Declining foliar and litter δ15N diverge from soil, epiphyte and input δ15N along a 120 000 yr temperate rainforest chronosequence


Author for correspondence:
Duncan N. L. Menge
Tel: +1 609 571 7773


  • Patterns in the natural abundance of nitrogen (N) isotopes (15N and 14N) can help in the understanding of ecosystem processes along environmental gradients, because some processes fractionate against the heavier isotope. We measured δ15N in many components of the Franz Josef soil chronosequence in New Zealand to see how each component varied along the sequence and within sites, and to see what this variation can tell us about how ecosystem processes such as N losses change with soil age.
  • We analyzed δ15N in foliage from 18 woody species, abscised leaves from seven woody species, three soil horizons, bryophytes, lichens, bulk deposition, and nodules from the N-fixing tree Coriaria arborea (Coriariaceae).
  • Foliar δ15N varied significantly across plant species. Foliage and bulk litter became 15N-depleted as soil age increased. Soil N from organic and mineral horizons was significantly more 15N-enriched than bulk litter N at each site. Increasing precipitation also decreased foliar and soil δ15N.
  • Comparing input and whole ecosystem δ15N revealed limited evidence for net fractionation during N losses. These trends are consistent with some combination of increasing fractionation during plant N uptake, mycorrhizal transfer, within-plant processing, and soil decomposition as soils age.


Nitrogen (N) availability has strong impacts on community (Tilman, 1987; Zavaleta et al., 2003) and ecosystem (Vitousek & Howarth, 1991; Elser et al., 2007; LeBauer & Treseder, 2008) processes in many systems. Yet despite its importance, ecosystem-scale N cycling is difficult to study because of the number and complexity of ecosystem N pools and fluxes. Increasingly, natural abundance stable N isotope ratios have been used to study N dynamics (Houlton et al., 2006, 2007; Craine et al., 2009). Many ecosystem processes fractionate against heavier isotopes, resulting in 15N-enriched source pools and 15N-depleted sink pools. Therefore, natural abundance N isotopes provide a valuable integrative record of N dynamics (Högberg, 1997; Robinson, 2001).

The degree of fractionation is better known for some N transformations than for others (Dawson et al., 2002; Tcherkez & Hodges, 2008). For example, biological N fixation (Dawson et al., 2002) and nitrate leaching (Houlton et al., 2006) fractionate little, but ammonium volatilization fractionates strongly (Dawson et al., 2002). However, the degree of intrinsic fractionation during some processes is unclear because the observed fractionation depends on the degree of completion of the reaction as well as enzymatic discrimination. If a discriminating enzyme consumes all of its substrate, there is no net fractionation (Dawson et al., 2002). For example, plant uptake of both nitrate and ammonium fractionates when available N concentrations are high or N is not limiting, but not when available N concentrations are low or N is limiting (Högberg, 1997; McKee et al., 2002; Clarkson et al., 2005).

One use of natural abundance N isotopes is the study of N cycling along environmental gradients such as long-term soil chronosequences. On very young soils, soil N is scarce (Chapin et al., 2002), N fertilization frequently stimulates net primary production (NPP; Vitousek et al., 1993), and symbiotic N fixers are relatively common (e.g. Walker, 1993) and actively fixing dinitrogen gas (e.g. Vitousek & Walker, 1989). By contrast, very old, undisturbed soils are thought to be primarily phosphorus (P)-limited because the major P input – rock weathering – decreases over time, whereas N inputs from atmospheric deposition and some N fixers continue indefinitely (Beadle, 1966; Walker & Syers, 1976; Vitousek & Farrington, 1997; Vitousek, 2004; Wardle et al., 2004). Within the range of P-limited soils, N dynamics can vary. Many old tropical forests are N-replete (Hedin et al., 2009), whereas older temperate forests are often N-poor (Vitousek & Howarth, 1991; Hedin et al., 1995; Perakis & Hedin, 2002), which can have implications for N isotope patterns.

To our knowledge, the patterns of natural abundance N isotopes in soils and plants have been investigated in five chronosequences in three different biomes: boreal forests in Alaska aged 55–225 yr (Hobbie et al., 1999a), boreal forests in Sweden aged hundreds to thousands of years (Hyodo & Wardle, 2009), tropical rainforests in Hawaii aged 28–67 000 yr (Vitousek et al., 1989) and 300–4 100 000 yr (Martinelli et al., 1999), and temperate annual grasslands in California aged 3000–3 000 000 yr (Brenner et al., 2001). Along both Hawaiian chronosequences, the older of which is known to switch from N limitation to co-limitation to P limitation (Vitousek & Farrington, 1997), both foliar and soil δ15N (see the Materials and Methods section for δ15N definition) increased with soil age (Vitousek et al., 1989; Martinelli et al., 1999), likely as the result of an increase in losses that fractionate against 15N across the chronosequence (Hedin et al., 2003). Foliar and litter δ15N increased with forest age in Sweden, likely because of increased N fixation by free-living N-fixers, increased reliance on dissolved organic N, or both (Hyodo & Wardle, 2009). By contrast, foliar δ15N decreased through early succession in Alaska (Hobbie et al., 1999a), likely because of a shift in the importance of fractionating mycorrhizal N transfer to plants (Hobbie et al., 1999b, 2000), although soil δ15N showed no consistent pattern (Hobbie et al., 1999a). On the California grassland chronosequence, foliar δ15N showed no trend with soil age and soil δ15N increased slightly (Brenner et al., 2001). These studies demonstrate that long-term variation in natural isotope abundance differs among systems and is driven by a range of processes.

