Ecology Letters, (2017) 20: 663–672
LETTER
Kai Yue,1,4 Dario A. Fornara,2 Wanqin Yang,1 Yan Peng,3 Changhui Peng,4,5 Zelin Liu4 and Fuzhong Wu1*
doi: 10.1111/ele.12767
Influence of multiple global change drivers on terrestrial carbon storage: additive effects are common Abstract The interactive effects of multiple global change drivers on terrestrial carbon (C) storage remain poorly understood. Here, we synthesise data from 633 published studies to show how the interactive effects of multiple drivers are generally additive (i.e. not differing from the sum of their individual effects) rather than synergistic or antagonistic. We further show that (1) elevated CO2, warming, N addition, P addition and increased rainfall, all exerted positive individual effects on plant C pools at both single-plant and plant-community levels; (2) plant C pool responses to individual or combined effects of multiple drivers are seldom scale-dependent (i.e. not differing from single-plant to plant-community levels) and (3) soil and microbial biomass C pools are significantly less sensitive than plant C pools to individual or combined effects. We provide a quantitative basis for integrating additive effects of multiple global change drivers into future assessments of the C storage ability of terrestrial ecosystems. Keywords Additive interaction, drought, elevated CO2, increased rainfall, nitrogen addition, phosphorus addition, warming. Ecology Letters (2017) 20: 663–672
Since the start of the Industrial Revolution human activities have significantly contributed to current global change with increasingly negative consequences for the natural environment and human well-being (Dillon et al. 2010; Frank et al. 2015). One of the most important human-induced global change effects is associated with the alteration of the biogeochemical cycle of carbon (C) (Liu & Greaver 2010; Lu et al. 2013; van Groenigen et al. 2014). Once ecosystems’ ability to cycle and store C is altered, this will have positive or negative feedback on a range of key global change drivers either reinforcing or diminishing their net effect on ecosystem functioning (Lu et al. 2013). Among other global change drivers, elevated atmospheric carbon dioxide (eCO2), rising air temperature (i.e. warming), atmospheric nitrogen (N) deposition, phosphorus (P) fertilisation and altered rainfall regimes remain the most pervasive (IPCC 2014). These global change drivers may significantly influence terrestrial ecosystem C storage and their individual effects have been extensively assessed over the past two decades. For example, several recent data syntheses have examined the impact of eCO2 on both plant and soil C storage (Luo et al. 2006; Norby & Zak 2011; van Groenigen et al. 2014), suggesting that eCO2 stimulates net accumulation of C in terrestrial
ecosystems. Likewise, N addition has been found to significantly increase both plant and soil C pools (Liu & Greaver 2010; Lu et al. 2011), but significantly decrease soil microbial biomass C (MBC) pools (Treseder 2008). Recent meta-analysis studies show how warming can also stimulate C storage in plant pools but not in soil pools (Lu et al. 2013), whereas increased P fertilisation can significantly increase both plant aboveground and belowground C pools (Li et al. 2016). A recent meta-analysis suggested that either drought or greater rainfall could increase soil C pools by 1.45 and 1.27%, respectively (Zhou et al. 2016b). Although numerous studies have examined how C storage in plants, soils or soil microbial biomass might change either under one or two global change drivers, very few studies have addressed how ecosystem C storage might respond to the simultaneous and interacting effects of multiple global change drivers. Poor understanding of how multiple global change drivers might influence terrestrial C storage will limit our ability to incorporate these potential effects into predictive biogeochemical models. Interactive effects among multiple concurrent global change drivers on ecosystem processes are very common (Dukes et al. 2005; Reich et al. 2006; Li et al. 2016) and can have complex significant impacts on ecosystem C cycling and storage. For example, eCO2 effects on plant C pools might depend on N and water availability (Reich et al. 2014), indicating that a
1
4
INTRODUCTION
Long-term Research Station of Alpine Forest Ecosystems, Provincial Key Labo-
Department of Biological Science, Institute of Environment Sciences,
ratory of Ecological Forestry Engineering, Institute of Ecology and Forestry,
University of Quebec at Montreal, Case Postale 8888, succursale Centre-Ville,
Sichuan Agricultural University, 211 Huimin Road, Wenjiang District, Chengdu
Montreal, QC H3C 3P8, Canada
611130, Sichuan, China
5
2
estry, Northwest A & F University, No. 3 Taicheng Road, Yangling 712100,
Sustainable Agri-Food Sciences Division, Agri-Food & Biosciences Institute
Laboratory for Ecological Forecasting and Global Change, College of For-
(AFBI), Newforge Lane, Belfast BT9 5PX, UK
China
3
*Correspondence: E-mail:
[email protected]
