Thirteen decades of foliar isotopes indicate declining nitrogen availability in central North American grasslands


Author for correspondence:
Kendra K. McLauchlan
Tel: +1 785 532 6155


  • Humans are increasing both the deposition of reactive nitrogen (N) and concentrations of atmospheric CO2 on Earth, but the combined effects on terrestrial ecosystems are not clear. In the absence of historical records, it is difficult to know if N availability is currently increasing or decreasing on regional scales.
  • To determine the nature and timing of past changes in grassland ecosystem dynamics, we measured the composition of stable carbon (C) and N isotopes in leaf tissue from 545 herbarium specimens of 24 vascular plant species collected in Kansas, USA from 1876 to 2008. We also parameterized a simple model of the terrestrial N cycle coupled with a stable isotope simulator to constrain processes consistent with observed patterns.
  • A prolonged decline in foliar N concentrations began in 1926, while a prolonged decline in foliar δ15N values began in 1940. Changes in the difference between foliar and atmospheric C isotopes reveal slightly increased photosynthetic water use efficiency since 1876.
  • The declines in foliar N concentrations and foliar δ15N suggest declining N availability in these grasslands during the 20th century despite decades of anthropogenic N deposition. Our results are consistent with progressive-nitrogen-limitation-type hypotheses where declines in N availability are driven by increased ecosystem N storage as a result of increased atmospheric CO2.


Over the past century, humans have drastically altered the global cycles of carbon (C) and nitrogen (N), yet the timing and magnitude of consequences for ecosystems are difficult to predict. Globally, the supply of nitrogen (N) to ecosystems has been increasing since the mid-1900s as a result of increased deposition of reactive N derived from fertilizer production and fossil fuel combustion (Galloway et al., 2008). In grasslands, N availability determines several aspects of ecosystem function, including plant species composition (Tilman et al., 2006), plant productivity (Elser et al., 2007; LeBauer & Treseder, 2008), nutritional quality (Van Soest, 1994; Ritchie, 2000), and retention of C and nutrients (Baer & Blair, 2008; Perring et al., 2008). Because N is a critical regulator of grassland ecosystem processes, and atmospheric N deposition is projected to increase globally by 10–15% within 25 yr (Dentener et al., 2006), determining the baseline and trajectory of N cycling is essential both for understanding past changes in grassland function and for predicting future grassland functioning.

The current rates of deposition of N to grasslands in North America are probably elevated relative to pre-Industrial times. Yet, we have a poor understanding of whether grassland N availability, that is, the supply of N relative to demands by plants and microbes, has been increasing or decreasing over the past few decades. Projecting back in time before the period of instrumental measurements is even more difficult (Galloway et al., 2004). There are two main possibilities. Nitrogen saturation theory predicts that N availability should have been increasing over the past few decades as rates of chronic atmospheric N deposition exceed ecosystem capacity to retain N (Aber et al., 2003). Inorganic N is deposited primarily as nitrate in wet deposition on the Great Plains of North America (National Atmospheric Deposition Program; Near our study area, rates of total inorganic N deposition averaged 4.9 kg ha−1 yr−1 from 1982 to 2008. Alternatively, elevated concentrations of atmospheric CO2 could reduce N availability by increasing sequestration of N in organic pools, including microbial biomass, wood, and recalcitrant plant litter (Luo et al., 2004). This phenomenon, termed ‘progressive nitrogen limitation’, has been observed in free air carbon enrichment (FACE) experiments (Reich et al., 2006a,b; Hovenden et al., 2008), but it has yet to be demonstrated in relatively unmanipulated systems.

