UC Berkeley
UC Berkeley Previously Published Works
Title
Carbon cycle confidence and uncertainty: Exploring variation among soil biogeochemical
models
Permalink
https://escholarship.org/uc/item/79d8n030
Journal
Global Change Biology, 24(4)
ISSN
1354-1013
Authors
Wieder, William R
Hartman, Melannie D
Sulman, Benjamin N
et al.
Publication Date
2018-04-01
DOI
10.1111/gcb.13979
Peer reviewed
Powered by the California Digital Library
University of California
Carbon cycle confidence and uncertainty: Exploring variation
among soil biogeochemical models
William R. Wieder
1,2
| Melannie D. Hartman
2,3
| Benjamin N. Sulman
4
| Ying-
Ping Wang
5,6
| Charles D. Koven
7
| Gordon B. Bonan
2
1
Institute of Arctic and Alpine Research, University of Colorado, Boulder, CO,
USA
2
Climate and Global Dynamics Laboratory, National Center for
Atmospheric Research, Boulder, CO, USA
3
Natural Resource Ecology
Laboratory, Colorado State University, Fort Collins, CO, USA
4
Program in
Atmospheric and Oceanic Sciences, Princeton University, Princeton, NJ, USA
5
South China Botanical Garden, Chinese Academy of Sciences, Guangzhou,
China
6
CSIRO Oceans and Atmosphere, Aspendale, Vic., Australia
7
Climate
and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory,
Berkeley, CA, USA
Correspondence William R. Wieder, Institute of Arctic and Alpine Research,
University of Colorado, Boulder, CO, USA. Email: wwieder@ucar.edu
Abstract
Emerging insights into factors responsible for soil organic matter stabilization
and decomposition are being applied in a variety of contexts, but new tools
are needed to facilitate the understanding, evaluation, and improvement of
soil biogeochemical theory and models at regional to global scales. To isolate
the effects of model structural uncertainty on the global distribution of soil
carbon stocks and turnover times we developed a soil biogeochemical
testbed that forces three different soil models with consistent climate and
plant productivity inputs. The models tested here include a first‐order,
microbial implicit approach (CASA‐CNP), and two recently developed
microbially explicit models that can be run at global scales (MIMICS and
CORPSE). When forced with common environmental drivers, the soil models
generated similar estimates of initial soil carbon stocks (roughly 1,400 Pg C
globally, 0–100 cm), but each model shows a different functional relationship
between mean annual temperature and inferred turnover times.
Subsequently, the models made divergent projections about the fate of
these soil carbon stocks over the 20
th
century, with models either gaining or
losing over 20 Pg C globally between 1901 and 2010. Single‐forcing
experiments with changed inputs, temperature, and moisture suggest that
uncertainty associated with freeze‐thaw processes as well as soil textural
effects on soil carbon stabilization were larger than direct temperature
uncertainties among models. Finally, the models generated distinct
projections about the timing and magnitude of seasonal heterotrophic
respiration rates, again reflecting structural uncertainties that were related
to environmental sensitivities and assumptions about physicochemical
stabilization of soil organic matter. By providing a computationally tractable
and numerically consistent framework to evaluate models we aim to better
understand uncertainties among models and generate insights about factors
regulating the turnover of soil organic matter.
1 INTRODUCTION
Soils represent the largest terrestrial carbon pool on Earth, storing nearly
five times as much carbon as vegetation (Jobbágy & Jackson, 2000). In the
new millennium, the theoretical understanding of factors responsible for soil
organic matter stabilization has undergone significant revisions (Lehmann &
Kleber, 2015; Schmidt et al., 2011). Driven by new measurements that afford
high resolution information on the chemical and physical nature of soil
organic matter, these emerging theories posit that microbial access to
otherwise decomposable substrates (as opposed to inherent chemical
recalcitrance) governs soil organic matter stabilization and turnover. Such
insights, however, remain poorly represented in global‐scale models that
investigate potential carbon cycle – climate feedbacks (Luo et al., 2016;
Wieder, Allison, et al. 2015), despite an expansion in the number and
diversity of soil biogeochemical models (Manzoni & Porporato, 2009; Sierra,
Müller, & Trumbore, 2012). Building the capacity to test emerging ecological
theories in global‐scale models is critical to informing future research needs,
testing soil biogeochemical theory, refining model features, and accelerating
advancements across scientific disciplines.
Earth system models (ESMs) are typically applied to project potential carbon
cycle – climate interactions and inform policy decisions (Ciais et al., 2013),
but these models also represent a scientific tool to test ecological insight at
larger spatial and longer temporal scales. In global‐scale applications where
ESMs are used to generate numerical projections, soil biogeochemical
models show large variation in estimates of present day soil carbon storage
and widely divergent projections of soil carbon response to environmental
change (Tian et al., 2015; Todd‐Brown et al., 2013). When propagated into
future scenarios, this creates uncertainties in the magnitude of terrestrial
carbon uptake (Anav et al., 2013; Arora et al., 2013; Friedlingstein
et al., 2014; Hoffman et al., 2014), and presents limitations for assessing the
allowable carbon emissions that are compatible with desired climate
outcomes (Jones et al., 2013, 2016; Zhang, Wang, Matear, Pitman, &
Dai, 2014). Troublingly, the soil biogeochemical models of these studies
share a common structure, and thus fail to incorporate process uncertainties
associated with factors regulating soil organic matter stabilization in soils. As
such, they potentially underestimate the true uncertainty associated with soil
carbon responses to environmental perturbations (Bradford, Wieder,
et al., 2016). Moreover, without applying these emerging soil biogeochemical
concepts into global scale models, opportunities to deepen ecological insight
by evaluating and refining theories are not being fully realized.
Building confidence in terrestrial carbon cycle projections, therefore, requires
consideration of the factors controlling the decomposition and formation of
soil organic matter (Bradford, Wieder, et al., 2016). This research priority
requires balancing demands between formulating model structures that
adequately represent theoretical understanding of processes relevant for
long‐term soil organic matter dynamics and avoiding undue complexity (Luo
et al., 2016; Wieder, Allison, et al. 2015). More practically, it requires a
numerically consistent, computationally efficient simulation framework that
can be used to compare and evaluate models at ecosystem‐ to global scales.
Overlying terrestrial models generate additional variation in the
biogeochemical and biophysical state upstream of the soil system—including
uncertainties in climate, hydrology, and plant productivity – and the potential
ecosystem responses of these factors to perturbations (Todd‐Brown
et al., 2013, 2014). Although such considerations are critical for assessing
the integrated terrestrial carbon cycle response to environmental change,
they present unnecessary impediments to assessing the soil biogeochemical
component of terrestrial models and advancing understanding of soil
systems. Moreover, as soils respond slowly to perturbations relative to many
of these upstream factors, modifications of soil model structures and
parameterizations often extend spin‐up time, which ultimately slows model
development (Exbrayat, Pitman, & Abramowitz, 2014; Koven, Chambers,
et al., 2015). To address these challenges, we developed a soil
biogeochemical testbed that facilitates the evaluation of and improvements
to the process‐level representation of global‐scale soil biogeochemical
models.
We compare three soil biogeochemical models that make distinct
assumptions about the processes and factors regulating the formation and
decomposition of soil organic matter. One of the models reflects traditional
ideas about the inherent chemical recalcitrance of soil organic matter. Thus,
it implicitly represents microbial activity and follows a conventional
decomposition cascade regulated by first‐order decay kinetics (Bradford &
Fierer, 2012; Schimel, 2001). The other two models explicitly represent soil
microbial activity and physiology, but make different assumptions about
interactions between microbial community activity and the physicochemical
soil environment. Recognizing that multiple sources of uncertainty generate
spread among models, in this paper we focus on quantifying model structural
uncertainty by comparing steady state soil carbon stocks, turnover times,
and their responses over a transient simulation with soil biogeochemical
models that are forced with identical inputs and environmental conditions.
2 MATERIALS AND METHODS
We created the biogeochemical testbed to conduct global‐scale soil
biogeochemistry simulations using a variety of forcing data sets without the
computational overhead and infrastructure necessary to run a full land
model. Here, we introduce the capabilities of the testbed by using a single
realization of climate and plant productivity estimates that serve as common
inputs to each of three soil organic matter models. In the subsections that
follow, we describe each component of the biogeochemical testbed in
greater detail, but briefly outline the workflow and configuration of the model
here (Figure 1).
Daily estimates of GPP, air temperature, soil temperature, and soil moisture
are needed as inputs to the testbed. The simulations presented here used
data from the Community Land Model (CLM version 4.5, discussed below).
Inputs force the Carnagie‐Aimes‐Stanford Approach terrestrial biosphere
model (CASA‐CNP; created by Potter et al., 1993), with modifications by
(Randerson, Thompson, Conway, Fung, & Field, 1997; Randerson, Thompson,
Malmstrom, Field, & Fung, 1996); and with N and P biogeochemistry as
implemented by (Wang, Law, & Pak, 2010). Here, we use the carbon‐only
version of CASA‐CNP vegetation model to calculate net primary productivity
(NPP) and carbon allocation to different plant tissues (roots, wood, and
leaves), as well as the timing of litterfall. Litterfall inputs are passed onto
three different soil biochemical models that include the CASA‐CNP model
that implicitly represents microbial activity using a first‐order decomposition
approach, as well as two recently developed microbially explicit models that
include the MIcrobial‐MIneralization Carbon Stabilization model (MIMICS;
Wieder, Grandy, Kallenbach, & Bonan, 2014; Wieder, Grandy, Kallenbach,
Taylor, & Bonan, 2015) and the Carbon, Organisms, Rhizosphere, and
Protection in the Soil Environment model (CORPSE; Sulman, Phillips, Oishi,
Shevliakova, & Pacala, 2014). For each model, we ran a spin up simulation to
bring soil organic matter pools to steady state and then conducted a
transient simulation including changes in climate and NPP over the historical
period (1901–2010) to compare the stocks and changes of soil C pools
simulated by each soil model. Below we summarize the data inputs, CASA‐
CNP vegetation model, the three soil carbon models applied in the testbed,
and the testbed configuration. More detailed information can be found in the
online user's manual and technical documentation that accompanies the
publically available model testbed code available
at github.com/wwieder/biogeochem_testbed_1.0.
