ArticlePDF Available

Asymmetric responses of primary productivity to precipitation extremes: A synthesis of grassland precipitation manipulation experiments



Climatic changes are altering Earth's hydrological cycle, resulting in altered precipitation amounts, increased inter-annual variability of precipitation, and more frequent extreme precipitation events. These trends will likely continue into the future, having substantial impacts on net primary productivity (NPP) and associated ecosystem services such as food production and carbon sequestration. Frequently, experimental manipulations of precipitation have linked altered precipitation regimes to changes in NPP. Yet, findings have been diverse and substantial uncertainty still surrounds generalities describing patterns of ecosystem sensitivity to altered precipitation. Additionally, we do not know whether previously observed correlations between NPP and precipitation remain accurate when precipitation changes become extreme. We synthesized results from 83 case studies of experimental precipitation manipulations in grasslands worldwide. We used meta-analytical techniques to search for generalities and asymmetries of aboveground NPP (ANPP) and belowground NPP (BNPP) responses to both the direction and magnitude of precipitation change. Sensitivity (i.e., productivity response standardized by the amount of precipitation change) of BNPP was similar under precipitation additions and reductions, but ANPP was more sensitive to precipitation additions than reductions; this was especially evident in drier ecosystems. Additionally, overall relationships between the magnitude of productivity responses versus the magnitude of precipitation change were saturating in form. The saturating form of this relationship was likely driven by ANPP responses to very extreme precipitation increases, although there were limited studies imposing extreme precipitation change and there was considerable variation among experiments. This highlights the importance of incorporating gradients of manipulations, ranging from extreme drought to extreme precipitation increases into future climate change experiments. Additionally, policy and land management decisions related to global change scenarios should consider how ANPP and BNPP responses may differ, and that ecosystem responses to extreme events might not be predicted from relationships found under moderate environmental changes. This article is protected by copyright. All rights reserved.
Asymmetric responses of primary productivity to precipitation
extremes: A synthesis of grassland precipitation manipulation
Kevin R. Wilcox
Zheng Shi
Laureano A. Gherardi
Nathan P. Lemoine
Sally E. Koerner
David L. Hoover
Edward Bork
Kerry M. Byrne
James Cahill Jr.
Scott L. Collins
Sarah Evans
Anna K. Gilgen
Petr Holub
Lifen Jiang
Alan K. Knapp
Daniel LeCain
Junyi Liang
Pablo Garcia-Palacios
Josep Pe~
William T. Pockman
Melinda D. Smith
Shanghua Sun
Shannon R. White
Laura Yahdjian
Kai Zhu
Yiqi Luo
Department of Microbiology and Plant Biology, University of Oklahoma, Norman, OK, USA
School of Life Sciences, Arizona State University, Tempe, AZ, USA
Department of Biology & Graduate Degree Program in Ecology, Colorado State University, Fort Collins, CO, USA
Department of Integrative Biology, University of South Florida, Tampa, FL, USA
U.S. Department of Agriculture, Agriculture Research Service, Fort Collins, CO, USA
Agricultural, Food & Nutritional Science, University of Alberta, Edmonton, AB, Canada
Department of Environmental Science and Management, Humboldt State University, Arcata, CA, USA
Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada
Department of Biology, University of New Mexico, Albuquerque, NM, USA
Department of Integrative Biology, Department of Microbiology and Molecular Genetics and Kellogg Biological Station, Michigan State University, Hickory
Corners, MI, USA
Department of Environmental Systems Science, ETH Zurich, Zurich, Switzerland
Global Change Research Institute, Czech Academy of Sciences, Brno, Czech Republic
Area de Biodiversidad y Conservaci
on, Departamento de Biolog
ıa, Geolog
ıa, F
ısica y Qu
ımica Inorg
anica, Escuela Superior de Ciencias Experimentales y
ıa, Universidad Rey Juan Carlos, M
ostoles, Spain
CSIC, Global Ecology Unit CREAF-CSIC-UAB, Bellaterra, Catalonia, Spain
CREAF, Cerdanyola del Vall
es, Catalonia, Spain
College of Forestry, Northwest A & F University, Yangling, China
Environment and Parks, Government of Alberta, Edmonton, AB, Canada
Facultad de Agronom
ıa, Instituto de Investigaciones Fisiol
ogicas y Ecol
ogicas Vinculadas a la Agricultura (IFEVA), Consejo Nacional de Investigaciones
ıficas y T
ecnicas, Universidad de Buenos Aires, Buenos Aires, Argentina
Department of BioSciences, Rice University, Houston, TX, USA
Department of Biology, University of Texas, Arlington, TX, USA
Kevin Wilcox, Department of Microbiology
and Plant Biology, University of Oklahoma,
Norman, OK, USA.
Zheng Shi, Department of Microbiology and
Plant Biology, University of Oklahoma,
Norman, OK, USA.
Emails: and zheng.
Climatic changes are altering Earths hydrological cycle, resulting in altered precipita-
tion amounts, increased interannual variability of precipitation, and more frequent
extreme precipitation events. These trends will likely continue into the future, hav-
ing substantial impacts on net primary productivity (NPP) and associated ecosystem
*Authors co-led this manuscript.
Received: 7 September 2016
Accepted: 23 February 2017
DOI: 10.1111/gcb.13706
Glob Change Biol. 2017;110. ©2017 John Wiley & Sons Ltd
Funding information
Division of Emerging Frontiers, Grant/Award
Number: 1137293; National Institute of
Food and Agriculture, Grant/Award Number:
AFRI 2016-67012-25169; Cordis, Grant/
Award Number: SyG-2013-610028
IMBALANCE-P; Division of Environmental
Biology, Grant/Award Number: 1027319,
services such as food production and carbon sequestration. Frequently, experimental
manipulations of precipitation have linked altered precipitation regimes to changes
in NPP. Yet, findings have been diverse and substantial uncertainty still surrounds
generalities describing patterns of ecosystem sensitivity to altered precipitation.
Additionally, we do not know whether previously observed correlations between
NPP and precipitation remain accurate when precipitation changes become extreme.
We synthesized results from 83 case studies of experimental precipitation manipula-
tions in grasslands worldwide. We used meta-analytical techniques to search for
generalities and asymmetries of aboveground NPP (ANPP) and belowground NPP
(BNPP) responses to both the direction and magnitude of precipitation change. Sen-
sitivity (i.e., productivity response standardized by the amount of precipitation
change) of BNPP was similar under precipitation additions and reductions, but
ANPP was more sensitive to precipitation additions than reductions; this was espe-
cially evident in drier ecosystems. Additionally, overall relationships between the
magnitude of productivity responses and the magnitude of precipitation change
were saturating in form. The saturating form of this relationship was likely driven by
ANPP responses to very extreme precipitation increases, although there were lim-
ited studies imposing extreme precipitation change, and there was considerable vari-
ation among experiments. This highlights the importance of incorporating gradients
of manipulations, ranging from extreme drought to extreme precipitation increases
into future climate change experiments. Additionally, policy and land management
decisions related to global change scenarios should consider how ANPP and BNPP
responses may differ, and that ecosystem responses to extreme events might not
be predicted from relationships found under moderate environmental changes.
aboveground net primary productivity, belowground net primary productivity, biomass
allocation, climate change, grasslands, meta-analysis, root biomass
Global warming has intensified many hydrological processes (Hunt-
ington, 2006), and general circulation models predict diverse
responses of the water cycle to climate change. These include
increases or decreases in precipitation amount depending on geo-
graphic region (Hartmann & Andresky, 2013; Zhang et al., 2007),
increased interannual variability of precipitation, and increased fre-
quency of extreme wet and dry years (Easterling et al., 2000;
Jentsch & Beierkuhnlein, 2008; Singh, Tsiang, Rajaratnam, & Diffen-
baugh, 2013; Smith, 2011), all of which will likely have large effects
on primary productivity (Breshears et al., 2005; Del Grosso et al.,
2008; Gherardi & Sala, 2015; Weltzin et al., 2003). It is especially
important to understand the magnitude of these impacts in grass-
lands, most of which are strongly water limited (Knapp, Briggs, &
Koelliker, 2001; Sala, Parton, Joyce, & Lauenroth, 1988), cover a
large proportion of the terrestrial land surface (Chapin, Chapin, Mat-
son, & Vitousek, 2011), and provide valuable ecosystem services
(e.g., forage production, soil C storage: Sala, Yahdjian, Havstad, &
Aguiar, 2017). Observational precipitation studies have shown robust
relationships between climatic context (e.g., mean annual precipita-
tionMAP) and the sensitivity of ecosystems to altered precipitation
(i.e., the magnitude of change in production standardized by the
magnitude of precipitation change; Huxman et al., 2004; Sala, Gher-
ardi, Reichmann, Jobb
agy, & Peters, 2012; Guo et al., 2012). Yet,
findings from individual experiments often conflict with these broad
patterns (Byrne, Lauenroth, & Adler, 2013; Cherwin & Knapp, 2012;
Koerner & Collins, 2014; White, Cahill, & Bork, 2014; Wilcox, Blair,
Smith, & Knapp, 2016; Wilcox, Fischer, Muscha, Petersen, & Knapp,
2015), highlighting the need for synthesis across experiments (Car-
penter et al., 2009; Knapp et al., 2004).
