Imprint of denitrifying on the global terrestrial biosphere

Benjamin Z. Houlton1 and Edith Bai

Department of Land, Air, and Water Resources, University of California, Davis, CA 95616

Communicated by William H. Schlesinger, Cary Institute of Ecosystem Studies, Millbrook, NY, October 21, 2009 (received for review June 6, 2009) Loss of (N) from land limits the uptake and storage of Results and Discussion. atmospheric CO2 by the biosphere, influencing Earth’s climate Across natural terrestrial environments, 15N/14N of bulk soils system and myriads of the global ecological functions and services (integrated to 50-cm depth) and foliage vary systematically with on which humans rely. Nitrogen can be lost in both dissolved and climate (2–4); hence, large-scale (i.e., regions, biomes) terrestrial gaseous phases; however, the partitioning of these vectors re- 15N/14N ratios can be estimated from precipitation and temper- mains controversial. Particularly uncertain is whether the bacterial ature fields (3). Although this model can introduce errors at conversion of plant available N to gaseous forms (denitrification) smaller scales, it reasonably captures the shift in plant and soil plays a major role in structuring global N supplies in the nonagrar- ␦15N across biomes, pointing to its robustness at the global scale ian centers of Earth. Here, we use the isotope composition of N (Fig. S1 and Tables S1 and S2). Taking advantage of this model (15N/14N) to constrain the transfer of this nutrient from the land to (3) and eliminating those areas of land that are managed by the water and atmosphere. We report that the integrated 15N/14N humans (Methods and SI Text), we estimate that the area- ␦15 of the natural terrestrial biosphere is elevated with respect to that integrated N of natural soil and vegetation leaves are equal to Ϯ Ϯ of atmospheric N inputs. This cannot be explained by preferential 5.4‰ ( 2.2) and 0.43‰ ( 2.5), respectively (Fig. 1A). Plants 15 14 loss of 14N to waterways; rather, it reflects a history of low 15N/14N are known to have a lower N/ N than soils owing to various gaseous N emissions to the atmosphere owing to denitrifying mechanisms (5, 6), consistent with the offset in Fig. 1. Although plants are not expected to alter the 15N/14N of bulk soils under

bacteria in the soil. Parameterizing a simple model with global N ECOLOGY quasiequilibrium conditions (7), we account for any such N isotope data, we estimate that soil denitrification (including N2) Ϸ recycling effects by combining components of the biosphere into accounts for 1/3 of the total N lost from the unmanaged terres- a single vector. Because roots, shoots, and leaves are similar in trial biosphere. Applying this fraction to estimates of N inputs, N2O their N isotope compositions (8) (Table S3), the 15N/14N of leaves Ϸ and NOx fluxes, we calculate that 28 Tg of N are lost annually via also reflects that of total vegetation. Using estimates (9) of the N2 efflux from the natural soil. These results place isotopic con- N stored in vegetation and soils worldwide, we show that the straints on the widely held belief that denitrifying bacteria account mass- and area-weighted ␦15N of the unmanaged terrestrial for a significant fraction of the missing N in the global N cycle. biosphere is equal to 5.3‰ (Fig. 1A). SCIENCES The 15N/14N of the natural terrestrial biosphere is elevated ͉ ͉ ͉ ͉ ENVIRONMENTAL denitrification isotope nitrogen ecosystem microbe compared with external N inputs (Fig. 1A). Biological N2 fixation, the biochemical conversion of N2 to ammonia, imparts a minor e present a simple nitrogen (N) isotope model to inves- isotope effect; its ␦15N ranges from 0 to Ϫ2‰ (10). The other major N input to natural systems occurs via atmospheric deposition, Wtigate the partitioning of gaseous (fgas) vs. leaching originating from natural and from fossil fuel sources, producing (fleaching) pathways of N removal from the nonagricultural (i.e., unmanaged) terrestrial biosphere. NЈs isotopes, 14N and 15N, are mainly NOx that rains down as on land (11). We show in Fig. ␦15 stable but display a high degree of variation in biogenic materials 1B that the N of bulk nitrate deposition observed for various owing to isotope fractionation, particularly kinetic fractionation, latitudes (Table S4), altitudes and biomes (from deserts to tropical rainforests; n ϭ 187) approach a median of Ϫ1.3‰, with most whereby organisms’ enzymes more rapidly transfer the light values clustered between Ϫ3 and 1. Although and isotope from substrates to products. 15N is rare (Ϸ0.37%), and 14 dissolved organic N compounds can also contribute deposition, small deviations in its concentration relative to N have been their 15N/14N ratios either overlap with or are somewhat 15N- widely used in the examination of N cycling on land and in the depleted with respect to that of nitrate in bulk precipitation (7, 12, sea (1). In the case of the N isotope composition of the terrestrial 13). We therefore infer that there must be a pathway of N loss that biosphere, we consider a steady-state approximation to examine preferentially removes 14N from the land, acting to elevate the the partitioning of global N losses: terrestrial biosphere’s 15N/14N over the long term. Two mechanisms could underlie terrestrial 15N enrichment. First, ␦15 Ϫ ␦15 ϩ␧ NTB NI L 15 14 f ϭ [1] leaching of N to waterways could remove low N/ N compounds gas ␧ Ϫ␧ 15 14 L G from soil (␧L) (Eq. 1), thereby elevating the N/ N of remaining ecosystem pools (1). In particular, can produce ␦15 ϭ 15 14 15 14 Ϫ ϫ 15 14 15 in which N (( N/ Nsample/ N/ Nstd 1) 1,000) of the low N/ N nitrate (␧ ϷϪ15 to Ϫ35‰) (1); if N-depleted nitrate terrestrial biosphere (TB) or atmospheric inputs (I); ␧ is the were to escape ecosystems via streams, it would increase the 15N/14N isotope effect associated with leaching (L) or gaseous (G) N removal; and 1 ϭ fgasϩ fleaching. We define ␧ as the ratio of the rate constants of the heavy and light isotopes (i.e., ␧ ϭ (15k/ Author contributions: B.Z.H. designed research; B.Z.H. and E.B. performed research; B.Z.H. 14k Ϫ1) ϫ 1,000) at steady state, whereby instantaneous isotope and E.B. analyzed data; and B.Z.H. and E.B. wrote the paper. effects are equivalent to equilibrium ones. Here, we constrain The authors declare no conflict of interest. global N losses by: (i) estimating the ␦15N of natural plants Freely available online through the PNAS open access option. and soils worldwide; (ii) compiling data on the ␦15N of N inputs, To whom correspondence should be addressed at: Department of Land, Air, and Water Resources, University of California, One Shields Avenue, Davis, CA 95616. Email: and potential fractionations via leaching vs. gaseous N losses; [email protected]. and (iii) parameterizing Eq. 1 to quantify the N lost via microbial This article contains supporting information online at www.pnas.org/cgi/content/full/ . 0912111106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0912111106 PNAS Early Edition ͉ 1of4 Downloaded by guest on September 28, 2021 Fig. 2. 15N/14N of N leaching losses to small streams across the terrestrial biosphere. These data were compiled from the primary literature (see SI Text). The small black symbols are individual observations. The large black symbols reflect site means for the ␦15N of stream nitrate; the gray symbols reflect the means for the ␦15N of total dissolved N in small streams. The regression line is drawn through the site means.

