Are there links between responses of soil microbes and ecosystem functioning to elevated CO2, N deposition and warming? A global perspective
Abstract
In recent years, there has been an increase in research to understand how global changes’ impacts on soil biota translate into altered ecosystem functioning. However, results vary between global change effects, soil taxa, and ecosystem processes studied, and a synthesis of relationships is lacking. Therefore, here we initiate such a synthesis to assess whether the effect size of global change drivers (elevated CO2, N deposition, and warming) on soil microbial abundance is related with the effect size of these drivers on ecosystem functioning (plant biomass, soil C cycle, and soil N cycle) using meta-analysis and structural equation modeling. For N deposition and warming, the global change effect size on soil microbes was positively associated with the global change effect size on ecosystem functioning, and these relationships were consistent across taxa and ecosystem processes. However, for elevated CO2, such links were more taxon and ecosystem process specific. For example, fungal abundance responses to elevated CO2 were positively correlated with those of plant biomass but negatively with those of the N cycle. Our results go beyond previous assessments of the sensitivity of soil microbes and ecosystem processes to global change, and demonstrate the existence of general links between the responses of soil microbial abundance and ecosystem functioning. Further we identify critical areas for future research, specifically altered precipitation, soil fauna, soil community composition, and litter decomposition, that are need to better quantify the ecosystem consequences of global change impacts on soil biodiversity.
Introduction
Elevated atmospheric carbon dioxide (CO2), nitrogen (N) deposition, and climate change (e.g., elevated temperatures and altered precipitation regimes) are among the major drivers of ongoing global change for terrestrial ecosystems worldwide (IPCC 2007). Increasing concern about the impacts of these drivers has boosted research on ecosystem processes such as plant productivity and global biogeochemical cycles (MEA, 2005). Although much of the ecosystem research has been devoted to understanding the role of plants as agents of ecosystem functioning (EF) responses to global change (Zavaleta et al., 2003; Reich et al., 2004; Cardinale et al., 2012; Hooper et al., 2012), in recent years there has been an increase in the number of studies focusing on soil communities (Allison & Martiny, 2008; Bardgett & Wardle, 2010). Like plants, soil biota are sensitive to global change, and the alterations in their abundance or diversity from climate change, N deposition, and elevated CO2 can feedback to affect the ecosystem processes they govern (van der Heijden et al., 2008; Bardgett & Wardle, 2010). Thus, a remaining challenge is to assess whether the effects of particular global change drivers on belowground communities are associated with the responses of multiple ecosystem processes.
Soil biota are structured in complex, highly diverse communities and their responses to global change include abundance, compositional, and physiological shifts (Eisenhauer et al., 2012; Wall et al., 2012; Frey et al., 2013). However, abundance measurements are most consistently used across taxa and studies (Treseder, 2004, 2008; Blankinship et al., 2011) and are simpler to standardize. Recent literature syntheses have shown unique and predictable effects of different global change drivers on soil biota abundance. For example, N deposition has been found to decrease microbial biomass (Treseder, 2008), whereas elevated CO2 has been found to increase it (Blankinship et al., 2011). Furthermore, warming often has positive effects on nematode abundance (Blankinship et al., 2011) but negative effects on total numbers of enchytraeids (Briones et al., 2007). Reduced precipitation has been found to decrease the abundance of fungi, enchytraeids and collembola (Blankinship et al., 2011). Whether these global change effects on belowground organisms’ abundance help to explain those on ecosystem functioning is still unclear, although many studies have found concurrent effects of global change on the abundance of particular soil taxa and the rates of magnitude of specific ecosystem processes (Allen et al., 2000; Allison & Treseder, 2008; Lamb et al., 2011).
Previous literature screenings have synthesized how the effect size of N deposition on the abundance of particular soil taxa (total microbial community) relates with the effect size of N deposition on carbon (C) cycle variables such as soil CO2 efflux (Treseder, 2008) or mineral soil C (Liu & Greaver, 2010). However, a synthesis of relationships between global change effect sizes across several soil taxa and ecosystem processes as we present here has been lacking. The evaluation of such linkages, and the identification of knowledge gaps, is fundamental to integrate soil organisms’ control of ecosystem responses to global change into large-scale models (Wall et al., 2012), guide future research efforts and mitigation plans (Jeffery et al., 2010), and support emerging soil policy initiatives (Koch et al., 2013).