Reconstructing past N availability requires retrospective techniques that analyze records of geochemical change. Plant specimens preserved in herbaria retain characteristics that can provide these reconstructions. Foliar N concentrations in modern plants have been used broadly as an index of N availability (Reich et al., 2001; Ollinger et al., 2002; Asner & Vitousek, 2005; Xia & Wan, 2008; Ordonez et al., 2009). In addition, multiple lines of evidence indicate that the standardized ratio of 15N to 14N (δ15N) in foliar tissue integrates a complicated terrestrial N cycle into a single metric (Robinson, 2001). Although the N cycle is complex and changes in the value of δ15N in different ecosystem N pools can reflect changes in many aspects of the N cycle (Kahmen et al., 2008), foliar δ15N assessed at the stand level in forests and grasslands often indexes terrestrial N availability (Craine et al., 2009a). Empirical correlations between traditional metrics of N supply to plants, such as net N mineralization and nitrification, and foliar δ15N have been demonstrated in several regions, such as the northeastern USA and Europe (Pardo et al., 2006). Because of the greater likelihood of 15N-enriching processes in ecosystems with high N availability, foliar δ15N is positively associated with terrestrial N availability on local to regional scales (Högberg, 1997).

To reconstruct changes in N availability in Great Plains grasslands, we examined changes in foliar tissue chemistry and stable N and C isotope composition for 24 widely distributed grassland plant species in Kansas between 1876 and 2008 common era (CE). Our main objective was to determine if N availability to grassland plants increased or decreased over the past 132 yr. If elevated rates of atmospheric N deposition have been causing increases in N availability, then foliar N concentrations and foliar δ15N for all plant species would have increased synchronously with the onset of anthropogenic N deposition. Alternatively, if N availability has been declining over time, as would be expected if progressive N limitation were occurring, then foliar N concentrations and foliar δ15N would have decreased over time. To identify specific changes in the N cycle that would lead to changes in foliar δ15N, we constructed a simple model of the N cycle coupled with a stable isotope simulator that tests the sensitivity of plant δ15N to changes in key components of the N cycle.

Materials and Methods

Herbarium specimens

Leaf tissue from 545 specimens of 24 species of vascular plants common in the grassland biome of Kansas – representing grass, forb, and woody plant functional groups – was selected from the Kansas State University (KSU) Herbarium for this study (Table 1). The KSU Herbarium houses c. 200 000 specimens, with specimen data accessible through an online searchable database ( The collection is especially strong in historical specimens from the central Great Plains of North America (Woods et al., 2005), which allowed us to access 13 decades of plant tissue for analysis. The duration of the record ranged from 113 to 131 yr among species, and between 17 and 30 individual plant specimens were sampled per species (Table 1).

Table 1.   Characteristics of the 24 vascular plant species used in the present study
SpeciesFunctional groupPhotosynthetic pathwayNumber of specimensOldest specimenYoungest specimen
  1. Names follow the Flora of North America Editorial Committee (1993+).

Agastache nepetoidesForbC31818762007
Ambrosia artemisiifoliaForbC32218872008
Chenopodium albumForbC31818852008
Echinacea angustifoliaForbC32618842005
Helianthus maximilianiForbC32018842008
Silphium laciniatumForbC32218872008
Verbesina alternifoliaForbC32318782007
Verbena strictaForbC32718872008
Andropogon gerardiiGrassC42118872008
Bouteloua curtipendulaGrassC43018792008
Bouteloua gracilisGrassC42418862007
Bouteloua dactyloidesGrassC42218842006
Elymus canadensisGrassC32518862008
Panicum virgatumGrassC42818862007
Poa prantensisGrassC31918882002
Schizachyrium scopariumGrassC42618862007
Celtis occidentalisWoodyC32318922008
Cornus drummondiiWoodyC32718792008
Juniperus virginianaWoodyC31718882008
Populus deltoidesWoodyC32118952008
Quercus macrocarpaWoodyC32218882008
Rhus glabraWoodyC32318872007
Rosa arkansanaWoodyC32118862008
Quercus muehlenbergiiWoodyC32218842008

