Ecological Applications, 20(7), 2010, pp. 1805–1819
Ó 2010 by the Ecological Society of America
Simulating greenhouse gas budgets of four California cropping
systems under conventional and alternative management
STEVEN DE GRYZE,1,2,5 ADAM WOLF,3 STEPHEN R. KAFFKA,1 JEFF MITCHELL,1 DENNIS E. ROLSTON,4
STEVEN R. TEMPLE,1 JUHWAN LEE,1 AND JOHAN SIX1
1
Department of Plant Sciences, University of California, One Shields Avenue, Davis, California 95616 USA
2
Terra Global Capital, One Ferry Building, Suite 255, San Francisco, California 94111 USA
3
Carnegie Institution, Department of Global Ecology, Stanford University, 260 Panama Street, Stanford, California 94305 USA
4
Department of Land Air and Water Resources, University of California, One Shields Avenue, Davis, California 95616 USA
Abstract. Despite the importance of agriculture in California’s Central Valley, the
potential of alternative management practices to reduce soil greenhouse gas (GHG) emissions
has been poorly studied in California. This study aims at (1) calibrating and validating
DAYCENT, an ecosystem model, for conventional and alternative cropping systems in
California’s Central Valley, (2) estimating CO2, N2O, and CH4 soil fluxes from these systems,
and (3) quantifying the uncertainty around model predictions induced by variability in the
input data. The alternative practices considered were cover cropping, organic practices, and
conservation tillage. These practices were compared with conventional agricultural
management. The crops considered were beans, corn, cotton, safflower, sunflower, tomato,
and wheat. Four field sites, for which at least five years of measured data were available, were
used to calibrate and validate the DAYCENT model. The model was able to predict 86–94%
of the measured variation in crop yields and 69–87% of the measured variation in soil organic
carbon (SOC) contents. A Monte Carlo analysis showed that the predicted variability of SOC
contents, crop yields, and N2O fluxes was generally smaller than the measured variability of
these parameters, in particular for N2O fluxes. Conservation tillage had the smallest potential
to reduce GHG emissions among the alternative practices evaluated, with a significant
reduction of the net soil GHG fluxes in two of the three sites of 336 6 47 and 550 6 123 kg
CO2-eqha1yr1 (mean 6 SE). Cover cropping had a larger potential, with net soil GHG flux
reductions of 752 6 10, 1072 6 272, and 2201 6 82 kg CO2-eqha1yr1. Organic practices
had the greatest potential for soil GHG flux reduction, with 4577 6 272 kg CO2-eqha1yr1.
Annual differences in weather or management conditions contributed more to the variance in
annual GHG emissions than soil variability did. We concluded that the DAYCENT model
was successful at predicting GHG emissions of different alternative management systems in
California, but that a sound error analysis must accompany the predictions to understand the
risks and potentials of GHG mitigation through adoption of alternative practices.
Key words: biogeochemical modeling; Central Valley, California, USA; conservation tillage; cover
cropping; DAYCENT model; greenhouse gases; low-input management; organic agriculture.
INTRODUCTION
Agriculture is an important source of biogenic
greenhouse gases (GHGs), especially in intensively
managed systems (Cole et al. 1993) as found in the
Central Valley of California, USA. Bemis et al. (2006)
estimated that in California, 8% of the total GHG
emissions originated from agriculture and forestlands.
Alternative management practices, such as winter cover
cropping (i.e., growing a second crop during the winter
that is incorporated in the spring), conservation tillage
(i.e., reducing the intensity of soil disturbance operations), or organic agriculture (here defined as replacing
Manuscript received 3 May 2009; revised 5 October 2009;
accepted 12 October 2009; final version received 14 December
2009. Corresponding Editor: K. K. Treseder.
5 E-mail: [email protected]
inorganic fertilizers by manure, cover crop residues, and
compost), have been proposed to reduce agricultural
GHG emissions significantly below baseline levels and
even to convert agricultural systems into net GHG sinks
(McCarl and Schneider 2001). The latter opens the
opportunity for farmers to actively participate in a
carbon credit market system (Pacala and Socolow 2004).
Before such a market system can be established,
however, a detailed analysis of the GHG emission
reductions from alternative management practices and
the associated uncertainties is necessary. Such an
analysis was, until now, not available for cropping
systems in the Central Valley of California.
It is practically impossible to continuously monitor
GHG fluxes across all possible permutations of crop
rotations, management practices, soils, and microclimates within the Central Valley. Biogeochemical process
models are useful tools to simulate gas exchange for
1805
1806
Ecological Applications
Vol. 20, No. 7
STEVEN DE GRYZE ET AL.
these different permutations (Del Grosso et al. 2006).
These models have been used successfully to predict
changes in soil C at the field scale (Paustian et al. 1997).
Nevertheless, the successful performance of these models
is strongly dependent on whether they were calibrated
for the specific environmental conditions of the systems
under investigation. In a comparative analysis of the
performance of nine different ecosystem models to
simulate seven long-term field sites, Smith et al. (1997)
concluded that model performance was strongly dependent upon (1) whether the models were developed for
soils and conditions similar to the tested field sites and
(2) how well they were calibrated for the site studied.
Similarly, Campbell et al. (2001) concluded that both
EPIC and CENTURY, two commonly used ecosystem
models, were unable to satisfactorily predict long-term
soil organic C (SOC) changes associated with different
management practices in conditions in southern
Saskatchewa, Canada, when no site-specific calibration
was conducted. No biogeochemical process model has
been calibrated specifically for the conditions and
practices in the Central Valley.
Even though results from biogeochemical models are
typically presented without quantification of the uncertainty around the estimates, valid model-based inferences are not possible without an estimate of the
accuracy of predictions (Ogle et al. 2006). The errors
associated with model estimates originate from either
variation within the input data or from the limited
representation of the mechanisms within the model
(Ogle et al. 2006). The first source of uncertainty is
classically quantified by performing multiple model runs
while varying input variables. For example, in a Monte
Carlo analysis, hundreds of simulation runs are carried
out in which the input variables are varied randomly
based on their probability density function (Saltelli et al.
2000). The second error is assessed by confronting
modeled and measured data.
The aims of this study were (1) to calibrate and
validate the DAYCENT model for conventional and
alternative management practices at four experimental
sites in California’s Central Valley, (2) to estimate net
GHG fluxes in these systems, and (3) to quantify the
uncertainty around model predictions due to variability
of input parameters and a limited model representation
of the cropping systems studied.
MATERIALS
AND
METHODS
Model description
The DAYCENT model is the daily time step version
of the well-known CENTURY biogeochemical process
model (Parton et al. 1987, 1994, Metherell et al. 1995).
DAYCENT was developed to simulate ecosystem C and
nutrient dynamics and trace gas fluxes. It includes
submodels for nitrification and denitrification (Parton et
al. 1996, Del Grosso et al. 2000), CH4 oxidation (Del
Grosso et al. 2000), as well as soil water and temperature
(Parton et al. 1998). It is a fully resolved biogeochemical
process model simulating the major processes that affect
soil organic matter (SOM), such as plant production,
water flow, nutrient cycling, and decomposition. The
model simulates SOM C and N stocks, which are
represented by two plant litter pools (i.e., structural and
metabolic litter) and three SOM pools (i.e., active, slow,
and passive SOM). These SOM pools are explicitly
defined by their turnover time: 1–5 years for the active
pool, 20–40 years for the slow pool, and 200–1500 years
for the passive SOM. Nutrient fluxes between pools are
further controlled by rate modifiers dependent upon
moisture, temperature, soil texture, and soil tillage. The
crop submodel simulates crop dry matter production
and yields as a function of light and temperature. The
crop submodel also simulates the influence of biomass
on the soil microenvironment (moisture, temperature,
and nutrients) and the amount and quality of crop
residues returned to the soil after harvest. A variety of
management options may be specified including crop
type, tillage, fertilization, manure and organic matter
addition, planting, harvesting (with variable residue
removal), drainage, and irrigation.
Site descriptions
The Central Valley of California consists of the cooler
Sacramento Valley in the north and the warmer San
Joaquin Valley in the south. Data from long-term
agricultural research experiments in California were
considered to calibrate and validate the DAYCENT
model. In the end, four sites were selected for which
sufficient historical management data and measurements
over time were available. Three sites are located in the
southern Sacramento Valley (the Long-term Research on
Agricultural Systems [LTRAS] project, the Sustainable
Agriculture Farming Systems [SAFS] project, and Field
74), and one site is located in the San Joaquin Valley (the
West Side Research and Extension Center [WSREC]
experiment). Table 1 summarizes the crop rotation
sequences for these experiments.
