Assessing net ecosystem carbon exchange of U.S. terrestrial ecosystems by integrating eddy covariance flux measurements and satellite observations

Abstract

More accurate projections of future carbon dioxide concentrations in the atmosphere and associated climate change depend on improved scientific understanding of the terrestrial carbon cycle. Despite the consensus that U.S. terrestrial ecosystems provide a carbon sink, the size, distribution, and interannual variability of this sink remain uncertain. Here we report a terrestrial carbon sink in the conterminous U.S. at 0.63 pg C yr−1 with the majority of the sink in regions dominated by evergreen and deciduous forests and savannas. This estimate is based on our continuous estimates of net ecosystem carbon exchange (NEE) with high spatial (1 km) and temporal (8-day) resolutions derived from NEE measurements from eddy covariance flux towers and wall-to-wall satellite observations from Moderate Resolution Imaging Spectroradiometer (MODIS). We find that the U.S. terrestrial ecosystems could offset a maximum of 40% of the fossil-fuel carbon emissions. Our results show that the U.S. terrestrial carbon sink varied between 0.51 and 0.70 pg C yr−1 over the period 2001–2006. The dominant sources of interannual variation of the carbon sink included extreme climate events and disturbances. Droughts in 2002 and 2006 reduced the U.S. carbon sink by ∼20% relative to a normal year. Disturbances including wildfires and hurricanes reduced carbon uptake or resulted in carbon release at regional scales. Our results provide an alternative, independent, and novel constraint to the U.S. terrestrial carbon sink.Research highlights▶ We report a terrestrial carbon sink in the conterminous U.S. at 0.63 pg C yr−1. ▶ U.S. carbon sink varied between 0.51 and 0.70 pg C yr−1 over the period 2001–2006. ▶ The severe droughts in 2002 and 2006 substantially reduced the U.S. carbon sink. ▶ Disturbances reduced carbon uptake or resulted in carbon release at regional scales. ▶ Our results provide an alternative and novel constraint to the U.S. carbon sink.

Full text

Page 1

Agricultural and Forest Meteorology 151 (2011) 60–69
Contents lists available at ScienceDirect
Agricultural and Forest Meteorology
journal homepage: www.elsevier.com/locate/agrformet
Assessing net ecosystem carbon exchange of U.S. terrestrial ecosystems by
integrating eddy covariance flux measurements and satellite observations
Jingfeng Xiaoa,∗, Qianlai Zhuangb, Beverly E. Lawc, Dennis D. Baldocchid, Jiquan Chene,
Andrew D. Richardsonf, Jerry M. Melillog, Kenneth J. Davish, David Y. Hollingeri, Sonia Whartonj,
Ram Orenk, Asko Noormetsl, Marc L. Fischerm, Shashi B. Verman, David R. Cooko, Ge Sunp,
Steve McNultyp, Steven C. Wofsyq, Paul V. Bolstadr, Sean P. Burnss, Peter S. Curtist,
Bert G. Drakeu, Matthias Falkj, David R. Fosterv, Lianhong Guw, Julian L. Hadleyx,
Gabriel G. Katulk, Marcy Litvaky, Siyan Mad, Timothy A. Martinz, Roser Matamalaaa,
Tilden P. Meyersbb, Russell K. Monsons, J. William Mungercc, Walter C. Oecheldd, U. Kyaw Tha Pawj,
Hans Peter Schmidee,ff, Russell L. Scottgg, Gregory Starrhh, Andrew E. Suykern, Margaret S. Tornii
aDepartment of Earth & Atmospheric Sciences, Purdue Climate Change Research Center, Purdue University, West Lafayette, IN 47907, USA
bDepartment of Earth & Atmospheric Sciences, Department of Agronomy, Purdue Climate Change Research Center, Purdue University, West Lafayette, IN 47907, USA
cCollege of Forestry, Oregon State University, Corvallis, OR 97331, USA
dEcosystem Science Division, Department of Environmental Science, Policy and Management, University of California, Berkeley, CA 94720, USA
eDepartment of Environmental Sciences, University of Toledo, Toledo, OH 43606, USA
fDepartment of Organismic and Evolutionary Biology, Harvard University, Cambridge, MA 02138, USA
gEcosystems Center, Marine Biological Laboratory, Woods Hole, MA, USA
hDepartment of Meteorology, Pennsylvania State University, University Park, PA 16802, USA
iUSDA Forest Service, Northeastern Research Station, Durham, NH 03824, USA
jDepartment of Land, Air and Water Resources, University of California, Davis, CA 95616, USA
kNicholas School of the Environment, Duke University, Durham, NC 27708, USA
lDepartment of Forestry and Environmental Resources and Southern Global Change Program, North Carolina State University, Raleigh, NC 27695, USA
mLawrence Berkeley National Laboratory, Environmental Energy Technologies Division, Atmospheric Science Department, Berkeley, CA 94720, USA
nSchool of Natural Resources, University of Nebraska-Lincoln, Lincoln, NE 68583, USA
oArgonne National Laboratory, Environmental Science Division, Argonne, IL 60439, USA
pUSDA Forest Service, Southern Research Station, Raleigh, NC 27606, USA
qDivision of Engineering and Applied Science/Department of Earth and Planetary Science, Harvard University, Cambridge, MA 02138, USA
rDepartment of Forest Resources, University of Minnesota, St. Paul, MN 55108, USA
sDepartment of Ecology and Evolutionary Biology, University of Colorado, Boulder, CO 80309, USA
tDepartment of Evolution, Ecology, and Organismal Biology, Ohio State University, Columbus, OH 43210, USA
uSmithsonian Environmental Research Center, Edgewater, MD 21037, USA
vHarvard Forest and Department of Organismic and Evolutionary Biology, Harvard University, Petersham, MA 01366, USA
wOak Ridge National Laboratory Environmental Sciences Division, Oak Ridge, TN 37831, USA
xHarvard Forest, Harvard University, Petersham, MA 01366, USA
yDepartment of Biology, University of New Mexico, Albuquerque, NM 87131, USA
zSchool of Forest Resources & Conservation, University of Florida, Gainesville, FL 32611, USA
aaArgonne National Laboratory, Biosciences Division, Argonne, IL 60439, USA
bbNOAA/ARL, Atmospheric Turbulence and Diffusion Division, Oak Ridge, TN 37831, USA
ccDepartment of Earth and Planetary Sciences, Harvard University, Cambridge, MA 02138, USA
ddDepartment of Biology, San Diego State University, San Diego, CA 92182, USA
eeDepartment of Geography, Indiana University, Bloomington, IN 47405, USA
ffAtmospheric Environmental Research, Institute of Meteorology and Climate Research, Research Center Karlsruhe (FZK/IMK-IFU), Kreuzeckbahnstr,
19, 82467 Garmisch-Partenkirchen, Germany
ggUSDA-ARS Southwest Watershed Research Center, Tucson, AZ 85719, USA
hhDepartment of Biological Sciences, University of Alabama, Tuscaloosa, AL 35487, USA
iiLawrence Berkeley National Laboratory, Earth Science Division, Berkeley, CA 94720, USA
∗Corresponding author at: Complex Systems Research Center, University of New Hampshire, Durham, NH 03824, USA. Tel.: +1 603 862 1873.
E-mail address: j.xiao@unh.edu (J. Xiao).
0168-1923/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.agrformet.2010.09.002

