Tang Et Al. 2019
Tang Et Al. 2019
Tang Et Al. 2019
https://doi.org/10.1007/s10533-019-00580-7 (0123456789().,-volV)
( 01234567
89().,-volV)
Received: 21 February 2019 / Accepted: 20 June 2019 / Published online: 25 June 2019
Ó Springer Nature Switzerland AG 2019
Abstract Climate change is expected to alter hourly dissolution. Multi-decadal modeling with varying
and daily rainfall regimes and, in turn, the dynamics of rainfall patterns was conducted on nine Australian
ecosystem processes controlling greenhouse gas emis- grasslands in tropical, temperate, and semi-arid
sions that affect climate. Here, we investigate the regions. Our results show that annual CO2 emissions
effects of expected twenty-first century changes in in the semi-arid grasslands increase by more than 20%
hourly and daily rainfall on soil carbon and nitrogen with a 20% increase in annual rainfall (with no
emissions, soil organic matter (SOM) stocks, and changes in the rainfall timing), but the tropical
leaching using a coupled mechanistic carbon and grasslands have opposite trends. A 20% increase in
nitrogen soil biogeochemical model (BAMS2). The annual rainfall also increases annual N2 O and NO
model represents various abiotic and biotic processes emissions in the semi-arid grasslands by more than
involving 11 SOM pools. These processes include 10% but decreases emissions by at least 25% in the
fungal depolymerization, heterotrophic bacterial min- temperate grasslands. When subjected to low fre-
eralization, nitrification, denitrification, microbial quency and high magnitude daily rainfall events with
mortality, necromass decomposition, microbial unchanged annual totals, the semi-arid grasslands are
response to water stress, protection, aqueous advection the most sensitive, but changes in annual CO2 emis-
and diffusion, aqueous complexation, and gaseous sions and SOM stocks are less than 5%. Intensification
of hourly rainfall did not significantly alter CO2
emissions and SOM stocks but changed annual NH3
Responsible Editor: Jan Mulder. emissions in the tropical grasslands by more than
300%.
Electronic supplementary material The online version of
this article (https://doi.org/10.1007/s10533-019-00580-7) con-
tains supplementary material, which is available to authorized Keywords Soil organic carbon Carbon cycle
users. Nitrogen cycle SOM model Precipitation
F. H. M. Tang (&) F. Maggi
Laboratory for Advanced Environmental Engineering
Research, School of Civil Engineering, The University of
Sydney, Bld. J05, Sydney, NSW 2006, Australia Introduction
e-mail: [email protected]
Climate change is predicted to increase rainfall
W. J. Riley
Earth Sciences Division, Lawrence Berkeley National
temporal variability, with a consensus of a shift
Laboratory, Berkeley, CA 94720, USA towards a higher frequency of droughts and heavier
123
198 Biogeochemistry (2019) 144:197–214
rainfall events (Easterling et al. 2000; Zhang et al. by reduced gas mobility in wet soil (Bouma and Bryla
2013). Although the uncertainty in rainfall predictions 2000); as a consequence, DIC is more prone to
is high and the predicted changes are spatially leaching.
heterogeneous (Maslin and Austin 2012), trend-de- Mineralization and organic carbon inputs to soil
tection studies based on global and regional rainfall through root exudation and plant litter are tightly
datasets have consistently reported an intensification linked to the availability of other nutrients, in partic-
in daily (Donat et al. 2013; Fischer and Knutti 2014) ular, nitrogen (Bengtson et al. 2012; Henriksen and
and hourly (Guerreiro et al. 2018) extremes. Changes Breland 1999; Manzoni and Porporato 2009). Varia-
in rainfall regimes can increase variations in soil water tions in soil water content can alter the microbial
content, which is a key driver of ecosystem processes activity of the nitrogen cycle, and its overall effect on
that affects vegetation growth (Porporato et al. 2003; nitrogen losses may be different from that of carbon
Yu et al. 2017; Tietjen et al. 2017), soil respiration (Gu and Riley 2010; Schimel 2018). The available
(Curiel Yuste et al. 2007; Schimel 2018), biogeo- inorganic nitrogen produced by increased SOM min-
chemical cycles (Delgado-Baquerizo et al. 2013; eralization after a rainfall pulse may be immobilized
Nielsen and Ball 2015), and greenhouse gas emissions into microbial biomass (Dijkstra et al. 2012), taken up
(e.g., CO2 , CH4 , NOx ; Harper et al. 2005; Kim et al. by plants (LLü et al. 2014), leached (Neilen et al.
2012). Hence, it is essential to analyze the extent to 2017), nitrified (Bateman and Baggs 2005; Stark and
which rainfall variability can affect terrestrial carbon Firestone 1995), or lost as nitrogen gases through
and nitrogen emissions. denitrification (Li et al. 1992; Sexstone et al. 1985;
Predicting the dynamics of soil organic matter Riley and Matson 2000). Microbial activity and plant
(SOM) as a result of rainfall intensification is complex nitrogen uptake may also have different responses to
and has been the target of many research efforts. On increased drying-rewetting cycles (Collins 2008;
the one hand, decreased rainfall amount can suppress Schwinning and Sala 2004) and the size of water
SOM depolymerization and mineralization due to pulses (Dijkstra et al. 2012). It is therefore difficult to
stronger microbial water stress (Schimel et al. 2007) predict the interactions and competitions between
and reduced nutrient mobility (Manzoni et al. 2012), these processes, and estimating their feedback on the
leading to a reduction in CO2 emissions. On the other carbon cycle can be even more challenging.
hand, rainfall extremes can increase the frequency of Owing to experimental studies that showed rapid
drying-rewetting cycles that result in CO2 pulses a few microbial response to soil moisture (Lundquist et al.
orders of magnitudes higher than background emis- 1999; Lee et al. 2004) and that soil microbes can
sions (known as ‘‘the Birch effect’’, Birch 1958; Li resuscitate and become active within hours after a
et al. 2010; Vargas et al. 2010). Studies based on rewetting event (Placella et al. 2012; Barnard et al.
single and multiple cycles of drying-rewetting exper- 2015), we question if the intensification in hourly
iments have arrived at very different conclusions rainfall extremes can have a more significant impact
regarding the carbon sources and mechanisms con- on SOM dynamics than daily variations. To this end,
tributing to the observed CO2 pulses (Schimel 2018). we aim to quantify the long-term impacts of hourly
The proposed mechanisms include contributions from and daily rainfall variations on carbon and nitrogen
dead microbial biomass (Kieft 1987), mobilization of emissions, leaching, and storage in grasslands with
stable carbon (Navarro-Garcı́a et al. 2012), microbial different seasonal rainfall regimes using a mechanistic
intracellular osmolytes (Warren 2014), and microbial model. We coupled the BAMS1 model developed in
resuscitation (Placella et al. 2012). Most of these Riley et al. (2014) to the nitrogen cycle model
experiments, however, were conducted at a time-scale developed in Maggi et al. (2008) by accounting for
of days to months and, hence it is difficult to C and N stoichiometric compositions of various SOM
extrapolate the observed drying-rewetting effects to pools. The C–N coupled model (called BAMS2;
long-term emissions and carbon storage. In addition to Biotic and Abiotic Model of SOM version 2) includes
microbial mediated processes, heavy rainfall pulses 11 SOM pools (four polymer pools and seven
can increase SOM losses through leaching in the form monomer pools), five microbial functional groups
of dissolved organic (DOC) and inorganic carbon (heterotrophic fungi and bacteria, ammonia oxidisers,
(DIC) (Liu et al. 2018). CO2 efflux may be suppressed nitrite oxidisers, and denitrifiers), plant nitrogen
123
Biogeochemistry (2019) 144:197–214 199
uptake, microbial growth, mortality and decomposi- NO, N2 O, N2 ); and five microbial functional groups
tion, protection, aqueous advection and diffusion, including heterotrophic fungi (FDEP ), heterotrophic
gaseous diffusion, aqueous complexation, and gaseous bacteria (BHET ), ammonia-oxidizing bacteria (BAOB ),
dissolution. BAMS2 was first benchmarked against nitrite-oxidizing bacteria (BNOB ), and denitrifying
field-observed heterotrophic soil respiration; N2 O and bacteria (BDEN ).