Most existing knowledge concerning patterns of ecosystem sen-
sitivity to precipitation is based on aboveground net primary produc-
tivity (ANPP) data, even though belowground net primary
productivity (BNPP) represents a large proportion of NPP in many
grasslands (Sims & Sing, 1978). Furthermore, recent evidence sug-
gests that BNPP responses to altered precipitation are often differ-
ent in magnitude from those of ANPP (Byrne et al., 2013; Wilcox
et al., 2015). Existing theory states that plants shift biomass alloca-
tion (above- vs. belowground) depending on soil resource availability
(Bloom, Chapin, & Mooney, 1985; Gao, Chen, Lin, Giese, & Brueck,
2011; Giardina, Ryan, Binkley, & Fownes, 2003). If soil moisture
decreases due to drought, plants may increase allocation of carbohy-
drates to roots to maximize resource uptake, thus minimizing BNPP
loss while exacerbating ANPP loss. Alternately, if soil moisture
increases due to high precipitation levels, plants may allocate growth
aboveground to maximize light capture, resulting in larger responses
above- vs. belowground. Under this framework, we would predict
allocation patterns to offset BNPP increases and decreases under
increased and decreased precipitation, respectively. If generalizable,
these allocation patterns should lead to higher ANPP sensitivity than
BNPP. Empirical evidence for optimal allocation theory concerning
soil nutrients is abundant (McConnaughay & Coleman, 1999; Poorter
& Nagel, 2000; Poorter et al., 2012), whereas a smaller number of
studies have shown such allocation responses under altered soil
moisture (Milchunas & Lauenroth, 2001; Wilcox, Blair, & Knapp,
2016). However, some experimental evidence has shown BNPP to
be more sensitive than ANPP to changes in precipitation (Frank,
2007; Wilcox et al., 2015).
Another critical knowledge gap is whether the sensitivity of net
primary productivity (ANPP+BNPP) differs under precipitation
increases vs. decreases. Knapp and Smith (2001) showed that ANPP
responded more strongly in wet vs. dry years, and they posited that
this was due to drought tolerance mechanisms of resident plants.
Yet, we lack similar information for BNPP responses; currently, our
synthetic knowledge of BNPP responses to altered precipitation in
grasslands consists of a few experiments conducted across 23 sites
(e.g., Byrne et al., 2013; Fiala, Tuma, & Holub, 2009; Wilcox et al.,
2015), and portions of two meta-analyses with limited numbers of
studies documenting BNPP responses (Wu, Dijkstra, Koch, Penuelas,
& Hungate, 2011; Zhou et al., 2016). Recently, a number of addi-
tional grassland precipitation studies have reported BNPP responses
in individual ecosystems, and this presents an opportunity to exam-
ine and identify trends of BNPP responses to increased and
decreased precipitation amounts across studies.
As precipitation extremes such as widespread drought (e.g., Mid-
western United States in 2012) and high precipitation years become
more frequent (IPCC, 2013), understanding patterns of ecosystem
responses in extreme wet and dry years will be vital for assessing
future provisioning of ecosystem services. Currently much of our
knowledge comes from ecosystem responses to naturally occurring
climatic variation (Huxman et al., 2004; Knapp, Ciais, & Smith, 2016;
Knapp & Smith, 2001; La Pierre, Blumenthal, Brown, Klein, & Smith,
2016), or from experiments implementing mild-to-moderate alter-
ations relative to the inherent interannual variation at the site (e.g.,
Miranda, Armas, Padilla, & Pugnaire, 2011; Cherwin & Knapp, 2012;
Byrne et al., 2013; Koerner & Collins, 2014; Wilcox et al., 2015; all
sensu Knapp et al., 2015). Extreme precipitation manipulations are
more rare (Evans & Burke, 2013; Hoover, Knapp, & Smith, 2014;
Yahdjian & Sala, 2006), and syntheses of extreme precipitation
experiments are even more uncommon. Understanding whether
ecosystem responses to mild/moderate precipitation change are pre-
dictive of ecosystem responses to larger magnitude precipitation
changes is necessary to assess and update projections of future
ecosystem functioning under climate change scenarios.
We synthesized results from 83 experimental case studies that
measured ANPP and/or BNPP responses to manipulated precipita-
tion amounts to address these knowledge gaps. Precipitation alter-
ations in these case studies ranged in magnitude from 86% to
+431% relative to control plots. We used meta-analytical techniques
with this compiled data set to test the following hypotheses: (1)
BNPP is less sensitive than ANPP to altered precipitation amount;
(2) both ANPP and BNPP have greater sensitivity to increased vs.
decreased precipitation; (3) ANPP and BNPP sensitivities vary across
temperature and precipitation gradients; and (4a) ANPP and BNPP
responses to precipitation change are linear across the magnitude of
precipitation change; or (4b) ANPP and BNPP responses to precipita-
tion change are saturating across precipitation magnitudes. Com-
pared with a linear relationship, a saturating relationship would
indicate larger responses to extreme drought and lesser responses to
extreme precipitation increases (Knapp et al., 2016). Assessment of
these hypotheses is integral for assessing climate change impacts on
ecosystem services across larger spatial scales, as well as identifying
where/when impacts of climatic extremes are likely to be severe.
Data compilation
We collected publications that reported on primary productivity
responses to experimental precipitation manipulations in grassland
ecosystems by searching Web of Science. This included both
increased (+PPT) and decreased (PPT) precipitation treatments. We
used the following search terms to obtain papers from January 1st,
1900, to November 14th, 2016: (plant growthOR primary pro-
duct*OR plant product*OR ANPPOR BNPP) AND (altered
precipitationOR droughtOR decreased precipitationOR in-
creased precipitationOR increased summer precipitationOR de-
creased summer precipitationOR water additionOR water
reductionOR water treatment*) AND (herbaceousOR grass*)
AND (experiment*OR treatment*). The search resulted in 322
peer-reviewed papers. We then went through these papers and
removed all that did not meet the following criteria:
1. Study described a unique experiment. In the case of multiple
publications of the same responses, we used the latest pub-
lished paper. However, if the newest paper did not present
annual responses, we used the most recent paper presenting
annual data.
2. Plant communities were not artificially constructed, with the
exception of species assemblages planted to approximate com-
munity abundances of a natural study site.
3. Experiment was conducted in the field, or using monolith plots
in a greenhouse.
4. Treatment was consistent in all years.
5. Raw productivity values were reported (not just proportional
change, or biomass with woody species).
6. Productivity was measured <2 months after treatment stopped.
7. Total community productivity was reported (not just species
8. Reported primary productivity in mass per area units.
9. A control precipitation treatment was present, and replication
was greater than one.
10. Reported the amount or proportion of precipitation change.
11. Reported the standard deviation or standard error and sample
We also added multiple studies fitting these criteria obtained
via personal communications and from literature cited sections of
published papers. Production responses were excluded when ANPP
incorporated previous year woody growth or if belowground stand-
ing crop root biomass was measured instead of BNPP in all peren-
nial ecosystems. We limited our analyses to results from plots that
solely manipulated precipitationresults from plots receiving pre-
cipitation combined with other resource manipulations were
excluded. We compiled annual means, standard deviations, and
sample sizes of ANPP and BNPP from the literature or directly
from the authors. We also compiled mean annual temperature
(MAT), mean annual precipitation (MAP), and the amount and/or
proportion of precipitation added or subtracted in each year of the
study, obtained from the papers or authors. When studies reported
results from experiments conducted in different locations or having
multiple distinct treatments, these components were treated as
individual case studies. In total, our meta-analysis included 47 pub-
lished papers providing 83 precipitation manipulation case studies.
Most (62 of the 83) of the case studies occurred in North America
and Europe (Table S1). See Table 1 for summary information
regarding the compiled data set and Text S1 for a bibliography of
the papers used.
Calculating sensitivity for meta-analysis
We employed a meta-analytic approach to assess the overall sensi-
tivity of ANPP and BNPP to altered precipitation (Hedges, Gurevitch,
& Curtis, 1999; Luo, Hui, & Zhang, 2006). Sensitivity (Sens) was cal-
culated to represent the magnitude of response relative to the
amount of precipitation change, as previously used by others (e.g.,
Huxman et al., 2004; Knapp et al., 2016; Sala et al., 2012; Smith,
Wilcox, Power, Tissue, & Knapp, 2017; Wilcox, Blair, Smith, et al.,
2016; Wu et al., 2011). The benefit of this calculation is that ecosys-
tem responses are made comparable by standardizing by the magni-
tude of precipitation change:
Sens ¼
Xcare the productivity means across replicates of
treatment and control groups, respectively, and PPT
and PPT
the precipitation amounts in treatment and control groups, respec-
tively. A variance (v
) associated with sensitivity was approximated
using Equation (2).
vsens ¼1
 (2)
where s
and s
are standard deviations of treatment and control
groups, respectively. We validated our calculated variance using
Monte Carlo simulations (Text S2).