To maximize continuity over time, plant specimens were preferentially selected from a well-sampled county (Riley County) in the Flint Hills of Kansas, from tallgrass prairie environments that were as little altered by human activity as possible. If a specimen with sufficient material for destructive sampling was not available from Riley County, we chose one that was as geographically near as possible, but samples were analyzed from across the state (Fig. 1). Over half of the specimens were collected from Riley County and the seven immediately adjoining counties. The longitude of each sample, a measure of geographic dispersal, was analyzed as a covariate and did not significantly influence the results. Soils and climate are broadly similar, although not identical, across the area sampled. Sampling consisted of removing c. 10 mg of green leaf material, with tissue sampled in a morphologically consistent manner for a single species. Each sample was then ground either with a Wig-L-Bug ball mill or with a mortar and pestle. As standard protocol, plant specimens were dried immediately after collection, so C and N concentrations in leaf tissue were preserved. To our knowledge, no N-containing compounds have been used in the KSU Herbarium for sample preparation or preservation.

Figure 1.

 A map of the collection location within the state of Kansas, USA for herbarium specimens used in this study. Specimens that possessed only county-level locality data were assigned county centroids as locations. Squares, grasses; stars, forbs; triangles, woody plant species.

Elemental and stable isotope analyses

Elemental and isotopic analyses for C and N were conducted at the Stable Isotope Mass Spectrometry Lab (SIMSL) at KSU using standard methods on a ThermoFinnigan Delta Plus mass spectrometer interfaced with a Carlo Erba 1110 elemental analyzer with Conflo II interface (Thermo Fisher Scientific Inc., Waltham, MA, USA). The isotopic ratio of N (δ15N) is expressed relative to the standard of atmospheric N2. The isotopic ratio of C (δ13C) is expressed relative to the standard of Pee Dee Belemnite (PDB). Accuracy, as determined by comparing measured values to the known value of an internal laboratory standard, was better than 0.18‰ for δ15N and 0.12‰ for δ13C across all batches of samples. Precision within each batch of samples analyzed, as estimated by the standard deviation of repeated measurements of a working standard, was lower than 0.2‰ for δ15N and 0.1‰ for δ13C across all batches of samples. The concentration of C could not be calculated for some specimens with limited material, but we obtained isotopic data for all samples.

Foliar δ13C data were used to calculate changes in carbon isotope discrimination:

image(Eqn 1)

13Ca and δ13Cp, the δ13C of air and plant tissue, respectively.) A cubic spline was used to determine annual δ13Ca values using published δ13Ca values from flask samples of atmospheric air (Keeling et al., 2005) and from air trapped in ice-cores (Francey et al., 1999).

Assuming that integrated and instantaneous discrimination values are substitutable, for C3 species, the ratio of the CO2 concentration inside the leaf (ci) to that of ambient air (ca) can be derived from Δ (see Eqn 1) (Farquhar et al., 1982)

image(Eqn 2)

(a, the fractionation resulting from diffusion through stomata (4.4‰); b, the fractionation resulting from carboxylation by Rubisco (c. 30)). We did not calculate cca ratios for C4 species because fractionations that occur during post-photosynthetic processes are large relative to the small Δ values of C4 species (Cernusak et al., 2009), which can lead to unrealistic values (Pedicino et al., 2002). For C3 species, ci : ca can then be used to calculate the intrinsic water use efficiency (iWUE) of plants, which is defined as rate of carbon assimilation (A) divided by stomatal conductance (g):

image(Eqn 3)

(1.6, the ratio of gaseous diffusivities of CO2 to water vapor.)