NO emissions; organic carbon inputs; and plant In BAMS2, NHþ 4 is a substrate in SOM decompo-
nitrogen uptake reported in the literature, and was sition reactions (R1–R8 in Fig. 1). All microbial
then used to conduct a suite of numerical experiments functional groups assimilate both carbon and nitrogen
on different hourly and daily rainfall variations in nine for growth, with fungi and bacteria having a C:N ratio
Australian grasslands located in tropical, temperate, of 8 and 5, respectively (Mouginot et al. 2014). In the
and semi-arid regions. mineralization of N-containing monomers (R9–R11),
a fraction of mineralized nitrogen is assimilated into
microbial biomass and the other fraction is released to
Methods the environment as free NHþ 4 , which can be used by
FDEP and BHET to decompose SOM, oxidized by BAOB
BAMS2 reaction network to NO 2 , and taken up by plants. The original
stoichiometric parameters of SOM decomposition
To account for the control of nitrogen availability on reactions in BAMS1 (Riley et al. 2014) were recalcu-
SOM dynamics, the BAMS1 carbon model described lated to account for the nitrogen immobilization into
in Riley et al. (2014) was coupled to the nitrogen cycle microbial biomass (Supplementary Information
model developed in Maggi et al. (2008). The C–N Table S.1).
coupled reaction network (BAMS2, Fig. 1) consists of Similarly, the stoichiometric parameters of nitrifi-
four SOM polymer pools (lignin, cellulose, hemicel- cation (R12–R13) and denitrification (R14–R17)
lulose, peptidoglycan); seven SOM monomer pools reactions reported in Maggi et al. (2008) were recal-
(monosaccharide, fatty acids, organic acids, phenols, culated to account for both carbon and nitrogen
nucleotides, amino acids, amino sugars); seven inor- assimilation into microbial biomass (Supplementary
ganic nitrogen molecules (NH3 , NHþ
4 , NO3 , NO2 , Table S.1). In addition to nitrification and
123
200 Biogeochemistry (2019) 144:197–214
denitrification, BAMS2 includes N2 fixation to NHþ 4 law. Equations used to model the transport of fluids
(R19). Although R19 represents biological fixation, and compounds in aqueous, gaseous, and biological
the N2 fixing microbial functional group is not phases are described in detail in Maggi (2019).
explicitly accounted for because N2 fixing microbes Aqueous complexation and gas dissolution (R40–
have a wide range of metabolic requirements; for R45, Supplementary Table S.1) are described in
example, they can be either aerobic or anaerobic and BRTSim-v3.1a using the mass action law (Maggi
can be either heterotrophic, autotrophic, chemolitho- 2019),
trophic, or methanogenic (Reed et al. 2011). Y Y
K¼ ½XR xR ½XP xP ; ð1Þ
Plants uptake both NHþ
4 and NO3 (R20–R21) and R P
produce aboveground (R28–R29, leaf and wood litter
with C:N ratio of 35, Moretto et al. 2001; Thomas and where K is the equilibrium constant, ½XR and ½XP are
Asakawa 1993) and belowground (R27, root exudates the reactant and product concentrations, respectively,
with C:N ratio of 12, Grayston et al. 1997; Mench and with xR and xP their corresponding stoichiometric
Martin 1991) SOM inputs. Litter decomposes into parameters. The values of K used in R40–R45 are
simpler organic polymers and monomers through obtained from Wolery (1992). Units for all variables
implicit exoenzyme activity (Riley et al. 2014), while are given in Supplementary Table S.2.
root exudates contain only organic monomers such as Chemical protection (R30–R39) is described using
monosaccharide, fatty acids, organic acids, and amino Langmuir kinetics to account for the protective
acids (Grayston et al. 1997). The carbon and nitrogen capacity of soil, such that (Atkins and De Paula 2005),
assimilated into microbial biomass are returned to the d½X(p)
SOM pools through microbial mortality (R22–R26). ¼ ka ðQmax ½X(p)Þ½X(aq) kd ½X(p);
dt
Here, microbial mortality and necromass decomposi- ð2Þ
tion are modeled as one lumped process.
In addition to biological processes, SOM mono- where [X(p)] and [X(aq)] are the concentrations of
mers and inorganic nitrogen also undergo abiotic chemical X in protected (p) and aqueous (aq) phases,
processes such as advection and diffusion, gas disso- respectively; ka and kd are the forward (protection) and
lution (R41–R45), and protection (e.g., via mineral reverse (un-protection) rate constants, respectively;
surface binding, R30–R39). SOM polymers are con- and Qmax is the maximum soil protective capacity. At
sidered to be non-soluble (in solid phases) organic equilibrium (d[X(p)]/dt = 0), KP ¼ ka =kd is the
carbon and do not undergo protection processes. protection equilibrium constant. Eq. 2 describes pro-
tection as a function of silt and clay content through
Biogeochemical and transport solver the variable Qmax . For SOM protection, Qmax is
estimated using the empirical relationship derived in
The BAMS2 reaction network (Fig. 1) was solved in Six et al. (2002) , i.e., Qmax [g-C protected kg soil1 ]
the general-purpose multi-phase and multi-component ¼ 0:32 Cfine ½% þ 16:33, where Cfine is the silt and
bioreactive transport simulator BRTSim-v3.1a clay content. For NHþ þ
4 protection, Qmax [g-NH4 -N
(Maggi 2019). BRTSim solves for the mass continuity protected kg soil1 ] ¼ 20:07 Cfine ½% (Alshameri
and conservation laws using hybrid explicit-implicit et al. 2018) is used, while Qmax [g-NO
3 /NO2 -N
finite volumes solvers. The water flow along a one- 1 4
protected kg soil ] ¼ 4:73 10 Cfine ½% (Black
dimensional variably saturated soil column is modeled and Waring 1979) is used for the protection of NO 3
using the Richards equation (Richards 1931) in and NO 2.