We aggregated sensitivity across studies by calculating a
weighted sensitivity estimate, similar to how previous meta-analyses
have aggregated response ratios (Hedges et al., 1999; Luo et al.,
2006). We calculated the weighted sensitivity (Sens
) as:
Sensþþ ¼Pk
where wis the weighting factor (w¼1
vsens) and kis the number of
studies. Standard error (SE) associated with Sens
was computed
using Equation (4).
SE Sensþþ
Significance of Sens
was assessed using a mixed-effects meta-
analytic model where individual studies are weighted by the follow-
ing equation:
where wis the weighting factor for an individual study, sis the
amount of variability not accounted for using the existing parameters
in the model, and v
is the study variance as calculated from Equa-
tion (2). Many of the experiments were conducted over multiple
years, and responses often varied interannually due, in large part, to
year-to-year variation in ambient rainfall. To account for this, case
study was designated as a random effect within the mixed-effects
model to account for pseudoreplication originating from studies
spanning multiple years.
TABLE 1 Summary information for experimental precipitation addition (+PPT) and reduction (PPT) treatments included in the meta-
DPPT (%)
DPPT (%)
Avg. duration
Range duration
MAP (mm)
MAP (mm)
MAT (°C)
MAT (°C)
+PPT 43.1 1.9431 3.2 123 551 1611526 7.5 4.816.3
PPT 48.7 18.186.0 2.0 14 572 1681632 10.4 1.622.0
All 49.7 86431 2.7 123 554 1611632 8.7 4.822.0
DPPT, percent change of precipitation manipulation relative to control plots; MAP, mean annual precipitation; MAT, mean annual temperature.
Sensitivity vs. climatic factors
To assess patterns of sensitivity across climatic gradients, we aver-
aged sensitivity values across years for ANPP or BNPP under
increased or decreased precipitation treatments within each case
study. This resulted in up to four sensitivity calculations per case
study, which occurred if a study measured ANPP and BNPP and
imposed both precipitation additions and reductions (one sensitivity
value each for PPT+ANPP, PPT- ANPP, PPT+BNPP, and PPT-
BNPP). Linear and various nonlinear models were compared using
AIC values to determine the most appropriate model structure for
correlating sensitivity with MAP and MAT (see Table S2 for identity
and form of relationships tested).
Magnitude of precipitation change vs.
magnitude of ANPP or BNPP response
Because the percentage change of precipitation varies in most stud-
ies from year to year, depending on ambient precipitation, we
assessed relationships between the percentage precipitation change
and the percentage productivity response for each year of each case
study. First, we did this to determine whether relationships between
the magnitude of production response and the magnitude of precipi-
tation change differed for ANPP vs. BNPP. We also determined
whether this relationship differed for precipitation additions vs. sub-
tractions. To this end, we calculated the percentage precipitation
change%DPPT =(PPT
, percentage productivity
response for ANPP%DANPP =(ANPP
, and
. We used percentage
change for this analysisinstead of the raw amount of precipitation
changebecause percentage change is comparable across ecosys-
tems spanning climatic gradients, whereas the absolute amount of
precipitation change may have very different implications in dry vs.
wet sites. We constructed a mixed-effects weighted-estimation
metaregression model (van Houwelingen, Arends, & Stijnen, 2002)
(Equation 5), with case study as a random factor, to look for signifi-
cant interaction terms between productivity type (ANPP vs BNPP)
and %DPPT as well as between treatment (increased vs decreased
precipitation) and %DPPT.
Second, we looked at whether these relationships were linear or
saturating. We did this through AIC comparisons of linear and natu-
ral log transformed (Table S4) models relating %DPPT with %DANPP
or %DBNPP. We again weighted the regressions using mixed-effects
weighted-estimation metaregression models. If the more appropriate
model is linear, this suggests that ecosystem responses to precipita-
tion extremes are proportional to their responses to mild or moder-
ate alterations in precipitation (i.e., levels of precipitation change
similar to those commonly found in historical precipitation records;
Knapp et al., 2015). If the more appropriate model is saturating, pri-
mary production responses under precipitation extremes may not
conform to patterns assessed under milder precipitation change.
Intercepts for %DPPT-%DANPP (or %DBNPP) regressions were set
at zero because, in an experimental framework, %DANPP (or %
DBNPP) should be zero when %DPPT is zero. Five outliers were
removed from the increased precipitation vs. %DANPP using a
threshold of a=0.05 (r-student: 6.29, 5.62, 4.36, 3.95, 3.80; all Bon-
ferroni p<.05). Results were qualitatively similar when these points
were included (Table S5). We did not detect publication bias when
examining plots showing the observed effect size and study variance
(funnel plots; Sterne & Egger, 2001).
All analyses were conducted in R(R Core Team, 2016), and
mixed-effects models were conducted using the NLME package
(Viechtbauer, 2010).
Across all studies, we found that the sensitivity of ANPP and BNPP
to both precipitation increases and decreases was greater than zero
(Figure 1; ANPP+:F
=32.2, p<.01; ANPP-: F
p<.01; BNPP+:F
=5.71, p=.02; BNPP-: F
=6.97, p=.02).
ANPP sensitivity to increased precipitation was 147% greater than
ANPP sensitivity to decreased precipitation (z=3.0, Tukey-adj.
p=.01). In contrast, BNPP sensitivity to precipitation increases and
decreases was not significantly different (z=1.8, Tukey-adj.
p=.28). Sensitivity to increased precipitation was 118% greater for
ANPP than BNPP (z=3.4, Tukey-adj. p<.01), but sensitivity to
decreased precipitation was not significantly different between
ANPP and BNPP (z=1.5, Tukey-adj. p=.44).
Precipitation sensitivity and background
The broad range of MAP and MAT of sites used in this meta-analysis
(Table 1) allowed us to examine patterns of precipitation sensitivity
FIGURE 1 Sensitivity of aboveground net primary productivity
(ANPP) and belowground net primary productivity (BNPP)
aggregated across experiments simulating increased (filled circles)
and decreased (open circles) precipitation. Sensitivity is calculated as
the amount of productivity response divided by the amount of
precipitation change. Numbers above symbols represent the number
of studies incorporated in each estimate. Different letters represent
different sensitivity at a=0.05, and error bars represent one
standard error from the mean
across large climatic gradients. We first tested whether sensitivity-
MAP and sensitivity-MAT relationships varied between precipitation
increases and decreases. We found that the sensitivity-MAP rela-
tionship was marginally different under precipitation increases vs.
decreases (F
=3.82, p=.05; different trend lines in Figure 2a).
The sensitivity of ANPP to precipitation additions was higher in arid
sites than in mesic sites. We found no relationship between MAP
and sensitivity of ANPP to precipitation reduction treatments (Fig-
ure 2a, Table S2). BNPP sensitivity-MAP relationships were not dif-
ferent for precipitation increases vs. decreases (F
=0.18, p=.67),
and sensitivity of BNPP to precipitation manipulations generally
decreased with MAP (Figure 2c). We did not find a significant inter-
action between MAT-sensitivity and precipitation increases vs.
decreases for ANPP (F
=0.04, p=.84) or BNPP (F
p=.36). We did not find a significant relationship between MAT
and ANPP sensitivity (Table S2), while BNPP sensitivity was greater
in colder sites (Figure 2b, d). See Table S2 for information about the
form, coefficients, and selection of each regression.
Comparing %DPPT and %DNPP linear vs.
saturating relationships
In our full models comparing the percentage change of productivity
(%DNPP) vs. percentage change of precipitation (%DPPT), we found
the natural log transformed model was a better fit than the linear
model (linear model AIC: 177.7, natural log model AIC: 175.2;
Table S3). Within the full natural log model, we found significant
interactions between %DPPT and precipitation direction
(precipitation increases vs. decreases) and between %DPPT and pro-
ductivity type (ANPP vs BNPP; Table S3). This was due to steeper %
DPPT-%DNPP slopes for ANPP (0.24 0.11; slope standard
error) vs BNPP (0.15 0.07) and for precipitation increases
(0.59 0.17) vs. decreases (0.22 0.07). Additionally, we found
significant interactions between %DPPT and precipitation direction
for both ANPP and BNPP analyzed separately (Table S3). These
interactions indicated that relationships between the magnitude of
productivity response and %DPPT may vary between ANPP and
BNPP as well as under precipitation increases vs. decreases.