The Franz Josef chronosequence on the west coast of New Zealand’s South Island is a well-studied chronosequence in temperate forest, a biome for which no chronosequence N isotope data have been reported to our knowledge. Another novel aspect of the Franz Josef chronosequence is that it spans the full range of ecosystem development stages (Peltzer et al., 2010), including young progressive sites (< 10 yr), maximal biomass sites, and relatively old retrogressive sites (120 000 yr) (Richardson et al., 2004). Total soil N : P and nonfixer foliar N : P generally increase with soil age (Stevens, 1968; Richardson et al., 2004), consistent with a switch from N to P limitation. However, soil available N and foliar N decline in the older sites (Richardson et al., 2004, 2005), suggesting that the older sites are N-poor even if they are not N-limited (see also Table 1).

Table 1.   Site characteristics and δ15N measurement species list
 Age (yr)
 5 (site 1)60 (site 2)130 (site 3)500 (site 5)12 000 (site 7)120 000 (site 9)
  1. Site numbers correspond to Richardson et al. (2004). Soil N : P is from the upper 10 cm of the mineral horizon. Foliar N : P is dominance-weighted. Bulk N deposition for 60–500 yr sites was assigned the same value as the 5 yr site. N fixation is from Coriaria arborea, bryophytes, lichens and bulk litter. Amax is the mean maximum photosynthetic rate of species indicated by *s. Dominance weightings were calculated from cover scores (see the Materials and Methods section).

  2. 1From Richardson et al. (2004); 2from Menge & Hedin (2009); 3from Whitehead et al. (2005).

  3. Species designations (superscripts): a, angiosperm; c, conifer; d, deciduous; e, evergreen; Nf, N fixer.

Annual precipitation1 (mm)652065766188627837063652
Soil pH16.
Total soil N : P1 (g g−1)
Foliar N : P1 (g g−1)16.815.38.29.813.916.1
Nonfixer foliar N : P1 (g g−1)8.414.78.29.813.916.1
Maximum tree height1 (m)
Bulk N deposition2 (kg ha−1 yr−1)
N fixation2 (kg ha−1 yr−1)
Amax3 (μmol m−2 s−1)19.616.312.
Dominance weighting1 (%)
 Coriaria arboreaNf,e,a89*25*    
 Olearia avicenniifoliae,a9*     
 Aristotelia serratad,a2*21*    
 Schefflera digitatae,a 46    
 Melicytus ramifloruse,a 8*1*   
 Griselinia littoralise,a < 147*2*  
 Coprosma lucidae,a  28   
 Pseudopanax colensoie,a  11   
 Coprosma ciliatae,a  2   
 Carpodetus serratuse,a  1   
 Weinmannia racemosae,a  10*20*30*13*
 Prumnopitys ferrugineae,c   12  
 Metrosideros umbellatae,a   66*< 1*3*
 Dacrydium cupressinume,c    44*14*
 Quintinia acutifoliae,a    2614
 Manoao colensoie,c     15
 Phyllocladus alpinuse,c     40
 Podocarpus halliie,c     1
% community occupied by sampled species949381686980

In this study, we examined natural abundance N isotopes across the Franz Josef chronosequence. Specifically, we report δ15N in foliage and abscised leaves; mineral, organic, and bulk litter soil horizons; lichens and bryophytes; and bulk deposition and actinorhizal root nodules, the last of which is an indication of symbiotic N fixation N inputs. We used these data to resolve two interlinked questions. First, how do each of these pools vary with soil age? And second, how do these pools compare within each site? These data will fill a gap in our knowledge of N isotope patterns along temperate forest chronosequences. Additionally, they will help us to determine the generality of N isotope patterns observed along chronosequences in other biomes.

Materials and Methods

Site description

The Franz Josef chronosequence (43°25′S 170°10′E) is composed of a series of schist outwash surfaces varying in soil age from 0 to > 120 000 yr, formed by repeated glacial advance and retreat over a narrow coastal strip in South Westland, New Zealand (Stevens, 1968; Walker & Syers, 1976; Almond et al., 2001). The climate is wet temperate, with a mean annual temperature at the valley mouth of 10.8°C (Hessell, 1982), and mean annual precipitation (MAP) that ranges from up to 6.5 m near the glacier (surfaces < 12 000 yr) to c. 3.5 m on the coastal plains (≥ 12 000 yr) (Table 1). Although rainfall estimates at the sites vary within uncertainty bounds between estimates derived from climate surfaces (Richardson et al., 2004) and meteorological studies (Henderson & Thompson, 1999), all estimates indicate a difference between the younger and older sites of up to 3 m, and precipitation measurements nearby suggest that such extreme gradients are plausible (Griffiths & McSaveney, 1983). Foliar and soil δ15N correlate negatively with MAP across a number of scales (Handley et al., 1999; Schuur & Matson, 2001; Amundson et al., 2003). Thus, to ensure a conservative test of effects of site age on δ15N, we included the more variable MAP estimates (Richardson et al., 2004) as an independent variable along with site age and sample type. We also note that surfaces younger than the last glacial period (≤ 12 000 yr) developed entirely in this climate, whereas older surfaces (> 12 000 yr) experienced cooler temperatures for a large part of their development (Moar & Suggate, 1996) and received atmospheric deposition of loess (Almond, 1996; Almond & Tonkin, 1999).

