Plant stoichiometric responses to elevated CO₂ vary with nitrogen and phosphorus inputs: Evidence from a global-scale meta-analysis

Abstract

Rising levels of atmospheric CO₂ have been implicated in changes in the nitrogen (N) and phosphorus (P) content of terrestrial vegetation; however, questions remain over the role of C, N and P interactions in driving plant nutrient stoichiometry, particularly whether N and P additions alter vegetation responses to CO₂ enrichment singly. Here we use meta-analysis of 46 published studies to investigate the response of plant N and P to elevated CO₂ alone and in combination with nutrient (N and P) additions across temperate vs. tropical biomes. Elevated CO₂ reduces plant N concentrations more than plant P concentrations in total biomass pools, resulting in a significant decline in vegetation N/P. However, elevated CO₂ treatments in combination with N additions increase plant P concentrations, whereas P additions have no statistical effect on plant N concentrations under CO₂ enrichment. These results point to compensatory but asymmetrical interactions between N, P and CO₂; that changes in N rapidly alter the availability of P, but not the converse, in response to increased CO₂. Our finding implies widespread N limitation with increasing atmospheric CO₂ concentrations alone. We also suggest that increased anthropogenic N deposition inputs could enhance plant N and P in a progressively CO₂-enriched biosphere.

Introduction

Atmospheric carbon dioxide concentrations (CO₂) have grown considerably due to human actions, with an anticipated peak concentration of greater than 700 μmol mol⁻¹ by the end of this century¹. The modern rise in CO₂ is thought to stimulate terrestrial productivity², resulting in negative feedback on atmospheric CO₂ levels and reductions in the pace of climate warming; however, the magnitude of this feedback is thought to be constrained by growth-limiting nutrients, especially nitrogen (N) and phosphorus (P) availability^3,4,5. Here we use meta-analysis to examine changes in N, P and N/P in plants in response to elevated CO₂ and nutrient fertilization across a range of sites and conditions.

The stoichiometry of plant N and P concentrations has provided insight in patterns of N versus P limitation across terrestrial ecosystems^6,7,8, including nutrient limitation responses to elevated CO₂⁹. Past work has demonstrated that elevated CO₂ reduces plant N concentrations generally^10,11, largely as a result of the carbohydrate dilution¹² and the inhibition of nitrate assimilation within plants¹³. The effect of CO₂ enrichment on plant P concentrations has been more variable than for N, with evidence for decreased¹⁴, increased¹⁵ or neutral effects on plant P¹⁶ in individual study systems. Using meta-analysis, Duval et al.¹⁷ showed that plant P responses to elevated CO₂ varied among plant functional groups. More recently, Yuan and Chen¹⁸ showed that elevated CO₂ on average decreases plant N/P across an array of sites; however, this study did not examine tropical plant N/P responses, despite tropical systems having different nutrient conditions than many temperate ones³. Further, past syntheses have not addressed the response of plant N/P to CO₂ in combination with other N, P or N plus P additions. Questions remain over whether N or P will become progressively more limiting under elevated CO₂ and how such limitations will be affected by changes in P and N inputs across temperate vs. tropical ecosystems. This is critical given that C, N and P are among the most anthropogenically altered biogeochemical cycles on Earth¹⁹.

It is increasingly important to recognize that plant responses to CO₂ are acting in combination with other global changes, particularly changes in anthropogenic N deposition and P inputs through cropland fertilization, which can also affect patterns of nutrient cycling and limitation on land. Nitrogen inputs can increase plant productivity and lead to over-enrichment with N²⁰ and can even result in what has been termed “anthropogenic P limitation”³–the progressive occurrence of P limitation to plant productivity via chronic N inputs from human sources. Indeed, past evidence indicates increased plant N/P with increased N inputs^18,21. On the other hand, the increasing use of P fertilizer is improving its availability in ecosystems⁴. The response of plant N/P to altered P availability can be expected, showing a decrease in plant N/P under P fertilization¹⁸.

