Author for correspondence: Joseph Craine Tel: +1 785 532 3062 Email: firstname.lastname@example.org
Ratios of nitrogen (N) isotopes in leaves could elucidate underlying patterns of N cycling across ecological gradients. To better understand global-scale patterns of N cycling, we compiled data on foliar N isotope ratios (δ15N), foliar N concentrations, mycorrhizal type and climate for over 11 000 plants worldwide. Arbuscular mycorrhizal, ectomycorrhizal, and ericoid mycorrhizal plants were depleted in foliar δ15N by 2‰, 3.2‰, 5.9‰, respectively, relative to nonmycorrhizal plants. Foliar δ15N increased with decreasing mean annual precipitation and with increasing mean annual temperature (MAT) across sites with MAT ≥ −0.5°C, but was invariant with MAT across sites with MAT < −0.5°C. In independent landscape-level to regional-level studies, foliar δ15N increased with increasing N availability; at the global scale, foliar δ15N increased with increasing foliar N concentrations and decreasing foliar phosphorus (P) concentrations. Together, these results suggest that warm, dry ecosystems have the highest N availability, while plants with high N concentrations, on average, occupy sites with higher N availability than plants with low N concentrations. Global-scale comparisons of other components of the N cycle are still required for better mechanistic understanding of the determinants of variation in foliar δ15N and ultimately global patterns in N cycling.
Despite its potential, interpreting global-scale patterns of foliar δ15N remains a challenge. Plants can acquire N from a number of sources, each of which could have a unique signature and vary over time. For example, the δ15N of soil organic matter (SOM) often increases with depth, ammonium is generally more enriched than nitrate, while depositional N is variable both spatially and temporally. The N cycle also contains a large number of fractionating steps, which generates multiple competing interpretations for any pattern of observed foliar signatures. In general, there are three sets of unresolved relationships that currently limit the use of foliar δ15N in inferring patterns of N cycling at the global scale. First, foliar δ15N values appear to consistently differ among plants that are associated with different types of mycorrhizal fungi. For example, at a given site, plant species associated with ericoid and ectomycorrhizal fungi can have foliar δ15N signatures that are 3–8‰ lower than nonmycorrhizal plants (Michelsen et al., 1996, 1998). It is uncertain whether arbuscular mycorrhizal fungi contribute to variation in foliar δ15N (Handley et al., 1999b), largely because they are thought to be involved mainly in plant P nutrition. In addition, although the results of individual studies comparing mycorrhizal types are robust, general patterns among mycorrhizal types have never been synthesized at global scales, and there are differences among studies in the general ranking of signatures among mycorrhizal types (Michelsen et al., 1998; Pardo et al., 2006).
Second, relationships between foliar δ15N and climate need to clarified. With respect to rainfall, Austin & Sala (1999) observed that foliar δ15N increased with decreasing mean annual precipitation (MAP) in data collected by Schulze et al. (1998) in northern Australia. The increase in foliar δ15N with decreasing MAP was later shown to exist at global scales (Handley et al., 1999a). With respect to temperature, Martinelli et al. (1999) showed that foliar δ15N was greater for tropical forests than temperate forests, suggesting a global relationship between temperature and foliar δ15N. Amundson et al. (2003) later quantified positive linear relationships between mean annual temperature (MAT) and soil δ15N, foliar δ15N and the difference between leaf and soil δ15N.
Although the climate patterns appear robust at multiple spatial scales, these relationships need to be improved in three ways. First, little data from cold ecosystems were included in these regional and global reviews. Coverage of foliar δ15N in ecosystems with low MAT has recently expanded (Welker et al., 2003), allowing for better examination of foliar δ15N patterns in cold ecosystems, i.e. MAT < 0°C. Second, although the relationship between foliar δ15N and temperature or precipitation seems well-established, no global synthesis of foliar δ15N has accounted for potential covarying changes in the types of mycorrhizal fungi along climate gradients. For example, increases in foliar δ15N with increasing MAT might be caused by a shift from ectomycorrhizal or ericoid mycorrhizal species to arbuscular mycorrhizal species and it is important to separate the influence of mycorrhizal fungi out of general climate relationships. Lastly, patterns of foliar δ15N for plants that potentially fix N2 have not, to our knowledge, been examined across climatic gradients. Because of the ability of these plants to obtain N from symbiotic N2-fixing bacteria, a process that does not discriminate against 15N, their acquisition of N can be somewhat independent from soil N dynamics (Högberg, 1997). As the 15N signatures of N2-fixing plants might be independent of climate and not reflect soil processes, these species are often excluded from syntheses, although it is unknown how the inclusion or exclusion of these species would bias the relationships between foliar δ15N and climate.