conjunction with the empirical relative permeability- Chemical and biochemical kinetic reactions are
potential-saturation relationship of the Brooks-Corey solved using the general framework of Michaelis–
model (Brooks and Corey 1964). The transport of Menten–Monod kinetics described in Maggi and Riley
dissolved compounds is described by the Darcy’s (2010). A biochemical kinetic reaction involving
advection velocity and the Fick’s diffusion. The growth of microbial functional group BX can be
advection of gaseous compounds is excluded, but written as,
gas diffusion is explicitly accounted for using Fick’s
123
Biogeochemistry (2019) 144:197–214 201
fS ¼ minff ðSB Þ; f ðSL Þ=maxff ðSL Þgg: ð5Þ where RNplant is the plant nitrogen uptake rate; kNHþ4 and
kNO3 are rate constants; and KM NHþ and KMNO are
The function f ðSB Þ describes the immobilization of 4 3
water into microbial biomass that has a specific water Michaelis–Menten constants for NHþ 4 and NO3
volume fraction fL and considers water as a resource uptake, respectively. The total amount of nitrogen
for microbial growth. Therefore, microbes can only taken up by plants (Nplant ) is used to regulate SOM
grow if there is enough water to immobilize and inputs (see R27–R29, Supplementary Table S.1) in
enough pore space to occupy. When microbes die and such a way that the total amount of organic N input to
decompose, water is re-mobilized and returned to the soil is always smaller than or equal to Nplant . Hence, in
soil. Following the approach in Maggi and Porporato instances when plant nitrogen uptake is low, the inputs
(2007), f ðSB Þ is defined as of SOM will also be low. Because plants also
experience water stress in dry conditions (Manzoni
and Porporato 2007; Porporato et al. 2003), a
123
202 Biogeochemistry (2019) 144:197–214
reduction factor of fS ¼ f ðSL Þ=maxff ðSL Þg (Supple- region starts from May to September with lower
mentary Fig. S.1b) is used in Eq. 8. annual rainfall but a higher number of wet days than
A summary of model parameters is reported in the tropical region. The semi-arid region generally has
Table S.1, and a list of inhibitions applied to each low annual rainfall with a small number of wet days
kinetic reaction is reported in Supplementary (Table 1).
Table S.3. Descriptions of mathematical equations, Soil characteristics at each site were obtained from
numerical methods, and solution convergence criteria the SoilGrids database (Hengl T et al. 2017) and were
used in BRTSim-v3.1a are detailed in Maggi (2019). used to estimate the hydraulic parameters (Supple-
An example of the input files for BAMS2 model is mentary Table S.4). The reactive transport model
provided along with the Supplementary Information described in ‘‘BAMS2 reaction network’’ and ‘‘Bio-
and the BRTSim solver can be downloaded from the geochemical and transport solver’’ sections was solved
links provided in the Acknowledgments. over a 2 m soil column with constant saturation as the
lower boundary condition. Water boundary fluxes
Site descriptions entering and leaving the soil column were defined by
rainfall and plant evapotranspiration (Supplementary
The BAMS2 reaction network was applied in nine Fig. S.2). Historical daily rainfall and temperature data
Australian grasslands in tropical, temperate, and semi- (from 1979 to 2017) at each site were obtained from
arid regions that have distinct seasonal rainfall the CPC US Unified Precipitation data provided by the
regimes. Site locations were determined based on the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA
Dynamic Land Cover Dataset (Lymburner et al. 2011) (Xie et al. 2010), and the Global Historical Climatol-
€
and the modified KOppen climate classification of the ogy Network-Daily dataset (Menne et al. 2012),
Bureau of Meteorology, Australia (Stern and Dahni respectively. The Richardson-type weather generator
2013) (Table 1). The tropical region is characterized developed by Chen et al. (2010) was then used to
by a pronounced dry season starting from May to produce 2000-year daily rainfall and temperature time
September and is followed by a period of heavy series with statistical properties similar to those of
rainfall between October and April with an average historical data. Plant actual evapotranspiration (ET) is
annual rainfall of 1289 mm y1 (Supplementary calculated as ET ¼ kc ET0 with the plant coefficient
Fig. S.2). In contrast, the wet season in the temperate kc ¼ 0:8 (Allen et al. 2005) and the potential
123
Biogeochemistry (2019) 144:197–214 203
evapotranspiration ET0 estimated using the FAO ETO intensified with decreasing Hwet (Supplementary
calculator (Allen et al. 1998). The root density along Fig. S.3, third row).
the soil profile is assumed here to be a negative
exponential distribution function with 50% of root Analyses and benchmarking
density located at 0.1 m (Christie 1978; Greenwood
and Hutchinson 1998). Plant water uptake (evapotran- Prior to the numerical experiments, baseline simula-
spiration), plant nitrogen uptake (R20–R21), and root tions (using historical rainfalls) were initialized with
exudation (R27) were allocated over the soil depth SOM concentrations close to the organic carbon
according to the root distribution. content reported in the SoilGrids database (Hengl T
et al. 2017) and the microbial biomass close to zero.
Rainfall scenarios The simulations were run for 2000 years for biochem-
ical reactions in the root zone to reach a steady state
Numerical experiments were conducted with three and to develop a steady microbial biomass profile. For
rainfall scenarios. The weather generator in Chen reporting our results, we considered the top 30 cm of
et al. (2010) was modified to generate rainfall time the soil as the root zone (RZ). The outputs of the
series with varying statistical properties specific for 2000-year simulations were then used as initial
each scenario, whereas no modification was applied to conditions in the numerical experiments. In all
the evapotranspiration time series. We discuss the numerical experiments, simulations were run for
possible implication of this simplification below. 1000 years and outputs from the last 50 years of
Scenario 1: change in annual cumulative rainfall simulation were averaged for analysis.
amount. Rainfall time series were modified so that the Baseline simulations were benchmarked against
annual cumulative rainfall amount (Pcum ) ranged field observations collected from the literature with
within ?/- 20% of the historical value, while the benchmark values reported in Table 2. Because
annual number of wet days (Dwet ) remained constant. BAMS2 includes only microbial heterotrophic respi-
In this scenario, the rainfall magnitude P in each ration, CO2 emissions in the baseline simulations were
quantile and the annual maximum daily rainfall (Rmax d ) compared against heterotrophic soil respiration flux
changed linearly with changes in Pcum , i.e., a 20% (RH ) of 353 natural and unmanaged grasslands
increase in Pcum led to 20% increases in P in all reported in the Soil Respiration Database Version
quantiles and Rmax
d (Supplementary Fig. S.3, first row). 4.0 (SRDB-V4 Bond-Lamberty and Thomson 2018).
Scenario 2: change in daily rainfall amount and In instances where the values of RH were not reported,
frequency. Rainfall time series were modified for Dwet we assumed that the ratio between heterotrophic and
to range within ?/- 50% of the historical value while autotrophic respiration is 1:1, i.e., RH ¼ 0:5Rs , with Rs
keeping Pcum unchanged. A decrease in Dwet caused a as the total carbon flux from soil respiration. N2 O
reduction in P at low quantiles and an increase in P at emissions were benchmarked against measurements in
high quantiles, implying fewer and heavier rainfalls. 40 grasslands reported in the database of Aronson and
Percent change in Rmax d increased non-linearly with Allison (2012), while NO emissions were compared
decreasing Dwet ; for example, a 50% decrease in Dwet against the dataset reported in Davidson and Kingerlee
resulted in a 70% increase in Rmax d (Supplementary (1997). The annual carbon inputs were compared
Fig. S.3, second row). against observations in 46 grasslands recorded in the
Scenario 3: change in hourly rainfall. Hourly Global Database of Litterfall Mass and Litter Pool
rainfall time series were constructed by exponentially Carbon and Nutrients (Holland et al. 2015), whereas
distributing the observed daily rainfall to a given the annual plant nitrogen uptake was benchmarked
number of wet hours Hwet in that day. Here, we used against field experiments of 16 grass species reported
Hwet ¼ 24 hours as the reference and we generated in Bessler et al. (2012).
hourly rainfall time series with decreasing Hwet by The correlation between two quantities x and y is
assuming the probability to rain in a given hour is calculated as R(x,y)= cov(x,y)/rx ry , where rx and ry
independent of the hour before. Hourly rainfall are the standard deviations of x and y, respectively.
