BGBiogeosciencesBGBiogeosciences1726-4189Copernicus GmbHGöttingen, Germany10.5194/bg-12-4385-2015Assessment of model estimates of land-atmosphere CO2 exchange across Northern EurasiaRawlinsM. A.rawlins@geo.umass.eduhttps://orcid.org/0000-0002-3323-8256McGuireA. D.KimballJ. S.DassP.https://orcid.org/0000-0003-3957-4055LawrenceD.https://orcid.org/0000-0002-2968-3023BurkeE.https://orcid.org/0000-0002-2158-141XChenX.https://orcid.org/0000-0002-3089-2260DelireC.KovenC.https://orcid.org/0000-0002-3367-0065MacDougallA.PengS.RinkeA.SaitoK.https://orcid.org/0000-0003-0846-0557ZhangW.https://orcid.org/0000-0001-9477-563XAlkamaR.BohnT. J.CiaisP.https://orcid.org/0000-0001-8560-4943DecharmeB.https://orcid.org/0000-0002-8661-1464GouttevinI.HajimaT.JiD.https://orcid.org/0000-0002-1887-887XKrinnerG.https://orcid.org/0000-0002-2959-5920LettenmaierD. P.MillerP.MooreJ. C.SmithB.SueyoshiT.Climate System Research Center, Department of Geosciences, University of Massachusetts, Amherst, MA, USAUS Geological Survey, Alaska Cooperative Fish and Wildlife Research Unit, University of Alaska, Fairbanks, Alaska 99775, USANTSG, University of Montana, Missoula, MT, USANational Center for Atmospheric Research, Boulder, CO, USAMet Office Hadley Centre, FitzRoy Road, Exeter, EX1 3PB, UKDepartment of Civil and Environmental Engineering, University of
Washington, Seattle, WA, USACRNM-GAME, Unité mixte de recherche CNRS/Meteo-France (UMR 3589), 42 av Coriolis, 31057 Toulouse, CEDEX, FranceLawrence Berkeley National Laboratory, Berkeley, CA, USASchool of Earth and Ocean Sciences, University of Victoria, Victoria, BC, CanadaLaboratoire des Sciences du Climat et de l'Environnement,
CEA-CNRS-UVSQ, UMR8212, 91191 Gif-sur-Yvette, FranceState Key Laboratory of Earth Surface Processes and Resource Ecology, College of Global Change and Earth System Science, Beijing Normal University, Beijing, ChinaAlfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, GermanyDepartment of Integrated Climate Change Projection Research, Japan Agency for Marine-Earth Science and Technology, Yokohama, Kanagawa, JapanDepartment of Physical Geography and Ecosystem Science, Lund University, Sölvegatan 12, SE 223 62 Lund, SwedenSchool of Earth and Space Exploration, Arizona State University, Tempe, AZ, USACNRS and Université Grenoble Alpes, LGGE, 38041, Grenoble, FranceIrstea, UR HHLY, 5 rue de la Doua, CS 70077, 69626 Villeurbanne, CEDEX, FranceDepartment of Geography, University of California, Los Angeles, CA, USANational Institute of Polar Research, Tachikawa, Tokyo, JapanM. A. Rawlins (rawlins@geo.umass.edu)28July201512144385440508January201503February201518May201501July2015This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://bg.copernicus.org/articles/12/4385/2015/bg-12-4385-2015.htmlThe full text article is available as a PDF file from https://bg.copernicus.org/articles/12/4385/2015/bg-12-4385-2015.pdf
A warming climate is altering land-atmosphere exchanges of carbon, with a
potential for increased vegetation productivity as well as the mobilization
of permafrost soil carbon stores. Here we investigate land-atmosphere carbon
dioxide (CO2) cycling through analysis of net ecosystem productivity (NEP)
and its component fluxes of gross primary productivity (GPP) and ecosystem
respiration (ER) and soil carbon residence time, simulated by a set of land
surface models (LSMs) over a region spanning the drainage basin of Northern
Eurasia. The retrospective simulations cover the period 1960–2009 at 0.5∘
resolution, which is a scale common among many global carbon and
climate model simulations. Model performance benchmarks were drawn from
comparisons against both observed CO2 fluxes derived from site-based eddy
covariance measurements as well as regional-scale GPP estimates based on
satellite remote-sensing data. The site-based comparisons depict a tendency
for overestimates in GPP and ER for several of the models, particularly at
the two sites to the south. For several models the spatial pattern in GPP
explains less than half the variance in the MODIS MOD17 GPP product. Across
the models NEP increases by as little as 0.01 to as much as 0.79 g C m-2 yr-2,
equivalent to 3 to 340 % of the respective model means, over the
analysis period. For the multimodel average the increase is 135 % of the mean
from the first to last 10 years of record (1960–1969 vs. 2000–2009), with a
weakening CO2 sink over the latter decades. Vegetation net primary
productivity increased by 8 to 30 % from the first to last 10 years,
contributing to soil carbon storage gains. The range in regional mean NEP
among the group is twice the multimodel mean, indicative of the uncertainty
in CO2 sink strength. The models simulate that inputs to the soil carbon
pool exceeded losses, resulting in a net soil carbon gain amid a decrease in
residence time. Our analysis points to improvements in model elements
controlling vegetation productivity and soil respiration as being needed for
reducing uncertainty in land-atmosphere CO2 exchange. These advances will
require collection of new field data on vegetation and soil dynamics, the
development of benchmarking data sets from measurements and remote-sensing
observations, and investments in future model development and intercomparison
studies.
Introduction
Northern boreal regions are known to play a major role in the land-atmosphere
exchange of CO2 at high latitudes . During the Holocene
the Arctic is believed to have been a net sink of carbon .
During modern times, often referred to as the anthropocene
, warming across the high northern latitudes has occurred
at a faster rate than the rest of the globe . The enhanced
warming is attributable to feedbacks involving biogeochemical and
biogeophysical processes . Warming may
increase soil microbial decomposition, placing the large permafrost carbon
pool at greater risk for being mobilized and transferred to the atmosphere as
greenhouse gases (GHGs), thus providing a positive feedback to global climate
. Warming may also lead to longer growing
seasons, contributing to increased plant productivity and ecosystem carbon
sequestration . At the same time, warming may
also lead to respiration increases through enhanced microbial activity and/or
increased input of plant photosynthates into the soil ,
offsetting any productivity increases and resulting in relatively low net
carbon uptake . Satellite observations show broad
greening trends in tundra regions ,
suggesting a potential increase in the land sink of atmospheric CO2. Some
areas, however, are browning .
Research studies point to uncertainty in the sign, magnitude and temporal
trends in contemporary land-atmosphere exchanges of CO2. A recent
synthesis of observations and models by suggests that
tundra regions across the pan-Arctic were a sink for atmospheric CO2 and a
source of CH4 from 1990–2009. However, a meta-analysis of 40 years of
CO2 flux observations from 54 studies spanning 32 sites across northern
high latitudes found that tundra was an annual CO2 source from the
mid-1980s until the 2000s, with the data suggesting an increase in winter
respiration rates, particularly over the last decade . In an
analysis of outputs from several models from recent terrestrial biosphere
model intercomparison projects, found that spatial patterns
in carbon stocks and fluxes over Alaska in 2003 varied widely, with some
models showing a strong carbon sink, others a strong carbon source, and some
showing the region as carbon neutral. It is critical to understand the net
carbon sink as recent studies suggest that with continued warming the Arctic
may transition from a net sink of atmospheric CO2 to a net source over the
coming decades .
In a study using a process model which included disturbances,
estimated a 73 % reduction in the strength of the pan-Arctic
land-based CO2 sink over 1997–2006 vs. previous decades in the late
20th century.
Recent studies have provided new insights into model uncertainties relevant
to our understanding of the land-based CO2 sink across Northern Eurasia.
Examining several independent estimates of the carbon balance of Russia
including two dynamic global vegetation models (DGVMs), two atmospheric
inversion methods, and a landscape-ecosystem approach (LEA) incorporating
observed data, concluded that estimates of heterotrophic
respiration were biased high in the two DGVMs, and that the LEA appeared to
give the most credible estimates of the fluxes. In an analysis of the
terrestrial carbon budget of Russia using inventory-based, eddy covariance,
and inversion methods, noted good agreement in net
ecosystem exchange among these bottom-up and top-down methods, estimating an
average CO2 sink across the three methods of 613.5 Tg C yr-1. Their
examination of outputs from a set of DGVMs, however, showed a much lower sink
of 91 Tg C yr-1. point to specification of vegetation
dynamics and nitrogen cycling in a subset of CMIP5 models as a potential
cause for their underestimation of changes in net productivity over the past
50 years. These analyses highlight the need for comprehensive assessments of
numerical model estimates of spatial and temporal variations in
land-atmosphere CO2 exchange against independent benchmarking data. A lack
of direct flux measurements across northern land areas presents considerable
challenges for model validation efforts .
In this study we examine model estimates of net ecosystem productivity (NEP)
and component fluxes gross primary productivity (GPP) and ecosystem
respiration (ER) across the arctic basin of Northern Eurasia from a series of
retrospective simulations for the period 1960–2009. Our analysis for the
region is unique in its synthesis of a large suite of land-surface models,
available site-level data, and a remote-sensing product. Study goals are
two-fold. First, using the available in situ data derived from tower-based
measurements and the remote-sensing GPP product we seek to assess model
efficacy in simulating spatial and temporal variations in GPP, ER, and NEP
across the region. In doing so we elucidate issues complicating evaluations
of model carbon cycle estimates across Northern Eurasia and, by extension,
other areas of the northern high latitudes. Second, we estimate time changes
in NEP and soil organic carbon (SOC) residence time and its controls as an
indicator of climate sensitivity and potential vulnerability of soil carbon
stocks. We focus the analysis and discussion on assessing how well the models
capture the seasonal cycle and spatial patterns in GPP and ER flux rates,
evaluating uncertainties in the net CO2 exchange given reported biases in
respiration rates, and in advancing understanding of the land–atmosphere
cycling of CO2 over recent decades.
MethodsStudy Region
The spatial domain is the arctic drainage basin of Northern Eurasia which
comprises all land areas draining to the Arctic Ocean, a region of some 13.5 million km2
(Fig. ). The basin covers roughly half of
the Northern Eurasian Earth Science Partnership Initiative (NEESPI) study
area, generally defined as the region between 15∘ E in the west, the
Pacific Coast in the east, 40∘ N in the south, and the Arctic Ocean
coastal zone in the north . Warming and associated
environmental changes to this region are among the most pronounced globally
. Tundra vegetation is common across northern
areas, with boreal forest and taiga comprising much of the remainder of the
region. Steppes and grasslands are found across a relative small area in the
extreme southwest. Continuous permafrost underlies over half of the region.
Sporadic and relic permafrost comprise the southwest portion of the domain.
West to east, the Ob, Yenisei, Lena, and Kolyma rivers drain a large fraction
of the total river discharge from the Northern Eurasian basin.
Study domain spanning the arctic drainage basin in
Northern Eurasia. Map panels show (a) plant functional types (PFTs) and
(b) permafrost classification along with tower sites used in the study:
(a) Chersky, (b) Chokurdakh, (c) Hakasija, and (d) Zotino locations
(Table ). Gridded PFTs are from the MODIS MOD12 product . Permafrost classes for
each grid are drawn from the CAPS data set ( (), 2003).
