UC Irvine
Faculty Publications
Title
The Impact of Boreal Forest Fire on Climate Warming
Permalink
https://escholarship.org/uc/item/2t07n8d3
Journal
Science, 314(5802)
ISSN
0036-8075 1095-9203
Authors
Randerson, J. T
Liu, H.
Flanner, M. G
et al.
Publication Date
2006-11-17
DOI
10.1126/science.1132075
Copyright Information
This work is made available under the terms of a Creative Commons Attribution License,
availalbe at https://creativecommons.org/licenses/by/4.0/
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
The Impact of Boreal Forest Fire
on Climate Warming
J. T. Randerson,
1
*
H. Liu,
2
M. G. Flanner,
1
S. D. Chambers,
3
Y. Jin,
1
P. G. Hess,
4
G. Pfister,
4
M. C. Mack,
5
K. K. Treseder,
1
L. R. Welp,
6
F. S. Chapin,
7
J. W. Harden,
8
M. L. Goulden,
1
E. Lyons,
1
J. C. Neff
9
E. A. G. Schuur,
5
C. S. Zender
1
We report measurements and analysis of a boreal forest fire, integrating the effects of greenhouse
gases, aerosols, black carbon deposition on snow and sea ice, and postfire changes in surface
albedo. The net effect of all agents was to increase radiative forcing during the first year (34 ± 31
Watts per square meter of burned area), but to decrease radiative forcing when averaged over an
80-year fire cycle (−2.3 ± 2.2 Watts per square meter) because multidecadal increases in surface
albedo had a larger impact than fire-emitted greenhouse gases. This result implies that future
increases in boreal fire may not accelerate climate warming.
A
rctic and boreal regions are warming
rapidly, with multiple consequences for
northern ecosystems and global climate
(1). In boreal ecosystems, future increases in air
temperatur e m ay lengthen the fir e season and
increase the probability of fires, leading some to
hypothesiz e a positive fe edback between w arming,
fire activity , carbon loss, and future climate ch ange
(2, 3). Although CO
2
and other greenhouse gases
emitted by fire contribute to climate warming,
understand ing t he net effect of a changing fire
regime on climate is challenging because of the
multiple ways by which fires influence atmospher-
ic composition and the land surface. Emissions of
aerosols, for example, can lead to either warming
or cooling at a region al scale, depending on factors
such as aero sol composition and the underly ing
albedo of both the Earth’s surface and clouds (4).
Subsequent deposition of black carbon aerosols on
glaciers, snow , sea ice, and the Greenland ice sheet
may reduce surface albedo (5), causing both
atmospheric heating (6) and enhanced surface
melting. Within a burn perimeter, combined
changes in ecosystem structure and species compo-
sition after fire cause net radiation and sensible heat
fluxes to decline substantially (7, 8). These changes
in the local surface energy budget persist for
decades and are probably regionally variable.
Concurrently , accumulation of carbon in organic
soils and vegetation during intermediate succes-
sional stages offsets the pulse of carbon released
during combustion (9).
Understanding the net effect of these processes
(and their temporal and spatial scales) is important
in managing northern forests to mitigate the
climate impacts of fossil fuel emissions. Although
changes in boreal forest albedo can have a
considerable cooling effectonNorthernHemi-
sphere climate (10, 11), these changes are offset by
accompanying changes in carbon accumulation
(12), so the net effect of land cover change on
climate may be close to neutral at a global scale
when both surface energy balance and CO
2
fluxes
are considered (13). Here we applied the concept
of radiative forcing (12) to assess quantitatively
the net effe ct of a boreal fore st fire on climate, on
the basis of carbon and surface energy budget
measurements that we made in a fire chronose-
quence of black spruce (Picea mariana)ininterior
Alaska. We considered two time scales: the year
immediately after fire and an 80-year period
during which species composition and ecosystem
structure returned to a prefire mature successional
state as defined by an adjacent unburned control
stand.