To begin to constrain the potential drivers of variation in foliar δ15N, we constructed a model of the N cycle and then conducted a series of sensitivity analyses. The model included four pools of N (soil organic matter, soil NH4+, soil NO3, and plant) and eight fluxes of N (net mineralization, net nitrification, deposition of N into NO3, plant uptake of NH4+, plant uptake of NO3, denitrification, NO3 leaching, and transfer of plant N into soil organic matter (SOM) (Fig. 2). This model differs from earlier models of stable N isotopes (Shearer et al., 1974; Garten, 1993; Koba et al., 2003) because it explicitly considers denitrification, an important fractionating pathway considered in recent ecosystem models (Bai & Houlton, 2009). The model was constructed using stella software (ISEE Systems, Lebanon, NH, USA) and stable N isotope values were obtained with nesis (Non-Equilibrium Stable Isotope Simulator) software (Rastetter et al., 2005). The acquisition of N and its transference to plants by mycorrhizal fungi are important determinants of plant δ15N (Hobbie & Colpaert, 2003). Yet, because of the nature of discrimination by mycorrhizal fungi, the potential of changes in reliance by plants on mycorrhizal fungi were estimated outside of this model. Initial sizes and signatures of pools were selected to be representative of temperate grasslands (Knapp et al., 1998) (Table 2). Discrimination factors for fluxes were set based on a survey of values reported in the literature (Högberg, 1997). Although there have been suggestions that N fractionation occurs during microbial assimilation and dissimilation (Dijkstra et al., 2006, 2008), it is difficult to understand ecosystem-level patterns of δ15N if N mineralization is assumed to fractionate (Houlton et al., 2006; Bai & Houlton, 2009).

Figure 2.

 A diagram of the nitrogen (N) cycle model from the present study, with four pools and eight fluxes. Parameter values are reported in Table 2. SOM, soil organic matter.

Table 2.   Parameters for the nitrogen (N) cycle model
ParameterParameter valueSignature or fractionation factor (‰)
  1. Mineralization and deposition are reported as g N m−2 yr−1, while SON is reported as g N m−2. All other parameters are fractions of pools. SON, soil organic nitrogen.

Nitrification0.5 or 0.925
Leaching0.1, 0.5, 0.9 
Denitrification0.1, 0.5, 0.915
Inorganic N deposition0.1, 0.5, 0.9−10, 0, 10
Turnover plant0.1, 0.5, 0.9 
Plant uptake0.8 

To determine equilibrium plant δ15N, the N cycle model was run for 100 yr with a time step of 0.1 yr, with results robust at different time steps. For the first set of runs, nitrification was set to 50% of the NH4+ pool, and for the second set it was set to 90%. For each set, plant δ15N signatures were determined 11 times – once for the base parameters and at two different levels for five parameters. For four fluxes (denitrification, leaching, deposition, and plant mortality), rates were adjusted to either 10% or 90% of the dependent pool in different runs of the model. Modeled denitrification rates were typically close to 0.5 g N m−2 yr−1, but could be as high as 0.9 g m−2 yr−1 in the high denitrification case, similar to Bai & Houlton (2009). The signature of N deposition was also increased and decreased by 10‰ in different runs.


To determine the pattern of foliar N concentrations and foliar δ15N over time, we performed piecewise linear regressions (Toms & Lesperance, 2003) using a nonlinear model algorithm. In some cases, the nonlinear model would not converge on a single set of parameters. In these cases, the inflection point was adjusted manually by single-year increments and then the model with the lowest SSE was used to select the best inflection point. Adding an additional inflection point did not lead to qualitative changes in the patterns of foliar N concentrations or foliar δ15N over time. All analyses were performed in jmp 7.0.1 (SAS Institute, Cary, NC, USA).


The observed values among herbarium samples in this study were wide-ranging, but generally consistent with previous surveys of foliar N concentrations and foliar δ15N. Among samples of vascular plant species, foliar N concentrations varied by over 55 mg g−1 (2.5 to 59.1 mg g−1) and foliar δ15N varied by 22‰ (−7.5 to 14.5‰). For all samples over the duration of the record, foliar δ15N increased with increasing foliar N concentrations (Fig. 3). Foliar N concentrations and δ15N were highest for forbs (25 mg g−1 and 2.9‰, respectively) and lowest in C4 grasses (15 mg g−1 and 0.4‰, respectively) (Table 3). In general, plant species that are associated with disturbed areas and high N availability, for example Chenopodium album, had high foliar N concentrations and high foliar δ15N, while species associated with low soil disturbance and N availability, for example Schizachyrium scoparium, had low foliar N concentrations and low foliar δ15N.

