Soil [N] modulates soil C cycling in CO2-fumigated tree stands: a meta-analysis


W. I. J. Dieleman. Fax: +32 3 265 22 71; e-mail:


Under elevated atmospheric CO2 concentrations, soil carbon (C) inputs are typically enhanced, suggesting larger soil C sequestration potential. However, soil C losses also increase and progressive nitrogen (N) limitation to plant growth may reduce the CO2 effect on soil C inputs with time. We compiled a data set from 131 manipulation experiments, and used meta-analysis to test the hypotheses that: (1) elevated atmospheric CO2 stimulates soil C inputs more than C losses, resulting in increasing soil C stocks; and (2) that these responses are modulated by N. Our results confirm that elevated CO2 induces a C allocation shift towards below-ground biomass compartments. However, the increased soil C inputs were offset by increased heterotrophic respiration (Rh), such that soil C content was not affected by elevated CO2. Soil N concentration strongly interacted with CO2 fumigation: the effect of elevated CO2 on fine root biomass and –production and on microbial activity increased with increasing soil N concentration, while the effect on soil C content decreased with increasing soil N concentration. These results suggest that both plant growth and microbial activity responses to elevated CO2 are modulated by N availability, and that it is essential to account for soil N concentration in C cycling analyses.


Terrestrial ecosystems play a pivotal role in the global carbon (C) cycle and are believed to sequester 30% of the anthropogenic CO2 emissions (Canadell et al. 2007; IPCC 2007). As the atmospheric CO2 concentration ([CO2]) continues to rise, plant productivity and C sequestration may further increase, either directly through enhanced photosynthesis or indirectly via increased water- or nutrient-use efficiency. This fertilizing effect of increasing [CO2] is well established (Eamus & Jarvis 1989; Ceulemans & Mousseau 1994; Idso & Idso 1994; Wullschleger, Norby & Gunderson 1997; Norby et al. 1999, 2002, 2005; Temperton, Millard & Jarvis 2003; Norby & Luo 2004; Ainsworth & Long 2005; Hyvönen et al. 2007). However, it has been hypothesized that as nutrients become progressively immobilized in plant biomass and soil organic matter (SOM) pools, nutrient limitations might eventually inhibit CO2-induced increases in plant productivity, and thus limit a further increase of C sequestration in terrestrial ecosystems (Kramer 1981; Ceulemans & Mousseau 1994; Oren et al. 2001; Temperton et al. 2003; Luo et al. 2004; de Graaff et al. 2006). In the short term, priming (a stimulation of microbial growth and activity, and decomposition of older, more recalcitrant organic matter as a consequence of increased labile C inputs (Cheng & Johnson 1998; Kuzyakov 2002; Fontaine et al. 2007) could serve as an alleviating mechanism, and in the longer term, nutrient-poor systems may adjust by steadily redistributing their acquired nitrogen (N) stocks (Rastetter, Agren & Shaver 1997; Cannell & Thornley 1998), but neither mechanism suffices to overcome the progressive N limitation completely.

Coinciding with the rising levels of atmospheric [CO2], atmospheric deposition of reactive N has also increased over the last century (Denman et al. 2007), originating mainly from fossil fuel burning and artificial fertilizer applications (Davidson 2009). Besides a number of negative effects on terrestrial ecosystems, nitrogen fertilization significantly stimulates above-ground biomass production (Ciais et al. 2008; Pregitzer et al. 2008; de Vries et al. 2009; Luyssaert et al. 2010; Thomas et al. 2010), and therefore could increase the amount of plant-derived C entering the soil. While a stronger CO2 effect on above-ground biomass in N fertilized systems compared to unfertilized ones is a general response (Curtis & Wang 1998; de Graaff et al. 2006), this is not the case for the below-ground biomass response (de Graaff et al. 2006). In addition, previous meta-analysis did not reveal a clear microbial C or microbial respiration response to N fertilization in CO2-fumigated tree stands (de Graaff et al. 2006).