For the third major area of improvement, foliar δ15N and foliar N concentrations appear to be correlated, but it is uncertain how robust this pattern is at the global scale and whether it is independent of climate. Multiple studies at various spatial scales have shown a consistent positive relationship between foliar N concentrations and δ15N: across 15 temperate and tropical forest sites (Martinelli et al., 1999); within and among plant species at Glacier Bay, Alaska (Hobbie et al., 2000), across 67 grasslands in four regions of the world (Craine et al., 2005), and within and among species and temperate forest stands (Pardo et al., 2006). The relationship has the potential to serve as another strong constraint on the pattern of N cycling across stands if it could be demonstrated to hold at the global scale.
Once the global patterns of foliar δ15N are established, they will be useful only inasmuch as they can be used to infer underlying characteristics of the N cycle. One of the most promising potential interpretations of foliar δ15N is that it might be correlated with soil N availability to plants – the supply of N to terrestrial plants relative to their N demands (McLauchlan et al., 2007). Although multiple studies have shown positive relationships between an index of N availability and foliar δ15N (Garten & Van Miegroet, 1994), it is currently unknown whether regional or global patterns of foliar δ15N consistently index N availability to plants. There are two main reasons why foliar δ15N might index soil N availability to plants. First, when N availability is high, N lost from the ecosystem is more likely to be depleted in 15N, which increases the δ15N of leaves. When N availability is low, N is cycled and lost primarily as organic N (Hedin et al., 1995; Neff et al., 2003). As N availability increases, the production of ammonium and then nitrate increases. Gaseous N loss during nitrification (Firestone & Davidson, 1989) and the leaching of 15N-depleted nitrate (Högberg, 1997; Koba et al., 2003) can cause the remaining N pool (and subsequently plants) to be enriched in 15N. With further increases in N availability, denitrification begins to consume a greater fraction of nitrate (Hall & Matson, 2003), further enriching inorganic N pools, provided that denitrification does not completely consume the nitrate pool (Houlton et al., 2006). Ammonia volatilization also discriminates strongly against 15N (Högberg, 1997). In ecosystems where ammonia volatilization is prevalent, an analogous pattern of increasing ammonia volatilization with increasing N availability would lead to a similar positive relationship between N availability and foliar δ15N, provided that the ammonia is not taken up by leaves (Frank & Evans, 1997).
The second main reason that foliar δ15N might index soil N availability is that plants experiencing low N availability might be more likely to be dependent on mycorrhizal fungi for N acquisition than at high N availability, and the N that mycorrhizal fungi transfer to plants is depleted in 15N (Taylor et al., 2000; Hobbie & Colpaert, 2003). Current evidence suggests that N-containing transfer compounds such as glutamine are depleted in 15N relative to bulk fungal N, presumably owing to the large discrimination against 15N associated with processes such as transaminase reactions (Macko et al., 1986; Stoker et al., 1996). In addition, mycorrhizal fungi at low N availability should transfer a lower proportion of the N that they take up than at high N availability (Hobbie et al., 2000; Hobbie & Colpaert, 2003). As N availability decreases, the benefit to plants to use mycorrhizal fungi for N acquisition increases. Hence, the dependence of plants on mycorrhizal fungi for N could increase consistently with decreasing N availability. Plant δ15N would decrease with decreased N availability if the proportion of nitrogen acquired by plants that is supplied by fungi increases or if the proportion of nitrogen taken up by fungi that is transferred to the host plants decreases (Hobbie et al., 2000).