The LTRAS experiment.—The Long-term Research
on Agricultural Systems (LTRAS) project is an ongoing
long-term field experiment established in 1993 to study
the sustainability of irrigated Mediterranean cropping
systems under conventional and alternative management
practices. It is located on 28.8 ha of land near Winters,
California (38832 0 300 N, 121852 0 2900 W). Two soil types
are present at the LTRAS site: (1) Yolo silt loam (finesilty, mixed, nonacid, thermic Typic Xerothent) and (2)
Rincon silty clay loam (fine, montmorillonitic, thermic
Mollic Haploxeralf ). Clay contents vary from 8% to
19% and sand contents range from 17% to 27% (Chen et
al. 1995).
The LTRAS experiment includes 10 cropping systems
(treatments), but here we focused on the three field corn
(Zea mays L.) and tomato (Lycopersicon esculentum var.
‘‘Halley’’) rotations: (1) conventional management
(using chemical fertilizer and pesticides; CCT); (2) a
system consisting of a legume cover crop preceding
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
1807
TABLE 1. Crop rotation sequences for four long-term field experiments in California, USA, used for model validation.
Site and
cropping system
Year 1
S
Year 2
W
S
Year 3
W
S
W
S
W
(wheat)
(CC)
beans
beans
(CC)
LTRAS (1994–2006)
Conventional
Cover cropping
Organic
tomato
tomato
tomato
(CC)
(CC)
SAFS (1989–2000)
Conventional, 4-year
Conventional, 2-year
Cover cropping
tomato
tomato
tomato
(wheat)
(CC)
safflower
(CC)
corn
corn
WSREC (2000–2006)
Conventional
Cover cropping
tomato
tomato
(CC)
cotton
cotton
(CC)
Field 74 (2003–2006)
Conventional
corn
corn
corn
corn
(CC)
safflower
sunflower
Year 4
chickpea
Notes: Crops in parentheses are grown in the fall and winter growing season. Sequences are repeated continuously, except for the
Field 74 experiment, for which three years were included. The Field 74 conventional plot was planted with wheat in the fall and
winter season before year 1. Blank cells indicate no crop grown during rotation; ellipses indicate years outside of the rotation
period. Abbreviations are: S, spring and summer growing season; W, fall and winter growing season; CC, winter cover crop added.
Three of the sites are located in the southern Sacramento Valley (the Long-term Research on Agricultural Systems [LTRAS]
project, the Sustainable Agriculture Farming Systems [SAFS] project, and Field 74), and one site is located in the San Joaquin
Valley (the West Side Research and Extension Center [WSREC] experiment).
unfertilized corn and followed by conventionally fertilized tomato (LCT); and (3) an organic system with
poultry manure amendments, no chemical fertilizer, and
a legume cover crop grown in the winter of each year
(OCT). The experiment is completely randomized; each
of these treatments is replicated three times on 0.4-ha
plots. Each crop in the two-year system is present each
year. Tomatoes grown under CCT and LCT were
fertilized with 45 kg N/ha at transplanting and 100 kg
N/ha as a side-dress application. In the CCT system,
corn received 45 kg N/ha pre-plant and 160 kg N/ha as a
side dress. The cover crop used as a green manure in the
LCT and OCT systems was sown as a mixture of 24%
pea (Pisum sativum L.) and 76% common vetch (Vicia
sativa L.) by seed mass. Initially, the CCT treatment
used a corn variety that matured in ;185 days (Pioneer
3162), whereas the LCT and OCT treatments used a
short-season corn variety that is planted later and
matured in ;150 days (NC þ 4616). Since 2003, a single
corn cultivar (ST 7570) has been used to accommodate
direct comparisons between standard and conservation
tillage subplots. Details of the experiment and yields
from the first nine years of the experiment are presented
in Denison et al. (2004). More recent results are reported
in Kaffka et al. (2005) and Mitchell et al. (2007a). Soil C
data were collected for all plots at the inception of the
trial in fall 1993 and occasionally thereafter (in 1995,
1998, 1999, 2003, and 2004) (Kong et al. 2005). From
2003, each of the 0.4-ha plots were split into a standard
tillage half and a conservation tillage half. Daily weather
data were available from the on-site weather station.
The SAFS experiment.—The Sustainable Agriculture
Farming Systems (SAFS) project was a large-scale field
experiment at the Agronomy Farm of the University of
California–Davis (38832 0 N, 121847 0 W) conducted from
1989 to 2000. The experiment was established on an 8.1ha site encompassing a Reiff loam (coarse-loamy, mixed,
nonacid, thermic Mollic Xerofluvents) and a Yolo silt
loam (fine-silty, mixed, nonacid, thermic Typic
Xerorthents). Clay contents ranged from 8% to 18%,
and sand contents ranged from 58% to 68%.
For three different cropping systems, sufficient input
data were available for modeling purposes: (1) a
conventionally managed system under a four-year
tomato (var. Brigade), safflower (Carthamus tinctorius
L.), corn, wheat (Triticum aestivum L.), and common
bean (Phaseolus vulgaris L.) rotation, (2) a four-year
cover cropped system under the same crop sequence as
the former system, but with legume cover crops
preceding each summer crop, and (3) a two-year
conventionally managed system under a tomato and
wheat rotation. The cover crop was a mixed culture of
oat (Avena sativa L.) and purple vetch (Vicia benghalensis L.), which was either harvested for hay or
incorporated as a green manure. Fertilizer application
rates varied throughout the experiment. Across all years,
an average of 166 kg Nha1yr1 was applied as
fertilizer in the four- and two-year rotation, conventionally managed treatments, compared to 42 kg N/ha
annually in the cover-cropped treatment. Additional
details of the experimental design are described in Clark
et al. (1998). Soil C was measured in 1988, 1993, 1995,
1998, and 2000 (Doane et al. 2003, Doane and Horwath
2004). Daily weather data were available from the
California Irrigation Management Information System
(CIMIS) station in Davis.
The WSREC experiment.—The University of
California West Side Research and Extension Center
1808
Ecological Applications
Vol. 20, No. 7
STEVEN DE GRYZE ET AL.
(WSREC) experiment in Five Points (36820 0 2900 N,
12087 0 1400 W) was designed to quantify the interactions
of tillage intensity and cover cropping on soil and air
quality. The study was conducted on a 3.2-ha parcel of
Panoche clay loam (fine-loamy, mixed, supernatic,
thermic Typic Haplocambids). Clay contents within
the study area ranged between 25% and 35%, and sand
contents ranged between 35% and 51%. The field
experiment had four tomato (variety ‘‘8892’’) and cotton
(Gossypium hirsutum L. var. ‘‘Riata’’) rotations comparing standard and conservation tillage practices with and
without winter cover cropping. The cover crop used was
a mixture of 30% Juan triticale (Tritosecal Wittm), 30%
Merced ryegrain (Secale cereale L.), and 40% common
vetch (Vicia sativa L.) (by mass) planted in the beginning
of November and chopped mid-March. The standard
tillage systems used tillage operations representative of
California row crops to break down and establish new
beds in the fall of each year. In contrast to the standard
tillage systems, within the conservation tillage system
beds were maintained for the duration of the experiment
and field traffic was restricted to midseason cultivation
within the furrows for weed control in the tomato
systems and undercutting after the harvest in the cotton
systems. In all treatments, cotton was fertilized with an
initial 11 kg Nha1yr1 and later side-dressed with 154
kg Nha1yr1. Tomato was fertilized with 11 kg
Nha1yr1 at the time of transplanting and later sidedressed with 138 kg Nha1yr1. A detailed analysis of
operations in conservation and standard tillage systems
is presented in Mitchell and Tu (2005). Daily weather
data were available from the CIMIS station in Five
Points.
The Field 74 experiment.—In 2003, a field experiment
was established within a 32-ha agricultural field (38836 0
N, 121850 0 W), in order to compare the effects of
standard and conservation tillage on CO2 and N2O
efflux from soils. The site is identified as ‘‘Field 74’’
according to the grower’s numbering system. Three soil
series occur on the site: Myers clay (fine, montmorillonitic, thermic Entic Chromoxererts), Hillgate loam (fine,
montmorillonitic, thermic Typic Palexeralfs), and San
Ysidro loam (fine, montmorillonitic, thermic Typic
Palexeralfs). The site has a shallow water table varying
between 50 and 100 cm depth during the rainy season
from late autumn to early spring. Clay contents ranged
from 11% to 29%, and sand contents ranged from 22%
to 45%.
The field was split into two halves of 16 ha, and
sampling points were established across the field using a
uniform grid with 64-m spacing. Soil properties (e.g.,
sand, clay, SOC, bulk density) were measured at all
sampling points in March 2004. Wheat was grown in the
winter of 2002 and 2003. In 2004, corn was grown.
Ammonium nitrate (UAN-32) was applied initially at a
rate of 50 kg N/ha and approximately 40 days later sidedressed at 150 kg Nha1yr1. In 2005, sunflower was
grown. Fertilizer was side-dressed at a rate of 90 kg N/
ha. During the last year of the experiment (2006), rainfed chickpea was grown without fertilizer application.