Here we compiled data from individual global change studies, and used meta-analytical tools and structural equation modeling (SEM) to evaluate whether the abundance responses of soil microbes to a particular global change driver explain variation in the EF responses to this driver. We argue that our SEM, which is a regression-based modeling technique and thus not causal in its essence, allows causal interpretation of these linkages based on prior ecological knowledge. We focused on the most commonly studied drivers (elevated CO2, N deposition, and warming), soil taxa (total microbial community, fungi, and bacteria) and ecosystem processes (plant biomass, soil C, and N cycle). Different metrics were used to measure the abundance of microbial taxa (e.g., PLFA biomass and chloroform-fumigation extractions for total microbial community; Table 1) and the ecosystem processes (e.g., NO3−-N, total N or N mineralization rate for N cycle; Table 1). Specifically, this study aimed to: (i) quantify the effect size of elevated CO2, N deposition, and warming on total microbial community, bacterial and fungal abundance, and on plant biomass, soil C cycle and soil N cycle, (ii) assess whether the effect size of elevated CO2, N deposition, and warming on soil microbes (across taxa) is related with the effect size of such drivers on EF (across processes), as well as evaluate the relative contribution of study length, magnitude of global change and climatic conditions, and (iii) analyze the correlations between the effect size of each global change driver on each microbial taxon and ecosystem process separately. Finally, the scope of our literature synthesis was drastically constrained by the infrequency of studies that examined both soil biota and EF responses to a global change driver, so we are using this study as an opportunity to identify knowledge gaps and discuss future research directions that will advance our understanding of how soil biota control ecosystem responses to global change.
| Target variable | Metric | n (Elevated CO2) | n (N deposition) | n (Warming) |
|---|---|---|---|---|
| Bacteria | Bacterial biomass | 1 | 2 | 1 |
| Bacteria | PLFA | 6 | 3 | 1 |
| Bacteria | MPN | 4 | 2 | 1 |
| Bacteria | DNA fingerprint | 2 | NA | NA |
| Bacteria | qPCR | 1 | NA | 5 |
| Bacteria | Biolog | NA | NA | 2 |
| Fungi | Fungal biomass | NA | 2 | 1 |
| Fungi | PLFA | 5 | 3 | 1 |
| Fungi | MPN | 3 | 1 | 1 |
| Fungi | Ergosterol | 3 | NA | 3 |
| Fungi | Root colonization | 10 | 6 | 1 |
| Fungi | qPCR | NA | NA | 4 |
| MC | PLFA | 1 | 2 | NA |
| MC | MB (soil weight) | 17 | 27 | 21 |
| MC | MB (soil surface) | 2 | 5 | NA |
| Plant biomass | Shoot biomass | 4 | 8 | NA |
| Plant biomass | Root biomass | 12 | 4 | 8 |
| Plant biomass | Total biomass | 12 | NA | 2 |
| C cycle | Soil respiration | 5 | 6 | 9 |
| C cycle | TOC | 12 | 19 | 13 |
| C cycle | DOC | 4 | 4 | 3 |
| N cycle | DIN | 8 | 7 | 12 |
| N cycle | DON | NA | 1 | NA |
| N cycle | N flux | 12 | 5 | 7 |
| N cycle | NH4 + -N | NA | NA | 6 |
| N cycle | NO3‒-N | 1 | NA | NA |
| N cycle | Total N | 2 | 7 | NA |
- n represents the sample size in terms of case studies.
- MC, total microbial community; MPA, most probable number counts; MB, microbial biomass; TOC, total organic carbon; DOC, dissolved organic carbon; DIN, dissolved inorganic nitrogen (NH4‒N + NO3‒N); DON, dissolved organic nitrogen; N flux (N mineralization, nitrification or ammonification rates), NA (not available).