Thus, altogether, in contrast to our knowledge of above-ground processes, below-ground processes, and the complex interactions between plants and microbes within the coupled carbon and nutrient cycles are still poorly understood, and it is not elucidated yet whether N constrains C cycling or C inputs stimulate N cycling (Luo, Field & Jackson 2006a). To improve projections of changes in below-ground C pools in response to global climate change, Zak et al. (2000) stressed that fine root production, longevity and biochemistry should serve as a starting point to resolve microbe-related research, because fine root dynamics and rhizodeposition are likely to exert a stronger influence than above-ground litter inputs. van Groenigen et al. (2006) and de Graaff et al. (2006) concluded that soil C can increase significantly in elevated CO2, but only when sufficient amounts of N fertilizer are added. In that case, the increased soil C inputs from plant material could overcome CO2-induced increases in decomposition, resulting in a net increase of soil C.

Using meta-analysis, we investigated how soil C inputs, C pools and C losses are affected by elevated atmospheric [CO2]. More specifically, we aimed to: (1) test the hypothesis that elevated atmospheric stimulates soil C inputs more than C losses, resulting in increasing soil C stocks; and (2) evaluate how N can modulate these responses.

Although some of the N fertilization effects in elevated CO2 studies that were found in previous studies are small or not existing, they do not necessarily imply that nutrient availability only plays a minor role, because unfertilized soils may in fact be nutrient rich, while nitrogen or even NPK-fertilized soils may be characterized by limitations of other nutrients. Therefore, in addition to the comparison of CO2 effects between fertilized and unfertilized experiments, we also compared the CO2 responses in poor and richer systems along a gradient of soil [N].


Data acquisition

We focus on tree stands only because many grassland sites are managed by grazing or through forage production, which plays a key role in the C balance as a large part of the primary production is removed (Soussana et al. 2007). Therefore, grasslands are often fertilized to sustain productivity, which is likely to modify their responses to CO2 fumigation.

Most of the data included in the analysis were extracted from figures and tables in published papers, although some data were not published in peer reviewed literature, but obtained directly from researchers. We collected data on above-ground biomass, fine and total root biomass, fine root production, root-to-shoot ratio, above-ground litterfall, microbial biomass C, heterotrophic respiration (Rh), soil CO2 efflux, net N mineralization and soil C content from 32 sites, resulting in 279 entries for the meta-analysis (Supporting Information Table S1). General information about the sites is given in Table 1.

Table 1.  General information for the sites used in the meta-analysis
SiteTreatmentSpeciesSoil [N] (g N kg−1 soil)LatitudeLongitudeMAP (mm)MAT (°C)
  1. MAP, mean annual precipitation; MAT, mean annual temperature.