Department of Geosciences and Natural Resource Management, University
of Copenhagen, Rolighedsvej 23, DK-1958 Frederiksberg C, Denmark
© 2017 John Wiley & Sons Ltd/CNRS
664 K. Yue et al.
large stimulation of plant C pools by eCO2 may especially occur under simultaneous N addition and increased rainfall. However, eCO2-induced positive effect on plant C pools may be suppressed by drought. If warming tends to increase plant C pools as well as microbial activity (Lu et al. 2013), it can also induce warming-associated droughts making the response of C cycle to increased temperatures more complex and difficult to predict. Moreover, statistical syntheses of ecological data from multiple experimental studies suggest that the stimulation effect of N addition on plant C pools can be enhanced by simultaneous P additions (Li et al. 2016). All together these results suggest that evaluating the interaction among multiple global change drivers is crucial to better understand and predict changes in C storage across different ecosystem pools (i.e. plants, soils and soil microbial biomass). At present, there is still a significant knowledge gap on how different terrestrial ecosystem C pools might respond to individual vs. combined effects of multiple global change drivers, and whether combined effects might or might not be additive. Additive interactions occur when the combined effect of multiple drivers is equal to or not significantly different from the sum of the individual effects. Net effects could also be synergistic (i.e. the combined effect is greater than the sum of the individual effects) or antagonistic (i.e. the combined effect is weaker than the sum of the individual effects) (Crain et al. 2008; Zhou et al. 2016a). The combined effect of eCO2 and warming is frequently assumed to be additive (Norby & Luo 2004), but non-additive effects of these two drivers were also observed in grassland ecosystems (Mueller et al. 2016). Previous experimental evidence shows how the magnitude of the response of different ecosystem functions to global change might decline with higher-order interactions of global change drivers (i.e. interactions of a larger number of individual factors), which would indicate a lack of additive effects (Leuzinger et al. 2011). This view is supported by the results of a recent synthesis, which shows how the response of plant biomass to the combined effects of elevated CO2 and warming was non-additive (Dieleman et al. 2012). However, a different meta-analysis study revealed that the combined effects of multiple global change drivers were generally the additive effects of single-driver treatments (Yuan & Chen 2015). Here, we use a meta-analysis approach and carry out a comprehensive data synthesis from 633 published field manipulative studies where we explicitly compare effect sizes of individual, combined and interactive effects of eCO2, warming, N addition, P addition, increased rainfall and drought on the size of C pools in plants, soils and soil microbial biomass. The main objectives of this study are to: (1) address individual vs. combined effects of multiple major global change drivers on important terrestrial C pools (i.e. plants, soils and soil microbial biomass), and (2) investigate whether the effects of multiple global change drivers on terrestrial C storage are additive or not. We hypothesise that (1) terrestrial ecosystem C pools will be significantly affected by the individual effects of different global change drivers; (2) the combined effects of two global change drivers on terrestrial C pools will not be significantly different from the sum of the corresponding individual effects (i.e. the interaction of two drivers will be additive rather than synergistic or antagonistic) and (3) the © 2017 John Wiley & Sons Ltd/CNRS
Letter
responses of terrestrial C pools to either individual or combined effects of key global change drivers will be influenced by moderator variables such as ecosystem type and experimental settings.
METHODS
Data compilation
We searched for peer-reviewed journal articles using ISI Web of Science and Google Scholar in October 2016 with no restriction on publication year. We focused on experimental studies which included key global change drivers: (1) eCO2, (2) warming, (3) N addition, (4) P addition, (5) increased rainfall, (6) drought and any combination of these six drivers. More than 4000 published articles, which were related to changes in C pools (plant, soil and/or microbial biomass) under experimental manipulations of these global change drivers, were collected. To minimise publication bias, we used six criteria to select appropriate studies, which were listed in Appendix S1. As to the soil C pool, we only included surface mineral soil samples with a maximum depth of 30 cm to increase sensitivity for detecting small effects (van Groenigen et al. 2006). To meet the statistical assumption of independence among observations in the meta-analysis (Hedges et al. 1999), we used values from the last measurement if multiple measurements were taken at different times during the study period. Different treatment levels, plant species and/or ecosystems within the same primary study were considered as different observations. However, these observations are not strictly independent and thus may be correlated. To deal with this kind of data compilation, we followed the approach used by previous studies (Vil a et al. 2011; Ferreira et al. 2015), which is described in detail in Appendix S1. Because of limited available data for three or more combined treatments, we considered only two-driver pairs in this study. After extraction, a total of 633 articles representing 3620 observations were included in our database (Table 1; Appendix S2). To address whether the effects of multiple global change drivers on plant C pools might vary at different hierarchical scales, we categorised these data into two groups: single-plant level vs. plant-community level. The single-plant level includes responses from whole individual plants or compartments (e.g. shoots, roots, leaves) within individual plants, whereas the plant-community level includes C pools responses from the whole-plant community. For assessing the influence of moderator variables (biotic and abiotic explanatory variables in meta-analysis), the data were further categorised into different subgroups according to ecosystem type, plant functional type (PFT), treatment magnitude, experimental facility and fertiliser chemical form. Continuous moderator variables such as latitude, longitude, mean annual temperature (MAT), mean annual precipitation (MAP), experimental duration and soil depth were also included in our database. When the data from selected primary studies were presented graphically, the figures were digitised to extract the numerical values using the free software Engauge Digitizer (Free Software Foundation, Inc., Boston, MA, USA). In addition, environmental factors
Letter
Global change and terrestrial carbon storage 665
Table 1 Number of observations associated with each global change driver and with two-driver pairs included in the meta-analysis
Carbon pool Plant (single-plant level)
Plant (community level)
Soil
Microbial biomass
Global change driver CO2 W N P R D CO2 W N P R D CO2 W N P R D CO2 W N P R D
CO2
W
N
371
55 233
47 18 408
114
21 121
96
44
P
R
D
0 0 80 92
0 2 6 0 15
6 9 5 0
18 16 350
1 0 179 143
5 5 18 2 36
6 82
24 12 327
0 0 62 42
0 9 2 0 21
6 59
6 7 201
0 0 26 44
0 6 2 1 17
50 1 3 1 0 27 3 4 2 0 30 3 6 0 0 12
CO2: elevated carbon dioxide; W: warming; N: nitrogen addition; P: phosphorus addition; R: increased rainfall; D: drought.
included in our database, such as MAT, MAP and latitude, were obtained directly from the primary studies or extracted from the WorldClim database (http//:www.worldclim.org) using the location information in case that these data were not reported.
DATA ANALYSIS
Individual and combined effects
The individual effect of a global change driver or the combined effect of a two-driver pair was defined as the response of a specific variable (e.g. soil C pool) in the treatment compared with the control (Crain et al. 2008), which is described by the natural logarithm of response ratio (lnRR) in this study (Hedges et al. 1999). The individual lnRR for each observation was calculated using eqn 1: Xt ð1Þ lnRR ¼ ln Xc where Xt is the experimental treatment mean and Xc is the control mean. The variance (m1) and weight (w1) associated with each lnRR value and the weighted mean lnRR (lnRR++) were also calculated, which were described in detail in Appendix S1. The individual or combined effect was significant if the bias-corrected 95% bootstrap-confidence interval (CI) of lnRR++ did not overlap with zero based on 999 iterations (Rosenberg et al. 2000).