2.1 Data inputs
Data inputs for the biogeochemical testbed can be modified from a variety of
sources, but for this study, data inputs were generated by the CLM using a
satellite phenology scheme forced with the CRU‐NCEP climate reanalysis
(Koven et al., 2013; Oleson et al., 2013; Figure 1). This standard
configuration of CLM generated globally gridded daily output of gross
primary productivity (GPP), air temperature, soil temperature, liquid soil
moisture and frozen soil moisture for the historical period (1901–2010). Soil
texture inputs to the testbed were depth‐weighted means in the top 50 cm of
soil from the CLM surface data set (Oleson et al., 2013). The testbed
assigned a single plant functional type (PFT) to each 2° × 2° grid cell,
computed as the mode from the 1‐km International Geosphere–Biosphere
Program Data and Information System (IGBP DISCover) data set with 18
vegetation types, including grassy tundra (Loveland et al., 2000; National
Center for Atmospheric Research Staff, 2017). CASA‐CNP defines biome‐
specific parameters corresponding to each PFT (Table S1). Results presented
here use output from the two‐degree version of CLM as input to the testbed,
although the testbed operates independent of resolution and can even be
configured to run for a single point or field site. Postprocessing of CLM
history files was required to format input data that could be read into the
testbed. Specifically, average soil temperature and liquid and frozen soil
moisture used by the testbed are depth‐weighted means in the rooting zone
according to the PFT‐specific root depth and root distribution (Table S1). Only
liquid soil moisture was considered when computing soil moisture limits on
growth for the vegetation model and decomposition in the CASA‐CNP and
CORPSE soil models. CORPSE also required information on frozen soil
moisture to calculate air‐filled pore space. MIMICS did not consider soil
moisture effects on decomposition.
2.2 CASA‐CNP vegetation model
The carbon‐only version of the CASA‐CNP terrestrial biosphere model
calculated daily net primary production (NPP) and subsequent plant litter
inputs to the soil. Daily NPP was calculated by subtracting the sum of plant
maintenance and growth respiration from the CLM‐derived GPP. Maintenance
respiration in CASA‐CNP was zero for leaves, and calculated as a function of
N content (g C g N
−1
day
−1
) for wood and fine roots (determined from fixed
biome‐specific C:N ratios, Table S1). These respiration rates were zero for air/
soil temperatures ≤250 K and increased exponentially with temperature
using a fixed biome‐specific Q
10
(Sitch et al., 2003). Growth respiration was a
fixed fraction (0.35) of the quantity GPP minus the sum of maintenance
respiration fluxes. The relative amounts of NPP allocated to leaves, wood, or
fine roots were fixed biome‐specific fractions that depended on leaf
phenology phase (Wang et al., 2010).
Turnover of live leaves, wood, and fine roots occurred daily at biome‐specific
age‐related death rates. The leaf turnover rate increased with cold and
drought stress, and was modeled following the approach of (Arora &
Boer, 2005). Nonwoody plant litter was partitioned into structural and
metabolic litter material as a function of the biome‐specific lignin:N ratio of
the plant litter (Table S1). Woody plant litter accumulated in the coarse
woody debris (CWD) pool, which decomposed as a function of temperature
and soil moisture for all models and included CO
2
respiration loss. Metabolic
litter, structural litter, and decomposing CWD comprised C inputs to all soil
carbon models in the testbed.
2.3 Soil carbon models
Previous publications document soil models applied in the testbed, but
Table 1 summarizes some of the key similarities and differences among the
soil models. Additional details are also available in the user's manual and
technical documentation available in the testbed's GitHub repository
(see Acknowledgements). The CASA‐CNP soil carbon model had two litter
pools (metabolic and structural) and three soil organic matter pools (fast,
slow, and passive). Live microbial biomass was not explicitly simulated as a
driver of decomposition, but the transfer of C from litter to soil pools or
among soil carbon pools produced CO
2
respiration losses. The decomposition
of pool
i
(
D
i
) is controlled pools size (
C
i
) and pool specific first‐order kinetics
(
k
i
) that are modified by environmental scalars calculated as a function of
soil temperature and moisture (T and θ, respectively).
Structural and metabolic litter pools decomposed into fast and slow pools as
a function of lignin fraction. The CWD pool decomposed to the fast and slow
SOM pools also as a function of the wood lignin fraction. Transfers of C from
the fast and slow pools formed the passive pool and were a function of soil
texture. The passive pool decomposed without transfers of C to other pools.
In CASA‐CNP the cropland PFTs had no moisture limitation on soil organic
matter decomposition and daily turnover rates for the fast, slow, and passive
pools were multiplied by 1.25, 1.5, and 1.5 respectively. Neither MIMICS nor
CORPSE modified decomposition rates for croplands.
MIMICS had two litter pools (metabolic and structural), two live microbial
biomass pools (copiotrophic and oligotrophic, referred to as r and K,
respectively), and three soil organic matter pools (available, chemically
protected, and physically protected). Nonwoody plant litter was partitioned
into metabolic and structural litter pools using a slightly different function of
the lignin:N ratio than the one in the CASA‐CNP model (see user's manual).
Decomposing CWD carbon was transferred to the structural litter pool. The
microbial decomposition of metabolic and structural litter and available SOM
pools were controlled by reverse Michaelis‐Menten kinetics and modified by
soil temperature:
where
D
i
was the decomposition of pool
i, V
max(
T
) was the temperature‐
sensitive maximum reaction velocity, K
es
(
T
) was the temperature‐sensitive
half‐saturation constant specific to the
r
or
K
microbial pool, C
i
was the
carbon pool, and MIC
r
/
K
was the
r
or
K
microbial pool. Decomposition fluxes
also controlled the growth of microbial biomass pools and had CO
2
respiration
losses that were determined by fixed (flux‐specific) microbial growth
efficiencies. Microbial turnover, which was proportional to annual NPP,
transferred C to physically protected, chemically protected, and available
SOM pools, without CO
2
respiration loss. Desorption of the physically
protected pool followed first‐order kinetics and was described as a function
of soil clay content, without CO
2
loss. Oxidation of the chemically protected
SOM, which transferred C to the available pool, followed reverse Michaelis‐
Menten kinetics and was therefore dependent on the size of standing
microbial biomass pools, but as none of the carbon is assimilated into
microbial biomass there are no associated CO
2
losses.
CORPSE had separate surface litter layer pools and SOM pools, each with
three chemically defined carbon species (labile, chemically resistant, and
dead microbes) and a live microbial biomass pool. The surface litter pools
were all considered unprotected while the SOM pools had unprotected and
protected counterparts. Metabolic and structural leaf litter was transferred to
the labile and chemically resistant surface litter pools, respectively, without
CO
2
respiration losses. Similarly, metabolic and structural root litter was
transferred to labile and chemically resistant unprotected soil carbon pools,
respectively. Root exudates, calculated as a fixed 2% of NPP, also
contributed to the labile unprotected soil pool. We reduced root litter input
by the amount of root exudate C added so total C inputs to CORPSE were
identical to those of the other soil models. Carbon from the decomposing
CWD pool was transferred to the chemically resistant litter pool. No carbon
was transferred between the surface litter and soil layers. The microbial
decomposition of unprotected labile, chemically resistant, and dead microbe
litter and SOM pools, CO
2
fluxes, and the growth of microbial biomass were
controlled by the existing microbial biomass and modified by soil
temperature and moisture:
where θ was volumetric liquid soil water content and θ
sat
was saturation soil
water content. Microbial growth efficiencies used fixed, pool‐specific
fractions, with labile C having a high associated growth efficiency and
chemically resistant C having a low efficiency. The model assumed that the
microbial biomass limitation on decomposition was related to the microbial
biomass as a fraction of total carbon. As a result, decomposition rate
responded linearly to total carbon content (similar to a first‐order model) but
was accelerated by greater labile C inputs (which stimulated microbial
biomass growth) and suppressed when labile C was depleted relative to
chemically resistant C. Microbial turnover, which was proportional to a fixed
turnover rate, transferred C to the unprotected dead microbes pool, with
CO
2
respiration loss. Carbon was transferred at fixed, first‐order rates from
the unprotected soil pools to their protected counterparts. These rates varied
with clay content and chemical species (with dead microbes having a
relatively higher protection rate), and occurred without CO
2
respiration
losses. Protected C was transferred back to unprotected pools at a different
fixed, first order rate.
2.4 Testbed configuration, simulations, & analyses
The simulations for each SOM model were carried out in three steps:
initialization, spinup, and transient simulations, which are described below.
We initialized CASA‐CNP vegetation pools by running the testbed with 1901
forcings for 100 years. This initialization created more stable vegetation
pools and litter inputs for subsequent simulations. The state of the CASA‐CNP
vegetation pools (but not SOM pools) from this initialization simulation were
used to initialize spinup runs for all SOM models.
Soil carbon pools were spun up by cycling over 1901–1920 forcings until
organic matter pools reached equilibrium. An SOM model was considered to
be in equilibrium when all three of the following criteria were met between
20‐year cycles: global litter plus soil carbon stocks changed <0.01 Pg, total
litter plus soil carbon in >98% of grid cells changed <1 g C/m
2
, and total
litter plus soil carbon in >98% of grid cells changed <0.1%. Spinup times
varied between models. CASA‐CNP required 10,000 years of an accelerated
spinup followed by 10,000 years of normal spinup in order to reach
equilibrium. For the accelerated spinup, the decomposition rate of the
passive pool was increased tenfold. Following accelerated spinup, the
passive carbon stock was multiplied tenfold before starting the normal
spinup phase. MIMICS organic matter pools required 12,000 years to reach
equilibrium, with the physically protected pool requiring the longest spinup
time. CORPSE organic matter pools required 50,000 years to reach
equilibrium, primarily due to slow continuing accumulation of chemically
resistant litter in high latitudes. In all models, these spinup times are still
prohibitively long for doing many repeated simulations or parameter
estimation, and highlight a research priority that must be addressed (Luo
et al., 2016) in this and other work.
We conducted full transient simulations from 1901 to 2010. For each of the
three soil models currently implemented in the testbed, we compared: (1)
initial conditions following model spinup; (2) changes in soil carbon pools
over the transient simulation; and (3) seasonal patterns of heterotrophic
respiration. Here, we focus on total soil carbon stocks that are simulated by
each model, which were calculated as the sum of all litter, microbial
biomass, and soil carbon pools. Beyond initial carbon stocks, estimates of
steady‐state soil carbon turnover times provide a metric to evaluate the
emergent relationship between climate the mean residence time of various C
stocks (Koven, Hugelius, Lawrence, & Wieder, 2017). Recognizing that
turnover times vary with model structure in transient simulations
(Rasmussen et al., 2016), turnover times were calculated by dividing initial
soil carbon stocks by heterotrophic respiration fluxes for each model,
masking out points with initial productivity <100 g C m
−2
year
−1
. Simulated
results were compared to an observationally derived functional relationship
with mean annual temperature from Koven et al. (2017) that was calculated
by dividing soil carbon stocks from the Harmonized World Soils Database
(HWSD; FAO et al., 2012) and Northern Circumpolar Soil Carbon Database
(Hugelius et al., 2013) by MODIS NPP estimates (Zhao, Heinsch, Nemani, &
Running, 2005). Although this turnover time vs. climate relationship is
derived from present day estimates of plant productivity, we contend that
these inferred turnover times represent important global‐scale patterns that
models should be expected to replicate.