Clear patterns of forest succession and retrogression are apparent through changes in species composition, tree height and vegetation cover. These forest characteristics are related to strong patterns of plant and soil chemistry that are consistent with a shift from N limitation to P limitation along the chronosequence (Richardson et al., 2004). For instance, N : P in total soil and in live foliage of nonfixing plants – which do not have direct access to atmospheric N2– increase with site age (Table 1). We selected six sites that ranged in age from 5 to 120 000 yr. These correspond to sites 1, 2, 3, 5, 7, and 9 from Richardson et al. (2004); throughout we use the site numbers from Richardson et al. (2004). The sites encompass the full range of ecosystem development stages along the chronosequence, from the progressive phase to the maximal biomass phase to the retrogressive phase (Peltzer et al., 2010). Soil and plant properties of these sites are summarized in Table 1.

Sample collection and processing

New, fully expanded, sunlit canopy leaves were sampled from three trees of the dominant species at each site in March 2002 using orchard cutters for short canopies or a shotgun for taller canopies. We hoped to sample all important species at all sites, but each species only occupies a subset of the sites (Table 1). Therefore, our sampling reflects the best possible coverage of dominant species at each site, and there is an overlap of at least two species between each pair of consecutive sites. Eight species were sampled at more than one site, including some from sites aged 5000 and 60 000 yr, corresponding to sites 6 and 8 in Richardson et al. (2004), and an additional 10 species were sampled at a single site each (Table 1). In addition, freshly abscised leaves were collected in March 2002 from the seven most common species (from underneath the same individuals used for canopy sampling) at sites where they formed a significant part of the canopy. We consider freshly abscised leaves – selected by species from the Oi horizon for comparison with fresh foliage – separately from bulk litter (the entire Oi horizon; see following paragraph). New canopy leaves from eight common species were resampled in January 2003 to test for interannual differences in N isotopic composition. Samples were dried at 70°C to constant mass and ground to a fine powder.

Soil samples from the upper 10 cm were collected in January 2002 from five subplots per site and divided into organic soil (Oa horizon, present only at the three oldest sites) and mineral soil (A horizon or below) (Richardson et al., 2004). Bulk litter (Oi horizon) was sampled in January 2003. Samples were moist-sieved (4 mm), dried at 105°C to constant mass, and ground for analysis.

Samples of Coriaria root nodules, bryophytes, and lichens (the latter two from both ground and epiphytic surfaces) were collected along randomly spaced, randomly oriented transects in a subset of sites (Menge & Hedin, 2009). Bryophytes included Ptychomnium aciculare (sites 5 and 9) and Hypnodendron spp.; lichen species included Pseudocyphellaria cinnamomea (sites 1, 3, and 5), Pseudocyphellaria homoeophylla (sites 3 and 5), and other Pseudocyphellaria and Sticta spp. Samples were dried at 60°C to constant mass and ground for analysis. Drying at 60 vs 105°C can alter δ15N values by up to 0.5‰ relative to each other and 1‰ relative to air-dried controls (Brearley, 2009), so comparisons across sample types (foliage vs soil vs other) should take this into account.

Event-based bulk deposition was sampled from July to December 2004 in open areas near sites 1, 7 and 9 using high-density polyethylene (HDPE) funnels 2 m off the ground connected to dark HDPE bottles on the ground (Menge & Hedin, 2009). Sites 2–5 are within 1.5 km of site one, so three collectors were sufficient. Samples were filtered (0.45 μm) immediately and frozen from ≤ 6 h after collection until they were chemically and isotopically analyzed. Our sampling design does not allow us to capture seasonal or within-canopy variation in δ15N. However, our main trends are much larger than observed seasonal changes in other systems – typically 1–2‰ at most (Ometto et al., 2006; Coletta et al., 2009; Bragazza et al., 2010) – and within-canopy variation reported elsewhere (Bergstrom & Tweedie, 1998) is negligible.