Nitrous oxide fluxes were measured using nonsteadystate portable chambers (Hutchinson and Livingston
2002) in 3–15 plots during 51 (standard tillage) or 50
(conservation tillage) campaigns from November 2003
to August 2006. At each sampling point, between one
and four samples were taken at positions in the middle
of the seedbed, middle of the furrow, over the crop row
between plants (during the growing season), and over
the side-dressed band of fertilizer N. However, there was
no consistent efflux pattern correlated with position
across the seedbed (Lee et al. 2009); therefore, fluxes
were averaged at each sample location. Daily weather
data were available from the CIMIS station in Davis.
Rainfall data were available from a tipping bucket
pluviometer on-site.
Modeling approach
Historical runs.—The three SOM pools used in
DAYCENT are conceptual. Therefore, their relative
size cannot be experimentally measured and historical
runs were performed to initialize the size of these pools
at the start of the agricultural experiments considered.
The historical runs represent the average history of land
use and management in the Central Valley. Therefore,
the history was identical for all experimental sites. The
climate history and soil types, however, were based on
the conditions of the individual experimental sites. We
assumed five broad periods in the history of land use and
management in the Central Valley: (1) native grassland
(between 0 and 1869; run until equilibrium), (2)
emergence of agriculture (between 1870 and 1920), (3)
introduction of irrigation (between 1921 and 1949), (4)
introduction of inorganic fertilizer (between 1950 and
1969), and (5) modern agriculture (from 1970). For the
first period, a medium-productivity grassland with a
mixture of annuals and perennials was simulated, with a
growing season from November until the end of April.
We included low-intensity grazing, affecting 10% of the
live shoots and 5% of the aboveground dead biomass.
The 1870 simulation years were sufficient to attain
equilibrium in all modeled C pools. The average
modeled C input to the soil at equilibrium was 165 6
13 g C/m2, which is 83% of the reported mean C input
values of ;200 g dry matterm2yr1 across grasslands
in the Central Valley (Bartolome and McClaran 1992,
Valentini et al. 1995, Potthoff et al. 2005). In the second
period (the emergence of agriculture) we simulated a
rain-fed, low-input winter wheat system with minimal
disturbance of the soil and a fallow period every five
years. In the third period (pre-modern agriculture), we
introduced irrigation and gradually diversified the crops
to include summer-grown corn. In the fourth period,
inorganic fertilizer was introduced. We used historical
records from the USDA to simulate the increase in the
amount of fertilizer used between 1950 and 1969. During
this period, we introduced tomatoes and increased the
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
degree of soil disturbance. In the last period, between
1970 and 1996, we simulated a random wheat, corn, and
tomato rotation with high-intensity tillage. Similarly to
the period before, the increasing use of fertilizer was
simulated based on historical records.
Model calibration and calculations.—We simulated all
experiments from their date of establishment until the
year 2006, except for the SAFS experiment, which was
discontinued in 2000. All annual variations in management conditions, such as planting date, harvesting date,
crop residue incorporation, and/or fertilizer amount,
were represented in the simulations. The soil microclimate was verified against measured soil temperature and
moisture contents (available at Field 74 and LTRAS in
Yolo County). If necessary, parameters such as the
saturated hydraulic conductivity, the volumetric water
content at field capacity or wilting point, and the
minimal soil water content were adjusted. Secondly, we
verified the relative size of the live biomass compartments (roots, shoots, and harvestable portion) using
published and measured root : shoot ratios, and harvest
indices (ratio of harvestable part over total aboveground
biomass). The C:N ratios of each of these biomass
compartments were verified with measured and literature values. These values were confirmed with model
results from the DSSAT/CERES plant growth models,
using average climate and soil conditions of Yolo
County (Jones et al. 2003). Only after the modeled
plant indices and ratios were correct, we adjusted the
photosynthetic rate parameter to match the modeled
harvestable biomass values with the recorded average
yield data at the different sites. No site-specific
adjustments to the crop parameterization were necessary, except for the maximal harvest index and maximal
rate of photosynthesis in corn and tomato cultivars,
which had to be adjusted to reflect the differences in the
length of the growing season for the cultivars used
across the sites.
Once the live biomass was simulated correctly, we
checked the sizes of the dead biomass and litter layer
compartments with measured data (at LTRAS and Field
74) and literature values. If necessary, parameters
controlling root or shoot death were adjusted. Next,
we verified soil C dynamics and adjusted the simulated
tillage intensity until changes in soil C corresponded to
those observed. Four different types of tillage events,
decreasing in intensity, were needed to simulate the
variety of tillage management practices in all the
experiments considered: a standard tillage type, a
conservation tillage type, a cover crop incorporation
type, and a within-season cultivation type. Standard
tillage was simulated by scheduling a high-intensity
tillage pass before planting and one post-harvest. This
single standard tillage pass scheduled in the model in
fact represents multiple passes done by producers,
including deep ripping, stubble disking, shallow disking,
grading, and listing beds. During the growing season, all
mechanical weed suppression was simulated by sched-
1809
uling one within-season cultivation pass during the
model run. To simulate conservation tillage, a lowintensity tillage pass in the spring and one post-harvest
sufficed. The conservation tillage passes were 30% less
intense than the standard tillage passes. In contrast to
the conventional tillage system, no within-season cultivation pass was scheduled in the conservation tillage
systems. For the cover-cropped treatments, one cover
crop incorporation pass was scheduled in between the
cover crop and main-crop growing seasons. It was
necessary to reduce the impact of the cover crop
incorporation type on SOM decomposition compared
to the other tillage types.
Last, we verified modeled N2O fluxes with measured
data. Daily N2O flux measurements were available for
Field 74 (Lee et al. 2009). If necessary, specific
parameters controlling soil moisture and parameters
highly influencing N2O production (e.g., existence of a
soil water table and minimal volumetric soil water
content per layer) were further adjusted.
Within each experiment, we calculated the net soil
GHG flux as
GHG ¼ 44
3 DSOC þ 296 3 ½N2 O þ 23 3 ½CH4 ð1Þ
12
where GHG is the net soil GHG flux in megagrams of
CO2 equivalents (CO2-eq; the amount of CO2 that
would have the same global warming potential as a
given mixture of greenhouse gases) per hectare per year;
DSOC is the change in SOC in megagrams of C per
hectare per year, [N2O] is the flux of N2O in megagrams
of N2O per hectare per year, and [CH4] is the flux of
CH4 in megagrams of CH4 per hectare per year. The
DAYCENT model does not simulate CH4 emission,
only CH4 oxidation; therefore all CH4 fluxes are
negative. The radiative forcing constants or global
warming potentials (GWPs) from IPCC (2001) were
used.
Uncertainty estimation, model performance, and statistical analysis.—For LTRAS, plot-level information of
soil properties and crop management practices was
available for each separate field replicate (n ¼ 3).
Therefore, each plot was simulated individually, and
we reported the standard deviation around the resulting
estimated C and N fluxes of the different field replicates.
For the other field experiments, no data on individual
field replicates were used, only treatment averages and
standard deviations. A Monte Carlo simulation approach was used for these sites to estimate variances for
modeled results. We generated ;100 different input data
files by randomly varying input parameters simultaneously using univariate normal distributions with
averages and standard deviations from measured data.
We then calculated the average and standard deviation
of the modeled outputs based on each of these input
data files.
The total mean square deviation was divided into
different components according to Gauch et al. (2003):
1810
Ecological Applications
Vol. 20, No. 7
STEVEN DE GRYZE ET AL.
TABLE 2. Literature (L) and model (M) values of critical plant parameters for the seven crops modeled in this study.
Beans
Parameter
L
M
C:N ratio
Harvested aboveground 10
biomass
Non-harvested
15
aboveground biomass
Roots
13
Harvest index
Shoot : root ratio
Corn
0.60
4.5
L
M
Safflower
Sunflower
Tomato
L
L
L
M
M
M
Wheat
L
M
Cotton
L
M
9
30
35
18
19
14
14
22
27
22
22
N/A
13
60
69
38
40
40
36
24
29
90
87
N/A 51
13
55
59
N/A
77
76
74
41
46
34
34
N/A 22
0.61
4.7
0.50
4.3
0.53
4.0
0.25
4.0
0.27
3.6
0.30
6.6
0.30
6.2
0.45
4.2
0.53
4.2
0.50
4.5
0.50 0.60
5.8 5.1
8
0.60
4.5
Note: N/A indicates a missing literature value.
nonzero intercept (or squared bias), non-unity slope,
and lack of correlation. These components have distinct
and transparent meanings. The nonzero intercept (or
squared bias) component represents the part of the total
deviation due to a nonzero intercept in the relation
between predicted and measured values; the non-unity
slope component represents the part of the total
deviation due to a slope difference; the lack of
correlation component represents the contribution of
the total deviation due to random scatter in both
predicted and measured variates.