To better understand the global patterns of foliar δ15N and N cycling, we compiled a data set of over 11 000 geo-referenced observations of foliar δ15N and foliar N concentrations from wild-grown, unfertilized plants. Data were matched with climate parameters and mycorrhizal types. We then examined the global relationships between foliar δ15N and the type of mycorrhizal fungi association, mean climate parameters, and plant N and P concentrations. The relationship between foliar δ15N and climate was determined separately for plants that are not associated with N2-fixing bacteria and from those species that are known to be or might be (e.g. plants of the Fabaceae family). We then examined the relationships between foliar δ15N and different measures of N availability for 15 studies to determine if there are consistent positive correlations between N availability and foliar δ15N. In general, we expect that sites with greater N availability, for example, warm, dry sites with plants with high foliar N concentrations, should have leaves that are enriched in 15N. That said, we are uncertain of the specific relationships between climate or foliar N concentration and foliar δ15N.
Materials and Methods
Data on foliar δ15N and foliar N concentrations of nonN2-fixing plants were identified from: previous published reviews; searching ISI Science Citation Index (http://www.isiknowledge.com) for papers that included terms such as ‘nitrogen isotopes’ or ‘15N’; and contacting individuals that were thought to have been likely to have collected data on foliar 15N. Data on non-N2-fixing plants were acquired from 91 studies distributed from around the world (Fig. 1) encompassing 900 sites, 1103 taxa and 9757 plants. Data for potential N2-fixing plants (those known to fix atmospheric N2 and all members of the Fabaceae family, unless verified not to fix atmospheric nitrogen) were found for 35 studies, 120 sites, and 283 taxa. The combined data set encompassed 95 studies, 916 sites, and 1386 taxa (see the Supporting Information, Notes S1, for a list of published studies used here). Data for each study were generally at the level of the individual plant, although for a few records, data from multiple individuals had been originally summarized at the site level. All data come from unfertilized wild-grown plants whose signatures are likely to reflect soil N cycling. Based on this criterion, nonvascular, wetland, agricultural, carnivorous, mycoheterotrophic and epiphytic plants were excluded. In addition, we also did not attempt to review nonepiphytic orchids because of their unique mycorrhizal symbiosis combined with a lack of data on the global scale. The 15N : 14N ratios are expressed using δ notation where foliar 15N : 14N ratios are expressed relative to the ratio of 15N : 14N in atmospheric N2 : δ15N (‰) = (Rsample/Ratm – 1) × 1000, where Ratm = 0.0036765. The δ15N of atmospheric N2 by definition is 0.0‰.
In addition to foliar δ15N, we compiled data on foliar N concentration ([NL]), the type of mycorrhizal association, MAT and MAP from published and unpublished studies. A total of 6627 of the plants were categorized as arbuscular, 1761 as ectomycorrhizal, 650 as ericoid and 699 as nonmycorrhizal. We note that in our dataset all mycorrhizal associations were found in ecosystems with MAT from −10 to 23°C and MAP from 200 mm yr−1 to 1400 mm yr−1 (Fig. 2).
Data on mean annual temperature and mean annual precipitation were derived from the original publications, if available. If no climate data were published, 30-yr means (1961–90) for annual temperature and precipitation were obtained from site location data and a global climate database at 10-min resolution (New et al., 2002). Mycorrhizal associations were occasionally assessed by the authors in some cases and derived from various published sources (Brundrett, 2004, 2008; Smith & Read, 2008). Arbutoid mycorrhizas were classified as ectomycorrhizal.
Foliar δ15N was first analysed with a general linear model that included MAT, MAP, mycorrhizal association and [NL]. Log-transformed MAP and [NL] were found to explain more variation in foliar 15N than the untransformed data and were used in all subsequent models. Observing relationships between MAT and foliar δ15N revealed a potential breakpoint around which there would be two different relationships between MAT and foliar δ15N. To determine the temperature around which to divide ecosystems, we first calculated the residuals of foliar δ15N with respect to [NL], mycorrhizal association, and log-transformed MAP. We then fitted a nonlinear model with the residuals and MAT that fits two linear segments end to end (i.e. a piecewise linear regression) (Toms & Lesperance, 2003). The nonlinear model solves for four parameters: a y-intercept, the slope of the first linear relationship, a breakpoint and the slope of the second linear relationship. Including a constant to alter the intercept of the second line did not markedly change its elevation. Because the model did not independently converge on a single answer, we ran multiple models that each had a different breakpoint temperature set a priori and selected the temperature that produced the lowest error sums of squares. Data from Falkengren-Grerup et al. (2004) were removed from the dataset as the data altered the breakpoint for MAT from −0.5 to 8.5°C and they appeared to be extreme outliers in other relationships (see below).