of dissolved N might also influence soil ␦15N, neither is there evidence for major 15N-discriminations against soil N isotopes by total N leaching losses (gray circles) (Fig. 2). Thus the available evidence points to a small leaching term; based on the intercept in Fig. 2, it seems that leaching could at most raise the ␦15N of the terrestrial biosphere by Ϸ1‰ above N inputs (see Fig. 1A) The only remaining mechanism involves isotope fractionation via gaseous pathways, particularly bacterial ones. Ammonia volatilization is not likely to affect the integrated 15N/14Nofthe terrestrial system; the short atmospheric lifespan of ammonia Fig. 1. 15N/14N of soil, vegetation and inputs to the terrestrial biosphere. (A) means that this N form mainly recycles within and among 15 ␦15N of soil and vegetation were estimated from multiple regression models terrestrial regions (17). Regional-scale anomalies in ␦ N, which (3) and integrated at the global scale in the absence of human-managed contain important information beyond the global mean, may be environments (See Methods and SI Text). The ␦15N of the terrestrial biosphere influenced by processes such as volatilization; but our integrated ␦15 ϫ ϩ ␦15 ϫ ϩ was estimated as follows: ( Nsoil Nsoil Nveg Nveg)/(Nsoil Nveg), where estimate would be minimally affected by this process. Although N is total N (i.e., 14Nϩ15N). The solid lines delineate the ␦15N range for emission and transport of ammonia from the terrestrial bio- biological dinitrogen fixation (10); the dashed line is the median of bulk sphere to the ocean has increased in response to anthropogenic nitrate deposition from B. Errors for soil and vegetation reflect the SE of the regression models, which were propagated to the terrestrial biosphere (i.e., activities (18), particularly N fertilizer applications, ammonia 2 ϩ 2 1/2 15 14 emissions from natural soils are relatively small (i.e., 2.4 Ϫ 10 Tg (SEsoil SEveg) ). (B) N/ N of bulk nitrate deposition from different loca- tions around the globe, including sites spanning temperate to tropical biomes, N/yr) (19, 20). By contrast, bacterial denitrifiers produce sub- and low to high altitudes (see SI Text). stantial quantities of long-lived gaseous compounds (N2,N2O) which mix evenly in the atmosphere before returning to land. Denitrification can also produce NOx, especially in dry environ- of remaining terrestrial matter. Alternatively, by discriminating ments (21), some of which rapidly washes out of the atmosphere 15 ␧ against N, gaseous N losses ( G) (Eq. 1) could impact the as deposition (17, 22). Nevertheless, this term is small compared 15 14 terrestrial N/ N system in manner similar to that of leaching (7). with the combined losses of NOx,N2O, and (especially) N2 (22). However, nitrate leaching does not seem to significantly discrim- Proximally, whether originating from denitrification or nitri- inate against ecosystem 15N (Fig. 2). Across major biomes and fier-denitrification, bacterial pathways of gaseous N removal latitudes, including arctic tundra, temperate and boreal forest, and would be expected to elevate the 15N/14N of nitrate relative to tropical lowland and montane forest (Table S5), there is good any upstream isotope effects (23, 24). Indeed, the average agreement between the ␦15N of nitrate in small drainage streams isotope effect of denitrification on nitrate is substantial in both and the ␦15N of soil particulate matter (Fig. 2). Together the slope pure culture (Ϫ20 Ϯ1.0‰) and in natural soil communities (0.98) and intercept (Ϫ0.85) of this significant (r2 ϭ 0.82; n ϭ 63; (Ϫ16 Ϯ1.6‰) (Table S6). The slightly lower (P Ͻ 0.1; two-tailed P Ͻ 0.001) linear relationship point to ␧L ϽϪ1‰ that is system- t test) value for soil might be due to heterogeneities in nitrate atically conserved across these diverse biomes. Missing from Fig. 2 consumption, which reduces expression of the isotope effect are arid sites with high average soil ␦15N(Ϸ4–15‰) (14, 15). relative to homogenous (culture) conditions (7, 25). Perhaps Although leaching is limited in such environments (i.e., evapora- more important, the isotope effect of denitrification seems to tion Ͼ precipitation), observations of high nitrate-␦15Ninbiolog- control entire ecosystems: the near Ϸ1:1 relationship in Fig. 2 ically active soils (8–15‰) (14) and in water that percolates from can only be explained by fractionating losses via denitrification soils to ground water (10–13‰) (16) suggest a similarly small under steady state conditions (7). The similarities between leaching effect in natural arid lands. Finally, although other forms nitrate ␦15N and soil ␦15N indicates that the rather high isotope