The Donnelly Flats crown fire occurred during
1 1 to 18 June 1999 in interior Alaska (63°55′N;
145°44′W)andburned~7600ha(14). The fire
was intense (e.g., figs. S1 and S2), causing stand-
replacing mortality of the black spruce within the
burn perimeter and consuming much of the soil
organic matter above the mineral horizon (15).
Aboveground fuel consumption from overstory
and understory vegetation was estimated with a
combinat ion of harves ting, allometry, and invento-
ry methods. Postfire soil respiration losses during
the first year after fire were estimated with a
combination of chamb er measuremen ts and eddy
covariance measurements. Pr ecision spectral pyrano-
meters (Eppley Laboratory , Inc., Newport, RI)
measured incoming and outgoing shortwave
radiation above the canopy (and thus surface
albedo) during July and August of 1999 within
the burn perimeter (7) and then mostly continuously
from October 1999 through September 2004 at
both the burn and control.
We converted field measurements of carbon
loss during the fire to CH
4
and CO
2
fluxes using
emission factors (16). Radiative forcing from thes e
greenhouse gases was estimated with equations
derived from a global radiative transfer model (17).
In our figures and table, we report global annual
mean radiative forcing (in W) per m
2
of burned
area, with radiative forcing defined following the
Intergover nmental Panel on Climate Change Third
Assessment Report convention as the change in net
radiation at the tropopause after stratospheric ad-
1
Department of Earth System Science, University of
California, Irvine, CA 92697, USA.
2
Department of Physics,
Atmospheric Science, and General Science, Jackson State
University, Jackson, MS 39217, USA.
3
Australian Nuclear
Science and Technology Organization, Environmental
Division, Menai, NSW 2234, Australia.
4
Atm ospheric
Chemistry Division, National Center for Atmospheric
Research, Boulder, CO 80301, USA.
5
Department of
Botany, University of Florida, Gainesville, FL 32611, USA.
6
Environmental Science and Engineering, California In-
stitute of Technology, Pasadena, CA 91125, USA.
7
Institute
of Arctic Biology, University of Alaska, Fairbanks, AK
99775, USA.
8
U.S. Geological Survey, Menlo Park, CA
94025, USA.
9
Geological Sciences and Environmental
Studies, University of Colorado at Boulder, Boulder, CO
80309, USA.
*To whom correspondence should be addressed. E-mail:
jranders@uci.edu
Fig. 1. Midday surface al-
bedo within the burn pe-
rimeter of the Donnelly
Flats fire (A)andfromthe
adjacent black spruce stand
that served as a control (B).
Summer albedo progressively
increased during each year
andexceededvaluesatthe
control site ~3 years after
fire. Snow events, including
one in late May of 2000,
caused spikes that are visible
at both the burn and con-
trol sites.
17 NOVEMBER 2006 VOL 314 SCIENCE www.sciencemag.org1130
REPORTS
justmen t (18). CH
4
was assumed to have a 10-year
atmospheric lifetime. The lifetime of the CO
2
anomaly from the fire was estimated with a
combination of ocean impulse-response functions
from the Joos and Siegenthaler ocean carbon
model (19) and a postfire trajectory of net eco-
system production (NEP) that we constructed
using mass balance constraints and eddy co-
variance measurements (figs. S3 and S4). W e used
the Column Radiation Model (20)toestimate
radiative forcing from changes in surface albedo
within the Donnelly Flats burn perimeter (figs. S5
to S8). The persistence of albedo changes in
postfire ecosy stems was as sessed from an analysis
of MODerate Resolution Imaging Spectro-
radiometer (MODIS) albedo measurements (21)
within burn perimeters of known ages (14) across
interior Alaska. We derived radiative forcing from
the fire-induced ozone anomaly using simulations
from the National Center for Atmospher ic Re-
search Community Atmosphe re Model version 3
(CAM 3) of the 2004 Alaska and Yukon fire
complex (22) scaled to the carbon emission levels
that we measured for the Donnelly Flats fire.