Materials and methods
Data collection
We quantitatively synthesized studies that evaluated the effects of global change on the abundance or biomass of at least one soil taxon and one ecosystem process. Briefly, ‘global change’ was regarded as elevated CO2, warming, N deposition, and/or altered precipitation; ‘soil biota’ was designated to include fungi, bacteria, total microbial community, and numerous other taxa of soil fauna (Fig. 1); and ‘EF’ included several processes (N, C and P cycles, plant productivity and litter decomposition). Any global change study addressing only soil biota or only EF was omitted. Searches were conducted using the ISI Web of Knowledge (http://apps.isiknowledge.com) on 26 April 2013, with no restriction on publication year, and were supplemented with references from previous reviews on the topic. See Appendix S1 for details on the term combinations used in the literature search, which yielded 8662 references.
To be included in our database, studies had to utilize global change rates no more extreme than those predicted under future scenarios, because they have the potential to be more informative to land managers and for parameterization of ecosystem models (Marshall et al., 2008). Maximum warming and CO2 levels were established based on estimates from the A1B Scenario of the Fourth IPCC Assessment (IPCC 2007). Elevated CO2 and warming studies were included if they provided less than 850 ppm of CO2 and increased air temperature up to 2.8 °C. N deposition studies were included if they provided less than 150 kg N ha−1 yr−1 (Dentener et al., 2006) from inorganic N sources (e.g., NH4NO3) or urea, which are the two most studied N forms in global change research and affect microbial activity in the same way (Ramirez et al., 2010). As predictions of changes in precipitation levels are not very accurate due to considerable regional variation (IPCC 2007), studies were only included if treatments represented a maximum of 50% increase or decrease in precipitation with respect to the control, as has been done in previous climate change experiments (Zavaleta et al., 2003). Other study selection criteria are described in Appendix S1. Only field studies were selected for warming and N deposition experiments, but controlled environments such as growth chambers and/or greenhouses were also included for CO2 studies, because they represent an important proportion of the conducted body of research (Blankinship et al., 2011).
Data extraction
A total of 75 articles, representing 330 cases studies, met the established criteria. However, not all topics were evenly represented, and therefore underrepresented global change drivers, soil taxa, and ecosystem processes were subsequently omitted from all downstream analyses. Thus, we focused our analysis on the most studied global change drivers (elevated CO2, warming, and N deposition) and target variables (total microbial community, bacterial and fungal abundance, plant biomass, soil C cycle and soil N cycle; Fig. 1), which rendered a final dataset of 238 case studies from 70 articles (Appendix S2).
For each study, the microbial and EF data were recorded for both the control and global change plots. Mean, standard deviation and sample size values were extracted directly from tables or from graphs using Dexter, an online tool provided by the German Astrophysical Virtual Observatory (http://dc.zah.uni-heidelberg.de/sdexter/). Microbial abundance measurements were always preferred, but biomass measurements were also included as a surrogate for abundance (Coleman et al., 2004), and hence both are referred to as ‘abundance’. Authors used several metrics to assess microbial abundance (Table 1). To represent the abundance of the total microbial community, microbial biomass was measured using chloroform-fumigation extractions (mg C kg−1) or measured as PLFA biomass (Treseder, 2008; Blankinship et al., 2011). Bacterial abundance was determined with incubations using selective inhibitors, bacteria-specific PLFA biomass, most probable number counts, DNA fingerprint (% of total clones from 16S rRNA gene clone libraries), qPCR and Biolog. Fungal abundance was determined with the same metrics, in addition to ergosterol concentration and root colonization. All of the EF metrics used (Table 1) were associated with one of the following processes: plant biomass (as a measure of productivity; Scurlock et al., 2002), soil C and soil N cycles. The metrics used described different aspects of the same process (e.g., root, shoot and total biomass as metrics for plant biomass, or NO3−-N, mineralization rate and total N as metrics for N cycle). All these processes are directly linked to the maintenance of primary production, biomass accumulation and nutrient cycling. According to this rationale, and taking into account the spatial extent of our study, we assumed that the higher the values for the different variables measured at a given study, the higher the overall rate of functioning at that site (Maestre et al., 2012). Methodological features of the experimental design (magnitude of global change, field vs. controlled environment for CO2 studies, study length, latitude and longitude) and the metric used (analytical method and units) were also recorded. We obtained the mean annual temperature and precipitation of each field study site from the WorldClim database (Hijmans et al., 2005), which provides average climatic values for the period 1950–2000.
