Georgia and Florida, along
the Mississippi River, and central Florida (Figure 7). Along
the Mississippi and in central Florida, rice, winter wheat,
and sugar cane dominate the croplands [USDA/JAWF,
1994]. Fire is a common management practice for these
three crops. The temporal variation of agricultural burning
corresponds well to the winter wheat, rice, and sugar cane
harvests. All 3 years showed an increase in burning between
Figure 6. Distribution of fire occurrence in the south-
eastern United States within various land covers in 2001–
2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
7of15
GB2021
June and July, corresponding to the winter wheat harvesting
(Figure 8). An increase in burning also occurred in October
and November, during the rice and sugar cane harvesting
period. Winter wheat is harvested in Arkansas and Mis-
sissippi during late May and June. Rice is harvested during
October and the beginning of November along the Mis-
sissippi River. The sugar cane harvest season commences in
late October and continues until March of the next year in
Louisiana and Florida. The border between Alabama, Geor-
gia, and Florida is a major cotton growing region [USDA/
NASS, 1997]. Burning for pest and weed management is a
common practice for cotton farmers before planting (March)
and after harvest (October).
[
26] It should be noted that the southeastern region of the
United States also contains states with little to no cropland
burning even though agricultural lands are present. The
upper south (Kentucky, Tennessee, and Virginia) contained
little agricultural burning (Figure 7). Some agricultural
burning was detected along the border between North
Carolina and South Carolina during all four seasons. This
is an area of soybean, cotton, winter wheat rotation cropping
[USDA/NASS, 1997], where burning in the spring for field
preparation, in the summer for pest management, in the fall
for harvest, and in the winter for weed control is a common
practice.
[
27] MODIS detected slightly few er fires in 2002 than the
other 2 years (Figure 8). In 2002, 1207 fires were detected
but in 2001 and 2003 over 1350 fires were detected. In
2002, there was a fairly steady decrease in summer and fall
burning, unlike 2001 and 2003, which exhibited a large
peak during October and November, the main months of
rice and sugar cane harvesting. The decrease in agricultural
burning during the sugar cane harvest (October and
November 2002) was followed by a slight increase in
burning in January 2003, possibly signifying a delay in
sugar harvest in 2002. The fall variability between the
3 years is dictated by the harvesting practices for rice and
sugar cane, as precipitation, the market, and labor control
when the exact harvest will take place.
3.2.2. Ukraine
[
28] MODIS detected a total number of 20,609, 14,137
and 2750 fires in the Ukraine during 2001, 2002, and 2003,
Figure 7. Spatiotemporal distribution of agricultural fire occurrences in the southeastern United States
in 2001 –2003. Each circle represents a 1-km MODIS fire pixel.
Figure 8. Interannual variability in monthly agricultural
fire occurrences in the southeastern United States.
Figure 9. Distribution of fire occurrence in Ukraine within
various land covers in 2001–2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
8of15
GB2021
respectively. Agricultural burning was the single largest
contributor to the detected fires in all 3 years (an average
of 86%) (Figure 9) and amounted to 2–13% of the global
agricultural fires. Grasslands, savannas, and shrublands
experienced insignificant levels of burning (<1%) in all
3 years. Forest fires increased by 4% from 2001 to 2003.
[
29] Cropland burning was remarkably ubiquitous across
the country, except in the Carpathian Mountains in the
southwestern corner (Figure 10). Cropland burning varied
substantially among years (Figure 11). Agricultural fires
decreased by roughly 32% in 2002 and by 87% in 2003
when compared with 2001. As the amount of agricultural
fires decreased from 2001 to 2003, the spatial distribution of
these fires remained rather simila r, except for a slight
decrease in burning in the croplands located in the north-
western (in 2002) and southeastern regions (in 2003) of the
country.