Figure 3.

 The relationship between foliar δ15N and foliar nitrogen (N) concentrations measured on leaves from 545 herbarium specimens. The line is an orthogonal linear regression (= 0.45).

Table 3.   Mean and standard error of chemical composition of leaf tissue from 24 vascular plant species categorized into four functional groups
 Sample sizeC concentration (mg g−1)N concentration (mg g−1)C : Nδ13C (‰)δ15N (‰)
Forbs175380.5 (5.4)25.1 (0.8)18.37 (0.66)−27.32 (0.12)2.91 (0.25)
C3 grasses44430.4 (7.8)20.9 (0.8)22.37 (1.21)−26.94 (0.16)2.52 (0.53)
C4 grasses150453.8 (10.0)15.1 (0.5)34.91 (1.48)−12.13 (0.09)0.42 (0.24)
Woody176470.7 (3.9)22.5 (0.6)23.27 (0.64)−26.45 (0.11)1.99 (0.21)

During the period from 1876 to 2008, both foliar N concentrations and foliar δ15N declined (Fig. 4). Between 1876 and 1926, foliar N concentrations increased nonsignificantly from 22.6 to 23.8 mg g−1 (0.026 mg g−1 yr−1). Starting in 1926, foliar N concentrations began to decline at a rate of 0.086 mg g−1 yr−1 (< 0.001). Over the next 8 decades, foliar N concentrations declined by 7.0 mg g−1 on average to a value of 16.8 mg g−1 in 2008. Patterns for foliar δ15N were similar to those for foliar N concentrations, although they did not begin to decline until over a decade later. Between 1876 and 1940, foliar δ15N increased from 2.0 to 3.4‰, which is a rate of 0.023‰ yr−1 (< 0.05). Between 1940 and 2008, foliar δ15N declined by 4.2‰, from 3.4 to −0.8‰, which is a rate of 0.062‰ yr−1 (< 0.001). The declines in foliar δ15N were still present after standardizing for changes in foliar N concentrations. Adjusting foliar δ15N based on the overall relationship between it and foliar N concentrations, plants with the mean foliar N concentration would still have declined at a rate of 0.010‰ yr−1 beginning in 1911 (< 0.001).

Figure 4.

 Piecewise linear regression between (a) foliar δ15N, (b) foliar nitrogen (N) concentrations, and (c) the residuals of the orthogonal regression between foliar δ15N and foliar N concentrations shown in Fig. 3 and the year in which each herbarium specimen was collected. The inflection points are marked with vertical dashed lines at (a) 1940, (b) 1926, and (c) 1911.

The decline in foliar δ15N during the latter half of the sampled period holds across all functional groups, although sample sizes are too small to adequately test this pattern with each species. Nonetheless, over half of the plant species (13 species) showed statistically significant declines in foliar δ15N over the duration of the record with simple linear regression (Supporting Information Fig. S1). The other 11 species did not exhibit a statistically significant decline in foliar δ15N over time. No species increased in foliar δ15N over the duration of its record.

The model of the N cycle indicated that small changes in denitrification had the largest effect among all the flux pathways on plant δ15N (Fig. 5). Plant δ15N was most sensitive to changes in denitrification rate at both medium (50%) and high (90%) rates of nitrification. Plant δ15N was sensitive to nitrate leaching only if ecosystems were experiencing intermediate rates of nitrification. With partial nitrification, increasing leaching from 0.5 to 0.9 g m−2 yr−1 increased plant δ15N values by 2.3‰, as the leached nitrate is isotopically depleted relative to bulk soil N when partial nitrification occurs. By contrast, with the higher nitrification rate, the nitrate pool was not depleted markedly relative to other sources of N, and changes in loss rate did not alter the signature of available N. When we set nitrification rates to 90% of the mineralized N, the same increase in nitrate leaching increased plant δ15N values by only 0.25‰. Changing the signature of atmospherically deposited N dramatically (± 10‰) did not produce concurrent dramatic changes in plant δ15N (2‰ at 50% nitrification and 2.2‰ at 90% nitrification), as deposited N is a small fraction of total N mineralized and the N quickly enters SOM pools.