Monthly GPP at sites (a) Chersky, (b) Chokurdakh, (c) Hakasija, and
(d) Zotino (Obs, Table ). Colored lines trace monthly GPP for
each model grid that encompasses the tower location. Site Hakasija includes
research areas Ha1 (filled circle), Ha2 (open circle), and Ha3 (triangle)
Modeled data
We used outputs from retrospective simulations of nine models participating
in the model integration group of the Permafrost Carbon Network. All
simulation outputs available at the time of writing were included in the
analysis (http://www.permafrostcarbon.org). The
simulation protocol allowed for the choice of a model's driving data sets for
atmospheric CO2, N deposition, climate, disturbance, and other forcings
(Tables and ). Simulations were run at
daily or sub-daily time steps in some models and at 0.5∘ resolution
over all land areas north of 45∘ N latitude. The present study
focuses on analysis of spatial patterns and temporal changes in
land-atmosphere CO2 fluxes over the period 1960–2009. Quantities analyzed
are GPP, ER, and NEP, defined here as NEP = GPP-ER, where a positive value
represents a net sink of CO2 into the ecosystem. ER is the sum of
heterotrophic respiration and autotrophic respiration as estimated by the
models. In this study we follow the conceptual framework for NEP and related
terms as described in . For this Permafrost Carbon Network
activity modeling groups are providing gridded data for permafrost regions of
the northern hemisphere. The nine models examined here (full model names in
Table 1) are the (1) CLM version 4.5 (hereafter CLM4.5, );
(2) CoLM ; (3) ISBA ; (4) JULES
; (5) LPJ Guess WHyMe (hereafter LPJG,
); (6) MIROC-ESM
; (7) ORCHIDEE-IPSL
; (8) UVic
; and (9) UW-VIC .
Table lists the model elements most closely related to
CO2 source and sink dynamics. These include model land cover
initialization, time series forcings, light use efficiency, and CO2 and
nitrogen fertilization. Among the models there is a wide range of accounting
for processes related to disturbances such as fire and land use change
(Table ). All but two of the nine models (ISBA and
UW-VIC) are considered to be dynamic global vegetation models (DGVMs),
possessing the ability for vegetation to change over the model simulation.
For ORCHIDEE, dynamic vegetation was not enabled in the simulation examined
in this study. While studies that examine the overall ecosystem carbon
balance (i.e. the net ecosystem carbon balance, NECB) are elemental to our
understanding of the carbon cycle of Northern Eurasia, the present study
focuses on the patterns in NEP and component fluxes GPP and ER, common in all
of the models, in order to avoid the uncertainties given the range of model
formulations related to the full carbon balance. Outputs from several of the
nine models have been examined in other recent studies. The LPJG and ORCHIDEE
were used in the synthesis of data and models presented by
. JULES, LPJG, ORCHIDEE, and CLM4.5 participated in the
TRENDY MIP . CLM4.5, ORCHIDEE, and LPJG were three of the
eight models examined in the study of .
Models participating in the Vulnerability of Permafrost Carbon
Research Coordination Network (RCN) retrospective simulations. Modeling
groups provided outputs for year 1960–2009, with the exception of CLM
(–2005); JULES (–1999); UW-VIC (–2006).
ModelInstitutionClimate Data SetCommunity Land Model (CLM4.5)National Center for Atmospheric Research, USACRUNCEP41Common Land Model (CoLM)Beijing Normal University, ChinaPrinceton2Interaction Sol-Biosphère-Atmosphere (ISBA)National Centre for Meteorological Research, FranceWATCH3 WFDEI6,10Joint UK Land Environment Simulator (JULES)Met Office, United KingdomWATCH3Lund-Potsdam-Jenna General Ecosystem Simulator (LPJG)Lund University, SwedenCRU TS 3.14Model for Interdisciplinary Research on Climate, Earth System Model (MIROC)Japan Agency for Marine-Earth Science and Technology, JapanCMIP55Organising Carbon and Hydrology In Dynamic Ecosystems (ORCHIDEE)Institute Pierre Simon Laplace (IPSL), FranceWATCH3 WFDEI6,10University of Victoria (UVic)University of Victoria, CanadaCRUNCEP41Variable Infiltration Capacity (UW-VIC)University of Washington, USACRU7, UDel8, NCEP-NCAR9
1
(http://dods.extra.cea.fr/data/p529viov/cruncep/readme.htm),
2
(http://hydrology.princeton.edu/data.pgf.php),
3
(http://www.waterandclimatechange.eu/about/watch-forcing-data-20th-century),
4,
5,
6http://www.eu-watch.org/gfx_content/documents/README-WFDEI.pdf,
7 for temperature,
8 Willmott and Matsura (2001) for precipitation; and for precipitation
adjustments,
9 for wind speed,
10 WATCH used for 1901–1978; WFDEI used for 1978–2009.
As in Fig. , for ER.
As in Fig. , for NEP. NEP = GPP-ER.
Observational dataFlux tower eddy covariance data
Model estimates for GPP, ER, and NEP are evaluated against data from six eddy
covariance flux towers in four research areas located across Russia. The data
are contained in the La Thuile global FLUXNET data set .
FLUXNET represents a global network of tower eddy covariance measurement
sites for monitoring land-atmosphere exchanges of carbon dioxide and water
vapor (http://daac.ornl.gov/FLUXNET/fluxnet.shtml). For these sites, GPP and
ER data records overlap in the years 2002–2005. Observations during colder
months are few. Tower sites are identified here by their locations: Chersky
(CHE), Chokurdakh (COK), Hakasija (HAK), and Zotino (ZOT). Data from three
towers are available for Hakasija; HAK1 is in an area of grassland-steppe;
HAK2 is grassland; HAK3 an abandoned agricultural field. Chersky and
Chokurdakh are in northeast Russia in the general zone of tundra vegetation.
Hakasija and Zotino are in an area of generally higher productivity in
southern Siberia (Fig. 1). Data are available for years 2002–2004 at
Chersky, Hakasija and Zotino, and 2003–2005 at Chokurdakh. General
characteristics of these sites are summarized in Table . In
this data set GPP and ER are derived from an empirical model driven by
field-based eddy covariance measurements of net ecosystem CO2 exchange
(NEE) using methodologies described in .
Properties in each model relevant to simulation of land-atmosphere
CO2 dynamics, particularly for the northern high latitude terrestrial
biosphere. Properties are indicated as present (✓), absent (×) or
otherwise (see footnote for details).
CLM4.5CoLMISBAJULESLPJGMIROCORCHIDEEUVicUW-VICTree mortality/senescence included?✓/✓✓/✓×/×✓/×✓/✓✓/✓✓/✓×/××/×Light limits photosynthesis?✓✓✓✓✓✓✓✓✓N limits photosynthesis?××××✓×××✓Vegetation competes for light/water/nitrogen?×/✓/×✓/✓/××/×/×✓/×/×✓/✓/×✓/✓/×✓/✓/×✓/✓/××/×/×No. of PFTs161495151312520CO2 fertilization?××✓✓✓×✓××Turnover time of carbon in heartwood (yr)50process dependent30–50PFT dependentPFT dependent2020–80PFT dependent33.3Turnover time of carbon in sapwood (yr)502930–50PFT dependentPFT dependent201PFT dependent33.3Turnover time of carbon in leaves (yr)10.5–20.4–1PFT dependentPFT dependent0.15–4.580 daysPFT dependent2.86Turnover time of carbon in coarse/fine roots50 yr1–2 yr150–365 daysPFT dependentPFT dependent20/1.1–6.25 yr80 daysPFT dependent33.3Time step of carbon cycle0.5 h1 h30 min–1 day0.5 h1 month1 day0.5 h–1 day1 h3 hDisturbance (F/L/I)a?F+LF××FF+L×L×Vegetation dynamic?✓✓×✓✓✓×✓×Vegetation dynamics time stepNA1 yrNA10 days1 month1 yr1 yr5 daysNALAIb dynamic?✓✓✓✓✓×✓✓×LAImax prescribed?×××✓×✓✓××LAI time step0.5 h1 day1 day1 day1 month1 day1 day5 days30 daysMax veg height prescribed?×✓✓✓×d✓××✓Max rooting depthvariable3.4 m2 m3 m2 m1 mvariable3.35 m1 mCsoilc layered? (Depth)✓(4 m)×(3.4 m)×(1 m)implicitimplicitimplicit✓(2–47 m)✓(3.35 m)×Soil layers for hydrology101014302611825Biogenic CH4 fluxes✓×××✓×××✓Depth of water extraction (m)PFT dependent3.4PFT dependentPFT dependent22Soil depth limited3.351Approach to soil thermal dynamicsheat diffusionheat diffusionmulti-layer (Fourier law)multi-layer finitemulti-layer finiteheat conduction1-D FourierAvis (2011)Finite differencedifference modeldifference modelEffect of vegetation on soil thermal dynamics?✓×✓ (only at surface)✓✓××✓ (water+albedo)×Snow insulation typemulti-layermulti-layermulti-layermulti-layerimplicitmulti-layerimplicit–bulkCapable of talik formation and dynamics?×××✓×✓×✓✓
a Fire; Land-use change; Insects,
b Leaf Area Index,
c Soil carbon,
d max height prescribed for shrubs.
Satellite-based estimates of GPP
Satellite-data-driven estimates of annual total GPP are also obtained from
the MODIS (Moderate Resolution Imaging Spectroradiometer) MOD17 operational
product . The MOD17 product has been derived
operationally from the NASA EOS MODIS sensors since 2000 and provides a
globally consistent and continuous estimation of vegetation productivity at
1-km resolution and 8-day intervals. MOD17 uses a light use efficiency
algorithm driven by global land cover classification and canopy fractional
photosynthetically active radiation (FPAR) inputs from MODIS. The product
also uses daily surface meteorology inputs from global reanalysis data
, and land cover class specific biophysical response functions
to estimate the conversion efficiency of canopy absorbed photosynthetically
active radiation to vegetation biomass (g C MJ-1) and GPP
. The MOD17 algorithms and productivity estimates have been
extensively evaluated for a range of regional and global applications,
including northern, boreal and Arctic domains
. We use the MOD17 Collection 5
product, which has undergone five major reprocessing improvements since 2000.
The MOD17 data are used in this study as a consistent satellite-derived
baseline for evaluating GPP simulations from the detailed carbon process
models.
ResultsModel evaluation and benchmarkingSite-level evaluations
Confident assessment of uncertainties in land-atmosphere CO2 fluxes is
dependent on robust comparisons of model estimates against consistent
benchmarking data. We begin by assessing the seven models which provided
estimates through 2005, along with MOD17 GPP product. Monthly GPP from the
models and MOD17 are compared with the cumulative monthly tower values by
extracting the model values for the grid cell encompassing each tower site.
Error measures that are based on absolute values of differences, like the
mean absolute error (MAE) and mean bias error (MBE) are preferable to those
based on squared differences . Model
performance is evaluated here using the MBE, defined as the
difference between the model and observed values: ϵj=Cj-Cobs, where Cj is GPP, ER or NEP for model j and Cobs is the
observed tower value.
As shown in (Fig. ), MOD17 GPP agrees well with the tower
estimates for Chersky and Chokurdakh, with MBE over the 3 years of -2
and -11 g C m-2 month-1, respectively (Table ).
MOD17 GPP broadly agrees with the observations at Hakasija and Zotino.
Average MBEs are 13 and 10 g C m-2 month-1, respectively, for these
sites with higher productivity than Chersky and Chokurdakh. Averaged across
all models the error in GPP is 7, 34, 34, and 13 g C m-2 month-1 for
Chersky, Chokurdakh, Hakasija and Zotino, respectively. The MBE for ER are 8,
35, 43, and 33 g C m-2 month-1, respectively.
Flux tower sites from the LaThuile data set
used in this study. Site Hakasija consists of records
from 3 sub-sites which all fall within the same RCN model grid. Each sub-site
is represented with a different symbol in Figs. c,
c, c. GPP and ER in the La Thuile
data set are calculated using methodologies described in
.
sitecoordinatesIGBP classstart/end yearsChersky (CHE)68.61∘ N, 161.34∘ Emixed forest2002–2004Chokurdakh (COK)70.62∘ N, 147.88∘ Eopen shrubland2003–2005Hakasija* (HAK)54.77∘ N, 89.95∘ Egrassland2002–2004Zotino (ZOT)60.80∘ N, 89.35∘ Eevergreen needleleaf forest2002–2004
* Data used from three research sites (HAK1, HA2, HAK3).