The influence of moderator variables on the magnitude and direction of the responses of terrestrial C pools to global change drivers were evaluated according to available data. The heterogeneity within (Qw) and between (Qb) moderator levels was compared using mixed models to assess the significance of each categorical moderator (Borenstein et al. 2009). The relationships between continuous moderator variables and lnRR were assessed by conducting meta-regressions. The net response to the individual or combined effects was calculated as the mean percentage change in C pool compared with the control (%) using the equation ðelnRRþþ 1Þ 100%, and the effects were considered insignificant at P < 0.05 if the 95% CI overlapped with zero. The mean response ratio (lnRR++) and its related 95% CI were calculated using the mixed model included in the meta-analytical software MetaWin 2.1 (Rosenberg et al. 2000). In addition, we evaluated the publication bias in the overall database for each variable included in this analysis with funnel plots, which are scatter plots of the effect sizes vs. the sample sizes of individual studies, using MetaWin 2.1 (Rosenberg et al. 2000). The funnel plots for all variables were symmetrical, which suggested absence of publication bias. Interactive effects
To further understand the interaction between individual drivers for a two-driver pair, we calculated interaction effect size for each observation using Hedges’ d, which is an estimate of the standardised mean difference not biased by small sample sizes (Gurevitch & Hedges 2001). The interaction effect size (dI) between drivers A and B was calculated by eqn 2 according to previous studies (Gurevitch et al. 1992; Crain et al. 2008; Zhou et al. 2016a). ðXAB XA Þ ðXB XC Þ JðmÞ ð2Þ 2s where XC, XA, XB and XAB are means of a variable in the control and treatment groups of A, B and their combination (A + B), respectively; s and J(m) are the pooled standard deviation and correction term for small sample bias, respectively, which were calculated by eqns 3 and 4, respectively. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðnc 1Þs2c þ ðnA 1Þs2A þ ðnB 1Þs2B þ ðnAB 1Þs2AB s¼ nc þ nA þ nB þ nAB 4 dI ¼
ð3Þ 3 ð4Þ JðmÞ ¼ 1 4m 1 where nc, nA, nB and nAB are the sample sizes, and sc, sA, sB and sAB are the standard deviations in the control and experimental groups of A, B and their combination (A + B), respectively; m is the degree of freedom (m = nc + nA + nB + nAB 4). The variance of dI (m2) was estimated by eqn 5, 1 1 1 1 1 d2I ð5Þ þ þ þ þ v2 ¼ 4 nc nA nB nAB 2ðnc þ nA þ nB þ nAB Þ The weighted mean dI (d++) was calculated according to eqn 6: © 2017 John Wiley & Sons Ltd/CNRS
666 K. Yue et al.
Pl
Letter
Pk
i¼1
dþþ ¼ Pl
j¼1
Pk
i¼1
wij dij
j¼1
wij
ð6Þ
where l is the number of groups, k is the number of comparisons in the ith group and w is weight, which is also the reciprocal of the variances (1/m2). The 95% CI of d++ was calculated as d++ Ca/2 9 s(d++), where Ca/2 is the twotailed critical value of the standard normal distribution. The interactions between two drivers were thus classified into three types, i.e. additive, synergistic and antagonistic (Crain et al. 2008). If the 95% CI overlapped with zero, the interactive effect was considered to be additive. For two-driver pairs whose individual effects were either both negative or have opposite directions, the interactions < 0 were synergistic and > 0 antagonistic. In cases where the individual effects were both positive, the interaction effect sizes > 0 were synergistic and < 0 antagonistic. RESULTS
Individual effects
We found that both eCO2 and warming significantly increased plant C pools at single-plant level by 9.8 and 4.8%, respectively (Fig. 1a). Similarly, N addition, P addition and increased rainfall, each taken individually, increased plant C pools at single-plant level by 20.1, 26.8 and 36.1%, respectively (Fig. 1a). In contrast, drought significantly decreased plant C pools at single-plant level by 5.1% (Fig. 1a). The individual effects of these global change drivers on plant C pools at the plant-community level were similar to their respective effects at single-plant level. However, drought showed insignificant effects on plant C pools at the community level (Fig. 1b). Soil C pools were significantly stimulated by the individual effects of eCO2 (+ 4.6%), N addition (+ 8.2%) and P addition (5.6%), but were significantly reduced by warming (4.3%; Fig. 1c). The individual effects of increased rainfall and drought had insignificant effects on soil C pools (Fig. 1c). The response of MBC pools to these drivers was less sensitive and was only significantly increased by eCO2 (+ 7.3%) and by greater rainfall (+ 11.2%; Fig. 1d). Combined effects
The combined effects of eCO2 + warming and eCO2 + N addition increased plant C pools at the single-plant level by 20.3 and 45.0%, respectively (Fig. 1e), which were similar to their effects on plant C pools at the community level (Fig. 1f). Likewise, plant C pools at both single-plant and plant-community levels were significantly increased by the combined effects of warming + N addition, N addition + P addition and N addition + increased rainfall, but were minimally affected by the combined effects of other two-driver pairs such as eCO2 + drought (Fig. 1e, f). Soil C pools were significantly stimulated by the combined effects of eCO2 + N addition and N addition + P addition, with average increases of 8.3 and 8.0%, respectively, but significantly decreased by the combined effects of warming + drought (11.1%; Fig. 1g). The MBC pool was only significantly stimulated by the combined © 2017 John Wiley & Sons Ltd/CNRS
effects of eCO2 + warming (Fig. 1h). The combined effects of other potential two-driver pairs were not presented because of lack of available data (see Table 1). Interactive effects
Across all two-driver pairs, additive interactions appeared to be much more common compared with synergistic and antagonistic interactions (Fig. 2). With the exception of synergistic effects of eCO2 9 warming and eCO2 9 N addition and antagonistic effects of N addition 9 drought, the interactive effects of other pairs on plant C pools at single-plant level were all additive (Fig. 2a). As to plant C pools at the community level, only the interactive effects of N addition 9 P addition were synergistic (Fig. 2b). Likewise, the interactive effects of two-driver pairs on soil C pools were generally additive except for warming 9 N addition, which was antagonistic (Fig. 2c). The interactive effects of two-driver pairs on MBC were found to be all additive (Fig. 2d). Although overall synergistic and antagonistic effects for some of the two-driver pairs were observed, additive interaction still exhibited a substantial predominance as shown by the frequency distribution of interaction types among individual observations (Fig. 2e– h). The interactive effects of other potential two-driver pairs were not presented because of lack of available data (see Table 1). Influence of moderator variables on individual and combined effects
Both individual and combined effects of the investigated global change drivers on terrestrial C pools were significantly influenced by moderator variables such as ecosystem type, experimental setting factors (e.g. duration, PFT and fertiliser form), experimental locations (i.e. latitude and longitude) and climate (i.e. MAT and MAP) (Appendix S3). For example, the individual effects of eCO2 on plant C pools at single-plant level varied significantly (Qb = 62.9, P < 0.00001) among different types of ecosystems (Appendix S3, Fig. S2a), and the responses of terrestrial C pools to N addition could be significantly influenced by the added fertiliser form (Appendix S3, Fig. S4). The influences of continuous moderator variables on both the individual and combined effects are shown in Appendix S3, Table S3 and Figs S11–S14.