Page 2

J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
61
a r t i c l ei n f o
Article history:
Received 19 July 2009
Received in revised form 20 August 2010
Accepted 6 September 2010
Keywords:
Net ecosystem carbon exchange
Eddy covariance
MODIS
Carbon sink
U.S.
Interannual variability
Drought
Disturbance
a b s t r a c t
More accurate projections of future carbon dioxide concentrations in the atmosphere and associated
climate change depend on improved scientific understanding of the terrestrial carbon cycle. Despite the
consensus that U.S. terrestrial ecosystems provide a carbon sink, the size, distribution, and interannual
variability of this sink remain uncertain. Here we report a terrestrial carbon sink in the conterminous U.S.
at 0.63pg C yr−1with the majority of the sink in regions dominated by evergreen and deciduous forests
and savannas. This estimate is based on our continuous estimates of net ecosystem carbon exchange
(NEE) with high spatial (1km) and temporal (8-day) resolutions derived from NEE measurements from
eddy covariance flux towers and wall-to-wall satellite observations from Moderate Resolution Imaging
Spectroradiometer (MODIS). We find that the U.S. terrestrial ecosystems could offset a maximum of 40%
of the fossil-fuel carbon emissions. Our results show that the U.S. terrestrial carbon sink varied between
0.51 and 0.70 pg C yr−1over the period 2001–2006. The dominant sources of interannual variation of
the carbon sink included extreme climate events and disturbances. Droughts in 2002 and 2006 reduced
the U.S. carbon sink by ∼20% relative to a normal year. Disturbances including wildfires and hurricanes
reduced carbon uptake or resulted in carbon release at regional scales. Our results provide an alternative,
independent, and novel constraint to the U.S. terrestrial carbon sink.
© 2010 Elsevier B.V. All rights reserved.
1. Introduction
More accurate quantification of net carbon dioxide (CO2)
exchange over regions, continents, or the globe can improve our
understanding of the feedbacks between the terrestrial biosphere
and the atmosphere in the context of global change and facilitate
climate policy-making (IPCC, 2007; Peters et al., 2007). Despite the
consensusthatU.S.terrestrialecosystemsprovideacarbonsink,the
size and distribution of the sink still remain uncertain (Houghton
et al., 1999; Caspersen et al., 2000; Schimel et al., 2000; Pacala et
al., 2001; SOCCR, 2007). More importantly, the interannual vari-
ability of this carbon sink is not well understood. Extreme climate
events (Ciais et al., 2005; Zeng et al., 2005; Xiao et al., 2009, 2010)
and disturbances (Law et al., 2004; Chambers et al., 2007; Amiro
et al., in press; Xiao et al., 2010) could substantially affect ecosys-
tem carbon fluxes and lead to significant year-to-year variations in
regional terrestrial carbon budgets. Here we integrate eddy covari-
ance flux measurements and wall-to-wall satellite observations to
assess recent U.S. net ecosystem carbon exchange (NEE) and year-
to-year variations.
Inventory studies of biomass (Clark et al., 2001; Goodale et al.,
2002) and soil carbon (Lal et al., 2001) are traditionally used to
quantifyNEEofanecosystemovermultipleyears(Baldocchi,2003).
The eddy covariance technique has emerged as an alternative way
to assess NEE (Baldocchi et al., 2001). Eddy covariance flux towers
havebeenprovidingcontinuousmeasurementsofecosystem-level
exchange of CO2spanning diurnal, synoptic, seasonal, and interan-
nualtimescalessincetheearly1990s(Wofsyetal.,1993;Baldocchi
et al., 2001). The AmeriFlux network consists of eddy covariance
fluxtowersencompassingalargerangeofclimateandbiometypes,
and provides the longest, most extensive, and most reliable mea-
surements of plot-scale NEE with high temporal resolution for the
U.S. These NEE estimates represent fluxes at the scale of the tower
footprint with longitudinal dimensions ranging between a hun-
dred meters and several kilometers depending on homogeneous
vegetationandfetch(Schmid,1994;Göckedeetal.,2008).Toexam-
ine terrestrial carbon cycling over regions or continents, therefore,
eddy flux measurements need to be upscaled to these large areas
(Xiao et al., 2008, 2010).
Satellite remote sensing provides observations of ecosystems
with spatially and temporally consistent coverage, and is a valu-
able tool for upscaling carbon fluxes to regional or continental
scales (Xiao et al., 2008, 2010). The Moderate Resolution Imaging
Spectroradiometer (MODIS) on board the NASA’ Terra and Aqua
satellites provides particularly useful observations as the MODIS
sensors view the entire Earth’s surface every one to two days and
acquiredatawith36spectralbandsand250m–1kmspatialresolu-
tion. Several recent studies have upscaled eddy flux measurements
to large areas using satellite data (e.g., Papale and Valentini, 2003;
Yamaji et al., 2007; Wylie et al., 2007; Xiao et al., 2008, 2010).
Moreover, some of the resulting flux estimates have been used to
assess regional terrestrial carbon uptake. For example, Papale and
Valentini (2003) estimated annual NEE for European forests using
the NEE estimates. Yamaji et al. (2007) used the NEE estimates
to assess the annual NEE of deciduous forests in Japan. Wylie et
al. (2007) examined the magnitude and interannual variability of
annual NEE for grasslands in the northern Great Plains. Xiao et al.
(2010) examined the magnitude, patterns, and interannual vari-
ability of gross primary productivity (GPP) for the conterminous
U.S.
Here we use our predictive NEE model developed from eddy
flux and MODIS data (Xiao et al., 2008) to produce continuous NEE
estimates with high spatial (1km) and temporal (8-day) resolu-
tions for the conterminous U.S. over the period 2000–2006. Our
continuous NEE estimates along with our previous GPP estimates
(Xiao et al., 2010) for the U.S. were both derived from eddy covari-
ance (EC) flux measurements and MODIS data, and are referred to
as EC-MOD. We then use our continuous NEE estimates to assess
the magnitude, distribution, and interannual variability of recent
U.S. ecosystem carbon exchange. One of the main innovations in
our estimates is the use of daily NEE measurements derived from
continuous observations from eddy covariance flux towers. These
measurements represent direct samples of net CO2exchange from
sitesencompassingawidevarietyofU.S.biomesandclimatetypes.
These data were not utilized in previous U.S. carbon budget stud-
ies (e.g., Houghton et al., 1999; Caspersen et al., 2000; Schimel et
al., 2000; Pacala et al., 2001; SOCCR, 2007). Our analysis provides
an alternative, independent, and novel perspective on recent U.S.
ecosystem carbon exchange.
2. Data and methods
2.1. AmeriFlux data
NEE is the difference of two large carbon fluxes – photosynthe-
sis (GPP) and ecosystem respiration (Re). To avoid compensating
errors the basic processes underlying ecosystem carbon uptake
and release should both be modeled well (Richardson et al., 2007).
Unlike most modeling methods, our approach directly estimates
NEE, which could avoid the compensating errors. We integrated

Page 3

62
J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
Fig. 1. Distribution of the 42 AmeriFlux sites used in this study. The base map is the Level I Ecoregions map of North America from Environmental Protection Agency (EPA,
2009). The broad, level I ecoregions for the conterminous United States are Northern Forests (5), Northwestern Forested Mountains (6), Marine West Coast Forest (7), Eastern
Temperate Forests (8), Great Plains (9), North American Deserts (10), Mediterranean California (11), Southern Semi-arid Highlands (12), Temperate Dry Forests (13), and
Tropical Wet Forests (15).
eddy flux measurements and MODIS data to estimate NEE of U.S.
ecosystems. We obtained the Level 4 NEE data for 42 AmeriFlux
sites over the period 2000–2006. These sites are distributed across
the conterminous U.S. (Fig. 1). The AmeriFlux NEE data we used
were described in detail by Xiao et al. (2008).
TheLevel4productconsistsofNEEdatawithfourdifferenttime
steps, including half-hourly, daily, 8-day, and monthly. NEE was
calculated using the storage obtained from the discrete approach
or using a vertical CO2profile system, and was gap-filled using
the artificial neural network method. We used 8-day NEE data
(gCm−2day−1) to match the compositing intervals of MODIS data.
2.2. Predictive model
The development of our predictive model has been fully
described by Xiao et al. (2008). Here we briefly summarize our
approach. We combined site-specific eddy flux measurements
and MODIS data to develop a predictive NEE model using piece-
wise regression models. The independent variables of the model
included a variety of MODIS data products: surface reflectance
(Vermote and Vermeulen, 1999), daytime and nighttime land sur-
facetemperature(LST)(Wanetal.,2002),andenhancedvegetation
index (EVI) (Huete et al., 2002) as well as the normalized difference
water index (NDWI; Gao, 1996) derived from surface reflectance.
These data can partly account for a variety of physical, physiologi-
cal, atmospheric, hydrologic, and edaphic variables that affect NEE
(Xiao et al., 2008). For each site, we obtained MODIS ASCII sub-
sets consisting of 7km×7km regions centered on the flux tower,
including surface reflectance, daytime and nighttime LST, and EVI
over the period 2000–2006 from the Oak Ridge National Labo-
ratory’s Distributed Active Archive Center. For each variable, we
extractedtheaveragevaluesforthecentral3km×3kmareawithin
the 7km×7km cutouts rather than the values of the 1km×1km
pixel corresponding to the tower to account for the inherent geolo-
cation error of MODIS data and better represent the flux tower
footprint.
We then developed the predictive NEE model using piecewise
regression models. We divided the site-level data set of Ameri-
Flux and MODIS data into a training set (2000–2004) and a test set
(2005–2006). The training and test sets were used to develop and
evaluate the predictive model, respectively. The performance of
the model was evaluated using three statistical measures: average
error,relativeerror,andcorrelationcoefficient.Themodelachieved
reasonable accuracy to estimate NEE at the site level (Xiao et al.,
2008).
In addition to the validation in the temporal domain using
the test data (Xiao et al., 2008), we also used cross-validation to
obtain an alternative measure of the predictive accuracy of our
model. We used k-fold cross-validation, in which the cases are
divided into k blocks of roughly the same size and target value
distribution. For each block, a predictive model is constructed
from the cases in the remaining blocks, while the cases in the
hold-out block is then used to test the performance of the model
(RuleQuest, 2008). The 10-fold cross-validation also showed that
our model predicted NEE fairly well (Fig. 2; R2=0.67, p<0.0001;
RMSE=1.45gCm−2day−1).
Fig.2. Observed8-dayNEEversuspredicted8-dayNEEfor10-foldcross-validation.