We then analyzed %DPPT-%DNPP relationships separately for
ANPP and BNPP under precipitation increases and decreases to
assess whether linear or saturating (natural logarithmic) models bet-
ter fit the data for each category. We found that the saturating
model was a better fit for ANPP under precipitation increases (linear
model AIC 47.1 vs ln model AIC 44.6; Figure 3; Table S4). The bet-
ter fit of the saturating model was maintained even after removal of
the point having very large %DPPT, but the AIC differentiation was
weaker (Table S5). For ANPP under precipitation decreases and
BNPP under precipitation increases and decreases, we found weak
or no evidence for saturating models as a better fit to the data (Fig-
ure 3; Table S4).
Plant growth accounts for a large fraction of the terrestrial carbon
cycle and acts as an important buffer against fossil fuel emissions (Le
FIGURE 2 Relationships between site-
level climate and sensitivity of ANPP (a, b)
and BNPP (c, d) to increased (filled circles)
and decreased (open circles) precipitation
treatments (Trt). Climatic variables tested
were mean annual precipitation (a, c) and
mean annual temperature (b, d). Trendlines
in (a) are split into precipitation increases
and decreases because slopes were
significantly different. Trendlines in (bd)
represent overall regressions because
sensitivity-MAP or MAT relationships were
not different between increased and
decreased precipitation treatments
(nonsignificant Trt 9MAP or Trt 9MAT
interactions). Relationships without
trendlines and the dotted trendline in (a)
are not significant at a=0.1
e et al., 2015). Making accurate assessments of future carbon
budgets depends upon understanding influences of altered precipita-
tion on primary productivity. Fortunately, many recent precipitation
studies have documented responses of various components of NPP,
allowing for synthesis to identify key general patterns of herbaceous
responses to precipitation change. We found that sensitivity to pre-
cipitation change often differed between ANPP and BNPP, and
depended on whether precipitation was increased or decreased. We
also found that drier and cooler sites had higher sensitivity to precip-
itation change, especially to precipitation additions. Lastly, we found
evidence that productivity responses to increased precipitation may
saturate under very wet conditions. In the paragraphs below, we dis-
cuss implications and potential mechanisms underlying these find-
Overall ANPP vs. BNPP sensitivity to
precipitation change
BNPP was less sensitive than ANPP to increased precipitation treat-
ments (Figure 1), coinciding with previous work (Wu et al., 2011).
One interpretation is that root:shoot plasticity may be strong under
wet conditions (Knapp, 1984) and may result in decreased root allo-
cation to facilitate greater light capture during periods of high soil
resources (Joslin & Wolfe, 1998). Additionally, saturated soil mois-
ture conditions may limit root development (Kozlowski, 1997). These
changes in allocation may limit increased BNPP under increased pre-
cipitation, as well as heighten ANPP responses, compared with over-
all NPP responses. Secondly, longevity of live roots is typically
greater under moderately wet soil conditions (Facette, McCully, &
Canny, 1999), which may reduce the need/space for increased BNPP
to replace root systems under wetter conditions (Hayes & Seastedt,
1987). However, under extremely wet conditions, root lifespan can
decrease (Kozlowski, 1997), which may result in a threshold
response at a certain magnitude of precipitation increase. The
methodology used to measure roots in most of these studies (root
ingrowth cores) would likely not detect this second mechanism as
competition for space is not typically a factor for roots growing in
root ingrowth cores during much of the growing season. For this
reason, we suggest that plasticity in root:shoot allocation may be the
important factor driving different above vs. belowground productiv-
ity responses to increased precipitation observed in this study. Con-
versely, we found that ANPP and BNPP sensitivities to drought
were similar in magnitude. This may be due to a limitation of carbo-
hydrates available for growth (and thus allocation shifts) during peri-
ods of low soil moisture in drought treatments. In the early portion
of the growing season, soil moisture is often high in both drought
and control conditions due to winter inputs and low evaporation
rates occurring with cooler spring temperatures. However, as soil
moisture is depleted later in the growing season and drought effects
become more evident (Denton, Dietrich, Smith, & Knapp, 2016), car-
bohydrates may be similarly deficient for both root and aboveground
growth, which may limit the potential for changes in allocation
above- or belowground.
PPT sensitivity across climatic gradients
Previous observational studies have assessed patterns of climatic
context (e.g., MAP) vs. the sensitivity of primary production to
altered precipitation amount by examining the slope between pri-
mary production and annual precipitation (Huxman et al., 2004; Sala
et al., 2012). We used a similar sensitivity metric to assess whether
similar patterns exist for ANPP and BNPP based on experimental
data. We found the sensitivity of ANPP and BNPP to altered precipi-
tation was negatively related to MAP (Figure 2a, c), coinciding with
these observational studies (Huxman et al., 2004; Sala et al., 2012).
To our knowledge, this pattern has not been previously identified
through synthesis of experimental findings. Wu et al. (2011) found
no relationship between sensitivity and MAP, potentially because
they limited their analysis to linear regression, a relationship we
found to be substantially less predictive than the negative exponen-
tial relationship shown in Figure 2 (Table S2). The nonlinearity of the
ANPP sensitivity-MAP relationship (Figure 2a, c) highlights the
importance of understanding precipitation impacts in more xeric
FIGURE 3 Relationships between percentage responses of (a)
ANPP or (b) BNPP and the magnitude of experimental precipitation
manipulation (DPPT; increased (filled circles and decreased (open
circles)). Circle sizes are inversely correlated with the estimated
sampling variance of the percentage response of productivity, which
was used to weight points within the metaregression (i.e., larger
circles influence the regression more, see Methods). In panel (a), the
far right point is included in the regression, but results are
qualitatively similar when this point is removed (Table S5)
ecosystems, due to their potential for much higher sensitivity to pre-
cipitation increases than more mesic systems. We found no signifi-
cant relationship between MAP and ANPP sensitivity to drought.
The different sensitivities to precipitation increases vs. decreases in
arid ecosystems may be due to buffering capacity of drought toler-
ant plant traits (Knapp & Smith, 2001) possessed by dominant plant
species in these ecosystems.
We found BNPP was generally less sensitive to precipitation
changes in warmer ecosystems. This may be due to longer residence
times of added soil moisture in cooler sites, resulting in a higher pro-
portion of soil water being utilized by plants vs. being evaporated
directly from the soil and cooler sites having higher water use effi-
ciency (Vermeire, Heitschmidt, & Rinella, 2009). In addition, this
could be driven by deeper rooting profiles in cooler, high latitude
sites. Root growth tends to occur more homogenously throughout
the soil profile due to more homogenous soil moisture levels across
soil depths (e.g., Schenk & Jackson, 2002; Wilcox et al., 2015), so it
may be that the additional soil depths available for root production
in cooler systems leads to greater BNPP sensitivity to water addi-
ANPP and BNPP responses across magnitudes
of precipitation manipulation
If relationships between %DPPT and %DNPP are nonlinear and satu-
rating, then using linear models from historical precipitation-produc-
tivity regressions will not accurately predict the impacts of extreme
drought or extreme precipitation increases (Knapp et al., 2016). In
our full models, we found that the saturating model relating %DPPT
and %DANPP/BNPP was a better fit to the data than the linear
model (Table S3). Past observational studies have looked for, but
have not been able to identify, nonlinear patterns of primary produc-
tivity and precipitation change through site-level historical records of
ANPP and annual precipitation (Hsu & Adler, 2014). This may stem
from the fact that, by definition, years having extreme precipitation
amounts occur very infrequently in the historical record. For exam-
ple, Hoover et al. (2014) examined a 27-year ANPP-precipitation
data set from the Konza Prairie Biological Station, and in the context
of a 111-year precipitation record from this same area, found only
one year of ANPP data that was linked with extreme precipitation.
This highlights the value of climate change experiments for quantify-
ing future ecosystem responses under novel climatic conditions, as
experimental manipulations are able to push systems beyond histori-
cal climatic limits within sites (e.g., Evans, Byrne, Lauenroth, & Burke,
2011; Zhu, Chiariello, Tobeck, Fukami, & Field, 2016).
The nature of the overall saturating relationship could be driven
by (1) lower magnitude of productivity responses under extreme pre-
cipitation increases, (2) greater magnitude of productivity responses
under extreme drought, or (3) both. When we analyzed %DANPP vs
%DPPT separately for +PPT and PPT, we only found convincing
evidence for a saturating relationship for +PPT (Table S3; Figure 3).