The lag time between two time series was quantified
123
204 Biogeochemistry (2019) 144:197–214
CO2 flux Range [18.34, 102.26] [0.84, 1505] Bond-Lamberty and Thomson (2018)
(g C m2 y1 ) Mean ± r 65.31 ± 32.36 397.48 ± 265.32
N2 O flux Range ½ 5:1 104 , 171.48] [- 13.33, 268.06] Aronson and Allison (2012)
2 1 Mean ± r 102.68 ± 78.07 98.05 ± 91.63
(mg N m y )
NO flux Range ½6:0 105 , 169.45] [0, 292] Davidson and Kingerlee (1997)
2 1 Mean ± r 91.56 ± 73.85 120.45 ± 18.25
(mg N m y )
SOM inputs Range [20.86, 167.36] [10.00, 835.00] Holland et al. (2015)
(g C m2 y1 ) Mean ± r 97.14 ± 54.82 253.89 ± 181.42
Plant N uptake Range [1.58, 12.26] ½\1, 20] Bessler et al. (2012)
(g N m2 y1 ) Mean ± r 7.38 ± 4.03 11.4 ± 0.9
using cross-correlation analysis (function xcorr in slightly lower than those in the temperate grasslands.
Matlab2017a). This pattern may be explained by the high denitrifi-
cation rates in the temperate grasslands that con-
tributed to CO2 emissions and the slightly lower
Results mineralization rates in the tropical grasslands. In all
grasslands, the depolymerization rates were substan-
Benchmarking of baseline simulations tially lower than the mineralization of SOM mono-
mers, suggesting that depolymerization is the rate-
The modeled CO2 , N2 O, and NO emissions; SOM limiting process that controls CO2 emissions.
input rates; and plant nitrogen uptake rates were within NO and N2 O emissions were highest in the
the range of field measurements reported in various temperate and semi-arid grasslands, respectively
databases (Table 2). In baseline simulations, the semi- (Fig. 2a). In the tropical grasslands, NO and N2 O
arid grasslands, which received the lowest amount and emissions were either negative (i.e., a sink) or close to
least frequent rainfall, had the lowest CO2 emissions zero. Although some studies have observed negative
and SOM inputs (Fig. 2). Although the tropical N2 O fluxes (da Silva Cardoso et al. 2017) and low
grasslands had the highest depolymerization and denitrification capacity (Xu et al. 2013) in tropical
SOM input rates, CO2 emissions in these sites were soils, other studies argued that a wetter soil would
Temperate Temperate
200 Semi-arid 200 Semi-arid
150
10
150 150
100
100 100
5
50 50 50
0 0
0 0
CO2 N2O NO NH3 SOM I DEP MIN NUP NIT DEN
Fig. 2 a CO2 , N2 O, NO, and NH3 emissions and b SOM inputs grasslands under historical rainfall patterns. Error bars represent
(SOMI ), depolymerization (DEP), monomer mineralization the standard deviations of the three sites in the same climatic
(MIN), plant nitrogen uptake (NUP ), nitrification (NIT), and region. Results are the averages of the last 50 years of the
denitrification (DEN) rates in tropical, temperate, and semi-arid simulation period
123
Biogeochemistry (2019) 144:197–214 205
have higher anaerobicity, and therefore should have 0.63, Fig. 3) during the dry season. The peaks in CO2
higher N2 O emissions (Skiba and Smith 2000). came approximately five to six days after the peak in S,
However, the process that limited N2 O and NO and N2 O and NO emissions came less than one day
emissions in our tropical grassland simulations is after the peak in S (Supplementary Fig. S.4). In the wet
nitrification rather than denitrification (Fig. 2b). In season, the correlations were slightly lower in the
BAMS2, NHþ 4 is the only source of inorganic nitrogen temperate grasslands and were substantially lower in
to the soil, mainly coming from N2 fixation (R19) and the semi-arid grasslands.
mineralization of N-containing monomers (R9–R11). In contrast to temperate and semi-arid grasslands,
NHþ
4 has to be first nitrified to NO2 or NO3 before
CO2 emissions in the tropical grasslands were nega-
BDEN can further convert the nitrogen into NO and tively correlated with S regardless of the season
N2 O. In tropical grasslands, the soil water content was (Fig. 3, first row). In all grasslands, NH3 emissions
relatively high (i.e., soil saturation S 0:6 0:8 ), generally had high negative correlations with S during
and therefore the NHþ 4 concentration in the root zone
the dry season.
was low. At low NHþ These correlation analyses suggest that soil mois-
4 , BAOB , which has a high KM
þ
value for NH4 , was out-competed by BHET and FDEP . ture has an important control on greenhouse gas
Because the transformation of NHþ emissions in both high and low annual rainfall
4 to NO3 by BAOB
grasslands.
was suppressed, denitrification could not occur and led
to negligible N2 O and NO emissions in the tropical
Scenario 1: impacts of annual rainfall amount
grasslands. We note however that, in wet soils that
have low NHþ 4 concentrations, the nitrifiers may have
Contrary to the general expectation that increasing
adapted to a KM value lower than that applied in
annual rainfall (Pcum ) would have a larger impact on
BAMS2, which was calibrated against temperate soils
drier lands, our simulations suggested that both dry
(Maggi et al. 2008).
and wet grasslands are very sensitive to changes in
Pcum , and they have distinctive responses (Fig. 4).
In the semi-arid grasslands, all carbon and nitrogen
Controls of soil moisture dynamics on C and N
emissions increased by 10% to 30% when Pcum was
emissions
increased by 20% (Fig. 4a–d). An increase in water
availability in the semi-arid grasslands increased all
To better understand how soil moisture dynamics and
biological processes, including plant nitrogen uptake
daily rainfall impact carbon and nitrogen emissions,
(Supplementary Fig. S.7f), SOM inputs to soil
we analyzed the correlations (R) between time-series
(Fig. 4h), heterotrophic respiration (Fig. 4f, g), nitri-
of soil saturation S; daily rainfall amount P; and CO2 ,
fication (Supplementary Fig. S.7d), and denitrification
N2 O, NO, and NH3 emissions (Fig. 3). In all grass-
(Supplementary Fig. S.7e). The increased biological
lands, the correlation R(S, P) was relatively weak with
activity, however, increased only slightly the SOM
slightly higher values observed in the tropical grass-
stocks (\5%, Fig. 4e). Together with increased water
lands in the wet season. In general, S had a better
advection at high Pcum , the increased biological
correlation with C and N emissions as compared to P.
activity also led to a substantial increase in DOC and
Simulations with BAMS2 were able to capture
DIC leaching to soils below the root zone (Supple-
relatively well the Birch effect resulting from drying-
mentary Fig. S.7a, b).