Figure 5.

 The modeled change in plant δ15N (‰) when five aspects of the nitrogen (N) cycle are changed under (a) medium (50%) and (b) high (90%) rates of nitrification. Values for the sensitivity analysis for each parameter (low, base, and high) are shown in Table 2. Closed circles, denitrification; open circles, nitrate leaching; closed squares, quantity of N deposited in precipitation; open squares, isotopic signature of N deposition; triangles, plant mortality.

Patterns of foliar δ13C over time suggest that photosynthetic water use efficiency has been steadily increasing over time. Plants with the C3 photosynthetic pathway had an average foliar δ13C value of −26.9‰ while grasses with the C4 photosynthetic pathway had an average foliar δ13C value of −12.1‰ (Table 3). For both C3 and C4 species, Δ increased linearly between 1876 and 2008 (Fig. 6) at rates of 0.0077 and 0.0070‰ yr−1, respectively. The addition of an inflection point between 1876 and 2008 with piecewise linear regression did not markedly alter the relationships (data not shown). For C3 species, increases in Δ were not enough to offset the increases in ca, leading to greater iWUE for C3 species. iWUE of the C3 plants measured in this study increased by c. 33% from 1876 to 2008 as global CO2 concentration increased by 35%.

Figure 6.

 Δ, a proxy for water use efficiency (WUE), calculated from Eqn 1 for each herbarium specimen (a) for C3 plants distributed among the grass, forb, and woody functional groups, = 4.96 + 0.0077x, and (b) for C4 grasses, = −8.65 + 0.0070x.


Nitrogen cycling in grasslands is complex, with many controls over this ecosystem response. Overall, our analysis of leaf tissue from a wide variety of Kansas grassland plant species over 132 yr suggests declines in N availability in the grasslands of the region. Foliar N concentrations began declining in 1926 and foliar δ15N began declining soon after in 1940. A number of distal mechanisms can be postulated as driving long-term declines in N availability, but available evidence supports the hypothesis that increasing atmospheric CO2 concentrations have been causing N to become progressively more limiting to ecosystem productivity. Since 1876, atmospheric CO2 concentrations have increased by c. 100 ppm (Carbon Dioxide Information Analysis Center, Oak Ridge, TN, USA). Experimental enrichment of atmospheric CO2 often causes reductions in foliar N concentrations and N availability to plants (Reich et al., 2006b).

Although elevating CO2 from a modern ambient (375 μmol mol−1) does not always cause foliar N concentrations to decline (Ainsworth & Long, 2005), many individual experiments provide evidence in support of the historical pattern we observed in herbarium foliar tissue. For example, 6 yr of elevated CO2 caused declines in soil N availability for tallgrass plant communities in Minnesota (Reich et al., 2006a). Similarly, soil N availability in Australian grasslands was reduced with 5 yr of experimental CO2 enrichment (Hovenden et al., 2008). Enrichment of atmospheric CO2 from 200 to 500 μmol mol−1 in a Texas grassland decreased foliar N concentrations and soil N availability (Gill et al., 2002). Additionally, experimental increases in CO2 concentration often decrease foliar δ15N in a variety of ecosystems (BassiriRad et al., 2003; Stock & Evans, 2006), although not always (Johnson et al., 2000; Billings et al., 2002; Williams et al., 2006). The only other historical record of archived leaf tissue in addition to the present study also documents declines in foliar δ15N and foliar N concentrations during the 20th century – in Spain from 1920 to 1995 (Peñuelas & Estiarte, 1997; Peñuelas & Filella, 2001).