2of4 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0912111106 Houlton and Bai Downloaded by guest on September 28, 2021 Table 1. Global isotopic constraints on N2 efflux from natural terrestrial soil (Tg N/yr) Inputs Gaseous N emissions

Fixation ϩ deposition N2ONOx N2*N2/N2O

134.64 12.0† 5.0† 26.6 2.2 130 6.1‡ 9.7‡ 27.9 4.6 8.0§ 7.4§ 28.3 3.5

Inputs to natural systems from Gruber and Galloway (130 Tg N) (22) and van Drecht et al. (134.64 Tg N/yr) (30). *This study, where denitrification ϭ 33% of total N losses; and N2 ϭ average N inputs (i.e., 132 Tg N) ϫ 0.33 Ϫ (N2O ϩ NOx). †Bowden (36). ‡Potter et al. (37). §Galloway et al. (17), preindustrial (unmanaged) soil.

are based on total terrestrial N, mainly reflect N losses via denitrification vs. leaching pathways. We conclude that the elevated 15N/14N of the natural (i.e., unmanaged) terrestrial biosphere bears the long-term imprint of microbial denitrification (including NOx,N2O, and N2), provid- ing a quantitative tracer of gaseous N emissions from the soil. The 15N/14N of the biosphere integrates many regional patterns thereby reflecting the relative strengths of leaching vs. denitri- Fig. 3. Steady-state solutions of the fraction of N lost via denitrification at fication pathways as a whole. Arid environments, for example,

the global scale. Isoclines are based on Eq. 1. The midpoint of inputs is based are among the most 15N enriched of the terrestrial biomes; yet ECOLOGY on our compilation in Fig. 2 and estimates of the ␦15N of biological dinitrogen denitrification rates are likely low in arid soils compared to other ␧ fixation. The isotope effect of leaching ( L) is based on the calculated intercept global ecosystems (30). Rather, our analysis suggests that gas- ␧ in Fig. 2. The isotope effect of denitrification ( G) is shown for culture studies eous N efflux is comparatively high relative to leaching in sites (square) and native soil communities (circle) (see SI Text). with high ␦15N, not necessarily large in absolute flux, consistent with global N cycle models (30, 31). Future work on regional effect of nitrification is not fully expressed (i.e., closed system) patterns will help in identifying the dominant source regions of and that any low 15N/14N nitrate produced during nitrification N release from the land to the water and atmosphere. SCIENCES passes through zones of denitrification in the soil before leaching For decades, N2 has been prominently purported in the (7). In sum, the elevated 15N/14N of the terrestrial biosphere (Fig. missing N of the global N cycle (9, 17, 22, 29, 32, 33). Using ENVIRONMENTAL ϭ 1A) points to substantial 14N removal via denitrifying bacteria— current estimates of global N inputs (mean 132 Tg N/yr; Table consistent with how this bacterial group elevates the 15N/14Nof 1) to natural ecosystems and subtracting fluxes of NOx and N2O the global nitrate reservoir in the oceanic water column (23, 26). from natural soils, we place isotopic constraints on N2 efflux Parameterizing Eq. 1 with the mid-point of inputs (Ϫ1‰) and from the unmanaged terrestrial biosphere (see Table 1). Spe- cifically, we estimate that Ϸ28 Tg of N are lost annually via the observed isotope effect of leaching (␧L ϭϪ0.8‰; Fig. 2) and denitrification to N in the terrestrial soil beneath natural denitrifying bacteria in pure culture (␧G ϭϪ20‰; Table S6)we 2 estimate that denitrification accounts for a substantial fraction vegetation, with an N2/N2O ratio ranging from 2.2 to 4.6 at the (mean ϭ 28%; maximum sensitivity range ϭ 26% to 31% given global scale. This implies that unmeasured N2 losses are indeed 15 15 substantial. The range in N /N O ratios is driven (largely) by by changing the mean ␦ NTB and ␦ NI by Ϯ10%) of fixed N loss 2 2 from the unmanaged terrestrial biosphere (square; Fig. 3). This variations in global estimates of NOx and N2O (Table 1); the estimate is probably a lower bound: any microzones of complete lower end of the range is approximately twice as high as the mean ϭ nitrate consumption in soil would not elevate the 15N/14Nof of point measures for upland ecosystems (i.e., N2/N2O 1), remaining pools (7). Using the mean isotope effect observed for although within the very wide range of the estimates reported Ϸ native denitrifier communities in various soils (Ϫ16‰) to help ( 0–32) (29). Moreover, the total denitrification flux (44 Tg account for such isotopic underexpression (circle; Fig. 3) in- N/yr) we estimate for the unmanaged terrestrial biosphere creases the contribution of denitrification to Ϸ36% (maximum agrees reasonably well with estimates for natural soils as based sensitivity range ϭ 32% to 40% given by changing the mean on numerical simulation models (e.g., 58 Tg N/yr) (31). Like other 15 15 approaches, our isotope-based calculation relies on knowledge of N ␦ NTB and ␦ NI by Ϯ10%) Additional information on denitri- fication’s isotope effect will help to constrain our estimate inputs, which suffer from their own uncertainties; the inputs we further (Fig. 4 and Table S6). Other uncertainties in our model used here fall on the lower end of the spectrum for natural N2 could arise from our steady-state assumption, because N may be fixation (100–290 Tg N/yr) (34). Combining fluxes with isotope- accumulating in terrestrial systems (17). For instance, given based models seems like a promising way forward in addressing modern rates of terrestrial C accumulation (Ϸ2 Pg C/yr) (27), a denitrification and N2 losses quantitatively (e.g., ref. 35), an ap- C/N ratio that partitions this C into soil (15/1) and plant materials proach that may be amenable to agricultural sites as well. (200/1) equally (28), and assuming that all of the additional C Methods results from new N inputs, we estimate that Ϸ18 Tg N/yr is ␦15 accumulating on land. This value drops to Ϸ10 Tg N/yr if all of The N of the terrestrial soil and plant matter were derived from global climate regressions, integrated across the terrestrial surface in the absence of the C is assumed to be stored in plant biomass, consistent with human-managed areas of land (see also SI Text). The integrated soil (or plant) other estimates (e.g., 9 Tg N/yr) (29). Even integrated over 100 ␦15N was obtained by applying the zonal tools algorithm in ArcView GIS years, however, such rates of N accumulation would only have Spatial Analyst to a raster estimate of the unmanaged soil (or plant) ␦15N altered total terrestrial N by Ͻ0.1% (9). Thus, our results, which surfaces. The global unmanaged surface was based on the biome classification