Annual GHG emissions were analyzed with a mixedANOVA model in which crop and tillage treatments (for
LTRAS and WSREC) and their interactions were
considered as fixed effects and year was considered as
a random effect. In addition, year was also a repeatedmeasures variable with the plot replicate (at LTRAS) or
the Monte Carlo replicate (at the other sites) as subjects
(Littell et al. 2006). The variance around annual
emissions was partitioned by using the following model
for individual GHG emissions:
Yijk ¼ l þ ai þ wj þ eijk
ð2Þ
where Yijk is the annual GHG flux of treatment i, year j,
and replicate k, l is the overall mean (fixed effect), ai is
the mean of treatment i (fixed effect), wj is the influence
of season j (random effect, mainly caused by weather
variations), and eijk is the residual of plot or Monte
Carlo replicate k (random effect). The average of
treatment i over all years and plots can then be expressed
as
l þ ai þ
Yi ¼
no:X
years no:
plots
X
j¼1
wj þ eijk
k¼1
ðno: yearsÞðno: plotsÞ
:
ð3Þ
The variance around the treatment mean becomes
r2
no: plots
no: years
r2w þ
varðYi Þ ¼
ð4Þ
where r2w is the interannual variance and r2 is the
residual variance. The proportion of the variance caused
by interannual differences is then
r2w
r2w þ
r2
no: plots
:
ð5Þ
RESULTS
Model validation
After calibration, modeled crop parameters were
comparable with typical values found in the literature
(Table 2). Per crop and field site, mean yields were
predicted reasonably well (Fig. 1), with variations
explained by the model ranging from 86% to 94%
(Table 3). In addition, SOC was predicted well (Fig. 2),
with variations explained by the model ranging from
69% to 87%, except for Field 74 (Table 3). Modeled
yearly N2O fluxes per crop were in the same order of
magnitude as the yearly fluxes reported in the literature
(Table 4).
For the LTRAS experiment, the model explained
;86% of the variation in measured yields, with most of
this nonexplained variation coming from the lack of
correlation (74%), indicating that no large bias existed
(Table 3). Within crops and across seasons and field plot
replicates, yield predictions were accurate at LTRAS
(Fig. 1). In addition, the modeled variability in yields
was in the same range as the measured variability. The
DAYCENT model accounted for 69% of the variations
in modeled SOC. Approximately 24% of the nonexplained variation was due to a non-unity slope,
indicating that SOC levels were slightly overpredicted
at higher SOC levels (Fig. 2).
For the SAFS experiment, the model explained ;92%
of the variation in yields; the 8% not explained came from
the lack of correlation component (96%) (Table 3).
Again, the modeled variability in yields was in the same
range as the measured variability (Fig. 1). Approximately
83% of the variation in SOC content was predicted by the
model. Similar to the LTRAS experiment, the nonexplained variation was partially due to a non-unity slope
(21%), as well as a nonzero intercept (23%). The nonzero
intercept indicates that SOC contents were slightly
overestimated both at small and larger SOC levels,
whereas the non-unity slope indicates that this bias was
greater for larger SOC levels (Fig. 2).
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
1811
FIG. 1. Modeled vs. measured yields by crop but across various years, replicates, and treatments at four long-term field
experiments in California, USA. For the Long-Term Research on Agricultural Systems (LTRAS) site, data per replicate plot that
were available were all modeled separately. For the Sustainable Agriculture Farming Systems (SAFS) and the West Side Research
and Extension Center (WSREC) sites, only means per treatment were available; the error bars show 6SD around modeled results,
as calculated by a Monte Carlo analysis. For the Field 74 site, both the mean and the SDs of yields were available; SDs are
indicated by horizontal error bars. The dot-dashed line is the 1:1 line. All yields are expressed as oven-dry mass, except for
tomatoes, which are expressed as fresh mass, assuming a moisture content of 94% (Perez-Quezada et al. 2003).
For WSREC, 94% of the variation in yields was
modeled. Again, the nonexplained variation was due to
a lack of correlation (96%) (Table 3). The modeled
uncertainties around the predicted yields were approximately one-third the size of the measured variation in
yields among field replicates (0.5 Mg/ha compared to 1.5
Mg/ha; Fig. 1). Generally, both modeled and measured
variation in SOC contents were substantial, which is
attributed to a limited number of SOC measurements on
the cover-cropped treatments (Fig. 2). In addition, we
were not able to simulate the 30% increase in SOC due
to conservation tillage across years in the conservation
tillage and cover-cropped treatments of the WSREC
experiment (Fig. 2).
For the Field 74 experiment, 92% of the variation in
yield was modeled; of the nonexplained variation, 91%
was coming from a lack of correlation (Table 3).
However, the measured variation in yields was larger
than the uncertainty around the modeled yields (Fig. 1).
No differences in either measured or modeled SOC were
1812
Ecological Applications
Vol. 20, No. 7
STEVEN DE GRYZE ET AL.
TABLE 3. Model performance statistics for predicting yields and soil organic carbon values at the four long-term field experiments
in California.
Prediction of yield
Site
LTRAS
SAFS
WSREC
Field 74
Prediction of soil organic carbon
Partitioning of the MSD
Partitioning of the MSD
Variation
Variation
explained
Nonzero
Non-unity
Lack of
explained
Nonzero
Non-unity
Lack of
by model (%) intercept (%) slope (%) correlation (%) by model (%) intercept (%) slope (%) correlation (%)
86
92
94
92
13
0
1
4
13
4
3
5
74
96
96
91
69
83
87
6
6
23
6
27
24
21
63
28
70
56
31
45
Notes: Definitions of the different partitions of the mean square deviation (MSD) are provided in Materials and methods:
Uncertainty estimation and model performance. Site abbreviations are: LTRAS, the Long-term Research on Agricultural Systems
project; SAFS, the Sustainable Agriculture Farming Systems project; WSREC, West Side Research and Extension Center
experiment.
found across the seasons or treatments (Fig. 2).
Differences in means were well within the error range.
Therefore, the portion of the variation in SOC explained
was small (6%). Generally, the range in modeled average
daily N2O fluxes at the Field 74 experiment was
comparable to the range in measured daily fluxes (Fig.
3). However, the solitary N2O flux peak measured on 22
May 2006 was not predicted by the model. In addition,
the model underestimated N2O emissions during May
and June of 2004. The modeled variability around daily
N2O fluxes was in general smaller than the measured
variability (Fig. 3).
Simulated greenhouse gas emissions
and mitigation potentials
At the LTRAS site, no significant change in SOC
levels was modeled for the standard and conservation
tillage treatments (Table 5). However, SOC levels
increased substantially when cover cropping or organic
management were implemented, regardless of whether
standard or conservation tillage was practiced. Annual
N2O fluxes were smaller in the conservation tillage
treatment than in the standard tillage treatment. Within
both the standard and conservation tillage treatments,
simulated annual N2O fluxes followed the order: cover
cropped , organic , conventional. Methane fluxes
among treatments followed a similar pattern as the N2O
fluxes. In both standard and conservation tillage plots,
the net soil GHG flux was greatest for the conventional
treatments, without any cover cropping or organic
management, smaller for the cover cropped and smallest
for the organic treatments. Conservation tillage management did not significantly change the net soil GHG
flux compared to standard tillage practices (Table 6).
Cover cropping decreased the net soil GHG flux by 1072
6 272 kg CO2-eqha1yr1. Organic management
decreased the net soil GHG flux by 4577 6 272 kg
CO2-eqha1yr1. Interannual differences among GHG
fluxes accounted for ;40–70% of the total variance.
For the SAFS experiment, the cover-cropped treatment sequestered ;577 6 21 kg Cha1yr1 more in
SOC compared to the two-year and four-year conventional treatments, which had similar SOC sequestration
rates. Cover cropping decreased N2O emissions with
;0.18 6 0.02 kg Nha1yr1. Methane fluxes were
similar in the conventional four-year rotation and covercropping treatments, but significantly smaller in the
conventional two-year rotation. The net soil GHG flux
for the cover cropping treatment, 2921 6 292 kg CO2eqha1yr1, was smallest of the three treatments. The
net soil GHG flux of the conventional four-year
rotation, 515 6 292 kg CO2-eqha1yr1, was larger
than the conventional two-year rotation, 925 6 298 kg
CO2-eqha1yr1. Cover cropping reduced the net soil
GHG flux by 2201 6 82 kg CO2-eqha1yr1 compared
to the conventional treatments. Approximately 90% of
the variance of GHG emissions was caused by
interannual differences.