Although elevated CO2 is a likely driver of the 20th century changes in foliar N characteristics, several other mechanisms are possible. For example, soil disturbance or grazing intensity could have declined during the past 80 yr, reducing N availability. Alternatively, declines in N availability might reflect prolonged ecosystem recovery after the Great Drought of the 1930s. It is unlikely that changes in atmospheric N deposition could be solely causing the observed changes. First, foliar N concentrations would tend to increase, not decrease, with increased deposition (McNeil et al., 2007). Second, the timing of the onset of declines in foliar N and δ15N predates elevated N deposition, which is considered in this region to be synchronous with the beginning of widespread inorganic fertilizer use during the 1950s. Third, our model suggests that decreases in the isotopic signature of N deposition would have to be implausibly large to cause foliar δ15N to decline > 4‰ (Elliott et al., 2007). Changes in climate could underpin changes in soil N cycling (Shaw & Harte, 2001; McCulley et al., 2009), but experimental effects are not always consistent (Verburg et al., 2009). Moreover, parsimonious pairing of climate records for Kansas and hypotheses to explain the patterns observed here are difficult to construct. Although instrumental and proxy records of Great Plains climate show high interannual variability in mean annual temperature and mean annual precipitation since 1876, including periods of drought in the 1930s and 1950s (Woodhouse & Overpeck, 1998), there are no regionally synchronous shifts in climate that coincide with the declines in N availability observed here.

According to the progressive N limitation hypothesis, declines in N availability are driven by an increase in ecosystem N storage through microbial immobilization, woody biomass, and recalcitrant plant litter (Luo et al., 2004). Although microbial immobilization per se might not be necessary to explain some of our observed patterns in foliar chemistry (Dewar et al., 2009), increased atmospheric CO2 can cause N availability to decline by reducing net N mineralization (Gill et al., 2002; Luo et al., 2004; Billings & Ziegler, 2005). Observed reductions in net mineralization were initially considered a response to lower litter quality (Luo et al., 2004), but could result from increased litter production alone (Liu et al., 2009). Based on our modeled sensitivity analysis of the N cycle, increases in microbial immobilization that are paired with decreases in the strongly fractionating nitrification or denitrification rates would be a single mechanism that would cause declines in foliar N concentrations and foliar δ15N. Reduced nitrification activity and numbers of ammonia-oxidizing bacteria have been linked to elevated CO2 (Horz et al., 2004; Bowatte et al., 2008). Despite low moisture availability and perceptions of highly aerated soils, the denitrification potential in upland soils of grasslands, including those in Kansas, is relatively high and strongly limited by NO3 availability (Groffman et al., 1993). Therefore, declines in NO3 availability would decrease denitrification, although it should be noted that elevated CO2 does not always reduce N2O production (Kammann et al., 2008) and that the specific responses of soil C and N dynamics to elevated CO2 extend beyond progressive nitrogen limitation (Finzi et al., 2006; Johnson, 2006).

Foliar depletion of δ15N can also be caused by mycorrhizal fungi. Changes in reliance on mycorrhizal fungi are unlikely to have caused the reduction in foliar δ15N we observed unless changing CO2 concentrations over time were strongly influencing the degree of association of mycorrhizal fungi with plant roots (BassiriRad et al., 2003). An increased rate of mycorrhizal infection under elevated CO2 would lead to lower foliar δ15N because mycorrhizal fungi transfer depleted N to plants (Hobbie & Colpaert, 2003). Most of the plants we sampled associate with arbuscular mycorrhizal fungi, which cause a relatively small decline in foliar δ15N (c. 1‰) relative to nonmycorrhizal plants (Craine et al., 2009b). Other factors that could be influencing foliar δ15N include within-plant transformations, including nitrate reduction, although these have generally been shown to be small relative to isotopic signature of source N at the whole-plant level (Evans, 2001). Systematic changes in inorganic source N are possible over this range of space and time. Foliar δ15N across German grassland sites is positively related to net N mineralization and within a site reflects inorganic N source (Kahmen et al., 2008). Elevated CO2 may also cause increases in nonstructural plant carbohydrates that influence foliar N concentrations (Körner et al., 1997; LeCain & Morgan, 1998). Differences between leaf δ15N and root δ15N are small (average < 0.5‰) across a wide range of grassland vascular plant species (Craine et al., 2005).