Houlton and Bai PNAS Early Edition ͉ 3of4 Downloaded by guest on September 28, 2021 scheme of VUB and VITO, derived from a full year cycle (1998–1999) of 10 daily soil communities were collated from the primary literature. Details of our composites of SPOT-VEGETATION (www.geosuccess.net/Geosuccess). Areas literature syntheses can be found in the SI Text. classified as croplands, urban and built-up, and cropland and natural vegeta- tion mosaic were considered as human-managed. ACKNOWLEDGMENTS. We thank Bryant A. Browne for influential discussions, Data on the ␦15N of N leaching losses to small streams and atmospheric N and Audrey Niboyet for comments. This work was supported by the Andrew inputs, and the isotope effect of denitrification in pure culture and in native W. Mellon Foundation.

1. Hogberg P (1997) Tansley review No 95 - N-15 natural abundance in soil-plant systems. 19. Bouwman AF, et al. (1997) A global high-resolution emission inventory for ammonia. New Phytol 137:179–203. Glob Biogeochem Cycles 11:561–587. 2. Handley LL, et al. (1999) The N-15 natural abundance (delta N-15) of ecosystem samples 20. Schlesinger WH, Hartley AE (1992) A global budget for atmospheric NH3. Biogeochem- reflects measures of water availability. Aust J Plant Physiol 26:185–199. istry 15:191–211. 3. Amundson R, et al. (2003) Global patterns of the isotopic composition of soil and plant 21. Hartley AE, Schlesinger WH (2000) Environmental controls on nitric oxide emission nitrogen Glob Biogeochem Cycles 17:1031. from northern Chihuahuan desert soils. Biogeochemistry 50:279–300. 4. Craine JM, et al. (2009) Global patterns of foliar nitrogen isotopes and their relation- 22. Gruber N, Galloway JN (2008) An Earth-system perspective of the global . ships with climate, mycorrhizal fungi, foliar nutrient concentrations, and nitrogen Nature 451:293–296. availability. New Phytol 183:980–992. 23. Brandes JA, Devol AH (2002) A global marine-fixed nitrogen isotopic budget: Impli- 5. Houlton BZ, Sigman DM, Schuur EA, Hedin LO (2007) A climate-driven switch in plant cations for Holocene nitrogen cycling. Glob Biogeochem Cycles 16:1120. nitrogen acquisition within tropical forest communities. Proc Natl Acad Sci USA 24. Perez T, et al. (2000) Isotopic variability of N2O emissions from tropical forest soils. Glob 104:8902–8906. Biogeochem Cycles 14:525–535. 6. Hobbie EA, Macko SA, Shugart HH (1999) Interpretation of nitrogen isotope signatures 25. Brandes JA, Devol AH (1997) Isotopic fractionation of oxygen and nitrogen in coastal using the NIFTE model. Oecologia 120:405–415. marine sediments. Geochim Cosmochim Acta 61:1793–1801. 7. Houlton BZ, Sigman DM, Hedin LO (2006) Isotopic evidence for large gaseous nitrogen 26. Deutsch C, Sigman DM, Thunell RC, Meckler AN, Haug GH (2004) Isotopic constraints losses from tropical rainforests. Proc Natl Acad Sci USA 103:8745–8750. on glacial/interglacial changes in the oceanic nitrogen budget. Glob Biogeochem 8. Pardo LH, et al. (2006) Regional assessment of N saturation using foliar and root delta Cycles 18:GB4012. N-15 Biogeochemistry 80:143–171. 27. Prentice IC, Heimann M, Sitch S (2000) The carbon balance of the terrestrial biosphere: 9. Schlesinger WH (1997) Biogeochemistry: An Analysis of Global Change (Academic, San Ecosystem models and atmospheric observations. Ecol Appl 10:1553–1573. Diego). 28. Hungate B, Dukes J, Shaw M, Luo Y, Field C (2003) Nitrogen and climate change. Science 10. Shearer G, Kohl DH (1986) N-2-fixation in field settings—Estimations based on natural 302:1512–1513. N-15 abundance. Aust J Plant Physiol 13:699–756. 29. Schlesinger WH (2009) On the fate of anthropogenic nitrogen. Proc Natl Acad Sci USA 11. Holland EA, et al. (1997) Variations in the predicted spatial distribution of atmospheric 106:203–208. nitrogen deposition and their impact on carbon uptake by terrestrial ecosystems. J 30. Van Drecht G, Bouwman AF, Knoop JM, Beusen AHW, Meinardi CR (2003) Global Geophys Res 102:15849–15866. modeling of the fate of nitrogen from point and nonpoint sources in soils, ground- 12. Cornell S, Rendell A, Jickells T (1995) Atmospheric inputs of dissolved organic nitrogen water, and surface water. Glob Biogeochem Cycles 17:1115. to the oceans. Nature 376:243–246. 13. Heaton THE, Spiro B, Madeline S, Robertson C (1997) Potential canopy influences on 31. Seitzinger S, et al. (2006) Denitrification across landscapes and waterscapes: A synthe- the isotopic composition of nitrogen and sulphur in atmospheric deposition. Oecolo- sis. Ecol Appl 16:2064–2090. gia 109:600–607. 32. Delwiche CC (1970) Nitrogen cycle. Sci Am 223:136–146. 14. Densmore JN, Bohlke JK (2000) Use of nitrogen isotopes to determine sources of nitrate 33. Galloway JN, Schlesinger WH, Levy IIH, Michaels A, Schnoor JL (1995) Nitrogen fixation: contamination in two desert basins in California. Interdisciplinary Perspectives Drink- Anthropogenic enhancement-environmental response. Glob Biogeochem Cycles ing Water Risk Assessment Management 260:63–73. 9:235–252. 15. Shearer G, et al. (1983) Estimates of N-2-fixation from variation in the natural abun- 34. Cleveland CC, et al. (1999) Global patterns of terrestrial nitrogen (N2) fixation in dance of N-15 in Sonoran Desert ecosystems. Oecologia 56:365–373. natural ecosystems. Glob Biogeochem Cycles 13:623–645. 16. McMahon PB, Bolke JK (2006) Regional patterns in the isotopic composition of natural 35. Bai E, Houlton BZ (2009) Coupled isotopic and process-based modeling of gaseous and anthropogenic nitrate in groundwater, high plains, USA. Environ Sci Technol nitrogen losses from tropical rainforests Global Biogeochem Cycles 23. 40:2965–2970. 36. Bowden WB (1986) Gaseous nitrogen emissions from undisturbed terrestrial ecosys- 17. Galloway JN, et al. (2004) Nitrogen cycles: Past, present, and future. Biogeochemistry tems—An assessment of their impacts on local and global nitrogen budgets. Biogeo- 70:153–226. chemistry 2:249–279. 18. Duce RA, et al. (2008) Impacts of atmospheric anthropogenic nitrogen on the open 37. Potter CS, Matson PA, Vitousek PM, Davidson EA (1996) Process modeling of controls ocean. Science 320:893–897. on nitrogen trace gas emissions from soils worldwide. J Geophys Res 101:1361–1379.