We did not find that the saturating curve was a substantially better
fit for %DBNPP under +PPT or PPT. We think this may be due to a
few factors. First, the range of %DPPT was much greater for studies
increasing precipitation and measuring ANPP, so perhaps saturating
relationships are only evident under very extreme changes in precipi-
tation (Knapp et al., 2016). Second, perhaps extreme drought
impacts require multiple successive years of precipitation reductions
to fully develop (Hoover et al., 2014) due to depletion of soil water
or carbohydrate reserves. The majority (51 of 83; Table S6) of our
case studies were only 12 years in length, which may be why we
failed to detect logarithmic relationships under drought aloneeven
though we included a number of experiments with large drought
magnitudes (Table 1). Third, extreme heat waves often co-occur with
extreme drought during real-world climatic extremes. This is likely to
cause larger productivity responses than typically found in single fac-
tor drought experiments through further depletion of soil moisture
(Hoover et al., 2014).
We found substantial variation surrounding the trends shown in
Figures 2 and 3. Much of the variation seen in these relationships
may stem from cross-site variation of nonclimate characteristics,
such as soil texture, soil fertility, plant species composition, fire
regime, or presence/absence of grazing. For example, nitrogen limita-
tion may constrain a sites sensitivity to increased precipitation (Lad-
wig et al., 2012), or drought tolerant plant species may reduce
sensitivity of an ecosystem to changes in water availability (Wilcox,
Blair, Smith, et al., 2016). Unfortunately, many studies did not report
sufficient site-level characteristics for robust assessment of these
factors as drivers of ecosystem sensitivity to precipitation change.
We encourage future precipitation studies to report ecosystem char-
acteristics such as soil available nutrients, soil texture, and plant spe-
cies/functional composition. We also see considerable value in
conducting experiments within single sites manipulating a gradient of
precipitation levelsranging from extreme precipitation increases to
extreme precipitation decreaseswhile controlling for other vari-
ables that may affect sensitivity (e.g., Gherardi & Sala, 2015; Luo,
Jiang, Niu, & Zhou, 2017).
To provide accurate projections of how ecosystems will respond
to future precipitation scenarios, generalities informing patterns of
precipitation impacts on ecosystem function are needed. Using
meta-analytic methods, we explored overall ANPP and BNPP sensi-
tivity to precipitation change, the climatic context of sensitivity, and
how patterns of primary productivity change as precipitation
changes become extreme. First, we suggest that shifts in allocation
of biomass above- vs. belowground may lower NPP during high rain-
fall years, compared with expectations based on ANPP responses
alone. Second, we identified drier ecosystems as being especially
sensitive to precipitation increases, while cooler ecosystems were
somewhat more sensitive to any changes in precipitation. Lastly, we
found that previously identified asymmetriesshowing greater pro-
ductivity responses in wet vs. dry years (Knapp & Smith, 2001)
may be reversed when precipitation alterations become very
extreme. In the future, we advocate for (1) increased attention to
BNPP responses to extreme precipitation changes, and (2) more
long-term experiments that implement multiple levels of increased
and decreased precipitation amount.
Thanks to B. Hungate and two anonymous reviewers for their
insightful input. Also, we would like to thank all the principal investi-
gators, graduate students, technicians, etc. who contributed to the
precipitation experiments used in this analysis, thus making this
study possible. We would like to acknowledge all the funding agen-
cies supporting past, current, and future climate change experiments
including DEB-1456597, DEB-1027319, ERC SyG-2013-610028
IMBALANCE-P, USDA NIFA-AFRI 2016-67012-25169, and the NSF
Macrosystems Biology funded Extreme Drought in Grasslands
(EDGE) project.
Bloom, A. J., Chapin, F. S., & Mooney, H. A. (1985). Resource limitation
in plants-an economic analogy. Annual Review of Ecology and System-
atics,16, 363392.
Breshears, D. D., Cobb, N. S., Rich, P. M., Price, K. P., Allen, C. D., Balice,
R. G., ... Anderson, J. J. (2005). Regional vegetation die-off in
response to global-change-type drought. Proceedings of the National
Academy of Sciences of the United States of America,102, 15144
Byrne, K. M., Lauenroth, W. K., & Adler, P. B. (2013). Contrasting effects
of precipitation manipulations on production in two sites within the
Central Grassland Region, USA. Ecosystems,16, 10391051.
Carpenter, S. R., Armbrust, E. V., Arzberger, P. W., Stuart Chapin III, F.,
Elser, J. J., Hackett, E. J., ... Mangel, M. (2009). Accelerate synthesis
in ecology and environmental sciences. BioScience,59, 699701.
Chapin, F. S. III, Chapin, M. C., Matson, P. A., & Vitousek, P. (2011). Prin-
ciples of terrestrial ecosystem ecology. New York, New York, USA:
Cherwin, K., & Knapp, A. (2012). Unexpected patterns of sensitivity to
drought in three semi-arid grasslands. Oecologia,169, 845852.
Del Grosso, S., Parton, W., Stohlgren, T., Zheng, D., Bachelet, D., Prince, S.,
... Olson, R. (2008). Global potential net primary production pre-
dicted from vegetation class, precipitation, and temperature. Ecology,
89, 21172126.
Denton, E. M., Dietrich, J. D., Smith, M. D., & Knapp, A. K. (2016).
Drought timing differentially affects above-and belowground produc-
tivity in a mesic grassland. Plant Ecology,218, 317328.
Easterling, D. R., Meehl, G. A., Parmesan, C., Changnon, S. A., Karl, T. R.,
& Mearns, L. O. (2000). Climate extremes: Observations, modeling,
and impacts. Science,289, 20682074.
Evans, S. E., & Burke, I. C. (2013). Carbon and nitrogen decoupling under
an 11-year drought in the shortgrass steppe. Ecosystems,16,2033.
Evans, S. E., Byrne, K. M., Lauenroth, W. K., & Burke, I. C. (2011). Defin-
ing the limit to resistance in a drought-tolerant grassland: Long-term
severe drought significantly reduces the dominant species and
increases ruderals. Journal of Ecology,99, 15001507.
Facette, M., McCully, M., & Canny, M. (1999). Responses of maize roots
to drying-limits of viability. Plant, Cell and Environment,22, 1559
Fiala, K., Tuma, I., & Holub, P. (2009). Effect of manipulated rainfall on
root production and plant belowground dry mass of different grass-
land ecosystems. Ecosystems,12, 906914.
Frank, D. A. (2007). Drought effects on above-and belowground produc-
tion of a grazed temperate grassland ecosystem. Oecologia,152,
Gao, Y. Z., Chen, Q., Lin, S., Giese, M., & Brueck, H. (2011). Resource
manipulation effects on net primary production, biomass allocation
and rain-use efficiency of two semiarid grassland sites in Inner Mon-
golia, China. Oecologia,165, 855864.
Gherardi, L. A., & Sala, O. E. (2015). Enhanced precipitation variability
decreases grass- and increases shrub-productivity. Proceedings of the
National Academy of Sciences,112, 1273512740.
Giardina, C. P., Ryan, M. G., Binkley, D., & Fownes, J. H. (2003). Primary
production and carbon allocation in relation to nutrient supply in a
tropical experimental forest. Global Change Biology,9, 14381450.
Guo, Q., Hu, Z., Li, S., Li, X., Sun, X., & Yu, G. (2012). Spatial variations in
aboveground net primary productivity along a climate gradient in Eur-
asian temperate grassland: Effects of mean annual precipitation and
its seasonal distribution. Global Change Biology,18, 36243631.
Hartmann, H., & Andresky, L. (2013). Flooding in the Indus River basin
a spatiotemporal analysis of precipitation records. Global and Plane-
tary Change,107,2535.
Hayes, D., & Seastedt, T. (1987). Root dynamics of tallgrass prairie in wet
and dry years. Canadian Journal of Botany,65, 787791.
Hedges, L. V., Gurevitch, J., & Curtis, P. S. (1999). The meta-analysis of
response ratios in experimental ecology. Ecology,80, 11501156.
Hoover, D. L., Knapp, A. K., & Smith, M. D. (2014). Resistance and resili-
ence of a grassland ecosystem to climate extremes. Ecology,95,
van Houwelingen, H. C., Arends, L. R., & Stijnen, T. (2002). Advanced
methods in meta-analysis: Multivariate approach and meta-regres-
sion. Statistics in Medicine,21, 589624.
Hsu, J. S., & Adler, P. B. (2014). Anticipating changes in variability of
grassland production due to increases in interannual precipitation
variability. Ecosphere,5,115.
Huntington, T. G. (2006). Evidence for intensification of the global water
cycle: Review and synthesis. Journal of Hydrology,319,8395.
Huxman, T. E., Smith, M. D., Fay, P. A., Knapp, A. K., Shaw, M. R., Loik,
M. E., ... Pockman, W. T. (2004). Convergence across biomes to a
common rain-use efficiency. Nature,429, 651654.
IPCC 2013. Climate Change 2013: The Physical Science Basis. Contribu-
tion of Working Group I to the Fifth Assessment Report of the Inter-
governmental Panel on Climate Change. In T. F. Stocker, D. Qin, G.-
K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V.