After determining the MAT breakpoint, the main regression model for foliar δ15N included log-transformed [NL], mycorrhizal association, log-transformed MAP, MAT and a categorical factor that divided ecosystems into those with MAT < −0.5°C and those with MAT ≥ −0.5°C. A more complicated model that included all pairwise interactions among the main factors only explained an additional 3% of the total variation in foliar δ15N and complicated the interpretations with little real additional insight into relationships. For MAP, MAT and [NL], we also calculated foliar δ15N residuals of multifactor regression models to more specifically examine the relationships between the variables and foliar δ15N.
After attempting to explain variation in foliar δ15N with the four main variables, we tested the ability of the identity of the continent, seasonality of precipitation, and soil pH to explain residual variation in foliar δ15N. Soil pH data was taken from the IGBP Global Soils Data Task (Belward et al., 1999). Relationships between foliar δ15N and foliar N and P concentrations were assessed for 1014 plant samples for which both N and P concentrations were provided. With fewer data points, the explanatory model only included [NL], foliar P concentration ([PL]), and the interaction between the two. Results were similar whether elemental concentrations were log-transformed or not (r2 = 0.34 for both) and additional covariates such as MAT, MAP or mycorrhizal type did not qualitatively alter the relationships between foliar δ15N and nutrient concentrations.
To determine whether patterns among potentially N2-fixing plants were similar to those that are not N2-fixing, the last set of analyses examined the patterns of foliar δ15N for potentially N2-fixing plants. We examined the 1604 records of foliar δ15N of this data set in a general linear model that was similar to the one used for nonN2-fixing species. Mean annual temperature, MAP and [NL] were included in the model, but mycorrhizal associations were not because all the plants are associated with arbuscular fungi. No categorical differences in the relationship between MAT and δ15N at different temperatures were found and therefore no categorical variable for MAT was included.
To better understand relationships between foliar δ15N and soil N availability, we examined two additional datasets. First, for studies that provided data on bulk soil δ15N as well as foliar δ15N, we examined the relationship between foliar δ15N and soil δ15N, and the difference between foliar and soil δ15N. Since foliar δ15N is initially dependent on the signature of soil organic nitrogen from which the N made available to plants is derived, the offset between soil and plant signatures is considered an index of the short-term enrichment or depletion of 15N available to plants and a potentially better index of short-term processes than foliar δ15N alone (Emmett et al., 1998; Amundson et al., 2003). For each data point, soil δ15N was calculated from 0 to 20 cm depth where possible. Second, 14 studies measured both an index of N supply and foliar or aboveground biomass δ15N over landscape to regional scales. In these studies, N supply data were measured as: in situ net N mineralization; laboratory incubations of soils under standard conditions; in situ net N mineralization with resin bags; or inorganic N leaching with lysimeters (Hogbom et al., 2002). For each study, we determined the relationships (generally linear) between N supply and δ15N of leaves or aboveground biomass. A few studies had measured N supply with multiple techniques (in situ and laboratory incubations) or for different portions of the soil (e.g. O vs A horizon). In these cases, the relationship with the highest coefficient of determination was selected for display after ensuring that this selectivity did not qualitatively alter the overall synthesis. Likely associated with the variability in how N supplies were assessed, there was little relationship between N supply and foliar δ15N when the data for all studies were joined into a single dataset (data not shown). All statistical analyses were conducted in jmp 5.1 (SAS Institute, Cary, NC, USA).
In the global data set of non-N2-fixing plants, the variables MAT, MAP, mycorrhizal association, and [NL] explained 56% of the total variation in foliar δ15N. In contrast to previous analyses, with this larger data set, foliar δ15N did not increase monotonically with MAT (Fig. 3a). Foliar δ15N showed no trend with MAT for ecosystems with MAT < −0.5°C, implying similar reliance on mycorrhizal fungi or similar relative importance of different loss pathways across the range of sites with MAT < −0.5°C. For ecosystems with MAT ≥ −0.5 C, foliar δ15N increased at a rate of 0.23‰ °C−1 (Fig. 3a). Foliar δ15N declined with increasing MAP with no evidence of a breakpoint in the relationship analogous to the one observed for the relationship between MAT and foliar δ15N (Fig. 3b). On average, foliar δ15N decreased by 2.6‰ for every order of magnitude increase in MAP.