[
30] The occurrence o f agricultural fires exhibited a
seasonal shift among the 3 years. More springtime burning
was observed in the northern parts of the country in 2002
than in the other 2 years (Figure 10). As illustrated in
Figure 11, in 2001 and 2002 the annual peak in agricultural
fire activity occurred during July and August with a fairly
small springtime peak in March and April, respectively. In
2003, agricultural fires were spread more evenly throughout
the spring, summer, and fall and the annual peak occurred in
April. The monthly time series of cropland fire detection
depicted in Figure 11 correspond to the planting and
harvesting seasons of Ukrainian crops. Farmers in the
Ukraine grow mainly cereal crops, such as winter wheat,
spring wheat, corn, barley, oats, and rye [USDA/JAWF,
1994]. Winter wheat is planted in the autumn and harvested
the next summer during July and August. Spring wheat is
planted in May and harvested in August and September.
Corn is also planted in May but is harvested later in the fall
during October. Wheat (winter and spring) and corn make
up the majority of grains cultivated. The springtime peak of
fires during April and May coincided with the field prep-
aration for spring barley and oats planting. The large
increase in fires between July and August overlapped with
the harvesting of winter and spring wheat. The substantial
decrease in August crop residue fires between 2001 and
2003 (Figure 11) was likely the result of variable weather
conditions. In 2002 warm weather in February triggered an
early sowing and harvesting of spring wheat and an early
harvest of winter wheat [USDA/PECAD/FAS, 2003e], and
Figure 10. Spatiotemporal distribution of agricultural fire occurrences in Ukraine in 2001 –2003. Each
circle represents a 1-km MODIS fire pixel.
Figure 11. Interannual variability in monthly agricultural
fire occurrences in Ukraine.
Figure 12. Distribution of fire occurrence in China within
various land covers in 2001–2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
9of15
GB2021
this was reflected in the shift to an earlier peak (March) in
spring agricultural fire activity. At the same time dryness
during the summer months resulted in drops in corn yields,
which could account for the reduced August fire activity in
2002 compared to 2001 [USDA/PECAD/FAS, 2002c]. In
2003 a cycle of thawing and refreezing smothered much of
the winter wheat crop, reducing production by more than
20% [USDA/PECAD/FAS, 2003f].
3.2.3. China
[
31] China is a major contributor to the total amount of
emissions caused by biomass burning in Asia [Streets et al.,
2003]. Nevertheless, fire occurrence in China has not been
studied in great deta il outside China an d is not we ll
characterized in the English scientific literature. This case
study presents the first results of the analysis of satellite fire
observations in China for the period of 2001– 2003.
[
32] MODIS registered a total number of 12,464, 14,023
and 20,501 fires over the mainland China in 2001, 2002 and
2003, respectively. The increase in the number of fires in
2003 was mainly du e to significant fire activity in the
forests of Northern China. Streets et al. [2003] identified
forests, agricultural residue and savanna fires as the largest
sources of burning biomass. The distribution of fire counts
within various land covers during 2001–2003 (Figure 12)
revealed that fires in croplands constituted approximately
30–40% of all fire detections. Fires in forests accounted for
20–30% and fires in savannas accounted for 13– 16% of
total fire occurrences. Agricultural burning in China
amounted to 3– 6% of the global agricultural fires.
[
33] Agricultural fires followed a fairly consistent pattern
in space (Figure 13) and time (Figure 14). The low
variability of agricultural fire occurrence in the spatiotem-
poral domain is indicative of the consistent application of
fire to crop management in China. From all major crops
cultivated in China, only groundnuts and sugar beets
growing techniques do not routinely include fire use due
to the low amounts of above ground biomass produced by
these crops [FA O , 2002b]. Other crops often require the use
of fire to clear residues including, corn, rapeseed, wheat,
cotton, rice and double cropping of soybean with rice or
wheat. In all cases, crop residue burnin g o ccurs either
immediately after harvesting or before new crop planting.
Figure 14. Interannual variability in monthly agricultural
fire occurrences in China.
Figure 13. Spatiotemporal distribution of agricultural fire occurrences in China in 2001–2003. Each
circle represents a 1-km MODIS fire pixel.
