Spatial and temporal variability of fires in relation to ecosystems, land tenure and rainfall in savannas of northern South America M. ROMERO-RUIZ *, A. ETTER w , A. SARMIENTO w and K . T A N S E Y * *Department of Geography, University of Leicester, University Road, Leicester LE1 7RH, UK, wDepartamento de Ecologı ´a y Territorio, Facultad Estudios Ambientales y Rurales, Universidad Javeriana, Tr. 4 Nr 42-00 piso 8, Bogota ´ DC, Colombia Abstract Fire is a predominant factor forcing global terrestrial biomass dynamics, with more than 30% of the land surface showing frequent burning, particularly in the tropics, where it mostly affects savannas ecosystems annually. Savannas, which cover approximately 269 million ha in South America, play a major role in the global carbon cycle. They are affected by increasing human pressures and global climate change. Using satellite data, this study quantifies vegetation burning in the Colombian Llanos savannas for the period 2000–2008, and analyzes how fire spatial pattern, frequency and extent vary with ecosystem type, land tenure and rainfall. On average 2.75 0.5 million ha (24 4.2%) of the savannas burn each year. Burned area is highly variable, with 3.4 million ha burned in 2002–2003 and o1.9 million ha in 2005– 2006. However, during the 2000–2008 period near of 3.7 million ha (33.5%) of the savannas never burned. Compared with the average 8–10 years of fire return time for the tropics and subtropics, these savannas burn twice as often. In addition, the average burn size figure for tropical and subtropical grassland savannas (with o5% trees) of 7000 ha (median 5000 ha), is about seven times the average burned patch size we found in our study. Fires predominate in the well-drained high plain savannas, lowest figures occurring along the Andean foothills, in forested areas and in pasture and croplands. Annual proportion burned varies with land tenure, being highest in National Parks. This study is the first complete regional map of fire disturbance in a South American savanna. This detailed regional data provides a unique opportunity for increasing the accuracy of global carbon emission calculations. Keywords: burned area, Colombia, fire, Llanos Orientales, Modis, satellite Received 23 June 2009 and accepted 17 August 2009 Introduction Human induced land cover change is considered to be one of the most important processes affecting ecosystem structure, carbon and nutrient cycling, greenhouse gas emissions and land–atmosphere energy exchange, all contributing to global ecological change (Foley et al., 2005). Factors driving land cover dynamics such as fire are key actors in these processes. More than 30% of the global land surface shows significant fire frequencies particularly in the tropical belt 201N–301S (Chuvieco et al., 2008a; Tansey et al., 2008). Between 2000 and 2007 more than 3900 million ha of burned area was detected globally, 80% occurring in woodlands and scrublands, and 17% in grasslands and croplands (Cochrane, 2002; Tansey et al., 2004, 2008). Most of these fires are asso- ciated with savanna vegetation taking place in Africa, Australia, and South America (Andreae, 1992; Hao & Liu, 1994). Savannas play a role in the global carbon cycle given their belowground storage capacity, seasonal burning, regrowth and tree-grass dynamics. Savanna fires are a major source of CO 2 release to the atmosphere (Mouillot et al., 2006), and fires associated with tropical grazing systems are thought to have been contributing to 440% of globally burned phytomass (Hall & Scurlock, 1991). Van der Werf et al. (2003) in turn, estimates carbon emissions from fires in the tropics to represent 9% of the NPP (4 Pg C yr 1 ), and the portion of NPP con- sumed by in for tropical savannas to be around 25%. Emissions from fires have increased over the last 100 years, from 30% to 50% mainly in the tropical areas (Mouillot et al., 2006). Fire is a seasonal phenomenon with peaks associated with dry seasons. During the wet season normally there Correspondence: M. Romero, tel. 1 44 571 2576639, fax 1 44 571 2018326, e-mail: [email protected]Global Change Biology (2010) 16, 2013–2023, doi: 10.1111/j.1365-2486.2009.02081.x r 2009 Blackwell Publishing Ltd 2013
11
Embed
Spatial and temporal variability of fires in relation to ecosystems, land tenure and rainfall in savannas of northern South America: SPATIAL AND TEMPORAL VARIABILITY OF FIRES