Meta-analytical procedure for data grouping
Data collected were inherently heterogeneous, as the studies used different metrics to measure the soil taxa and ecosystem processes (Table 1). Thus, a data grouping procedure was performed to deal with such heterogeneity and to test its influence on the global change effect sizes. Two method features were evaluated: (i) environmental conditions (controlled vs. field environments, only for CO2 studies) and (ii) metric used (see Table 1 for the different categories analyzed in each target variable). We ran two separate weighted random-effects meta-analyses (Gurevitch & Hedges, 1999) for each global change driver, one for microbial abundance and one for EF. We calculated Hedge's d as an estimate of global change effect size, and assessed its heterogeneity between the categories of the method feature studied. Positive values indicate that the response variable (the microbial taxa abundances and the ecosystem processes) in the global change plot has a larger value than in the control. To test the effects of global change drivers on each target variable, we assessed whether the bias-corrected 95% bootstrap confidence intervals (CI) from 999 iterations overlapped zero (Rosenberg et al., 2000). We used estimated nonparametric variances because most of our experimental data did not follow a normal distribution (Adams et al., 1997). See Appendix S1 for a full description of the meta-analytical procedure. The results of the random-effects models confirmed that the responses of each microbial taxon and ecosystem process to elevated CO2, N deposition, and warming did not depend on the experimental conditions or metric used (Table S1; Prandom > 0.05 in all cases).
Relationship between the responses of microbes and ecosystem functioning to global change
To assess the relationship between the effect size of global change on microbial abundance and on rates of EF, we used structural equation modeling (Grace, 2006). However, the low number of case studies jointly examining both variables (Table S2) drastically constrained our ability to build robust models for each specific combination of microbial taxon and ecosystem process. Thus, we analyzed such relationship among effect sizes across microbial taxa (total microbial community, fungal and bacterial abundance) and ecosystem processes (plant biomass, soil C and N cycles), an approach that has been previously followed when synthesizing the effects of multiple global change drivers (Blankinship et al., 2011). We tested whether the effect sizes (Hedge's d) of global change on overall microbial abundance modulated the effect sizes on overall EF. The contribution of site-specific factors, such as climatic conditions, study length and magnitude of global change to the effect sizes was also evaluated. We also tested separate models for each of elevated CO2, N deposition, and warming. In addition, we calculated Pearson correlations among the effect sizes of global change on each microbial taxon and ecosystem process separately. The evaluation of such relationships, although not directional, allowed us to interpret whether the two effect sizes were linked at the microbial taxon and ecosystem process level. Correlations were not performed when microbial abundance and EF were measured in fewer than eight case studies. Although we conducted a large number of statistical tests, P values were not adjusted for multiple testing as this approach is considered overly conservative (Gotelli & Ellison, 2004).
On the basis of current ecological knowledge, we hypothesized a hierarchy of relationships in a path diagram (Grace, 2006). Our conceptual a priori global SEM (Fig. 2), which was tested separately for elevated CO2, N deposition, and warming, predicted the effect size of global change on microbial abundance to explain variation in the effect size of global change on EF. This path can be interpreted as the ability of overall microbes (across taxa) to control overall EF (across processes) responses to global change, and is the key aspect of our study. Previous theoretical, modeling and experimental studies have underlined the important role played by microbes controlling ecosystem responses to global change (Allison & Treseder, 2008; Allison et al., 2010aa; Bardgett & Wardle, 2010; Wagg et al., 2014). Thus, the model structure proposed was supported by current ecological knowledge, justifying a causal interpretation of the model outputs (Shipley, 2002). Nevertheless, as all SEM, they are contingent on the structure imposed by the modelers. Therefore, our model only allows causal interpretation of the directional relations introduced (e.g., effect size on microbes explaining variation in effect size on EF), but does not deny the possibility of other type of relations existing within the study system (e.g., effect size on EF explaining effect size on microbes), which were beyond the scope of our synthesis and therefore not tested.