Figure 15. Distribution of fire occurrence in Mexico
within various land covers in 2001 –2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
10 of 15
GB2021
Croplands of China can be grouped into three large areas,
southern, c entral and northern, characterized by similar
climatic regimes and crop types [USDA/JAWF, 1994]. The
main crops of the southern agricultural area are sugar cane
and double-crop rice. The increase in agricultural fire
activity in this area during the winter period can be
explained by the sugar cane harvesting in the period of
November through April and late double-crop rice harvest-
ing in October and November. Summer fires are most likely
associated with the clearing of the l itter following the
harvesting of early double-crop rice in June and July. The
major crops of northern China, single-crop rice, corn,
soybean and wheat, are generally harvested in the period
of August through October and planted in late April through
May with the clearing of the fields occurring during winter
and early spring. The central agricultural zone presents the
most complex combination of crop types including wheat,
soybeans, corn, cotton, rapeseed, and single- and double-
crop rice, and is therefore characterized by the most
complex pattern of fire occurrence.
[
34] There were thr ee distinct peaks in monthly agricul-
tural fire occurrence (Figure 14). The first peak in early
spring (March –April) is most likely explained by the
clearing of the fields for planting of nearly all major crops
in China. The largest peak in July coincides with the
harvesting of early double crop rice and winter wheat.
The last peak in the number of fires occurred i n late
September and October, which is generally the harvesting
period for the majority of crops planted in the spring.
3.2.4. Mexico
[
35] MODIS detected 22,035 total fires in 2001, 27,912 in
2002, and 28,751 in 2003. Savanna fires accounted for the
majority of the burning in Mexico with an average of 35%
during the 3-year period (Figure 15). Forest fires were
similarly important with an average of 33%. Agricultural
burning was the third largest contributor and accounted for
an average of 12% of all fires in Mexico from 2001 to 2003.
[
36] Much of Mexico’s agriculture burning is situated
near the coast, with the exception of the cluster of burning
in the state of Guanajuato in central Mexico (Figure 16).
Mexican agriculture includes crops such as wheat, maize,
vegetables, coffee, sugar cane, sorghum, and citrus [USDA/
JAWF, 1994]. About half of Mexico’s agriculture is grown
in the central plateau that includes the states of Me´xico,
Pueblo, and Guanajuato [Liverman, 1992]. Winter burning
occurs during the sugar cane harvest, which begins in
November and continues until April. Between April and
June, in addition to the ending of the sugar cane harvest, the
Figure 16. Spatiotemporal distribution of agricultural fire occurrences in Mexico in 2001–2003. Each
circle represents a 1-km MODIS fire pixel.
Figure 17. Interannual variability in monthly agricultural
fire occurrences in Mexico.
Figure 18. Distribution of fire occurrence in India within
various land covers in 2001–2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
11 of 15
GB2021
entire wheat harvest takes place throughout Mexico. Be-
tween the months of July and August no crops are officially
harvested [USDA/JAWF, 1994]. The late summer fire detec-
tions illustrated in Figure 16 may correspond to fiel d
preparation for the autumn planting or field maintenance
(i.e., pest control, weed control).
[
37] The peak agricultural fire activity occurred between
April and May in all 3 years. (Figure 17). This burning
coincided with the harvest of the first rice crop, the end of
the sugar cane harv est and the beginning of the wheat
harvest. In May 2002, 140% more fires were detected than
in May 2001. Approximately 232% more agricultural fires
were detected in May 2003 than in May 2001. The May
2003 increased fire activity may be associated with higher
crop production [USDA/PECAD/FAS, 2003g, 2003h].
3.2.5. India
[
38] Together with China, India dominates biomass burn-
ing emissions in Asia [Streets et al., 2003 ; Yevich and
Logan, 2003]. A total number of 15,175, 13,377 and
12,637 were detected by MODIS in 2001, 2002, and 2003,
respectively. Agricultural fires were the single largest con-
tributor, and accounted for 43–57% of all fires during the
3-year period (Figure 18). Savanna and forests fires were
similarly important with an average of 18%. Globally,
India contributed 6% of all agricultural fires.