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Spatial and temporal variability of fires in relation toecosystems, land tenure and rainfall in savannas ofnorthern South America
M . R O M E R O-R U I Z *, A . E T T E R w , A . S A R M I E N T O w and K . T A N S E Y *
*Department of Geography, University of Leicester, University Road, Leicester LE1 7RH, UK, wDepartamento de Ecologıa
y Territorio, Facultad Estudios Ambientales y Rurales, Universidad Javeriana, Tr. 4 Nr 42-00 piso 8, Bogota DC, Colombia
Abstract
Fire is a predominant factor forcing global terrestrial biomass dynamics, with more than 30%of the land surface showing frequent burning, particularly in the tropics, where it mostlyaffects savannas ecosystems annually. Savannas, which cover approximately 269 million ha inSouth America, play a major role in the global carbon cycle. They are affected by increasinghuman pressures and global climate change. Using satellite data, this study quantifiesvegetation burning in the Colombian Llanos savannas for the period 2000–2008, and analyzeshow fire spatial pattern, frequency and extent vary with ecosystem type, land tenure andrainfall. On average 2.75� 0.5 million ha (24� 4.2%) of the savannas burn each year. Burnedarea is highly variable, with 3.4 million ha burned in 2002–2003 and o1.9 million ha in 2005–2006. However, during the 2000–2008 period near of 3.7 million ha (33.5%) of the savannasnever burned. Compared with the average 8–10 years of fire return time for the tropics andsubtropics, these savannas burn twice as often. In addition, the average burn size figure fortropical and subtropical grassland savannas (with o5% trees) of 7000 ha (median 5000 ha), isabout seven times the average burned patch size we found in our study. Fires predominate inthe well-drained high plain savannas, lowest figures occurring along the Andean foothills, inforested areas and in pasture and croplands. Annual proportion burned varies with landtenure, being highest in National Parks. This study is the first complete regional map of firedisturbance in a South American savanna. This detailed regional data provides a uniqueopportunity for increasing the accuracy of global carbon emission calculations.
assessment of burns, at least for medium to large size
fires.
However, annual burning patterns exhibit high varia-
bility, with 43.4 million ha burned in 2002–2003 and
o1.9 million ha in 2004–2005 period (Fig. 3a and b). The
map of burning recurrence shows areas of high recur-
rence such as the southwest and northeast of the study
area, while other areas closer to the mountain range and
savannas with high replacement by pastures and plan-
tations show much lower fire recurrence (Fig. 3c). Dur-
ing the 2000–2008 period some 3.7 million ha (33.5%) of
savannas were never burned, while 1.95 million ha
(18%) was burned more than four times in 8 years.
A further important characteristic of fires and their
impact concerns the spatial patterns of scars and the
variability of their metrics such as total NP, the mean
Fig. 3 Burned area maps: (a) Burn scars (black) for a strong (2002–2003) fire season; (b) burn scars for a weak (2004–2005) fire season;
(c) fire recurrence for the period 2000–2008 (hotspots in circles are areas with high concentration of >5 in 8 years fire recurrence).
2018 M . R O M E R O-R U I Z et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, 16, 2013–2023
burned patch size and patch density. Table 2 shows a
number of average spatial metrics for the burned areas
of the period 2000–2008. The average number of an-
nually burned patches in the region was 3486 patches,
with a mean size of around 1000 ha, and a CV of 990 ha.
This gives an annual average of about one burned patch
for every 5000 ha. In general, mean patch size is well
correlated with the proportion of area burned (Fig. 4).
Relations of the burning patterns with rainfall, ecosystemtype and land tenure
The monthly average burned area and rainfall for the dry
season showed a negative relationship for all four stations
analyzed as shown in Fig. 5. The average monthly rainfall
for the dry months of December through March in the cen-
ter of the Llanos is 28.5 mm, and 13.6 mm only for January–
February. Burned area strongly decreases with monthly
rainfall values 450 mm across the whole study area (Fig. 6).