Bex & P. M. Midgley (Eds.). 1535 pp. Cambridge, UK and New York,
NY, USA: Cambridge University Press.
Jentsch, A., & Beierkuhnlein, C. (2008). Research frontiers in climate
change: Effects of extreme meteorological events on ecosystems.
Comptes Rendus Geoscience,340, 621628.
Joslin, J. D., & Wolfe, M. H. (1998). Impacts of water input manipulations
on fine root production and mortality in a mature hardwood forest.
Plant and Soil,204, 165174.
Knapp, A. (1984). Water relations and growth of three grasses during
wet and drought years in a tallgrass prairie. Oecologia,65,3543.
Knapp, A., Briggs, J., & Koelliker, J. (2001). Frequency and extent of
water limitation to primary production in a mesic temperate grass-
land. Ecosystems,4,1928.
Knapp, A., Ciais, P., & Smith, M. (2016). Reconciling inconsistencies in
precipitation-productivity relationships: Implications for climate
change. New Phytologist,214,4147.
Knapp, A. K., Hoover, D. L., Wilcox, K. R., Avolio, M. L., Koerner, S. E.,
La Pierre, K. J., ... Smith, M. D. (2015). Characterizing differences in
precipitation regimes of extreme wet and dry years: Implications for
climate change experiments. Global Change Biology,21, 26242633.
Knapp, A. K., & Smith, M. D. (2001). Variation among biomes in temporal
dynamics of aboveground primary production. Science,291, 481484.
Knapp, A. K., Smith, M. D., Collins, S. L., Zambatis, N., Peel, M., Emery, S.,
... Andelman, S. J. (2004). Generality in ecology: Testing North
American grassland rules in South African savannas. Frontiers in Ecol-
ogy and the Environment,2, 483491.
Koerner, S. E., & Collins, S. L. (2014). Interactive effects of grazing,
drought, and fire on grassland plant communities in North America
and South Africa. Ecology,95,98109.
Kozlowski, T. T. (1997). Responses of weedy plants to flooding and salin-
ity. Tree Physiology Monograph,1,129.
La Pierre, K. J., Blumenthal, D. M., Brown, C. S., Klein, J. A., & Smith, M.
D. (2016). Drivers of variation in aboveground net primary productiv-
ity and plant community composition differ across a broad precipita-
tion gradient. Ecosystems,19, 521533.
Ladwig, L. M., Collins, S. L., Swann, A. L., Xia, Y., Allen, M. F., & Allen, E.
B. (2012). Above-and belowground responses to nitrogen addition in
a Chihuahuan Desert grassland. Oecologia,169, 177185.
Le Qu
e, C., Moriarty, R., Andrew, R. M., Canadell, J. G., Sitch, S., Kors-
bakken, J. I., ... Houghton, R. A. (2015). Global carbon budget 2015.
Earth System Science Data,7, 349396.
Luo, Y., Hui, D., & Zhang, D. (2006). Elevated CO2 stimulates net accu-
mulations of carbon and nitrogen in land ecosystems: A meta-analy-
sis. Ecology,87,5363.
Luo, Y., Jiang, L., Niu, S., & Zhou, X. (2017). Nonlinear responses of land
ecosystems to variation in precipitation. New Phytologist,207,57.
McConnaughay, K., & Coleman, J. (1999). Biomass allocation in plants:
Ontogeny or optimality? A test along three resource gradients. Ecol-
ogy,80, 25812593.
Milchunas, D., & Lauenroth, W. (2001). Belowground primary production
by carbon isotope decay and long-term root biomass dynamics.
Ecosystems,4, 139150.
Miranda, J. D., Armas, C., Padilla, F., & Pugnaire, F. (2011). Climatic
change and rainfall patterns: Effects on semi-arid plant communities
of the Iberian Southeast. Journal of Arid Environments,75, 1302
Poorter, H., & Nagel, O. (2000). The role of biomass allocation in the
growth response of plants to different levels of light, CO
, nutrients
and water: A quantitative review. Functional Plant Biology,27, 595
Poorter, H., Niklas, K. J., Reich, P. B., Oleksyn, J., Poot, P., & Mommer, L.
(2012). Biomass allocation to leaves, stems and roots: Meta-analyses
of interspecific variation and environmental control. New Phytologist,
R Core Team (2016). R: A language and environment for statistical com-
puting. Vienna, Austria.
Sala, O. E., Gherardi, L. A., Reichmann, L., Jobb
agy, E., & Peters, D.
(2012). Legacies of precipitation fluctuations on primary production:
Theory and data synthesis. Philosophical Transactions of the Royal
Society B: Biological Sciences,367, 31353144.
Sala, O. E., Parton, W. J., Joyce, L., & Lauenroth, W. (1988). Primary pro-
duction of the central grassland region of the United States. Ecology,
Sala, O. E., Yahdjian, L., Havstad, K., & Aguiar, M. R. (2017). Rangeland
ecosystem services: Nature0s supply and humans0demand. In D. D.
Briske (Ed.), Rangeland systems: Processes, management and challenges
(pp. 467489). New York, NY: Springer.
Schenk, H. J., & Jackson, R. B. (2002). The global biogeography of roots.
Ecological Monographs,72, 311328.
Sims, S. L., & Sing, J. S. (1978). The structure and function of ten western
North American grasslands: III. Net primary production, Turnover and
efficiencies of energy capture and water use. Journal of Ecology,66,
Singh, D., Tsiang, M., Rajaratnam, B., & Diffenbaugh, N. S. (2013). Precipi-
tation extremes over the continental United States in a transient,
high-resolution, ensemble climate model experiment. Journal of Geo-
physical Research: Atmospheres,118, 70637086.
Smith, M. D. (2011). An ecological perspective on extreme climatic
events: A synthetic definition and framework to guide future
research. Journal of Ecology,99, 656663.
Smith, M. D., Wilcox, K. R., Power, S. A., Tissue, D. T., & Knapp, A. K.
(2017). Assessing community and ecosystem sensitivity to climate
change toward a more comparative approach. Journal of Vegetation
Science,28, 235237.
Sterne, J. A. C., & Egger, M. (2001). Funnel plots for detecting bias in
meta-analysis: Guidelines on choice of axis. Journal of Clinical Epi-
demiology,54, 10461055.
Vermeire, L. T., Heitschmidt, R. K., & Rinella, M. J. (2009). Primary pro-
ductivity and precipitation-use efficiency in mixed-grass prairie: A
comparison of northern and southern US sites. Rangeland Ecology and
Management,62, 230239.
Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor
package. Journal of Statistical Software,36,148.
Weltzin, J. F., Loik, M. E., Schwinning, S., Williams, D. G., Fay, P. A.,
Haddad, B. M., ... Pockman, W. T. (2003). Assessing the response of
terrestrial ecosystems to potential changes in precipitation. BioScience,
53, 941952.
White, S. R., Cahill, J. F., & Bork, E. W. (2014). Implications of precipita-
tion, warming, and clipping for grazing resources in Canadian prairies.
Agronomy Journal,106,3342.
Wilcox, K. R., Blair, J. M., & Knapp, A. K. (2016). Stability of grassland soil
C and N pools despite 25 years of an extreme climatic and distur-
bance regime. Journal of Geophysical Research: Biogeosciences, B,121,
Wilcox, K. R., Blair, J. M., Smith, M. D., & Knapp, A. K. (2016). Does
ecosystem sensitivity to precipitation at the site-level conform to
regional-scale predictions? Ecology,97, 561568.
Wilcox, K. R., Fischer, J. C., Muscha, J. M., Petersen, M. K., & Knapp, A.
K. (2015). Contrasting above-and belowground sensitivity of three
Great Plains grasslands to altered rainfall regimes. Global Change Biol-
ogy,21, 335344.
Wu, Z., Dijkstra, P., Koch, G. W., Penuelas, J., & Hungate, B. A. (2011).
Responses of terrestrial ecosystems to temperature and precipitation
change: A meta-analysis of experimental manipulation. Global Change
Biology,17, 927942.
Yahdjian, L., & Sala, O. E. (2006). Vegetation structure constrains primary
production response to water availability in the Patagonian steppe.
Ecology,87, 952962.
Zhang, X., Zwiers, F. W., Hegerl, G. C., Lambert, F. H., Gillett, N. P., Solo-
mon, S., ... Nozawa, T. (2007). Detection of human influence on
twentieth-century precipitation trends. Nature,448, 461465.
Zhou, X., Zhou, L., Nie, Y., Fu, Y., Du, Z., Shao, J., ... Wang, X. (2016).
Similar responses of soil carbon storage to drought and irrigation in
terrestrial ecosystems but with contrasting mechanisms: A meta-ana-
lysis. Agriculture, Ecosystems and Environment,228,7081.
Zhu, K., Chiariello, N. R., Tobeck, T., Fukami, T., & Field, C. B. (2016).