We accounted for the fact that data for global change effect sizes on overall microbes related to different microbial taxa (bacteria, fungi, and microbes), and those effects on overall EF related to different processes (plant biomass, soil C cycle, and soil N cycle). Thus, 2 two-indicator latent variables, ‘Microbial taxon’ and ‘Ecosystem process’, were introduced in the model. Latent variables have multiple uses, but here they function to sum together the effects of the levels of a categorical variable, which are represented by dummy variables (Grace, 2006). When modeling categorical dummy variables, it is necessary to omit one indicator (Grace, 2006), and we omitted the soil taxon or ecosystem process that allowed us to interpret the latent variable in a more straightforward way. A significant individual path coefficient from a category composing ‘Microbial taxon’ (e.g., bacteria) or ‘Ecosystem process’ (e.g., soil C cycle) means a larger effect size of global change for that specific category with respect to the reference.
The a priori model also predicted a direct effect of climate, study length, and magnitude of global change on the effect size of global change (Hedge's d) on both microbes and EF. Study length and magnitude of global change were introduced in the model as exogenous variables. Climate was modeled as a composite variable, which allows an additive combination of the effects of multiple conceptually related variables (mean annual temperature and mean annual precipitation) upon a response variable (the effect sizes of global change). Composite variables are primarily a graphical and numerical interpretation tool, and do not change the underlying model (Grace, 2006). Climatic influence was not included in the elevated CO2 model, because some studies were performed in controlled environments.
We examined the distributions of the endogenous variables and tested their normality. To increase the degrees of freedom, any path with a coefficient <0.10 was removed from the model when not significant. Overall goodness-of-fit of the models was tested against the dataset and checked following Schermelleh-Engel et al. (2003). We used the traditional χ2 goodness-of-fit test, but, because of its sensitivity to sample size, the RMSEA index was also considered (Grace, 2006). SEM analyses were performed with AMOS Software Version 22.0 (Amos Development Co.).
Results
Summary of selected studies and responses to global change drivers
The most studied global change drivers were elevated CO2, N deposition, and warming (95% of the case studies collected, Fig. 1). The most studied soil taxa were fungi, bacteria, and the total microbial community (90% of the case studies collected, Fig. 1). The most studied ecosystem processes were plant biomass, soil C cycle and soil N cycle (86% of the case studies collected, Fig. 1). A positive effect size of elevated CO2 was observed for total microbial abundance and plant biomass (Fig. 3a). N deposition had a positive effect on bacterial abundance and the soil N cycle (Fig. 3b), meaning that N deposition plots showed higher values of the variables describing the N cycle (e.g., NO3−-N, N mineralization rate or total N) than the control plots. Warming increased fungal abundance, and had positive, but nonsignificant, effects on plant biomass, soil C cycle, and soil N cycle (Fig. 3c).
Relationships between the effect sizes of global change on microbes and ecosystem functioning
Goodness-of-fit tests for all SEM evaluated indicated acceptable fits (Figures S1–S3), as the χ2 tests were not significant (P > 0.05 in all models) and the RMSEA fit measure was <0.08 (P > 0.1 in all the models), indicating that the data fitted the a priori model hypothesized for the three global change drivers (Fig. 2).
The effect size of elevated CO2 on overall (across taxa) microbes was not related with the overall (across processes) CO2 effect size on EF (Fig. 4a and Figure S1). The two site-specific factors had contrasting relationships with CO2 effect size on microbes, with a positive influence of the magnitude of the CO2 treatment and a negative one of study length. The two latent variables introduced in the model, ‘Microbial taxon’ and ‘Ecosystem process’, affected both effect sizes, which indicates that the responses to elevated CO2 were different between taxa and processes, as also demonstrated in Fig. 3a. At the microbial taxon and ecosystem process level, fungal abundance responses to elevated CO2 were positively correlated with those of plant biomass but negatively correlated with those of the N cycle (Table 2). The effect size of elevated CO2 on total microbial community abundance and plant biomass varied in the same direction.