In relation to the variability in land tenure/manage-
ment and biophysical conditions, fire patterns of the
Colombian Llanos savannas show marked differences.
Replacement of savannas by pastures and plantations
drastically reduces fires, which practically disappear in
improved pastures and plantations. In savannas the
proportion of area that is burned each year varies greatly
with land tenure, being highest within National Parks
with over 28%, while in ranches this was 19% (Fig. 7a).
Variability is also highest in National Parks. During 2000–
2008, 30–32% of the area of ranches and indigenous
reserves were never burned, while for National Parks
this figure was only 16%. Overall, savanna fires are more
frequent in Indigenous reserves than in ranches. The
patterns of fire recurrence are also variable with ranches
showing the lowest recurrence, and National Park the
highest recurrence of 3 in 8 years or more (Fig. 7b).
Table 2 Spatial metrics of burned areas in the Colombian
Llanos for the period 2000–2008
Year
Percent
land area
Numbers
of scars
Average
scar
size (ha)
Area
coefficient
of variation
2000–2001 0.26 3953 952 1050.6
2001–2002 0.24 3513 925 1144.2
2002–2003 0.30 3034 1408 982.0
2003–2004 0.26 3061 1189 1122.4
2004–2005 0.16 4005 565 898.8
2005–2006 0.24 3770 899 938.7
2006–2007 0.23 3185 1019 898.2
2007–2008 0.25 3369 1048 889.0
Average 0.24 3486 1005 990.5
Fig. 4 Relationship between the total burned area and the
average burn scar size for the Colombian Llanos savannas.
Fig. 5 Monthly rainfall in the dry season and area burned within 50 km radius for climate station Las Gaviotas (see Fig. 1).
Fig. 6 Regression between the dry season monthly rainfall and
area burned within 50 km radius for four rainfall stations (see
Fig. 1) across the Colombian Llanos in the period 2000–2008.
S PA T I A L A N D T E M P O R A L VA R I A B I L I T Y O F F I R E S 2019
r 2009 Blackwell Publishing Ltd, Global Change Biology, 16, 2013–2023
Differences in savanna ecosystem type also had a
strong impact on fire patterns, with the average annual
proportion of burned area by ecosystem varying be-
tween 3% and 38% per year. The highest figures occur in
the well-drained high-plain savannas south of the River
Meta especially in the sandy savannas (39%) (Fig. 8a).
The lowest figures occur along the foothills of the
Andes. Also, between 22% and 30% of the area of
well-drained high-plain savannas were not burned dur-
ing the 2000–2008 period. In contrast more than 80% of
the alluvial overflow plain savannas were free from fire
in the same period (Fig. 8b). The ecosystem type with
the highest fire recurrence was again the sandy high-
plain savanna type, where 230 000 ha (36%) burned in
four or more years during the study period.
Discussion
Recent global studies have confirmed that the savanna
biome is the most continuously affected by burning
with strong environmental impacts (Chuvieco et al.,
2008b; Tansey et al., 2008). According to the figures of
Van der Werf et al. (2003) and Tansey et al. (2004) our
results indicate that burned areas of the Colombian
Llanos savannas correspond to between 3% and 6% of
the globally burned herbaceous vegetation area, and
about 4% of the annually burned area in South America.
Multi-satellite data has become a major means to
monitor changes in land cover, including fires at the
global, regional and local scales (Roy et al., 2005).