Page 4

J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
63
2.3. Network representativeness
TheAmeriFluxsitesweusedinvolveawidevarietyofecosystem
types (Fig. 1) and are fairly representative of typical U.S. climate
types (Xiao et al., 2008). Moreover, some forest sites are at dif-
ferent stages since stand replacing disturbance, which are located
in disturbance clusters of sites, such as Metolius intermediate
aged ponderosa pine (Oregon), Metolius new young pine (Oregon),
Intermediate hardwood (Wisconsin), and Mature red pine (Wis-
consin). In addition, some sites have received treatment, including
the Howland Forest West Tower (Maine: nitrogen fertilizer) and
the Mead cropland sites (Nebraska; irrigation versus rainfed; con-
tinuous maize versus maize/soybean rotation). Hargrove et al.
(2003) carried out an analysis of multivariate environmental “data
space” (including such factors as elevation, means and extremes of
temperature, monthly precipitation, soil characteristics and other
factors)andconcludedthatthecentral,Midwestern,andnortheast-
ernpartoftheU.S.werewellrepresentedbytheAmeriFluxnetwork
at that time, with the southern, southwestern, and Pacific North-
west environments less well represented by existing tower sites.
SincethatanalysisnewsiteshavebeenestablishedinFlorida,Texas,
New Mexico, Arizona, and Southern California, which substantially
increased the coverage and representativeness of the south and
southwestern parts of the U.S.
We assessed the representativeness of the AmeriFlux network
using the Euclidean distance between each grid cell and the eddy
flux sites in the attribute space, following Yang et al. (2008). The
attribute space we used consists of all the numerical variables
included in the predictive model: bands 1–6, NDWI, daytime LST,
nighttime LST, and EVI. The Euclidean distance measures the simi-
larity between each grid cell and eddy flux sites (Yang et al., 2008).
The Euclidian distance (ED) between two points X=(x1, x2, ..., xn)
and Y=(y1, y2, ..., yn) is defined as:
??
where X is the attribute vector of a grid cell, while Y is the mean
attribute vector of eddy flux sites within a given vegetation type.
Daytime and nighttime LST were scaled to the range of −1 and
+1 using their minimum and maximum values over the period
2000–2006 to minimize the influence of the magnitude of these
variables on the calculation of ED. We calculated ED for each
grid cell for each 8-day interval. For each year, the mean ED was
calculated from all 8-day ED values for each grid cell. We then
calculated the minimum ED for each grid cell over the period
2000–2006.
Theresultingsimilaritymap(Fig.3)showedthattheGreatLakes
region, the Great Plains, a part of the Pacific West, and northeast-
ern forests exhibited highest similarity to the AmeriFlux network;
the Southeast and the Gulf Coast region exhibited intermediate
similarity, while a large part of the Rocky Mountain region, north-
easternNorthDakotaandwesternMinnesotahadlowestsimilarity.
This indicates that the Great Lakes region, the Great Plains, a part
of Pacific West, and northeastern forests are fairly represented by
AmeriFlux sites, while a large part of the Rocky Mountain region,
northeastern North Dakota and western Minnesota were under-
represented.
ED =
n
(xi− yi)2
(1)
2.4. MODIS data
To estimate NEE at the continental scale, we obtained wall-
to-wall MODIS data including surface reflectance, daytime and
nighttimeLST,andEVIfromtheEarthObservingSystem(EOS)Data
Gateway for each 8-day interval from February 2000 to Decem-
ber 2006. For each variable, the quality of the value of each pixel
Fig. 3. Representativeness of the AmeriFlux network measured by the similarity
between each grid cell and eddy flux sites using the Euclidean distance in the
attribute space. The similarity is proportional to the magnitude of the distance.
(1km) was determined using the quality assurance (QA) flags, and
the bad-quality value was replaced using a linear interpolation
approach (Xiao et al., 2008). The NDWI was also calculated from
band 2 (near-infrared) and band 6 (shortwave infrared) of the sur-
face reflectance product. Each 16-day EVI composite was used for
two 8-day intervals corresponding to the composting interval of
other MODIS products. We also obtained the land cover type for
each cell from the 1km MODIS land cover map (Friedl et al., 2002),
and reclassified the vegetation classes to seven broader classes:
evergreen forests, deciduous forests, mixed forests, shrublands,
savannas, grasslands, and croplands (Xiao et al., 2008).
2.5. Model prediction
Wethenusedourpredictivemodel(Xiaoetal.,2008)toestimate
NEE for each 1km×1km cell across the conterminous U.S. for each
8-day interval from February 2000 to December 2006 using the
wall-to-wall MODIS data. The land-cover type for each pixel was
obtainedfromthereclassifiedMODISland-covermap.Foreachgrid
cell, we calculated annual NEE (gCm−2yr−1) for each year from our
8-day NEE estimates.
We produced the probability distribution of our mean annual
NEE values over the period 2001–2006 across the conterminous
U.S., and then compared this histogram with that of the com-
pileddatabaseofpublishededdyfluxmeasurementsovertheglobe
(Baldocchi, 2008) (Fig. 4). This database contains 504 site-years of
data from 125 study sites over the globe. Both probability distribu-
tionswerenegativelyskewed,indicatingthatoverall,theterrestrial
ecosystemsabsorbedcarbonfromtheatmosphere.Moreover,these
two probabilistic histograms had similar ranges and almost iden-
tical means and standard deviations. The similarity of these two
distributions indicated that our annual estimates generally cap-
turedtheexpectedmean,range,andspatialvariabilityofpublished
annual NEE.
We then examined the magnitude, distribution, and year-to-
year variations of annual NEE. The coefficient of variation (CV) and
annual anomalies were used to assess the interannual variability of
NEE. CV is defined as the ratio of standard deviation to the mean.
Foreachgridcell,theannualNEEvaluesovertheperiod2001–2006
were used to calculate the mean and standard deviation of annual
NEE, and CV was then calculated as the standard deviation divided
by the mean. The CV value of each grid cell indicates the variabil-
ity of annual NEE over the 6-year period. For grid cells with mean
annual NEE within −1 to +1gCm−2yr−1, CV was assigned as 0 to
avoidunreasonablylargeCVvaluesinabsolutemagnitude.Foreach