| Global change driver | Microbial taxon | Ecosystem process | n | r | P |
|---|---|---|---|---|---|
| CO2 | Fungi | Plant biomass | 15 | 0.540 | 0.038 |
| CO2 | Fungi | N cycle | 9 | −0.855 | 0.003 |
| CO2 | MC | Plant biomass | 14 | 0.562 | 0.037 |
| CO2 | MC | C cycle | 14 | −0.157 | 0.592 |
| CO2 | MC | N cycle | 21 | −0.165 | 0.474 |
| N deposition | MC | C cycle | 17 | 0.824 | <0.001 |
| N deposition | MC | N cycle | 12 | 0.078 | 0.809 |
| Warming | Bacteria | C cycle | 8 | 0.635 | 0.091 |
| Warming | Bacteria | N cycle | 8 | 0.204 | 0.629 |
| Warming | Fungi | C cycle | 11 | 0.476 | 0.139 |
| Warming | Fungi | N cycle | 8 | 0.229 | 0.586 |
| Warming | MC | C cycle | 17 | 0.419 | 0.094 |
| Warming | MC | N cycle | 19 | 0.185 | 0.449 |
- P values below 0.05 are in bold. n represents the sample size in terms of case studies. MC: total microbial community.
The N deposition effect size on overall microbes was significantly related (r = 0.31, P = 0.002) with the N deposition effect size on overall EF (Fig. 4b and Figure S2). Study length significantly affected the effect size on EF but not on microbes. Climatic conditions were associated with the N deposition effect size on microbes, with a higher effect size in the warmer sites. As shown in Fig. 3b, the two effect sizes were different between taxa and processes. Separate correlations showed that total microbial community abundance responses to N deposition were highly related to those of C cycle (Table 2).
The warming effect size on overall microbes accounted for an important part of the variance (r = 0.19, P = 0.041) in the warming effect size on overall EF (Fig. 4c and Figure S3). Although not significantly, the magnitude of warming and the study length were associated with both effect sizes. The higher the contrast between the warmed and control plots, the higher the warming effect size on microbes, but the longer the warming study duration, the lower the warming effect size on EF. The influence of climate on warming effect size on EF was due to both a positive effect of total precipitation and a negative effect of mean temperature. As found in Fig. 3c, the warming effect size on microbes was different between taxa. However, the effect sizes of warming on total microbial community, bacterial, and fungal abundances were not significantly correlated with the effect sizes of this global change driver on soil C and N cycles (Table 2).
Discussion
Soil microbial responses to global change are linked with responses of ecosystem functioning
Here, we present an assessment of the relationships between multiple global change drivers, and the responses of soil microbes and ecosystem processes. While such correlative approaches are not the ultimate test to quantify the ecosystem consequences of global change impacts on soil biodiversity, they can indicate general trends and direct future research efforts. When evaluating the effect size of elevated CO2 on overall responses across microbial taxa and ecosystem processes, the responses of microbes did not explain those of EF. Nevertheless, at the microbial taxon and ecosystem process level, the increases in total microbial community and fungal abundance found with elevated CO2 were positively correlated with the increase in plant biomass. Elevated CO2 effects on belowground organisms are more likely to occur through altered plant root production and exudation than via direct effects of aboveground CO2 (Paterson et al., 1997; Zak et al., 2000), due to the high CO2 concentrations in the soil pore space (Drigo et al., 2008). However, such increase in root biomass can affect soil C and N cycling by altering soil microbial biomass and activity (de Graaff et al., 2006). For example, arbuscular mycorrhizae, which are linked with low rates of N cycling (van der Heijden et al., 2008), are typically stimulated by such root growth increases under elevated CO2 (Treseder, 2004). Since all metrics used to measure the effects of elevated CO2 on the N cycle were related with rapid increases in soil N availability (DIN or N flux rates; Table 1), the negative correlation found between the effect sizes of elevated CO2 on fungal abundance and on soil N cycle supports such plant-mediated mechanism. Thus, increased mycorrhizal biomass as a consequence of root growth with elevated CO2 may decrease soil N transformation rates compared with control plots, which are likely more bacterial-dominated (Fig. 3a). The fact that different microbial taxa may have different effects on EF could explain the absence of an important functional role for soil microbes, when responses to elevated CO2 were evaluated across microbial taxa and ecosystem processes.