DISCUSSION
Individual effects
Overall we found that each of different global change drivers (i.e. eCO2, warming, N addition, P addition, increased rainfall) had significant positive individual effects on plant C pools at both single-plant level and plant-community level. These findings partly confirm our first hypothesis that different global change drivers will significantly affect plant C pools. However, global change effects on soil C and MBC pools were much weaker and less frequent. Our evidence of significant individual effects of multiple global change drivers on plant C pools is supported by the findings of previous meta-analyses (Luo et al. 2006; Wu et al. 2011; Lu et al.
Letter
Global change and terrestrial carbon storage 667
Com bined effects Plant C (single plant level)
Individual effects Plant C (single plant level)
(a) CO2
CO2 + warming
(371)
Warming
(e)
(55) (47)
CO2 + N
(233)
CO2 + drought
N
(408)
(6) (18)
Warming + N
P
(9)
Warming + drought
(92) (15)
Rainfall
(80)
N+P N + rainfall
Drought
(50)
–10
0
(6)
N + drought 10
20
30
40
50
60
(5)
–75
–50
Plant C (community level) CO2 + warming
(114)
Warming
(f)
(143)
Warming + rainfall
(27)
20
30
40
50
60
(5) (3)
N+P
(179)
N + rainfall
(18)
–75
–50
–25
0
Soil C CO2 + warming
(96)
Warming
(g)
N
(327)
P
(42)
CO2 + drought
Drought
(30)
0
10
30
40
50
60
125
100
125
(62)
–75
–50
–25
0
25
50
Microbial biomass C CO2 + warming
(44) (59) (201)
N
100
(4)
N+P
(d) Warming
75
(12)
Microbial biomass C CO2
125
(9)
Warming + drought
20
100
(3)
Warming + N
(21)
75
(24)
Warming + rainfall Rainfall
50
(6)
CO2 + N
(82)
–10
25
Soil C
(c) CO2
125
(18)
Warming + drought
(36)
Rainfall
100
(21)
(16)
Warming + N
P
10
75
(5)
(350)
0
50
CO2 + rainfall
N
–10
25
CO2 + N
(121)
Drought
0
Plant C (community level)
(b) CO2
–25
(h)
(6)
CO2 + N
(6)
CO2 + drought
(3) (7)
Warming + N (44)
P
(6)
Warming + rainfall (17)
Rainfall
(12)
Drought –10
0
10
20
30
40
Changes in percentage (%)
50
60
Warming + drought
(6)
N+P
(26)
–75
–50
–25
0
25
50
75
Changes in percentage (%)
Figure 1 Individual effects of different global change drivers on (a) plant C pools at the single-plant level, (b) plant C pools at the plant-community level, (c) soil C pools and (d) soil microbial biomass C (MBC) pools, and the combined effects of different global change drivers on (e) plant C pools at the single-plant level, (f) plant C pools at the plant-community level, (g) soil C pools and (h) soil MBC pools. Results are expressed as the percentage change relative to the control (%). Values indicate the means with 95% confident intervals (CIs) and sample size numbers are shown in parentheses. The effects of global change drivers are significant when the 95% CIs does not overlap with zero. Legend: CO2 = elevated CO2; N = nitrogen addition; P = phosphorus addition; Rainfall = increased rainfall.