rewetting cycles in the semi-arid and temperate
CO2 emissions in the temperate grasslands
grasslands, with a peak in CO2 emission observed
increased by less than 3% with increasing Pcum
after rainfall events (Supplementary Fig. S.5). Except
(Fig. 4a). Depolymerization and mineralization rates
for SA1 that has a wet season between October to
increased only slightly with increased water availabil-
April (Supplementary Fig. S.2), SA2 and SA3 are
ity, but this was associated with a higher increase in
relatively dry throughout the year and are considered
SOM inputs, hence, resulting in SOM stocks approx-
to have only a dry season. CO2 , N2 O, and NO
imately 10% greater than those in the baseline
emissions in the semi-arid and temperate grasslands
simulations (Fig. 4). Although heterotrophic respira-
had relatively high positive correlations with S (R [
tion was enhanced with increasing soil water,
123
206 Biogeochemistry (2019) 144:197–214
Fig. 3 Correlations of
average soil saturation S in
the root zone (first row) and
daily rainfall amount
P (second row) against CO2 ,
N2 O, NO, and NH3
emissions in the dry and wet
seasons. Error bars represent
the standard deviations of
the three sites in the same
climatic region. Among the
three semi-arid grasslands,
only SA1 has a wet season
Fig. 4 Effects of changes in annual cumulative rainfall amount ization rates (DEP), g mineralization rates of SOM monomers
Pcum (Scenario 1) on annual a CO2 emissions, b N2 O emissions, (MIN), and h SOM inputs rates. Shaded areas represent the
c NO emissions, d NH3 emissions, e SOM stocks, f depolymer- standard deviations
123
Biogeochemistry (2019) 144:197–214 207
nitrification and denitrification rates in the temperate White (2008) that showed an increase in aboveground
grasslands decreased substantially with increasing net primary productivity when semi-arid ecosystems
Pcum , leading to the reduction in N2 O and NO were subjected to rainfall events that were larger in
emissions (Fig. 4b, c). The increased water content size but fewer in number. The balance between
may have diluted the concentration of NHþ 4 in the root increased SOM inputs and decomposition caused a
zone, causing the nitrifiers and denitrifiers to be out- slight increase in SOM stocks (\2%, Fig. 5e) and a
competed by heterotrophic bacteria and fungi. substantial increase in DOC and DIC leaching to
Increased water content also decreased the volatiliza- below the root zone (Supplementary Fig. S.8a, b). In
tion of ammonia (Fig. 4d). contrast to SOM depolymerization and mineralization,
In contrast to the semi-arid and temperate grass- the nitrification and denitrification rates in the semi-
lands, the wet tropical grasslands generally featured a arid grasslands were reduced with decreasing Dwet ,
decrease in biological activity with increasing Pcum . leading to a reduction in N2 O emissions (Fig. 5b).
CO2 emissions decreased with increasing soil water Although biological denitrification was reduced,
content (Fig. 4a) because high water content reduced chemodenitrification increased with decreasing Dwet
oxygen availability and decreased SOM and NHþ 4
and contributed to the increasing NO emissions
concentrations, leading to decreasing heterotrophic (Fig. 5c). The effects of increased rainfall intensity
respiration. In particular, the mineralization rates and reduced frequency on nitrogen emissions in the
decreased two times more than the depolymerization semi-arid grasslands matched relatively well with the
(Fig. 4f, g) because the soluble SOM monomers numerical-experiments tested in Gu and Riley (2010).
tended to be advected out of the root zone at high Gu and Riley (2010) also found that, when applied
Pcum . An increasing Pcum also reduced plant nitrogen with a low total rainfall amount, high intensity and low
uptake (Supplementary Fig. S.7f) and SOM inputs frequency rainfall events reduced N2 O emissions in
(Fig. 4h), but the overall balance between inputs and sandy loams soils, but increased NO emissions.
decomposition resulted in a net SOM storage Less frequent and more intense events did not alter
(Fig. 4e). Although DOC leaching increased with CO2 emissions in the temperate grasslands but
increasing soil water content, the decreased biological substantially reduced N2 O and NO emissions
activity had substantially reduced the DIC leaching (Fig. 5a–c). Big pulses of water diluted and trans-
(Supplementary Fig. S.7a, b). ported inorganic nitrogen out of the root zones, and
hence decreased the nitrification and denitrification
Scenario 2: impacts of daily rainfall amount rates.
and frequency In the tropical grasslands, CO2 , N2 O, and NO
emissions were not sensitive to the decrease in Dwet ,
We investigated the response of C and N dynamics to but the NH3 volatilization was greatly reduced
variations in daily rainfall amount and frequency by (Fig. 5a–d). CO2 emissions, however, increased
changing the number of wet days Dwet in a year while slightly with increasing Dwet , suggesting that more
keeping the total annual rainfall constant; that is, a frequent and less intense rainfall events can increase
time-series with a smaller Dwet value has fewer but heterotrophic respiration in grasslands with tropical
larger rainfall events. rainfall regimes.
Among all grasslands, the semi-arid grasslands
were the most sensitive to variations in Dwet . CO2 Scenario 3: impacts of hourly rainfall
emissions in the semi-arid grasslands increased by intensification
approximately 7% with a 50% decrease in Dwet
(Fig. 5a). Fewer and larger rainfall events increased CO2 emissions, SOM decomposition rates, and SOM
the plant nitrogen uptake (Supplementary Fig. S.8f), stocks were relatively insensitive to hourly rainfall
and therefore increased the SOM inputs to soil amounts in all grasslands with CO2 emissions
(Fig. 5h). Upon the assumption that plant nitrogen increased only slightly in the tropical and semi-arid
uptake is proportional to plant biomass growth, similar grasslands (\2%, Fig. 6a, e–g). DOC and DIC
experimental observations were reported in Heisler- leaching to below the root zone, however, increased
123
208 Biogeochemistry (2019) 144:197–214
Fig. 5 Effects of changes in the number of wet days Dwet rates (DEP), g mineralization rates of SOM monomers (MIN),
(Scenario 2) on annual a CO2 emissions, b N2 O emissions, c NO and h SOM inputs rates. Shaded areas represent the standard
emissions, d NH3 emissions, e SOM stocks, f depolymerization deviations.
Fig. 6 Effects of changes in the number of wet hours Hwet rates (DEP), g mineralization rates of SOM monomers (MIN),
(Scenario 3) on annual a CO2 emissions, b N2 O emissions, c NO and h SOM inputs rates. Shaded areas represent the standard
emissions, d NH3 emissions, e SOM stocks, f depolymerization deviations
with a decreasing number of wet hours Hwet in the emissions of nitrogen gases. In the tropical grasslands,
semi-arid grasslands (Supplementary Fig. S.9a, b). the NH3 volatilization was largely reduced (i.e.,
Although SOM decomposition was not signifi- [ 300% reduction, Fig. 6d) with decreased Hwet . In
cantly affected, fewer and larger hourly rainfall events the temperate grasslands, denitrification rates slightly
(i.e., decreasing Hwet ) altered substantially the decreased with decreasing Hwet and caused a decline in
123
Biogeochemistry (2019) 144:197–214 209
N2 O and NO emissions (Fig. 6b, c). Although the models, there are still some other mechanisms that are
denitrification rates were not substantially altered, the currently not accounted for here. In BAMS2, we
variation in hourly rainfall amounts changed the ratio considered a simplified nitrogen cycle that includes
of N2 O:NO production in the semi-arid grasslands only N2 fixation, nitrification, and denitrification.