Although this has made possible to map areas affected
by fires and analyze burning dynamics, the differences
in the spatial, temporal and spectral resolutions of
satellite data, as well as in the processing algorithms
often leads to large differences in results (Roy et al.,
2005). A recent study by Chuvieco et al (2008b) using 32
day composition MODIS data, calculated that 1.2% of
the Colombian territory is burned annually, which is
only about half of the 2.75 million ha we found for the
Colombian Orinoco region alone. This discrepancy can
be attributed to the algorithms used for detection, the
definition of the dry season, the geographical location of
the affected areas and the type of sensor used (Loboda &
Csiszar, 2007). The algorithms used by these studies do
not take into account factors such as the meteorological
conditions where the fire occurs, the type and structure
of the landscape, land use and cover, biomass, distribu-
tion and heterogeneity and combustion rates (Loboda &
Csiszar, 2007). Our results agree with Roy et al. (2005)
who point to the need to develop algorithms that are
sensitive to ecosystem variability and to the character-
istics of the fire regime of specific environments.
The commission and omission errors in our results of
validation may be attributed to the timing of Landsat
and CBERS image acquisition dates and the periods of
MODIS image composition. Some burned areas were
not identified by our algorithm because the fire occur-
rence was at the beginning of the period of the 8-day
Fig. 7 Burning patterns for different land tenure types for the
period 2000–2008: (a) average annual proportion burned and
standard deviation; (b) fire recurrence in 8 years.
Fig. 8 Burning recurrence for different ecosystem types:
(a) average proportion annually burned and standard deviation;
(b) fire recurrence in 8 years.
2020 M . R O M E R O-R U I Z et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, 16, 2013–2023
MODIS composition, while the high-resolution images
(Landsat and Cbers) were taken at the end of the 8-day
composition. The precision in large burned areas shows
a high accuracy level leading to an acceptable overall
Kappa (k) value (0.79), which compensates for the
omission errors that arise from the fact that MODIS
images do not detect small scars (o10 ha).
Temporal and spatial variability of fires
Our results show that the fire activity in the Colombian
Llanos savannas produces a changing seasonal mosaic
of burned and unburned vegetation patches, such as
found in other savannas of the world (Laris, 2002).
However, we found that the fire recurrence has a strong
spatial pattern, with some areas showing consistently
high burning rates while other areas exhibiting consis-
tently low fire recurrence. As observed by Chuvieco
et al. (2008a) for other high fire frequency areas in the
tropics, the Colombian Llanos savannas are character-
ized by low accessibility, mostly extensive low invest-
ment uses, low-density population and dominance of
grass cover, and fires are predominantly of high density,
low duration and high spatial variability.
Compared with the average 8–10 years of fire return
time for the tropics and subtropics that was estimated
by Van der Werf et al. (2003), the savannas in the
Colombian Llanos burn twice as often. Compared
with the savannas of northern Brazil the average
4 years fire return time in the Colombian Llanos
savannas is almost twice the 2.5 years reported by
Barbosa & Fearnside (2005b). However, if the compar-
ison is made with only the well-drained sandy savan-
nas, the figures of the Llanos savannas and Northern
Brazil turn to be very similar.
Van der Werf et al. (2003) give an average burn size
figure for tropical and subtropical grassland savannas
(with o5% trees) of 7000 ha (median 5000 ha), which is
about seven times the average burned patch size we
found in our study. We found that except for the 2004–
2005 fire season which was a low burn year, the
interannual average burned patch size was very con-
stant. Several studies have shown that fire intensity
changes throughout the dry season, with early fires
burning vegetation more superficially and leaving
a more fragmented pattern in the landscape, while
with the dry season progressing fires become more
intense, more continuous and larger (Louppe et al.,
1995; Russell-Smith et al., 1997). We found that fire
areas tend to be smaller at the beginning of the dry
season, becoming much larger towards the middle of
the dry season and again smaller at the end of the
season.
Relation of the burning patterns with rainfall, land tenureand ecosystem type
Fire frequency and pattern is one of the dominant
factors defining heterogeneity of savanna environ-
ments, and man has been identified as the main cause
of fire production in savanna ecosystems (Hudak et al.,
2004). This phenomenon has significant ecological im-
plications in terms of species composition and ecosys-
tem functionality. In a broad context it has been shown
that the extension, frequency and severity of fires,
depends factors such as climate, topography and soil,
is much influenced by the structure and composition of
the vegetation (Shvidenko & Nilsson, 2000). Our results
show a marked realtionship with ecosystem types (ve-
getation, soil, hydrology), confirming the above state-
ment. Structure and composition of vegetation tend to
determine fire activity in the long-term, while changes
in climatic factors have a strong relation with fire
activity in the short-term.