© 2017 John Wiley & Sons Ltd/CNRS
668 K. Yue et al.
Letter Plant C (single plant level) (e) CO2 × warming
(a)
(29)
CO2 × N
75.9%
10.3%
(47)
CO2 × drought
(6)
Warming × N
83.3%
88.9%
(9)
N×P
88.9%
11.1%
70.0%
(6)
N × drought –1
0
22.0%
8.0%
100%
(5) –2
2.1%
11.1%
(50)
N × rainfall
14.9%
16.7%
(18)
Warming × drought
13.8%
83.0%
1
60.0% 40.0%
2
0
10
20
Hedges' d++
30
40
50
Sample size numbers Plant C (community level) (f)
CO2 × warming
(b)
(16)
CO2 × N
(18)
CO2 × rainfall
(5)
Warming × N
100% 100% 100%
(16)
Warming × rainfall
(5)
N×P
18.8%
81.3% 100%
(136)
N × rainfall
(15) –2
–1
12.5%
73.5%
14.0%
100%
0
1
2
0
50
Hedges' d++
100
150
Sample size numbers Soil C (g)
CO2 × warming
(c)
(4)
CO2 × N
100%
(24)
91.7%
CO2 × drought
(3)
Warming × N
11.1%
88.9% 100%
(4)
N×P
(16) –2
–1
0
87.5%
1
2
0
10
15
20
25
Sample size numbers Microbial biomass C (h)
(d)
(4)
CO2 × N
100%
(6)
100%
CO2 × drought
(3)
Warming × N
(7)
Warming × rainfall
(6)
Warming × drought
100%
100%
(6)
66.7%
(13) –1
0
28.6%
71.4%
N×P –2
12.5%
5
Hedges' d++
CO2 × warming
8.3%
91.7%
(9)
Warming × drought
4.2%
100%
(12)
Warming × rainfall
4.2%
1
30.8%
2
0
Hedges' d++ Antagonistic
33.3% 7.7%
5
61.5%
10
15
Sample size numbers Synergistic
Additive
Figure 2 Interactive effects of multiple global change drivers on (a) plant C pools at the single-plant level, (b) plant C pools at the plant-community level, (c) soil C pools and (d) soil microbial biomass C pools revealed by the weighted mean Hedges’d (Hedges’d++), and the corresponding frequency distribution of interaction types among individual observations (e, f, g, h). Values represent means with 95% confident intervals (CIs) and sample size numbers are shown in parentheses. If the 95% CI overlapped with zero, the interactive effect was considered to be additive, otherwise the interactive effect was synergistic or antagonistic. Because many studies only reported combined effects, sample sizes may be smaller than the corresponding sample sizes in Figure 2. Legend: CO2 = elevated CO2; N = nitrogen addition; P = phosphorus addition; Rainfall = increased rainfall.
© 2017 John Wiley & Sons Ltd/CNRS
Letter
2013; Li et al. 2016; Zhou et al. 2016b). The potential underlying mechanisms responsible for these effects were well synthesised in these meta-analyses and are further discussed in this study (see Appendix S3). In addition to previous studies, our findings reveal that the individual effects of multiple global change drivers on plant C pools are seldom scale dependent as only the effects of eCO2 and rainfall were found to significantly vary between single-plant and plant-community levels (Table S3). Evidence of scale-dependent effects is, for example, provided by the response of photosynthetic rates to eCO2, which are consistently larger at leaf level than at the whole-plant level (de Graaff et al. 2006; Ainsworth & Rogers 2007). Compared with plant C pools, responses of soil C and MBC pools to the individual effects of the same global change drivers were much weaker. We found, for example, that soil C pools significantly increased under eCO2 but decreased under warming. These opposite responses of soil C pools to eCO2 and warming may be attributed to the different effects that these two global change drivers may have on soil C influx and efflux processes (Lu et al. 2013; van Groenigen et al. 2014), which are discussed in detail in Appendix S3. The observed positive N addition effect on soil C pools agrees with the findings of previous meta-analysis studies (Lu et al. 2011). Phosphorus additions showed a similar but weaker effect on soil C pools. Soil MBC pool was only significantly increased by eCO2 and increased rainfall, indicating weaker MBC sensitivity to future global change scenarios, which include the effects of multiple drivers. Combined and interactive effects
Confirming our second hypothesis, we found that the interactive effects of multiple global change drivers were generally additive (i.e. not differing from the sum of their individual effects) rather than synergistic or antagonistic. Understanding the interactive effects of multiple global change drivers on terrestrial C storage is crucial for validating Earth system models and for predicting future C storage responses to global changes. Here, we show how plant C pools were significantly stimulated by the combined effects of eCO2 + warming and eCO2 + N addition at both single-plant and plant-community levels (Fig. 1e, f). Interactive effects of eCO2 9 warming and eCO2 9 N addition on plant C pools at the plant-community level were additive (Fig. 2b, f). At the single-plant level these interactive effects were also predominantly additive among individual observations (Fig. 2e), although the overall interactive effects were synergistic (Fig. 2a). Synergistic responses of plant biomass to the interactive effects of eCO2 9 warming or eCO2 9 N addition at the single-plant level may have similar explanations. For example, eCO2-induced increases in water use efficiency could reduce water limitation caused by warming thereby allowing plant growth to benefit from the combination of eCO2 9 warming (Morgan et al. 2011). In this case the positive warming effect on plant growth would be larger in the combined treatment than in the individual warming treatment. Similarly, for the eCO2 9 N addition interaction, the eCO2 fertilisation effect on plant growth may be weaker when plant growth is limited by nutrient availability. The
Global change and terrestrial carbon storage 669
simultaneous N addition allows the full eCO2 fertilisation effect to be expressed resulting in a synergistic interaction (Reich et al. 2014). However, additive interactions of eCO2 9 warming or eCO2 9 N addition on plant C pools remained predominant (75.9 and 83.0%, respectively) when compared with synergistic and antagonistic effects (Fig. 2e). Overall, our findings suggest that additive interactive effects of eCO2 9 warming and eCO2 9 N addition on plant C pools are very common and that such effects are consistent across different hierarchical scales (from single-plant to plantcommunity levels). Nevertheless, a recent quantitative review shows evidence of non-additive effects of eCO2 and warming on plant C pools (Dieleman et al. 2012). We further discuss in Appendix S3 what factors (e.g. difference in methodological approach) could explain the discrepancy between our results and those from Dieleman et al.(2012). Plant C pools were significantly stimulated by the combined effects of warming + N addition, N addition + P addition and N addition + increased rainfall either at the single-plant level or at the plant-community level and these interactions were mainly additive. The positive effect of N addition + P addition on plant C pools could be explained by the fact that N addition stimulates plant biomass accumulation, which in turn increases P demand potentially leading to P limitation for plant growth (Li et al. 2016). Thus, additional P inputs will strengthen the individual effects of N addition resulting in a synergistic interaction between N and P additions. Although the combined effects of N addition + drought showed insignificant effects on plant C pools at the single-plant level (Fig. 1e), the interaction between these two drivers were found to be antagonistic (Fig. 2a). It could be that additional N inputs cannot counteract the significantly negative effects that drought has on plant growth, however, low statistical power may have limited the strength of this analysis (Loladze 2014). Despite our evidence of synergistic and antagonistic interactive effects of N addition 9 P addition and N addition 9 drought on plant C pools, additive interactions of two-driver pairs appeared to be consistently predominant among individual observations regardless of plant C pool scales (Fig. 2e, f). Soil C pools were only significantly stimulated by the combined effects of eCO2 + N addition and N addition + P addition and decreased by warming + drought, whereas two-driver pair interactions were all found to be additive except for an antagonistic interaction of warming 9 N addition (Fig. 2c). Previous studies found that eCO2 only caused accumulation of soil C when additional N was added at rates well above typical atmospheric N inputs (van Groenigen et al. 2006). The eCO2-induced increase in plant growth usually contributes to reduce soil N availability especially in the long term and in the absence of exogenous N additions (Finzi et al. 2002). Therefore, N fertilisation under eCO2 can sustain increases in plant growth and thus maintain C inputs to soils (Oren et al. 2001). In our study we found positive combined effects of eCO2 + N addition on plant C pools at the community level (Fig. 1f). Likewise, the significant increase in soil C pools stimulated by the combined effects of N addition + P addition may be attributed to increases in plant C pools (Fig. 1f). Among interactive effects on soil C pools, only the interaction © 2017 John Wiley & Sons Ltd/CNRS