In the WSREC experiment, adding a cover crop led to
a much larger simulated increase in SOC, 752 6 10 kg
Cha1yr1, than adopting conservation tillage, 66 6 10
kg Cha1yr1. Cover cropping increased annual N2O
emissions in both standard and conservation tillage
treatments. Averaged over both tillage treatments, cover
cropping increased annual N2O emissions by 0.55 6
0.03 kg Nha1yr1. Conservation tillage decreased the
net soil GHG flux with 336 6 47 kg CO2-eqha1yr1,
whereas cover cropping decreased the net soil GHG flux
with 2499 6 47 kg CO2-eqha1yr1. Approximately
91% and 82% of the variance of annual SOC differences
and N2O emissions, respectively, was explained by
interannual differences, while this was only 38% for
the variance of annual CH4 emissions.
At Field 74, we simulated a small but significant SOC
sequestration rate of 128 6 28 kg Cha1yr1 in the
conservation tillage treatment. Conservation tillage did
not significantly affect N2O emissions, while conservation tillage did decrease CH4 fluxes with ;0.20 6 0.05
kg Cha1yr1. Conservation tillage decreased the net
soil GHG flux with 550 6 123 kg CO2-eqha1yr1 less
compared to the standard tillage treatment. Interannual
differences explained ;50% of the variance of annual
SOC differences and N2O emissions and ;20% in
annual CH4 emissions.
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
1813
FIG. 2. Modeled vs. measured soil organic carbon (SOC) levels by treatments across various crops and years at four long-term
field experiments in California. For the Long-Term Research on Agricultural Systems (LTRAS) site, data per replicate plot that
were available were all modeled separately. For the Sustainable Agriculture Farming Systems (SAFS), only averages per treatment
were available, and the vertical error bars show 6SD around modeled results, as calculated by a Monte Carlo analysis. For the
West Side Research and Extension Center (WSREC) and Field 74 sites, both mean and the SD of soil organic carbon were
available; SDs are indicated by horizontal error bars. The dot-dashed line is the 1:1 line.
DISCUSSION
Alternative agricultural management practices, such
as conservation tillage, winter cover cropping, or
organic farming, have been proposed as ways to reduce
soil GHG emissions from cropping systems. Despite
their potential for atmospheric GHG mitigation, these
practices have been only very limitedly adopted by
international carbon offset protocols and standards,
mainly due to concerns about the large uncertainties
around estimates of GHG mitigation potentials and cost
of sound measurements. Biogeochemical models are the
tool of choice to minimize measurement uncertainty in a
cost-effective way and extrapolate results from a limited
set of field experiments to a large geographical region
with varying climatic conditions. However, calibrating a
biogeochemical process model in California is challenging due to the great diversity of crops, cropping systems,
microclimates, and soil conditions within the state.
Because carbon trading could form a source of revenue
for California farmers, an urgent need has emerged to
collect data from agricultural experiments and employ
these data to calibrate biogeochemical process models
for California agriculture.
1814
Ecological Applications
Vol. 20, No. 7
STEVEN DE GRYZE ET AL.
TABLE 4. Crop-wise comparison of modeled annual N2O emissions (mean 6 SD) from four long-term field experiments in
California with measured annual N2O emissions from the literature.
Annual N2O emissions from this study
Site
N2O emission
(kg Nha1yr1)
Corn
Field 74
SAFS
LTRAS
4.1 6 0.7
0.6 6 0.3
3.0 6 0.1
Cotton
WSREC
3.6 6 0.2
Sunflower
Tomato
Field 74
WSREC
LTRAS
SAFS
Field 74
SAFS
3.2
4.4
4.2
2.1
1.3
1.9
Crop
Wheat
6
6
6
6
6
6
0.7
2.1
0.4
0.4
0.2
0.5
Annual N2O emissions from the literature
Location
N2O emission
(kg Nha1yr1)
Source
Belgium
Germany
Madison, USA
Colorado, USA
Costa Rica
France
Australia
Pakistan
Germany
Northern China
1.5
2.1
3.6–5.2
4.0
7.1
11.0
1.6–2.6
3.6
9.4–12.9
5.5
Goossens et al. (2001)
Mogge et al. (1999)
Cates and Keeney (1987)
Hutchinson and Mosier (1979)
Weitz et al. (2001)
Jambert et al. (1997)
Rochester (2003)
Mahmood et al. (2000)
Flessa et al. (1995)
He et al. (2000)
0.7–1.2
1.0
3.5
Flessa et al. (1998)
Kaiser and Heinemeyer (1996)
Kaiser et al. (1998)
Germany
Germany
Germany
Notes: Only studies with measurements for at least 300 days/year and conventional fertilization practices were retained. Site
abbreviations are: LTRAS, the Long-term Research on Agricultural Systems project; SAFS, the Sustainable Agriculture Farming
Systems project; WSREC, West Side Research and Extension Center experiment.
We used data from a number of long-term field
experiments to calibrate the DAYCENT model. The
calibration sites encompass a wide range of management
practices (standard and conservation tillage management, winter cover cropping, and organic farming) and
crops (beans, corn, cotton, safflower, sunflower, tomatoes, and wheat). The soils and climatic conditions at the
sites are representative of California’s Central Valley. To
retain the model as geographically widely applicable, the
crop parameterization and model input files were kept as
general as possible and non-site dependent. The
calibration process was performed in a sequential
manner. First, parameters related to soil temperature
and soil moisture dynamics were adjusted based on
temperature and moisture measurements outside of the
growing season. Secondly, all biomass growth parameters were adjusted based on recorded yields and plant
parameters found in the literature. Third, parameters
related to decomposition of dead plant biomass were
adjusted until measurements of the litter layer corresponded to modeled values. Fourth, parameters related
to the impact of tillage on SOC decomposition were
adjusted until measured changes in SOC corresponded
with modeled changes in SOC. Last, some specific soil
moisture parameters affecting N2O production were
altered based on observed daily measurements of N2O
fluxes. Although model calibration remains a subjective
procedure, given the large number of parameters to be
calibrated for the DAYCENT model and the often dual
or ambiguous effects of changes in parameters, the
sequential manner of calibrating model parameters
minimizes this subjectivity.
The GHG flux changes calculated and reported in this
study are net soil GHG fluxes and not comprehensive
cropping system emissions. The latter would require a
rigorous life cycle analysis that includes the accounting
of emissions from manufacturing of farm inputs such as
fertilizers or fuel use during farm operations. A life cycle
analysis is, however, beyond the scope of the current
paper.
After a careful calibration, the DAYCENT model
predicted mean yields of most crops and sites satisfactorily. In contrast, the standard deviations around
measured yields (based on field replicates) were underestimated by the model (quantified in a Monte Carlo
analysis). The model underestimated the variability of
yields for corn and wheat crops in the Field 74
experiment and for cotton at the WSREC site. For the
latter, the standard deviation of simulated yields was
approximately one-third the size of the observed
standard deviation of cotton yields. We attribute this
to several factors. First, each model is always a
simplification of reality; a substantial amount of factors
that may influence crop yield are not simulated by the
model (e.g., pests, seedling emergence problems, micronutrient deficiencies, temperature at anthesis, fruit set,
etc.). Second, our Monte Carlo analysis did not take
into account variations in management (such as
fertilization amounts or exact planting or harvesting
dates). Integrating variations in management in an
uncertainty analysis is challenging since crop and soil
management are strongly correlated with weather.
Finally, some processes are naturally stochastic. For
example, the harvest index of cotton is very variable and
unpredictable under water stress conditions. The
DAYCENT model is deterministic and will therefore
underestimate the variability associated with such
processes. Although modeled variabilities were smaller
than observed variabilities, the differences in observed
variability among sites were well reflected by differences
in modeled variability. For example, the variability of
both observed and modeled SOC contents was highest at
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
1815
FIG. 3. Modeled and measured N2O emissions vs. time for the standard and conservation tillage treatments at the Field 74 site.
The gray area around the model results shows 6SD of the mean, as calculated by a Monte Carlo analysis. The solitary N2O flux
peak measured on 22 May 2006 was not predicted by the model.
Field 74. This is attributed to the well-known considerable textural variability at this site (Lee et al. 2006),
compared to the other sites. This correspondence
between observed and modeled variabilities demonstrates that the DAYCENT model captures the most
important sources of variability among sites, even if not
all sources of variability are simulated.