Several additional experiments were conducted that demonstrate the utility
of the testbed in rapidly assessing and understanding variation among
models. Initial simulations suggested that soil texture potentially mediated
soil C responses among models. Thus, we repeated the spinup and fully
transient simulations with globally consistent soil texture (20% clay, 40% silt,
and 40% sand). This global loam experiment only changed the soil texture
effects on particular transfer coefficients and turnover times that were
simulated by each soil biogeochemical model and did not concurrently
modify the soil hydraulic conditions. Second, to decompose the effects of
particular forcings on soil carbon stocks we conducted three isolated‐forcing
experiments where plant productivity, soil temperature, and soil moisture
individually changed over the 20
th
century, but the remaining input variables
were held constant (cycling over 1901–1920 values as in the spinup). We
compared the time series of soil carbon changes from isolated forcing
experiments to the fully transient 20
th
century simulations.
3 RESULTS
3.1 Initial conditions
When forced with CRU‐NCEP climate, simulated global mean annual soil
temperatures were 15.6°C and mean liquid soil moisture was 42.1% of
saturation (Figure S1a,b, averaged over the initialization period, 1901–1920).
GPP estimates from CLM4.5sp totaled 117 ± 1.1 Pg C/year (mean ± 1 σ) and
initial NPP estimates from CASA‐CNP averaged 48 ± 0.8 Pg C/year
(Figure S2a). With these inputs, the biogeochemical testbed generated total
carbon stocks (including litter, soil organic matter and microbial biomass)
totaling 1,360, 1,420, and 1,410 Pg carbon for CASA‐CNP, MIMICS, and
CORPSE respectively (Figure 2a–c; Figure S3). For comparison, soil C
estimates from the HWSD totaled 1,260 Pg C globally (Figure 2d; 0–100 cm
depth, as regridded by (Wieder, Boehnert, Bonan, & Langseth, 2014). Our
aim here is not to evaluate the spatial distribution of soil carbon stocks
simulated by any of the models, although the testbed offers opportunities for
parameter estimation in single point and global simulations (e.g., Hararuk,
Smith, & Luo, 2015) We note, however, that MIMICS was calibrated against
the HWSD (Wieder, Grandy, et al. 2015), whereas CASA‐CNP and CORPSE
were not similarly calibrated. We also recognize that global stocks of ‘litter’ C
are not clearly defined in globally gridded soil carbon estimates, and that the
HWSD likely underestimates high latitude soil C stocks (Todd‐Brown
et al., 2013). Thus, we also present permafrost soil C estimates from the
NCSCD (0–100 cm depth), which shows larger soil carbon stocks in
permafrost regions (Figure 3, Figure S3). The three soil models implemented
in the testbed adequately represented global soil carbon stocks, falling
within benchmark ranges for global soil carbon stocks given an
observationally consistent field of plant productivity (Todd‐Brown
et al., 2014).
Despite general agreement of global soil C stocks among models, they
exhibited notably different spatial distributions. Across high latitudes, CASA‐
CNP and CORPSE generated steady‐state soil C densities that were closer to
observations from the NCSCD and notably larger than those simulated by
MIMICS or observed in the HWSD (Figures 2 and 3, Figure S3). Conversely, at
low latitudes, CASA‐CNP and CORPSE displayed soil carbon densities well
below estimates from MIMICS and the HWSD. The global loam experiment
indicated that steady‐state carbon stocks simulated in CASA‐CNP and MIMICS
showed a greater sensitivity to soil texture (−95 and −178 Pg C,
respectively, compared to control simulation) than CORPSE (+ 27 Pg C).
Whereas CASA‐CNP showed relatively homogenous reductions in steady‐
state soil carbon stocks, MIMICS showed substantially larger soil C
differences in regions of high clay content (e.g., much of the tropics, the
southeastern United States, and SE Asia, Figure S4). All three models
generally showed larger carbon stocks in tundra regions with loam soils,
especially CORPSE.
Although the soil models used similar temperature functions, they showed
large differences in patterns of inferred turnover times and temperature
(Figure 4). Models and observations showed the longest turnover times in
grid cells with colder mean annual temperatures. Observations suggested
that over the cold domain (mean annual temperature <0°C) soil carbon
turnover had a higher temperature sensitivity (steeper slope), whereas over
the warm domain (mean annual temperature >15°C) turnover times had a
lower temperature sensitivity (shallow slope; Koven et al., 2017). The CASA‐
CNP soil model simulated a log‐linear relationship between temperature and
the logarithm of turnover time, with variation among individual grid cells
largely attributed to differences in soil moisture (Figure 4a). In the cold
domain, CASA‐CNP matched the higher temperature sensitivity of soil carbon
turnover better than the two microbially explicit models. In warmer sites,
however, CASA‐CNP showed a linear decrease in log turnover times
(especially in mesic and wet systems), that was not consistent with
observation‐based estimates. (The cluster of grid cells with very low turnover
times are agricultural grid cells, mainly in India, that had high productivity,
but very low soil carbon stocks owning to how agricultural decomposition
rates are handled in CASA‐CNP). By contrast, MIMICS failed to represent high
temperature sensitivity in the cold‐domain, but over the warm‐domain
MIMICS captured the lower temperature sensitivity (flat slope) of inferred
turnover times, although the intercept may be too high (Figure 4b). Finally,
CORPSE showed a stronger than observed temperature sensitivity in all
cases (Figure 4c), with long turnover times simulated by CORPSE in the cold‐
domain resulting in large carbon stocks at high latitudes. Thus, despite
similarities in the overall soil C stocks represented by these models we find
strong differences in the spatial distribution and potential temperature
sensitivities among CASA, MIMICS, and CORPSE that may influence
projections of soil carbon change over the historical period.
Figure 4
Inferred soil carbon turnover times vs. mean annual temperature for each grid cell in CASA‐CNP,
MIMICS, and CORPSE (a–c respectively). Points are colored by mean annual soil moisture (percent
saturation of liquid water), and binned according to the color bar below the figure. Black lines show the
observationally derived relationship between inferred turnover times and temperature ±50%
prediction interval (calculated by Koven et al., 2017)
3.2 Transient response
By the end of the transient simulation period, global mean annual soil
temperature increased by 1.1°C and mean annual soil moisture (calculated
as percent saturation) increased by 0.5%, relative to the initial conditions
(Figure 5a). Notably, high latitude soils showed the greatest changes,
generally becoming warmer and wetter (Figure S1c–d), with higher
wintertime soil temperatures increasing liquid water availability for longer
periods of time. By the start of the 21
st
century, GPP increased by
19 Pg C/year (+16%); meanwhile NPP increased 7 Pg C/year (+15%;
Figure 5a; Figure S2b), and similar in magnitude to an ensemble of CMIP5
Earth system models (Wieder, Cleveland, Smith, & Todd‐Brown, 2015).
Higher plant productivity increased global vegetation carbon stocks
simulated by CASA‐CNP by 36 Pg C, whereas coarse woody debris stocks
declined by 0.7 Pg C.
Figure 5
Globally averaged changes in (a) environmental conditions: soil temperature (°C), soil moisture (%
saturation), and plant litter inputs (red, blue and black lines, respectively); and the cumulative change
(b) soil carbon stocks simulated by: CASA‐CNP, MIMICS, and CORPSE (green, purple, and brown lines,
respectively) in the full transient simulation. Isolated forcing experiments showing changes in soil
carbon stocks following changes in only (c) GPP, (d) soil temperature, and (e) soil moisture. For all
plots, annual values were weighted by land area and differenced from initial conditions averaged over
the spin‐up period
Changes in productivity and climate drove a net accumulation of soil carbon
in CASA‐CNP and MIMICS by the end of the simulation (+18.1 and +24.1 Pg
C, respectively), whereas CORPSE lost soil carbon over the same period
(−21.7 Pg C; Figure 5b). Despite receiving identical litter inputs and climate
forcing, the three soil models tested here showed dramatically different
spatial patterns of soil carbon gains and losses (Figure 6). Particular changes
in soil carbon stocks largely depended on the balance of changes in plant
productivity and soil conditions, along with different assumptions made by
each model. For example, in tundra ecosystems plant productivity increased
by 20%–30%, whereas soil temperature warmed by <1°C
(Figures S1 and S2). In CASA‐CNP and MIMICS this increased plant
productivity overwhelmed soil carbon losses from the increased
heterotrophic respiration, leading to net soil carbon accumulations – mainly
in the litter pools simulated by both models. By contrast, CORPSE lost large
amounts of soil carbon in these regions (Figure 6). Soil texture largely
modulated the initial soil carbon stocks simulated by each model (Figure S4),
but had a more muted effect on transient soil C dynamics. In the global loam
experiment, soil carbon accumulations in CASA‐CNP and MIMICS were
dampened (+17.7 and +19.0 Pg C, respectively), whereas CORPSE lost
slightly more soil carbon over the same period (−22.1 Pg C). MIMICS
assumed that clay rich soils preferentially stabilize microbial residues in
physically protected soil organic matter pools; thus, in the global loam
experiment soil carbon accumulations were approximately 200 g C/
m
2
(roughly 20%) less across the tropics in MIMICS (data not shown).
Figure 6
Spatial distribution of changes in soil carbon stocks (g C/m
2
) simulated by the end of the historical
period (mean of 2001–2010) in the biogeochemical testbed for (a) CASA‐CNP (+18 Pg C), (b) MIMICS
(+24 Pg C), and (c) CORPSE (−21 Pg C)
The testbed allowed us to parse out gross changes among models from
isolated forcing experiments, rather than just seeing the net changes over
the fully transient simulation. Isolated forcing experiments showed that
MIMICS had a higher sensitivity to changes in plant productivity and
temperature than the other models—accumulating twice the amount of C as
CORPSE in the isolated GPP experiment, and losing twice as much C in the
isolated soil temperature simulation (Figure 5c,d, Figure S5). Most of these
differences, however, took place in mid‐to‐low latitudes (<50°N), where
MIMICS simulated significantly larger initial carbon stocks than the other two
models (Figure 3). In MIMICS, microbial turnover increased with higher plant
productivity (Wieder, Grandy, et al. 2015). This served as a density
dependent control over decomposition rates (Buchkowski, Bradford, Grandy,
Schmitz, & Wieder, 2017), but it also increased the inputs of microbial
residues to soil organic matter pools.