[
39] Agricultural fires in India are extensive through out
most of the country (Figure 19) and they are used to manage
a number of crops. Agriculture in India includes crops, such
as rice, maize, sorghum, millet, sugar cane, wheat, soybeans
and rapeseed [USDA/JAWF, 1994]. Agricultural fire activity
exhibited similar magnitude and seasonal patterns during
the 3 years (Figure 20). Two distinctive peaks in agricultural
fire activity occurred during the sprin g (May) and fall
(October) of every year. The high concentration of fires in
northwestern India from March through May correspond
with the harvesting of winter wheat, rapeseed and sugar
cane. The fires during the same time in the southern states
are associated with sugar cane harvesting, in the southeast-
ern states with rice straw burning, and in the central states
with sorghum harvesting. The heightened fire application
during the fall throughout the country is associated with the
harvesting for a number of crops planted during the spring
and early summer, including maize, soybeans and cotton.
3.2.6. South Africa
[
40] The major crops cultivated in South Africa are
predominantly maize, wheat and sugar cane [USDA/
JAWF, 1994]. Within South Africa a total number of
13,419, 15,626 and 12,609 fires were detected in 2001,
2002 and 2003, respectively. Similar fire distributions by
land cover type were found in all f ire seasons and
savanna fires accounted for the majority of the burns
(Figure 21). Although fires in croplands accounted for a
relatively small percentage (7%) of all fires, pu blic
pressure to reduce burning in agricultural areas and avoid
the resulting smoke pollution hazard near towns is
increasing [South African Sugar Association Experiment
Station, 2003].
Figure 19. Spatiotemporal distribution of agricultural fire occurrences in India in 2001–2003. Each
circle represents a 1-km MODIS fire pixel.
Figure 20. Interannual variability in monthly agricultural
fire occurrences in India.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
12 of 15
GB2021
[41] Figure 22 shows monthly time series of agricultural
fire activity in South Africa in 2001, 2002 and 2003. In all
years, the fire occurrences showed a bimodal distribution
with two maxima, the first occurring during March to
April and the second during July to September. A distinct
seasonality was also revealed in the spatial distribution of
fire activity illustrated in Figure 23 and was related to
seasonal agric ultural practices. The crop planting and
harvesting periods vary throughout South Africa related
to rainfall seasonality. There are two important patterns in
the annual rainfall progression, namely, summer and win-
ter maxima regimes [Desmet and Cowling, 1999]. The
Western Cape has winter rainfall, whereas most of the
eastern parts of the country have summer rainfall [Schulze ,
1997].
[
42] Application of fires on agricultural lands in the
southwestern part of the Western Cape, mainly during
the months of March and April, is mostly related to
burning of winter wheat crop residues before the new
planting season (usually April to July) [USDA/JAWF,
1994]. The croplands in the eastern parts of the country
are mainly burned during the winter and early spring. The
fires along the eastern seaboard, extending from Northern
Pondoland in the Eastern Cape to the Mpumalanga Low-
veld, are associated with sugar cane burning pr ior to
harvesting. Sugar cane is an important export crop and
South Africa in the world’s tenth largest sugar producer.
The fires concentrated farther inland in the Maize Triangle
(describes the region comprised of the North West Prov-
ince, the northwestern, northern and eastern Free State, the
Mpumalanga Highveld and the KwaZulu-Natal Midlands)
during the late winter and spring are mostly applied to
burn maize residues after harvesting.
[
43] The monthly distribution of fires was generally
similar among years, with the exception of August 2001
(Figure 22). In August 2001, 340% more fires were detected
compared with August 2002 and 215% more fires compared
with August 2003. Similar spatial patterns were observed
among the 3 years in August, with fire activity occurring
mainly in the Mpumalanga province and along the east
coast of KwazaZulu-Natal and northern Eastern Cape. The
increased fire activity in August 2001 was also concentrated
in these maize and sugar producing regions. The drop in fire
activity in these regions compared with August 2001 was
likely related to the 2002/2003 El Nin˜o drought [USDA/
PECAD/FAS, 2002a, 2003b].