The main groups of mycorrhizal fungi differed in foliar δ15N by almost 6‰ (Fig. 4). The type of mycorrhizal association explained the second-largest fraction of the explained variation in foliar δ15N (29%, Table 1), even though 68% of the species were associated with arbuscular mycorrhizal fungi. Standardized to a common MAT (13.2°C, arithmetic mean), MAP (751 mm yr−1, geometric mean) and [NL] (15.8 mg g−1, geometric mean), nonmycorrhizal plants had the highest average foliar δ15N (0.9 ± 0.2‰). Among mycorrhizal plants, arbuscular mycorrhizal plants had the highest mean foliar δ15N (−1.1 ± 0.1‰), ericoid mycorrhizal plants the lowest (−5.0 ± 0.2‰) and ectomycorrhizal plants were intermediate in mean foliar δ15N (−2.3 ± 0.2‰).
Table 1. Regression results for foliar δ15N of nonN2-fixing plants and potentially N2-fixing plants
MAT, mean annual temperature; MAT Break, a categorical separation of sites based on a MAT of −0.5°C. MAP, mean annual precipitation; [NL], leaf N concentration.
For nonN2-fixing plants, r2 = 0.56, n = 9757. For potentially N2-fixing plants, r2 = 0.23; n = 1604; SS = sums of squares.
After controlling for mycorrhizal type and climate, foliar δ15N increased linearly with the logarithm of [NL] (Fig. 3c) with [NL] accounting for 44% of the explainable variation in foliar δ15N (Table 1). Doubling [NL] increased foliar δ15N by 2.1‰. Among the 1014 foliar samples from nonN2-fixing species where both N and P had been measured, after controlling for variation in P concentrations, foliar δ15N increased at a rate of 4.5‰ for every 10 mg g−1 increase in N concentrations (Table 2), similar to the increase for the greater dataset that constrained for other factors. By contrast, after controlling for variation in N concentrations, foliar δ15N decreased at a rate of 4.7‰ for every 1 mg g−1 increase in P concentration. Relationships between N concentrations and foliar δ15N were independent of P concentrations and vice versa.
Table 2. Relationships between foliar δ15N and concentrations of nitrogen ([NL]) and phosphorus ([PL]) in leaves
Coefficient (‰ per mg g−1)
Model r2 = 0.34, n = 1014.
−3.05 ± 0.35
0.45 ± 0.02
−4.72 ± 0.32
[NL] × [PL]
−0.038 ± 0.026
Almost half of the unexplained variation in foliar δ15N was found within sites (16% of the total variance, data not shown). Among sites, little additional variation could be explained by incorporating other factors, although there often were significant differences. For example, foliar δ15N differed significantly among continents, with leaves from Australia and Europe depleted in 15N relative to the global mean (1.1 and 2.0‰, respectively) and North America enriched by 0.6‰ after accounting for the other four variables. Sites with primarily winter rainfall had 0.7‰ lower site-mean δ15N than other sites. Soil pH explained only 0.2% of the residual variation in foliar δ15N, but was not a significant predictor of foliar δ15N after taking into account the seasonality of precipitation.
Excluding N2-fixing plants did not strongly bias relationships between climate and foliar δ15N. Among potentially N2-fixing plants, patterns of foliar δ15N fell within the general envelopes delineated by the relationships between foliar δ15N of non-fixing plants and MAT, MAP and [NL]. Potentially N2-fixing species were more prevalent at high MAT than low MAT but spanned most of the precipitation gradients. By contrast, potentially N2-fixing species did not span the full range of foliar N concentrations, having relatively few observations at low [NL]. Mean annual temperature, MAP and [NL] explained 23% of the variation in foliar δ15N for potentially N2-fixing plants, with about half of the total variation explained by those three factors for the larger dataset. For these plants, foliar δ15N increased with increasing MAT at approximately the same rate as non-N2-fixing plants from ecosystems with MAT ≥ −0.5°C (0.24 vs 0.23°C−1; Fig. 5a), while MAP and foliar δ15N were unrelated (P > 0.92; Fig. 5b). Increasing [NL] an order of magnitude for potentially N2-fixing species increased foliar δ15N by 5.8‰ (Fig. 5c).