4of4 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0912111106 Houlton and Bai Downloaded by guest on September 28, 2021 Supporting Information

Houlton and Bai 10.1073/pnas.0912111106 SI Text To estimate the ␦15N of the unmanaged terrestrial biosphere, Estimating Terrestrial ␦15N. After ref. 1, we estimate of the ␦15Nof we integrated the ␦15N of plants and soils at the global scale and soil (Fig. 1; Fig. S1A) and plant (Fig. 1; Fig. S1B) matter by weighted each pool by total N mass. As a sensitivity check on this applying multiple regression models to climate data in the analysis, we compared the ␦15N of our model-only integration terrestrial biosphere: (Fig. S1 a and c; based on Eqs. s1 and s2) to a hybrid estimate that combined the model and empirical means in the temperate/ ϩ ء Ϫ ء ϭ 15␦ Nsoil 0.2048 MAT 0.0012 MAP 4.32 [S1] boreal and tropical forest biomes (from Table S1; Fig. S1 b and d). This shows good agreement: the model only global ␦15Nis [MAP ϩ 0.069 [S2ءMAT Ϫ 0.0016ء15N ϭ 0.0697␦ leaves almost identical to that of the model-empirical hybrid (Table S2). These models are based on empirical relationships observed for It thus seems that the model-based integration offers a rather various climnosequences, spanning different biomes and climatic robust approximation of ␦15N at the global scale. conditions on Earth’s surface. To develop an integrative ␦15N for Finally, based on past evidence (4) and our comparison in the entire land surface, we parameterized these models with Table S3, we assume that the ␦15N of plant leaves roughly reflects mean annual temperature (MAT) and precipitation (MAP) data that of total plant N. Consequently, we used the N content of (0.5 ϫ 0.5 latitude/longitude grids) (2). The integrated soil (or entire vegetation biomass (ref. 5) to estimate of the ␦15Nofthe plant) ␦15N was then obtained by applying the zonal tools terrestrial biosphere. algorithm in ArcView GIS Spatial Analyst to a raster estimate of the unmanaged soil (or plant) ␦15N surfaces. The global unman- Data Syntheses. We used ISI’s Web of Science to explore the aged surface was based on the biome classification scheme of literature for data on the isotope effect of denitrification in pure VUB and VITO, derived from a full year cycle (1998–1999) of culture and in native soil communities (Table S6), and the ␦15N 10 daily composites of SPOT-VEGETATION (www.geosuccess- of N inputs via deposition (Table S4) and N leaching losses .net/Geosuccess). Areas classified as croplands, urban and built- (Table S5). Many of the ␦15N data were not readily available in up, and cropland and natural vegetation mosaic were considered tables; rather, there were presented graphically. In these cases we as human-managed. used the Data Thief shareware (www.datathief.org/) to digitize We examined the validity of the model by comparing modeled the graphs and thereby acquire data on ␦15N for our synthesis. ␦15 and measured (3) N distributions across temperate/boreal vs. The geographic results of our ␦15N syntheses are presented in tropical biomes. Recognizing that the modeled values are Tables S4 and S5. weighted by area (as described above), whereas the empirical Data on the isotope effect of denitrification are shown in means neither include all areas nor are they weighted by area Table S6. Overall, we collated 38 separate observations for this within biomes (3), we show in Table S1 that modeled and ϭ measured ␦15N agree reasonably well, generally falling within 1 analysis. The mean of the culture data is 20 per mil (n 31), to 2 per mil of each other. An exception to such good agreement measured in both steady-state batches and under closed condi- may be seen in the case of tropical leaf ␦15N, where the model tions without continuous replenishment of nitrate substrates. ϭ appeared to underestimate the empirical mean by Ͼ3 per mil. The native soil community data (n 7), which included tem- The variation around the empirical mean is substantial, however, perate vs. tropical soils, was 16 per mil on average. Isotopic pointing to considerable variation in plant ␦15N within the effects in natural soils with native communities are quantified by tropical biome. The area-weighted model may be capturing some tracking concurrently changes in nitrate concentrations and ␦15N low ␦15N areas in the tropics (i.e., wet, montane forests) that are over the time course of incubations. Overall, denitrification in not well described by the empirical synthesis. Likewise, the culture seemed to impart a slightly higher isotope effect (P Ͻ model may be underestimating the ␦15N of soils in the tropics, 0.1), consistent with underexpression of the isotope effect of perhaps pointing to the conservative nature of our estimates of denitrification in heterogeneous environments; although more gaseous N fractions (Eq. 1, main text) in this region. data are needed to fully explore/resolve this hypothesis.