Stable isotope analysis

Leaf and soil samples were analyzed for N content and isotopes at the Waikato Stable Isotope Unit in Waikato, New Zealand, using a Europa Scientific 20/20 (Europa Scientific 20/20, Crewe, UK) isotope ratio mass spectrometer (IRMS). Coriaria nodule, bryophyte, and lichen samples were analyzed for N content and isotopes at the Boston University Stable Isotope Laboratory in Boston, MA, USA, using a GV Instruments Isoprime IRMS (GV Instruments, Manchester, UK). The N isotopic composition of bulk deposition TDN was analyzed in the Sigman laboratory at Princeton University (Princeton, NJ, USA), using a modified Finnigan GasBench (Thermo Scientific, Pittsburgh, PA, USA) and DeltaPlus IRMS (Thermo Scientific, Pittsburgh, PA, USA), following the denitrifier method (Sigman, 2001) and persulfate oxidation (Knapp et al., 2005). All machines were calibrated to international standards. All isotope ratios are expressed as δ15N relative to the atmospheric N2 standard: δ15N = ((15N/14N)sample/(15N/14N)standard− 1) × 1000, in ‰ units.

Canopy species dominance weighting

The proportion of the upper sunlit canopy occupied by the species sampled at each site was estimated from visual cover assessments for each species made by Richardson et al. (2004). Average cover scores from five subplots of 5 m radius were combined with average maximum tree height for each species to give average cover scores for all species present within the upper third of the maximum tree height for the site. At each site the species sampled formed > 68% of the total cover (Table 1) and > 90% of the upper canopy cover (data not shown). The dominance weighting for each species sampled was then calculated from cover scores as the proportion of total sampled species cover. Dominance weighting values (Table 1) are used to estimate weighted mean foliar δ15N at each site.

Statistical analyses

All statistical analyses were conducted in R 2.8 (R Development Core Team, 2009), and, unless otherwise stated, were linear regression models performed using the lm function. Two-tailed tests were used throughout, and post-hoc comparisons used Tukey’s HSD.


Foliar and abscised leaf δ15N

Foliar δ15N (δ15Nf) ranged from very 15N-depleted (−14‰) to slightly 15N-enriched (4‰). Overall, site age and species explained 83% of the variation in the foliar δ15N (R2 = 0.83, < 0.0001). Including MAP in the model increased the R2 to 0.86, but the MAP main effect was not significant (= 0.58) (see Fig. 1 caption for significance of all effects and interactions). Individual species tended to become more 15N-depleted with soil age, especially in sites of intermediate to old age (Fig. 1). δ15Nf of Melicytus ramiflorus (df = 14; = 0.004), Dacrydium cupressinum (df = 29; < 0.001), Metrosideros umbellata (df = 20; = 0.006), Weinmannia racemosa (df = 46; < 0.0001), and Quintinia acutifolia (df = 4; = 0.008) all decreased with soil age, whereas Coriaria arborea (df = 14; = 0.64) and Griselinia littoralis (df = 17; P = 0.33) showed no trend, and Aristotelia serrata increased (df = 14; = 0.02). D. cupressinum (= 0.02) and M. umbellata (= 0.02) δ15Nf were elevated at lower rainfall, whereas W. racemosaδ15Nf was not (> 0.05). However, the MAP × age interaction was significant for W. racemosa (= 0.007), with a sharper decrease with age at lower rainfall than at higher rainfall (Fig. 1).

Figure 1.

Foliar δ15N across the Franz Josef chronosequence. Symbols represent species mean ± SE at each site. Some species at certain sites are displayed with slightly altered ages for readability, but come from the same sites listed in the Materials and Methods section. Note the logarithmic scale of the horizontal axis. Effects and significance of the overall foliar model were A****, S****, P ns, A × S****, A × P***, S × P*** and A × S × P ns, where A is log site age, S is species, P is mean annual precipitation, ‘×’ indicates the interaction term, and the significance codes are: ns, > 0.05; *, < 0.05; **, < 0.01; ***, < 0.001; ****, < 0.0001. The species effect was significant within each site (< 0.001 for ANOVAs at sites aged 5, 60, 130, 500, 5000, 12 000, 60 000, and 120 000 yr, with df within, between species = 21, 2; 25, 4; 27, 6; 23, 3; 7, 1; 31, 3; 6, 1; 36, 6). Letters denote significant (< 0.05) species differences within each site, and thus have no meaning across sites.

Within each site, δ15Nf of individual species tended to differ (Fig. 1). Species differences within each site were also consistent across sites (Fig. 1). At each site where they coexisted, the symbiotic N fixer Coriaria arborea was more 15N-enriched than A. serrata; M. ramiflorus was more 15N-enriched than G. littoralis; M. umbellata was more 15N-enriched than W. racemosa; and D. cupressinum was more 15N-enriched than M. umbellata, W. racemosa and Q. acutifolia. The sole exception was that G. littoralis was more 15N-enriched than W. racemosa at the 500 yr site but they did not differ statistically at the 130 yr site.

There was a strong correlation between δ15Nf and abscised leaf δ15N within individual trees from 2002, and the fit was very close to the 1 : 1 line (Fig. 2a). Within individual species, there was some variation between 2002 and 2003 foliage (Fig. 2b).

Figure 2.