Page 5

64
J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
Fig. 4. Probabilistic histograms of estimated and published annual NEE. Dark gray
bars indicate estimated annual NEE values, and light gray bars indicate published
annual NEE from 506 site-years of data over the globe. Superimposed (black curve)
is a Gaussian probability distribution. The means of estimated and published annual
NEE are −189.5 and −182.4gCm−2yr−1, respectively; their standard deviations are
272.3 and 269.6, respectively.
year, we also calculated the anomaly of annual NEE relative to the
mean over the 6-year period.
2.6. Climate and disturbance data
We used precipitation, fire severity, and wind fields data to
assess the impacts of drought, fire, and hurricane on NEE. We
obtained monthly precipitation data over the period 1970–2006
from the PRISM (Parameter-elevation Regressions on Independent
SlopesModel)group(PRISM,2004).PRISMproducesahigh-quality
gridded climate data set using station observations of precipita-
tion, temperature, and other climatic factors. The spatial resolution
of this data set is 4km×4km. We calculated the average annual
precipitation over the period 1970–1999, and then calculated the
anomaly of annual precipitation for each year over the period
2001–2006 relative to the long-term mean (1970–1999).
We also used fire severity data for the Biscuit fire in Oregon.
The Biscuit Fire was among the largest forest fires in modern
United States history, encompassing >2000km2primarily within
the Rogue-Siskiyou National Forest (RSNF) in southwest Oregon
(Thompson et al., 2007). Fire severity was based on the difference
normalizedburnratio(dNBR)(Lutesetal.,2004)fromLandsatThe-
matic Mapper (TM) data acquired before and immediately after the
fire.dNBRisaquantitativeindicatorofabovegroundgreenbiomass
and changes in soil moisture and color as well as consumption
of down fuels (Lutes et al., 2004). dNBR has proven an effective
measure of burn severity of forest ecosystems (Brewer et al., 2005;
Miller and Yool, 2002). Different degrees of fire severity were iden-
tified: little or no change, green and dead mixed, dead trees with
needles, and dead trees without needles (Thompson et al., 2007).
Finally, we obtained tropical cyclone surface wind field data for
hurricane Katrina (Powell et al., 1998) from the Atlantic Oceano-
graphic and Meteorological Laboratory (AOML), National Oceanic
and Atmospheric Administration (NOAA) (AOML, 2007). The wind
field data was produced using the Hurricane Research Division
(HRD) surface wind analysis system (Powell et al., 1998). The
HRD approach to hurricane wind analysis requires the input of all
available surface weather observations (e.g., ships, buoys, coastal
platforms, surface aviation reports, etc.). We converted the wind
Fig.5. MeanannualNEEfortheconterminousU.S.overtheperiod2001–2006.Units
aregCm−2yr−1.Positivevaluesindicatecarbonrelease,andnegativevaluesindicate
carbon uptake. Gray lines indicate state boundaries. Black lines indicate boundaries
of geographical regions: Northeast (NE), Southeast (SE), North Central (NC), South
Central (SC), Pacific Northwest (PNW), and Pacific Southwest (PSW).
fields for hurricane Katrina to isotaches, and extracted the isotachs
representing tropical storm (39–73mph), category 1 (74–95mph),
and category 2 (96–110mph) wind fields.
3. Results and discussion
3.1. Annual NEE
The 6-yr mean annual pattern of NEE is shown in Fig. 5. This
representstheterrestrialpartofthecarboncyclewithoutfireemis-
sions and immediate carbon loss due to the removal and burning
of biomass. The temperate forests in the eastern U.S. absorbed car-
bon mainly because of forest regrowth following the abandonment
of agricultural lands (Caspersen et al., 2000). Some regions in the
west including the Rocky Mountains and the Pacific Coast also
assimilated carbon. The dominant evergreen forests in the Pacific
Coast region are highly productive as these ecosystems assimi-
lated carbon even in the winter due to mild temperature and moist
conditions (Waring and Franklin, 1979; Anthoni et al., 2002). For
example, Douglas-fir, a major species in the Pacific Northwest and
California, is known to be highly plastic and able to photosynthe-
size in winter when temperatures are above freezing. By contrast,
many areas in the western U.S. including the Great Basin, the Col-
orado Plateau, and the western Great Plains were nearly carbon
neutral due to sparse vegetation and large precipitation deficits.
Drought could reduce GPP and net primary productivity (NPP), and
lead to changes in heterotrophic respiration (Rh) occurring in the
same direction but with a smaller magnitude (Xiao et al., 2009).
These concurrent changes in NPP and Rhlikely led to reduced
annual NEE in absolute magnitude. A part of the Southwest pro-
videdminorcarbonsources,whilethesmallregioninnortheastern
Arizona and southeastern Utah exhibited annual NEE greater than
200gCm−2yr−1. This region was affected by moderate to severe
drought in 2000–2003 and 2005, which likely resulted in a net
carbon release over the 6-year period. In addition, our predictive
model utilized LST to account for Re, and the explanatory variables
could not account for the sizes of soil carbon pools and litter. The
relatively high temperatures in this region in the summer and the
inability of our predictive model to account for soil carbon pools
and litter may lead to overestimation of Re, which in turn can
lead to reduced net carbon uptake (Re<GPP) or net carbon release
(Re>GPP).

Page 6

J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
65
Fig. 6. Mean annual NEE for each vegetation type within the conterminous U.S.
over the period 2001–2006: evergreen forests (EF), deciduous forests (DF), mixed
forests (MF), shrublands (Sh), savannas (Sa), and grasslands (Gr). Units are pgCyr−1.
The bars are the estimated mean annual NEE. The error bars indicate the standard
deviation from the mean.
We examined the total annual NEE for U.S. and for each broad
vegetation type over the period 2001–2006 (Fig. 6). On average,
the total annual NEE for U.S. is −1.21pgCyr−1, with −0.63 for
natural ecosystems and −0.58pgCyr−1for croplands. We then
examinedthemagnitudeoftheU.S.terrestrialcarbonsinkusingour
annual NEE estimates. Croplands were not included here because
the carbon absorbed by crops will be released back into the atmo-
sphere due to the consumption of yields and the burning of
biomass. We estimated total carbon sink in the conterminous U.S.
at 0.63pgCyr−1with the majority of the sink in regions dominated
by evergreen and deciduous forests and savannas in nearly equal
proportions of the total sink (Fig. 6). Our estimate did not include
the immediate emissions from the burning of biomass in wildfires.
We assumed that wildfire released 0.06pgCyr−1as estimated by
Wiedinmyer and Neff (2007), and then deducted fire emissions
from our sink estimate. With fire emissions taken into considera-
tion, our estimate of the U.S. carbon sink (0.57pgCyr−1) was at the
higher end of the range (0.30–0.58gCyr−1) estimated by Pacala et
al. (2001) and between a recent estimate (∼0.63pgCyr−1) based
on an inverse modeling approach (Deng et al., 2007) and the esti-
mate(0.49pgCyr−1)bythefirstNorthAmericanStateoftheCarbon
Cycle Report (SOCCR, 2007). Compared to previous approaches
basedonforestinventorymeasurements,ecosystemmodels,book-
keeping models, or atmospheric inverse modeling (e.g., Houghton
et al., 1999; Caspersen et al., 2000; Schimel et al., 2000; Pacala et
al., 2001; Deng et al., 2007; SOCCR, 2007), our approach integrated
eddy flux measurements and wall-to-wall satellite observations,
and provided an independent, alternative, and novel constraint to
the recent U.S. terrestrial carbon sink. A quantitative breakdown
of the 6-yr mean annual NEE map by geographical regions (Fig. 7)
showed that the greatest carbon uptake occurred in the South Cen-
tral region, followed by Southeast and Northeast. North Central
and the Pacific Northwest had intermediate carbon sinks, while
the Pacific Southwest had the lowest carbon uptake. The Rocky
Mountain region was a minor carbon source.
A better understanding of the magnitude and spatial distri-
bution of the U.S. carbon sink is helpful for successful carbon
management strategies to mitigate fossil-fuel emissions or stabi-
lize concentrations of greenhouse gases in the atmosphere (SOCCR,
2007). Our results show that the U.S. terrestrial ecosystems could
offset 40% of the U.S. fossil fuel emissions (Fig. 7), suggesting that
the U.S. ecosystems play an important role in slowing down the
buildup of CO2in the atmosphere. This is higher than the SOCCR
estimateof30%,whichwasbasedoninventorydata(SOCCR,2007).
TheNorthCentral(0.42pgCyr−1)andSouthCentral(0.40pgCyr−1)
regions had the largest carbon emissions due to consumption of
fossil fuels (EIA, 2008), whereas ∼19% and 55% were offset by
Fig. 7. Mean annual NEE per region for the period 2001–2006. White bars are
the fossil fuel fluxes (Energy Information Administration, Department of Energy,
http://www.eia.doe.gov),andgraybarsareannualNEE.UnitsarepgCyr−1.Theerror
bars denote the standard deviation from the mean. The labels refer to the regions:
Northeast(NE),Southeast(SE),North-Central(NC),South-Central(SC),RockyMoun-
tain (RM), Pacific Northwest (PNW), and Pacific Southwest (PSW). Numbers are the
percentages that the fossil fuel fluxes were offset by ecosystem carbon uptake. No
percentage was provided for the Rocky Mountain region because this region was a
minor carbon source.
regional ecosystem carbon uptake, respectively. The Northeast and
Southeast regions had intermediate fossil-fuel carbon emissions,
and ecosystem carbon uptake in these regions offset 40% and 64%
of the emissions, respectively. The Pacific Southwest also offset
40% of the regional carbon emissions. The Rocky Mountain region,
however, provided a minor carbon source that was ∼0.1pgCyr−1
greater than fossil fuel emissions in the region. The Pacific North-
west region absorbed about twice as much carbon as emitted by
the consumption of fossil fuels in the region, which likely puts
the region at an advantage in carbon policy-making and trading.
The quantification of the proportions of carbon emissions offset
by ecosystem carbon uptake and their magnitudes have implica-
tions for carbon-management and climate policy-decision making,
including the commitments of different states to the reduction of
carbon emissions. It should be noted that our estimates represents
the upper bounds of the capacity of U.S. ecosystems for sequester-
ing carbon from the atmosphere because our NEE did not include
fire emissions, immediate loss of carbon due to other disturbances
and management, and other greenhouse gas emissions from crop-
lands and pastures.
3.2. Year-to-year variations
Fig. 8 shows the interannual variability of NEE characterized by
CV. Negative values indicate the interannual variability of carbon
uptake, and positive values indicate the interannual variability of
carbon release. CV generally exhibited larger spatial variability in
regions with carbon release than in regions with carbon uptake.
This is likely because ecosystems providing carbon sources are
mainly distributed in arid and semi-arid regions, while ecosystems
with carbon uptake are primarily distributed in humid and semi-
humid regions. The magnitude of annual NEE in arid and semi-arid
region is often smaller than that in humid and semi-humid regions.
Moreover, arid and semi-arid regions are characterized by larger
temporal variability in water availability than humid and semi-
humid regions due to the variation in climate and weather at
multiple scales and vegetation-soil water feedbacks (Snyder and
Tartowski,2006).Amongregionswithcarbonrelease,thesouthern
part of the western Great Plains, including western Texas, eastern
New Mexico, and southeastern Colorado exhibited largest CV val-