Yuan and Chen¹⁸ suggested that multiple global change treatments including elevated CO₂, precipitation, warming and N deposition result in additive effects on plant N/P ratios. However, they did not study the effects of elevated CO₂ with P additions or with N and P additions combination. Evidence for widespread nutrient co-limitation suggests that the cycling of one nutrient can influence the availability of another²². Past work in Hawaii has shown that N fertilization increased plant-root and soil phosphatase levels²³, a finding that was later demonstrated across a wide variety of terrestrial ecosystems²⁴. P fertilization was found to enhance N uptake by plants and make plant roots competitive for N against free-living microbes in forest ecosystems²⁵. These interactions between N and P have also been shown to play a role in global patterns of symbiotic N₂-fixation²⁶ and allow plants and mycorrhizal fungi to allocate biomass and energy towards the acquisition of limiting resources in general²⁷.

Here, we analyze plant N/P in response to elevated CO₂ alone and in combination with N, P and N + P additions to test the hypotheses that (1) elevated CO₂ decreases plant N and P concentrations;(2) nutrient inputs prevent declines in plant nutrient contents of the fertilized nutrient under elevated CO₂; and (3) inputs of N or P alters plant nutrient contents of the other nutrient under elevated CO₂ due to compensatory interactions among C, N and P cycles.

Results

We collected 133 observations from 46 separate studies to examine CO₂ treatment effects on plant N/P (see Supplementary References). The majority of the data was from the temperate regions (70%) as opposed to the (sub-)tropics (28%) (see Fig. 1 and Supplementary Table S1). Of these observations, 28 measured plant N/P in FACE; 97 measured plant N/P in chambers; and the remaining 8 were classified as others including branch bag techniques, natural CO₂ springs and screen-aided CO₂ control. Of these species, 86 observations were collected from woody plants and 47 from non-woody plants. There were 19 observations for legumes and 114 for non-legumes. Most studies focused on aboveground plant nutrient concentrations (102 observations) compared to 27 belowground observations and 4 whole-plant concentrations reported. Of the compiled studies, there were 34 observations for elevated CO₂ with N alone, 22 observations for elevated CO₂ with added P alone and 15 observations for elevated CO₂ with added N and P in combination (see Supplementary Table S1).

Figure 1

Map of sites of the collected studies.

Figure 1 was created by R (R Core Team, 2014)⁵⁴ using the maps package.

Effects of elevated CO₂ on plant N, P and N/P

Plant N concentrations decreased significantly with elevated CO₂ (~12%), regardless of the climatic zones, the kind of CO₂ delivery, plant functional group or plant tissue examined (Fig. 2a). Across all observations in this analysis, elevated CO₂ was associated with an averaged 4% decrease in plant P concentrations, but not consistently among factors as observed for decreased N concentrations (Fig. 2b). Rather, elevated CO₂ decreased plant P concentration in temperate regions but increased it in (sub-)tropics. CO₂-incuced declines in plant P concentrations were observed for non-woody plants and aboveground tissues, but not for woody plants and belowground tissues. Hence, elevated CO₂ significantly decreased plant N/P by 11% compared to control conditions, but this response varied with climatic zones (Fig. 2c). Overall, elevated CO₂ decreased plant N/P to a greater extent in (sub-)tropics (23%) than in temperate regions (6%). CO₂ enrichment caused significant declines in plant N/P for chambers (14%) but not FACE experiments, for woody species (13%) but not non-woody species and for legumes (11%) but not for non-legumes.

Figure 2

Effects of elevated CO₂ on plant N and P stoichiometry.

(a) plant N concentrations; (b) plant P concentrations; (c) plant N/P. The error bars show the 95% confidence interval of LnRR. LnRR, the natural logarithm of response ratio that is calculated as the experimental mean divided by the control mean. The asterisk (*) denotes the effect of elevated CO₂ was significant. Different lowercases in the left indicate significant differences between groups. The number of observations for each category is given in the right.