Temperature and precipitation are the two most
highly correlated climatic factors with the levels of
presence and hazard fires (Finney, 1998). Barbosa &
Fearnside (2005b) found a high correlation between
the area burned and rainfall in the Roraima savannas
of the Brazilian Amazon. These authors also discuss the
importance of precipitation and the different types of
ecosystems in the occurrence of a fire. The results of our
study show the same tends, where a sharp decrease in
fires occur in months where rainfall exceeds 50 mm.
Fire has been recognized as an indicator of anthro-
pogenic activity to help identify critical areas of defor-
estation and changes in land use (Malingreau &
Gregoire, 1996; Ehrlich et al., 1997). Barbosa & Fearnside
(2005b) found that the majority of the fires in the
savannas of northern Brazil are started by humans,
but that the spread of fires is independent of human
presence. We found that fires vary in terms of their
frequency and pattern for different savannas types in
the Colombian Llanos. The unexpectedly overall pro-
portion and large size of fires in uninhabited National
Park, can be explained by the periodic movement of
indigenous people living in Reservations around it that
move across the conservation area for communication
and hunting (Correa et al., 2006). According to our
results the fire frequency and burned proportion is
somewhat higher in indigenous reserves than in cattle
ranches. The constant burning in ranches is associated
with the need to deliver fresh regrowth for livestock
due to the low palatability and forage quality of grasses
in the native savannas and the control of woody plants.
In indigenous lands however, fire is more associated
to hunting activities and to the savannas ‘cleaning’ from
ticks and other. Studies concerning the effects of land
S PA T I A L A N D T E M P O R A L VA R I A B I L I T Y O F F I R E S 2021
r 2009 Blackwell Publishing Ltd, Global Change Biology, 16, 2013–2023
tenure and land use differences on fire patterns in
savanna environments are very scarce. In other ecosys-
tem contexts such as the Amazon basin indigenous
reserves and national parks have contributed to the
reduction of deforestation rates associated to burning,
with fires four (indigenous reserves) to nine (parks)
times higher outside than inside them (Nepstad et al.,
2006).
In the Colombian savannas, transformation has been
relatively slow. However, during the last few years
increasingly large areas of savannas are being converted
into improved pastures, plantations (oil palm and tim-
ber) and annual crops, where fire activity is being
reduced or eliminated from these landscapes and their
immediate surroundings. The changes in the magni-
tude and frequency of disturbance variables such as
fires and biotic composition that normally drive the
state and cycles of these ecosystems is expected to have
large impacts on the local and regional ecological func-
tioning.
An unexplained finding of this study that needs
further study is the unexpectedly high proportion of
savannas not directly related with intensive land uses
that were not burned during the 2000–2008 period,
while some areas had a consistently high frequency.
This should be expected to have important conse-
quences for the local savanna vegetation dynamics
(Campos & Nepstad, 2006).
Conclusions
Currently, change modelling is demanding increasingly
accurate and spatially explicit regional and local data
and knowledge on land cover dynamics and change
processes such as deforestation, burning, irrigation,
agricultural intensification and urbanization. This is
the more urgent in dynamically changing landscapes
of developing countries with fast population growth
and industrialization processes. Our study contributes
the first complete regionally accurate mapping for al-
most a decade of fire activity in the Colombian Llanos
savannas. Specifically, we show how fire occurrence
(pattern, frequency and extent) vary at the sub-regional
level due to ecosystem, land tenure and rainfall, essen-
tial to the understanding of fire patterns at the regional
scale. We have shown that the increasing trend of the
savanna clearing and intensification process is changing
the fire dynamics and burning patterns. However, these
land use and land cover changes need also be analyzed
in a broader context of the impacts on ecological com-
ponents and processes, including biodiversity, soils,
water and fauna, such as termites and ants. Currently,
Savannas are considered the most important biome in
terms of fire events in South America. Between 2000 and
2008 the savannas of northern South America contrib-
uted with 25% of the overall burned areas of the
continent (Chuvieco et al., 2008b). Of this, the Colom-
bian Llanos savannas contributed with 65% whereas
Venezuelan Llanos savannas with 35%. Burning in this
region constitutes a globally significant 3–6% of herb-
aceous vegetation, an annual average of 2.75 million ha.