4. Summary and Conclusions
[44] Agricultural fires have been largely overlooked in
the framework of fire monitoring systems. Our assessment
from 3 years of MODIS Terra fire data from 2001 to
2003 indicates that fire is regularly used in agricultural
practices around the world and is a significant component
of global fires (8–11%). Extensive cropland burning takes
place in the Russian Federation, Ukraine, India and
China, which together contribute more than 40% of the
global agricultural fires. Distinct differences in the sea-
sonality of agricultural fire activity were observed among
regions, related to different crop types and crop calendars.
Year-to-year changes i n agricultural fire activity were
substantial in Eastern Europe and European Russia and
had a significant influence on the global-scale agricultural
fire occurrences.
[
45] This study provides a consistent spatially and tem-
porally explicit characterization of global agricultural fires,
despite the inherent inaccuracies of the underlying remotely
sensed data sets. The comparison of our results with
multiyear agricultural reports provided by the USDA for
several regions and countries reveals a general consistency
between interannual fire trends and reported agricultural
production. Additional work is evidently required to further
examine this relationship and to assess agricultural burned
area and diurnal variations in agricultural burning. A
globally comprehensive estimation of greenhouse gas emis-
sions from different crop type burns is needed in the
framework of national emissions estimation and regional
air quality studies. Agricultural burning is anthropogenic
and in principle, the timing and extent of fires can be
modified through improved management and legislation.
Satellite fire observations, such as those used in this study,
have the potential for providing an independent means to
Figure 22. Interannual variability in monthly agricultural
fire occurrences in South Africa.
Figure 21. Distribution of fire occurrence in South Africa
within various land covers in 2001–2003.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
13 of 15
GB2021
monitor trends in agricultural burning and to assess the
effectiveness of fire legislation.
[46] Acknowledgments. We acknowledge useful discussions about
the MODIS active fire product with Louis Giglio. Tony Knowles and Mike
Wallace provided useful information on agricultural fire use in South
Africa. The reviewers are thanked for their helpful comments. This work
was funded by the National Aeronautics and Space Administration (NASA)
MODIS grant NNG04HZ18C.
References
Andreae, M. O. (1991), Biomass burning: Its history, use, and distribution
and its impact on environmental quality and global change, in Global
Biomass Burning: Atmospheric, Climatic and Biospheric Implications,
edited by J. S. Levine, pp. 3 – 21, MIT Press, Cambridge, Mass.
Chidumayo, E. N. (1987), A shifting cultivation land use system under
population pressure in Zambia, Agrofor. Syst., 5, 15 –25.
Csiszar, I., L. Denis, L. Giglio, C. O. Justice, and J. Hewson (2005), Global
fire activity from two years of MODIS data, Int. J. Wildland Fire, 14,
117– 130.
DeFries, R. S., R. A. Houghton, M. C. Hansen, C. Field, D. L. Skole, and
J. Townshend (2002), Carbon emissions from tropical deforestation and
regrowth based on satellite observations for the 1980s and 90s, Proc.
Natl. Acad. Sci., 99, 14,256 –14,261.
Desmet, P. G., and R. M. Cowling (1999), The climate of the karoo—A
functional approach, in The Karoo: Ecological Patterns and Processes,
edited by W. R. J. Dean and S. J. Milton, pp. 3 – 16, Cambridge Univ.
Press, New York.
Ekboir, J. (Ed.) (2002), CIMMYT 2000 – 2001 World Wheat Overview and
Outlook: Developing No-Till Packages for Small-Scale Farmers, 74 pp.,
Int. Maize and Wheat Improve. Cent. (CIMMYT), Mexico City, Mexico.
Fischer, G., M. Shah, H. van Velthuizen, and F. O. Nachtergaele (2001),
Global Agro-ecological Assessment for Agriculture in the 21st Century,
44 pp., Int. Inst. for Appl. Syst. Anal., Laxenburg, Austria.
Food and Agriculture Organization (2002a), Global Forest Fire Assessment
1990 – 2000, For. Resour. Assess. Working Pap. 55, Rome.
Food and Agriculture Organization (2002b), FAOSTAT 2001: FAO Statis-
tical Databases [CD ROM], Rome, Italy.