Overall the models simulate fairly well the seasonal cycle in GPP
(Fig. ) and ER (Fig. ), including the
timing of peak CO2 drawdown. Modest overestimates are noted near growing
season peak at Hakasija and Zotino. However, for all four sites significant
over- and under-estimates in GPP and ER are also noted
(Table ). For the two sites in the south there is a tendency
for overestimation in GPP and ER. All models overestimate both GPP and ER at
Hakasija. Seven of the nine models overestimate GPP and ER at Zotino, with ER
overestimated by a considerable degree. Overestimates in ER for Hakasija and
Zotino during late summer and autumn are particularly noteworthy. An ANOVA
test was carried out to determine whether model errors in ER exceed the
errors in GPP. The tests confirm that ER errors are greater on average
than the GPP errors for comparisons where (i) ER errors for all sites are
pooled together and compared against GPP pooled across all sites and (ii) ER
and GPP errors for the two northern sites are pooled and compared against ER
and GPP errors from the two southern sites.
Mean annual gross primary productivity (GPP) from the
permafrost RCN models and from the MOD17 product. The averaging period is
2000–2009 for GPP from the MOD17 product and all models with the exception
of CLM4.5 (1995–2004); CoLM (1991–2000); and JULES (1991–2000). Spatial
correlations between MOD17 GPP and each model GPP for all grids is shown at
upper left in each map panel. Map panel at upper right is coefficient of
variation (CV) for GPP. At each grid the CV is estimated from the mean and
standard deviation across the nine models (MOD17 not included).
Distributions for mean annual GPP from the models and
the MOD17 product over the averaging period listed in
Fig. . The rectangles bracket the 25th
and 75th percentiles. Whiskers extend to the
5th and 95th percentiles. Thick and thin
horizontal lines mark the mean and median respectively.
Average model error in g C m-2 month-1 for site-level
comparisons over the years 2002–2005 shown in Figs. 2–4. Errors are
calculated as the average (ϵj^) over all years and months
for which a model estimate and site estimate are available at a given site.
Thus, for each site and month, the mean bias error (MBE) is calculated as the
average difference between the model and observed values: ϵj=Cj-Cobs, where Cj is GPP, ER or NEP for model j and Cobs
is the observed value from the La Thuile FLUXNET observations
. The last column lists mean NEP error
(NEP‾) across all sites. Model estimates for years
2002–2005 are not available for CoLM and JULES. Differences were evaluated
using a 2-way repeated measures ANOVA test. Test design was a comparison of
GPP vs ER t tests for (i) each area separately; (ii) GPP and ER pooled for
the two tundra sites and across the two forest sites; and (iii) GPP
errors pooled across the four sites vs. ER errors pooled across the four
sites.
CHE COK HAK ZOT ModelGPPERNEPGPPERNEPGPPERNEPGPPERNEPNEP‾MOD17-2––-11––13––10––CLM4.5-25-19-6-42-23-19822-157881-3-11ISBA272523441-7827838298-16-5LPJG-10-5-5-5-1-45374-22-34-13-20-13MIROC20182494362837-10-421-25-7ORCHIDEE2312114932171621-6-30-6-24-1UVic-14-7-71636-203038-9-731-38-19UW-VIC2734-6140119191833-16220-18-5Average78-13435-13443-111333-20-8
Spatially averaged ER vs. GPP over the period
1960–2009. Horizontal and vertical lines span the range across the
5th and 75th percentiles for GPP and ER,
respectively. The GPP 5th and 75th percentiles
are shown in Fig. . NEP is equal to the difference GPP
minus ER.
The tendency to overestimate ER leads to discrepancies in net CO2 source
(negative NEP) at Hakasija and Zotino, particularly in autumn
(Fig. ). Average NEP errors are -11 and
-20 g C m-2 month-1 for Hakasija and Zotino, respectively
(Table ). Errors in the magnitude and timing of NEP prior to
and following the dormant season are much smaller at Chersky, and to some
extent Chokurdakh. However, a lack of available tower-based data during the
colder months limits the robustness of our assessments during that time of
year.
(a) Annual NEP (1960–2009) averaged across the nine
models. Areas in blue are a net annual source of CO2. (b) Coefficient of
variation as estimated from the across model mean and standard deviation for
each grid.
We further evaluate model performance through two additional error metrics,
the refined index of agreement (dr) and the
Nash-Sutcliffe coefficient of efficiency (E) . As described
by , the refined index of agreement (dr) involves the
sum of the magnitudes of the differences between the model-predicted and
observed deviations about the observed mean, relative to the sum of the
magnitudes of the perfect-model (model predicted = observed) and observed
deviations about the observed mean. It is bounded between -1 and +1. When
dr equals 0.0, it signifies that the sum of the magnitudes of the errors
and the sum of the perfect-model-deviation and observed-deviation magnitudes
are equivalent. Like dr, the Nash-Sutcliffe E considers observed
deviations within the basis of comparison. For both metrics, values closer to
1 indicate higher model accuracy. Nash-Sutcliffe's E is also positively
correlated with dr. Values of E less than zero occur when the residual
model variance is larger than the data variance.
A wide range of model performance is evident from
Table . As with the mean errors shown in
Table , agreements with observations are generally better at
Chersky and Chokurdakh than Hakasija and Zotino. ER errors are also
greater than GPP errors. Nash-Sutcliffe Es are negative for all models for
both GPP and ER at Hakasija, and for most of the comparisons at Chokurdakh.
Models CLM4.5, ISBA and UW-VIC exhibit the largest disagreements among the
seven models for which estimates are available over the 2002–2005 period.
Nash-Sutcliffe coefficient of efficiency (E) and
Willmott's refined index of agreement (dr) for
comparison of GPP and ER errors derived from comparisons at sites shown in
Table .
CHE COK HAK ZOT ModelGPPERGPPERGPPERGPPERCLM4.50.15,0.67-0.09,0.50-0.74,0.44-1.52,0.15-1.20,0.39-2.77,-0.03-0.19,0.66-5.34,-0.19ISBA0.43,0.67-0.79,0.34-0.04,0.54-5.64,-0.26-10.25,-0.24-19.44,-0.55-0.82,0.62-10.56,-0.34LPJG0.64,0.770.68,0.760.86,0.830.62,0.71-5.37,-0.09-26.99,-0.640.76,0.850.64,0.76MIROC0.49,0.76-0.38,0.48-1.23,0.33-8.02,-0.29-2.69,0.24-2.85,-0.010.95,0.940.35,0.60ORCHIDEE0.44,0.690.45,0.66-1.08,0.32-3.37,-0.04-2.39,0.33-1.29,0.210.80,0.870.74,0.83UVic0.35,0.680.69,0.760.59,0.74-3.98,-0.14-1.93,-0.44-9.50,-0.410.91,0.87-0.17,0.50VIC0.14,0.67-3.41,0.10-14.88,-0.45-60.73,-0.74-2.04,0.30-0.32,0.610.83,0.87-0.27,0.56Regional-level evaluation of model GPP
Estimates from the MOD17 product provide a temporally and spatially
continuous benchmark to assess model simulated GPP over the study domain.
Average annual-total GPP from MOD17 over the period 2000–2009 is shown in
Fig. . The MOD17 product clearly captures three distinct
land cover zones over the region, representing: (i) grasslands across the
south; (ii) boreal forests in the center of the region; and (iii) tundra to
the north. Highest production occurs in the western forests where mean annual
temperatures are higher. Both the steppe and tundra areas show annual GPP of
less than 300 g C m-2 yr-1. Areas of low productivity in high
elevation areas to the north are well delineated. The spatially averaged mean
across the region is approximately 470 g C m-2 yr-1. In most of the
models the patterns in GPP broadly represent the major biome areas captured
in the MODIS land cover product (Fig. a). The east to west
gradient is broadly captured in most of the models. However, grid-based
correlations with the MOD17 GPP estimates (upper left of map panels in Fig. 5)
show a wide range of agreement across the models. Spatial averages of the
correlations across the domain range from r=0.92 (ISBA) to r=0.48 (ORCHIDEE).
Four of the nine (LPJG, MIROC, ORCHIDEE, UVic) simulate a GPP field that
explains less than 44 % of the variability in GPP found within the MOD17
product. Annual GPP in the LPJG is notably low across the eastern half of the
region. The CLM4.5 tends to predict lower GPP than MOD17 over tundra areas
and higher productivity in the boreal zone. As estimated by the coefficient
of variation (CV, upper right panel of Fig. ), agreement in
GPP is best across the higher productivity taiga biome.
Figure shows the distribution of GPP for all grids of
each model. In general, the models bracket the MOD17 estimates, with several
models showing a larger spread and several showing a reduced spread. Regional
averages from each model fall within ±20 % of the MOD17
average of 468 g C m-2 yr-1, with the exception of
the LPJG model for which annual GPP is 40 % lower than MOD17.
Distributions for mean annual NEP from the models over the averaging
period listed in Fig. . Boxplot quartiles are as described
in caption for Fig. .
Annual NEP as a spatial average across the region
for each year 1960–2009.
For each model the spatial pattern in ER (not shown) closely matches the
pattern in GPP, consistent with the strong dependence of autotrophic
respiration and litterfall on vegetation productivity
. Area-averaged GPP and ER are highly correlated
(r=0.99, Fig. ). That is, models which simulate low
(high) GPP also simulate low (high) ER.
Spatial patterns and area averages
In this study net ecosystem productivity (NEP) represents the net exchange of
CO2 between the land surface and the atmosphere. NEP is defined as the
difference between GPP and ER. We do not examine other emission components of
land-atmosphere CO2 exchange , as several of the models
possess limited representation of disturbance processes important for carbon
cycling in boreal forest regions (e.g. fire and forest harvest). The
multimodel mean NEP is highest over the south-central part of the region and
lowest in the tundra to the north (Fig. a). Only 0.3 %
of the region is a net annual source of CO2, notably two small areas in
Scandinavia. Tundra areas are a net sink of approximately 15 g C m-2 yr-1
based on the multimodel mean NEP. As measured by the coefficient of
variation (CV), the agreement in NEP among the models is highest across the
boreal region and lowest in the tundra to the north and grasslands to the
south (Fig. b). The multimodel mean NEP is
approximately 20 g C m-2 yr-1 or 270 Tg C yr-1 over the
simulation period (Fig. ). Among the models, NEP varies
from 4 (UVic) to 48 (JULES) g C m-2 yr-1, a range that is double
the multimodel mean. The UVic simulates a negative NEP (CO2 source) for
nearly half of the region, and the CoLM and MIROC for nearly 25 % of the
region.
Cumulative NEP in Pg C over the simulation period
for each model.
Temporal changes over the period 1960–2009
Figure shows the time series of regionally
averaged annual NEP each year over the period 1960–2009 for each model.
Across the model group annual NEP is positive in most but not all years.
Several models show a net source of CO2 in some years, primarily during
the earlier decades of the period. Among the models NEP increases by 0.01 to
0.79 g C m-2 yr-2, (3 to 340 % of the respective model means)
based on a linear least squares (LLS) regression (Table ).
Seven of the models (CLM4.5, CoLM, ISBA, JULES, LPJG, MIROC, ORCHIDEE) show
statistically significant trends at the p<0.01. Taking averages over the
first decade (1960–1969) and last decade (2000–2009) we estimate that the
NEP change ranges from 10 to 400 % of the first decade mean, with a nine
model average of 135 %. For each model the GPP trend magnitude exceeds the ER
trend magnitude (Table ), hence the increase in NEP over
time. The increases from the first to last decade of the simulations range
from 9–35 % of the early decade average for GPP and 8–30 % for ER. Total
cumulative NEP over the 50-year period and averaged across all models is
approximately 12 (range 3–20) Pg C (Fig. ).