Similarly , we estimated radiative forcing from the
direct effect of aero sols and deposition of black
carbon on snow and sea ice by injecting emissions
from the Donnelly Flats fire into CAM 3 (23, 24).
In the Supporting Online Material we provide
more information about our methods for estimat-
ing radiative forcing, as well as an additional set
of forcing estimates that take into account the
efficacy of the different agents (25).
During the fire event, 206 ± 110 g C m
−2
were emitted by combustion from the bl ack
spruce overstory, 107 ± 74 g C m
−2
from the
vascular plant understory , and 1246 ± 600 g C m
−2
from the duff layer composed of mosses, lichens,
roots, partially decomposed plant litter, and
humus. Total fuel consumption for the Donnelly
Flats fire (1560 ± 610 g C m
−2
)wassimilarto
other estimates for boreal North America, in-
cluding 1580 g C m
−2
for moderately severe fires
in boreal North America (26) and 1300 g C m
−2
for the mean of Canadian boreal forests (27). In-
cluding additional soil respiration losses of 202 ±
53gCm
−2
year
−1
during the first year after fire,
the ecosystem lost a total of 1760 ± 620 g C m
−2
.
Radiative forcing from long-lived greenhouse
gases (CH
4
and CO
2
) contributed a total of 8 ±
3Wm
−2
during the first year. Deposition of
black carbon on snow and sea ice added another
8±5Wm
−2
. An increase in tropospheric ozone
from fire-emitted trace gases generated a positive
radiative forcing of 6 ± 4 W m
−2
. Fire-emitted
aerosols mixed widely across arctic and boreal
regions (fig. S9), decreased net radiation at the
surface (−90 ± 35 W m
−2
), but did not substan-
tially change radiative forcing (17 ± 30 W m
−2
).
Changes in surface albedo within the fire perim-
eter offset positive radiative forcing from the
other agents. Specifically, the loss of overstory
canopy after fire led to increased snow exposure
during spring and fall (fig. S10), higher albedo
(Fig. 1), and a negative annual radiative forcing
(−5±2Wm
−2
). The combined effect of all
forcing agents was 34 ± 31 W m
−2
during year
1(Table1).
After the first year, the short-lived effects of
ozone, aerosols, and black carbon deposition
were no longer important, so the net effect of
the fire on radiative forcing reflected the
balance between the persistence of postfire
changes in surface albedo and the effects from
the remaining greenhouse gases in the atmo-
sphere. During the first 5 years after fire, summer
albedo progressively increased (Fig. 1), probably
from an increase in grass and shrub cover and
partial loss of black carbon that initially coated
soil surfaces and dead black spruce boles. This
strengthened the negative radiative forcing
from postfire albedo changes, with this quan-
tity decreasi ng from −5±2Wm
−2
during the
firs t year to −8±3Wm
−2
during the period 3
to 5 years after fire. Analysis of MODIS sat-
ellite data from nearby forest stands provided
evidence that spring and summer albedo typ-
ically remains elevated for about three decades
after fire and that recovery to prefire albedo
levels requires ~55 years (Fig. 2).
We predicted that the greenhouse gas pulse
from the Donnelly Flats fire should gradually
decline over a period of 5 decades, owing to CH
4
oxidation and CO
2
uptake by the oceans and
regrowing vegetation within the burn perimeter
(fig. S3D). During this interval, the greenhouse
gases will contribute to a positive radiative forcing
Table 1. Radiative forcing associated with the Donnelly Flats fire.