δ15N variation during abscission and from year to year. (a) Foliar and abscised δ15N from the 2002 sampling were highly correlated (df = 45, r2 = 0.94, < 0.001). Each point represents a single tree, with species indicated by symbol shape. The line plotted is the 1 : 1 line, which is statistically indistinguishable from the fit (slope ± SE = 0.95 ± 0.04, intercept ± SE = −0.15 ± 0.23). Divergence below this line would indicate fractionation during N resorption, for which there is no strong evidence. (b) Foliar δ15N in 2002 (horizontal axis) vs 2003 (vertical axis). The species means at each site were well correlated (df = 16, r2 = 0.65, < 0.001), but the fit diverged from the 1 : 1 line (slope ± SE = 0.75 ± 0.13, intercept ± SE = −0.47 ± 0.79). Each symbol represents a species mean ± SE. Species are the same as in Fig. 1.

Dominance-weighted foliar δ15N

Dominance-weighted δ15Nf decreased from −0.70‰ at site 1 to −8.45‰ at site 9 (Fig. 3a). Dominance-weighted δ15Nf (df = 189) decreased with site age (< 0.0001) and MAP (< 0.001), with a significant interaction between the two (= 0.008). For the model with MAP, R2 = 0.38, whereas R2 = 0.31 for the model without MAP (< 0.0001 for both).

Figure 3.

δ15N patterns across the Franz Josef chronosequence. (a) Mineral soil, organic soil, and bulk litter mean ± SE are shown by open symbols, and the dominance-weighted foliar mean is shown by closed circles. Effects and significance on soils (excluding foliar means, which are not soil) were A**, H****, P****, A × H**, A × P**, H × P ns, and A × H × P ns, where A is log site age, H is the soil horizon, P is mean annual precipitation, ‘×’ indicates the interaction term, and the significance codes are: ns, > 0.05; *, < 0.05; **, < 0.01; ***, < 0.001; ****, < 0.0001. Letters denote significant (< 0.05) horizon differences within each site following ANOVAs (< 0.0001 at each site) with df within, between soil horizons = 8, 1; 8, 1; 8, 1; 9, 2; 11, 2; and 9, 2 for sites aged 5, 60, 130, 500, 12 000, and 120 000 yr. (b) Coriaria arborea nodule (A ns), bulk deposition (A ns, P ns), bryophyte (A ns, P ns, A × P ns), and lichen (A ns, P****, A × P ns) means and SE are shown as a function of soil age. Note the logarithmic scale of the horizontal axis of both panels.

Soil δ15N

In our statistical analysis, soil δ15N decreased with site age and MAP and varied with soil horizon, with significant age × horizon and age × MAP interactions (Fig. 3a; R2 = 0.84 and 0.73 with and without including MAP; < 0.0001 for both). Soil from the mineral horizon was 15N-enriched (range −0.02–4.10‰) relative to the atmosphere (df = 27; = 0.002). Soil horizons differed from each other within each site. As with plant species, individual soil horizon differences within sites were fairly consistent across sites. Soil from both the mineral (range −0.02‰–4.10‰) and organic (−2.28‰–1.17‰) horizons were more 15N-enriched than bulk litter (−7.11‰ to −0.30‰) at each site, and the mineral horizon was more 15N-enriched than the organic horizon at one site (Fig. 3a). Mineral soil δ15N (df = 25) had a strong interaction between MAP and site age (< 0.001), decreasing with site age at high precipitation but not changing at low precipitation; main effects showed a strong decrease at elevated MAP (< 0.0001) but no change with site age (= 0.40) (Fig. 3a). Organic soil δ15N (df = 6; = 0.64) showed no consistent trends with soil age (= 0.64) or MAP (= 0.49). By contrast, bulk litter δ15N (df = 26) decreased with both soil age (< 0.001) and MAP (= 0.04), similar to dominance-weighted plant δ15N (Fig. 3a) but without a significant interaction (= 0.54).

N input δ15N

Coriaria arborea nodule δ15N ranged from −2.30‰ to 1.51‰, and bulk deposition δ15N ranged from −2.40‰ to 2.90‰ (Fig. 3b). The mean of the two was −0.47‰. There were no significant differences between the two (= 0.78) or effects of site age (= 0.34) or MAP (= 0.75). The full model, with R2 = 0.13, was not significant (= 0.56). Excluding the top two and bottom two points gives a range of −2.3‰ to 1.5‰, which we use as our reasonable input range.

Bryophyte and lichen δ15N

Bryophytes were the most 15N-enriched pool (δ15N range 0.78‰–5.96‰), and did not change with soil age (df = 6; = 0.17) or MAP (= 0.67) (Fig. 3b). Lichens ranged around the atmospheric standard (−3.00‰–2.13‰), did not change with soil age (df = 17; = 0.85), but decreased with MAP (< 0.0001) (Fig. 3b). For the full lichen δ15N model, R2 = 0.72 and < 0.0001.


Precipitation effects

Increasing MAP lowers the δ15N of dominance-weighted foliage, foliage of two of three species, mineral soil, bulk litter, and lichens across the Franz Josef chronosequence, agreeing with published negative correlations between MAP and foliar and soil δ15N across a range of scales (Austin & Sala, 1999; Handley et al., 1999; Schuur & Matson, 2001; Amundson et al., 2003; Craine et al., 2009; Liu & Wang, 2010). Because our older sites (≥ 12 000 yr) have lower MAP (3500 mm) than our younger sites (6000–6500 mm; Table 1), the trends with site age are even stronger than they appear, because the precipitation effect shifts δ15N up at the older sites.