Averaged across the models, NEP exhibits an increase during mainly the
earliest decades that tends to weaken over the latter decades
(Fig. ). The uncertainty range for the
multimodel mean shows that the region has been a net sink for CO2 over
the simulation period. Interestingly the uncertainty range reflects
relatively better model agreement in annual NEP (lower variance) during the
years 1960–1965 and in the low NEP years 1978 and 1996. Amid this increase
there is evidence of a deceleration in NEP. The deceleration is apparent
when examining trend magnitude and significance across all time intervals
(minimum 20-year interval) over the simulation period
(Fig. ). Here several models (ISBA, LPJG, ORCHIDEE)
exhibit weaker linear trends over time and all models show a lack of
significant positive trends for time intervals spanning the latter decades
(e.g. 1980–1999 or 1982–2009). While temporal trends in NEP are highly
variable across the models, it is clear that the greatest increases in NEP
occurred during the earliest decades of the simulation period. The LLS trend
is significant for 20 of 42 (48 %) possible time periods beginning in 1975 or
later, whereas 72 of 107 (67 %) are significant for periods starting in
1960–1962.
Residence Time
Annual estimates of residence time (RT) are calculated for each model and at
each grid cell over the period 1960–2009 using model soil carbon storage and
the rate of heterotrophic respiration (Rh). Among the models RT
(long-term climatological mean) varies from 40 (CoLM) to 400 years (CLM4.5),
and largely by model soil carbon amount, which varies by an order of
magnitude across the models. Over the period examined all of the models
simulate a statistically significant (p< 0.01) decrease in the
regionally-averaged RT. Across the models the decrease from first to last
decade of the study period ranges from -5 to -16 % of each model’s
mean. The decline occurs amid an increase in SOC storage over time. All
models with the exception of CoLM simulate a statistically significant
increase in soil carbon and all exhibit an increase in Rh. The increases
in carbon storage range from 0.2 to 3.6 % while the increases in Rh
range from 7 to 22 %. Likewise the models simulate an increase in the
rate of net primary productivity (NPP) of 8 to 30 %. Across the model group
the change in RT is highly correlated (r= 0.99) with change in Rh. In
essence, higher rates in Rh and NPP led to a decrease in soil carbon RT,
with increased soil carbon storage resulting from enhanced vegetation
productivity and litterfall inputs.
Trend in GPP, ER, and NEP over simulation period for each model.
Trend slopes (g C m-2 yr-2) are estimated using an auto-regressive
AR[1] model to account for temporal autocorrelation. Standard error for the
regression is indicated in ( ). Standard deviation of the model means is
shown in [ ]. Significant trends (p< 0.01) are denoted with an asterisk
(*).
The spatial pattern in RT changes suggests that controlling influences are
leading to both decreases and increases over different parts of the region.
The largest decreases are found across north-central Russia and the eastern
third of the domain (Fig. a). The decline in RT is
statistically significant (p< 0.01) for just over 46 % of the region.
exceeding -20 % for approximately 16 % of the region. An increase in RT is
noted for less than 5 % of the region, including a small area in the far
north and across extreme southern parts of the region. The change, however,
is not significant in those areas. The CV map (Fig. b)
lends further confidence to the RT decreases across much of the center of the
region. High uncertainties (CVs > 10) are noted in the areas where the
multimodel average suggests an increase in RT.
Spatially averaged annual NEP as an average across
the nine models. Gray region marks the 95th confidence
interval, where CI =μ± (SE × 1.96), where μ is the nine
model average and SE is the standard error. Standard deviation (σ)
used to estimate SE is obtained each year from the set of nine model NEP
values used to obtain the yearly average.
Magnitude of linear trend in NEP over given time
interval for all trends significant at p< 0.05. For each model, linear
trends are calculated for all time intervals of 20 years or more. For
example, 1960–1979, 1960–1980, …, 1990–2009. Intervals for which the
trend is significant are marked with a line from the start to end year of the
interval and shaded by the trend magnitude. As an example, one time interval
is identified with a significant NEP trend for UW-VIC, from 1964–1993.
DiscussionUncertainties in tower-based measurements
The potential for alterations to the terrestrial sink of atmospheric CO2
across the high northern latitudes motivates our examination of model
estimates of land-atmosphere exchanges of CO2 across the arctic drainage
basin of Northern Eurasia. Validation of model estimates through comparisons
to measured flux tower data is hindered by several factors. The limited
extent of available measurements from a sparse regional tower network clearly
challenges the validation of model estimates and, in turn, identification of
model processes which require refinement. There are also inherent
uncertainties in GPP and ER data derived from net ecosystem exchange (NEE)
measurements at the eddy covariance tower sites. ER is generally assumed to
equal NEE during nighttime hours . An empirical
relationship is derived to estimate ER during that time and it is
extrapolated into the daylight hours. GPP is then generally calculated as the
difference between NEE and ER (accounting for appropriate signs). Since there
is generally daylight for photosynthesis during the middle of the summer, ER
could potentially be underestimated if primary production had occurred during
the hours used for ER model calibration. Direct validation of the
partitioning of measured NEE flux to GPP and ER is not possible. However, in
a recent sensitivity study compared two independent
methods for partitioning and found general agreement in the results. This
agreement across methods increases our confidence in the partitioned GPP and
ER estimates in the LaThuile FLUXNET data set. When measurements come from
nearly-ideal sites the error bound on the net annual exchange of CO2 has
been estimated to be less than ±50 g C m-2 yr-1. Systematic errors in eddy covariance fluxes due to
non-ideal observation conditions are uncertain at this time. Total error is
likely below the value of 200 g C m-2 yr-1 that has been
conservatively estimated . The model errors estimated in
this present study often exceed that level for site Hakasija and, for a few
models, Zotino as well. Lastly, any conclusions about the CO2 sink
strength drawn from such a limited number of eddy covariance sites should be
viewed with caution.
(a) Change in soil organic carbon (SOC) residence time (RT) averaged
across all nine models. Change is significant for 46 % of the region,
predominantly negative changes (decreases). (b) CV for RT as estimated from
the across-model mean and standard deviation at each grid.
Model uncertainties contributing to errors in net CO2 sink/source activity
Regionally averaged GPP is within 20 % of the MOD17 average
(470 g C m-2 yr-1) for 8 of the 9 models. While the models broadly capture the three
major biomes across the region, a wide range in spatial GPP estimates is
evident. This result may reflect differences in model forcings, initial
conditions, parameterization and the dynamic vs static nature of vegetation
and LAI (Table ). While these differences make it
difficult to unambiguously determine the underlying causes for many of the
mismatches, the evaluations, in the context of prior studies, point to
particular biases. The timing of peak summer GPP is generally well captured
in most of the models (Fig. ). Despite the agreement in
peak GPP (and ER) timing, several models overestimate the small source of
CO2 before, and to some degree after, winter dormancy at the Hakasija
sites and Zotino. Overestimates in GPP and ER are more common than
underestimates (Table ). Indeed, all errors are positive for
site Hakasija and five of the seven models show relatively large
overestimates in ER at Zotino. The tendency to overestimate GPP suggests that
parametrizations and process specifications controlling primary production
(e.g. # 1, 2, 3, 4, 6, 8 in Table 2) may require refinement. It should be
noted that large seasonal flux errors (e.g.
) will appear as more modest
monthly errors such as those noted in our analysis. While it is not possible
to evaluate sources of error separately for Rh and autotrophic respiration
(Ra), our results and those from prior studies implicating Rh in the
model uncertainties suggest a need for further
investigation of model processes controlling respiration. Only one of the
nine models, the CLM4.5, simulated limits on productivity due to nitrogen
availability. None account for competition for nitrogen. Lack of accounting
for nitrogen limits on photosynthesis may be leading to overestimates in
simulated GPP, since nitrogen availability limits terrestrial carbon
sequestration in boreal regions . While accounting for fire
is important for estimates of impacts on recently disturbed areas, and may be
contributing to the wide range in GPP exhibited by CLM4.5, CoLM, and LPJG
(Fig. ), climate variability is a more dominant
influence on regional fluxes . Regarding errors in respiration
rates, models with the highest soil carbon amounts (CLM4.5 and UW-VIC)
exhibit relatively high ER rates when compared to the observations at several
sites (Fig. ). This tendency is consistent with results
described by , who suggest that initial carbon pool size
is the main driver of the response to warming, with the magnitude of the
carbon pool strongly controlling the sensitivity of Rh to changes in
temperature and moisture. While all of the models incorporate temperature and
moisture in their formulations for Rh, only three of the nine account for
the effect of vegetation type on soil thermal dynamics. A wide range in
process specifications for soil thermal dynamics is present across the
models.
In a study of nine models from the TRENDY project, found that
the models overestimate both GPP and ER, and underestimate NEE at most of the
flux sites examined, and for the Northern Hemisphere based on upscaled
measurements. A low NEE, or NEP, may be attributable to model biases in
respiration exceeding those in productivity. Averaged across the nine models
and the region of the present study, NEP of approximately 20 g C m-2 yr-1
(Fig. ) (270 Tg C yr-1) is broadly
consistent with inventory assessments for Eurasian forests, which range
between 93 and 347 Tg C yr-1.
concluded that NPP simulated by two DGVMs examined was nearly balanced by the
models' estimate of Rh. found that GPP increased
during the years 1920 to 2008, with the GPP increase in the DGVMs balanced equally by
increases in respiration. They reported NEP over the Russian territory as an
average of three methods at nearly 30 g C m-2 yr-1. The DGVM
average, however, was only 4.4 g C m-2 yr-1 and so low that the
authors chose to remove it from their final carbon budget. This underestimate
was attributed to an excess in Rh. While the mean NEP of
20 g C m-2 yr-1 in the present study is more consistent with the three-method
average of than their lower DGVM estimates, our comparisons
against tower-based data and results of other studies suggest the sink
strength is underestimated. Of the three models common to that study and the
present one, the CLM4.5 and ORCHIDEE rank on the low end of model NEP
magnitudes (Fig. ).
Recent research points to phenology as one of the principle sources of error
in model simulations of land-atmosphere exchanges of CO2.
found that the change in NEP simulated by a set of CMIP5
models could not account for the observed increase in the seasonal cycle
amplitude in atmospheric CO2 concentrations. They point to data showing
that boreal regions have experienced greening and shifting age composition
which strongly influence NEP and suggest that process models under-represent
the observed changes. Model inability to capture canopy phenology has been
identified as a major source of model uncertainty leading to large seasonal
errors in carbon fluxes such as GPP
. Indeed, evaluated against flux
tower data across the eastern USA, current state-of-the-art terrestrial
biosphere models have been found to mis-characterize the temperature
sensitivity of phenology, which contributes to poor model performance
. Two recent studies using eight land surface models from
the TRENDY comparison (several examined in the present
study) and 11 coupled carbon-climate models have found that
models consistently overestimate leaf area index (LAI) and have a longer
growing season, mostly due to a later autumn dormancy, compared to satellite
data. However, when estimated using model GPP, dormancy was much earlier than
previously predicted using LAI. The authors conclude that the models are
keeping inactive leaves for longer than they should, but with little impact
on carbon cycle fluxes. further suggested that it was
unlikely that differences in climate in the coupled models were solely
responsible for the positive bias. also concluded that
variability in land model fluxes was driven primarily by differences in
model physics rather than differences in forcing data.