Forcing agent
Radiative forcing*
[W (m
2
burned)
−1
]
Year 1 Years 0 to 80 (mean)
Long-lived greenhouse gases (CH
4
and CO
2
) 8 ± 3 1.6 ± 0.8
Ozone 6 ± 4 0.1 ± 0.1
Black carbon deposition on snow 3 ± 3 0.0 ± 0.0
Black carbon deposition on sea ice 5 ± 4 0.1 ± 0.1
Aerosols (direct radiative forcing)† 17 ± 30 0.2 ± 0.4
Impact at the surface: −90 W ± 35 m
−2
Changes in post-fire surface albedo −5±2 −4.2 ± 2.0
Total‡ 34 ± 31 –2.3 ± 2.2
*All the radiative forcing estimates reported here represent annual mean values (in W) for the global atmosphere associated
with burning of a 1-m
2
area within the perimeter of the Donnelly Flats fire. We report values averaged over year 1 and for the
mean of the 0- to 80-year period after fire (and including the fire event). †We did not estimate indirect effects of aerosols
on radiative forcing as mediated, for example, by cloud drop sizes or cloud lifetime (4). Although uncertain, indirect aerosol
effects are thought to contribute to negative radiative forcing (18, 25) and would offset other positive radiative forcing agents
during year 1. ‡Accounting for the efficacy of the different forcing agents (25), the total effective forcing of the Donnelly
Flats fire was 18 ± 42 W m
−2
during year 1 and −2.4 ± 2.3 W m
−2
during years 0 to 80 (tables S1 and S2).
Fig. 2. Postfire albedo dur-
ing (A) spring (Julian Days
33 to 113) and (B) summer
(Julian Days 145 to 241)
from MODIS satellite obser-
vations extracted from burn
scars of different ages in
interior Alaska (circles and
solid line, left axis). A control
was constructed from the
mean of evergreen conife r
vegetation that did not burn
in the last 55 years (dashed-
dotted line, left axis). Annual
radiative forcing as esti-
mated from tower measure-
ments of albedo from burn
and control sites during 2002
to 2004 was −8±3Wm
−2
[(A), triangle, right axis]. The
longer-term postfire trajectory
of albedo-driven annual radi-
ative forcing was assumed to
follow the MODIS albedo
pattern [(A), dashed line,
right axis]. Years with lim-
ited burned area were ex-
cluded from the analysis.
www.sciencemag.org SCIENCE VOL 314 17 NOVEMBER 2006 1131
REPORTS
(Fig. 3A). After ~60 years, continued uptake by
the postfire ecosystem should cause atmospheric
CO
2
to decrease below background levels,
subsequent withdrawal of CO
2
from the ocean,
and a negative radiative forcing. As a result of this
trajectory and concurrent changes in surface
albedo, the influence of the fire on radiative
forcing depe nds on the averaging period (Fig.
3B). Averaged over years 0 to 80, net radiative
forcing from the different forcing agents was
−2.3 ± 2.2 W m
−2
(Table 1).
A change in fire return times will have
consequences for climate forcing (Fig. 3C), based
on the time-since-fire trajectories of the different
forcing agents estimated from the Donnelly Flats
fire, combined with a stand age model (28). If the
fire return time decreases [as has been suggested
from future warming and drying in continental
interiors (29)], loss of carbon will increase
radiative forcing (Fig. 3C). Accounting for all
forcing agents, however, leads to a small negative
radiative forcing at the global scale (Fig. 3C) and
calls into question the positive feedback that has
been suggested in past work. The cooling from a
decrease in fire return times is likely to be sub-
stantially larger in the Northern Hemisphere,
taking into account the spatial pattern of the
temperature anomalies resulting from the dif-
ferent forcing agents. Specifically , radiative forcing
from greenhouse gases has a widely distributed
impact on global temperature (18), whereas the
influence of postfire changes in surface albedo
will be concentrated almost entirely in northern
regions (10, 11, 13, 18).
For the boreal biome as a whole, key factors
that are likely to determine the balance between
negative and positive radiative forcing associated
with fire include burn severity, species estab-
lishment in postfire ecosystems, and the duration
of winter snow cover. Increased burn severity, for
example, may increase aerosol and greenhouse
gas emissions, but it is not clear to what extent
this may be canceled by greater loss of canopy
overstory and consequently higher albedo val-
ues during winter and spring. Another unre-
solved question involves the extent to which
fire in Siberian larch forests, which are needle-
leaf deciduous, has the same influence on post-
fire surface albedo as reported here for North
American needleleaf evergreen forests. Decreases
in spring snow cover (30) may weaken neg-
ative feedbacks associated with postfire in-
creases in surface albedo documented in North
America.