It is unlikely that water limits NPP at Franz Josef, but periods of soil anoxia are possible. The δ15N decrease along a rainfall gradient in Maui (Schuur & Matson, 2001) has been attributed to an increasing degree of completion of denitrification resulting from such anoxia (Houlton et al., 2006), such that the fractionation is expressed more in the less wet (2000–3500 mm yr−1) sites than in the wetter (4000–5000 mm yr−1) sites where nearly all nitrate is consumed. At Franz Josef the evidence for net fractionation during N losses is weak (see following section) so this mechanism has limited support, although we cannot rule it out. Another potential explanation for the change with MAP – changes in the δ15N of N inputs (Heaton, 1987) – seems unlikely given that δ15N of our bulk deposition inputs (Fig. 3b) does not change with precipitation.

Little evidence for fractionation during N losses

Fractionation during N losses is thought to play a major role in N isotope patterns in some ecosystems (Martinelli et al., 1999; Houlton et al., 2006; Hobbie & Ouimette, 2009), and strongly affects our expectations for our other patterns, so we derived a calculation to estimate net N loss fractionation with our data. We estimated the δ15N of the entire ecosystem – essentially the weighted average of plant δ15N and soil δ15N – using foliar and soil data from Franz Josef, published corrections for soil depth and within-plant δ15N differences, and a range of ratios of soil N to plant N (see Supporting Information, Methods S1 for details). The δ15N of the entire ecosystem is equal to the δ15N of inputs modified by net N loss fractionation during the history of the site, so we compared our ecosystem-level δ15N estimate with our measured input δ15N range to estimate net N loss fractionation.

Our reasonable range of net input δ15N (−2.3‰–1.5‰) is similar to other findings (Handley et al., 1999; Houlton et al., 2006). Total plant δ15N is quite negative at our older sites, whereas weighted total soil N is somewhat 15N-enriched. Based on our best guesses for soil N to plant N ratios, ecosystem δ15N is solidly within the input range until site 7, at which point it is slightly outside (0.6‰) the input range (Fig. 4). Therefore, based on our calculation, the evidence for net N loss fractionation is weak, although it is slightly stronger in the older sites. This is consistent with estimates from a global database of soil and plant δ15N showing little evidence for fractionating N losses in cool, wet biomes (Amundson et al., 2003), suggesting that fractionating losses such as denitrification either do not occur or are not expressed in these biomes. Our observations are also consistent with dissolved organic N leaching being the dominant loss pathway near the Franz Josef sites (McGroddy et al., 2008), although strong P limitation in the oldest sites may make inorganic N losses more likely.

Figure 4.

Calculated ecosystem-level δ15N relative to input δ15N. Ecosystem δ15N is calculated as a weighted average of plant N and soil N. Plant δ15N is weighted foliar N (see Fig. 3) adjusted by +1.6‰ to account for fractionation within the plant. Soil δ15N is the weighted average of the soil horizons adjusted by +1.4‰ at sites with deep soils to account for 15N-enriched deep soil (see Supporting Information, Methods S1 for details). The seven symbols at each site reflect different ratios of total soil N to total plant N, ranging from 0 : 1 (plants only) to ∞ : 1 (soils only). The thin line is our best guess at ecosystem-level δ15N. The thick horizontal lines are the upper and lower reasonable values of input δ15N from our measurements of bulk deposition and N fixation. Values above this range indicate net fractionation during N losses over the course of ecosystem development. At sites 1 and 2 both plant δ15N and soil δ15N are within the input range of −2.3‰ to 1.5‰, so no ratio of soil N : plant N suggests net N loss fractionation. At sites 3 and 5, soil N : plant N ratios between 1 : 1 and 8 : 1 all yield ecosystem δ15N values within the input range, but more extreme ratios would yield values outside the input range. At sites 7 and 9 soil N : plant N ratios of 4 : 1 and greater (site 7) and 6 : 1 and greater (site 9) produce ecosystem δ15N values more 15N-enriched than the input range. Note the logarithmic scale of the horizontal axis.

Litter is depleted relative to organic and mineral soil

The 15N-enrichment from less decomposed (bulk litter) to more decomposed (mineral) soil horizons is consistent within each of our sites, and becomes exaggerated as soil age increases. This pattern is consistent with observations from temperate deciduous forests (Templer et al., 2007), boreal forests (Hyodo & Wardle, 2009), tropical forests (Martinelli et al., 1999), temperate grasslands (Brenner et al., 2001; Baisden et al., 2002), and elsewhere (Hobbie & Ouimette, 2009). All of the δ15N patterns we observe at Franz Josef depend on complex interactions between multiple processes, discriminations, and degrees of completion, and a full understanding of these drivers would require careful experimentation. However, we can speculate about whether previously identified mechanisms are consistent with our results.