Simulated Rh estimates among the DGVMs analyzed by vary
in the range between 200 to 225 g C m-2 yr-1. In the present study
the nine model average is 190 g C m-2 yr-1. point
to lower estimates from of 139 g C m-2 yr-1
and of 174 g C m-2 yr-1 as being more
representative for the region. Our benchmark comparisons of ER against
tower-based data are consistent with these recent studies and suggest that
several models are overestimating Rh, particularly over the boreal forest
zone. Among the model examined in this study a wide range in soil carbon
parameterizations is noted (Table ). Not surprisingly the
effects of active layer depth on the availability of soil organic carbon for
decomposition and combustion has been recognized as a key sensitivity in
process models . Regarding below-ground processes, model
parameterizations and processes controlling carbon storage and turnover such
as litter decomposition rates and biological activity in frozen soils
require close examination as well. Model simulations of
Rh during the non-growing season are sensitive to the presence or absence
of snow , suggesting that future studies of mechanisms
controlling winter CO2 emissions in tundra may help resolve uncertainties
in processes within land surface models and provide a means to connect a
warming climate with vegetation changes, permafrost thaw and CO2 dynamics.
Uncertainties in temporal trend estimates
Uncertainties exist as to whether tundra areas are presently a net sink or
source of CO2. Across tundra regions, process models indicate a stronger
sink in the 2000s compared with the 1990s, attributable to a greater increase
in vegetation net primary production than heterotrophic respiration in
response to warming (. The spatial pattern in
multimodel mean NEP in this study points to small areas in Scandinavia
(< 1 % of the domain) as sources of CO2. Broadly, areas classified as tundra
are a modest CO2 sink of approximately 15 g C m-2 yr-1.
Across-model standard deviations in areas of small positive and negative NEP
are a factor of ten or more greater than the multimodel mean in some areas,
and are generally high across the tundra (Fig. b).
Estimates of NEP sink magnitudes must be interpreted with caution given that
the models in general possess inadequate representation of disturbances which
are an important component of the overall carbon balance .
Among this model group, four of the nine account for fire. The nature of
model initialization and spinup is also a strong influence on simulated NEP
changes. For example, spin-up procedures can explain some of the
discrepancies. ISBA, for instance, was equilibrated using the 10 coldest
years of the WATCH forcing repeatedly to emulate preindustrial climate. As a
result, soil and vegetation carbon were fairly low at the beginning of the
20th century run, much lower than the equilibrium that would result from the
1960s climate. Due to the large characteristic timescale of soil carbon,
part of ISBA's large trend during the 1960–2009 period
(Fig. ) can be traced to the climate used for the
model spinup procedure.
Previous studies have pointed to changes in the seasonal drawdown and release
of CO2 across the northern high latitudes . A change in
the seasonal cycle of GPP and ER is also noted (figure not shown), with the
models analyzed in this study simulating a relatively higher productivity
rate from late spring to mid-summer. Indeed, increased productivity did not
occur uniformly across the growing season, as most of the models show little
change in August or September NEP over time. The models also simulate little
change in NEP over the cold season. Greater productivity in spring and early
summer may be due in part to earlier spring thawing and temporal advance in
growing season initiation , whereas GPP and NEP are more
strongly constrained by moisture limitations later in the growing season
. Extension of the growing season is therefore attributed more
to a regional warming driven advance in spring thaw than a delay in autumn
freeze-up which correlates with
regional annual evapotranspiration for the region above 40∘ N
. There are, however, signs of a delay in the timing of the
fall freeze (-5.4 days decade-1) across Eurasia over the period
1988–2002 consistent with fall satellite snow cover (SCE)
increases, and attributed to greater fall/winter snowfall and regional
cooling . Consistent with the advance in spring thaw, the
models examined here show a greater NEP increase in spring compared to
autumn.
Soil carbon storage across the region increased significantly over the study
period in eight of the nine models. A relatively larger increase in Rh is
correlated strongly with the associated decline in soil carbon residence
time. This suggests that amid recent warming, vegetation carbon inputs to the
soil were greater than the enhancement in decomposition. In a recent study
involving CMIP5 models, found that while the coupled
climate/carbon-cycle models reproduce the latitudinal patterns of carbon
turnover times, differences between the models of more than one order of
magnitude were also noted. The authors suggest that more accurate
descriptions of hydrological processes and water–carbon interactions are
needed to improve the model estimates of ecosystem carbon turnover times. The
reduction in soil carbon residence time may at least partially be a direct
response to increasing NEP, rather than through warming effects on
respiration. A recent study using a set of simulations from
five CMIP5 models found that, because heterotrophic respiration equilibrates
faster to the increasing NPP than the soil carbon stocks, increased
productivity leads to reductions in inferred residence times even when there
are no changes to the environmental controls on decomposition rates, a
process they refer to as false priming. Because the experimental protocol
analyzed here does not include a fixed-climate simulation, it is not possible
to unambiguously separate the contribution from the false priming effect from
that due to warming-related respiration increases, but the fact that soil C
stocks increase over the period of simulation suggests that it is the
dominant effect. Apart from climatological factors, vegetation growth is also
dependent on biological nitrogen availability. Failure to account for
nitrogen limitation may thus impart a bias in the modeled carbon flux
estimates. However, more process models are incorporating linkages between
carbon and nitrogen dynamics . Given the broad range in
spatial patterns in GPP across the models, a closer examination of processes
related to nitrogen limitations and primary production is needed. The lower
rate of NEP increase over the latter decades of the simulation period
suggests a weakening of the land CO2 sink, driven by increased Rh from
warming, associated permafrost thaw, and an upward trend in fire emissions
.
As the climate warms, the amount of carbon emitted as CH4 and CO2 will
depend on whether soils become wetter or drier. A synthesis of observations
and models points to intensification of the pan-Arctic hydrological cycle
over recent decades , manifested prominently by increasing
river discharge from Northern Eurasia . In addition to
hydrological cycle intensification and deepening soil active layer
, rapid thaw and ground collapse will also likely alter
the landscape and impact land-atmosphere carbon exchanges. Land surface
models are now beginning to implement new process formulations to account for
these fine scale perturbations. Several of the models examined in this study
incorporate the effect of soil freeze-thaw state on decomposition of organic
carbon (Table ). Only four of the nine models, however,
account for methane emissions. Six simulate talik formation, and among these
a variety of approaches are employed to compute snow insulation type.
Conclusions
Outputs from a suite of land surface models were evaluated against
independent data sets and used to investigate elements of the land-atmosphere
exchange of CO2 across Northern Eurasia over the period 1960–2009. The
models exhibit a wide range in spatial patterns and regional mean magnitudes.
Compared to tower-based data, overestimates in both GPP and ER are noted in
several of the models, with larger errors in ER relative to GPP, particularly
for the comparisons at the southern higher productivity sites. Regarding
agreement in the spatial pattern in GPP, less than half of the variance in
GPP expressed in the MOD17 product is explained by the GPP pattern from four
of the nine models. The NEP increases range from 3 to 340 % of the
model means, further illustrating uncertainties in sink strength.
The models exhibit a decrease in
residence time of the soil carbon pool that is driven by an increase in
Rh, simultaneous with an increase in soil carbon storage. This result
suggests that net primary productivity (NPP) inputs to the pool increased
more than Rh fluxes out. Among the quantities examined, uncertainties are
lowest for GPP across the forest/taiga biome and highest for residence time
over tundra and steppe areas. Amid the uncertainty in NEP magnitude, the
results of this study and others suggests that the CO2 sink of the region
is underestimated.
Several recommendations are made as a result of this analysis. The range in
area and climatological mean NEP across the models, more than double the mean
value, illustrates the considerable uncertainty in the magnitude of the
contemporary CO2 sink. The results of the site-level comparison point to a
need to better understand the connections between model-simulated
productivity rates, soil dynamics controlling heterotrophic respiration
rates, and associated uncertainties in total ER. Given the strong connections
between soil thermal and hydrological variations and soil respiration, we
recommend that model improvements are targeted at processes and
parameterizations controlling respiration with depth in the soil profile.
These validation efforts are especially important given the likelihood of net
carbon transfer from ecosystems to the atmosphere from permafrost thaw
. Model responses to CO2 fertilization and
nitrogen limitation, processes largely underrepresented in the models, should
be evaluated in the context of ecosystem productivity. While insights have
been gained by examining the model estimates of GPP, ER, and NEP, an improved
understanding of net CO2 sink/source dynamics will require the continued
development and application of model formulations for carbon emissions from
fire and other disturbances. The limited number of measured site data across
this important region clearly hampers model assessments, highlighting the
critical need for new field, tower, and aircraft data for model validation
and parametrization. Specifically, new observations in the boreal zone are
required to better evaluate model biases documented in this and in other
recent studies. Moreover, our finding of biases in CO2 source activity
during the shoulder seasons points to a critical need for observations during
autumn, winter, and spring. Given our results, conclusions drawn from studies
which use a single model should be viewed cautiously in the absence of
rigorous validation against observations across the region of interest.
New observations from current and upcoming field campaigns such as Carbon in
Arctic Reservoirs Vulnerability Experiment (CARVE) and the Arctic Boreal
Vulnerability Experiment (ABoVE) should be used to confirm the results of
this study. Future model evaluations will benefit from continued development
of consistent benchmarking data sets from field measurements and remote
sensing. Regarding tower data, any new measurements must be supported by
refinements in the models used to partition the measured NEE flux into GPP
and ER components. Regarding these and similar model intercomparisons,
investments must be made which will minimize or eliminate differences in a
priori climate forcings used in the simulations. At a programmatic level
support for these activities should lead to well-designed model
intercomparisons which minimize, to the extent possible, differences in model
spinup, forcings and other elements which confound model intercomparisons.
M. A. Rawlins conceived the study with input from A. D. McGuire,
J. K. Kimball and P. Dass. Co-authors D. Lawrence, E. Burke, X. Chen, C. Delire,
C. Koven, A. MacDougall, S. Peng, A. Rinke, K. Saito, W. Zhang, R. Alkama, T. J. Bohn,
P. Ciais, B. Decharme, I. Gouttevin, T. Hajima, D. Ji, G. Krinner,
D. P. Lettenmaier, P. Miller, J. C. Moore, B. Smith, and T. Sueyoshi provided
model simulation outputs. M. A. Rawlins analyzed the outputs and other data.
M. A. Rawlins prepared the manuscript with contributions from all co-authors.
Acknowledgements
This research was supported by the US National Aeronautics and Space
Administration NASA grant NNX11AR16G and the Permafrost Carbon Network
(http://www.permafrostcarbon.org/) funded by the National Science
Foundation. The MODIS Land Cover Type product data was obtained through the
online Data Pool at the NASA Land Processes Distributed Active Archive Center
(LP DAAC), USGS/Earth Resources Observation and Science (EROS) Center, Sioux
Falls, South Dakota (https://lpdaac.usgs.gov/data_access). We thank Hans
Dolman and a second reviewer for their insightful comments which helped
improve the manuscript. We thank the researchers working at FLUXNET sites for
making available their CO2 flux data. We also thank Eugenie Euskirchen and
Dan Hayes for comments on an earlier version of the manuscript, and Yonghong
Yi for assistance with the FLUXNET data. Charles Koven was supported by the
Director of the Office of Biological and Environmental Research, Office of
Science, US Department of Energy, under Contract DE-AC02-05CH11231 as part
of the Regional and Global Climate Modeling Program (RGCM). Eleanor J. Burke
was supported by the Joint UK DECC/Defra Met Office Hadley Centre Climate
Programme (GA01101) and the European Union Seventh Framework Programme
(FP7/2007-2013) under grant agreement no. 282700. Bertrand Decharme and
Christine Delire were supported by the French Agence Nationale de la
Recherche under agreement ANR-10-CEPL-012-03. Several of the authors were
funded by the European Union 7th Framework Programme under
project Page21 (grant 282700). Any use of trade, firm, or product names is
for descriptive purposes only and does not imply endorsement by the US
Government.