Overall responses of microbes to N deposition were explained by the responses of EF to this global change driver. The functional role of microbial abundance was larger than the one played by site-specific factors such as the duration of the study and the magnitude of N enrichment. Total microbial community and fungal, but not bacterial, abundance have been found to decrease as N load and duration of the N deposition treatment increase (Treseder, 2008). To simulate realistic future N deposition rates, we limited our studies to 150 kg N ha−1 (Dentener et al., 2006). This restrictive rate excluded unrealistically high N loads and long-term agricultural studies that may have promoted the pattern found by Treseder (2008). The effects of N deposition on total microbial community abundance were highly correlated with the responses of the soil C cycle to such N enrichment. This is an interesting result because it indicates that N deposition effects on soil C cycling, but not on N cycling, are associated with those of soil microbial biomass. Soil C cycle was positively affected by N deposition, although the confidence intervals barely included zero (Fig. 3b), which indicates an increase in the variables describing such ecosystem process (total organic C in the 65% of the case studies; Table 1). The enhancement of soil organic C with N deposition has also been found in previous meta-analyses (Nave et al., 2009; Liu & Greaver, 2010), and reductions of litter decomposition rates via changes in either plant litter quality (Knorr et al., 2005), microbial communities (Sinsabaugh et al., 2002) or the extent of litter decay (Whittinghill et al., 2012) have been hypothesized as potential mechanisms. Our correlative approach supports the microbial-driven hypothesis. Although plant biomass increases have been identified as one of the main N deposition contributions to the C cycle (LeBauer & Treseder, 2008; Liu & Greaver, 2010), we did not find enough case studies to facilitate addressing the links between any microbial taxon and plant biomass. Thus, we cannot elucidate whether the relationship found between the abundance of the total microbial community and C cycle responses to N deposition will explain ecosystem C sequestration. However, our study does highlight that more research investigating the litter decomposition-microbial mechanism is needed to better understand the N deposition effects on C cycling.
The effect sizes of warming on microbes and EF evaluated across taxa and ecosystem processes, respectively, were also positively and significantly associated, suggesting that microbial abundance and EF respond in parallel to elevated temperatures. We found an interesting matching in the temporal and treatment rate responses to warming, where short-term studies and higher temperature treatments promoted larger positive warming effects on both microbes and EF, which may have facilitated the previous link found. This pattern may be a product of long-term microbial thermal adaptation to elevated temperature (Bradford et al., 2008), but should be interpreted with caution as the relationships were not significant. Warming generally has weak effects at a global scale on net ecosystem C exchange, due to the offset of plant production with C losses (Lu et al., 2013). However, the absence of significant correlations between specific microbial taxa and soil C cycle, and the low number of case studies measuring plant biomass, prevented us from demonstrating whether changes in microbial abundance with warming contribute to the balance between ecosystem C efflux and influx. In general, our elevated CO2, N deposition, and warming results, which are based on a review of field studies, provide empirical support to theoretical and modeling efforts advocating for an explicit inclusion of the microbial component of soils into ecosystem models (Allison & Martiny, 2008; Allison et al., 2010aa; Treseder et al., 2012; Wieder et al., 2013).
Research gaps and guidelines for future ecosystem studies
The ambitious scope of this literature screening allowed us to identify major standardization and research gaps on the linkages between global change, soil communities and EF. The study selection criteria were based on predicted rates for global change drivers available from the IPCC (2007; Appendix S1), which resulted in the exclusion of many altered precipitation (e.g., Van Gestel et al., 1992; Williams & Rice, 2007) and N deposition studies (e.g., Allison et al., 2008; Zheng et al., 2008; Allison et al., 2010bb) due to the magnitude of treatments falling beyond our specified cutoffs. While the importance of standardized research methods to facilitate cross-site comparisons is increasingly recognized (e.g., Wall et al., 2008; Powers et al., 2009; Sylvain et al., 2014), there is little consensus on the magnitude of treatments to simulate responses of ecosystems to global change. If we hope to prediction future ecosystem scenarios, experiments should be designed to account for changes in global change rates predicted over the next 50–100 years. Altered precipitation studies were underrepresented in our assembled database. This constitutes a major gap to understand how soil biota modulates ecosystem responses to global change, because altered precipitation has a larger influence on soil biota abundance across taxa than elevated CO2 or warming, as found by the most recent review on the topic (Blankinship et al., 2011). Soil fauna were also infrequently measured in the available literature, despite their functional role (Bardgett & Chan, 1999; Eisenhauer et al., 2011; Garcia-Palacios et al., 2013) and sensitivity to warming and altered precipitation at global scales (Blankinship et al., 2011), hindering a full assessment of the functional implications of soil biodiversity under global change. Regarding the ecosystem processes measured, the current underrepresentation of studies assessing litter decomposition complicates the understanding of how soil biota mediates global change effects on nutrient dynamics and C cycling.