670 K. Yue et al.
of warming 9 N addition was antagonistic rather than additive (Fig. 2c), but additive interactions were still dominant (91.7%, Fig. 2g) among individual observations. The interactive effects of all available two-driver pairs on soil MBC pools were found to be additive, and only the combined effects of eCO2 + warming had significant (and positive) effects on MBC, which may be attributed to the positive response that microbes usually show to eCO2 and warming (Carney et al. 2007; Lu et al. 2013). Our results further indicate that soil C and MBC pools are significantly less sensitive than plant C pools to both individual and combined effects of multiple global change drivers. Because the pool size of soil C is approximately four times larger than plant C pool (Nieder & Benbi 2008), it could be that large soil C pools are more resilient to global change when compared with smaller plant C pools (Batjes 2014). It is also difficult to detect changes in soil C pools under short-term global change experimental manipulations, despite changes in soil C pool could be biogeochemically significant (Hungate et al. 2009). Influences of moderator variables
According to our third hypothesis, responses of terrestrial C pools to individual effects of multiple global change drivers were mediated by moderator variables. These included plant compartments (e.g. aboveground and belowground), ecosystem type, environmental factors (e.g. latitude, longitude, MAT and MAP) and experimental settings (e.g. CO2 enrichment facility, warming method, N or P addition form, treatment magnitude, experimental duration, etc.; see Appendix S3). Potential mechanisms associated with the net effect of different moderator variables have been discussed in previous synthesis studies (Lu et al. 2013; Li et al. 2016; Zhou et al. 2016b). In our study, for example, N addition effects on plant C pools at either single-plant or plant-community levels were most manifest in grassland ecosystems (P < 0.0001), perhaps due to high plant responses to the addition of urea vs. other N chemical forms added to grassland soils (Appendix S3, Fig. S4a, b). We found that moderator variables such as ecosystem type, PFT, climate and experimental duration can also significantly influence the combined effects of multiple two-driver pairs on terrestrial ecosystem C pools (Appendix S3). For instance, changes in the duration of the experiment are usually crucial for assessing ecosystem response to global change drivers (Isbell et al. 2013). The combined effects of eCO2 + N addition on plant C pools at the community level showed a significant positive correlation with experimental duration (Appendix S3, Fig. S12d), whereby plant C pools increased with longer-term studies. However, a significant (P = 0.015) negative correlation was found between N addition + P addition and experimental duration, probably because N addition tends to stimulate plant growth initially (H€ oegberg et al. 2006) rather than in the long term. Uncertainty analysis and improvement needed
Our meta-analysis shows how interactive effects of multiple global change drivers might influence terrestrial C pools, © 2017 John Wiley & Sons Ltd/CNRS
Letter
however, uncertainties still remain due to inherent limitations in experimental and statistical methodologies and the lack of large, complete datasets. First, the observation-weighted approach that we used here may potentially overestimate the amount of additive interactions associated with the large variance of some observations (Crain et al. 2008; Zhou et al. 2016a). Nevertheless, our statistical analyses suggest that this may not be the case for the large dataset used in this study. We found that average weights of the interaction (dI) for significant (antagonistic and synergistic) results were 4.05, 4.28, 3.77 and 4.24 for single-plant C pools, plant-community C pools, soil C and MBC pools, respectively. Because these weights’ values were not significantly different from those for the insignificant (additive) results (i.e. 4.21, 5.56, 4.03 and 5.88, respectively), we suggest that the overestimation of additive interactions may not be an issue in this study (Zhou et al. 2016a). Second, some of the sample sizes for available twodriver pairs were very small thus may conceal potential synergistic or antagonistic effects. Here, our results indicated that additive interactions were more frequent among all the available individual observations (Fig. 2e–h). Third, the compiled database results were mainly from studies in Europe, North American and China (see Appendix 2, Fig. S1), thus the lack of a sufficient number of studies from other critical regions reduces our ability to integrate responses of terrestrial C pools globally. Learning from our meta-analysis study we suggest that: (1) well-designed experiments including multi-drivers are urgently needed to better capture the response of terrestrial C pools to future global change scenarios, and (2) a more ecologically relevant metric needs to be developed to better incorporate and compare the interactive effects of multiple drivers’ treatments. More experimental manipulation studies are also needed from other critical regions including Africa and South America to get a better global-scale perspective. CONCLUSIONS