Nonlinear, interacting responses to climate limit grassland production
under global change. Proceedings of the National Academy of Sciences,
113, 1058910594.
Additional Supporting Information may be found online in the sup-
porting information tab for this article.
How to cite this article: Wilcox KR, Shi Z, Gherardi LA, et al.
Asymmetric responses of primary productivity to precipitation
extremes: A synthesis of grassland precipitation manipulation
experiments. Glob Change Biol. 2017;00:110.
... In addition, our results in this article may ignore the effects of anthropogenic activities and climatic change owing to the large-scale spatial resolution, especially the effects of anthropogenic activities. Finally, prior studies showed that aboveground productivity is more sensitive to climate change than belowground productivity; thus, climate change may affect the proportion of aboveground and belowground ANPP [44]. However, these uncertainties may be decreased by utilizing localized simulation parameters and high-resolution remote sensing images in the future [4]. ...
Full-text available
Climate change and anthropogenic activities have had a profound effect on the variation in grassland productivity in the Tibetan Plateau in recent decades. Quantifying the impacts of climatic and anthropogenic variables on grassland productivity is a necessary step in making the management policies of a sustainable grassland ecosystem. Net primary productivity (NPP) is an important part of the terrestrial carbon cycle and can be used to assess vegetation growth. Based on the Carnegie–Ames–Stanford Approach model and statistical analysis method, in this study we estimated the variations in grassland potential NPP (PNPP), actual NPP (ANPP) and human-induced NPP (HNPP) in the Northwest Sichuan Plateau (NWSP) of the Southeast Tibetan Plateau from 2001 to 2020. Also, we assessed the contribution of climatic change and anthropogenic activities to grassland ANPP. The results showed that the average values of grassland ANPP, PNPP and HNPP in the whole NWSP increased at the rates of 3.81, 9.14 and 7.18 g C m−2 a−1, respectively. Grassland ANPP increased in 91.7% of the total area. Climate-oriented impacts led grassland ANPP to increase in 82.6% of the area, and temperature increase was the dominant factor. Additionally, anthropogenic activity was the major reason for the grassland ANPP’s decline (5.4% of the total area). Overall, our findings are beneficial for the formulation of practical countermeasures regarding climate change adaption and damaged grassland recovering in the plateau.
... However, many ecosystems are becoming more sensitive, in terms of productivity and greenness, to changes in water availability [172][173][174] , with dry-region sensitivity to precipitation increasing at a rate of 0.624% yr −1 from 1981 to 2015 175 . If vegetation productivity becomes more sensitive to water availability, droughts and dry anomalies could result in large losses to the carbon land sink 176,177 . In high-latitude systems, potential increases in the land carbon sink from longer growing seasons could be limited by increasingly severe and frequent seasonal water deficits 110,178 . ...
... We suggested that the ANPP gains in rainy years were larger than reduced in lower rainfall years and intensified with the MAP in desert steppes. Conversely, ANPP gains in wet years and reductions in dry years could be offset, causing A ANPP to decrease with the MAP in meadow steppes (Wilcox et al., 2017;Wu et al., 2018). In addition, the spatial pattern of A ANPP and A PUE was inconsistent with A P in temperate and alpine grasslands, respectively ( Figure S3). ...
Full-text available
In the face of accelerated global dryland expansion and grassland degradation, signaling grassland ecosystem state transitions is an ongoing challenge in ecology. However, there is still a lack of effective indicators and understanding of the mechanisms of grassland ecosystem state transitions at the continental scale. Here, we propose a framework that links ecosystem function-based indicators and critical slowing down (CSD) theory to reveal grassland state transitions. Across precipitation gradients, we quantified the statistical characteristics and spatial patterns in ANPP and PUE dynamics (variability, asymmetry, and sensitivity to precipitation and temperature) in Eurasian grasslands. We show that the CV ANPP , CV PUE , A ANPP , A PUE , S PUE-P , and S ANPP-P of temperate steppes were significantly higher than those of alpine steppes, while the S PUE-T and S ANPP-T were the inverse. In temperate grasslands, A ANPP , A PUE , and S ANPP-P indicated the transition of typical steppes, and CV ANPP , A PUE , and S PUE-T indicated the transition from meadow to typical steppes. In alpine grasslands, A PUE indicated the transition between alpine deserts and alpine steppes, and A ANPP and S ANPP-P indicated the transition between alpine steppes and meadow steppes. The interannual variability of precipitation strongly affected xerophyte proportion and demographic processes, which control state transitions in low-resilience grasslands. Community structures and limiting factors (nutrient, light, and/or temperature) regulate state transitions in high-resilience grasslands. Our results demonstrate that function-based indicators are predictive of impending state transitions of temperate and alpine grasslands, highlighting the complementation of ANPP and PUE dynamics that have the potential for predicting grassland ecosystem regime shifts and their underlying mechanisms.
... Observed effects of drought on plants include reductions in net primary productivity and species richness as well as altered carbon cycling (Hoover et al. 2018). However, results remain often inconclusive and mechanisms of drought responses across ecosystems remain difficult to identify (Wu et al. 2011, Wilcox et al. 2017, Hoover et al. 2018. Based on this prevalent ambiguity, the question of whether system-specific peculiarities or experimental artifacts are contributing to this inconsistency between studies is repeatedly raised in literature (Felton andSmith 2017, Hoover et al. 2018). ...
Lacking comparability among rainfall manipulation studies is still a major limiting factor for generalizations in ecological climate change impact research. A common framework for studying ecological drought effects is urgently needed to foster advances in ecological understanding the effects of drought. In this study, we argue, that the soil–plant–atmosphere‐continuum (SPAC), describing the flow of water from the soil through the plant to the atmosphere, can serve as a holistic concept of drought in rainfall manipulation experiments which allows for the reconciliation experimental drought ecology. Using experimental data, we show that investigations of leaf water potential in combination with edaphic and atmospheric drought – as the three main components of the SPAC – are key to understand the effect of drought on plants. Based on a systematic literature survey, we show that especially plant and atmospheric based drought quantifications are strongly underrepresented and integrative assessments of all three components are almost absent in current experimental literature. Based on our observations we argue, that studying dynamics of plant water status in the framework of the SPAC can foster comparability of different studies conducted in different ecosystems and with different plant species and can facilitate extrapolation to other systems, species or future climates.
It is important to elucidate the changing distribution pattern of net primary productivity (NPP) to mechanistically understand the changes in aboveground and belowground ecosystem functions. In water-scarce desert environments, snow provides a crucial supply of water for plant development and the spread of herbaceous species. Yet uncertainty persists regarding how herbaceous plants' NPP allocation responds to variation in snow cover. The goal of this study was to investigate how variation in snow cover in a temperate desert influenced the NPP allocation dynamics of herbaceous species and their resistance to environmental change in terms aboveground and belowground productivity. In the Gurbantunggut Desert, wintertime snow cover depth was adjusted in plots by applying four treatments: snow removal (-S), ambient snow, double snow (+S), and triple snow (+2S). We examined their species richness, aboveground NPP (ANPP), belowground NPP (BNPP), and the resistance of ANPP and BNPP. We found that species diversity of the aboveground community increased significantly with increasing snow cover and decreased significantly Pielou evenness in plots. This resulted in greater ANPP with increasing snow cover; meanwhile, BNPP first increased and then decreased with increasing snow cover. However, this productivity in different soil layers responded differently to changed snow cover. In the 0-10 cm soil layer, productivity first rose and then declined, while it declined linearly in both the 10-20 cm and 20-30 cm soil layers, whereas in the 30-40 cm soil layer it showed an increasing trend. Belowground resistance would increase given that greater snow cover improved the BNPP in deeper soil and maintained the resource provisioning for plant growth, thus improving overall belowground stability. These results can serve as a promising research foundation for future work on how the functioning of desert ecosystems becomes altered due to changes in plant community expansion and, in particular, changes in snow cover driven by global climate change.