Page 7

66
J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
Fig. 8. Interannual variability of annual NEE across the conterminous U.S. over the
period 2001–2006 characterized by the coefficient of variation (CV).
ues. These areas had relatively low mean annual NEE in absolute
magnitude and were frequently affected by moderate to severe
drought (2000–2003 and 2005), leading to relatively large year-
to-year variations. Among regions with carbon uptake, a swath of
land in Texas and eastern Great Plains exhibited large CV values in
absolute magnitude. Most highly productive regions including the
Southeast and the Pacific Northwest had relatively low CV, show-
ingthatannualNEEoftheseecosystemsexhibitedlowyear-to-year
variations. These regions are characterized by favourable climate
conditions, and ecosystems in these regions may not be subject
to water stress despite minor to moderate negative precipitation
anomalies.
ThemagnitudeoftheU.S.terrestrialcarbonsinkvariedbetween
0.51 and 0.70pgCyr−1from 2001 to 2006 due to climate variabil-
ity, disturbances, and management practices (Fig. 9). In our 6-yr
estimate, 2002 and 2006 were the lowest net uptake years (0.56
and 0.51pgCyr−1, respectively). The U.S. carbon sink in these two
yearsdecreasedby16.3%and27.1%relativetopreviousyears(2001
and 2005, respectively) due to the effects of drought and wildfires.
Moderate to extreme drought affected more than 50% of the con-
terminous U.S. during the summer in 2002 and 2006 (NCDC, 2008),
modifyingthebalanceofecosystemphotosynthesisandrespiration
in these years. Annual NEE anomaly maps showed large positive
anomalies (smaller sinks or sources) in many regions affected by
drought (Fig. 9). The impact of the 2002 drought on ecosystem car-
bon uptake is generally consistent with the drought effects on the
terrestrial carbon cycle observed for Europe and North America
(Ciais et al., 2005; Peters et al., 2007). Our results also show that
Fig. 9. Effects of extreme climate events and disturbances on annual NEE. (a) Impact of the Biscuit Fire on annual NEE: burned area, fire severity, and anomalies of annual
NEE in 2003. Fire severity was based on the difference normalized burn ratio (dNBR) from Landsat Thematic Mapper (TM) data acquired before and immediately after the
fire: little or no change (I), green and dead mixed (II), dead trees with needles (III), and dead trees without needles (IV). (b) Anomalies of annual NEE relative to the 6-year
period 2001–2006 and anomalies of annual precipitation relative to the 30-year period 1970–1999 taken from the PRISM climate database in 2002 and 2006. (c) Impact of
hurricane Katrina on annual NEE in 2006. The white lines indicate the isotachs, including tropical storm, hurricane category 1, and hurricane category 2.

Page 8

J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
67
the 2006 drought had a larger impact on the U.S. ecosystem carbon
exchange than the 2002 drought.
Atthestandorregionallevel,NEEissignificantlyaffectedbydis-
turbances (Law et al., 2004). Disturbances can substantially alter
ecosystem carbon fluxes and regional carbon budgets (McCarthy
et al., 2006; Chambers et al., 2007), by reducing the aboveground
biomass and increase litter, thereby leading to a decrease in GPP
and an increase in Rh. Our results show that disturbances includ-
ing wildfires and hurricanes could affect regional annual NEE. In
addition, numerous wildfires occurred over the western U.S. due
to drought conditions related to large-scale atmospheric circula-
tion patterns, burning 1.5–4.0×104km2of forests from 2000 to
2006(U.S.FireAdministration,2008).Ourresultsindicatethatfires
had significant impacts on regional carbon budgets. For example,
the 2002 Biscuit Fire in Oregon led to large positive NEE anoma-
lies for the forest ecosystems within the fire polygon (Fig. 9). The
burn severity of the fire varied over space (Thompson et al., 2007),
leading to the significant spatial variability of NEE anomalies. Our
results also indicate that other disturbances such as hurricanes
affected the interannual variability of NEE at the regional scale.
For example, hurricane Katrina that occurred in late August 2005
affectedover2×104km2offorestacrossMississippi,Louisianaand
Alabama, with damage ranging from broken branches to downed
trees. Our results show large positive NEE anomalies in areas clas-
sified as hurricane category I and II (Fig. 9).
Modeling interannual variation in NEE has proven challenging.
The interannual variation in NEE represents an extreme test for
models because it is much smaller than seasonal or spatial vari-
ation in photosynthesis and respiration (Richardson et al., 2007).
Our approach made use of 8-day MODIS products that provided
real-time observations of vegetation before and after disturbances,
andthereforecouldpartlyaccountfortheeffectsofdisturbanceson
NEE. The current efforts of developing continental-scale dynamic
land cover and disturbances products by USGS EROS Data Center
and the University of Maryland (Goward et al., 2008) and their
incorporation into the model could improve the estimation of NEE.
This could affect the ranking of NEE among regions because of
different disturbance regimes among regions. For example, forest
harvest in the southeastern US is on a very short rotation and dis-
turbance from fire is more prevalent in the interior west. To better
characterize disturbance with this approach, an even distribution
of disturbance clusters among ecoregions would be beneficial in
combination with disturbance mapping at an appropriate spatial
and temporal scale for capturing disturbances (Cohen et al., 2002).
Itischallengingtoestimatelitterfallandsoilcarbonpoolsasaresult
of disturbance. The availability of disturbance data and the training
of the model with eddy flux measurements from flux towers rep-
resenting different stages following disturbances may improve the
accuracy for estimating NEE of forest ecosystems.
3.3. Sources of uncertainty
Despite the encouraging performance of our predictive model
in estimating NEE and capturing the spatiotemporal patterns and
magnitude of U.S. ecosystem carbon exchange, we recognize that
our NEE estimates contain significant uncertainties. There are sev-
eral sources of uncertainty associated with our flux estimates:
uncertainties in eddy flux measurements, uncertainties in input
data(e.g.,landcover),modelstructuraluncertainty,anduncertain-
ties arising from the representativeness of the AmeriFlux network.
The carbon flux measurements derived from eddy flux towers con-
tain significant uncertainties (Richardson et al., 2008), while the
gap-filling techniques used to fill the data gaps introduced addi-
tionaluncertaintiesof∼±25gCm−2yr−1(Moffatetal.,2007).Land
cover is likely the predominant source of uncertainty because dif-
ferent land-cover types are associated with different parameters
in our predictive model and the classification uncertainty directly
introduces errors into flux estimates. The use of higher-resolution
land cover data (e.g., NLCD) may help assess the accuracy of the
MODIS land cover map and the effect on NEE estimates.
The model structure could also introduce uncertainty into our
NEE estimates. The explanatory variables included in our predic-
tive model could not account for all factors that affect NEE. For
example, these variables could not account for nitrogen availabil-
ity that limits plant growth. Moreover, the explanatory variables
could not sufficiently account for the factors affecting Rh: substrate
availability, soil temperature, and soil moisture (Xiao et al., 2008).
The LST derived from MODIS is a measure of the soil temperature
at the surface, and is a good indicator of Re(Rahman et al., 2005)
as both autotrophic and heterotrophic respiration are significantly
affectedbyair/surfacetemperature(LloydandTaylor,1994).NDWI
can partly account for soil moisture (Fensholt and Sandholt, 2003).
However, surface reflectance and EVI can only partly account for
aboveground biomass and non-photosynthetic material (e.g., lit-
ter), and are not sensitive to soil carbon pools. Root and associated
mycorrhizal respiration produce roughly half of soil respiration,
withmuchofthereminderderivedfromdecompositionofrecently
produced root and leaf litter (Ryan and Law, 2005). The inability of
our model to account for transient carbon pools could introduce
uncertainties to our NEE estimates (Richardson et al., 2007).
The representativeness of the AmeriFlux network will also pre-
sumably affect our spatial estimates of carbon fluxes. We speculate
that our flux estimates exhibited larger uncertainties in regions
under-represented by eddy flux towers. Although the AmeriFlux
networkisgenerallyrepresentativeoftheconterminousU.S.ecore-
gions (Figs. 1 and 2; Hargrove et al., 2003) and the 42 sites used in
this study included most of the active flux sites in the network,
some geographical regions, ecoregions, and biome types are still
underrepresented (Fig. 3; Xiao et al., 2008), which could affect the
accuracy of our NEE estimates. In particular, no flux towers have
been established for savannas (tree cover 10–30%), and therefore
we merged savannas (tree cover 10–30%) and woody savannas
(tree cover 30–60%) together in the development of the predictive
model (Xiao et al., 2008). Savannas and woody savannas occupy
∼4.2×105km2of land, and the treatment of savannas as woody
savannas could lead to biases to our NEE estimates for savannas.
Similarly,nofluxtowershavebeenestablishedforopenshrublands
(shrubcanopycover10–60%),andwethusmergedopenshrublands
with closed shrublands (shrub canopy cover >60%). The treatment
of open shrublands as closed shrublands could also lead to biases
in our NEE estimates. The current AmeriFlux network should be
augmented by establishing more sites for the western U.S., more
disturbance clusters of sites representing early stages following
disturbances, and more sites for certain biomes in the UMD clas-
sification system (e.g., open shrublands and savannas) (Xiao et al.,
2008).
A comprehensive, quantitative analysis of the uncertainties
associated with our flux estimates is beyond the scope of this
study. Future upscaling studies, however, should gauge the uncer-
tainty in flux estimates by considering uncertainties of eddy
flux measurements, using different upscaling methods, evaluating
uncertaintiesanddifferencesininputdata(e.g.,classificationaccu-
racy of land-cover maps), propagating the probability distributions
ofparametersthroughthemodels,andcomparingchangesinfluxes
caused by systematically removing individual flux tower data from
the development of predictive models.
4. Conclusions
We produced continuous estimates of NEE with high spatial
(1km)andtemporal(8-day)resolutionsovertheperiod2000–2006
for the conterminous U.S. by integrating eddy flux measurements