Bily KrizCO2Picea abies49.5 N18.53 E14005
Birmensdorf (calcareous)CO2 × fertilization × soilP. abies, Fagus sylvatica0.7647.35 N8.43 E69110
Birmensdorf (acidic)CO2 × fertilization × soilP. abies, F. sylvatica0.51247.35 N8.43 E69110
ChristchurchCO2Pinus radiata0.148543.53 N172.7 E61612.2
DukeFACECO2Pinus taeda0.7935.95 N79.15 W114015.5
DukeFACECO2 × fertilizationP. taeda0.7935.95 N79.15 W114015.5
EUROFACECO2 × fertilizationPopulus alba, Populus nigra, Populus × euramericana1.242.37 N11.8 E70014.1
FACTS II FACECO2 × ozonePopulus tremuloides, Betula papyrifera, Acer saccharum1.249.67 N89.57 W83311.5
Forestry CommissionCO2 × waterFraxinus excelsior, Pinus sylvestris, Quercus petraea35.9 N84.33 W137814.3
GlencorseCO2Betula pendula55.52 N3.2 W8508.2
GlendevonCO2 × fertilizationAlnus glutinosa, B. pendula, Picea sitchensis, P. sylvestris56.2 N4 W14166.8
Mekrijarvi Research StationCO2 × warmingP. sylvestris62.78 N30.95 E6672.1
Merrit IslandCO2Quercus myrtifolia, Quercus geminata, Quercus chapmanii0.128.63 N80.7 W1310-
Montalto di CastroCO2Quercus ilex forest1.142.37 N11.53 E61215
Oak Ridge FACECO2Liquidambar styraciflua1.1235.9 N84.33 W137113.9
Oak Ridge OTC (maple)CO2 × warmingA. saccharum, Acer rubrum35.9 N84.33 W137814.3
Oak Ridge OTC (oak)CO2Quercus alba35.9 N84.33 W137814.3
Oak Ridge OTC (yellow-poplar)CO2Liriodendron tulipifera35.9 N84.33 W137814.3
POPFACECO2P. alba, P. nigra, P. euramericana1.242.37 N11.8 E70014.1
SuonenjokiCO2B. pendula ROTH 4 and 800.4662.65 N27.05 E5793.59
Swiss Treeline FACECO2Larix decidua, Pinus uncinata46.8 N9.83 E  
UA OTCCO2P. sylvestris1.251.17 N4.38 E77012
UMBS – AlderCO2A. glutinosa0.18345.57 N84.67 W  
UMBS – Aspen (eur.)CO2 × fertilizationPopulus × euramericana45.57 N84.67 W  
UMBS – Aspen (grand.)CO2Populus grandidentata0.07945.55 N84.78 W  
UMBS – Aspen (trem.)CO2 × fertilizationP. tremuloides0.97 (Rich soil), 0.21 (poor soil)45.57 N84.67 W  
UMBS – Aspen (trem.2)CO2 × fertilizationP. tremuloides0.97 (Rich soil), 0.21 (poor soil)45.57 N84.67 W  
UMBS – MapleCO2 × fertilizationA. saccharum0.97 (Rich soil), 0.21 (poor soil)45.57 N84.67 W  
USDA – OrangeCO2Citrus aurantium33.43 N112.07 W  
USDA PlacervilleCO2 × fertilizationPinus ponderosa0.85638.73 N120.8 W100018
VielsalmCO2 × fertilizationP. abies50.28 N5.92 E9727.5
WebFACECO2Mixed deciduous forest47.47 N7.5 E990 

We included CO2-enriched studies, using free air carbon enrichment (FACE) or open top chamber (OTC) technology, where roots could proliferate freely (i.e. pot or growth chamber studies were not included). Experimental conditions were summarized by a number of variables (Table 2). Studies were categorized as fertilized when any N-based fertilizer was added during the experiment, or unfertilized when no fertilizer was added. Experiments were classified as irrigated when water was added, and not irrigated when no water was added during the experiment. Some studies also included other manipulations (e.g. temperature, ozone, different soil types or used multiple species in the same experiment). Results from different treatments, plant species, soils or measurement protocols within the same experiment were considered independent measurements. Sampling methods are described in Supporting Information Table S1. We refer to the manuscripts cited in the tables and appendices for detailed methodologies in the specific experiments.

Table 2.  Treatment conditions at the experimental sites
SiteCO2 increase (µmol mol−1)Fumigation typeFertilizationIrrigationAge at start (years)
  1. F, fertilized; FACE, free-air CO2 enrichment; I, irrigated; NF, not fertilized; NI, not irrigated; OTC, open top chamber.

Bily Kriz350OTCNFNI13
Birmensdorf (calcareous)200OTCF and NFI2
Birmensdorf (acidic)200OTCF and NFI2
EUROFACE180FACEF and NFICuttings (3 years SRC)
Forestry Commission300OTCNFI and NI1
Glendevon350OTCF and NFI1 or 2
Mekrijarvi Research Station200OTCNFNI20–30
Merrit Island350OTCNFNIPost-burn (3 months)
Montalto di Castro350OTCNFNI30
Oak Ridge OTC (maple)300OTCNFNI1
Oak Ridge OTC (oak)300OTCNFNI1
Oak Ridge OTC (yellow-poplar)300OTCNFNI1
POPFACE180FACENFICuttings (3 yearsSRC)
Swiss Treeline FACE180FACENFNI29
UMBS – alder345OTCNFNICuttings
UMBS – aspen (eur.)345OTCF and NFICuttings
UMBS – aspen (grand.)350OTCNFICuttings
UMBS – aspen (trem.)350OTCF and NFICuttings
UMBS – aspen (trem.2)200OTCF and NFNICuttings
UMBS – maple200OTCF and NFNICuttings
USDA – orange300OTCFI3
USDA Placerville350OTCF and NFI3
Vielsalm350OTCF and NFI11