Our transient simulations highlighted uncertainties in understanding
temperature and moisture sensitivity in cold regions. Warmer temperatures
ultimately drove the high latitude soil C losses simulated over the
20
th
century; but the isolated forcing experiments demonstrated that CASA‐
CNP and MIMICS had stronger direct sensitivities to changing temperatures
(Figures 5 and 6, Figure S5). By contrast, CORPSE showed the largest
sensitivity to isolated soil moisture forcings (including thawing of frozen soil
water), and lost more than three times the amount of C as the comparable
CASA‐CNP simulation (Figure 5e, Figure S5). Nearly all of the simulated C
losses came from high latitude ecosystems—where soil moisture changes
are mainly controlled by freeze/thaw state and the thawing of frozen soils
allowed the large C stocks built up in frozen conditions to decompose. Thus,
actual temperature sensitivity may be a combination of metabolic
sensitivities to temperature, as well as interactions between temperature
and moisture via controls over liquid water availability in soils subject to
freezing (Commane et al., 2017; Koven, Lawrence, & Riley, 2015).
To further explore differences among models we looked at mean annual
cycles of heterotrophic respiration from the testbed (Figure 7). By design, at
the beginning of the simulations litter inputs equaled heterotrophic
respiration rates for all models (48.1 Pg C/year). A climatology of annual soil
respiration rates averaged across latitudinal bands, therefore, illustrates
differences in the seasonal cycle of carbon fluxes from each model. As each
soil model in the testbed was driven by a common climate and vegetation
model, differences among the left panels of Figure 7 reflect distinctions in
the seasonal amplitude of terrestrial net ecosystem exchange with the
atmosphere. Across midlatitudes in the northern hemisphere CASA‐CNP
showed the lowest amplitude in seasonal CO
2
fluxes (Figure 7a). Over this
same region, MIMICS showed higher summertime respiration than CASA‐CNP,
but both models simulated similar wintertime respiration rates (Figure 7c).
By contrast, CORPSE had very low midlatitude heterotrophic respiration
fluxes in winter, but much larger summertime rates—generating the highest
amplitude seasonal cycle of all the models (Figure 7e). The stronger seasonal
cycle shown by CORPSE is consistent with the high transient sensitivity to
freeze/thaw state by that model. These distinctions were amplified over time
(Figure 7, right panels), showing a global intensification of heterotrophic
CO
2
fluxes between the first and last decades of the simulation. By the end of
the transient simulation annual CO
2
fluxes were no longer equal among
models, however, as soil carbon losses were greater for CORPSE, which
simulated heterotrophic respiration fluxes that were roughly 1 Pg C/year
higher than CASA‐CNP and MIMICS. By the end of the transient simulations,
we also note a qualitative difference in the latitude‐seasonal responses of HR
between CORPSE and the other models in the mid‐ to high‐ latitude regions,
where CORPSE tends to show respiratory increases earlier in the season and
more northerly than the baseline climatological cycle, while the other two
models tend to show increases that are more closely aligned in seasonality
and latitude with the baseline climatology (Figure 7b,d,f).
Figure 7
Hovmöller diagram showing the climatological mean daily respiration rate (g C m
−2
day
−1
) averaged
over each latitude band for the initialization period (1901–1920; left column), and the difference
between the final (2001–2010) and initial (1901–1920) mean daily respiration rates (right column).
Results from each model are shown for (a, b) CASA‐CNP, (c, d) MIMICS, and (e, f) CORPSE
To clarify differences among models we focused on fluxes from a single
latitudinal band (here 54˚N) over the last decade of the simulation.
Figure 8 illustrates the seasonal cycle of environmental drivers (temperature,
soil moisture, and litter inputs), as well as the annual evolution of
heterotrophic respiration fluxes and microbial biomass represented by each
model. Again, CASA‐CNP and MIMICS produced similar wintertime fluxes.
With warming in spring (and greater availability of liquid water) heterotrophic
respiration rates quickly accelerated in all models, but this occurs sooner in
the year for both CASA‐CNP and CORPSE (Figure 8). The annual respiration
rates simulated by CASA‐CNP generally tracked soil temperature changes,
with maximum fluxes corresponding to periods with the warmest soil
temperatures. By contrast, the maximum respiration rates simulated by the
microbially explicit models were somewhat lagged from the CASA‐CNP fluxes
—corresponding to periods when litter inputs and temperature were also
highest. Moreover, MIMICS and CORPSE both simulated higher maximum
heterotrophic respiration rates, leading to a higher amplitude in the seasonal
cycle of soil CO
2
fluxes. Some of this temporal shift in respiration rates was
likely related to changes in microbial biomass stocks, which broadly tracked
the seasonal cycle of litter inputs.
Figure 8
Mean annual cycle of (a) soil temperature, soil moisture and litter inputs (red, blue, and black lines,
respectively) at 54°N over the last decade of the simulation (2001–2010). The lower panel (b) shows
heterotrophic respiration fluxes (solid lines) and microbial biomass stocks (dashed lines) from CASA‐
CNP, MIMICS, and CORPSE (green, purple, and brown lines, respectively) for the same region and time
period
4 DISCUSSION
Our results suggest that the actual uncertainty related to soil carbon
projections may be larger than previously realized. Todd‐Brown et al.
(2013, 2014) reported a wider range of initial soil carbon stocks and
trajectories over the 21
st
century from an ensemble of CMIP5 models, but
each of these models was forced with spatially varying and highly model‐
idiosyncratic climate and productivity estimates. By using a consistent
forcing among models, our results better capture the variation in soil carbon
stocks and their potential response to environmental change that is caused
by different model assumptions, which is translated into model structures,
and particular model parameterizations. Indeed, given their common forcing,
global similarities in testbed results are not surprising (Ahlström, Schurgers,
Arneth, & Smith, 2012; Friend et al., 2014). Models in the biogeochemical
testbed, however, more broadly sample the theoretical space related to soil
organic matter decomposition and stabilization (Wieder, Allison, et al. 2015).
This variation in model form (and parameterization) translated into
differences among models in the: distribution of steady state soil carbon
stocks (Figures 2and 3); functional relationship of turnover time with mean
annual temperature (Figure 4); transient response of soil carbon stocks to
environmental perturbations (Figures 5 and 6) and seasonal dynamics of
heterotrophic respiration (Figures 7 and 8). We acknowledge that some
model spread is likely explained by differences in calibration approaches;
specifically, MIMICS was calibrated against the global pattern of C stocks
estimated by HWSD, while CORPSE and CASA‐CNP were not (Figure 2).
Future calibration of all three models against the same benchmark (e.g.,
Figure 4) may reduce uncertainty in the transient responses among models
(Figure 5).
Through the historical period, CASA‐CNP and MIMICS show similar changes in
global soil carbon stocks (+18 and +24 Pg C, respectively), which were
opposite in sign from the soil carbon changes simulated by CORPSE (−21 Pg
C; Figure 5). When combined with changes to vegetation C stocks from the
CASA‐CNP simulations (+36 Pg C) projected terrestrial carbon uptake would
fall well short of terrestrial carbon sink estimated by the Global Carbon
Project (62–142 Pg C between 1959–2010, assuming uncertainty of 0.8 Pg C/
year; Houghton et al., 2012; Le Quéré et al., 2014). Although our simulations
lack representation of land use and land cover change, results from the
testbed demonstrate that in order to capture inferred trends in terrestrial
carbon uptake over the end of the 20
th
century much less carbon would have
to accumulate in vegetation pools of land models that applied CASA‐CNP and
MIMICS than would be necessary in a model using CORPSE. Here, we focus
on understanding the structural uncertainties among models that broadly
relate to differences among models in their representation of
physicochemical stabilization of soil organic matter, temperature
sensitivities, and moisture sensitivities. Notably, we found that uncertainties
regarding the physicochemical stabilization of soil organic matter and freeze‐
thaw dynamics were greater than uncertainties related to direct temperature
sensitivities among models.
4.1 Physicochemical stabilization
Physical limitation of microbial access to otherwise decomposable substrates
plays a critical role in preserving soil organic matter (Conant et al., 2011;
Cotrufo, Wallenstein, Boot, Denef, & Paul, 2013; Dungait, Hopkins, Gregory,
& Whitmore, 2012; Lehmann & Kleber, 2015; Schimel & Schaeffer, 2012).
Concurrently, microbial biomass serves as both the catalyst for soil organic
matter decomposition and the source of soil organic matter formation,
through the mineral stabilization of microbial residues and necromass
(Grandy & Neff, 2008; Kallenbach, Frey, & Grandy, 2016; Liang, Cheng,
Wixon, & Balser, 2011). While the three models included in the testbed all
represented this process, their implementations and assumptions differed
substantially, reflecting important uncertainties in how to appropriately
represent pore‐scale physicochemical stabilization mechanisms in global‐
scale models. Our global loam experiment illustrated that steady‐state soil
carbon dynamics in CASA‐CNP and MIMICS showed a greater sensitivity to
soil texture than CORPSE (Figure S4). While the appropriateness of soil
texture to describe diverse stabilization mechanisms on mineral surfaces and
within aggregates is in itself debatable (Doetterl et al., 2015; Mikutta, Kleber,
Torn, & Jahn, 2006), texture still serves as a useful proxy for which gridded
input data sets are available for global‐scale simulations (Bailey et al., 2017).
We also note that few of the ESMs represented in the CMIP5 archive use any
information about edaphic properties (texture, mineralogy, or pH) in their soil
biogeochemical submodels.