Future interactions between the land sur-
face and climate in northern regions may
involve both negative feedbacks within the
boreal interior (via mechanisms outlined here)
and positive feedbacks involving shrub and
forest expansion in arctic tundra ecosystems
(31) and loss of snow cover. Our analysis
illustrates how ecosystem processes that gen-
erate carbon sources and sinks have insep-
arable consequences for other forcing agents
(12, 13, 32, 33). To the extent that the con-
temporary Northern Hemisphere carbon sink
originates from changes in northern forest
cover and age (34), its value from a climate
perspective requires a more nuanced view that
encompasses all agents of radiative forcing.
Important next steps include reducing uncer-
tainties associated with direct and indirect
aerosol effects and disturbance-linked changes
in albedo, exploring the combined impacts of
feedbacks of the forcing agents estimated here
within climate models, and extending this
approach to assess the radiative forcing asso-
ciated with land-cover transitions in temperate
and tropical ecosystems.
References and Notes
1. ACIA, “Arctic Climate Impact Assessm ent” (Cambridge
Univ. Press, Cambridge, UK, 2005).
2. W. A. Kurz, M. J. Apps, B. J. Stocks, W. J. A. Volney,
in Biotic Feedbacks in the Global Climate System:
Will the Warming Speed the Warming? G. M. Woodwell,
F. Mackenzie, Eds. (Oxford Univ. Press, Oxford, UK,
1995), pp. 119–133.
3. E. S. Kasischke, B. J. Stocks, Fire, Climate Change, and
Carbon Cycling in the Boreal Forest. M. M. Cadwell et al.,
Eds., Ecological Studies (Springer, New York, 2000).
4. V. Ramanathan, P. J. Crutzen, J. T. Kiehl, D. Rosenfeld,
Science 294, 2119 (2001).
5. S.G.Warren,W.J.Wiscombe,J. Atmos. Sci. 37, 2734 (1980).
6. J. Hansen, L. Nazarenko, Proc. Natl. Acad. Sci. U.S.A.
101, 423 (2004).
7. S. D. Chambers, F. S. Chapin, J. Geophys. Res. 108,
8145 (2002).
8. H. Liu, J. T. Randerson, J. Lindfors, F. S. Chapin, J. Geophys.
Res. 110, D13101 (2005).
9. J. W. Harden et al., Global Change Biol. 6, 174 (2000).
10. G. B. Bonan, D. Pollard, S. L. Thompson, Nature 359,
716 (1992).
11. P. K. Snyder, C. Delire, J. A. Foley, Clim. Dyn. 23, 279 (2004).
12. R. A. Betts, Nature 408, 187 (2000).
13. V. Brovkin et al., Global Change Biol. 10, 1253 (2004).
14. The Alaska Fire Service maintains a database of annual
fire statistics and geographic information system (GIS)
burn perimeters (http://agdc.usgs.gov/data/blm/fire/).
15. J. C. Neff, J. W. Harden, G. Gleixner, Can. J. For. Res. Rev.
Can. Rech. For. 35, 2178 (2005).
16. M. O. Andreae, P. Merlet, Global Biogeochem . Cycles 15,
955 (2001).
17. G. Myhre, E. J. Highwood, K. P. Shine, F. Stordal,
Geophys. Res. Lett. 25, 2715 (1998).
18. V. Ramaswamy et al.,inClimate Change 2001: The
Scientific Basis. Contributions of Working Group 1 to the
Third Assessment Report of the Intergovernmental Panel
on Climate Change, J. T. Houghton et al., Eds.
(Cambridge Univ. Press, Cambridge, UK, 2001),
pp. 350–416.
19. I. G. Enting, T. M. L. Wigley, M. Heimann, “Future
Emissions and Concentrations of Carbon Dioxide: Key
Ocean/Atmosphere/Land Analyses” (Technical paper no.