Generally, simulated SOC values corresponded well
with measured SOC values. The model explained
between 69% and 87% of the variance in SOC, except
for the Field 74 site, for which the explained variance
was smaller. The model overpredicted SOC levels by
;10% for the cover-cropped treatment of SAFS and the
organic treatment of LTRAS. Since the amount of C
input (plant residues or manure) to the SOC was
simulated correctly, DAYCENT slightly underestimated
SOC decomposition rates, especially when C inputs were
high. The variability of measured and simulated SOC
values was substantial at Field 74, with coefficients of
variation of ;25%. This variability is again attributed to
the textural variability at this site (Lee et al. 2006), but
also to the smaller size and non-replicated nature of this
experiment compared to the other experiments. The
model simulated that conservation tillage did not
increase SOC significantly at LTRAS, while significant
SOC sequestration rates were modeled at WSREC (66 6
10 kg Cha1yr1) and Field 74 (128 6 28 kg
Cha1yr1). The simulated changes in SOC are smaller
than values reported in the literature for conservation
tillage management outside of California. Franzluebbers
1816
Ecological Applications
Vol. 20, No. 7
STEVEN DE GRYZE ET AL.
TABLE 5. Changes in soil organic carbon (DSOC) and annual greenhouse gas (GHG) emissions for various alternative agricultural
management treatments at four long-term field experiments in California.
Site and treatment
or property
LTRAS
Standard tillage
Standard tillage and cover cropping
Standard tillage and organic
Percentage of variation due
to seasonal differences
Conservation tillage
Conservation tillage and cover cropping
Conservation tillage and organic
Percentage of variation due
to seasonal differences
SAFS
Conventional 4-year rotation
Conventional 2-year rotation
Cover cropping
Percentage of variation due
to seasonal differences
WSREC
Standard tillage
Standard tillage and cover cropping
Conservation tillage
Conservation tillage and cover cropping
Percentage of variation due
to seasonal differences
Field 74
Standard tillage
Conservation tillage
Percentage of variation due
to seasonal differences
DSOC
(kg Cha1yr1)
N2O
(kg Nha1yr1)
CH4
(kg Cha1yr1)
Net soil GHG
(kg CO2-eqha1yr1)
95 6 46
315 6 46
1324 6 46
74%
3.18 6 0.10
2.60 6 0.10
3.02 6 0.10
37%
1.52 6 0.02
1.44 6 0.02
1.49 6 0.02
46%
1081 6 192
9 6 192
3496 6 192
72%
47 6 87
321 6 87
1279 6 87
65%
3.01 6 0.18
2.21 6 0.18
2.98 6 0.18
53%
1.51 6 0.05
1.46 6 0.05
1.49 6 0.05
68%
1182 6 391
192 6 391
3349 6 391
61%
407 6 77
436 6 78
999 6 77
94%
2.21 6 0.08
1.54 6 0.08
1.70 6 0.08
80%
1.62 6 0.02
1.44 6 0.02
1.63 6 0.02
89%
515 6 292
925 6 298
2921 6 292
96%
38
38
38
38
3.44 6 0.10
4.01 6 0.10
3.26 6 0.10
3.79 6 0.10
82%
2.00 6 0.02
1.93 6 0.02
1.99 6 0.02
1.94 6 0.02
38%
128 6 20
256 6 20
51%
2.62 6 0.08
2.43 6 0.08
49%
1.54 6 0.04
1.33 6 0.04
19%
90 6
677 6
9 6
729 6
91%
1866 6
675 6
1487 6
969 6
92%
147
147
147
147
700 6 87
150 6 87
43%
Notes: Positive values for DSOC indicate an increase in SOC. For N2O and CH4 fluxes, positive values indicate a net flux from
the soil to the atmosphere. Carbon dioxide equivalents is a quantity that describes the amount of CO2 that would have the same
global warming potential as a given mixture of greenhouse gases. For N2O, a radiative forcing constant of 296 was used; for CH4, a
radiative forcing constant of 23 was used. Standard errors of the group mean are based on a mixed ANOVA model; the standard
error tests the null hypothesis of whether the absolute value is equal to zero. Proportions of the standard error are those that are
due to differences among different seasons rather than differences among the replicates within the season, based on a mixed
ANOVA model. For example, 74% of the standard error of 46 is due to seasonal differences for the standard tillage SOC effects at
LTRAS. Site abbreviations are: LTRAS, the Long-term Research on Agricultural Systems project; SAFS, the Sustainable
Agriculture Farming Systems project; WSREC, West Side Research and Extension Center experiment.
(2005) reported increases of 420 6 460 kg Cha1yr1 in
the southeastern USA, the review of West and Post
(2002) reported an increase of ;600 6 100 kg
Cha1yr1 globally, and the review of Six et al.
(2004) reported 200 kg Cha1yr1 for a 10-year-old
no-tillage system in a humid climate. However, most of
these numbers are based on systems in which tillage is
almost completely eliminated. In California systems,
conservation tillage systems are still fairly intensive. In
general, the number of tillage passes is halved in
conservation tillage systems in California (Mitchell et
al. 2007b, 2009). In the Sacramento Valley, tillage passes
are reduced from 10 to ;5 and in the San Joaquin
Valley from 20 to 10. In addition, the intensity of an
individual conservation tillage pass is smaller than the
intensity of an individual conventional tillage pass.
Compared to highly reduced tillage systems elsewhere,
potential increases in SOC due to conservation tillage
will be modest in California. In contrast to conservation
tillage, adding a cover crop during the winter led to an
increase in SOC of 220 6 65 kg C/ha at LTRAS, 577 6
21 kg C/ha at SAFS, and 752 6 10 kg Cha1yr1 at
WSREC. These values are close to the mean SOC
increase for no-tillage winter cover-cropping systems
reported by Franzluebbers (2005), 530 kg Cha1yr1,
and are somewhat greater than the 100–300 kg
Cha1yr1 range reported by Lal et al. (1998).
Simulated annual N2O fluxes are within the range
reported by other authors, both for the absolute values
(Table 4) and for the relative proportion of N2O to the
total GHG flux (Bemis et al. 2006). The relatively small
N2O flux for corn at SAFS (0.6 kg Nha1yr1) was
related to the relatively low fertilizer rates at this site.
The model simulated that all conventionally managed
systems had a positive net soil GHG flux, despite an
increase in SOC, due to the dominance of N2O
emissions. The conclusion that N2O emissions are
dominating the greenhouse gas contribution of agricultural systems in California was also reported by Bemis et
al. (2006). They found that N2O emissions accounted for
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
1817
TABLE 6. Impacts of alternative management treatments on soil organic carbon (SOC), N2O and CH4 fluxes, and net soil
greenhouse gas (GHG) flux at four long-term field sites in California.
Site and treatment
DSOC
(kg Cha1yr1)
DN2O
(kg Nha1yr1)
LTRAS
Conservation tillage
Cover cropping Manure application 36 6 31
220 6 65
1229 6 65
0.07 6 0.08
0.58 6 0.14
0.16 6 0.14
0.00 6 0.01
0.09 6 0.03
0.04 6 0.03
SAFS
Cover cropping
577 6 21
0.18 6 0.02
WSREC
Conservation tillage
Cover cropping
66 6 10
752 6 10
Field 74
Conservation tillage
128 6 28
DCH4
(kg Cha1yr1)
DGHG
(kg CO2-eqha1yr1)
Contribution of
DN2O to DGHG (%)
168 6 131
1072 6 272
4577 6 272
20
25
2
0.10 6 0.01
2201 6 82
4
0.20 6 0.03
0.55 6 0.03
0.00 6 0.01
0.06 6 0.01
336 6 47
2499 6 47
28
10
0.19 6 0.11
0.20 6 0.05
550 6 123
16
Notes: Values are mean differences 6 SE relative to conventional practices, as calculated by a statistical contrast. These values
are calculated as statistical estimates of a linear mixed model. Because the experiments are not fully balanced across all treatments
and crops, the value of the estimate may be different from the difference of the individual averages. Positive values of DSOC
indicate that the treatment increases SOC. For N2O and CH4 fluxes, positive values indicate that the treatment increases the
emissions. Site abbreviations are: LTRAS, the Long-term Research on Agricultural Systems project; SAFS, the Sustainable
Agriculture Farming Systems project; WSREC, West Side Research and Extension Center experiment.
The effect of low-input management and manure application is calculated only based on values from the standard tillage
treatment, since the standard tillage treatment has been implemented for a longer time.
50% of the total net soil GHG flux. Modeled average
daily N2O fluxes over time are within the range of
measured values from Field 74 (Fig. 3). However, during
the spring (May–June) of 2004, the model clearly
underestimated N2O emissions in the conservation
tillage treatment. We attribute this to the increase in
bulk density and an associated decrease in pore space
over time in the conservation tillage system (Lee et al.
2006). The DAYCENT model does not simulate
compaction or loosening of the soil, and bulk densities
are assumed to remain constant during the experiment.