1. Amundson R, et al. (2003) Global patterns of the isotopic composition of soil and plant 11. Templer PH, Arthur MA, Lovett GM, Weathers KC (2007) Plant and soil natural nitrogen Glob Biogeochem Cycles 17:1031. abundance delta N-15: Indicators of relative rates of nitrogen cycling in temperate 2. Willmott CJ, Matsuura K (2000) Terrestrial Air Temperature and Precipitation: Monthly forest ecosystems. Oecologia 153:399–406. and Annual Climatologies, Version 3.01. Center for Climate Research (University of 12. Mariotti A, et al. (1981) Experimental-determination of nitrogen kinetic isotope Delaware, Newark, NY). fractionation—Some principles—Illustration for the denitrification and nitrification 3. Martinelli LA, et al. (1999) Nitrogen stable isotopic composition of leaves and soil: processes. Plant Soil 62:413–430. Tropical versus temperate forests. Biogeochemistry 46:45–65. 13. Barford CC, Montoya JP, Altabet MA, Mitchell R (1999) Steady-state nitrogen isotope 4. Pardo LH, et al. (2006) Regional assessment of N saturation using foliar and root delta effects of N-2 and N2O production in Paracoccus denitrificans. Appl Environ Microbiol N-15 Biogeochemistry 80:143–171. 65:989–994. 5. Schlesinger WH (1997) Biogeochemistry: An Analysis of Global Change (Academic, San 14. Miyake Y, E. W (1971) The isotope effect on the nitrogen in biochemical oxidation- Diego). reduction reactions. Records of Oceanographic Work in Japan 11:1–6. 6. Ometto JPHB, et al. (2006) The stable carbon and nitrogen isotopic composition of 15. Olleras T (1983) Kinetische Isotopeneffekte der Arginase- und Nitratreduktase- vegetation in tropical forests of the Amazon Basin, Brazil. Biogeochemistry 79:251– Reaktion: Ein Beitrag zur Aufklärung der entsprechenden Reaktionsmechanismen. 274. PhD dissertation (Technical University Munchen-Weihenstephan, Germany). 7. Kolb KJ, Evans RD (2002) Implications of leaf nitrogen recycling on the nitrogen isotope 16. Delwiche CC, Steyn PL (1970) Nitrogen isotope fractionation in soils and microbial composition of deciduous plant tissues. New Phytol 156:57–64. reactions. Environ Sci Technol 4:929–935. 8. Shearer G, Kohl DH (1986) N-2-fixation in field settings—Estimations based on natural 17. Granger J, Sigman DM, Lehmann MF, Tortell PD (2008) Nitrogen and oxygen isotope N-15 abundance. Aust J Plant Physiol 13:699–756. fractionation during dissimilatory nitrate reduction by denitrifying bacteria. Limnol 9. Gebauer G, Schulze ED (1991) Carbon and nitrogen isotope ratios in different com- Oceanogr 53:2533–2545. partments of a healthy and a declining Picea-Abies forest in the Fichtelgebirge, NE 18. Houlton BZ, Sigman DM, Hedin LO (2006) Isotopic evidence for large gaseous nitrogen Bavaria Oecologia 87:198–207. losses from tropical rainforests. Proc Natl Acad Sci USA 103:8745–8750. 10. Koopmans CJ, vanDam D, Tietema A, Verstraten JM (1997) Natural N-15 abundance in 19. Chien SH, Shearer G, Kohl DH (1977) Nitrogen isotope effect associated with nitrate two nitrogen saturated forest ecosystems. Oecologia 111:470–480. and nitrite loss from waterlogged soils. Soil Sci Soc Am J 41:63–69.