Edited by: U. Seibt
ReferencesAdam, J. C. and Lettenmaier, D. P.: Adjustment of global gridded
precipitation for systematic bias, J. Geophys. Res.-Atmos., 108, 4257, 10.1029/2002JD002499, 2003.Adam, J. C., Clark, E. A., Lettenmaier, D. P., and Wood, E. F.: Correction of
global precipitation products for orographic effects, J. Climate, 19,
15–38, 10.1175/JCLI3604.1, 2006.
Anav, A., Murray-Tortarolo, G., Friedlingstein, P., Sitch, S., Piao, S., and
Zhu, Z.: Evaluation of land surface models in reproducing satellite Derived
leaf area index over the high-latitude northern hemisphere. Part II: Earth
system models, Remote Sensing, 5, 3637–3661, 2013.
Avis, C. A., Weaver, A. J., and Meissner, K. J.: Reduction in areal extent
of
high-latitude wetlands in response to permafrost thaw, Nat. Geosci., 4,
444–448, 2011.
Baldocchi, D.: TURNER REVIEW No. 15. 'Breathing' of the terrestrial biosphere:
lessons learned from a global network of carbon dioxide flux measurement
systems, Aust. J. Bot., 56, 1–26, 2008.
Baldocchi, D. D.: Assessing the eddy covariance technique for evaluating
carbon dioxide exchange rates of ecosystems: past, present and future,
Glob. Change Biol., 9, 479–492, 2003.Belshe, E., Schuur, E., and Bolker, B.: Tundra ecosystems observed to be
CO2
sources due to differential amplification of the carbon cycle, Ecol.
Lett., 16, 1307–1315, 2013.Best, M. J., Pryor, M., Clark, D. B., Rooney, G. G., Essery, R .L. H.,
Ménard, C. B., Edwards, J. M., Hendry, M. A., Porson, A., Gedney, N.,
Mercado, L. M., Sitch, S., Blyth, E., Boucher, O., Cox, P. M., Grimmond, C.
S. B., and Harding, R. J.: The Joint UK Land Environment Simulator (JULES),
model description – Part 1: Energy and water fluxes, Geosci. Model Dev., 4,
677–699, 10.5194/gmd-4-677-2011, 2011.Bohn, T. J., Podest, E., Schroeder, R., Pinto, N., McDonald, K. C., Glagolev,
M., Filippov, I., Maksyutov, S., Heimann, M., Chen, X., and Lettenmaier, D.
P.: Modeling the large-scale effects of surface moisture heterogeneity on
wetland carbon fluxes in the West Siberian Lowland, Biogeosciences, 10,
6559–6576, 10.5194/bg-10-6559-2013, 2013.
Bond-Lamberty, B., Wang, C., and Gower, S. T.: A global relationship between
the heterotrophic and autotrophic components of soil respiration?, Glob.
Change Biol., 10, 1756–1766, 2004.Carvalhais, N., Forkel, M., Khomik, M., Bellarby, J., Jung, M., Migliavacca,
M., Mu, M., Saatchi, S., Santoro, M., Thurner, M., Weber, U., Ahrens, B.,
Beer, C., Cescatti, A., Randerson, J. T., and Reichstein, M.: Global
covariation of carbon turnover times with climate in terrestrial ecosystems,
Nature, 514, 213–217, 10.1038/nature13731, 2014.Chapin III, F. S., Sturm, M., Serreze, M. C., McFadden, J. P., Key,
J. R., Lloyd, A. H., McGuire, A. D., Rupp, T. S., Lynch, A. H., Schimel, J. P.,
Beringer, J., Chapman, W. L., Epstein, H. E., Euskirchen, E. S., Hinzman,
L. D., Jia, G., Ping, C.-L., Tape, K. D., Thompson, C. D. C., Walker, D. A.,
and Welker, J. M.: Role of Land-Surface Changes in Arctic Summer Warming,
Science, 310, 657–660, 10.1126/science.1117368, 2005.Clark, D. B., Mercado, L. M., Sitch, S., Jones, C. D., Gedney, N., Best, M.
J., Pryor, M., Rooney, G. G., Essery, R. L. H., Blyth, E., Boucher, O.,
Harding, R. J., Huntingford, C., and Cox, P. M.: The Joint UK Land
Environment Simulator (JULES), model description – Part 2: Carbon fluxes and
vegetation dynamics, Geosci. Model Dev., 4, 701–722,
10.5194/gmd-4-701-2011, 2011.Cohen, J. L., Furtado, J. C., Barlow, M. A., Alexeev, V. A., and Cherry,
J. E.: Arctic warming, increasing snow cover and widespread boreal winter cooling,
Environ. Res. Lett., 7, 014007, 10.1088/1748-9326/7/1/014007, 2012.Crutzen, P. J.: The “Anthropocene”, in: Earth System Science in the
Anthropocene, edited by Ehlers, P. D. E. and Krafft, D. T., 13–18,
Springer, Berlin, Heidelberg,
10.1007/3-540-26590-2_3,
2006.Decharme, B., Boone, A., Delire, C., and Noilhan, J.: Local evaluation of
the Interaction between Soil Biosphere Atmosphere soil multilayer diffusion
scheme using four pedotransfer functions, J. Geophys. Res.-Atmos., 116, D20126, 10.1029/2011JD016002, 2011.Dolman, A. J., Shvidenko, A., Schepaschenko, D., Ciais, P., Tchebakova, N.,
Chen, T., van der Molen, M. K., Belelli Marchesini, L., Maximov, T. C.,
Maksyutov, S., and Schulze, E.-D.: An estimate of the terrestrial carbon
budget of Russia using inventory-based, eddy covariance and inversion
methods, Biogeosciences, 9, 5323–5340, 10.5194/bg-9-5323-2012, 2012.Dutta, K., Schuur, E. A. G., Neff, J. C., and Zimov, S. A.: Potential carbon
release from permafrost soils of Northeastern Siberia, Glob. Change
Biol., 12, 2336–2351, 10.1111/j.1365-2486.2006.01259.x, 2006.
Euskirchen, E., McGuire, A. D., Kicklighter, D. W., Zhuang, Q., Clein, J. S.,
Dargaville, R., Dye, D., Kimball, J. S., McDonald, K. C., Melillo, J. M.,
Romanovsky, V. E., and Smith, N. V.: Importance of recent shifts in soil thermal dynamics on growing
season length, productivity, and carbon sequestration in terrestrial
high-latitude ecosystems, Glob. Change Biol., 12, 731–750, 2006.Exbrayat, J.-F., Pitman, A. J., Zhang, Q., Abramowitz, G., and Wang, Y.-P.:
Examining soil carbon uncertainty in a global model: response of microbial
decomposition to temperature, moisture and nutrient limitation,
Biogeosciences, 10, 7095–7108, 10.5194/bg-10-7095-2013, 2013.Fisher, J. B., Sikka, M., Oechel, W. C., Huntzinger, D. N., Melton, J. R.,
Koven, C. D., Ahlström, A., Arain, M. A., Baker, I., Chen, J. M., Ciais,
P., Davidson, C., Dietze, M., El-Masri, B., Hayes, D., Huntingford, C., Jain,
A. K., Levy, P. E., Lomas, M. R., Poulter, B., Price, D., Sahoo, A. K.,
Schaefer, K., Tian, H., Tomelleri, E., Verbeeck, H., Viovy, N., Wania, R.,
Zeng, N., and Miller, C. E.: Carbon cycle uncertainty in the Alaskan Arctic,
Biogeosciences, 11, 4271–4288, 10.5194/bg-11-4271-2014,
2014.
Goetz, S. J., Bunn, A. G., Fiske, G. J., and Houghton, R.:
Satellite-observed
photosynthetic trends across boreal North America associated with climate and
fire disturbance, P. Natl. Acad. Sci. USA, 102, 13521–13525, 2005.
Goetz, S. J., Fiske, G. J., and Bunn, A. G.: Using satellite time-series
data sets to analyze fire disturbance and forest recovery across Canada,
Remote Sens. Environ., 101, 352–365, 2006.Gouttevin, I., Menegoz, M., Dominé, F., Krinner, G., Koven, C., Ciais,
P., Tarnocai, C., and Boike, J.: How the insulating properties of snow affect
soil carbon distribution in the continental pan-Arctic area, J.
Geophys. Res.-Biogeo., 117, G02020, 10.1029/2011JG001916, 2012.Graven, H., Keeling, R., Piper, S., Patra, P., Stephens, B., Wofsy, S., Welp,
L., Sweeney, C., Tans, P., Kelley, J., Daube, B. C., Kort, E. A., Santoni, G. W., and Bent, J. D.:
Enhanced Seasonal Exchange of CO2 by Northern Ecosystems Since 1960, Science, 341, 1085–1089, 2013.Groisman, P. and Soja, A. J.: Ongoing climatic change in Northern Eurasia:
justification for expedient research, Environ. Res. Lett., 4,
045002, 10.1088/1748-9326/4/4/045002, 2009.
Groisman, P. Y. and Bartalev, S. A.: Northern Eurasia Earth Science
Partnership Initiative (NEESPI), science plan overview, Global
Planet. Change, 56, 215–234, 2007.
Groisman, P. Y., Clark, E. A., Lettenmaier, D. P., Kattsov, V. M., Sokolik,
I. N., Aizen, V. B., Cartus, O., Chen, J., Schmullius, C. C., Conard, S.,
Katzenberger, J., Krankina, O., Kukkonen, J., Sofiev, M. A., Machida, T., Maksyutov, S.,
Ojima, D., Qi, J., Romanovsky, V. E., Walker, D., Santoro, M., Shiklomanov, A. I.,
”Vörösmarty, C., Shimoyama, K., Shugart, H. H., Shuman, J. K., Sukhinin, A. I.,
and Wood, E. F.: The Northern Eurasia earth science partnership: an example of
science applied to societal needs, B. Am. Meteorol.
Soc., 90, 671–688, 2009.
Harris, I., Jones, P., Osborn, T., and Lister, D.: Updated high-resolution
grids of monthly climatic observations–the CRU TS3. 10 Dataset,
Int. J. Climatol., 34, 623–642, 2014.
Hayes, D. and Turner, D.: The need for “apples-to-apples” comparisons of
carbon dioxide source and sink estimates, Eos Trans. AGU, 93, 404–405, 2012.Hayes, D. J., McGuire, A. D., Kicklighter, D. W., Gurney, K., Burnside, T.,
and
Melillo, J. M.: Is the northern high-latitude land-based CO2 sink
weakening?, Global Biogeochem. Cy., 25, GB3018,
10.1029/2010GB003813, 2011.Hayes, D. J., Kicklighter, D. W., McGuire, A. D., Chen, M., Zhuang, Q., Yuan,
F., Melillo, J. M., and Wullschleger, S. D.: The impacts of recent
permafrost thaw on land–atmosphere greenhouse gas exchange, Environ.
Res. Lett., 9, 045005, 10.1088/1748-9326/9/4/045005, 2014.