Regional differences in initial soil carbon stocks highlight the need to better
resolve factors regulating physicochemical stabilization of soil organic matter
in models. For example, CASA‐CNP and CORPSE simulated lower than
observed steady‐state soil carbon densities in warmer ecosystems
(Figures 2 and 3). This suggests that the physicochemical stabilization
mechanisms implicitly represented in these models may not be strong
enough to counteract environmental conditions that would otherwise favor
rapid decomposition (Figure 4). By contrast, MIMICS simulated higher soil
carbon stocks in warm regions that were more consistent with observation‐
based estimates. Similarly, variation among models in transient simulations
reflects uncertainty related to the ultimate fate of new carbon that enters
terrestrial ecosystems. In first order models, like CASA‐CNP, variation in
carbon inputs largely determines the variation in soil carbon changes,
reflecting the linear relationship between inputs and turnover times (Koven,
Chambers, et al., 2015; Todd‐Brown et al., 2014). Accordingly, increased
productivity in the transient simulation increased soil carbon stocks in CASA‐
CNP, especially in colder climates with longer base turnover times
(Figures 5c and 6a, Figure S5b). In the microbially explicit models, increased
plant productivity and litter inputs also build proportionally larger microbial
biomass pools (Figure S2c–d). These larger microbial biomass pools can
simultaneously accelerate the decomposition of organic matter and build soil
carbon stocks. The balance of these factors depends on assumptions about
the catalytic capacity of larger microbial biomass pools vs. the potential fate
of microbial residues.
Increased plant productivity over the 20
th
century increased the rate at which
microbial residues contributed to soil organic matter pools. MIMICS assumes
that finely textured soils have a much greater capacity to stabilize microbial
residues (Wieder, Grandy, et al., 2014), accounting for the larger tropical soil
C accumulation (Figure 6b, Figure S5b). In contrast, larger microbial biomass
pools simulated by CORPSE (as well as increased root exudation) accelerated
the decomposition of unprotected soil organic matter and litter stocks
resulting in smaller increases in C stocks globally (Figures 5c and 6c). The
rapid turnover times simulated by CORPSE in temperate and tropical
ecosystems (Figure 4) suggest that little of the new carbon will be retained in
CORPSE simulations, an interpretation supported by results from the isolated
GPP simulation (Figure S5b).
Indeed, losses of soil carbon have been observed with increasing plant
productivity in high‐latitude ecosystems (Hartley et al., 2012). In temperate
forests, multidecadal litter manipulation studies generally show modest
carbon accumulation in organic soil horizons, but no change in the carbon
stocks of mineral soils (Bowden et al., 2014; Lajtha, Bowden, &
Nadelhoffer, 2014; Lajtha, Townsend, et al., 2014). This suggests a more
nuanced relationship between plant productivity and soil carbon storage may
be necessary to understand and simulate likely terrestrial carbon responses
to changes in plant productivity. The models in the biogeochemical testbed
take a step in this direction, but our results highlight the need to refine the
representation of factors affecting microbial access to otherwise
decomposable substrates in soils.
4.2 Temperature sensitivities
Uncertainties in observed soil biogeochemical responses to temperature
present notable challenges for projecting terrestrial carbon dynamics in a
warming world (Conant et al., 2011; Davidson & Janssens, 2006; Jones, Cox,
& Huntingford, 2003). Although theory predicts that warmer temperatures
should accelerate soil organic matter decomposition and lead to soil carbon
losses, experimental evidence for these assumptions remains unclear
(Bradford, Wieder, et al., 2016). Recent syntheses, however, demonstrate
that experimental warming consistently increases soil respiration rates
(Carey et al., 2016) and leads to soil carbon losses in sites where initial soil
carbon stocks were large (Crowther et al., 2016). Models in the testbed
reflected these general expectations (Figure 5), but extending the insight
provided from these relatively short‐term experimental findings to decadal‐
and centennial‐scales increases the uncertainty associated with societally
relevant carbon cycle projections. Moreover, these syntheses cannot
decompose the changes in productivity vs. turnover times associated with
warming; however, they do corroborate field studies suggesting that warmer
summertime temperature may be accelerating the decomposition of soil
organic matter in the Alaskan tundra and thereby turning Arctic landscapes
into a source of carbon dioxide to the atmosphere (Commane et al., 2017;
Schuur et al., 2009). Collectively, these observations highlight the
importance of capturing the appropriate soil carbon temperature sensitivity
for understanding potential carbon cycle – climate feedbacks, especially in
carbon‐rich, high latitude ecosystems.
Differences in base decomposition rates and temperature sensitivities largely
describe differences in steady state and transient responses among first‐
order models (Todd‐Brown et al., 2014), but understanding apparent
temperature response functions that emerge from microbially explicit
models is somewhat more complicated. Decomposition rates of organic
matter in MIMICS and CORPSE were controlled by reverse Michaelis‐Menten
based kinetics (Equations 2 and 3), and both models applied temperature
functions to calculate maximum reaction velocities (
V
max
) with similar
temperature sensitivities (Q
10
, data not shown). MIMICS, however, also
calculates a temperature sensitive half‐saturation constant (K
es
). This likely
dampened the climate sensitivity of soil carbon turnover times (German,
Marcelo, Stone, & Allison, 2012) and decreased the apparent Q
10
of simulated
reaction rates (Davidson & Janssens, 2006). These factors may explain the
shallow slope in the MIMCS log turnover time – temperature relationships in
warmer domains (Figure 4b). By contrast, CORPSE used a fixed half‐
saturation constant, applied an Arrhenius equation to
calculate
V
max
(resulting in higher temperature sensitivities at lower
temperatures), and assumed that the chemical quality of different substrate
pools conferred different temperature sensitivities. Additionally, CORPSE
strongly limited decomposition when soil water was moistly frozen while
MIMICS did not include an explicit soil moisture dependence. As a result, the
inferred turnover times simulated by CORPSE in temperate and tropical
ecosystems were very fast, but a strong moisture limitation to decomposition
rates in frozen soils drove the change in slope of the log turnover times with
air temperature in Figure 4c.
Model structure also determines variation in the transient responses among
models (Jones et al., 2005; Rasmussen et al., 2016). For example, steady
state turnover times simulated by MIMICS showed the lowest temperature
sensitivity (Figure 4), but the model also had the largest soil C losses in the
isolated soil warming experiment (Figure 5d); whereas the opposite was true
for CORPSE. At high latitudes, most soil carbon simulated by MIMICS was in
pools that were vulnerable to microbial degradation and, therefore, sensitive
to changes in temperature (Figure 5d, Figure S5c). By contrast, much of the
soil carbon simulated by CASA‐CNP was in pools with slower decomposition
rates, thus extending the time needed for temperature sensitivities to
emerge. Indeed, previous work indicates that over decadal times scales
MIMICS has a faster response to experimental warming, compared to a first
order model, but over centennial time scales ultimately loses less carbon
(Wieder, Grandy, et al., 2014). Moreover, local effects like edaphic
properties, substrate quality, microbial community composition, soil
moisture, and redox conditions compound uncertainty in assessing the
vulnerability of soil carbon stocks to temperature change (Bradford, Berg,
Maynard, Wieder, & Wood, 2016; Bradford et al., 2014; Davidson &
Janssens, 2006). Interactions between soil moisture and temperature
resulted in more modest C losses from CORPSE in the isolated soil
temperature experiment (Figure 5d,e; discussed next). Articulating the true
uncertainty associated with any projection of soil carbon change, therefore,
requires a deeper investigation into the structural assumptions represented
in models—which extends beyond temperature sensitivity of carbon turnover
times.
4.3 Moisture sensitivities
At multiple scales of interest, measuring and modeling soil water availability
remains highly uncertain (Clark et al., 2015; Loescher, Ayres, Duffy, Luo, &
Brunke, 2014). Subsequently, translating the effects of the soil hydrologic
state into biogeochemical models also presents enormous challenges
(Carvalhais et al., 2014; Manzoni & Katul, 2014; Moyano, Manzoni, &
Chenu, 2013). Yet, water availability fundamentally determines microbial
activity in all soils. Limited liquid water availability notably preserves soil
organic matter in high‐latitude permafrost systems, where soil water can be
frozen for most or all of the year. The transition from liquid to frozen water
rapidly reduces decomposition rates in the field (Commane et al., 2017) and
models (Koven, Lawrence, et al., 2015), albeit with varied sensitivities
(Figure 5e). Because it lacks structures that consider the effects of liquid
water availability on decomposition rates, MIMICS simulated rapid turnover
times and low soil carbon stocks in permafrost regions (Figures 3 and 4b). In
contrast, CORPSE was especially sensitive to freezing because it strongly
limited decomposition at low soil moisture (Equation 3; Sulman et al., 2014).
This accentuated the strong threshold behavior in steady state turnover
times around mean annual temperatures of 0°C (Figure 4c) and resulted in
much lower wintertime respiration fluxes from CORPSE (Figures 7 and 8).
We recognize that the abrupt changes in turnover times with frozen soils
reflected in CORPSE simulations are at least partially due to the single‐layer
implementation of the soil models here. Indeed, all of the models may
benefit from explicitly resolving profiles of soil temperature and moisture in
their representation of biogeochemical processes to better capture
permafrost soil carbon dynamics (Koven et al., 2013, 2017). Nevertheless,
lengthening of the nonfrozen season in permafrost soils has been shown to
significantly increase soil carbon emissions (Commane et al., 2017); and
these contrasting model outcomes (Figures 5and 6) highlight real and
important sources of uncertainty in projecting carbon cycle responses to
warming and associated hydrologic changes, especially at high latitudes. The
results from CORPSE projecting larger global soil carbon changes to soil
moisture (which is mainly an indirect temperature effect) than to the direct
temperature effect, as well as the larger disagreement between CORPSE and
the other models in the testbed for moisture than temperature responses
(Figures 5e and 6, Figure S5), underscores both the importance and lack of
model agreement on this critical process. Again, however, finding
appropriate data streams to parameterize soil moisture effects on substrate
availability for a global‐scale model remains a challenge. More broadly,
uncertainties among models and observational data sets related to
permafrost soil carbon densities and vulnerability to environmental change
remain an outstanding challenge for global‐scale models (Burke, Jones, &
Koven, 2013; Koven, Riley, & Stern, 2012; Koven, Lawrence, et al., 2015)
that reflects the difficulty in representing interactions between the physical
soil systems and the biotic agents responsible for soil organic matter
formation and decomposition.
This work addresses a particular challenge in comparing, evaluating and
ultimately improving global‐scale soil biogeochemical models under a
common experimental framework. The biogeochemical testbed provides a
computationally tractable, numerically consistent framework to begin
exploring the effects of different model structures and parameterizations on
soil carbon stocks and fluxes at global scales. Variation in soil carbon
projections among models were caused by differences in the steady state
turnover times simulated by each model, and the turnover time responses to
environmental changes over the 20
th
century. These can be simplified into
uncertainties among models related to the physicochemical stabilization
limiting microbial access to otherwise decomposable carbon substrates,
temperature sensitivities of soil organic matter turnover, and effects of liquid
water availability on microbial activity. An important application of the
testbed is motivating improvements in model structures and
parameterizations. Based on our initial results we suggest that improved
parameterization of temperature sensitivities in CORPSE and implementation
of water availability effects on decomposition (especially in frozen soils) in
MIMICS could improve the fidelity of simulations using those models.