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 1of9 Downloaded by guest on September 28, 2021 20. Blackmer AM, Bremner JM (1977) Nitrogen isotope discrimination in denitrification of 30. Michalski G, Bohlke JK, Thiemens M (2004) Long term atmospheric deposition as the nitrate in soils. Soil Biol Biochem 9:73–77. source of nitrate and other salts in the Atacama Desert, Chile: New evidence from 21. Heaton THE, Spiro B, Madeline S, Robertson C (1997) Potential canopy influences on mass-independent oxygen isotopic compositions Geochim Cosmochim Acta 68:4023– the isotopic composition of nitrogen and sulphur in atmospheric deposition. Oecolo- 4038. gia 109:600–607. 31. Carrillo JH, Hastings MG, Sigman DM, Huebert BJ (2002) Atmospheric deposition of 22. Durka W, Schulze ED, Gebauer G, Voerkelius S (1994) Effects of forest decline on uptake inorganic and organic nitrogen and base cations in Hawaii Glob Biogeochem Cycles and leaching of deposited nitrate determined from N-15 and O-18 measurements. 16:1076. Nature 372:765–767. 32. Tye AM, Heaton THE (2007) Chemical and isotopic characteristics of weathering and 23. Buzek F, Cerny J, Paces T (1998) The behavior of nitrogen isotopes in acidified forest nitrogen release in non-glacial drainage waters on Arctic tundra. Geochim Cosmochim soils in the Czech Republic. Water Air Soil Pollut 105:155–164. Acta 71:4188–4205. 24. Spoelstra J, Schiff SL, Jeffries DS, Semkin RG (2004) Effect of storage on the isotopic 33. Oelmann Y, Kreutziger Y, Bol R, Wilcke W (2007) Nitrate leaching in soil: Tracing the composition of nitrate in bulk precipitation. Environ Sci Technol 38:4723–4727. 25. Hales HC, Ross DS, Lini A (2007) Isotopic signature of nitrate in two contrasting NO3- sources with the help of stable N and O isotopes. Soil Biol Biochem 39:3024–3033. watersheds of Brush Brook, Vermont, USA. Biogeochemistry 84:51–66. 34. Campbell JL, Mitchell MJ, Mayer B (2006) Isotopic assessment of NO3- and SO42- 26. Burns DA, Kendall C (2002) Analysis of delta N-15 and delta O-18 to differentiate NO3- mobility during winter in two adjacent watersheds in the Adirondack Mountains, New sources in runoff at two watersheds in the Catskill Mountains of New York Water York. J Geophys Res 111:GB4007. Resour Research 38:1051. 35. Mitchell MJ, et al. (2001) Nitrogen biogeochemistry of three hardwood ecosystems in 27. Campbell DH, Kendall C, Chang CCY, Silva SR, Tonnessen KA (2002) Pathways for the Adirondack region of New York. Biogeochemistry 56:93–133. nitrate release from an alpine watershed: Determination using delta N-15 and delta 36. Schuur EAG, Matson PA (2001) Net primary productivity and nutrient cycling across a O-18. Water Resour Research 38:1052. mesic to wet precipitation gradient in Hawaiian montane forest. Oecologia 128:431– 28. Garten CT (1996) Stable nitrogen isotope ratios in wet and dry nitrate deposition 442. collected with an artificial tree. Tellus Ser B 48:60–64. 37. Brandes JA, McClain ME, Pimentel TP (1996) N-15 evidence for the origin and cycling 29. Hastings MG, Sigman DM, Lipschultz F (2003) Isotopic evidence for source changes of of inorganic nitrogen in a small Amazonian catchment. Biogeochemistry 34:45–56. nitrate in rain at Bermuda. J Geophys Res Atmos 108:4790.

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 2of9 Downloaded by guest on September 28, 2021 Fig. S1. Global scale estimates of the ␦15N of soil (A and B) and vegetation (C and D). A and C are based on the multiple regression model Eqs s1 and s2. B and D are derived from S1 and S2 for all areas other than temperate and tropical forest biomes, which are based on the empirical means of plants and soils (Table S1). Black areas were classified as managed (see SI Text) and thus excluded from our estimate of the ␦15N of the unmanaged terrestrial surface in all cases.

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 3of9 Downloaded by guest on September 28, 2021 Table S1. Measured vs. modeled ␦15N across temperate vs. tropical biomes Soil ␦15N Plant leaf ␦15N

Measured (top 10 cm) Measured (Ͼ10 cm) Modeled (area-integrated) Measured Modeled (area-integrated)

Tropical 9.0(Ϯ2.0) 12.0(Ϯ3.5) 6.8(Ϯ2.3) 3.7(Ϯ3.5) 0.43(Ϯ2.5) Temperate 2.0(Ϯ4.0) 4.0(Ϯ2.0) 3.5(Ϯ2.3) Ϫ2.8(Ϯ2.0) Ϫ1.0 (Ϯ2.3)

Errors are standard deviations of the population means and standard errors of the multiple regression models used to estimate ␦15N. Empirical data from ref. 3.

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 4of9 Downloaded by guest on September 28, 2021 Table S2. Comparison of model-based estimates of ␦15N and the model-empirical hybrid Model with measured tropical Global model and temperate forest means

Soil ␦15N 5.43‰ 5.46‰ Leaf ␦15N 0.43‰ 0.53‰

Measured data from ref. 3.

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 5of9 Downloaded by guest on September 28, 2021 Table S3. Literature comparison of ␦15N (per mil vs. air) among leaves, wood, and roots Plant system Leaf Stem Wood Root Bark References

Amazon forests 5.8 4.6 3.5 Ometto et al. (6) Drought deciduous shrubs 1.7 0.75 Ϫ1.0 Kolb et al. (7) Winter deciduous shrubs Ϫ0.4 0.00 1.1 Kolb et al. (7) Desert shrub Ϫ1.5 Ϫ3.1 Ϫ3.0 Shearer et al. (8) Perennial shrub Ϫ1.6 Ϫ2.6 Ϫ2.2 Shearer et al. (8) Norway spruce Ϫ3.0 Ϫ4.0 Ϫ2.0 Gebauer and Schulze (9) Douglas fir Ϫ4.84 Ϫ5.25 Ϫ5.54 Ϫ5.60 Koopmans et al. (10) Scots pine Ϫ2.12 Ϫ1.34 Ϫ1.40 Ϫ3.01 Koopmans et al. (10) Beech Ϫ0.5 1.0 1.0 Templer et al. (11) Hemlock Ϫ0.6 0.3 0.1 Templer et al. (11) Sugar maple Ϫ1.0 0.9 2.7 Templer et al. (11) Red oak Ϫ1.3 0.4 0.6 Templer et al. (11) Global forests Ϫ5to2 Ϸleaf Pardo et al. (4)

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 6of9 Downloaded by guest on September 28, 2021 Table S4. Literature compilation of the ␦15N of bulk nitrate deposition across different terrestrial biomes Site Latitude, longitude Leaching ␦15N reference Soil ␦15N reference Biome type