Page 9

68
J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
and wall-to-wall MODIS data. Our continuous NEE estimates along
with our previous GPP estimates (Xiao et al., 2010), referred to as
EC-MOD, were both derived from eddy covariance (EC) and MODIS
data. The EC-MOD dataset has high temporal and spatial resolu-
tions, and are highly constrained by eddy covariance data. EC-MOD
provides alternative, independent gridded flux estimates for U.S.,
and is useful for evaluating simulations of ecosystem models and
atmospheric inversions.
We examined the spatial patterns, magnitude, and interannual
variability of U.S. ecosystem carbon exchange using our continu-
ous NEE estimates. We estimated the terrestrial carbon sink in the
conterminous U.S. at 0.63pgCyr−1with the majority of the sink in
regions dominated by evergreen and deciduous forests and savan-
nas. Our results show that U.S. ecosystems play an important role
in slowing down the buildup of CO2in the atmosphere. Our results
also show that recent U.S. annual NEE exhibited significant year-
to-year variations. The dominant sources of the recent interannual
variation included extreme climate events (e.g., drought) and dis-
turbances (e.g., wildfires, hurricanes). Our results also highlight
the need to improve our understanding of the impacts of stand-
replacing disturbances on the forest carbon budget. Our study
provides an alternative, independent, and novel constraint to the
net ecosystem carbon exchange of U.S. terrestrial ecosystems.
Acknowledgements
This study was supported by grants from the National Science
Foundation (NSF) and Department of Energy (DOE). We thank the
principal investigators and contributors of the MODIS data prod-
ucts, the Oak Ridge National Laboratory (ORNL) Distributed Active
Archive Center (DACCC), and the Earth Observing System (EOS)
Data Gateway for making these MODIS data products available.
The Level I Ecoregions map of North America was obtained from
Environmental Protection Agency (EPA), the Biscuit fire severity
data from J. Thompson, Harvard University, and the PRISM climate
database from the PRISM Group, Oregon State University. Com-
puting support was provided by the Rosen Center for Advanced
Computing, Purdue University. We also thank anonymous review-
ers and Dr. Anne Verhoef for their valuable comments on earlier
versions of the manuscript. [The EC-MOD dataset is available upon
request.]
Contributors: J.X. and Q.Z. designed the study; J.X. conducted the
research and analyzed the results; J.X. and Q.Z. wrote the paper;
B.E.L., D.D.B., J.C., A.D.R., K.J.D., D.Y.H., S.W., R.O., A.N., M.L.F., S.B.V.,
D.R.C., G.S., S.M., S.C.W., P.V.B., S.P.B., P.S.C., B.G.D., M.F., D.R.F., L.G.,
J.L.H., G.G.K., M.L., S.M., T.A.M., R.M., T.P.M., R.K.M., J.W.M., W.C.O.,
K.T.P.U., H.P.S., R.L.S., G.S., A.E.S., and M.S.T. contributed eddy flux
data; B.E.L., D.D.B., J.C., A.D.R., J.M.M., K.J.D., D.Y.H., S.W., R.O., A.N.,
M.L.F., S.B.V., and D.R.C. provided comments on the manuscript.
Authors from P.V.B. to M.S.T. are listed alphabetically.
References
Amiro, B.D., Barr, A.G., Barr, J.G., Black, T.A., Bracho, R., Brown, M., Chen, J., Clark, K.L.,
Davis, K.J., Desai, A.R., Dore, S., Engel, V., Fuentes, J.D., Goldstein, A.H., Goulden,
M.L., Kolb, T.E., Lavigne, M.B., Law, B.E., Margolis, H.A., Martin, T., McCaughey,
J.H., Misson, L., Montes-Helu, M., Noormets, A., Randerson, J.T., Starr, G., Xiao, J.,
in press. Ecosystem carbon dioxide fluxes after disturbance in forests of North
America. J. Geophys. Res., doi:10.1029/2010JG001390.
Anthoni, P.M., Unsworth, M.H., Law, B.E., Irvine, J., Baldocchi, D.D., Tuyl, S.V., Moore,
D., 2002. Seasonal differences in carbon and water vapor exchange in young and
old-growth ponderosa pine ecosystems. Agric. Forest Meteorol. 111, 203–222.
AtlanticOceanographic andMeteorological
http://www.aoml.noaa.gov/hrd/Storm pages/katrina2005.
Baldocchi, D.D., 2003. Asssessing the eddy covariance technique for evaluating car-
bon dioxide exchange rates of ecosystems: past, present, and future. Global
Change Biol. 9, 479–492.
Baldocchi, D.D., Falge, E., Gu, L., Olson, R., Hollinger, D., Running, S., Anthoni, P.,
Bernhofer, C., Davis, K., Evans, R., Fuentes, J., Goldstein, A., Katul, G., Law, B.,
Lee, X., Malhi, Y., Meyers, T., Munger, W., Oechel, W., Paw, U.K.T., Pilegaard,
Laboratory(AOML),2007.
K., Schmid, H.P., Valentini, R., Verma, S., Vesala, T., Wilson, K., Wofsy, S., 2001.
FLUXNET: a new tool to study the temporal and spatial variability of ecosystem-
scale carbon dioxide, water vapor, and energy flux densities. Bull. Am. Meteorol.
Soc. 82, 2415–2434.
Baldocchi, D., 2008. ‘Breathing’ of the terrestrial biosphere: lessons learned from a
global network of carbon dioxide flux measurements systems. Aust. J. Bot. 56,
1–26.
Brewer, C.K., Winne, J.C., Redmond, R.L., Opitz, D.W., Mangrich, M.V., 2005. Classify-
ing and mapping wildfire severity: a comparison of methods. Photogram. Eng.
Remote Sens. 71, 1311–1320.
Caspersen, J.P., Pacala, S.W., Jenkins, J.C., Hurtt, G.C., Moorcroft, P.R., Birdsey, R.A.,
2000. Contributions of land-use history to carbon accumulation in U.S. forests.