Here, we evaluated the responses of each ecosystem process (e.g., C cycle) to global change across different metrics (e.g., soil respiration, dissolved organic C or total organic C), as the low number of studies found prevented us from conducting a specific analysis for each metric, and acknowledge that our analysis cannot discriminate among particular outcomes of each process (e.g., soil C losses vs. soil C accumulation). An appropriate procedure to overcome this issue, and scale up from particular ecosystem processes to whole EF, would be the use of multifunctionality indexes, which address the ability to maintain multiple functions simultaneously (Zavaleta et al., 2010). However, current global change research lacks the homogeneity needed to calculate such indexes across studies. Finally, our literature review only focused on soil microbial abundance measurements. Similar meta-analytical approaches will greatly benefit from the inclusion of community compositional metrics because this will allow examination of changes in diversity/function relationships in response to changing environmental pressures. A good example of how to quantitatively synthesize bacterial community diversity and compositional data derived from the sequencing of the 16S rRNA gene is the meta-analysis by Shade et al. (2013), techniques from which could be implemented when more soil biodiversity data become available in global change studies. High-throughput sequencing is also opening promising avenues for process-based ecosystem models by linking soil organisms’ phylogeny, physiological traits, and responses to global change disturbances (Fierer et al., 2013; Luo et al., 2013; Evans & Wallenstein, 2014).
Strengths and limitations of the approach followed to relate global change effect sizes on microbes and ecosystem functioning
Our synthesis effort goes beyond the assessment of global change effects on soil microbes or EF separately, which has already been done (Treseder, 2008; Liu & Greaver, 2010; Blankinship et al., 2011; Lu et al., 2013). Specifically, we studied whether the responses of soil microbial abundance (fungi, bacteria, and total microbial community) explained variation in those of plant biomass, soil C cycling and soil N cycling. We acknowledge that our analysis is based on statistical associations from the structural equation modeling, and that it does not enable us to estimate ultimate causality such as in controlled experimental designs. However, structural equation modeling allows the assessment of multivariate hypotheses predicting multiple drivers of a treatment effect size (e.g., climate, methodological features, the effect size upon other variables), and thus its use in ecological meta-analysis is growing (Grace et al., 2007; Eldridge et al., 2011; Garcia-Palacios et al., 2013). Our approach allowed us to synthesize current literature, find general patterns and identify key areas for future global change research.
Conclusions
Our literature synthesis demonstrated the existence of strong general links between the responses of soil microbes and EF (plant biomass, soil C cycling and soil N cycling) to global change (elevated CO2, warming, and N deposition). This suggests that soil microbes are important mediators of global change effects on EF. In the case of N deposition and warming, these links were strong and consistent across taxa and ecosystem processes, whereas for elevated CO2 such links were more taxon and ecosystem process specific. The links found between global change, EF and soil microbes support the explicit consideration of soil organisms in ecosystem models. To do so, we need to understand the mechanisms underlying, for example, the effects of plant-soil interactions on N cycle responses to elevated CO2, or how changes in soil C sequestration with N deposition are modulated by microbially driven shifts in litter decomposition. Important gaps (altered precipitation, soil invertebrates, soil community composition, and litter decomposition) prevented us to conduct a broader assessment of the soil biodiversity-ecosystem functioning relationship under global change, and deserve attention in future studies.
Acknowledgements
We thank Barbara Fricks for her help extracting data from the papers. We also thank four anonymous reviewers for improving the manuscript. PGP was supported by a Fulbright postdoctoral contract from the Spanish Ministerio de Educación and by a European Commission's FP7 Marie Curie IEF grant (DECOMFORECO-2011-299214).