Heinsch, F. A., Zhao, M., Running, S. W., Kimball, J. S., Nemani, R. R.,
Davis, K. J., Bolstad, P. V., Cook, B. D., Desai, A. R., Ricciuto, D. M.,
Law, B. E., Oechel, W. C., Hyojung, K., Hongyan, L., Wofsy, S. C., Dunn, A. L.,
Munger, J. W., Baldocchi, D. D., Liukang, X., Hollinger, D. Y., Richardson, A. D.,
Stoy, P. C., Siqueira, M. B. S., Monson, R. K., Burns, S. P., Flanagan, and L. B.:
Evaluation of remote sensing based terrestrial productivity from MODIS using
regional tower eddy flux network observations, IEEE T. Geosci. Remote,
44, 1908–1925, 2006.Hobbie, S. E., Schimel, J. P., Trumbore, S. E., and Randerson, J. R.:
Controls
over carbon storage and turnover in high-latitude soils, Glob. Change
Biol., 6, 196–210, 10.1046/j.1365-2486.2000.06021.x,
2000.Högberg, P., Nordgren, A., Buchmann, N., Taylor, A. F. S., Ekblad, A.,
Högberg, M. N., Nyberg, G., Ottosson-Löfvenius, M., and Read, D. J.:
Large-scale forest girdling shows that current photosynthesis drives soil
respiration, Nature, 411, 789–792, 10.1038/35081058,
2001.International Permafrost Association Standing Committee on Data Information
and Communication (comp.): Circumpolar Active-Layer Permafrost System
(CAPS), Boulder, Colorado USA, National Snow and Ice Data Center,
10.7265/N5SF2T3B, (last access: 08 July 2014), 2003.Ji, D., Wang, L., Feng, J., Wu, Q., Cheng, H., Zhang, Q., Yang, J., Dong, W.,
Dai, Y., Gong, D., Zhang, R.-H., Wang, X., Liu, J., Moore, J. C., Chen, D.,
and Zhou, M.: Description and basic evaluation of Beijing Normal University
Earth System Model (BNU-ESM) version 1, Geosci. Model Dev., 7, 2039–2064,
10.5194/gmd-7-2039-2014, 2014.Kalnay, E., Cai, M., Li, H., and Tobin, J.: Estimation of the impact of
land-surface forcings on temperature trends in eastern United States, J.
Geophys. Res.-Atmos., 111, D06106, 10.1029/2005JD006555, 2006.Keenan, T., Baker, I., Barr, A., Ciais, P., Davis, K., Dietze, M., Dragoni,
D., Gough, C. M., Grant, R., Hollinger, D., Hufkens, K., Poulter, B., McCaughey, H.,
Raczka, B., Ryu, Y., Schaefer, K., Tian, H., Verbeeck, H., Zhao, M., and Richardson, A. D:
Terrestrial biosphere model
performance for inter-annual variability of land-atmosphere CO2 exchange,
Glob. Change Biol., 18, 1971–1987, 2012.Keenan, T. F., Gray, J., Friedl, M. A., Toomey, M., Bohrer, G., Hollinger,
D. Y., Munger, J. W., O'Keefe, J., Schmid, H. P., Wing, I. S., Yang, B.,
and Richardson, A. D.:
Net carbon uptake has increased through warming-induced changes in temperate
forest phenology, Nature Climate Change, 4, 598–604, 10.1038/nclimate2253, 2014.
Kim, Y., Kimball, J., Zhang, K., and McDonald, K.: Satellite detection of
increasing Northern Hemisphere non-frozen seasons from 1979 to 2008:
Implications for regional vegetation growth, Remote Sens. Environ.,
121, 472–487, 2012.
Kimball, J., McDonald, K., and Zhao, M.: Spring thaw and its effect on
terrestrial vegetation productivity in the western Arctic observed from
satellite microwave and optical remote sensing, Earth Interact., 10,
1–22, 2006.Koven, C., Friedlingstein, P., Ciais, P., Khvorostyanov, D., Krinner, G., and
Tarnocai, C.: On the formation of high-latitude soil carbon stocks: Effects
of cryoturbation and insulation by organic matter in a land surface model,
Geophys. Res. Lett., 36, L21501, 10.1029/2009GL040150, 2009.
Koven, C. D., Ringeval, B., Friedlingstein, P., Ciais, P., Cadule, P.,
Khvorostyanov, D., Krinner, G., and Tarnocai, C.: Permafrost carbon-climate
feedbacks accelerate global warming, P. Natl. Acad. Sci. USA, 108, 14769–14774, 2011.Koven, C. D., Chambers, J. Q., Georgiou, K., Knox, R., Negron-Juarez, R.,
Riley, W. J., Arora, V. K., Brovkin, V., Friedlingstein, P., and Jones, C.
D.: Controls on terrestrial carbon feedbacks by productivity vs. turnover in
the CMIP5 Earth System Models, Biogeosciences Discuss., 12, 5757–5801,
10.5194/bgd-12-5757-2015, 2015.
Kurganova, I. and Nilsson, S.: Carbon dioxide emission from soils of Russian
terrestrial ecosystems, Interim Report, IR-02, 70, 2003.
Lasslop, G., Reichstein, M., Papale, D., Richardson, A. D., Arneth, A., Barr,
A., Stoy, P., and Wohlfahrt, G.: Separation of net ecosystem exchange into
assimilation and respiration using a light response curve approach: critical
issues and global evaluation, Glob. Change Biol., 16, 187–208, 2010.MacDougall, A. H., Eby, M., and Weaver, A. J.: If Anthropogenic CO2
Emissions
Cease, Will Atmospheric CO2 Concentration Continue to Increase?, J.
Climate, 26, 9563–9576, 2013.McDonald, K. C., Kimball, J. S., Njoku, E., Zimmermann, R., and Zhao, M.:
Variability in Springtime Thaw in the Terrestrial High Latitudes: Monitoring
a Major Control on Biospheric Assimilation of Atmospheric CO2 with
Spaceborne Microwave Remote Sensing, Earth Interact., 8, 1–23, 2004.McGuire, A., Melillo, J., Randerson, J., Parton, W., Heimann, M., Meier, R.,
Clein, J., Kicklighter, D., and Sauf, W.: Modeling the effects of snowpack
on heterotrophic respiration across northern temperate and high latitude
regions: Comparison with measurements of atmospheric carbon dioxide in high
latitudes, Biogeochemistry, 48, 91–114, 10.1023/A:1006286804351,
2000.McGuire, A. D., Christensen, T. R., Hayes, D., Heroult, A., Euskirchen, E.,
Kimball, J. S., Koven, C., Lafleur, P., Miller, P. A., Oechel, W., Peylin,
P., Williams, M., and Yi, Y.: An assessment of the carbon balance of Arctic
tundra: comparisons among observations, process models, and atmospheric
inversions, Biogeosciences, 9, 3185–3204, 10.5194/bg-9-3185-2012,
2012.
Melillo, J. M., McGuire, A. D., Kicklighter, D. W., Moore, B., Vorosmarty,
C. J., and Schloss, A. L.: Global climate change and terrestrial net primary
production, Nature, 363, 234–240, 1993.
Miller, P. A. and Smith, B.: Modelling tundra vegetation response to recent
arctic warming, Ambio, 41, 281–291, 2012.
Mitchell, T. D. and Jones, P. D.: An improved method of constructing a
database
of monthly climate observations and associated high-resolution grids,
Int. J. Climatol., 25, 693–712, 2005.
Murray-Tortarolo, G., Anav, A., Friedlingstein, P., Sitch, S., Piao, S., Zhu,
Z., Poulter, B., Zaehle, S., Ahlström, A., Lomas, M., Levis, S., Viovy, N., and Zeng, N.: Evaluation
of Land Surface Models in Reproducing Satellite-Derived LAI over the
High-Latitude Northern Hemisphere. Part I: Uncoupled DGVMs, Remote Sensing,
5, 4819–4838, 2013.Myneni, R. B., Keeling, C. D., Tucker, C. J., Asrar, G., and Nemani, R. R.:
Increased plant growth in the northern high latitudes from 1981 to 1991,
Nature, 386, 698–702, 10.1038/386698a0, 1997.
Nash, J. and Sutcliffe, J. V.: River flow forecasting through conceptual
models part – A discussion of principles, J. Hydrol., 10, 282–290,
1970.Oak Ridge National Laboratory: MODIS (MCD12Q1) Land Cover Product,
Land Processes
Distributed Active Archive Center (LP DAAC), available at: https://lpdaac.usgs.gov (last access: 01 July 2014),
Sioux Falls, South Dakota, USA, 2014.Oechel, W. C., Laskowski, C. A., Burba, G., Gioli, B., and Kalhori, A. A. M.:
Annual patterns and budget of CO2 flux in an Arctic tussock tundra
ecosystem, J. Geophys. Res.-Biogeo., 119, 323–339,
10.1002/2013JG002431, 2014.Oleson, K., Lawrence, D., Bonan, G., Drewniak, B., Huang, M., Koven, C.,
Levis,
S., Li, F., Riley, W., Subin, Z., Swenson, S., Thornton, P., Bozbiyik, A.,
Fisher, R., Kluzek, E., Lamarque, J., Lawrence, P., Leung, L., Lipscomb, W.,
Muszala, S., Ricciuto, D., Sacks, W., Sun, Y., Tang, J., and Z. L., Y.:
Technical Description of version 4.5 of the Community Land Model (CLM). Ncar
Technical Note NCAR/TN-503+STR, Tech. rep., National Center for Atmospheric
Research, boulder, CO, 422 pp., 10.5065/D6RR1W7M, 2013.Parmentier, F., Van Der Molen, M., Van Huissteden, J., Karsanaev, S.,
Kononov, A., Suzdalov, D., Maximov, T., and Dolman, A.: Longer growing
seasons do not increase net carbon uptake in the northeastern Siberian tundra, J.
Geophys. Res.-Biogeo., 116, G04013, 10.1029/2011JG001653, 2011.Peng, S., Ciais, P., Chevallier, F., Peylin, P., Cadule, P., Sitch, S., Piao,
S., Ahlström, A., Huntingford, C., Levy, P., Li, X., Liu, Y., Lomas, M., Poulter, B.,
Viovy, N., Wang, T., Wang, X., Zaehle, S., Zeng, N., Zhao, F., and Zhao, H.: Benchmarking the
seasonal cycle of CO2 fluxes simulated by terrestrial ecosystem models,
Global Biogeochem. Cy., 29, 46–64, 10.1002/2014GB004931, 2015.Piao, S., Sitch, S., Ciais, P., Friedlingstein, P., Peylin, P., Wang, X.,
Ahlström, A., Anav, A., Canadell, J. G., Cong, N., Huntingford, C., Jung, M.,
Levis, S., Levy, P. E., Li, J., Lin, X., Lomas, M. R., Lu, M., Luo, Y., Ma, Y.,
Myneni, R. B., Poulter, B., Sun, Z., Wang, T., Viovy, N., Zaehle, S., and Zeng, N.: Evaluation of
terrestrial carbon cycle models for their response to climate variability and
to CO2 trends, Glob. Change Biol., 19, 2117–2132, 2013.Pries, C. E. H., Schuur, E. A. G., and Crummer, K. G.: Holocene Carbon
Stocks and Carbon Accumulation Rates Altered in Soils Undergoing Permafrost Thaw,
Ecosystems, 15, 162–173, 10.1007/s10021-011-9500-4,
2012.
Quegan, S., Beer, C., Shvidenko, A., McCallum, I., Handoh, I. C., Peylin, P.,
Roedenbeck, C., Lucht, W., Nilsson, S., and Schmullius, C.: Estimating the
carbon balance of central Siberia using a landscape-ecosystem approach,
atmospheric inversion and Dynamic Global Vegetation Models, Glob. Change
Biol., 17, 351–365, 2011.Peterson, B. J., Holmes, R. M., McClelland, J. W., Vörösmarty, C. J.,
Lammers, R. B., Shiklomanov, A. I., Shiklomanov, I. A., and Rahmstorf, S.:
Increasing river discharge to the Arctic Ocean, Science, 298, 2171–2173,
10.1126/science.1077445,
2002.
Rawlins, M. A., Steele, M., Holland, M. M., Adam, J. C., Cherry, J. E., Francis, J. A., Groisman, P. Y., Hinzman, L. D., Huntington, T. G., Kane, D. L., Kimball, J. S., Kwok, R., Lammers, R. B., Lee, C. M., Lettenmaier, D. P., McDonald, K. C., Podest, E., Pundsack, J. W., Rudels, B., Serreze, M. C., Shiklomanov, A., Skagseth, Ø, Troy, T. J., Vörösmarty, C. J., Wensnahan, M., Wood, E. F., Woodgate, R., Yang, D., Zhang, K., Zhang, T.: Analysis of the Arctic System for Freshwater
Cycle Intensification: Observations and Expectations, J. Climate.,
23, 5715–5737, 2010.