Science 290, 1148–1151.
Chambers, J.Q., Fisher, J.I., Zeng, H., Chapman, E.L., Baker, D.B., Hurtt, G.C., 2007. Hur-
ricane Katrina’s carbon footprint on U.S. Gulf Coast Forest Sci. 318, 1107–11107.
Ciais, P.H., Reichstein, M., Viovy, N., Granier, A., Ogée, J., Allard, V., Aubinet, M.,
Buchmann, N., Bernhofer, C., Carrara, A., Chevallier, F., Noblet, N.D., Friend, A.D.,
Friedlingstein, P., Grünwald, T., Heinesch, B., Keronen, P., Knohl, A., Krinner, G.,
Loustau, D., Manca, G., Matteucci, G., Miglietta, F., Ourcival, J.M., Papale, D., Pile-
gaard, K., Rambal, S., Seufert, G., Soussana, J.F., Sanz, M.J., Schulze, E.D., Vesala,
T., Valentini, R., 2005. Europe-wide reduction in primary productivity caused by
the heat and drought in 2003. Nature 437, 529–533.
Clark, D.A., Brown, S., Kicklighter, D.W., Chambers, J.Q., Thomlinson, J.R., Ni, J., 2001.
Measuring net primary production in forests: concepts and field methods. Ecol.
Appl. 11, 356–370.
Cohen, W.B., Spies, T.A., Alig, R.J., Oetter, D.R., Maiersperger, T.K., Fiorella, M., 2002.
Characterizing23years(1972–1995)ofstandreplacementdisturbanceinwest-
ern Oregon forests with Landsat imagery. Ecosystems 5, 122–137.
Deng, F., Chen, J.M., Ishizawa, M., Yuen, C.-W., Mo, G., Higuchi, K., Chan, D., Maksyu-
tov,S.,2007.GlobalmonthlyCO2fluxinversionwithafocusoverNorthAmerica.
Tellus 59B, 179–190.
Energy Information Administration,Department
http://www.eia.doe.gov.
Environmental Protection Agency (EPA), 2009, http://www.epa.gov/wed/pages/
ecoregions.htm.
Fensholt,R.,Sandholt,I.,2003.Derivationofashortwaveinfraredwaterstressindex
from MODIS near- and shortwave infrared data in a semiarid environment.
Remote Sens. Environ. 87, 111–121.
Friedl, M.A., McIver, D.K., Hodges, J.C.F., Zhang, X.Y., Muchoney, D., Strahler, A.H.,
Woodcock, C.E., Gopal, S., Schneider, A., Cooper, A., Baccini, A., Gao, F., Schaaf,
C., 2002. Global land cover mapping from MODIS: algorithms and early results.
Remote Sens. Environ. 83, 287–302.
Gao, B.C., 1996. NDWI—a normalized difference water index for remote sensing of
vegetation liquid water from space. Remote Sens. Environ. 58, 257–266.
Göckede, M., Foken, T., Aubinet, M., Aurela, M., Banaz, J., Bernhofer, C., Boonefond,
J.M., Brunet, Y., Carrara, A., Clement, R., Dellwik, E., Elbers, J., Eugster, W., Fuhrer,
J., Granier, A., Grünwald, T., Heinesch, B., Janssens, I.A., Knohl, A., Koeble, R., Lau-
rila, T., Longdoz, B., Manca, G., Marek, M., Markkanen, T., Mateus, J., Matteucci,
G., Mauder, M., Migliavacca, M., Minerbi, S., Moncrieff, J., Montagnani, L., Moors,
E., Ourcival, J.-M., Papale, D., Pereira, J., Pilegaard, K., Pita, G., Rambal, S., Reb-
mann, C., Rodrigues, A., Rotenberg, E., Sanz, M.J., Sedlak, P., Seufert, G., Siebicke,
L., Soussana, J.F., Valentini, R., Vesala, T., Verbeeck, H., Yakir, D., 2008. Quality
control of CarboEurope flux data. Part 1. Coupling footprint analyses with flux
data quality assessment to evaluate sites in forest ecosystems. Biogeosciences
5, 433–450.
Goodale, C.L., Apps, M.J., Birdsey, R.A., Field, C.B., Heath, L.S., Houghton, R.A., Jenkins,
J.C., Kohlmaier, G.H., Kurz, W., Liu, S., Nabuurs, G.-J., Nilsson, S., Shvidenko, A.Z.,
2002. Forest carbon sinks in the Northern Hemisphere. Ecol. Appl. 12, 891–899.
Goward, S.N., Masek, J.G., Cohen, W., Moisen, G., Collatz, G.J., Healey, S., Houghton,
R.A., Huang, C., Kennedy, R., Law, B., Powell, S., Turner, D., Wulder, M.A.,
2008. Forest disturbance and North American carbon flux. EOS Trans. AGU 89,
doi:10.1029/2008EO110001.
Hargrove, W.W., Hoffman, F.M., Law, B.E., 2003. New analysis reveals representa-
tiveness of the AmeriFlux network. EOS Trans. 84, 529–544.
Houghton, R.A., Hackler, J.L., Lawrence, K.T., 1999. The U.S. carbon budget: contribu-
tions from land-use change. Science 285, 574–578.
Huete,A.,Didan,K.,Miura,T.,Rodriguez,E.P.,Gao,X.,Ferreira,L.G.,2002.Overviewof
the radiometric and biophysical performance of the MODIS vegetation indices.
Remote Sens. Environ. 83, 195–213.
Intergovernmental Panel on Climate Change, Climate Change 2007 – The Physical
Science Basis, 2007. Contribution of Working Group I to the Fourth Assessment
Report of the IPCC. Cambridge University Press, New York.
Lal, R., Kimble, J.M., Follett, R.F., Stewart, B.A., 2001. Assessment Methods for Soil
Carbon. Advances in Soil Science. Lewis Press, Boca Raton, FL, pp. 676.
Law, B.E., Turner, D., Campbell, J., Sun, O.J., Tuyl, S.V., Ritts, W.D., Cohen, W.B., 2004.
Disturbances and climate effects on carbon stocks and fluxes across Western
Oregon USA. Global Change Biol. 10, 1429–1444.
Lloyd,J.,Taylor,J.A.,1994.Onthetemperaturedependenceofsoilrespiration.Funct.
Ecol. 8, 315–323.
Lutes,D.C.,Keane,J.F.,Caratti,C.H.,Key,C.H.,Benson,N.C.,Gangi,L.J.,2004.FIREMON:
Fire Effects Monitoring and Inventory System (US Department of Agriculture
Forest Service, Rocky Mountain Research Station, Ogden, UT), Vol. RMRS-GTR-
164-CD, p. 400.
McCarthy, H.R., Oren, R., Kim, H.-S., Johnsen, K.H., Maier, C., Pritchard, S.G.,
Davis, M.A., 2006. Interaction of ice storms and management practices on
current carbon sequestration in forests with potential mitigation under
ofEnergy(DOE),2008.