Effects of elevated CO₂ with nutrient fertilization on plant N, P and N/P

In the dataset of elevated CO₂ with N fertilization, elevated CO₂ alone (without N fertilization) significantly decreased plant N concentration (3%) and increased plant P concentration (16%), thus resulting in a decrease of 20% in plant N/P (see Supplementary Fig. S1a). However, elevated CO₂ with N addition did not affect plant N/P, as it led to statistically equivalent increases in plant N concentration (20%) and in plant P concentration (16%) (Fig. 3a). The response ratios of plant N concentrations, P concentrations and N/P with CO₂ enrichment and N fertilization were not related to the amount of N added (see Supplementary Fig. S2).

Figure 3

Effects of elevated CO₂ and nutrient fertilization on plant N and P stoichiometry.

(a) elevated CO₂ with N fertilization (CO₂ + N); (b) elevated CO₂ with P fertilization (CO₂ + P); (c) elevated CO₂ with N and P fertilizations (CO₂ + NP). The error bars show the 95% confidence interval of LnRR. LnRR, the natural logarithm of response ratio that is calculated as the experimental mean divided by the control mean. The asterisk (*) denotes the effect of treatments was significant. The number of observations for each category is given in the right.

Elevated CO₂ without P fertilization tended to negatively affect plant N concentrations, P concentrations and N/P (see Supplementary Fig. S1b). However, the patterns were altered when P was added. CO₂ plus P treatments consistently decreased plant N concentrations (7%), increased plant P concentrations (24%) and decreased plant N/P ratios (26%) (Fig. 3b). The responses of plant N and P concentrations and plant N/P to elevated CO₂ with P fertilization were not related to the amount of P added (see Supplementary Fig. S2).

In the dataset of elevated CO₂ with N and P fertilizations, N and P fertilizations did not significantly affect plant N concentration, P concentration and N/P responses to elevated CO₂ (see Supplementary Fig. S1c and Fig. 3c).

Relationships between plant N and P concentrations

The response ratios (the experimental mean divided by the control mean) of plant N concentrations to elevated CO₂ were significantly positively related to those of plant P concentrations (R² = 0.1716, P < 0.01) (Fig. 4a). There was a significant negative relationship between the response ratios of plant N and P concentration under elevated CO₂ when N added (Fig. 4b), while no significant relationship was observed under elevated CO₂ when P or NP added (Fig. 4c,d).

Figure 4

Relationships between the response ratios of plant N and P concentrations to elevated CO₂ alone and elevated CO₂ with nutrient fertilization.

LnRR, the natural logarithm of response ratio that is calculated as the experimental mean divided by the control mean. The circles indicate the values which were greater than 1 or smaller than −1. Linear regressions were estimated using the points excluding the circles. (a) Elevated CO₂ alone (CO₂); (b) elevated CO₂ with N fertilization (CO₂ + N); (c) elevated CO₂ with P fertilization (CO₂ + P); (d) elevated CO₂ with N and P fertilizations (CO₂ + NP).

Discussion

This cross-system analysis supports expectations for our first hypothesis – that plant N and P contents decline systematically under CO₂ enrichment singly. This result was clear and systematic in the case of plant N concentrations and generally confirmed results of previous research^10,11,28,29. In contrast, plant P responses to CO₂ have been less-examined. Our results showed that P exhibited a more complex CO₂ response than N across a broad range of terrestrial ecosystems, which resulted in some differences in plan N/P. For example, plant P was increased under elevated CO₂ in (sub-)tropics rather than temperate regions, which imply that relatively high soil N availability in (sub-)tropics could help plants acquire P with high energy (C) inputs²⁶. While aboveground P concentrations declined with CO₂ enrichment, P in belowground tissues did not change with elevated CO₂. This suggests that plants may be allocating more P to roots under elevated CO₂, or actively mining soil P pools, perhaps via increased investment in mycorrhizal fungi and fine roots^30,31. In addition, woody plants showed almost no change in P concentrations under elevated CO₂, whereas P concentrations of herbaceous vegetation declined significantly, revealing strong functional-group dependencies in plant P responses. The relative stability in legume N/P under elevated CO₂ indicate that N₂-fixing species has the ability to balance N and P through direct access to N from atmosphere and investment in phosphatase and mycorrhizae to acquire P^26,32.