Note the difference between fertilized and ‘high soil N concentration’. Although both annotations can be interpreted as nutrient rich, we tested their effects with different analyses: categorical analysis (fertilized versus unfertilized) for the former, and continuous (regression with soil [N] as variable) for the latter one.


Data were analysed with meta-analytical techniques using MetaWin 2.1 software (Rosenberg, Adams & Gurevitch 2000). In conventional meta-analysis, each individual observation is weighted by the reciprocal of the mixed-model variance (Hedges, Gurevitch & Curtis 1999). We used standard deviation (SD) values reported in the individual studies, or calculated the SD from the standard error and the number of replicates (number of FACE rings or OTCs). Studies that did not report standard error or deviation were not included in the database. The natural log of the response ratio (r = response in treatment plots/response in untreated plots) was used as metric in the analyses, and is reported as the percentage change in elevated CO2. The use of the natural logarithm instead of the Hedges d-index has the advantage of linearizing the metric, thereby being less sensitive to changes in a small control group.

A mixed model was used to assess the overall treatment effect of CO2 enrichment, and the influence of fertilizer addition and soil N concentration. We also tested for differences between irrigation treatments, seasonal growth strategy (deciduous or evergreen trees), fumigation type used, amount of CO2 increase and duration of the treatments. When several years of data were reported for one experiment, we calculated a weighted mean, using the reciprocal of the measurement variance.

If the number of studies used to calculate a mean and confidence interval (CI) is lower than 20, the CI can be too narrow (Hedges et al. 1999). Therefore, we used the CI based on resampling methods for the assessment of statistical differences (2500 iterations). Confidence limits based on bootstrapping tests are wider than standard confidence limits, implying that resampling estimates are more conservative (Adams, Gurevitch & Rosenberg 1997). If the calculated 95% CI did not overlap with zero, a significant response to elevated CO2 was accepted. Significant differences between groups (=categorical analyses for treatment comparisons, different seasonal strategies and fumigation technologies) were identified on the basis of the within- and between-group heterogeneity. Significant differences are reported at P < 0.05. Analyses with continuous variables (soil N concentration, and amount and duration of CO2 increase) were performed when the number of studies was larger than 10. Both a weighted regression using MetaWin as an unweighted regression [using Matlab (R2007a) (MathWorks, Natick, MA, USA)] were performed and used for comparison. Significant correlations were reported at P < 0.05.


Soil C inputs

Above-ground litterfall and fine root production responded to elevated CO2 with an increase of 14 and 44%, respectively (Fig. 1; Table 3).

Figure 1.

Overall CO2 effects on soil C inputs, C pools, C losses and N availability. The effects on litterfall (LF), fine root production (FRP), above-ground biomass (AB), total root biomass (TRB), fine root biomass (FRB), root-to-shoot ratio (R/S), microbial biomass C (MBC), soil C content (soilC), heterotrophic respiration (Rh), soil CO2 efflux (SCE) and net N mineralization (Nmin) are indicated as percentage response to elevated CO2. Overall means and confidence intervals (CIs) are given, which means a significant CO2 effect is apparent when the zero line is not crossed. The number of studies used for the analysis is indicated above the x-axis.