Kongsfjorden, Svalbard, 78° 54Ј N, 11° 53Ј E Tye and Heaton (32) Tye and Heaton (32) Arctic tundra, permafrost Norway Bavaria, Germany 50° 5Ј N, 11° 51Ј E Durka et al. (22) Gebauer and Schulze (9) Temperate coniferous forest Jena, Germany 50° 55Ј N, 11° 35Ј E Oelmann et al. (33) Oelmann et al. (33) Temperate grassland, (Biodiversity experiment) Green Mts., Vermont 44°19Ј N, 72°53Ј W Hales et al. (25) Hales et al. (25) Temperate deciduous/conifer forest Adirondack Mts., New York 44° 0Ј N, 74° 14Ј W Campbell et al. (34) Mitchell et al. (35) Temperate mixed deciduous forest Catskill Mts., New York 41° 59Ј N, 74° 30Ј W Burns and Kendall (26) Burns and Kendall (26) Temperate mixed deciduous forest Maui, Hawaii 20° 48Ј 55Љ N, 156° 13Ј 47Љ W Houlton et al. (18) Schuur and Matson (36) Tropical montane forest Maui, Hawaii 20° 48Ј 48Љ N, 156° 14Ј 49Љ W Houlton et al. (18) Schuur and Matson (36) Tropical montane forest Maui, Hawaii 20° 48Ј 48Љ N, 156° 14Ј 25Љ W Houlton et al. (18) Schuur and Matson (36) Tropical montane forest Maui, Hawaii 20° 48Ј 30Љ N, 156° 15Ј 0Љ W Houlton et al. (18) Schuur and Matson (36) Tropical montane forest Maui, Hawaii 20° 48Ј 26Љ N, 156° 15Ј 10Љ W Houlton et al. (18) Schuur and Matson (36) Tropical montane forest Maui, Hawaii 20° 48Ј 21Љ N, 156° 15Ј 19Љ W Houlton et al. (18) Schuur and Matson (36) Tropical montane forest Central Amazon, Brazil 2° 56Ј S, 60° 58Ј W Brandes et al. (37) Brandes et al. (37) Tropical lowland forest

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 7of9 Downloaded by guest on September 28, 2021 Table S5. Literature compilation of the ␦15N of N leaching losses relative to soil ␦15N across different biomes Site Latitude, longitude ␦15N nitrate deposition reference Biome type

Yorkshire, U.K. 54° 13Ј N, 01° 54Ј W Heaton et al. (21) Temperate coniferous forest Bavaria, Germany 50° 5Ј N, 11° 51Ј E Durka et al. (22) Temperate coniferous forest Erzgebirge Mts., Czech Republic 50° 35Ј N, 13° 00Ј E Buzek et al. (23) Temperate coniferous forest Turkey Lakes Watershed, Canada 46° 32Ј N 84° 20 W Spoelstra et al. (24) Temperate deciduous forest Green Mts., Vermont 44° 19Ј N, 72° 53Ј W Hales et al. (25) Temperate deciduous/conifer forest Catskill Mts., New York 41° 59Ј N, 74° 30Ј W Burns and Kendall (26) Temperate mixed deciduous forest Rocky Mts., Colorado 40° 17Ј 17Љ 105° 39Ј 43Љ Campbell et al. (27) Alpine tundra/forest/meadow Smokey Mts., Tennessee 35° 58Ј N, 84° 17Ј W Garten (28) Temperate coniferous/deciduous forest Bermuda 32 18Ј N, 64 47Ј W Hastings et al. (29) Subtropical flora Atacama Desert, Northern Chile 20° to 26° S 70° W Michalski et al. (30) Desert Maui, Hawaii 20° 48Ј 55Љ N, 156° 13Ј 47Љ W Houlton et al. (18) Tropical montane forest Maui, Hawaii 20° 48Ј 48Љ N, 156° 14Ј 25Љ W Houlton et al. (18) Tropical montane forest Maui, Hawaii 20° 48Ј 21Љ N, 156° 15Ј 19 Љ W Houlton et al. (18) Tropical montane forest Hawaii, Hawaii 19° 25Ј N, 155° 12Ј W Carrillo et al. (31) Tropical montane forest

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 8of9 Downloaded by guest on September 28, 2021 Table S6. Literature compilation of the isotope effect of denitrification in pure culture and in native soil communities Reference ␧ (‰ vs. air) Type/matrix Biome

Mariotti et al. (12) 29 Pure culture N/A Mariotti et al. (12) 25 Pure culture N/A Barford et al. (13) 29 Pure culture N/A Miyake and Wada (14) 20 Pure culture N/A Olleras (15) 30 Pure culture N/A Delwiche and Steyn (16) 20 Pure culture N/A Delwiche and Steyn (16) 16 Pure culture N/A Delwiche and Steyn (16) 19 Pure culture N/A Delwihe and Steyn (16) 21 Pure culture N/A Delwiche and Steyn (16) 17 Pure culture N/A Granger et al. (17) 23 Pure culture N/A Granger et al. (17) 7 Pure culture N/A Granger et al. (17) 22 Pure culture N/A Granger et al. (17) 23 Pure culture N/A Granger et al. (17) 15 Pure culture N/A Granger et al. (17) 18 Pure culture N/A Granger et al. (17) 21 Pure culture N/A Granger et al. (17) 23 Pure culture N/A Granger et al. (17) 21 Pure culture N/A Granger et al. (17) 23 Pure culture N/A Granger et al. (17) 17 Pure culture N/A Granger et al. (17) 18 Pure culture N/A Granger et al. (17) 20 Pure culture N/A Granger et al. (17) 25 Pure culture N/A Granger et al. (17) 27 Pure culture N/A Granger et al. (17) 24 Pure culture N/A Granger et al. (17) 19 Pure culture N/A Granger et al. (17) 10 Pure culture N/A Granger et al. (17) 5 Pure culture N/A Granger et al. (17) 20 Pure culture N/A Granger et al. (17) 18 Pure culture N/A Houlton et al. (18) 13 Soil/native community Tropical Houlton et al. (18) 13 Soil/native community Tropical Chien et al. (19) 19 Soil/native community Temperate Chien et al. (19) 6 Soil/native community Temperate Blackmer and Bremner (20) 14 Soil/native community Temperate Blackmer and Bremner (20) 23 Soil/native community Temperate Blackmer and Bremner (20) 22 Soil/native community Temperate

Houlton and Bai www.pnas.org/cgi/content/short/0912111106 9of9 Downloaded by guest on September 28, 2021