Modeled N2O emissions in this period were underestimated as they were based on a smaller bulk density than
the actual value. The apparent variability of N2O
emissions was not at all times accurately estimated
using DAYCENT. We attribute this to the limited
representation of the mechanisms involved in N2O
emissions within the model rather than an incorrect
representation of the variability of the input data. A
solitary peak of emission in the spring of 2006 (22 May),
occurring in both the standard and conservation tillage
treatments, could not be simulated. The measurement
error around this peak was quite large (30 6 46 and 76
6 83 g Nha1d1 in the standard and conservation
tillage treatments, respectively). This peak occurred a
day after a mild rain (7 mm) during the growth of the
chickpea crop, to which no fertilizer was applied. Since
the rainfall was very mild and occurred after a dry
period of 20 days, DAYCENT could not model the
resulting short-term soil saturation conditions which,
most likely, triggered the peaks in N2O emissions.
The simulated difference in net soil GHG flux
mitigation potential of alternative practices compared
to conventional practices (Table 6) was the smallest or
even insignificant for conservation tillage (336 6 47 and
550 6 123 kg CO2-eqha1yr1), followed by cover
cropping (1072 6 272, 2201 6 82, and 2499 6 47 kg
CO2-eqha1yr1) and the greatest under manure
application (4577 kg CO2-eqha1yr1). The large
difference in net soil GHG flux after manure application
is explained by the significant amount of organic carbon
that is added to the system every year. Manure
application is a much more effective way of increasing
soil carbon content compared to conservation tillage or
cover cropping. However, not only should the total
mitigation potential be of interest, it is also important to
know how much of the total reduction is attributed to
increases in soil C vs. reductions in N2O fluxes. The
capacity of a soil to store C is limited (VandenBygaart et
al. 2002, Six et al. 2004), and if the proper soil
management is not maintained, all or part of the
sequestered C will be eventually released again to the
atmosphere. In contrast, reductions in N2O emissions
are permanent (VandenBygaart et al. 2004, Smith et al.
2007). We found that although N2O was typically the
most important gas contributing to the total net soil
GHG flux (Table 5), changes in SOC were key to
achieving a negative net soil GHG flux for mitigation
potentials of alternative practices (Table 6). For
example, cover cropping at LTRAS decreased the net
soil GHG flux by ;1072 6 272 kg CO2-eqha1yr1,
but only 25% of this reduction was due to reductions in
N2O fluxes. Similar or smaller contributions in decreases
of N2O fluxes to the total mitigation in net soil GHG
flux were observed for the other treatments and sites. At
WSREC, winter cover cropping with conservation
tillage led to an increase in N2O emissions of ;0.55 6
0.03 kg Nha1yr1. The model simulated that soil
1818
STEVEN DE GRYZE ET AL.
mineral N levels were higher for the cover-cropped
treatments than the non-cover-cropped treatments
during late spring (before planting) and during the
decomposition of crop residue during the fall. The
higher mineral N levels led to greater nitrification and
denitrification during these periods in the cover-cropped
treatments.
We found that year-to-year differences in weather or
management dominated the total variance around
predictions of annual net soil GHG flux. Consequently,
an error analysis of predicted GHG mitigation potentials
will have to take interannual variability into account. It
can be expected that the error in estimating GHG
mitigation related to interannual differences will decrease
with the duration of the practice. Therefore, a sound
quantification of this error is necessary to determine the
contract duration and the risks associated with short-term
vs. long-term adaptation of alternative practices for
mitigation of GHG emissions. To extrapolate these
results to a regional scale, it will be necessary to simulate
GHG fluxes across a range of soils, land uses, and
climates based on a calibrated model such as the one
presented here. Such a simulation will also have to
quantify the error around the predicted mitigation
potentials.
In conclusion, the results from this study suggest that
the potentials for GHG mitigation by implementing
alternative agricultural practices are smallest for conservation tillage, larger for winter cover cropping, and
the greatest for organic inputs (cover crops þ manures).
Within the limitations discussed here, the calibrated
DAYCENT model can be used as a tool to forecast
GHG fluxes in the alternative cropping systems in
California, but only when combined with a rigorous
error analysis.
ACKNOWLEDGMENTS
We acknowledge funding by the California Energy
Commission and the Kearney Foundation of Soil Science. We
thank Raymond Chan from the Biocomputing Center of the
Plant Sciences department at UC–Davis for technical help. We
thank Keith Paustian, Steve Ogle, Changsheng Li, and William
Salas for their technical advice and support. We thank Miet
Boonen for research assistance. Finally, we are very grateful to
William Horwath from SAFS and Dennis Bryant from LTRAS
for their collaboration and their generous sharing of data.
LITERATURE CITED
Bartolome, J. W., and M. P. McClaran. 1992. Composition and
production of California oak savanna seasonally grazed by
sheep. Journal of Range Management 45:103–107.
Bemis, G., P. Perez, R. Shapiro, and B. B. Blevins. 2006.
Inventory of California greenhouse gas emissions and sinks:
1990 to 2004. California Energy Commission, Sacramento,
California, USA.
Campbell, C. A., R. P. Zentner, F. Selles, B. C. Liang, and B.
Blomert. 2001. Evaluation of a simple model to describe
carbon accumulation in a Brown Chernozem under varying
fallow frequency. Canadian Journal of Soil Science 81:383–
394.
Cates, R. L., and D. R. Keeney. 1987. Nitrous-oxide
production throughout the year from fertilized and manured
maize fields. Journal of Environmental Quality 16:443–447.
Ecological Applications
Vol. 20, No. 7
Chen, J. Y., J. W. Hopmans, and G. E. Fogg. 1995. Sampling
design for soil-moisture measurements in large field trials.
Soil Science 159:155–161.
Clark, M. S., W. R. Horwath, C. Shennan, and K. M. Scow.
1998. Changes in soil chemical properties resulting from
organic and low-input farming practices. Agronomy Journal
90:662–671.
Cole, C. V., K. Flach, J. Lee, D. Sauerbeck, and B. Stewart.
1993. Agricultural sources and sinks of carbon. Water, Air
and Soil Pollution 70:111–122.
Del Grosso, S. J., W. J. Parton, A. R. Mosier, D. S. Ojima,
A. E. Kulmala, and S. Phongpan. 2000. General model for
N2O and N2 gas emissions from soils due to denitrification.
Global Biogeochemical Cycles 14:1045–1060.
Del Grosso, S. J., W. J. Parton, A. R. Mosier, M. K. Walsh,
D. S. Ojima, and P. E. Thornton. 2006. DAYCENT
national-scale simulations of nitrous oxide emissions from
cropped soils in the United States. Journal of Environmental
Quality 35:1451–1460.
Denison, R. F., D. C. Bryant, and T. E. Kearney. 2004. Crop
yields over the first nine years of LTRAS, a long-term
comparison of field crop systems in a Mediterranean climate.
Field Crops Research 86:267–277.
Doane, T. A., O. C. Devevre, and W. R. Horwath. 2003. Shortterm soil carbon dynamics of humic fractions in low-input
and organic cropping systems. Geoderma 114:319–331.
Doane, T. A., and W. R. Horwath. 2004. Annual dynamics of
soil organic matter in the context of long-term trends.
Global Biogeochemical Cycles 18GB3008. [doi: 10.1029/
2004GB002252]
Flessa, H., P. Dorsch, and F. Beese. 1995. Seasonal variation of
N2O and CH4 fluxes in differently managed arable soils in
southern Germany. Journal of Geophysical Research,
Atmospheres 100:23115–23124.
Flessa, H., U. Wild, M. Klemisch, and J. Pfadenhauer. 1998.
Nitrous oxide and methane fluxes from organic soils under
agriculture. European Journal of Soil Science 49:327–335.
Franzluebbers, A. J. 2005. Soil organic carbon sequestration
and agricultural greenhouse gas emissions in the southeastern
USA. Soil and Tillage Research 83:120–147.
Gauch, H. G., J. T. G. Hwang, and G. W. Fick. 2003. Model
evaluation by comparison of model-based predictions and
measured values. Agronomy Journal 95:1442–1446.
Goossens, A., A. De Visscher, P. Boeckx, and O. Van
Cleemput. 2001. Two-year field study on the emission of
N2O from coarse and middle-textured Belgian soils with
different land use. Nutrient Cycling in Agroecosystems 60:
23–34.
He, Y. W., Y. Inamori, M. Mizuochi, H. N. Kong, N. Iwami,
and T. H. Sun. 2000. Measurements of N2O and CH4 from
the aerated composting of food waste. Science of the Total
Environment 254:65–74.
Hutchinson, G. L., and G. P. Livingston. 2002. Soil–
atmosphere gas exchange. Pages 1159–1182 in J. H. Dane
and C. Topp, editors. Methods of soil analysis. Part 4.
Physical methods. Soil Science Society of America, Madison,
Wisconsin, USA.
Hutchinson, G. L., and A. R. Mosier. 1979. Nitrous-oxide
emissions from an irrigated cornfield. Science 205:1125–1127.
IPCC [Intergovernmental Panel on Climate Change]. 2001.
Climate change 2001: the scientific basis. Cambridge University Press, Cambridge, UK.