(Fig. 6b, c), with N2 O:NO ratio increased as Hwet However, the nitrogen cycle in soil is much more
decreased. complicated than that, and many new metabolic
capabilities of N-transforming microorganisms are
continuously being discovered (Kuypers et al. 2018;
Discussion Schreiber et al. 2012). Biotic N-transformation path-
ways not considered in BAMS2 include dissimilatory
The BAMS2 model represents the highly complex nitrate reduction to ammonium (DNRA,
interplay between many biotic and abiotic mecha- NO 3 ! NO
2 ! NH þ
4 , Tiedje et al. 1983), anaerobic
nisms hypothesized to be important for carbon and ammonium oxidation (anammox, NO 2 ! NO ?
nitrogen cycles, including depolymerization, SOM NHþ 4 ! N 2 H 4 ! N 2 , Mulder et al. 1995), complete
mineralization, microbial mortality, necromass ammonia oxidation (comammox, NHþ
4 ! NO3 ,
decomposition, N2 fixation, nitrification, denitrifica- Daims et al. 2015), hydroxylamine oxidation to nitric
tion, protection, advection, and diffusion. These oxide (NH2 OH ! NO, Caranto and Lancaster 2017),
mechanisms have different responses to soil water and nitric oxide dismutation to dinitrogen (NO ! N2 ,
content, and therefore a detailed description of their Ettwig et al. 2010). We note that BAOB and BNOB can
interactions is pivotal to this study that explicitly aims also reduce NO 2 to NO and N2 O (Schreiber et al.
at assessing the impact of rainfall variability on soil 2012); however, this capability was not included in
carbon and nitrogen dynamics. We note however that BAMS2. Even though complex, accounting for a more
the determination of model parameter values can be detailed description of the nitrogen cycle may improve
difficult for a model with high complexity, and this can the estimation of greenhouse gas emissions and SOM
introduce additional uncertainties. In this work, we stocks as our simulation analysis shows that the
used the validation by construct approach (McCarl interactions between soil carbon and nitrogen cycles
and Apland 1986) to design and test our model. The have non-linear responses to rainfall variability.
model parameters relative to the carbon cycle were By having fixed C:N ratios of litter and root
estimated against 618 SOM profiles of grasslands exudates, we used a simplified approach to regulate
located across Nebraska and Colorado (detailed in the above- and belowground SOM inputs through plant
Riley et al. 2014); those corresponding to the nitrifi- nitrogen uptake in such a way that the total organic
cation and denitrification processes were estimated nitrogen inputs to the soil cannot exceed the total
against field measurements of CO2 , N2 O, and NO inorganic nitrogen (NHþ
4 and NO3 ) taken up by plants.
fluxes (detailed in Maggi et al. 2008); and the other This approach assumes that all nitrogen taken up by
parameters were estimated against field and laboratory plants is assimilated into plant biomass and eventually
experiments reported in the literature (detailed in returned to the soil. The assimilation of carbon into
Supplementary Table S.1). We then benchmarked the plant biomass was not explicitly modeled, and hence
model outputs against field observed CO2 , N2 O and we did not consider a dynamic litter C:N ratio.
NO emissions; SOM inputs; and plant nitrogen uptake Improvements to the description of plant-soil interac-
rates compiled in various databases (detailed in tions in BAMS2 may be implemented in future work
Table 2). Although the sensitivity analysis of model to account for plant carbon assimilation, flexible C:N
parameters had been conducted separately for carbon ratios for litter and root exudates, and the effects of
(Riley et al. 2014) and nitrogen (Maggi et al. 2008) nutrient limitation on photosynthesis capacity follow-
cycles, we note that the parameter sensitivity may ing suggestions in Achat et al. (2016).
change after coupling the two models, and therefore a We observed in our simulations that, when switch-
global sensitivity analysis of BAMS2 is needed, and it ing to a new rainfall pattern, the microbial population
is the target of our next work. took a few decades to reach a steady profile and a
Although the reaction network in BAMS2 is steady bacterial to fungal ratio. This observation aligns
comparably or more complex than many other SOM
123
210 Biogeochemistry (2019) 144:197–214
with experimental studies that showed the dependency dynamics observed in grasslands, with model outputs
of soil respiration on historical rainfall, which can be falling within the range of field observations compiled
explained by the shift in microbial community com- in various published databases. Dry and wet grass-
position and activity (Lau and Lennon 2012; Hawkes lands responded differently to variations in rainfall
et al. 2017). Hence, field studies that spanned across patterns and rainfall variability had a different impact
time-scales of months may capture only the transient on carbon and nitrogen emissions. An increasing
effects. Although limited by the need to simplify an annual rainfall generally increased both microbial and
ecosystem, long-term simulations with models such as plant activities in the semi-arid grasslands, leading to
BAMS2 allow assessment of cumulative impacts of increases in CO2 , N2 O, NO, and NH3 emissions; yet, it
rainfall variability on soil C and N dynamics and reduced inorganic N availability in the tropical
identification of interactions between C and N cycles, grasslands, decreased biological activities, and caused
which are difficult to capture in field studies. In a reduction in CO2 emissions. The balance between
particular, our simulations featured a tight link SOM inputs and decomposition, however, always
between soil respiration and nitrogen availability. resulted in increasing SOM stocks with increasing
Aligned with field data analysis in Wang and Fang annual rainfall in all grasslands. High rainfall amounts
(2009), we observed a reduction in CO2 emissions can dilute NHþ 4 concentrations to below optimal
with increasing annual rainfall in wet tropical grass- values for nitrification, thus reducing N2 O and NO
lands, and we can explain this observation as a emissions in the temperate grasslands. Fewer and
consequence of N limitation. Although increased larger daily rainfall events slightly increased CO2
rainfall amount releases plants and soil microbes from emissions and SOM stocks in the semi-arid grasslands,
water stress, high soil water content also reduces the but caused a substantial increase in NO emissions as a
concentrations of inorganic N, putting soil microbes in result of increased chemodenitrification. Changes in
an N-limiting condition and causing decreased soil hourly rainfall amounts and frequency did not signif-
respiration. In dry semi-arid grasslands, the observed icantly alter soil carbon emissions and stocks in all
increase in soil respiration with increasing rainfall grasslands. Although the biotic processes in the
amount can be attributed to the direct moisture effect tropical grasslands are relatively insensitive to hourly
on soil microbes that increases microbial activity and rainfall variability, the high magnitude hourly rainfall
the indirect effect through increased plant litter (Lau events can substantially reduce NH3 volatilization.
and Lennon 2012). While the short-term impacts of
drying and rewetting cycles on soil activity has been Acknowledgements FHMT and FM are supported by the
studied in many field experiments (e.g., Kieft 1987; SREI2020 EnviroSphere research program of the University of
Sydney. FM is also supported by the Mid Career
Harper et al. 2005; Xiang et al. 2008), our simulations Research Award and Sydney Research Accelerator Fellowship
confirmed that, with no change in annual rainfall, (SOAR) of the University of Sydney. WJR is supported by the
prolonged droughts and increased high rainfall pulses Director, Office of Science, Office of Biological and
can increase cumulative CO2 emissions in dry grass- Environmental Research of the U.S. Department of Energy
under Contract No. DE-AC02-05CH11231 as part of the LBNL
lands in the long-term, which we attribute to increased TES Belowground Biogeochemistry Scientific Focus Area. The
substrate availability as a result of accumulation authors thank Giulia Ceriotti for the many conversations on
(resulting from plant residuals and microbial lysis) topics presented here. The authors acknowledge the Sydney
during the droughts. Informatics Hub and the University of Sydney’s high
performance computing cluster Artemis for providing the high
performance computing resources that have contributed to the
results reported within this work. The BRTSim solver package
Conclusions can be downloaded at https://sites.google.com/site/
thebrtsimproject/home or from the mirror https://www.
dropbox.com/sh/wrfspx9f1dvuspr/
We present a C–N coupled mechanistic SOM model AAD5iA9PsteX3ygAJxQDxAy9a?dl=0.