Table 3.  Elevated CO2 effects on above-ground biomass, litterfall, total root biomass, fine root biomass, fine root production, root-to-shoot ratio (R/S), microbial biomass, heterotrophic respiration (Rh), soil CO2 efflux, net N mineralization and soil C, indicated as percentage response to elevated CO2
 OverallFertilizedNot fertilized
  1. Numbers in bold italics indicate statistically significant CO2 effects. The CO2 effect is considered significant when 0 is not included in the confidence interval (CI).

C inputs
 Fine root production441952
C pools
 Above-ground biomass213019
 Root biomass395038
 Fine root biomass433646
 Microbial biomass−22–5
 Soil C014–5
C losses
 Soil CO2 efflux192417
N availability
 Net N mineralization36243

Both parameters were not affected differently by elevated CO2 in fertilized and unfertilized plots (Table 4). The CO2 effect on fine root production was positively related to soil N concentration, for both the weighted meta-analysis regression (Table 4) as for the unweighted regression (Fig. 2a). No significant effect of amount or duration of the CO2 increase in the treated plots was observed (Table 4).

Table 4. P values for the meta-analytical comparisons of CO2 effects in different experimental treatments or conditions
 FertilizationCO2 increaseDurationSoil N
  1. Categorical comparison is conducted for fertilization (fertilized versus not fertilized). A continuous regression analysis was performed using the amount and duration of CO2 increase and the soil N concentration as explanatory variables. Numbers in bold italics indicate statistically significant differences between categories or significant correlations. Differences are significant at P < 0.05.

C inputs
 Fine root production0.220.130.28<0.001
C pools
 Above-ground biomass0.690.180.890.5
 Root biomass0.40.470.870.27
 Fine root biomass0.630.480.17<0.001
 Microbial biomass0.820.550.58
 Soil C0.120.390.790.02
C losses
 Heterotrophic respiration (Rh)0.540.060.0450.02
 Soil CO2 efflux0.440.060.090.85
N availability
 Net N mineralization0.250.110.1
Figure 2.

Unweighted regressions relating individual CO2 responses (y-axis) to soil N concentration (x-axis). Responses of fine root production (a), fine root biomass (b), soil C content (c) and heterotrophic respiration (Rh) (d) are depicted as the log response ratio [ln(elevated CO2/ambient CO2)]. Soil [N] is given in g N kg−1 soil. Differences are significant at P < 0.05.

C pools

Above-ground biomass, and total and fine root biomass responded positively to elevated CO2 with a 21, 39 and 43% increase, respectively (Fig. 1; Table 3). This did not result in a statistically significant increase in root-to-shoot ratio in CO2-fumigated studies, although there was a clear positive trend (Fig. 1; Table 3). Microbial biomass C and soil C were not significantly affected by elevated CO2 (Fig. 1; Table 3).

There was no significant difference in CO2 response between fertilized and unfertilized plots for any of the studied C pools (Table 4). However, above-ground biomass and soil C were significantly stimulated by elevated CO2 in the N fertilized studies, but not in the unfertilized experiments (Table 3). The CO2 effect on fine root biomass was positively related to soil N concentration, for both the weighted meta-analysis regression (Table 4) as for the unweighted regression (Fig. 2b). The CO2 effect on soil C content demonstrated a negative relationship with increasing soil N concentration in both regressions (Fig. 2c; Table 4).

There was no significant effect of amount or duration of the CO2 increase in the treated plots (Table 4).

Soil C losses and net N mineralization

Rh and soil CO2 efflux increased by 37 and 19%, respectively (Fig. 1; Table 3). Net N mineralization decreased by 36% in elevated CO2 (Fig. 1; Table 3).

There was no significant difference in CO2 response between fertilized and unfertilized plots for any of the studied parameters (Table 4). Both Rh and net N mineralization were significantly affected by elevated CO2 only in the unfertilized experiments (Table 3). Increasing soil N concentration had a positive effect on the CO2 response of Rh according to the meta-analysis regression (Table 4), while this relationship was borderline insignificant for the unweighted regression (Fig. 2d).

The CO2 effect on microbial respiration increased with treatment duration (Table 4). None of the other parameters exhibited significant relationships with the amount of elevated CO2 added, nor with the duration of the CO2 treatments (Table 4).