Friedl, M. A., et al. (2002), Global land cover mapping from MODIS:
Algorithms and early results, Remote Sens. Environ., 83, 287 – 302.
Giglio, L., J. Descloitres, C. O. Justice, and Y. J. Kaufman (2003), An
enhanced contextual fire detection algorithm for MODIS, Remote Sens.
Environ., 87, 273 – 282.
Giglio, L., I. Csiszar, and C. O. Justice (2006), Global distribution and
seasonality of active fires as observed with the Terra and Aqua MODIS
sensors, J. Geophys. Res., 111, G02016, doi:10.1029/2005JG000142.
Hansen, M. C., R. S. Defries, J. R. G. Townshend, and R. Sohlberg (2000),
Global land cover classification at 1 km spatial resolution using a classi-
fication tree approach, Int. J. Remote Sens., 21, 1331 – 1364.
Hao, W. M., and M.-H. Liu (1994), Spatial and temporal distribution of
tropical biomass burning, Global Biogeochem. Cycles, 8, 495– 503.
Jenkins, B., S. Q. Turn, and R. B. Williams (1992), Atmospheric emissions
from agricultural burning in California: Determination of burn fractions,
distribution factors, and crop-specific contributions, Agric. Esosyst.
Environ., 38, 313 – 330.
Justice, C. O., J. P. Malingreau, and A. Setzer (1993), Satellite remote
sensing of fires: Potential and limitation, in Fire in the Environment:
Its Ecological, Climatic and Atmospheric Chemical Importance, edited
by P. Crutzen and J. Goldammer, pp. 77 – 88, John Wiley, Hoboken,
N. J.
Justice, C. O., L. Giglio, S. Korontzi, J. Owens, J. T. Morisette, D. P. Roy,
J. Descloitres, S. Alleaume, F. Petitcolin, and Y. Kaufman (2002a), The
MODIS fire products, Remote Sens. Environ., 83, 244 – 262.
Justice, C. O., J. R. G. Townshend, E. F. Vermote, E. Masuoka, R. E. Wolfe,
N. Saleous, D. P. Roy, and J. T. Morisette (2002b), An overview of
MODIS land data processing and product status, Remote Sens. Environ.,
83, 3 – 15.
Kakareka, S. V., and T. I. Kukharchyk (2003), PAH emission from the open
burning of agricultural debris, Sci. Total Environ., 308, 257 – 261.
Kaufman, Y. J., C. O. Justice, L. P. Flynn, J. D. Kendall, E. M. Prins,
L. Giglio, D. E. Ward, W. P. Menzel, and A. W. Setzer (1998), Potential
global fi re monitor ing from EOS-M ODIS, J. Geophys. Res., 103,
32,215 – 32,238.
Leff, B., N. Ramankutty, and J. A. Foley (2004), Geographic distribution of
major crops across the world, Global Biogeochem. Cycles, 18, GB1009,
doi:10.1029/2003GB002108.
Liverman, D. M. (1992), Global warming and Mexican agriculture: Some
preliminary results, in Economic Issues in Global Climate Change: Agri-
culture, Forestry, and Natural Resources, edited by J. M. Reilly and
M. Anderson, pp. 332 –352, Westview, Boulder, Colo.
Marris, E. (2005), Conservation in Brazil: The forgotten ecosystem, Nature,
437, 944 –945.
Morisette, J. T., L. Giglio, I. Csiszar, and C. O. Justice (2005), Validation of
the MODIS active fire product over Southern Africa with ASTER data,
Int. J. Remote Sens., 26, 4239 – 4264.
Morton, D., R. DeFries, and Y. Shimabukuro (2005), Quantifying cropland
expansion in cerrado and transition forest ecosystems with MODIS
satellite image time series, in Cerrado Land-Use and Conservation: As-
sessing Trade-Offs Between Human and Ecological Needs—XIX Annual
Meeting of the Society for Conservation Biology – Conservation Biology,
Capacity Building, and Practice in a Globalized World [CD-ROM],
Univ. de Bras. and Conserv. Int., Washington, D. C.