Overall our study shows that the interactive effects of two global change drivers across multiple two-driver pairs are mainly additive. This suggests that additive effects of multiple global change drivers on terrestrial ecosystem C pools may be more common than synergistic or antagonistic effects. Our results also show that (1) eCO2, warming, N addition, P addition and increased rainfall, all exerted positive individual effects on plant C pools at both the single-plant and plant-community levels; (2) the magnitude of these effects on plant C pools was seldom scale dependent (i.e. did not vary significantly between single-plant and plant-community levels); (3) soil C and soil MBC pools are much less sensitive to global change than plant C pools; (4) plant C pool responses to the combined effects of multiple two-driver pairs are also seldom scale dependent, whereas soil C and MBC pools still show weaker responses to combined effects of multiple global change drivers and (5) ecosystem type, plant compartment, environmental and experimental factors are all important moderator variables mediating the responses of terrestrial C storage to both individual and combined effects of multiple global change drivers. Our study provides new insights into the development of improved terrestrial C cycle models and the
Letter
design of manipulative experiments across world regions, which could capture the complex response of terrestrial ecosystems to global change. ACKNOWLEDGEMENTS
We thank all the scientists whose data and work were included in this meta-analysis. We are also grateful to three anonymous reviewers and the corresponding editor for their insightful comments and very helpful suggestions. This research was financially supported by the National Natural Science Foundation of China (31622018, 31670526 and 31570445), the joint Ph.D. programme grant from China Scholarship Council (201506910002) and Natural Sciences and Engineering Research Council of Canada (NSERC) Discover Grant. AUTHORSHIP
KY and FW planned the research; KY and YP collected raw data and KY analysed output data and performed all statistical analyses; KY and DAF wrote the first draft of the manuscript and all authors contributed to revisions. REFERENCES Ainsworth, E.A. & Rogers, A. (2007). The response of photosynthesis and stomatal conductance to rising [CO2]: mechanisms and environmental interactions. Plant Cell Environ., 30, 258–270. Batjes, N. (2014). Total carbon and nitrogen in the soils of the world. Eur. J. Soil Sci., 65, 10–21. Borenstein, M., Hedges, L.V., Higgins, J.P. & Rothstein, H.R. (2009). Introduction to Meta-Analysis. John Wiley & Sons Ltd, Chichester. Carney, K.M., Hungate, B.A., Drake, B.G. & Megonigal, J.P. (2007). Altered soil microbial community at elevated CO2 leads to loss of soil carbon. Proc. Natl Acad. Sci. USA, 104, 4990–4995. Crain, C.M., Kroeker, K. & Halpern, B.S. (2008). Interactive and cumulative effects of multiple human stressors in marine systems. Ecol. Lett., 11, 1304–1315. Dieleman, W.I., Vicca, S., Dijkstra, F.A., Hagedorn, F., Hovenden, M.J., Larsen, K.S. et al. (2012). Simple additive effects are rare: a quantitative review of plant biomass and soil process responses to combined manipulations of CO2 and temperature. Glob. Change Biol., 18, 2681–2693. Dillon, M.E., Wang, G. & Huey, R.B. (2010). Global metabolic impacts of recent climate warming. Nature, 467, 704–706. Dukes, J.S., Chiariello, N.R., Cleland, E.E., Moore, L.A., Shaw, M.R., Thayer, S. et al. (2005). Responses of grassland production to single and multiple global environmental changes. PLoS Biol., 3, e319. Ferreira, V., Castagneyrol, B., Koricheva, J., Gulis, V., Chauvet, E. & Gracßa, M.A. (2015). A meta-analysis of the effects of nutrient enrichment on litter decomposition in streams. Biol. Rev., 90, 669–688. Finzi, A.C., DeLucia, E.H., Hamilton, J.G., Richter, D.D. & Schlesinger, W.H. (2002). The nitrogen budget of a pine forest under free air CO2 enrichment. Oecologia, 132, 567–578. Frank, D., Reichstein, M., Bahn, M., Thonicke, K., Frank, D., Mahecha, M.D. et al. (2015). Effects of climate extremes on the terrestrial carbon cycle: concepts, processes and potential future impacts. Glob. Change Biol., 21, 2861–2880. de Graaff, M.A., Van Groenigen, K.-J., Six, J., Hungate, B. & van Kessel, C. (2006). Interactions between plant growth and soil nutrient cycling under elevated CO2: a meta-analysis. Glob. Change Biol., 12, 2077–2091.