Overall CO2 responses

The observed overall CO2-induced stimulation of above- and below-ground biomass and production agrees well with previous experimental findings (Rogers, Runion & Krupa 1994; Curtis & Wang 1998; Pendall et al. 2004; de Graaff et al. 2006). Interestingly, elevated CO2 increased fine and total root biomass and production in all possible experimental conditions addressed in this analysis (Table 3), while this was not the case for the above-ground biomass response to elevated CO2. Together with the positive trend in the root-to-shoot ratio, this provides a strong signal for a C allocation shift towards below-ground biomass compartments in CO2-fumigated systems. This is a common response in an elevated CO2 world (Rogers et al. 1994; Luo, Hui & Zhang 2006b) as plants need more resources to sustain the enhanced growth (Bryant, Chapin & Klein 1983).

Increased above- and below-ground litterfall in elevated CO2 enhances the soil C input. As soil organisms tend to be C limited (Zak et al. 1993; Hu et al. 2006), one would expect an increase in microbial biomass C as a consequence. However, this is not observed. Although microbes probably profit from the improved C availability initially, their biomass turns over relatively quickly (Heath et al. 2005; Lukac et al. 2009), possibly in part because of enhanced grazing by other soil organisms in elevated CO2 (Zak et al. 2000). Moreover, the higher N immobilization in the increasing plant biomass (Luo et al. 2004) may impose a concomitant N limitation of microbial growth (Hu et al. 2001, 2006). We hypothesize that N limitation is a more plausible explanation for the lack of response of microbial biomass to elevated CO2. Indeed, microbial biomass did not increase in spite of the increased plant C inputs. Moreover, the overall response of net N mineralization (a measure for the available inorganic N in the soil) to elevated CO2 was negative, and net N mineralization decreased even more in elevated CO2, when only the unfertilized experiments were included. As larger quantities of C entering the soil normally result in more N uptake, even in N-limited systems (Finzi et al. 2007), our results thus suggest that elevated CO2 makes trees more efficient in immobilizing N, and that microbial growth likely becomes N limited in elevated CO2, at least where N availability is not very high.

Our observed increase in microbial respiration is counterintuitive, considering the lack of microbial biomass response in elevated CO2 studies. However, besides the increase of biomass C inputs in the soil, plants also tend to increase root exudation in elevated CO2 (Fitter et al. 1997; Drigo, Kowalchuk & van Veen 2008; Lukac et al. 2009). This labile C input could fuel the microbial community (Zak et al. 2000; Heath et al. 2005), but is mainly respired because the N necessary to convert these C inputs into microbial biomass is lacking. Therefore, Rh can increase despite the lack of change in microbial biomass. Another possible mechanism is a shift in microbial community composition towards a more fungal-dominated community, which is less N demanding (Hu et al. 2001; Zhang et al. 2005; Carney et al. 2007). This shift may occur, but would still be expected to increase microbial biomass, albeit less pronounced. Both mechanisms could play a role in explaining the positive response of Rh in elevated CO2, but data are lacking to test which of these mechanisms is more important.

As both plant litter production and Rh in CO2-fumigated experiments increase to a similar extent, the lack of response of soil C is not unexpected (Fig. 3). Similar results for forests were already reported by de Graaff et al. (2006), who reported a positive response of soil C only for grasslands. Thus, any increase in C accumulation in tree stands subjected to elevated CO2 will likely be confined to increased woody biomass production.

Figure 3.

Conceptual representation of the overall CO2 effects on C cycling in tree stands. Pools are given in boxes; fluxes are given in lines. Blue lines are C-related processes; green dashed lines are N related. Black circles indicate a statistically significant CO2 effect; grey circles indicate a statistically non-significant trend. Results are considered statistically significant at P < 0.05. SOM, soil organic matter, Nmin, net N mineralization; Rh, microbial respiration; Rr, root respiration; SCE, soil CO2 efflux. Elevated CO2 induces a C allocation shift towards below-ground biomass, where the increased C inputs (fine root production/turnover) increase the CO2 response of microbial respiration (Rh), leaving net change in soil C unaffected in elevated CO2. A strong negative CO2 response of net N mineralization indicates a lower N availability in elevated CO2.