Our results showed net declines in plant N/P under elevated CO₂, which were in line with those reported for temperate ecosystems by Yuan and Chen¹⁸. Our study advances their analysis by revealing that this decline was driven by much larger reductions in plant N concentrations (12%) than plant P concentrations (4%), including both temperate and tropical ecosystems. In fact, the coefficient of the linear regression (slope) of the responses of plant P vs. N contents to elevated CO₂ was less than unity (P < 0.01) (Fig. 4a). This means that the changes in plant N contents exceeded those of plant P across our global data synthesis. These findings are consistent with a recent study of C3 plants in which a 15% reduction in N but only 9% reduction in P occurred with enhanced CO₂ concentrations³³.

Generally, the decline in plant N/P we observed points to more substantial N vs. P limitation with CO₂ enrichment, which can explained by factors that differentially affect plant metabolism of N vs. P^11,12. Elevated CO₂ may increase the efficiency of photosynthesis or metabolically down-regulate photosynthetic enzymes, thus causing increasing photosynthetic N use efficiency with decreasing N supplies^34,35. Further, higher ATP requirements in response to elevated CO₂ would disproportionally increase P vs. N demands in ecosystems^36,37. In addition, elevated CO₂ could increase plant growth rates, which requires P-rich ribosomal RNA (rRNA)^38,39. Any combination of these mechanisms would lower N/P in response to elevated CO₂.

Our results also support the second hypothesis – that nutrient inputs alter plant N/P response to elevated CO₂ (Fig. 4). This result implies that plant nutrient responses to rising CO₂ will depend on the magnitude of changes in N and P inputs in ecosystems⁹. This is important given that to anthropogenic alterations of N and P that have occurred in concert with atmospheric CO₂. The positive responses of plant N and P to N and P inputs, respectively, have the potential to compensate for their declines induced by elevated CO₂. These results contrast with Cotrufo et al.¹⁰, who found no evidence for an effect of N fertilization on tissue N concentrations beyond the declines observed with elevated CO₂. This discrepancy could reflect our differing approaches; Cotrufo et al.¹⁰ divided their observations into arbitrary classes of N addition whereas we used a combination of meta-analysis and regression to examine nutrient by CO₂ effects. Elevated CO₂ frequently stimulates rates of photosynthesis, yet N or P supply can modulate the magnitude of the CO₂-fertilization effect via its effect on the carboxylation capacity^29,40,41. Wholesale reductions or longer-term acclimation of photosynthesis to elevated CO₂ can be associated with declines in tissue N concentrations^42,43. Our results imply that either N or P supplies can offset any down-regulation of photosynthesis in response to CO₂ enrichment.

Finally, we found support for the third hypothesis – that input of N or P alters plant nutrient contents of the other nutrient– in the case of N by CO₂ additions but not for P by CO₂ treatments. Plant P concentrations increased under elevated CO₂ with N addition, thereby stabilizing plant N/P across terrestrial ecosystems. Nitrogen additions have been shown to enhance P uptake in roots from P-deficient soluble sources under elevated CO₂ by inducing a set of morphological, physiological and molecular adaptive strategies⁴⁴. Numerous studies have shown that N additions can facilitate P acquisition via alterations in root development⁴⁵ and rhizosphere pH⁴⁶ and via organism investments in phosphatase enzymes²⁴. However, there was no apparent trade-off in responses of plant N and P to elevated CO₂ or N addition in our meta-analysis. On the contrary, we found no evidence for a compensating effect of CO₂ plus P fertilization treatments on the declines in plant N concentrations under CO₂. Previous analysis on the effect of P additions on plant N concentrations have been inconsistent, with no changes in plant N observed under N-rich conditions⁴⁷, decreases observed in N-poor soils⁴⁸ and increases under conditions of P limitation²⁵. Our results suggest that, on average, P additions do not alter the pattern of systematically declining plant N concentrations with CO₂ enrichment, consistent with several past studies from individual sites^49,50.