Global change and terrestrial carbon storage 671
van Groenigen, K.-J., Six, J., Hungate, B.A., de Graaff, M.-A., Van Breemen, N. & Van Kessel, C. (2006). Element interactions limit soil carbon storage. Proc. Natl Acad. Sci. USA, 103, 6571–6574. van Groenigen, K.J., Qi, X., Osenberg, C.W., Luo, Y. & Hungate, B.A. (2014). Faster decomposition under increased atmospheric CO2 limits soil carbon storage. Science, 344, 508–509. Gurevitch, J. & Hedges, L.V. (2001). Meta-analysis: combining the results of independent experiments. In: Design and Analysis of Ecological Experiments (eds Scheiner, S. & Gurevitch, J.). Oxford University Press, New York, NY, pp. 347–369. Gurevitch, J., Morrow, L.L., Wallace, A. & Walsh, J.S. (1992). A metaanalysis of competition in field experiments. Am. Nat., 140, 539–572. Hedges, L.V., Gurevitch, J. & Curtis, P.S. (1999). The meta-analysis of response ratios in experimental ecology. Ecology, 80, 1150–1156. H€ oegberg, P., Fan, H., Quist, M., Binkley, D. & Tamm, C.O. (2006). Tree growth and soil acidification in response to 30 years of experimental nitrogen loading on boreal forest. Glob. Change Biol., 12, 489–499. Hungate, B.A., van Groenigen, K.J., Six, J., Jastrow, J.D., Luo, Y., de Graaff, M.A. et al. (2009). Assessing the effect of elevated carbon dioxide on soil carbon: a comparison of four meta-analyses. Glob. Change Biol., 15, 2020–2034. IPCC (2014). Climate change 2014: synthesis report. In: Contribution of Working Groups I, II, and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (eds Pachauri, R.K. & Meyer, L.A.). IPCC, Geneva, Switzerland, pp. 1–151. Isbell, F., Reich, P.B., Tilman, D., Hobbie, S.E., Polasky, S. & Binder, S. (2013). Nutrient enrichment, biodiversity loss, and consequent declines in ecosystem productivity. Proc. Natl Acad. Sci. USA, 110, 11911– 11916. Leuzinger, S., Luo, Y., Beier, C., Dieleman, W., Vicca, S. & K€ orner, C. (2011). Do global change experiments overestimate impacts on terrestrial ecosystems? Trends Ecol. Evol., 26, 236–241. Li, Y., Niu, S. & Yu, G. (2016). Aggravated phosphorus limitation on biomass production under increasing nitrogen loading: a meta-analysis. Glob. Change Biol., 22, 934–943. Liu, L. & Greaver, T.L. (2010). A global perspective on belowground carbon dynamics under nitrogen enrichment. Ecol. Lett., 13, 819–828. Loladze, I. (2014). Hidden shift of the ionome of plants exposed to elevated CO2 depletes minerals at the base of human nutrition. Elife, 3, e02245. Lu, M., Zhou, X., Luo, Y., Yang, Y., Fang, C., Chen, J. et al. (2011). Minor stimulation of soil carbon storage by nitrogen addition: a metaanalysis. Agr. Ecosyst. Environ., 140, 234–244. Lu, M., Zhou, X., Yang, Q., Li, H., Luo, Y., Fang, C. et al. (2013). Responses of ecosystem carbon cycle to experimental warming: a metaanalysis. Ecology, 94, 726–738. Luo, Y., Hui, D. & Zhang, D. (2006). Elevated CO2 stimulates net accumulations of carbon and nitrogen in land ecosystems: a metaanalysis. Ecology, 87, 53–63. Morgan, J.A., LeCain, D.R., Pendall, E., Blumenthal, D.M., Kimball, B.A., Carrillo, Y. et al. (2011). C4 grasses prosper as carbon dioxide eliminates desiccation in warmed semi-arid grassland. Nature, 476, 202–205. Mueller, K., Blumenthal, D.M., Pendall, E., Carrillo, Y., Dijkstra, F., Williams, D. et al. (2016). Impacts of warming and elevated CO2 on a semi-arid grassland are non-additive, shift with precipitation, and reverse over time. Ecol. Lett., 19, 956–966. Nieder, R. & Benbi, D.K. (2008). Carbon and Nitrogen in the Terrestrial Environment. Springer Science & Business Media, New York. Norby, R.J. & Luo, Y. (2004). Evaluating ecosystem responses to rising atmospheric CO2 and global warming in a multi-factor world. New Phytol., 162, 281–293. Norby, R.J. & Zak, D.R. (2011). Ecological lessons from free-air CO2 enrichment (FACE) experiments. Annu. Rev. Ecol. Evol. Syst., 42, 181. Oren, R., Ellsworth, D.S., Johnsen, K.H., Phillips, N., Ewers, B.E., Maier, C. et al. (2001). Soil fertility limits carbon sequestration by
© 2017 John Wiley & Sons Ltd/CNRS
672 K. Yue et al.
forest ecosystems in a CO2-enriched atmosphere. Nature, 411, 469– 472. Reich, P.B., Hobbie, S.E., Lee, T., Ellsworth, D.S., West, J.B., Tilman, D. et al. (2006). Nitrogen limitation constrains sustainability of ecosystem response to CO2. Nature, 440, 922–925. Reich, P.B., Hobbie, S.E. & Lee, T.D. (2014). Plant growth enhancement by elevated CO2 eliminated by joint water and nitrogen limitation. Nat. Geosci., 7, 920–924. Rosenberg, M.S., Adams, D.C. & Gurevitch, J. (2000). MetaWin: Statistical Software for Meta-Analysis, Version 2. Sinauer Associates, Sunderland, MA. Treseder, K.K. (2008). Nitrogen additions and microbial biomass: a metaanalysis of ecosystem studies. Ecol. Lett., 11, 1111–1120. Vil a, M., Espinar, J.L., Hejda, M., Hulme, P.E., Jarosık, V., Maron, J.L. et al. (2011). Ecological impacts of invasive alien plants: a metaanalysis of their effects on species, communities and ecosystems. Ecol. Lett., 14, 702–708. Wu, Z., Dijkstra, P., Koch, G.W., Pe~ nuelas, J. & Hungate, B.A. (2011). Responses of terrestrial ecosystems to temperature and precipitation change: a meta-analysis of experimental manipulation. Glob. Change Biol., 17, 927–942. Yuan, Z. & Chen, H.Y. (2015). Decoupling of nitrogen and phosphorus in terrestrial plants associated with global changes. Nat. Clim. Chang., 5, 465–469.
© 2017 John Wiley & Sons Ltd/CNRS
Letter
Zhou, L., Zhou, X., Shao, J., Nie, Y., He, Y., Jiang, L. et al. (2016a). Interactive effects of global change factors on soil respiration and its components: a meta-analysis. Glob. Change Biol., 22, 3157–3169. Zhou, X., Zhou, L., Nie, Y., Fu, Y., Du, Z., Shao, J. et al. (2016b). Similar responses of soil carbon storage to drought and irrigation in terrestrial ecosystems but with contrasting mechanisms: a meta-analysis. Agr. Ecosyst. Environ., 228, 70–81.
SUPPORTING INFORMATION
Additional Supporting Information may be found online in the supporting information tab for this article.
Editor, Shuli Niu Manuscript received 22 November 2016 First decision made 8 January 2017 Second decision made 24 February 2017 Manuscript accepted 6 March 2017