Influence of N fertilization

We did not find any significant differences in the response of above-ground biomass and soil C storage to elevated CO2 between fertilized and unfertilized studies, which is in contrast to earlier studies (de Graaff et al. 2006; van Groenigen et al. 2006; Luo et al. 2006a). However, while both above-ground biomass and soil C are significantly stimulated by elevated CO2 in the fertilized studies, they are not in the unfertilized ones, suggesting that the lack of a statistically significant difference might be caused by low statistical power because of a smaller data set (in this analysis, only tree stands were considered). Our focus on tree stands only may also explain part of the difference, because forests and grasslands have very different C use patterns (Schulze et al. 2009).

van Groenigen et al. (2006) and de Graaff et al. (2006) indicated that soil C accumulation is significantly enhanced in elevated CO2 only when sufficient amounts of N were added. They hypothesize that as nitrogen fertilization enhances plant productivity, it therefore could increase the amount of C entering the soil, resulting in a net increase of soil C. The reason for this would be that the stimulated plant production in fertilized stands would overcome the increased decomposition in elevated CO2. While this hypothesis also fits our results, we suggest a prominent role for fine root dynamics as soil C inputs (Zak et al. 2000). As de Graaff et al. (2006) suggested, we found a stronger CO2 response of woody biomass compartments in the fertilized stands. However, while the fine root biomass and production response to elevated CO2 was significantly affected in both fertilized and unfertilized tree stands, the CO2 response is larger in the unfertilized ones. Indeed, N fertilization is known to generally stimulate woody biomass increase, without affecting soil C inputs (Pregitzer et al. 2008). At the same time, we observed a strongly increased Rh response to elevated CO2 in the unfertilized stands, while the CO2 effect is not significant in the fertilized stands. Our results therefore suggest that, because of an increased soil exploration in unfertilized stands, fine root dynamics and rhizodeposition will be more pronounced, and will serve as a direct substrate for the microbial community. As a consequence, we see an increased soil C storage in the fertilized stands subject to elevated CO2, while a negative trend is apparent in the unfertilized stands.

Aside from this C input-related feedback, retarded rates of SOM decomposition in N-fertilized systems are common (Fog 1988; Berg & Matzner 1997), which could also contribute to an increased soil C storage in tree stands (Janssens et al. submitted). The inhibitory effects of N fertilization on SOM decomposition can be obscured in CO2-fumigated experiments (Janssens et al. 2010), as soil C inputs typically increase under CO2-fumigated systems (DeLucia et al. 1999; Pendall et al. 2004; Subke, Inglima & Cotrufo 2006; Liu et al. 2007; Soussana & Luscher 2007), and CO2 elevation stimulates root exudation and rhizodeposition, all of which affect microbial activity (Norby, O'Neill & Wullschleger 1995; Canadell, Pitelka & Ingram 1996; Lipson, Wilson & Oechel 2005). This could also explain why N fertilization only stimulates soil C accumulation in elevated CO2 when very large amounts of N are applied (van Groenigen et al. 2006).

Relationship with soil N concentration

Soil N concentration was significantly correlated to responses of fine root biomass and –production, microbial respiration and soil C to elevated CO2 (Table 4). Figure 2 illustrates that roots and Rh show only minor responses when soil N is low, which is to be expected, as the elevated CO2-induced growth stimulation cannot be sustained without sufficient available N. In N-rich soil, however, we see that elevated CO2 strongly increases plant productivity, which affects below-ground C cycling through a stimulation of both C inputs and losses. In N-rich soils, this accelerated C cycling under elevated CO2 even results in a negative effect of elevated CO2 on soil C storage with increasing soil N concentration (Table 4; Fig. 2). Overall, these results confirm that the CO2 effect on soil C inputs is the driving factor in soil C cycling, and can be modulated by N. However, these effects of soil N on the elevated CO2 responses in our analysis differ from the approach where N fertilization is used as a measure for soil N availability in tree systems. Therefore, responses to elevated CO2 and interactions with N are summarized in Box 1.