Full-text available
Assessing the dynamics of grassland functioning is critical for gaining an understanding of their feedback on rising aridity. In attempting to understand the response of grassland ecosystem functioning to aridity, the (i) relationships between biomass productivity (above- and belowground biomass: AGB and BGB, and their partitioning: BGB:AGB) and seasonal and annual aridity, and (ii) biomass allocation pattern between the AGB and BGB of C3- and C4-dominated grasslands in humid temperate, humid savanna, cold steppe, and savanna ecoregions were assessed. Results reveal that biomass productivity and its partitioning responded significantly to differences in growing season aridity, but the response patterns were not consistent for ecoregions. The decreased annual and seasonal biomass partitioning in humid savanna and cold steppe was associated with increased AGB and decreased BGB with accelerated aridity. There was a significant positive correlation in the biomass allocation pattern between the AGB and BGB of plants in three ecoregions, which supports the optimal partitioning theory. This study reveals that growing season aridity, rather than annual aridity, is the primary factor of biomass productivity and partitioning in the studied grasslands. These findings have significant repercussions for predicting ecosystem functioning and stability, restoring degraded ecosystems, and ensuring the sustainable management of grassland biodiversity. HIGHLIGHTS Ecosystem functioning under aridity has been assessed for four grassland ecoregions.; Significant changes in growing season biomass resulted from increasing growing season aridity.; Above- and belowground biomass showed a positive correlation and supported optimal partitioning theory.;
Precipitation and temperature are the two major drivers of species distribution on the earth. Change in precipitation has severe effects on the species composition of all ecosystems including grassland. In the present study, we have tried to assess the effect of precipitation on two major functional groups of tropical grassland i.e. graminoid (grasses and sedges) and forbs (herbaceous flowering plants). The study was performed in three rainout shelters with three different rain doses (16T, 11T, and 8T) and one unsheltered plot (open C) with ambient rain. Each sheltered and unsheltered plot has three 1 × 1 m randomly assigned subplots of uninvaded indigenous grassland plots (NIG). The study revealed that the aboveground net primary production (ANPP) of graminoids + forbs of the tropical grassland increase with the increase in precipitation. A significant positive correlation of ANPP was found with total inorganic – N (TIN) and soil moisture (SM) and a significant negative with microbial biomass nitrogen (MBN). Regression analysis reveals that after soil moisture, total inorganic – N and N – mineralization are the major determinants for the ANPP. However, when graminoid and forb species are studied separately, graminoids showed a positive response to increased precipitation while forbs did not show such a response. Indicating that the major contributor to the ANPP response toward precipitation increase is graminoid species in a dry tropical grassland. The study indicates the sensitivity of Indian grassland to the change in rainfall quantity, as studied forbs species decrease in both low and high precipitation. Showing that in tropics, forbs species may extinct as of their narrow range of tolerance due to precipiation change, in turn affecting the biodiversity of the area. This is the new possibility of research for researchers around the world. Moreover, to draw any conclusion a detailed study considering the nature of resource acquisition, root length, root architect, and competitive behavior among graminoids and forbs must be done separately, in relation to the precipitation.
Full-text available
Understanding precipitation controls on functional diversity is important in predicting how change in rainfall patterns will alter plant productivity in the future. Trait‐based approaches can provide predictive knowledge about how certain species will behave and interact with the community. However, how functional diversity relates to above‐ and below‐ground biomass production in variable rainfall conditions remains unclear. Here, we tested the role of mass ratio and niche complementarity hypotheses in shaping above‐ and below‐ground biomass–functional diversity relationships in seasonal drought. We implemented a fully crossed experiment that manipulated drought timing (fall dry, spring dry, consistent dry and ambient rainfall) and community composition (grass‐dominated, forb‐dominated and mixed grass–forb) in a California annual grassland. Plant communities with mixed functional groups showed higher above‐ and below‐ground biomass than either the grass‐ or forb‐dominant communities. We found divergent functional diversity–biomass relationships for above‐ and below‐ground biomass. Above‐ground biomass decreased with community weighted means (CWMs) of specific leaf area (SLA) and height, supporting the mass ratio hypothesis, which posits that dominant species with specific traits drive biomass production of the community. Below‐ground biomass showed no evidence of either mass ratio hypothesis or niche complementarity. While biomass was largely unaffected by the timing of drought in one season, we found community‐wide functional trait shifts in response to rainfall treatments. Above‐ground traits shifted to higher SLA in consistent dry compared to ambient. Below‐ground traits shifted to longer, finer and denser roots in fall and consistent dry, and shorter and coarser roots in spring dry. Functional diversity buffered biomass production by enabling shifts in above‐ and below‐ground functional traits across variable rainfall conditions. Read the free Plain Language Summary for this article on the Journal blog.
Full-text available
(1) Estimation of grazing livestock intake is the basis for studying animal–plant relationships and the nutritional status of grazing livestock and has important implications for grassland composition and productivity. (2) We used the saturated alkanes method to determine the feed intake and vegetation nutrient digestibility of livestock at different grazing intensities and in different months. (3) We found that C31 had the highest concentration in both pasture and fecal output, and the average recovery of C31 was 77.99%. The different grazing intensities significantly affected livestock intake. As the grazing intensity increased, there was a decreasing trend of livestock intake and the highest livestock feed intake was 6.11 kg DM/day in light grazing. With the increase in grazing season months, the highest livestock intake was 6.67 kg DM/day in the cold period in September. The month also had a significant effect on the digestibility of livestock for all nutrient variables when compared to the grazing intensity. Livestock weight and medium palatability species are more important for livestock intake. (4) Our study provides a more accurate measurement of grazing livestock intake, which can be used as a reference for the scientific management of grazing livestock and the rational use of grazing pastures.
1. Global warming intensifies the hydrological cycle and may result in changes in the frequency and intensity of precipitation events. However, a knowledge gap still exists about the general influences of intra-annual precipitation variability on grassland plant diversity and ecosystem function at the global scale. 2. Here, we synthesized field manipulative experiments from 66 publications to quantify the effects of intra-annual precipitation variability increases (IPVI) on community biomass and plant diversity in grasslands worldwide. 3. At the global scale, we found that IPVI generally increased grassland community above-ground biomass (AGB) by 6.90%, and decreased grass biomass and soil ammonium content by 10.38% and 25.35% respectively. IPVI increased and decreased total AGB in arid and humid regions, respectively, and IPVI enhanced grassland below-ground biomass and plant species richness in arid regions, but showed no effect in humid regions. Furthermore, moderate IPVI stimulated plant species richness and community biomass in arid grasslands, while extreme IPVI impaired them in mesic grasslands. Changes in total AGB under IPVI were related to changes in the biomass of plant functional groups, species richness and soil moisture. Structural equation modelling demonstrated that climate conditions (mean annual temperature and mean annual precipitation) and background soil properties (soil sand content and soil organic carbon content) jointly regulated grassland total AGB responses to IPVI across climate regions. 4. Synthesis. Overall, our study shows divergent responses of grassland productivity and diversity to IPVI in arid and humid regions. Accounting for climate, edaphic properties and precipitation variability can help improve our ability to predict the responses of grassland ecosystems to future changes in precipitation at global scales.
Full-text available
Ecosystem services are the benefits that society receives from nature, including the regulation of climate, the pollination of crops, the provisioning of intellectual inspiration and recreational environment, as well as many essential goods such as food, fiber, and wood. Rangeland ecosystem services are often valued differently by different stakeholders interested in livestock production, water quality and quantity, biodiversity conservation, or carbon sequestration. The supply of ecosystem services depends on biophysical conditions and land-use history, and their availability is assessed using surveys of soils, plants, and animals. The demand for ecosystem services depends on educational level, income, and location of residence of social beneficiaries. The demand can be assessed through stakeholder interviews, questionnaires, and surveys. Rangeland management affects the supply of different ecosystem services by producing interactions among them. Trade-offs result when an increase in one service is associated with a decline in another, and win–win situations occur when an increase in one service is associated with an increase in other services. This chapter provides a conceptual framework in which range management decisions are seen as a challenge of reconciling supply and demand of ecosystem services.
Full-text available
Plant communities can vary widely in their sensitivity to changing precipitation regimes, as reported by Byrne et al., Mulhouse et al., and Sternberg et al. in this issue of Journal of Vegetation Science. But to understand why communities differ in their sensitivity, we argue that clearly defined metrics of sensitivity and coordinated research approaches are needed to elucidate mechanisms. This article is protected by copyright. All rights reserved.
Full-text available
This article is a Commentary on Knapp et al., 214: 41–47.
Full-text available
Climate models forecast an intensification of the global hydrological cycle with droughts becoming more frequent and severe, and shifting to times when they have been historically uncommon. Droughts, or prolonged periods of precipitation deficiency, are characteristic of most temperate grasslands, yet few experiments have explored how variation in the seasonal timing of drought may impact ecosystem function. We investigated the response of above- and belowground net primary production (ANPP & BNPP) to altered drought timing in a mesic grassland in NE Kansas. Moderate drought treatments (25% reduction from the mean growing season precipitation [GSP]) were imposed by erecting rainout shelters in late spring (LSP), early summer (ESM), and mid-summer (MSM, n = 10 plots/treatment). These treatments were compared to two controls (long-term average GSP [LTA] and ambient GSP [AMB]) and a wet treatment (+30% of the long-term average GSP [WET]). We found that LSP drought did not significantly reduce ANPP relative to control plots while the ESM and MSM drought did despite equivalent reductions in soil moisture. In contrast, the WET treatment did not affect ANPP. Neither the WET nor the drought treatments altered BNPP as compared to the controls. Our results suggest that forecasts of ecosystem responses to climate change will be improved if both the seasonal timing of alterations in precipitation as well as differential responses of above- and belowground productivity to drought are incorporated into models.