Moreover, none of the carbon‐only, single layer models implemented in the
testbed consider the effects of vertical resolution in regulating SOM turnover
—highlighting gaps that should be addressed with future model
development. Continuing to resolve these key uncertainties will require
greater communication between empirical and modeling communities. As
models begin to more faithfully reflect theoretical understanding of factors
responsible for soil organic matter formation and decomposition we see the
testbed as a tool to facilitate regional‐ to global‐scale model comparison and
evaluation, while developing understanding of soil biogeochemical
processes.
ACKNOWLEDGEMENTS
The testbed model code, user's manual and technical documentation are
publically available at github.com/wwieder/biogeochem_testbed_1.0. This
work was supported by the US Department of Energy, Office of Science,
Biological and Environmental Research (BER) under award numbers TES DE‐
SC0014374 and BSS DE‐SC0016364, US Department of Agriculture NIFA
2015‐67003‐23485, and US Department of Energy‐Biological and
Environmental Research, RUBISCO SFA. B. Sulman was supported under
award NA14OAR4320106 from the National Oceanic and Atmospheric
Administration, U.S. Department of Commerce. The statements, findings,
conclusions, and recommendations are those of the author(s) and do not
necessarily reflect the views of the National Oceanic and Atmospheric
Administration, or the U.S. Department of Commerce.
References
Ahlström, A.
,
Schurgers, G.
,
Arneth, A.
, &
Smith, B.
(
2012
).
Robustness and
uncertainty in terrestrial ecosystem carbon response to CMIP5 climate
change projections
.
Environmental Research
Letters
,
7
,
044008
. https://doi.org/10.1088/1748-9326/7/4/044008
Anav, A.
,
Friedlingstein, P.
,
Kidston, M.
,
Bopp, L.
,
Ciais, P.
,
Cox, P.
, …
Zhu,
Z.
(
2013
).
Evaluating the land and ocean components of the global carbon
cycle in the CMIP5 earth system models
.
Journal of
Climate
,
26
,
6801
–
6843
. https://doi.org/10.1175/jcli-d-12-00417.1
Arora, V. K.
, &
Boer, G. J.
(
2005
).
A parameterization of leaf phenology for the
terrestrial ecosystem component of climate models
.
Global Change
Biology
,
11
,
39
–
59
. https://doi.org/10.1111/j.1365-2486.2004.00890.x
Arora, V. K.
,
Boer, G. J.
,
Friedlingstein, P.
,
Eby, M.
,
Jones, C. D.
,
Christian, J. R.
,
…
Hajima, T.
(
2013
).
Carbon‐concentration and carbon‐climate feedbacks in
CMIP5 earth system models
.
Journal of
Climate
,
26
,
5289
–
5314
. https://doi.org/10.1175/jcli-d-12-00494.1
Bailey, V. L.
,
Bond‐Lamberty, B.
,
Deangelis, K.
,
Grandy, A. S.
,
Hawkes, C.
V.
,
Heckman, K.
, …
Wallenstein, M. D.
(
2017
).
Soil carbon cycling proxies:
Understanding their critical role in predicting climate change
feedbacks
.
Global Change Biology
,
1
–
11
. https://doi.org/10.1111/gcb.13926
Bowden, R. D.
,
Deem, L.
,
Plante, A. F.
,
Peltre, C.
,
Nadelhoffer, K.
, &
Lajtha,
K.
(
2014
).
Litter input controls on soil carbon in a temperate deciduous
forest
.
Soil Science Society of America
Journal
,
78
,
S66
–
S75
. https://doi.org/10.2136/sssaj2013.09.0413nafsc
Bradford, M. A.
,
Berg, B.
,
Maynard, D. S.
,
Wieder, W. R.
, &
Wood, S.
A.
(
2016
).
Understanding the dominant controls on litter
decomposition
.
Journal of
Ecology
,
104
,
229
–
238
. https://doi.org/10.1111/1365-2745.12507
Bradford, M. A.
, &
Fierer, N.
(
2012
).
The biogeography of microbial
communities and ecosystem processes: Implications for soil and ecosystem
models
. In
D. H. Wall
,
R. D. Bardget
,
V. Behan‐Pelletier
,
J. E. Herrick
,
T. H.
Jones
,
K. Ritz
,
J. Six
,
D. R. Strong
&
W. H. Van Der Putten
(Eds.),
Soil ecology
and ecosystem services
(pp.
189
–
200
).
Oxford, UK
: Oxford University
Press. https://doi.org/10.1093/acprof:oso/9780199575923.003.0017
Bradford, M. A.
,
Warren Ii, R. J.
,
Baldrian, P.
,
Crowther, T. W.
,
Maynard, D.
S.
,
Oldfield, E. E.
, …
King, J. R.
(
2014
).
Climate fails to predict wood
decomposition at regional scales
.
Nature Climate
Change
,
4
,
625
–
630
. https://doi.org/10.1038/nclimate2251
Bradford, M. A.
,
Wieder, W. R.
,
Bonan, G. B.
,
Fierer, N.
,
Raymond, P. A.
,
&
Crowther, T. W.
(
2016
).
Managing uncertainty in soil carbon feedbacks to
climate change
.
Nature Climate
Change
,
6
,
751
–
758
. https://doi.org/10.1038/nclimate3071
Buchkowski, R. W.
,
Bradford, M. A.
,
Grandy, A. S.
,
Schmitz, O. J.
, &
Wieder,
W. R.
(
2017
).
Applying population and community ecology theory to advance
understanding of belowground biogeochemistry
.
Ecology
Letters
,
20
,
231
–
245
. https://doi.org/10.1111/ele.12712
Burke, E. J.
,
Jones, C. D.
, &
Koven, C. D.
(
2013
).
Estimating the permafrost‐
carbon climate response in the CMIP5 climate models using a simplified
approach
.
Journal of Climate
,
26
,
4897
–
4909
. https://doi.org/10.1175/jcli-d-
12-00550.1
Carey, J. C.
,
Tang, J.
,
Templer, P. H.
,
Kroeger, K. D.
,
Crowther, T. W.
,
Burton,
A. J.
, …
Jiang, L.
(
2016
).
Temperature response of soil respiration largely
unaltered with experimental warming
.
Proceedings of the National Academy
of Sciences of the
USA
,
113
,
13797
–
13802
. https://doi.org/10.1073/pnas.1605365113
Carvalhais, N.
,
Forkel, M.
,
Khomik, M.
,
Bellarby, J.
,
Jung, M.
,
Migliavacca, M.
,
…
Weber, U.
(
2014
).
Global covariation of carbon turnover times with climate
in terrestrial
ecosystems
.
Nature
,
514
,
213
–
217
. https://doi.org/10.1038/nature13731
Ciais, P.
,
Sabine, C.
,
Bala, G.
,
Bopp, L.
,
Brovkin, V.
,
Canadell, J.
, …
Jones,
C.
(
2013
).
Carbon and other biogeochemical cycles
. In
T. F. Stocker
,
D.
Qin
,
G.‐K. Plattner
,
M. Tignor
,
S. K. Allen
,
J. Boschung
,
A. Nauels
,
Y. Xia
,
V.
Bex
&
P. M. Midgley
(Eds.),
Climate change 2013: The physical science basis.
Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change
(pp.
465
–
570
).
Cambridge, UK
and New York, NY, USA
: Cambridge University Press.
Clark, M. P.
,
Nijssen, B.
,
Lundquist, J. D.
,
Kavetski, D.
,
Rupp, D. R.
,
Woods, R.
A.
, …
Arnold, J. R.
(
2015
).
A unified approach for process‐based hydrologic
modeling: 1. Modeling concept
.
Water Resources
Research
,
51
,
2498
–
2514
. https://doi.org/10.1002/2015WR017198
Commane, R.
,
Lindaas, J.
,
Benmergui, J.
,
Luus, K. A.
,
Chang, R.
,
Daube, B. C.
,
…
Miller, S. M.
(
2017
).
Carbon dioxide sources from Alaska driven by
increasing early winter respiration from Arctic tundra
.
Proceedings of the
National Academy of Sciences of the
USA
,
114
,
5361
–
5366
. https://doi.org/10.1073/pnas.1618567114
Conant, R. T.
,
Ryan, M. G.
,
Ågren, G. I.
,
Birge, H. E.
,
Davidson, E. A.
,
Eliasson,
P. E.
, …
Hyvönen, R.
(
2011
).
Temperature and soil organic matter
decomposition rates – Synthesis of current knowledge and a way
forward
.
Global Change
Biology
,
17
,
3392
–
3404
. https://doi.org/10.1111/j.1365-2486.2011.02496.x
Cotrufo, M. F.
,
Wallenstein, M. D.
,
Boot, C. M.
,
Denef, K.
, &
Paul,
E.
(
2013
).
The Microbial Efficiency‐Matrix Stabilization (MEMS) framework
integrates plant litter decomposition with soil organic matter stabilization: Do
labile plant inputs form stable soil organic matter?
Global Change
Biology
,
19
,
988
–
995
. https://doi.org/10.1111/gcb.12113
Crowther, T. W.
,
Todd‐Brown, K. E. O.
,
Rowe, C. W.
,
Wieder, W.
,
Carey, J.
C.
,
Machmuller, M. B.
, …
Blair, J. M.
(
2016
).
Quantifying global soil carbon
losses in response to
warming
.
Nature
,
540
,
104
–
108
. https://doi.org/10.1038/nature20150
Davidson, E. A.
, &
Janssens, I. A.
(
2006
).
Temperature sensitivity of soil
carbon decomposition and feedbacks to climate
change
.
Nature
,
440
,
165
–
173
. https://doi.org/10.1038/nature04514
Doetterl, S.
,
Stevens, A.
,
Six, J.
,
Merckx, R.
,
Van Oost, K.
,
Pinto, M. C.
,
…
Boeckx, P.
(
2015
).
Soil carbon storage controlled by interactions between
geochemistry and climate
.
Nature
Geoscience
,
8
,
780
–
783
. https://doi.org/10.1038/ngeo2516
Dungait, J. J.
,
Hopkins, D. W.
,
Gregory, A. S.
, &
Whitmore, A. P.
(
2012
).
Soil
organic matter turnover is governed by accessibility not recalcitrance
.