Box 1. Interactive effects of elevated CO2 and N on C cycling in tree stands.

N fertilization and soil N concentration are two different ways to approach N availability, yet interact differently with elevated CO2 (see grey coloured part in the table). These contrasting effects are mainly a function of the direct availability of the N. In fertilized stands, the N is added in mineral form, while in N-rich systems, the N is still embedded in organic molecules or bound to the soil matrix. CO2-fumigated tree stands will respond to both conditions in a different way, starting with fine root dynamics. In N-fertilized stands, the readily available N reduces the need for soil exploration by fine roots, and the associated reduction in rhizodeposition decreases the stimulation of Rh in elevated CO2. This decreased decomposition response in N-fertilized tree stands provides a larger potential for soil C accumulation. When tree stands are not fertilized, the larger demand for N in elevated CO2 elicits an increased soil exploration by fine roots. Therefore, only in the N-rich systems, plants can sustain the increased growth responses in elevated CO2. In response to the exacerbated fine root dynamics in elevated CO2, the Rh response to elevated CO2 increases, resulting in a decreasing soil C response to elevated CO2 in systems with high soil N concentrations.

inline image

Our findings indicate that an approach where treatments are simply opposed is a convenient way for statistical analysis, but fails to cover important information that is often not accounted for. Unfertilized sites can be nutrient rich, and therefore partly miss the expected fertilization treatment response (e.g. POPFACE or EUROFACE experiments). Similarly, it is normal that nutrient-poor sites that are fertilized demonstrate the largest relative responses (e.g. the Birmensdorf experiments). This could be one of the reasons why we do not always find a clear effect of fertilization treatments on soil C cycling in elevated CO2. The response to CO2 is an issue of definition, time and resource supply (Körner 2006). Based on our results, we suggest that more attention should be given to proper descriptions and reporting of experimental characteristics and soil properties in manipulation experiments. Parameters such as soil nutrient or water availability, or stand leaf area index (LAI), age and management are often not described properly, although they play a very important role in regulating plant responses to elevated CO2 and would make evaluating tree stand responses more accurate. Therefore, we underline that a better understanding of terrestrial ecosystem responses to global change could be obtained from better or more standardized reporting of experimental conditions.


Our results confirm the important role of fine root dynamics in soil C cycling in elevated CO2, as the increased fine root activity induced an acceleration of SOM decomposition processes. At the same time, N availability can limit plant growth responses and can therefore influence soil C cycling responses in elevated CO2. While we failed to indicate differences between N-fertilized and -unfertilized tree stands in elevated CO2, we clearly showed that soil N concentration can modulate soil C cycling. In elevated CO2, fine root biomass and production, and Rh all increase with increasing soil N, while soil C decreases with total soil N concentration, regardless of N fertilization. We can therefore conclude that soil C cycling rates and soil C sequestration potential in elevated CO2 will be influenced by initial soil properties and fertility, because (woody) plant growth responses to elevated CO2 are dependent on N availability, while below-ground responses are more dependent on changing soil C availability.


We thank L.E. Henry and K. Pregitzer for providing data, as well as C. Körner for useful information about their experimental sites. Special thanks go to the researchers of the ECOCRAFT [Framework programmes of the EC (EC contracts within 5FP and 6FP, Environment and Research)] and the EUROFACE projects, who provided the large number of data that formed the foundation for this paper. M.L. is a postdoctoral research associate of the Flemish Science Foundation FWO (Fonds Wetenschappelijk onderzoek, FWO Vlaanderen). I.J. holds an FWO research grant, and S.L. holds a European Research Council (ERC) grant. Part of this research has been funded through the UA-Research Centre of Excellence ECO.


  • 1

    References used in the meta-analysis (see Supporting Information for full references).