(BAMS2) to investigate the effects of hourly and daily
rainfall variations on soil carbon and nitrogen emis-
sions, stocks, and leaching in grasslands with different
seasonal rainfall regimes. BAMS2 captured relatively
well the Birch effect and the carbon and nitrogen
123
Biogeochemistry (2019) 144:197–214 211
123
212 Biogeochemistry (2019) 144:197–214
three stocking rates for 30 years. Grass Forage Sci Lü XT, Dijkstra FA, Kong DL, Wang ZW, Han XG (2014) Plant
53(2):120–128 nitrogen uptake drives responses of productivity to nitro-
Gu C, Riley WJ (2010) Combined effects of short term rainfall gen and water addition in a grassland. Sci Rep 4:4817
patterns and soil texture on soil nitrogen cycling: a mod- Lundquist EJ, Scow KM, Jackson LE, Uesugi SL, Johnson CR
eling analysis. J Contam Hydrol 112(1–4):141–154 (1999) Rapid response of soil microbial communities from
Guerreiro SB, Fowler HJ, Barbero R, Westra S, Lenderink G, conventional, low input, and organic farming systems to a
Blenkinsop S, Lewis E, Li XF (2018) Detection of conti- wet/dry cycle. Soil Biol Biochem 31(12):1661–1675
nental-scale intensification of hourly rainfall extremes. Nat Lymburner L, Tan P, Mueller N, Thackway R, Lewis A,
Clim Chang 8(9):803 Thankappan M, Randall L, Islam A, Senarath U (2011) The
Harper CW, Blair JM, Fay PA, Knapp AK, Carlisle JD (2005) National Dynamic Land Cover Dataset. Geoscience Aus-
Increased rainfall variability and reduced rainfall amount tralia. http://data.bioregionalassessments.gov.au/dataset/
decreases soil CO$_2$ flux in a grassland ecosystem. Glob 1556b944-731c-4b7f-a03e-14577c7e68db. Accessed 16
Chang Biol 11(2):322–334 August 2018
Hawkes CV, Waring BG, Rocca JD, Kivlin SN (2017) Histor- Maggi F (2019) BRTSim, a general-purpose computational
ical climate controls soil respiration responses to current solver for hydrological, biogeochemical, and ecosystem
soil moisture. Proc. Natl Acad. Sci. 114(24):6322–6327 dynamics. arXiv preprint arXiv:1903.07015
Heisler-White JL, Knapp AK, Kelly EF (2008) Increasing pre- Maggi F, Porporato A (2007) Coupled moisture and microbial
cipitation event size increases aboveground net primary dynamics in unsaturated soils. Water Resour Res. https://
productivity in a semi-arid grassland. Oecologia doi.org/10.1029/2006WR005367
158(1):129–140 Maggi F, Riley WJ (2010) Mathematical treatment of isotopo-
Hengl T, de Jesus JM, Heuvelink GB, Gonzalez MR, Kilibarda logue and isotopomer speciation and fractionation in bio-
M, Blagotić A (2017) SoilGrids250m: global gridded soil chemical kinetics. Geochim Cosmochim Acta
information based on machine learning. PLoS ONE 74(6):1823–1835
12(2):e169748 Maggi F, Gu C, Riley WJ, Hornberger GM, Venterea RT, Xu T,
Henriksen TM, Breland TA (1999) Nitrogen availability effects Oldenburg CM et al (2008) A mechanistic treatment of the
on carbon mineralization, fungal and bacterial growth, and dominant soil nitrogen cycling processes: model develop-
enzyme activities during decomposition of wheat straw in ment, testing, and application. J Geophys Res. https://doi.
soil. Soil Biol Biochem 31(8):1121–1134 org/10.1029/2007JG000578
Holland EA, Post WM, Matthews E, Sulzman J, Staufer R, Manzoni S, Porporato A (2007) A theoretical analysis of non-
Krankina O (2015) A global database of litterfall mass and linearities and feedbacks in soil carbon and nitrogen cycles.
litter pool carbon and nutrients. Data set. http://daac.ornl. Soil Biol Biochem 39(7):1542–1556
gov from Oak Ridge National Laboratory Distributed Manzoni S, Porporato A (2009) Soil carbon and nitrogen min-
Active Archive Center, Oak Ridge, Tennessee, USA eralization: theory and models across scales. Soil Biol
https://doi.org/10.3334/ORNLDAAC/1244 Biochem 41(7):1355–1379
Kieft TL (1987) Microbial biomass response to a rapid increase Manzoni S, Schimel JP, Porporato A (2012) Responses of soil
in water potential when dry soil is wetted. Soil Biol Bio- microbial communities to water stress: results from a meta-
chem 19(2):119–126 analysis. Ecology 93(4):930–938
Kim DG, Vargas R, Bond-Lamberty B, Turetsky MR (2012) Manzoni S, Moyano F, Kätterer T, Schimel J (2016) Modeling
Effects of soil rewetting and thawing on soil gas fluxes: a coupled enzymatic and solute transport controls on
review of current literature and suggestions for future decomposition in drying soils. Soil Biol Biochem
research. Biogeosciences 9(7):2459–2483 95:275–287
Kuypers MM, Marchant HK, Kartal B (2018) The microbial Maslin M, Austin P (2012) Uncertainty: climate models at their
nitrogen-cycling network. Nat Rev Microbiol 16(5):263 limit? Nature 486(7402):183
Lau JA, Lennon JT (2012) Rapid responses of soil microor- McCarl BA, Apland J (1986) Validation of linear programming
ganisms improve plant fitness in novel environments. Proc models. J Agric Appl Econ 18(2):155–164
Natl Acad Sci 109(35):14058–14062 Mench M, Martin E (1991) Mobilization of cadmium and other
Lee X, Wu HJ, Sigler J, Oishi C, Siccama T (2004) Rapid and metals from two soils by root exudates of Zea mays L.,
transient response of soil respiration to rain. Glob Chang Nicotiana tabacum L. and Nicotiana rustica L. Plant Soil
Biol 10(6):1017–1026 132(2):187–196
Li C, Frolking S, Frolking TA (1992) A model of nitrous oxide Menne MJ, Durre I, Vose RS, Gleason BE, Houston TG (2012)
evolution from soil driven by rainfall events: 1. Model An overview of the global historical climatology network-
structure and sensitivity. J Geophys Res daily database. J Atmos Ocean Technol 29(7):897–910
97(D9):9759–9776 Menne MJ, Durre I, Korzeniewski B, McNeal S, Thomas K, Yin
Li X, Miller AE, Meixner T, Schimel JP, Melack JM, Sickman X, Anthony S, Ray R, Vose RS, Gleason BE, Houston TG
JO (2010) Adding an empirical factor to better represent (2012) Global historical climatology network—Daily
the rewetting pulse mechanism in a soil biogeochemical (GHCN-Daily), Version 3. NOAA National Climatic Data
model. Geoderma 159(3–4):440–451 Center. https://doi.org/10.7289/V5D21VHZ. Accessed
Liu T, Wang L, Feng X, Zhang J, Ma T, Wang X, Liu Z (2018) September 2018
Comparing soil carbon loss through respiration and Monod J (1949) The growth of bacterial cultures. Annu Rev
leaching under extreme precipitation events in arid and Microbiol 3(1):371–394
semiarid grasslands. Biogeosciences 15(5):1627–1641
123
Biogeochemistry (2019) 144:197–214 213
Moretto AS, Distel RA, Didoné NG (2001) Decomposition and Schwinning S, Sala OE (2004) Hierarchy of responses to
nutrient dynamic of leaf litter and roots from palatable and resource pulses in arid and semi-arid ecosystems.