Foliar δ15N and the difference in δ15N between leaf and soil were positively correlated (Fig. 6), reinforcing the idea that the general patterns of N availability can be assessed with foliar δ15N alone. When averaged for each site, soil δ15N correlates with foliar δ15N, with foliar δ15N increasing at a faster rate than soil δ15N as soil δ15N increases (Fig. 6a). As such, foliar δ15N was positively correlated with the difference between the δ15N of leaves and soil (Fig. 6b).
Foliar δ15N also increased with increasing N supply in 13 of 15 multi-site studies of unmanaged ecosystems (Fig. 7, Table 3), averaging a 4.9‰ increase in foliar δ15N across the range of N availability in each study. For each of 13 studies, foliar δ15N increased monotonically with N availability regardless of whether it was measured as in situ mineralization, ex situ potential net mineralization or in situ mineralization from resin bags. In the two studies that did not fit the trend, one examined herbaceous understory plants in Swedish forests (Falkengren-Grerup et al., 2004). In this study, foliar δ15N increased with increasing potential N mineralization for the nine sites with the lowest mineralization rates, but declined for the three sites with the highest mineralization rates. In one grassland study where N availability was associated more with differences in grazing than supply (Craine et al., 2009), potential N mineralization (PotNmin) explained little variation in foliar δ15N (δ15NL = 3.53 – 0.52 * PotNmin, P = 0.03, r2 = 0.01).
Table 3. Details on the studies of N availability and foliar δ15N presented in Fig. 3
The analyses presented here provide new insights into global patterns of foliar δ15N and the N cycle. First, this study is the first review to incorporate mycorrhizal fungi into global-scale analyses of foliar δ15N, thereby explicitly addressing a potential covariate influencing relationships between foliar δ15N and climate. Second, this review showed different relationships between temperature and foliar δ15N for cold and warm ecosystems. Third, the positive relationships reported here between [NL] and foliar δ15N at the global scale across a number of ecosystem types extends the generality of previous relationships quantified on smaller scales. Fourth, although relationships between foliar δ15N and climate are not biased by excluding potentially N2-fixing plants, MAP and foliar δ15N were uncorrelated for potentially N2-fixing plants. Lastly, foliar δ15N consistently increases along natural N supply gradients from landscape to regional scales.
In past reports of global patterns of foliar δ15N, the ability of mycorrhizal fungi to deliver 15N-depleted N to plants has been acknowledged (Amundson et al., 2003) but had not been explicitly incorporated in analyses that determined relationships between climate and foliar δ15N. The effect of mycorrhizal type on foliar δ15N observed in this study extends the patterns of foliar δ15N dependence upon mycorrhizal associations previously observed in tundra, boreal, alpine and heath ecosystems (Michelsen et al., 1998; Schmidt & Stewart, 2003) to the global scale. The differences among plants with different mycorrhizal fungi could be influenced by differences in sources of N among types (Smith & Read, 2008). Yet, these differences in foliar δ15N among plants with different mycorrhizal fungi associations are broadly congruent with the degree of plant reliance on the fungi for N – ericoid plants are more reliant on mycorrhizal fungi for N than ectomycorrhizal plants than arbuscular plants (Hobbie et al., 2005; Jin et al., 2005). The lower average foliar δ15N in arbuscular mycorrhizal plants compared with nonmycorrhizal plants suggests that N transfer from arbuscular mycorrhizal fungi to host plants favors 15N-depleted N in a similar fashion to the well-established transfer pathways in ectomycorrhizal symbioses. In addition, the absence of proteolytic capabilities in arbuscular mycorrhizal fungi (Chalot & Brun, 1998) suggests that the potential N sources for nonmycorrhizal and arbuscular mycorrhizal plants should be similar. If true, then 15N depletion in arbuscular mycorrhizal plants relative to co-occurring nonmycorrhizal plants may reflect preferential transfer by arbuscular mycorrhizal fungi of 15N-depleted compounds. Recent experiments support arbuscular mycorrhizal fungi transferring N to plants (Leigh et al., 2009), but culture studies investigating the effects of arbuscular mycorrhizal colonization on plant δ15N have been inconclusive (Hobbie & Hobbie, 2008).