Global
Change Biology
,
18
,
1781
–
1796
. https://doi.org/10.1111/j.1365-
2486.2012.02665.x
Exbrayat, J. F.
,
Pitman, A. J.
, &
Abramowitz, G.
(
2014
).
Disentangling
residence time and temperature sensitivity of microbial decomposition in a
global soil carbon
model
.
Biogeosciences
,
11
,
6999
–
7008
. https://doi.org/10.5194/bg-11-6999-
2014
FAO
,
IIASA
,
ISRIC
,
ISSCAS
,
JRC
(
2012
).
Harmonized world soil database
(version 1.2)
.
Rome, Italy
: FAO; and Laxenburg, Austria: IIASA.
Friedlingstein, P.
,
Meinshausen, M.
,
Arora, V. K.
,
Jones, C. D.
,
Anav,
A.
,
Liddicoat, S. K.
, &
Knutti, R.
(
2014
).
Uncertainties in CMIP5 climate
projections due to carbon cycle feedbacks
.
Journal of
Climate
,
27
,
511
–
526
. https://doi.org/10.1175/jcli-d-12-00579.1
Friend, A. D.
,
Lucht, W.
,
Rademacher, T. T.
,
Keribin, R.
,
Betts, R.
,
Cadule, P.
,
…
Ito, A.
(
2014
).
Carbon residence time dominates uncertainty in terrestrial
vegetation responses to future climate and atmospheric CO2
.
Proceedings of
the National Academy of Sciences of the
USA
,
111
,
3280
–
3285
. https://doi.org/10.1073/pnas.1222477110
German, D. P.
,
Marcelo, K. R. B.
,
Stone, M. M.
, &
Allison, S. D.
(
2012
).
The
Michaelis‐Menten kinetics of soil extracellular enzymes in response to
temperature: A cross‐latitudinal study
.
Global Change
Biology
,
18
,
1468
–
1479
. https://doi.org/10.1111/j.1365-2486.2011.02615.x
Grandy, A. S.
, &
Neff, J. C.
(
2008
).
Molecular C dynamics downstream: The
biochemical decomposition sequence and its impact on soil organic matter
structure and function
.
Science of the Total
Environment
,
404
,
297
–
307
. https://doi.org/10.1016/j.scitotenv.2007.11.013
Hararuk, O.
,
Smith, M. J.
, &
Luo, Y.
(
2015
).
Microbial models with data‐driven
parameters predict stronger soil carbon responses to climate change
.
Global
Change Biology
,
21
,
2439
–
2453
. https://doi.org/10.1111/gcb.12827
Hartley, I. P.
,
Garnett, M. H.
,
Sommerkorn, M.
,
Hopkins, D. W.
,
Fletcher, B.
J.
,
Sloan, V. L.
, …
Wookey, P. A.
(
2012
).
A potential loss of carbon associated
with greater plant growth in the European Arctic
.
Nature Climate
Change
,
2
,
875
–
879
. https://doi.org/10.1038/nclimate1575
Hoffman, F. M.
,
Randerson, J. T.
,
Arora, V. K.
,
Bao, Q.
,
Cadule, P.
,
Ji, D.
,
…
Obata, A.
(
2014
).
Causes and implications of persistent atmospheric
carbon dioxide biases in earth system models
.
Journal of Geophysical
Research:
Biogeosciences
,
119
,
141
–
162
. https://doi.org/10.1002/2013JG002381
Houghton, R. A.
,
House, J. I.
,
Pongratz, J.
,
Van Der Werf, G. R.
,
DeFries, R.
S.
,
Hansen, M. C.
, …
Ramankutty, N.
(
2012
).
Carbon emissions from land use
and land‐cover
change
.
Biogeosciences
,
9
,
5125
–
5142
. https://doi.org/10.5194/bg-9-5125-
2012
Hugelius, G.
,
Tarnocai, C.
,
Broll, G.
,
Canadell, J. G.
,
Kuhry, P.
, &
Swanson, D.
K.
(
2013
).
The Northern Circumpolar Soil Carbon Database: Spatially
distributed datasets of soil coverage and soil carbon storage in the northern
permafrost regions
.
Earth System Science
Data
,
5
,
3
–
13
. https://doi.org/10.5194/essd-5-3-2013
Jobbágy, E. G.
, &
Jackson, R. B.
(
2000
).
The vertical distribution of soil
organic carbon and its relation to climate and vegetation
.
Ecological
Applications
,
10
,
423
–
436
. https://doi.org/10.1890/1051-
0761(2000)010[0423:tvdoso]2.0.co;2
Jones, C. D.
,
Ciais, P.
,
Davis, S. J.
,
Friedlingstein, P.
,
Gasser, T.
,
Peters, G. P.
,
…
Jackson, R. B.
(
2016
).
Simulating the Earth system response to negative
emissions
.
Environmental Research
Letters
,
11
,
095012
. https://doi.org/10.1088/1748-9326/11/9/095012
Jones, C. D.
,
Cox, P.
, &
Huntingford, C.
(
2003
).
Uncertainty in climate–carbon‐
cycle projections associated with the sensitivity of soil respiration to
temperature
.
Tellus Series B
,
55
,
642
–
648
. https://doi.org/10.1034/j.1600-
0889.2003.01440.x
Jones, C.
,
Mcconnell, C.
,
Coleman, K.
,
Cox, P.
,
Falloon, P.
,
Jenkinson, D.
,
&
Powlson, D.
(
2005
).
Global climate change and soil carbon stocks;
predictions from two contrasting models for the turnover of organic carbon in
soil
.
Global Change Biology
,
11
,
154
–
166
. https://doi.org/10.1111/j.1365-
2486.2004.00885.x
Jones, C.
,
Robertson, E.
,
Arora, V.
,
Friedlingstein, P.
,
Shevliakova, E.
,
Bopp, L.
,
…
Liddicoat, S.
(
2013
).
Twenty‐first‐century compatible CO2 emissions and
airborne fraction simulated by CMIP5 earth system models under four
representative concentration pathways
.
Journal of
Climate
,
26
,
4398
–
4413
. https://doi.org/10.1175/jcli-d-12-00554.1
Kallenbach, C. M.
,
Frey, S. D.
, &
Grandy, A. S.
(
2016
).
Direct evidence for
microbial‐derived soil organic matter formation and its ecophysiological
controls
.
Nature
Communications
,
7
,
13630
. https://doi.org/10.1038/ncomms13630
Koven, C. D.
,
Chambers, J. Q.
,
Georgiou, K.
,
Knox, R.
,
Negron‐Juarez, R.
,
Riley,
W. J.
, …
Jones, C. D.
(
2015
).
Controls on terrestrial carbon feedbacks by
productivity versus turnover in the CMIP5 earth system
models
.
Biogeosciences
,
12
,
5211
–
5228
. https://doi.org/10.5194/bg-12-
5211-2015
Koven, C. D.
,
Hugelius, G.
,
Lawrence, D. M.
, &
Wieder, W. R.
(
2017
).
Higher
climatological temperature sensitivity of soil carbon in cold than warm
climates
.
Nature Climate
Change
,
7
,
817
–
822
. https://doi.org/10.1038/nclimate3421
Koven, C. D.
,
Lawrence, D. M.
, &
Riley, W. J.
(
2015
).
Permafrost carbon–
climate feedback is sensitive to deep soil carbon decomposability but not
deep soil nitrogen dynamics
.
Proceedings of the National Academy of
Sciences of the
USA
,
112
,
3752
–
3757
. https://doi.org/10.1073/pnas.1415123112
Koven, C. D.
,
Riley, W. J.
, &
Stern, A.
(
2012
).
Analysis of permafrost thermal
dynamics and response to climate change in the CMIP5 earth system
models
.
Journal of Climate
,
26
,
1877
–
1900
. https://doi.org/10.1175/jcli-d-12-
00228.1
Koven, C. D.
,
Riley, W. J.
,
Subin, Z. M.
,
Tang, J. Y.
,
Torn, M. S.
,
Collins, W. D.
,
…
Swenson, S. C.
(
2013
).
The effect of vertically‐resolved soil
biogeochemistry and alternate soil C and N models on C dynamics of
CLM4
.
Biogeosciences
,
10
,
7109
–
7131
. https://doi.org/10.5194/bg-10-7109-
2013
Lajtha, K.
,
Bowden, R. D.
, &
Nadelhoffer, K.
(
2014
).
Litter and root
manipulations provide insights into soil organic matter dynamics and
stability
.
Soil Science Society of America
Journal
,
78
,
S261
–
S269
. 10.2136/sssaj2013.08.0370nafsc
Lajtha, K.
,
Townsend, K. L.
,
Kramer, M. G.
,
Swanston, C.
,
Bowden, R. D.
,
&
Nadelhoffer, K.
(
2014
).
Changes to particulate versus mineral‐associated
soil carbon after 50 years of litter manipulation in forest and prairie
experimental
ecosystems
.
Biogeochemistry
,
119
,
341
–
360
. https://doi.org/10.1007/
S10533-014-9970-5
Le Quéré, C.
,
Peters, G. P.
,
Andres, R. J.
,
Andrew, R. M.
,
Boden, T. A.
,
Ciais, P.
,
…
Sitch, S.
(
2014
).
Global carbon budget 2013
.
Earth System Science
Data
,
6
,
235
–
263
. https://doi.org/10.5194/essd-6-235-2014
Lehmann, J.
, &
Kleber, M.
(
2015
).
The contentious nature of soil organic
matter
.
Nature
,
528
,
60
–
68
. https://doi.org/10.1038/nature16069
Liang, C.
,
Cheng, G.
,
Wixon, D.
, &
Balser, T.
(
2011
).
An Absorbing Markov
Chain approach to understanding the microbial role in soil carbon
stabilization
.
Biogeochemistry
,
106
,
303
–
309
. https://doi.org/10.1007/
s10533-010-9525-3
Loescher, H.
,
Ayres, E.
,
Duffy, P.
,
Luo, H.
, &
Brunke, M.
(
2014
).
Spatial
variation in soil properties among North American ecosystems and guidelines
for sampling designs
.
PLoS
One
,
9
,
e83216
. https://doi.org/10.1371/journal.pone.0083216
Loveland, T. R.
,
Reed, B. C.
,
Brown, J. F.
,
Ohlen, D. O.
,
Zhu, Z.
,
Yang, L.
,
&
Merchant, J. W.
(
2000
).
Development of a global land cover characteristics
database and IGBP DISCover from 1 km AVHRR data
.
International Journal of
Remote Sensing
,
21
,
1303
–
1330
. https://doi.org/10.1080/014311600210191
Luo, Y. Q.