unpalatable grasses in a semi-arid grassland. Appl Soil Oecologia 141(2):211–220
Ecol 18(1):31–37 Sexstone AJ, Parkin TB, Tiedje JM (1985) Temporal response
Mouginot C, Kawamura R, Matulich KL, Berlemont R, Allison of soil denitrification rates to rainfall and irrigation 1. Soil
SD, Amend AS, Martiny AC (2014) Elemental stoi- Sci Soc Am J 49(1):99–103
chiometry of fungi and bacteria strains from grassland leaf Six J, Conant RT, Paul EA, Paustian K (2002) Stabilization
litter. Soil Biol Biochem 76:278–285 mechanisms of soil organic matter: implications for
Moyano FE, Manzoni S, Chenu C (2013) Responses of soil C-saturation of soils. Plant Soil 241(2):155–176
heterotrophic respiration to moisture availability: an Skiba U, Smith KA (2000) The control of nitrous oxide emis-
exploration of processes and models. Soil Biol Biochem sions from agricultural and natural soils. Chemosphere
59:72–85 2(3–4):379–386
Mulder A, Van de Graaf AA, Robertson LA, Kuenen JG (1995) Stark JM, Firestone MK (1995) Mechanisms for soil moisture
Anaerobic ammonium oxidation discovered in a denitri- effects on activity of nitrifying bacteria. Appl Environ
fying fluidized bed reactor. FEMS Microbiol Ecol Microbiol 61(1):218–221
16(3):177–183 Stern H, Dahni RR (2013) The distribution of climate zones
Navarro-Garcı́a F, Casermeiro MÁ, Schimel JP (2012) When across Australia: identifying and explaining changes dur-
structure means conservation: effect of aggregate structure ing the past century, in 25th Conference on Climate Vari-
in controlling microbial responses to rewetting events. Soil ability and Change. American Meteorological Society,
Biol Biochem 44(1):1–8 Austin
Neilen AD, Chen CR, Parker BM, Faggotter SJ, Burford MA Tang JY, Riley WJ (2019) A theory of effective microbial
(2017) Differences in nitrate and phosphorus export substrate affinity parameters in variably saturated soils and
between wooded and grassed riparian zones from farmland an example application to aerobic soil heterotrophic res-
to receiving waterways under varying rainfall conditions. piration. J Geophys Res Biogeosci 124(4):918–940
Sci Total Environ 598:188–197 Thomas RJ, Asakawa NM (1993) Decomposition of leaf litter
Nielsen UN, Ball BA (2015) Impacts of altered precipitation from tropical forage grasses and legumes. Soil Biol Bio-
regimes on soil communities and biogeochemistry in arid chem 25(10):1351–1361
and semi-arid ecosystems. Glob Chang Biol Tiedje JM, Sexstone AJ, Myrold DD, Robinson JA (1983)
21(4):1407–1421 Denitrification: ecological niches, competition and sur-
Placella SA, Brodie EL, Firestone MK (2012) Rainfall-induced vival. Antonie van Leeuwenhoek 48(6):569–583
carbon dioxide pulses result from sequential resuscitation Tietjen B, Schlaepfer DR, Bradford JB, Lauenroth WK, Hall
of phylogenetically clustered microbial groups. Proc Natl SA, Duniway MC, Wilson SD et al (2017) Climate change-
Acad Sci 109(27):10931–10936 induced vegetation shifts lead to more ecological droughts
Porporato A, Laio F, Ridolfi L, Caylor KK, Rodriguez-Iturbe I despite projected rainfall increases in many global tem-
(2003) Soil moisture and plant stress dynamics along the perate drylands. Glob Chang Biol 23(7):2743–2754
Kalahari precipitation gradient. J Geophys Res. https://doi. Vargas R, Detto M, Baldocchi DD, Allen MF (2010) Multiscale
org/10.1029/2002JD002448 analysis of temporal variability of soil CO$_2$ production
Reed SC, Cleveland CC, Townsend AR (2011) Functional as influenced by weather and vegetation. Glob Chang Biol
ecology of free-living nitrogen fixation: a contemporary 16(5):1589–1605
perspective. Annu Rev Ecol Evol Syst 42:489–512 Wang W, Fang J (2009) Soil respiration and human effects on
Richards LA (1931) Capillary conduction of liquids through global grasslands. Glob Planet Chang 67(1–2):20–28
porous mediums. J Appl Phys 1(5):318–333 Warren CR (2014) Response of osmolytes in soil to drying and
Riley WJ, Matson PA (2000) NLOSS: a mechanistic model of rewetting. Soil Biol Biochem 70:22–32
denitrified N$_2$O and N$_2$ evolution from soil. Soil Wickland KP, Neff JC (2008) Decomposition of soil organic
Sci 165(3):237–249 matter from boreal black spruce forest: environmental and
Riley WJ, Maggi F, Kleber M, Torn MS, Tang JY, Dwivedi D, chemical controls. Biogeochemistry 87(1):29–47
Guerry N (2014) Long residence times of rapidly decom- Wolery TJ (1992) EQ3/6: A software package for geochemical
posable soil organic matter: application of a multi-phase, modeling of aqueous systems: package overview and
multi-component, and vertically resolved model (BAMS1) installation guide (version 7.0). Lawrence Livermore
to soil carbon dynamics. Geosci Model Dev National Laboratory Livermore, CA
7(4):1335–1355 Xiang SR, Doyle A, Holden PA, Schimel JP (2008) Drying and
Schimel JP (2018) Life in dry soils: effects of drought on soil rewetting effects on C and N mineralization and microbial
microbial communities and processes. Annu Rev Ecol Evol activity in surface and subsurface California grassland
Syst 49:409–432 soils. Soil Biol Biochem 40(9):2281–2289
Schimel J, Balser TC, Wallenstein M (2007) Microbial stress- Xie P, Chen M, Shi W (2010) CPC unified gauge-based analysis
response physiology and its implications for ecosystem of global daily precipitation. In Preprints, 24th conference
function. Ecology 88(6):1386–1394 on hydrology, Atlanta, GA, American Meteor Society, vol
Schreiber F, Wunderlin P, Udert KM, Wells GF (2012) Nitric 2. https://www.esrl.noaa.gov/psd/
oxide and nitrous oxide turnover in natural and engineered Xu Y, Xu Z, Cai Z, Reverchon F (2013) Review of denitrifi-
microbial communities: biological pathways, chemical cation in tropical and subtropical soils of terrestrial
reactions, and novel technologies. Front Microbiol 3:372 ecosystems. J Soils Sediments 13(4):699–710
123
214 Biogeochemistry (2019) 144:197–214
Yan Z, Liu C, Todd-Brown KE, Liu Y, Bond-Lamberty B, Zhang X, Wan H, Zwiers FW, Hegerl GC, Min SK (2013)
Bailey VL (2016) Pore-scale investigation on the response Attributing intensification of precipitation extremes to
of heterotrophic respiration to moisture conditions in human influence. Geophys Res Lett 40(19):5252–5257
heterogeneous soils. Biogeochemistry 131(1–2):121–134
Yan Z, Bond-Lamberty B, Todd-Brown KE, Bailey VL, Li S,
Publisher’s Note Springer Nature remains neutral with
Liu C, Liu C (2018) A moisture function of soil hetero-
regard to jurisdictional claims in published maps and
trophic respiration that incorporates microscale processes.
institutional affiliations.
Nat Commun 9(1):2562
Yu K, Saha MV, D’Odorico P (2017) The effects of interannual
rainfall variability on tree-grass composition along Kala-
hari rainfall gradient. Ecosystems 20(5):975–988
123