The relationships between MAT and foliar δ15N for ecosystems with MAT ≥ −0.5°C were similar to previous results (Martinelli et al., 1999; Amundson et al., 2003), but with a greater slope than was found for temperate forests alone (Pardo et al., 2006). In contrast to sites with MAT ≥ −0.5°C, foliar δ15N did not change with increasing MAT across sites with MAT < −0.5°C. Two facts are important in interpreting the patterns of foliar δ15N for cold sites. First, soil δ15N values in cold regions are typically close to those of atmospheric N2 (0‰) (Amundson et al., 2003). Second, dissolved organic N (DON) likely dominates the cycling of N in these soils (Neff et al., 2003) and losses of N in ecosystems with MAT of < −0.5°C are likely to be dominated by DON. With little fractionation during solubilization, DON loss should not lead to enrichment of the available N pool. As such, the lack of increase in δ15N with MAT below −0.5°C implies that the signature of N being lost from the ecosystem or the dependence of plants on mycorrhizal fungi for N does not change with temperature over this range.
As observed before (Handley et al., 1999a; Schuur & Matson, 2001; Amundson et al., 2003), foliar δ15N declined with increasing MAP for nonN2-fixing plants. In addition to finding a better fit between log-transformed MAP and foliar δ15N than untransformed MAP, and extending relationships to higher precipitation sites than before, there is little qualitative difference between our results here regarding relationships with MAP and earlier studies. That said, more research is necessary to understand whether the high foliar δ15N can be directly linked to high gaseous N loss, lower dependence on mycorrhizal fungi and/or changes in cycling of N within the plant. For example, for nonN2-fixing plants, high foliar δ15N at low precipitation suggests that these sites might experience relatively high rates of gaseous N loss through ammonium volatilization, gaseous N loss during nitrification or denitrification. Although no study has yet compared gaseous N loss rates across a precipitation gradient, denitrification rates can be high in xeric ecosystems (Peterjohn & Schlesinger, 1991) and are higher for tropical forests than temperate forests (Stehfest & Bouwman, 2006). Although requiring anaerobic microsites, denitrification is often limited by nitrate availability (Groffman et al., 1993), which could explain why dry sites that might have soils that are less anaerobic can have greater denitrification. The lack of elevated foliar δ15N at high precipitation might reflect strong N limitation and low N availability (Schuur & Matson, 2001), but could also reflect complete denitrification of soil nitrate pools, which does not leave behind 15N-enriched substrate (Houlton et al., 2006), or consistently high reliance by plants on mycorrhizal acquisition of N. Future research on the dual controls of nitrification and denitrification across precipitation gradients on the δ15N of available N could provide new insights into patterns of foliar δ15N with increasing MAP.
At the local to regional scale, foliar δ15N consistently increased with increasing N supply. The one study that did not show a monotonically positive relationship between foliar δ15N and N availability (Falkengren-Grerup et al., 2004) only measured understory species, and the patterns observed may not have represented stand-level N dynamics. As such, understory plants in sites with high rates of mineralization might have relied more on nitrate than sites with lower mineralization, while other singular explanations such as signatures of N deposition could also explain this anomaly. With multiple ways to quantify aspects of the supply of N to plants, how best to represent N supply, much less N availability to plants, is still an open question. For example, in some cases Pardo et al. (2006) found better relationships for temperate forests between foliar δ15N and the fraction of mineralized N that was nitrified than laboratory or field measures of mineralization rate. Also, soil N supply likely has to be coupled with plant N demand to best represent N availability. In Craine et al. (2009), variation in N availability among 330 South African grassland sites was driven more by variation in plant demand than supply. Sites with high foliar δ15N did not necessarily have higher N supplies, but instead were associated with higher grazing intensity. Sodic sites and private protected areas known to have high herbivore densities had grasses with high foliar δ15N. These sites were associated with lower grass biomass and species that tend to increase with grazing pressure.