Amundson et al. (2003) and Amundson & Baisden (2000) presented simplified ecosystem N cycling using a two-pool model containing plants and soil. In that approach, fractionation may be associated with external losses of N and uptake of N from soil to plants. As noted previously, little or no evidence exists for fractionating external N losses along this chronosequence, so that a two-pool model implies the results we observe are caused by increasing fractionation associated with plant uptake. However, the simplified two-pool approach does not attempt to account for differences among co-occurring litter, organic soil and mineral soil. Some processes within the ecosystem can be expected to cause N isotope discrimination within these layers (Templer et al., 2007). Among these processes, mineralization imparts only limited fractionation against 15N, while nitrification and plant uptake can discriminate more strongly against 15N, leaving behind a pool of 15N-enriched available N (Högberg, 1997; Robinson, 2001; Dawson et al., 2002; Amundson et al., 2003). Successive processes of mineralization, nitrification, and plant uptake of ammonia or nitrate are therefore likely to explain the progressive increase in δ15N values from litter to organic soil to mineral soil.

Decreasing foliar δ15N (δ15Nf) with soil age

The δ15Nf observed at Franz Josef spans a large part of the global range (Amundson et al., 2003; Craine et al., 2009), and is among the largest ranges recorded at a single location (Nadelhoffer et al., 1996; Clarkson et al., 2005). One of the most striking isotopic patterns at Franz Josef is the decrease in δ15Nf across the chronosequence, which is seen in both the dominance-weighted foliage (Fig. 3a) and many individual species (Fig. 1). Because N retranslocation out of foliage does not fractionate against 15N in these sites (Fig. 2a), δ15Nf reflects the plant N source pool (e.g. nitrate, ammonium and small organic N molecules in the soil) at each site modified by fractionation against 15N between the soil and foliage. The plant N source δ15N and fractionation between soil and foliage depend on myriad interacting processes, but it is possible to speculate about the likelihood of potential drivers.

The plant N source pool could become 15N-depleted across the chronosequence if it comes primarily from decomposition of litter. In support of this interpretation, Baisden et al. (2002) found that plant N resembled soil pools with rapid turnover in California grasslands. The interpretation that plant N becomes depleted because litter N becomes depleted, however, suffers from a chicken-and-egg problem, as litter N is depleted because plant N is depleted. Therefore, although this feedback can contribute to the foliar pattern, it cannot wholly explain it.

Fractionation during N uptake could also increase with site age. If N is strongly limiting in the early sites, N uptake would be nearly complete, resulting in little uptake fractionation regardless of root or mycorrhizal isotopic discrimination. By contrast, if N is plentiful relative to plant demand in the older, presumably P-limited sites, the expressed fractionation may be greater (Högberg, 1997; Fry et al., 2000; McKee et al., 2002; Clarkson et al., 2005), which could deplete δ15Nf with increasing soil age. Mycorrhizal N transfer to plants can fractionate strongly (Hobbie et al., 1999a; Craine et al., 2009), and the decreasing overall fertility in the older sites might encourage greater mycorrhizal colonization, which would be consistent with decreasing δ15Nf across the sites. Finally, although fractionation during within-plant processing can be large (up to 11‰) (Handley & Raven, 1992), it is typically small (< 2‰) (Högberg, 1997; Houlton et al., 2007; Templer et al., 2007), so there is little evidence to suggest change in within-plant fractionation across the sites.

Different isotopic values for different species

The consistent and dramatic isotopic separation of different species’δ15Nf that we observed at each site is not unique to Franz Josef (Michelsen et al., 1996, 1998; Hobbie et al., 2005; Templer et al., 2007; Kahmen et al., 2008). However, the substantially greater variation in species δ15N in the older sites, suggesting a possible divergence in plant strategies for acquiring and processing N, differs from other chronosequences (Vitousek et al., 1989; Martinelli et al., 1999; Hobbie et al., 2000; Hyodo & Wardle, 2009). Tree preferences for different forms of nitrogen such as nitrate vs ammonium, (McKane et al., 2002; Templer & Dawson, 2004; Kahmen et al., 2008) – which can differ widely in δ15N at a given site (Houlton et al., 2007) – could lead to species differences in δ15Nf (Kahmen et al., 2008). Species differences within sites do not result from different types of mycorrhizal associates – which can influence δ15Nf (Michelsen et al., 1996, 1998; Craine et al., 2009) – because all species we sampled form arbuscular mycorrhizal associations (McNabb, 1958; Baylis et al., 1963; Hall, 1975; Hurst et al., 2002; Russell et al., 2002). However, different degrees of mycorrhizal infection can also influence δ15Nf (Hobbie et al., 2005), and might differ between species. If this is the dominant mechanism, we would expect the 15N-enriched species such as D. cupressinum and M. ramiflorus to have less mycorrhizal infection than the 15N-depleted species such as Q. acutifolia and W. racemosa. The grouping of plant species by δ15Nf is intriguing, as it does not correspond to groupings by traits such as leaf mass per unit area or growth rate (S. Richardson, unpublished).