Reichstein, M., Falge, E., Baldocchi, D., Papale, D., Aubinet, M., Berbigier,
P., Bernhofer, C., Buchmann, N., Gilmanov, T., Granier, A., Grünwald, T., Havránková, K., Ilvesniemi, H., Janous, D., Knohl, A., Laurila, T., Lohila, A., Loustau, D., Matteucci, G., Meyers, T., Miglietta, F., Ourcival, J. M., Pumpanen, J., Rambal, S., Rotenberg, E., Sanz, M., Tenhunen, J., Seufert, G., Vaccari, F., Vesala, T., Yakir, D., and Valentini, R.: On the
separation of net ecosystem exchange into assimilation and ecosystem
respiration: review and improved algorithm, Glob. Change Biol., 11,
1424–1439, 2005.Reichstein, M., Papale, D., Valentini, R., Aubinet, M., Bernhofer, C., Knohl,
A., Laurila, T., Lindroth, A., Moors, E., Pilegaard, K., Seufert, G.:
Determinants of terrestrial ecosystem carbon balance inferred from European
eddy covariance flux sites, Geophys. Res. Lett., 34, 1402, 10.1029/2006GL027880, 2007.
Richardson, A. D., Anderson, R. S., Arain, M. A., Barr, A. G., Bohrer, G.,
Chen, G., Chen, J. M., Ciais, P., Davis, K. J., Desai, A. R., Dietze, M. C., Dragoni, D.,
Garrity, S. R., Gough, C. M., Grant, R., Hollinger, D. Y., Margolis, H. A., McCaughey, H.,
Migliavacca, M., Monson, R. K., Munger, J. W., Poulter, B., Raczka, B. M., Ricciuto, D. M.,
Sahoo, A. K., Schaefer, K., Tian, H., Vargas, R., Verbeeck, H., Xiao, J., and Xue, Y.:
Terrestrial biosphere models need better representation of vegetation
phenology: results from the North American Carbon Program Site Synthesis,
Glob. Change Biol., 18, 566–584, 2012.Romanovsky, V. E., Smith, S. L., and Christiansen, H. H.: Permafrost thermal
state in the polar Northern Hemisphere during the international polar year
2007–2009: a synthesis, Permafrost Periglac., 21, 106–116,
10.1002/ppp.689,
2010.
Running, S., Nemani, R., Heinsch, F., Zhao, M., Reeves, M., and Hashimoto,
H.: A continuous satellite-derived measure of global terrestrial primary
production, BioScience, 54, 547–560, 2004.Schaefer, K., Zhang, T., Bruhwiler, L., and Barrett, A.: Amount and timing
of permafrost carbon release in response to climate warming, Tellus B, 63, 165–180,
10.1111/j.1600-0889.2011.00527.x, 2011.Schaefer, K., Schwalm, C. R., Williams, C., Arain, M. A., Barr, A., Chen,
J. M., Davis, K. J., Dimitrov, D., Hilton, T. W., Hollinger, D. Y., Humphreys, E.,
Poulter, B., Raczka, B., Richardson, A. D., Sahoo, A., Thornton, P., Vargas, R.,
Verbeeck, H., Anderson, R., Baker, I., Black, T. A.., Bolstad, P., Chen, J., Curtis, P. S.,
Desai, A., Dietze, M., Dragoni, D., Gough, C., Grant, R., Gu, L., Jain, A., Kucharik, C.,
Law, B., Liu, S., Lokipitiya, E., Margolis, H. A., Matamala, R., McCaughey, J. H., Monson,
R., Munger, J. W., Oechel, W., Peng, C., Price, D. T., Ricciuto, D., Riley, W. J., Roulet, N.,
Tian, H., Tonitto, C., Torn, M., Weng, E., and Zhou, X.:
A model-data comparison of gross primary productivity: Results from the
North American Carbon Program site synthesis, J. Geophys.
Res.-Biogeo., 117, G03010, 10.1029/2012JG001960, 2012.
Schepaschenko, D., Mukhortova, L., Shvidenko, A., and Vedrova, E.: The pool
of organic carbon in the soils of Russia, Eurasian Soil Sci., 46, 107–116,
2013.Schuur, E., Vogel, J. G., Crummer, K. G., Lee, H., Sickman, J. O., and
Osterkamp, T. E.: The effect of permafrost thaw on old carbon release and
net carbon exchange from tundra, Nature, 459, 556–559, 10.1038/nature08031, 2009.Schuur, E. A. G. and Abbott, B.: Climate change: High risk of
permafrost thaw, Nature, 480, 32–33, 10.1038/480032a, 2011.Schuur, E. A. G., McGuire, A. D., Schadel, C., Grosse, G., Harden,
J. W.,
Hayes, D. J., Hugelius, G., Koven, C. D., Kuhry, P., Lawrence, D. M.,
Natali, S. M., Olefeldt, D., Romanovsky, V. E., Schaefer, K.,
Turetsky,
M. R., Treat, C. C., and Vonk, J. E.: Climate change and the permafrost
carbon feedback, Nature, 520, 171–179,
10.1038/nature14338, 2015.
Serreze, M. C. and Barry, R. G.: Processes and impacts of Arctic
amplification: A research synthesis, Global Planet. Change, 77,
85–96, 2011.Serreze, M. C., Barrett, A. P., Slater, A. G., Woodgate, R. A., Aagaard, K.,
Lammers, R. B., Steele, M., Moritz, R., Meredith, M., and Lee, C. M.: The
large-scale freshwater cycle of the Arctic, J. Geophys. Res., 111, C11010,
10.1029/2005JC003424, 2006.
Sheffield, J., Goteti, G., and Wood, E. F.: Development of a 50-Year
High-Resolution Global Dataset of Meteorological Forcings for Land Surface
Modeling, J. Climate, 19, 3088–3111, 2006.
Smith, B., Prentice, I. C., and Sykes, M. T.: Representation of vegetation
dynamics in the modelling of terrestrial ecosystems: comparing two
contrasting approaches within European climate space, Global Ecol.
Biogeogr., 10, 621–637, 2001.Smith, N. V., Saatchi, S. S., and Randerson, J. T.: Trends in high northern
latitude soil freeze and thaw cycles from 1988 to 2002, J.
Geophys. Res.-Atmos., 109, D12101, 10.1029/2003JD004472, 2004.Thornton, P. E., Doney, S. C., Lindsay, K., Moore, J. K., Mahowald, N.,
Randerson, J. T., Fung, I., Lamarque, J.-F., Feddema, J. J., and Lee, Y.-H.:
Carbon-nitrogen interactions regulate climate-carbon cycle feedbacks: results
from an atmosphere-ocean general circulation model, Biogeosciences, 6,
2099–2120, 10.5194/bg-6-2099-2009, 2009.
Turner, D. P., Ritts, W. D., Cohen, W. B., Gower, S. T., Running, S. W.,
Zhao, M., Costa, M. H., Kirschbaum, A. A., Ham, J. M., Saleska, S. R., and Ahl, D. E.:
Evaluation of MODIS NPP and GPP products across multiple biomes, Remote
Sens. Environ., 102, 282–292, 2006.Viovy, N. and Ciais, P.: CRUNCEP data set for 1901–2008, Tech. Rep. Version
4, Tech. rep., Laboratoire des Sciences du Climat et de l'Environnement,
available at: http://dods.extra.cea.fr/data/p529viov/cruncep/readme.htm (last access: 20 July 2014), 2011.Vogel, J., Schuur, E. A. G., Trucco, C., and Lee, H.: Response of CO2
exchange in a tussock tundra ecosystem to permafrost thaw and thermokarst
development, J. Geophys. Res.-Biogeo., 114, G04018,
10.1029/2008JG000901, 2009.Wania, R., Ross, I., and Prentice, I.: Integrating peatlands and permafrost
into a dynamic global vegetation model: 2. Evaluation and sensitivity of
vegetation and carbon cycle processes, Global Biogeochem. Cy., 23, GB3015, 10.1029/2008GB003413,
2009a.Wania, R., Ross, I., and Prentice, I.: Integrating peatlands and permafrost
into a dynamic global vegetation model: 1. Evaluation and sensitivity of
physical land surface processes, Global Biogeochem. Cy., 23, GB3014, 10.1029/2008GB003412,
2009b.Wania, R., Ross, I., and Prentice, I. C.: Implementation and evaluation of a
new methane model within a dynamic global vegetation model: LPJ-WHyMe v1.3.1,
Geosci. Model Dev., 3, 565–584, 10.5194/gmd-3-565-2010, 2010.
Waring, R., Landsberg, J., and Williams, M.: Net primary production of
forests: a constant fraction of gross primary production?, Tree Physiol.,
18, 129–134, 1998.Watanabe, S., Hajima, T., Sudo, K., Nagashima, T., Takemura, T., Okajima, H.,
Nozawa, T., Kawase, H., Abe, M., Yokohata, T., Ise, T., Sato, H., Kato, E.,
Takata, K., Emori, S., and Kawamiya, M.: MIROC-ESM 2010: model description
and basic results of CMIP5-20c3m experiments, Geosci. Model Dev., 4,
845–872, 10.5194/gmd-4-845-2011, 2011.
Weedon, G., Gomes, S., Viterbo, P., Shuttleworth, W., Blyth, E., Österle,
H., Adam, J., Bellouin, N., Boucher, O., and Best, M.: Creation of the WATCH
forcing data and its use to assess global and regional reference crop
evaporation over land during the twentieth century, J.
Hydrometeorol., 12, 823–848, 2011.Willmott, C. J. and Matsuura, K.: Advantages of the mean absolute error (MAE)
over the root mean square error (RMSE) in assessing average model
performance, Clim. Res., 30, 79–82, 10.3354/cr030079, 2005.
Willmott, C. J., Robeson, S. M., and Matsuura, K.: A refined index of model
performance, Int. J. Climatol., 32, 2088–2094, 2011.Yi, Y., Kimball, J. S., Jones, L. A., Reichle, R. H., Nemani, R., and
Margolis, H. A.: Recent climate and fire disturbance impacts on boreal and arctic
ecosystem productivity estimated using a satellite-based terrestrial carbon
flux model, J. Geophys. Res.-Biogeo., 118, 606–622, 10.1002/jgrg.20053, 2013.Yi, Y., Kimball, J. S., and Reichle, R. H.: Spring hydrology determines
summer net carbon uptake in northern ecosystems, Environ. Res. Lett., 9,
064003, 10.1088/1748-9326/9/6/064003, 2014.Zaehle, S.: Terrestrial nitrogen–carbon cycle interactions at the global
scale, Philos. T. Roy. Soc. B, 368, 20130125, 10.1098/rstb.2013.0125,
2013.Zhang, K., Kimball, J. S., Hogg, E. H., Zhao, M., Oechel, W. C., Cassano,
J. J., and Running, S. W.: Satellite-based model detection of recent
climate-driven changes in northern high-latitude vegetation productivity,
J. Geophys. Res., 113, g03033, 10.1029/2007JG000621, 2008.
Zhao, M. and Running, S. W.: Drought-induced reduction in global terrestrial
net primary production from 2000 through 2009, Science, 329, 940–943, 2010.
Zhao, M., Heinsch, F. A., Nemani, R. R., and Running, S. W.: Improvements of
the MODIS terrestrial gross and net primary production global data set,
Remote Sens. Environ., 95, 164–176, 2005.Zhang, X., He, J., Zhang, J., Polaykov, I., Gerdes, R., Inoue, J., and
Wu, P.: Enhanced poleward moisture transport and amplified northern
high-latitude wetting, Nature, 3, 47–51, 10.1038/nclimate1631, 2013.