The amount of variation in foliar δ15N in this study of Kansas grassland plants (22‰) is approximately two-thirds of the maximum variation in foliar δ15N (32‰) measured globally in natural ecosystems (Craine et al., 2009b). Thus, even though our samples came from a relatively small area of the Great Plains, they include a broad array of plants experiencing a variety of environmental conditions. Positive correlations between foliar N concentration and foliar δ15N are similar to those found across grasslands in multiple continents (Craine et al., 2005). The observed slopes in this study between foliar N concentrations and foliar δ15N were almost identical to those in the global relationship (Craine et al., 2009b): 0.16 and 0.17 ‰ (mg N g−1)−1, respectively.

Consistent with results from controlled experiments, the iWUE of C3 plants increased linearly with CO2 concentration from 1876 to 2008 (Fig. 6). This increase in iWUE is probably a result of decreased stomatal conductance in response to rising CO2 concentrations (Polley et al., 1993; Anderson et al., 2001; Lee et al., 2001). As atmospheric CO2 concentrations increase, stomatal conductance can decrease without limiting photosynthetic rates, which reduces the amount of water lost per unit CO2 assimilated. The response of iWUE to increasing CO2 concentrations found in this study is nearly identical to the response found for several C3 plants grown in controlled experiments (Polley et al., 1993). With Δ for C4 species only increasing c. 1‰ from 1876 to 2008, iWUE should also be increasing for these species as this increase in Δ, which signifies increases in ci, is still small proportional to the increase in ca (see Eq. 3). The observed increases in ci : ca observed here agree with other historical foliage samples for both C3 (Penuelas & Estiarte, 1997) and C4 plants (Pedicino et al., 2002). However, experimental increases in CO2 can lead to a variety of responses in ci : ca– a decrease (Polley et al., 1996), no change (Polley et al., 1993; Marshall & Monserud, 1996; Ward et al., 2005) or an increase (Lee et al., 2001).

While grassland ecosystems of the Great Plains are changing as a consequence of multiple anthropogenic effects, we did not document a trend toward N saturation over the past 132 yr. The quantity of N deposited from the atmosphere onto these grassland ecosystems did not lead to a net increase in N availability to plants. Rather, our results are consistent with the progressive nitrogen limitation hypothesis, which indicates that changes in global atmospheric CO2 concentrations are an important driver of ecosystem function. Although it would seem that increased N deposition would increase N availability, perhaps the rate of inorganic N deposition (4.9 kg ha−1 y−1 from 1982 to 2008 at Konza Prairie Biological Station in Riley County, Kansas) is not yet sufficient to lead to a net increase in N availability. Evaluating the relative roles of changes in land use, grazing, and disturbance regimes will help to clarify how the grassland ecosystems of the Great Plains are responding to global change. Grassland ecosystems may have a greater capacity to retain N than previously assumed, and future management of this threatened ecosystem will need to account for decreased N availability along with other human impacts.


This work was supported by the National Science Foundation (EPSCoR 0553722) and the Kansas Technology Enterprise Corporation. We greatly appreciate technical assistance from Johanna Burniston and Timothy Burrell. Many herbarium specimens were collected and identified by Steve Rolfsmeier and Mark Mayfield. We thank Shiva Mohandass and Jesse Nippert for helpful discussions. Erik Hobbie and reviewers provided useful comments that improved the manuscript. This is publication number 10-012-J of the Kansas Agricultural Experiment Station.