Acknowledgements
The Alban Program of the European Union, the Colfuturo Fund(Colombia) and The Nature Conservancy – Northern AndesProgram provided funding to Milton Romero. The authors wouldlike to thank Juan Carlos Berrio for providing support in all thisprocess and Javier Medina for his comments to this document.
References
Andreae MO (1992) Biomass burning: its history, use and distribution and
its impact on environmental quality and global climate. In: Global
Biomass Burning, Atmospheric, Climatic, and Biospheric Implications (ed.
Levine J), pp. 3–21. MIT press, Cambridge, MA.
Asner GP, Elmore AJ, Olander LP, Martin RE, Harris AT (2004) Grazing
systems, ecosystem responses, and global change. Annual Review of
Environment and Resources, 29, 261–299.
Barbosa RI, Fearnside PM (2005a) Above-ground biomass and the fate of
carbon after burning in the savannas of Roraima, Brazilian Amazonia.
Forest Ecology and Management, 216, 295–316.
Barbosa RI, Fearnside PM (2005b) Fire frequency and area burned in the
Roraima savannas of Brazilian Amazonia. Forest Ecology and Manage-
ment, 204, 371–384.
Brannstrom C, Jepson W, Filippi AM, Redo D, Xu Z, Ganesh S (2008) Land
change in the Brazilian Savanna (Cerrado), 1986–2002: comparative ana-
lysis and implications for land-use policy. Land Use Policy, 25, 579–595.
Campos M, Nepstad D (2006) Smallholders, the Amazon’s new conserva-
tionists. Conservation Biology, 20, 1553–1556.
Cochrane M (2002) Wildfire Tropical Forest Fires in Latin America and the
Caribbean. Prevention, Assessment and Early Warning. United Nations
Environment Programme – UNEP, Washington.
Correa HD, Ruiz SL, Arevalo LM (2006) Plan de accion en Biodiversidad de la
cuenca del Orinoco – Colombia (2005–2015): Propuesta tecnica. Corporino-
quia, Cormacarena, Instituto Humbodt, Unitropico, Fundacion Oma-
cha, Fundacion Horizonte Verde, Universidad Javeriana, Unillanos,
WWF, Colombia GTZ, Bogota.
Coutinho LM (1990) Fire in the ecology of the Brazilian cerrado. In: Fire in
the Tropical Biota – Ecosystem Processes and Global Challenges (ed. Gold-
ammer JG), pp. 82–105. Springer-Verlag, Berlin.
Chuvieco E, Giglio L, Justice C (2008a) Global characterization of fire
activity: toward defining fire regimes from Earth observation data.
Global Change Biology, 14, 1488–1502.
Chuvieco ES, Opazo W, Sione H et al. (2008b) Global burned-land
estimation in Latin America using modis composite data. Ecological
Applications, 18, 64–79.
Delmas R, Lacaux JP, Brocard D (1995) Determination of biomass burning
emission factors: methods and results. Environmental Monitoring and
Assessment, 38, 181–204.
Dwyer E, Pereira JMC, Gregoire J-M, DaCamara CC (2000) Characteriza-
tion of the spatio-temporal patterns of global fire activity using satellite
imagery for the period April 1992 to March 1993. Journal of Biogeogra-
phy, 27, 57–69.
2022 M . R O M E R O-R U I Z et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, 16, 2013–2023