Page 10

J. Xiao et al. / Agricultural and Forest Meteorology 151 (2011) 60–69
69
future CO2 atmosphere. J. Geophys. Res. 111, doi:10.1029/2005JD006428,
D15103.
Miller, J.D., Yool, S.R., 2002. Mapping forest post-fire canopy consumption in several
overstory types using multi-temporal Landsat TM and ETM data. Remote Sens.
Environ. 82, 481–496.
Moffat, A.M., Papale, D., Reichstein, M., Hollinger, D.Y., Richardson, A.D., Barr,
A.G., Beckstein, C., Braswell, B.H., Churkina, G., Desai, A.R., Falge, E., Gove, J.H.,
Heimann, M., Hui, D., Jarvis, A.J., Kattge, J., Noormets, A., Stauch, V.J., 2007. Com-
prehensive comparison of gap-filling techniques for eddy covariance net carbon
fluxes. Agric. Forest Meteorol. 147, 209–232.
Pacala, S.W., Hurtt, G.C., Baker, D., Peylin, P., Houghton, R.A., Birdsey, R.A., Heath,
L., Sundquist, E.T., Stallard, R.F., Ciais, P., Moorcroft, P., Caspersen, J.P., Shevli-
akova, E., Moore, B., Kohlmaier, G., Holland, E., Gloor, M., Harmon, M.E., Fan,
S.M., Sarmiento, J.L., Goodale, C.L., Schimel, D., Field, C.B., 2001. Consistent land-
and atmosphere-based U.S. carbon sink estimates. Science 292, 2316–2320.
Papale, D., Valentini, A., 2003. A new assessment of European forests carbon
exchange by eddy fluxes and artificial neural network spatialization. Global
Change Biol. 9, 525–535.
Peters, W.P., Jacobson, A.R., Sweeney, C., Andrews, A.E., Conway, T.J., Masarie, K.,
Miller, J.B., Bruhwiler, L.M.P., Pétron, G., Hirsch, A.I., Worthy, D.E.J., van der Werf,
G.R., Randerson, J.T., Wennberg, P.O., Krol, M.C., Tans, P.P., 2007. An atmospheric
perspective on North American carbon dioxide exchange: carbontracker. Proc.
Natl. Acad. Sci. U.S.A. 104, 18925–18930.
Powell,M.D.,Houston,S.H.,Amat,L.R.,Morisseau-Leroy,N.,1998.TheHRDreal-time
hurricane wind analysis system. J. Wind Eng. Indus. Aero. 77–78, 53–64.
PRISM Group, 2004. Oregon State University, http://www.prismclimate.org.
Rahman, A.F., Sims, D.A., Cordova, V.D., El-Masri, B.Z., 2005. Potential of MODIS EVI
and surface temperature for directly estimating per-pixel ecosystem C fluxes.
Geophys. Res. Lett. 32, doi:10.1029/2005GL024127, L19404.
Richardson, A.D., Hollinger, D.Y., Aber, J.D., Ollinger, S.V., Braswell, B.H., 2007. Envi-
ronmentalvariationisdirectlyresponsibleforshort-butnotlong-termvariation
in forest-atmosphere carbon exchange. Global Change Biol. 13, 788–803.
Richardson, A.D., Mahecha, M.D., Falge, E., Kattge, J., Moffat, A.M., Papale, D., Reich-
stein, M., Stauch, V.J., Braswell, B.H., Churkina, G., Kruijt, B., Hollinger, D.Y., 2008.
StatisticalpropertiesofrandomCOfluxmeasurementuncertaintyinferredfrom
model residuals. Agric. Forest Meteorol. 148, 38–50.
RuleQuest, 2008. http://www.rulequest.com. Visited on 10/18/2007.
Ryan, M.G., Law, B.E., 2005. Interpreting, measuring, and modeling soil respiration.
Biogeochemistry 73, 3–27.
Schimel, D., Melillo, J., Tian, H., McGuire, A.D., Kicklighter, D., Kittel, T., Rosenbloom,
N., Running, S., Thornton, P., Ojima, D., Parton, W., Kelly, R., Sykes, M., Neilson,
R., Rizzo, B., 2000. Contribution of increasing CO2and climate to carbon storage
by ecosystems in the United States. Science 287, 2004–2006.
Schmid, H.P., 1994. Source areas for scalars and scalar fluxes. Boundary Layer Mete-
orol. 67, 293–318.
Snyder, K.A., Tartowski, S.L., 2006. Multi-scale temporal variation in water availabil-
ity: implications for vegetation dynamics in arid and semi-arid ecosystems. J.
Arid Environ. 65, 219–234.
SOCCR, 2007. In: King, A.W., Dilling, L., Zimmerman, G.P., Fairman, D.M., Houghton,
R.A., Marland, G.A., Rose, A.Z., Wilbanks, T.J. (Eds.), The First State of the Carbon
Cycle Report (SOCCR). The North American Carbon Budget and Implications for
the Global Carbon Cycle. US Climate Change Science Program, Washington, DC,
p. 19.
Thompson, J.R., Spies, T.A., Ganio, L.M., 2007. Reburn severity in managed and
unmanaged vegetation in a large wildfire. Proc. Natl. Acad. Sci. U.S.A. 104,
10743–10748.
U.S. Fire Administration, 2008. http://www.usfa.dhs.gov.
Vermote, E.F., Vermeulen, A., 1999. MODIS Algorithm Technical Background Doc-
ument – Atmospheric Correction Algorithm: Spectral Reflectances (MOD09),
Version 4.0. http://modis.gsfc.nasa.gov/data/atbd/atbd mod08.pdf.
Wan, Z., Zhang, Y., Zhang, Q., Li, Z.-L., 2002. Validation of the land-surface
temperature products retrieved from Terra Moderate Resolution Imaging Spec-
troradiometer data. Remote Sens. Environ. 83, 163–180.
Waring, R.H., Franklin, J.F., 1979. Evergreen coniferous forests of the Pacific North-
west. Science 204, 1380–1386.
Wiedinmyer, C., Neff, J.C., 2007. Estimates of CO2 from fires in the United
States: implications for carbon management. Carbon Balance Manage. 2,
doi:10.1186/1750-0680-2-10.
Wofsy,S.C.,Goulden,M.L.,Munger,J.W.,Fan,S.-M.,Bakwin,P.S.,Daube,B.C.,Bassow,
S.L., Bazzaz, F.A., 1993. Net exchange of CO in a mid-latitude forest. Science 260,
1314–1317.
Wylie, B.K., Fosnight, E.A., Gilmanov, T.G., Frank, A.B., Morgan, J.A., Haferkamp, M.R.,
Meyers, T.P., 2007. Adaptive data-driven models for estimating carbon fluxes in
the Northern Great Plains. Remote Sens. Environ. 106, 399–413.
Xiao, J., Zhuang, Q., Baldocchi, D.D., Law, B.E., Richardson, A.D., Chen, J., Oren, R.,
Starr, G., Noormets, A., Ma, S., Verma, S.B., Wharton, S., Wofsy, S.C., Bolstad,
P.V., Burns, S.P., Cook, D.R., Curtis, P.S., Drake, B.G., Falk, M., Fischer, M.L., Fos-
ter, D.R., Gu, L., Hadley, J.L., Hollinger, D.Y., Katul, G.G., Litvak, M., Martin, T.A.,
Matamala,R.,McNulty,S.,Meyers,T.P.,Monson,R.K.,Munger,J.W.,Oechel,W.C.,
Paw, U.K.T., Schmid, H.P., Scott, R.L., Sun, G., Suyker, A.E., Torn, M.S., 2008. Esti-
mationofnetecosystemcarbonexchangefortheconterminousUnitedStatesby
combining MODIS and AmeriFlux data. Agric. Forest Meteorol. 148, 1827–1847,
doi:10.1016/j.agrformet.2008.06.015.
Xiao, J., Zhuang, Q., Liang, E., McGuire, A.D., Moody, A., Kiclighter, D.W., Shao, X.,
Melillo, J.M., 2009. Twentieth century droughts and their impacts on terrestrial
carbon cycling in China. Earth Interactions 13, 1–31, doi:10.1175/2009EI275.1
(010).
Xiao, J., Zhuang, Q., Law, B.E., Chen, J., Baldocchi, D.D., Cook, D.R., Oren, R., Richard-
son, A.D., Wharton, S., Ma, S., Martin, T.A., Verma, S.B., Suyker, A.E., Scott, R.L.,
Monson, R.K., Litvak, M., Hollinger, D.Y., Sun, G., Davis, K.J., Bolstad, P.V., Burns,
S.P., Curtis, P.S., Drake, B.G., Falk, M., Fischer, M.L., Foster, D.R., Gu, L., Hadley,
J.L., Katul, G.G., Matamala, R., McNulty, S., Meyers, T.P., Munger, J.W., Noormets,
A., Oechel, W.C., Paw, U.K.T., Schmid, H.P., Starr, G., Torn, M.S., Wofsy, S.C., 2010.
A continuous measure of gross primary production for the conterminous U.S.
derived from MODIS and AmeriFlux data. Remote Sens. Environ. 114, 576–591,
doi:10.1016/j.rse.2009.10.013.
Yamaji,T.,Sakai,T.,Endo,T.,Baruah,P.J.,Akiyama,T.,Saigusa,N.,Nakai,Y.,Kitamura,
K., Ishizuka, M., Yasuoka, Y., 2007. Scaling-up technique for net ecosystem pro-
ductivity of deciduous broadleaved forests in Japan using MODIS data. Ecol. Res.
23, doi:10.1007/s11284-007-0438-0.
Yang,F.,Zhu,A.-X.,Ichii,K.,White,M.A.,Hashimoto,H.,Nemani,R.R.,2008.Assessing
the representativeness of the AmeriFlux network using MODIS and GOES data.
J. Geophys. Res. 113, G04036, doi:10.1029/2007JG000627.
Zeng, N., Qian, H., Roedenbeck, C., Heimann, M., 2005. Impact of 1998–2002 mid-
latitude drought and warming on terrestrial ecosystem and the global carbon
cycle. Geophys. Res. Lett. 32, L22709, doi:10.1029/2005GL024607.