31, Commonwealth Scientific and Industrial Research
Organization Division of Atmospheric Research, 2001).
20. B. P. Briegleb, J. Geophy. Res. 97, 7603 (1992).
21. C. B. Schaaf et al., Remote Sens. Environ. 83, 135 (2002).
22. G. Pfister et al., Geophys. Res. Lett. 32, L11809 (2005).
23. P. J. Rasch, W. D. Collins, B. E. Eaton, J. Geophys. Res.
106, 7337 (2001).
24. M. G. Flanner, C. S. Zender, J. Geophys. Res. 111,
D12208 (2006).
25. J. Hansen et al., J. Geophys. Res. 110, D18104 (2005).
26. E. S. Kasischke et al., Global Biogeochem. Cycles 19,
GB1012 (2005).
27. B. D. Amiro et al., Can. J. For. Res. Rev. Can. Rech. For.
31, 512 (2001).
28. E. A. Johnson, Fire and Vegetation Dynamics: Studies
from the North American Boreal Forest, H. J. B. Birks, Ed.,
Cambridge Studies in Ecology (Cambridge Univ. Press,
Cambridge, UK, 1992).
29. M. D. Flannigan, K. A. Logan, B. D. Amiro, W. R. Skinner,
B. J. Stocks, Clim. Change 72, 1 (2005).
30. Z. M. Kuang, Y. L. Yung, Geophys. Res. Lett. 27, 1299 (2000).
31. F. S. Chapin et al., Science 310, 657 (2005).
32. G. P. Robertson, E. A. Paul, R. R. Harwood, Science 289,
1922 (2000).
33. R. A. Pielke et al., Philos. Trans. R. Soc. London Ser. A
360, 1705 (2002).
34. C. L. Goodale et al., Ecol. Appl. 12, 891 (2002).
35. This work was supported by NSF and NASA grants (OPP-
0097439 and NNG04GK49G, respectively). We thank
M. Prather for advice and C. Dunn, J. Henkelman,
J. Raymond, and J. Garron for technical assistance.
Supporting Online Material
www.sciencemag.org/cgi/content/full/314/5802/1130/DC1
Materials and Methods
SOM Text
Figs. S1 to S10
Tables S1 and S2
References and Notes
5 July 2006; accepted 3 October 2006
10.1126/science.1132075
Fig. 3. (A) Annual radiative forcing from long-
lived greenhouse gases and the postfire trajectory
of surface albedo. (B) Cumulative annual radiative
forcing for the different forcing agents averaged
over the time since the fire (or equivalently, the age
of the stand). (C) Climate forcing of the different
components as a function of the fire return time
relative to a distribution of stands at steady state
with a mean fire return time of 80 years. (C) was
constructed with postfire trajectories for the in-
dividual agents measured or predicted for the
Donnelly Flats fire [e.g., (A)] and the forest stand
age distribution model described in the Supporting
Online Material. For (C), by definition, each forcing
agent had a zero mean at steady state (at a mean
fire return time of 80 years).
17 NOVEMBER 2006 VOL 314 SCIENCE www.sciencemag.org
1132
REPORTS
![Fig. 3. (A) Annual radiative forcing from longlived greenhouse gases and the postfire trajectory of surface albedo. (B) Cumulative annual radiative forcing for the different forcing agents averaged over the time since the fire (or equivalently, the age of the stand). (C) Climate forcing of the different components as a function of the fire return time relative to a distribution of stands at steady state with a mean fire return time of 80 years. (C) was constructed with postfire trajectories for the individual agents measured or predicted for the Donnelly Flats fire [e.g., (A)] and the forest stand age distribution model described in the Supporting Online Material. For (C), by definition, each forcing agent had a zero mean at steady state (at a mean fire return time of 80 years).](/figures/fig-3-a-annual-radiative-forcing-from-longlived-greenhouse-b8qtjd35.webp)