Separating proximal and distal drivers of variation in δ15N is a long-standing question that ultimately cannot be answered without additional data and/or modeling. That said, the patterns that we are attempting to explain span approx. 5‰ (MAT), 4‰ (MAP), 20‰ ([NL]) and 6‰ (types of mycorrhizal fungi). Although still a valid hypothesis, there is currently no evidence that within-plant fractionations change markedly along these gradients. For example, for grasses collected from a wide range of grasslands (Craine et al., 2005), the difference in δ15N between leaves and roots changed by only 1‰ across an order of magnitude of foliar N concentrations (J. Craine, unpublished). With regard to other aspects of within-plant fractionation, often no discrimination is observed during resorption of N from leaves (Kolb & Evans, 2002), while differences between leaves and stems are often c. 1‰ (Gebauer & Schulze, 1991; Hobbie et al., 2008). Differences in signatures of depositional N might add variation to our dataset, but do not appear to vary consistently with climate at regional scales (Bragazza et al., 2005; Elliott et al., 2007). Differences in plant preference for forms of N can explain variations within a site, but not stand-level differences in signatures. Other distal factors might be controlled by climate, but it is unknown whether climate directly affects the N cycle or does so indirectly by altering patterns of disturbance. For example, there are unresolved debates about the relative importance in xeric grasslands of low precipitation per se or greater grazing in enriching plants in 15N (Schulze et al., 1998; Austin & Sala, 1999; Cook, 2001).
If regional-scale variation in N supply is generally tied to foliar δ15N, one cannot necessarily assume that global-scale patterns in foliar δ15N or differences between any small number of sites within a region can be interpreted as representing N availability. Yet, if the general regional relationships between N availability and foliar δ15N extend to the global scale, then the global foliar δ15N patterns suggest that: warm and dry sites have high N availability relative to cold and wet sites; plants with high N concentrations occupy sites with high N supply; and N availability might be relatively high in sites with low P availability. Whether this greater N availability also translates to less N limitation to plant production is uncertain. Foliar δ15N increased with increasing foliar N : P ratio, regardless of whether a result of increasing N concentrations or decreasing P concentrations. Although foliar N : P ratio has been considered to index the relative limitation of N and P to plants in terrestrial ecosystems (Güsewell, 2004), at least for grasslands this is not necessarily true (Craine et al., 2008).
With important questions still remaining about the patterns of foliar δ15N, future research should turn to understanding the potential underlying determinants of plant δ15N that can be used to interpret the global foliar δ15N patterns. Because multiple levels of causation may exist for a given pattern, future research should begin to identify what underlying processes are correlated with signatures. More comparative data at the global scale are needed to evaluate the potential drivers of foliar δ15N with climate or increasing [NL], such as data on the signatures of available N (Houlton et al., 2006; Kahmen et al., 2008). Measurements of the biomass or colonization rates of mycorrhizal fungi and their isotopic signatures would assist in testing the importance of changes in the reliance of plants on mycorrhizal fungi, while measurements of root δ15N would evaluate whether within-plant fractionation is responsible for any of the changes in foliar δ15N seen across gradients. The patterns laid out here will help constrain the potential mechanisms that underlie changes in N cycling along ecological gradients, such as the directional change in the reliance of plants on mycorrhizal fungi or the signature of N lost from ecosystems. Knowing the patterns of foliar δ15N at the global scale might help constrain models of gaseous N loss, while there is now real potential to routinely use plant δ15N to reconstruct past N availability as well as to monitor changing soil N availability.
J.M.C. was supported by the Andrew W. Mellon Foundation and A.E. was supported by the Henry Luce Foundation. We extend our appreciation to the those that provided published and unpublished data for this review: Mary Beth Adams, Sharon Billings, John Campbell, Scott Chang, Jacqui Codron, Daryl Codron, Corli Coetsee, Jana Compton, Garry Cook, Paul Dijkstra, Susana Echeverria-Rodriguez, Ursula Falkengren-Grerup, Joe Fargione, Arthur Gessler, Christine Goodale, Paul Grogan, Per Gundersen, Heidi Hawkins, Peter Hietz, Will Hoffman, Lars Högbom, Ben Houlton, Keisuke Koba, Peter Leavitt, Mikhail Makarov, Steve McNulty, Amy Miller, Lars-Ola Nilsson, Rafael Oliveira, Scott Ollinger, Jean Ometto, Heather Rueth, Louis Santiago, Fabio Scarano, Patrick Schleppi, Ernst-Detlef Schulze, Sah Shambhu, Keirith Snyder, Ryunosuke Tateno, and Peter Vitousek. Peter Groffman, Ben Houlton, Jeb Barrett. We also thank the anonymous reviewers who provided comments and helpful discussion on the manuscript.