,
Ahlstrom, A.
,
Allison, S. D.
,
Batjes, N. H.
,
Brovkin, V.
,
Carvalhais,
N.
, …
Georgiou, K.
(
2016
).
Toward more realistic projections of soil carbon
dynamics by earth system models
.
Global Biogeochemical
Cycles
,
30
,
40
–
56
. https://doi.org/10.1002/2015gb005239
Manzoni, S.
, &
Katul, G.
(
2014
).
Invariant soil water potential at zero
microbial respiration explained by hydrological discontinuity in dry
soils
.
Geophysical Research
Letters
,
41
,
7151
–
7158
. https://doi.org/10.1002/2014GL061467
Manzoni, S.
, &
Porporato, A.
(
2009
).
Soil carbon and nitrogen mineralization:
Theory and models across scales
.
Soil Biology and
Biochemistry
,
41
,
1355
–
1379
. https://doi.org/10.1016/j.soilbio.2009.02.031
Mikutta, R.
,
Kleber, M.
,
Torn, M.
, &
Jahn, R.
(
2006
).
Stabilization of soil
organic matter: Association with minerals or chemical
recalcitrance?
Biogeochemistry
,
77
,
25
–
56
. https://doi.org/10.1007/s10533-
005-0712-6
Moyano, F. E.
,
Manzoni, S.
, &
Chenu, C.
(
2013
).
Responses of soil
heterotrophic respiration to moisture availability: An exploration of processes
and models
.
Soil Biology and
Biochemistry
,
59
,
72
–
85
. https://doi.org/10.1016/j.soilbio.2013.01.002
National Center for Atmospheric Research Staff
(Eds). (
2017
).
The climate
data guide: CERES: IGBP land classification
. Retrieved
from https://climatedataguide.ucar.edu/climate-data/ceres-igbp-land-
classification
Oleson, K.
,
Lawrence, D. M.
,
Bonan, G. B.
,
Drewniak, B.
,
Huang, M.
,
Koven,
C.
, …
Yang, Z. L.
(
2013
).
Technical description of version 4.5 of the
Community Land Model (CLM)
. NCAR Technical Note NCAR/TN‐
503+STR.
420
pp, https://doi.org/10.5065/d6rr1w7m
Potter, C. S.
,
Randerson, J. T.
,
Field, C. B.
,
Matson, P. A.
,
Vitousek, P.
M.
,
Mooney, H. A.
, &
Klooster, S. A.
(
1993
).
Terrestrial ecosystem production
– a process model‐based on global satellite and surface data
.
Global
Biogeochemical Cycles
,
7
,
811
–
841
. https://doi.org/10.1029/93gb02725
Randerson, J. T.
,
Thompson, M. V.
,
Conway, T. J.
,
Fung, I. Y.
, &
Field, C.
B.
(
1997
).
The contribution of terrestrial sources and sinks to trends in the
seasonal cycle of atmospheric carbon dioxide
.
Global Biogeochemical
Cycles
,
11
,
535
–
560
. https://doi.org/10.1029/97GB02268
Randerson, J. T.
,
Thompson, M. V.
,
Malmstrom, C. M.
,
Field, C. B.
, &
Fung, I.
Y.
(
1996
).
Substrate limitations for heterotrophs: Implications for models that
estimate the seasonal cycle of atmospheric CO2
.
Global Biogeochemical
Cycles
,
10
,
585
–
602
. https://doi.org/10.1029/96GB01981
Rasmussen, M.
,
Hastings, A.
,
Smith, M. J.
,
Agusto, F. B.
,
Chen‐Charpentier, B.
M.
,
Hoffman, F. M.
, …
Luo, Y.
(
2016
).
Transit times and mean ages for
nonautonomous and autonomous compartmental systems
.
Journal of
Mathematical Biology
,
73
,
1379
–
1398
. https://doi.org/10.1007/s00285-016-
0990-8
Schimel, J.
(
2001
).
1.13 ‐ Biogeochemical models: Implicit versus explicit
microbiology
. In
E.‐D. Schulze
,
M. Heimann
,
S. Harrison
,
E. Holland
,
J. Lloyd
,
I.
C. Prentice
&
D. S. Schimel
(Eds.),
Global biogeochemical cycles in the
climate system
(pp.
177
–
183
).
San Diego, IL
: Academic Press.
Schimel, J. P.
, &
Schaeffer, S. M.
(
2012
).
Microbial control over carbon cycling
in soil
.
Frontiers in
Microbiology
,
3
,
348
. https://doi.org/10.3389/fmicb.2012.00348
Schmidt, M. W.
,
Torn, M. S.
,
Abiven, S.
,
Dittmar, T.
,
Guggenberger,
G.
,
Janssens, I. A.
, …
Nannipieri, P.
(
2011
).
Persistence of soil organic matter
as an ecosystem
property
.
Nature
,
478
,
49
–
56
. https://doi.org/10.1038/nature10386
Schuur, E. A.
,
Vogel, J. G.
,
Crummer, K. G.
,
Lee, H.
,
Sickman, J. O.
,
&
Osterkamp, T. E.
(
2009
).
The effect of permafrost thaw on old carbon
release and net carbon exchange from
tundra
.
Nature
,
459
,
556
–
559
. https://doi.org/10.1038/nature08031
Sierra, C. A.
,
Müller, M.
, &
Trumbore, S. E.
(
2012
).
Models of soil organic
matter decomposition: The SoilR package, version 1.0
.
Geoscientific Model
Development
,
5
,
1045
–
1060
. https://doi.org/10.5194/gmd-5-1045-2012
Sitch, S.
,
Smith, B.
,
Prentice, I. C.
,
Arneth, A.
,
Bondeau, A.
,
Cramer, W.
,
…
Venevsky, S.
(
2003
).
Evaluation of ecosystem dynamics, plant geography
and terrestrial carbon cycling in the LPJ dynamic global vegetation
model
.
Global Change Biology
,
9
,
161
–
185
. https://doi.org/10.1046/j.1365-
2486.2003.00569.x
Sulman, B. N.
,
Phillips, R. P.
,
Oishi, A. C.
,
Shevliakova, E.
, &
Pacala, S.
W.
(
2014
).
Microbe‐driven turnover offsets mineral‐mediated storage of soil
carbon under elevated CO
2
.
Nature Climate
Change
,
4
,
1099
–
1102
. https://doi.org/10.1038/nclimate2436
Tian, H. Q.
,
Lu, C. Q.
,
Yang, J.
,
Banger, K.
,
Huntzinger, D. N.
,
Schwalm, C. R.
,
…
Huang, M.
(
2015
).
Global patterns and controls of soil organic carbon
dynamics as simulated by multiple terrestrial biosphere models: Current
status and future directions
.
Global Biogeochemical
Cycles
,
29
,
775
–
792
. https://doi.org/10.1002/2014gb005021
Todd‐Brown, K. E. O.
,
Randerson, J. T.
,
Hopkins, F.
,
Arora, V.
,
Hajima,
T.
,
Jones, C.
, …
Zhang, Q.
(
2014
).
Changes in soil organic carbon storage
predicted by earth system models during the 21st
century
.
Biogeosciences
,
11
,
2341
–
2356
. https://doi.org/10.5194/bg-11-
2341-2014
Todd‐Brown, K. E. O.
,
Randerson, J. T.
,
Post, W. M.
,
Hoffman, F. M.
,
Tarnocai,
C.
,
EaG, S.
, &
Allison, S. D.
(
2013
).
Causes of variation in soil carbon
predictions from CMIP5 earth system models and comparison with
observations
.
Biogeosciences
,
10
,
1717
–
1736
. https://doi.org/10.5194/bg-
10-1717-2013
Wang, Y. P.
,
Law, R. M.
, &
Pak, B.
(
2010
).
A global model of carbon, nitrogen
and phosphorus cycles for the terrestrial
biosphere
.
Biogeosciences
,
7
,
2261
–
2282
. https://doi.org/10.5194/bg-7-
2261-2010
Wieder, W. R.
,
Allison, S. D.
,
Davidson, E. A.
,
Georgiou, K.
,
Hararuk, O.
,
He,
Y.
, …
Todd‐Brown, K.
(
2015
).
Explicitly representing soil microbial processes
in earth system models
.
Global Biogeochemical
Cycles
,
29
,
1782
–
1800
. https://doi.org/10.1002/2015gb005188
Wieder, W. R.
,
Boehnert, J.
,
Bonan, G. B.
, &
Langseth, M.
(
2014
).
Regridded
harmonized world soil database v1.2. from Oak Ridge National Laboratory
Distributed Active Archive Center, Oak Ridge, Tennessee, USA
. Retrieved
from http://daac.ornl.gov/, https://doi.org/10.3334/ornldaac/1247
Wieder, W. R.
,
Cleveland, C. C.
,
Smith, W. K.
, &
Todd‐Brown,
K.
(
2015
).
Future productivity and carbon storage limited by terrestrial
nutrient availability
.
Nature Geoscience
,
8
,
441
–
444
. https://doi.org/10.1038/
ngeo2413
Wieder, W. R.
,
Grandy, A. S.
,
Kallenbach, C. M.
, &
Bonan, G.
B.
(
2014
).
Integrating microbial physiology and physio‐chemical principles in
soils with the MIcrobial‐MIneral Carbon Stabilization (MIMICS)
model
.
Biogeosciences
,
11
,
3899
–
3917
. https://doi.org/10.5194/bg-11-3899-
2014
Wieder, W. R.
,
Grandy, A. S.
,
Kallenbach, C. M.
,
Taylor, P. G.
, &
Bonan, G.
B.
(
2015
).
Representing life in the Earth system with soil microbial functional
traits in the MIMICS model
.
Geoscientific Model
Development
,
8
,
1789
–
1808
. https://doi.org/10.5194/gmd-8-1789-2015
Zhang, Q.
,
Wang, Y. P.
,
Matear, R. J.
,
Pitman, A. J.
, &
Dai, Y.
J.
(
2014
).
Nitrogen and phosphorous limitations significantly reduce future
allowable CO2 emissions
.
Geophysical Research
Letters
,
41
,
632
–
637
. https://doi.org/10.1002/2013GL058352
Zhao, M.
,
Heinsch, F. A.
,
Nemani, R. R.
, &
Running, S.
W.
(
2005
).
Improvements of the MODIS terrestrial gross and net primary
production global data set
.
Remote Sensing Of
Environment
,
95
,
164
–
176
. https://doi.org/10.1016/j.rse.2004.12.011