In summary, our extensive meta-analysis shows that elevated CO₂ decreases plant N/P, implying a tendency for N rather than P limitation of terrestrial productivity to arise in response to elevated CO₂ alone. However, when CO₂ and N increase in concert, as is occurring over much of the terrestrial biosphere, compensatory interactions among C, N and P cycles stabilize total plant N/P. These results suggest that, ecosystems exposed to elevated CO₂ and N deposition will not necessarily progress rapidly into conditions of P limitation, owing to a suite of non-symmetrical interactions between N and P. Longer-term effects remain unclear, however, models used to forecast the N and P cycles in response to CO₂ and climate change would be well-served to consider such compensatory interactions in determining the capacity for additional CO₂ sequestration in the future

Methods

We searched ISI Web of Science, using the terms “elevated CO₂”, “CO₂ enrichment” and “CO₂ enriched”, in combination with “nutrient”, “nitrogen” and “phosphorus”, to create a database for our meta-analysis. The data were restricted to studies performed in natural terrestrial ecosystems, i.e., excluding agricultural or other managed ecosystems. All studies included in this analysis measured N/P or N and P concentrations (from which we could calculate N/P) under ambient CO₂ and elevated CO₂. We also limited our data to studies where means, standard deviations of the mean and number of replicates were reported or could be calculated. For studies that measured N and P concentrations at multiple time-points, only the final value was used to maintain the statistically independence between individual observations. DataThief⁵¹ was used to acquire numbers from figures where data were not presented in tables. The units for N and P concentrations were all converted to mg g⁻¹. All N/P are presented here on a mass basis.

We categorized the data by climatic zones, experiment types, plant characteristics and presence any kind of nutrient fertilization involving N and P. The climatic zones included arctic, temperate and (sub-)tropical zones. Research facilities used to increase CO₂ concentration were classified as Free-Air Carbon dioxide Enrichment (FACE) and chambers; plants were categorized based on functional type (woody vs. non-woody plants and non-legume vs. legume) and plant tissues examined (i.e., aboveground vs. belowground).

To examine interactions between elevated CO₂ and nutrient (N and P) fertilization, we also collected data on the level N and/or P additions where available, including elevated CO₂ with N fertilization (from 1 g N m⁻² yr⁻¹ to 20 g N m⁻² yr⁻¹) (CO₂ + N), elevated CO₂ with P fertilization (from 1.5 g P m⁻² yr⁻¹ to 16 g P m⁻² yr⁻¹) (CO₂ + P) and elevated CO₂ with N and P fertilizations (CO₂ + NP). For each of the three datasets, we made two comparisons: elevated CO₂ without nutrient fertilization vs. control (ambient CO₂ without nutrient fertilization) and elevated CO₂ with nutrient fertilization vs. control.

The treatment effects were examined by calculating response ratios (RR) from each individual study. RR is used as an indicator of effect size and is calculated as the experimental mean divided by the control mean (ambient CO₂ without nutrient fertilization)⁵². The natural logarithm of response ratio (LnRR) was then used to perform statistical analysis as it equally weighs the negative and positive responses⁵². The mean effect size was calculated using a mixed-effects model of the meta-analytical software MetaWin 2.0⁵³. The 95% confidence intervals (CI) for effect-size estimates were calculated using resampling techniques. The effects of experimental treatments of a variable were considered to be significant if the 95% CI values did not overlap with zero⁵². We compared the responses to experimental treatments between groups in each category, in which each group should have at least five studies⁵². The responses to experimental treatments among groups in each category were considered to be significantly different when the between-group heterogeneity was significant (P < 0.05)⁵². The differences between means of the groups were considered statistically significant if their 95% CI values did not overlap.