Jambert, C., R. Delmas, D. Serca, L. Thouron, L. Labroue, and
L. Delprat. 1997. N2O and CH4 emissions from fertilized
agricultural soils in southwest France. Nutrient Cycling in
Agroecosystems 48:105–114.
Jones, J. W., G. Hoogenboom, C. H. Porter, K. J. Boote, W. D.
Batchelor, L. A. Hunt, P. W. Wilkens, U. Singh, A. J.
Gijsman, and J. T. Ritchie. 2003. The DSSAT cropping
system model. European Journal of Agronomy 18:235–265.
October 2010
GREENHOUSE GASES IN CROPPING SYSTEMS
Kaffka, S. R., D. C. Bryant, and R. F. Denison. 2005.
Comparison of organic and conventional maize and tomato
cropping systems from a long-term experiment in California.
Pages 218–221 in Proceedings of the First Scientific
Conference of the International Society for Organic Farming
Research. Institute of Organic Agriculture, Bonn, Germany.
Kaiser, E. A., and O. Heinemeyer. 1996. Temporal changes in
N2O-losses from two arable soils. Plant and Soil 181:57–63.
Kaiser, E. A., K. Kohrs, M. Kucke, E. Schnug, O. Heinemeyer,
and J. C. Munch. 1998. Nitrous oxide release from arable
soil: importance of N-fertilization, crops and temporal
variation. Soil Biology and Biochemistry 30:1553–1563.
Kong, A. Y. Y., J. Six, D. C. Bryant, R. F. Denison, and C. van
Kessel. 2005. The relationship between carbon input,
aggregation, and soil organic carbon stabilization in sustainable cropping systems. Soil Science Society of America
Journal 69:1078–1085.
Lal, R., J. Kimble, R. F. Follett, and C. V. Cole. 1998. The
potential of U.S. cropland to sequester carbon and mitigate
the greenhouse effect. Ann Arbor Press, Ann Arbor,
Michigan, USA.
Lee, J., J. W. Hopmans, C. van Kessel, A. P. King, K. J. Evatt,
D. Louie, D. E. Rolston, and J. Six. 2009. Tillage and
seasonal emissions of CO2, N2O and NO across a seed bed
and at the field scale in a Mediterranean climate. Agriculture,
Ecosystems and Environment 129:378–390.
Lee, J., J. Six, A. P. King, C. Van Kessel, and D. E. Rolston.
2006. Tillage and field scale controls on greenhouse gas
emissions. Journal of Environmental Quality 35:714–725.
Littell, R. C., G. A. Milliken, W. W. Stroup, R. D. Wolfinger,
and O. Schabenberber. 2006. SAS for mixed models. Second
edition. SAS Institute, Cary, North Carolina, USA.
Mahmood, T., R. Ali, M. I. Sajjad, M. B. Chaudhri, G. R.
Tahir, and F. Azam. 2000. Denitrification and total fertilizerN losses from an irrigated cotton field. Biology and Fertility
of Soils 31:270–278.
McCarl, B. A., and U. A. Schneider. 2001. Greenhouse gas
mitigation in U.S. agriculture and forestry. Science 294:2481–
2482.
Metherell, A. K., C. A. Cambardella, W. J. Parton, G. A.
Peterson, L. A. Harding, and C. V. Cole. 1995. Simulation of
soil organic matter dynamics in dryland wheat-fallow
cropping systems. Pages 1–20 in R. Lal, J. Kimble, E.
Levine, and B. A. Stewart, editors. Soil management and
greenhouse effect. Advances in soil sciences. CRC Press,
Boca Raton, Florida, USA.
Mitchell, A. E., Y.-J. Hong, E. Koh, D. M. Barrett, D. E.
Bryant, R. F. Denison, and S. Kaffka. 2007a. Ten-year
comparison of the influence of organic and conventional crop
management practices on the content of flavonoids in
tomatoes. Journal of Food and Agricultural Chemistry 55:
6154–6159.
Mitchell, C. C., and S. X. Tu. 2005. Long-term evaluation of
poultry litter as a source of nitrogen for cotton and corn.
Agronomy Journal 97:399–407.
Mitchell, J. P., K. Klonsky, A. Shrestha, R. Fry, A. DuSault, J.
Beyer, and R. Harben. 2007b. Adoption of conservation
tillage in California: current status and future perspectives.
Australian Journal of Experimental Agriculture 47:1383–
1388.
Mitchell, J. P., G. S. Pettygrove, S. Upadhyaya, A. Shrestha, R.
Fry, R. Roy, P. Hogan, R. Vargas, and K. Hembree. 2009.
Classification of conservation tillage practices in California
irrigated row crop systems. Publication 8364. University of
California, Division of Agricultural and Natural Resources,
Oakland, California, USA.
Mogge, B., E. A. Kaiser, and J. C. Munch. 1999. Nitrous oxide
emissions and denitrification N-losses from agricultural soils
1819
in the Bornhoved Lake region: influence of organic fertilizers
and land-use. Soil Biology and Biochemistry 31:1245–1252.
Ogle, S. M., F. J. Breidt, and K. Paustian. 2006. Bias and
variance in model results associated with spatial scaling of
measurements for parameterization in regional assessments.
Global Change Biology 12:516–523.
Pacala, S., and R. Socolow. 2004. Stabilization wedges: solving
the climate problem for the next 50 years with current
technologies. Science 305:968–972.
Parton, W. J., M. Hartman, D. Ojima, and D. Schimel. 1998.
DAYCENT and its land surface submodel: description and
testing. Global and Planetary Change 396:1–14.
Parton, W. J., A. R. Mosier, D. S. Ojima, D. W. Valentine,
D. S. Schimel, K. Weier, and A. E. Kulmala. 1996.
Generalized model for N2 and N2O production from
nitrification and denitrification. Global Biogeochemical
Cycles 10:401–412.
Parton, W. J., D. S. Ojima, C. V. Cole, and D. S. Schimel. 1994.
A general model for soil organic matter dynamics: sensitivity
to litter chemistry, texture and management. Soil Science
Society of America Journal 39:147–167.
Parton, W. J., D. S. Schimel, C. V. Cole, and D. S. Ojima. 1987.
Analysis of factors controlling soil organic matter levels in
Great Plains grasslands. Soil Science Society of America
Journal 51:1173–1179.
Paustian, K., O. Andren, H. H. Janzen, R. Lal, P. Smith, G.
Tian, H. Tiessen, M. Van Noordwijk, and P. L. Woomer.
1997. Agricultural soils as a sink to mitigate CO2 emissions.
Soil Use and Management 13:230–244.
Perez-Quezada, J. F., G. S. Pettygrove, and R. E. Plant. 2003.
Spatial–temporal analysis of yield and soil factors in two
four-crop rotation fields in the Sacramento Valley, California. Agronomy Journal 95:676–687.
Potthoff, M., L. E. Jackson, K. L. Steenwerth, I. Ramirez,
M. R. Stromberg, and D. E. Rolston. 2005. Soil biological
and chemical properties in restored perennial grassland in
California. Restoration Ecology 13:61–73.
Rochester, I. J. 2003. Estimating nitrous oxide emissions from
flood-irrigated alkaline grey clays. Australian Journal of Soil
Research 41:197–206.
Saltelli, A., K. Chan, and E. M. Scott. 2000. Sensitivity
analysis. John Wiley and Sons, New York, New York, USA.
Six, J., S. M. Ogle, F. J. Breidt, R. T. Conant, A. Mosier, and
K. Paustian. 2004. The potential to mitigate global warming
with no-tillage management is only realized when practiced in
the long term. Global Change Biology 10:155–160.
Smith, P., et al. 1997. A comparison of the performance of nine
soil organic matter models using datasets from seven longterm experiments. Geoderma 81:153–225.
Smith, P., et al. 2007. Policy and technological constraints to
implementation of greenhouse gas mitigation options in
agriculture. Agriculture, Ecosystems and Environment 118:
6–28.
Valentini, R., J. A. Gamon, and C. B. Field. 1995. Ecosystem
gas exchange in a California grassland: seasonal patterns and
implications for scaling. Ecology 76:1940–1952.
VandenBygaart, A. J., X. M. Yang, B. D. Kay, and J. D.
Aspinall. 2002. Variability in carbon sequestration potential
in no-till soil landscapes of southern Ontario. Soil and Tillage
Research 65:231–241.
Weitz, A. M., E. Linder, S. Frolking, P. M. Crill, and M.
Keller. 2001. N2O emissions from humid tropical agricultural
soils: effects of soil moisture, texture and nitrogen availability. Soil Biology and Biochemistry 33:1077–1093.
West, T. O., and W. M. Post. 2002. Soil organic carbon
sequestration rates by tillage and crop rotation: a global data
analysis. Soil Science Society of America Journal 66:1930–
1946.