Robinson, J. M. (1991), Fir e from space: Global fire evaluation usi ng
infrared remote sensing, Int. J. Remote Sens., 12, 3 – 24.
Russell-Smith, J., Y. Cameron, A. Edwards, G. E. Allan, G. D. Cook,
P. Cooke, R. Craig, B. Heath, and R. Smith (2003), Contemporary fire
regimes of northern Australia, 1997 – 2001: Change since Aboriginal
occupancy, challenges for sustainable management, Int. J. Wildland Fire,
12, 283 –297.
Sastry, N. (2002), Forest fires, air pollution, and mortality in Southeast
Asia, Demography, 39(1), 1 – 23.
Schulze, R. E. (1997), Climate, in Vegetation of Southern Africa, edited by
R. M. Cowling, D. M. Richardson, and S. M. Pierce, pp. 21 –42, Cam-
bridge Univ. Press, New York.
Figure 23. Spatiotemporal distribution of agricultural fire occurrences in South Africa in 2001–2003.
Each circle represents a 1-km MODIS fire pixel.
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
14 of 15
GB2021
South African Sugar Association Experiment Station (2003), Managing
Natural Resources: A Canegrowers Handbook, 14 pp., KwaZulu-Natal,
South Africa.
Streets, D. G., K. F. Yarber, J. H. Woo, and G. R. Carmichael (2003),
Biomass burning in Asia: Annual and seasonal estimates and atmospheric
emissions, Global Biogeoch em. Cycles, 17(4 ), 1099, doi:10.1029/
2003GB002040.
U.S. Department of Agriculture/Joint Agricultural Weather Facility (1994),
Major World Crop Areas and Climatic Profiles, Agric. Handb., vol. 666,
279 pp., Washington, D. C.
U.S. Department of Agriculture/National Agricul tural Statistics Service
(1997), 1997 U.S. Summary and State Data,vol.1,1997 C ensus of
Agriculture, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2002a), Southern Africa: 2001/02
grain update, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assess-
ment Division/Foreign Agricultural Service (2002b), Zimbabwe: Gov-
ernment cuts 2001/02 corn production estimate, bulletin, Washington,
D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2002c), Ukraine harvests second
consecutive bumper crop, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003a), Russia: Wheat output to
drop 33 percent, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003b), Poor planting conditions
for 2002/03 corn in South Africa, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003c), Zimbabwe: Grain produc-
tion shortfall will continue, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003d), Australian wheat produc-
tion may down under, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003e), Ukraine: Winter wheat
situation, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003f), Ukraine: Winter wheat
area drops nearly 20 percent, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003g), Mexico: 2003 rice crop
slightly better than 2002, bulletin, Washington, D. C.
U.S. Department of Agriculture/Production Estimates and Crop Assessment
Division/Foreign Agricultural Service (2003h), Mexico: 2003 monsoon
not overwhelming Mexico, but crop prospects re main good, bulletin,
Washington, D. C.
U.S. Environmental Protection Agency (USEPA) (2003), Inventory of U.S.
greenhouse gas emissions and sinks: 1990– 2001, USEPA #430-R-03-
004, Washington, D. C.
Vallis, I., W. J. Parton, B. A. Keating, and A. W. Wood (1996), Simulation
of the effects of trash and N fertilizer management on soil organic matter
levels and yields of sugar cane, Soil Tillage Res., 38, 115– 132.
World Health Organization (2000), Vegetation fires, Fact Sheet 254, Geneva.
Yevich, R., and J. A. Logan (2003), An assessment of biofuel use and
burning of agricultural waste in the developing world, Global Biogeo-
chem. Cycles, 17(4), 1095, doi:10.1029/2002GB001952.
C. Justice, S. Korontzi, S. Kumar, T. Loboda, and J. McCarty,
Department of Geography, University of Maryland at College Park, 2181
LeFrak Hall, College Park, MD 20742, USA. (stef@hermes.geog.umd.edu)
GB2021 KORONTZI ET AL.: GLOBAL DISTRIBUTION OF AGRICULTURAL FIRE
15 of 15
GB2021