Bryophyte 15N-enrichment

The 15N-enrichment of bryophytes in Franz Josef contrasts with bryophyte δ15N in Europe (Pearson et al., 2000; Bragazza et al., 2005; Solga et al., 2005; Skinner et al., 2006; Zechmeister et al., 2008) and Central America (Hietz et al., 2002; Wania et al., 2002), which is typically depleted (generally −12‰ to −2‰, but up to 6‰ near NOx sources). The bryophytes in many of these other studies were not under a canopy, and atmospheric deposition rates were generally high (up to 20 kg N ha−1 yr−1); there was a strong correlation between bryophyte δ15N and atmospheric deposition δ15N. The bryophytes in our study occur as epiphytes, on the forest floor, and on fallen logs, and atmospheric deposition is low (0.9–1.5 kg N ha−1 yr−1 (Menge & Hedin, 2009)). Thus, N sources also include biological N fixation by cyanobacteria within the mats, leaching from foliage above, and litter deposition directly onto mats. All of these potential sources are 15N-depleted relative to bryophytes (Figs 1–3). Therefore, the 15N-enriched bryophyte signal could result from fractionating N losses from bryophyte mats themselves or fractionating N uptake into aerial plant roots scavenging bryophyte mats. At least one species at Franz Josef –Metrosideros umbellata – has aerial roots (Dawson, 1967).

Comparison of soil and foliar results to other chronosequences

Bulk soil shows 15N enrichment with site age in California (Brenner et al., 2001) and Hawaii (Martinelli et al., 1999), but no apparent change with site age in Sweden (Hyodo & Wardle, 2009) and Alaska (Hobbie et al., 1999a). At Franz Josef there is no strong soil δ15N pattern; there appears to be slight enrichment across the sites in bulk soil δ15N when soil pools are weighted by N content (Fig. 4), but neither mineral nor organic soil changes with soil age on its own. An increase in net N loss fractionation with site age has been proposed as the explanation for the Hawaiian (Vitousek et al., 1989) and Californian (Brenner et al., 2001) sites. This explanation is consistent with our N loss calculation, which suggests no fractionation during N losses at the younger Franz Josef sites and small amounts of fractionation during N losses at the older sites. All the chronosequences display more 15N enrichment with degree of soil decomposition (Hobbie et al., 1999a; Martinelli et al., 1999; Brenner et al., 2001; Hyodo & Wardle, 2009) (Fig. 3). Discrimination against 15N during some aspect of decomposition, as hypothesized by Templer et al. (2007), is the most universal candidate for this broad pattern. Foliage becomes more 15N-depleted with soil age in Franz Josef (Figs 1, 3a), similar to Alaska (Hobbie et al., 1999a) but unlike Hawaii (Vitousek et al., 1989; Martinelli et al., 1999), Sweden (Hyodo & Wardle, 2009) (both more 15N-enriched with site age), and California (Brenner et al., 2001) (no change).

Isotopically, the Franz Josef chronosequence is most similar to the Alaskan chronosequence, and they are similar in other ways as well. Both are primary successional sequences initiated by glacial retreat, both have actinorhizal N fixers that dominate early successional habitats (Chapin et al., 1994; Richardson et al., 2004), and both become nutrient-poor in the older sites. In Alaska mycorrhizal transfer has been implicated as the dominant mechanism explaining decreasing foliar δ15N with soil age (Hobbie et al., 1999b, 2000, 2005). This could contribute to the pattern at Franz Josef, but we have no hard evidence to evaluate this mechanism, and other factors such as fractionation during N uptake, within-plant processing, or fractionation during soil organic matter decomposition might play roles. The ways in which the Franz Josef chronosequence differs from the Alaskan chronosequence represent an intermediate between the Alaskan and Hawaiian sites, consistent with its intermediate latitudinal location. For example, the old sites at Franz Josef are thought to be P-limited like the old Hawaiian sites but N-poor like the old Alaskan sites, and the bulk soil δ15N pattern with age lies between the clear rise in Hawaii and the stasis in Alaska.

This study represents the first chronosequence study of N isotopes in temperate forests, helping to fill in the global map of natural abundance N isotope patterns. The clear decrease in foliar and litter δ15N with soil age, as observed in boreal but not in tropical forests, could result from a number of mechanisms, including fractionation during decomposition or fractionation between the plant N source and foliage, although we currently lack the data to evaluate the likelihood of these mechanisms. The divergence of foliar and litter δ15N from slower turnover soil N pools appears ubiquitous across forests worldwide, and could result from fractionation during some aspect of decomposition. Soil δ15N does not rise clearly as it does in tropical forests, corresponding to a lower likelihood of net fractionation during N losses.


We thank A. Rajendram for isotopic analyses at the Waikato Stable Isotope Unit (New Zealand) and J. Brookshire for isotopic analyses at Princeton University (USA). This project benefited from useful discussions with Michelle Mack, Ted Schuur, and members of Lars Hedin’s lab. D.N.L.M. was supported in part as a Postdoctoral Associate at the National Center for Ecological Analysis and Synthesis, a Center funded by National Science Foundation (grant #EF-0553768), the University of California, Santa Barbara, and the State of California; and in part by the Carbon Mitigation Initiative, with funding from BP and Ford. W.T.B. was funded by the New Zealand Foundation for Research, Science and Technology (FRST). S.J.R. and D.A.P. were funded by the Ecosystem Resilience Outcome-Based Investment (contract C09X0502) through the New Zealand FRST.