全球变化知识资源中心

We must make progress in our ability to represent nitrogen (N) in global models if we are to reduce uncertainty in climate change projections and develop more insightful impact scenarios for decision-makers. Nitrogen can both warm and cool the climate system, depending on its form, phase and flux, and interaction with the biosphere's natural CO₂ sinks¹, with non-trivial effects on Earth's heat balance^{2, 3}. For instance, gaseous N emissions from the soil limit the availability of this nutrient for plant CO₂ capture — an indirect warming effect — yet can simultaneously cool global temperatures via the N-based aerosols that alter the planet's reflectance⁴. Once in the atmosphere, gaseous N species can directly increase the Earth's greenhouse effect, particularly when incomplete soil denitrification releases nitrous oxide (N₂O), the third most important greenhouse gas in modern climate change². Moreover, downstream and downwind transport of N accelerates eutrophication, decreases aquatic biodiversity, impairs water- and air-quality for human health, and contributes to N₂O emissions in coastal ecosystems^{1, 5, 6}. A recent assessment⁷ in the European Union (EU27) showed that the externality damages associated with excess N spillovers are roughly equivalent to the gross profits attributable to enhanced food production via N-based fertilizers, at around 100 billion annually.

Terrestrial N fates are therefore vital to many aspects of the environment, society and climate system; but the models used by the IPCC have been criticized for their lack of constraint on terrestrial N balances and loss pathways⁸. We suggest that including the ratios of natural N isotopes (¹⁵N/¹⁴N or δ¹⁵N = [(¹⁵N/¹⁴N_sample)/(¹⁵N/¹⁴N_standard) ] – 1 where the standard is atmospheric N₂) can improve the efficacy of Earth system models generally, and N-based projections of modern climate change in particular. As a case study, we demonstrate here how natural N isotope composition can be used to validate and advance N cycle predictions in the Community Land Model with Coupled Carbon Nitrogen (CLM-CN, hereafter just CLM)⁹. We focused on this model because of its historical importance in setting climate science and policy: CLM was the only model to consider the effect of N in CO₂ and climate change simulations in the Fifth Assessment Report from the IPCC (ref. 2).

We conducted our investigation in two sequential steps. First, we used empirical relationships to project patterns of soil δ¹⁵N throughout the land surface and thereby develop an observation against which the efficacy of global models can be quantitatively appraised. The δ¹⁵N of plant and soil pools varies systematically as a function of mean annual temperature and precipitation (r²= 0.39)¹⁰; hence climate correlations have been widely used to estimate soil δ¹⁵N globally, capturing biome-scale patterns to within ~1‰ of empirical observations and latitudinal differences in soil δ¹⁵N equal to ~10‰ (ref. 11). Such patterns in soil δ¹⁵N reflect N losses to fractionating (denitrification) relative to non-fractionating (leaching) pathways¹¹, with the highest proportions of denitrification (relative to total N losses) observed for desert ecosystems, and lowest denitrification proportions in high-latitude boreal regions where N leaching losses are generally high (Fig. 1a,b).

a,d,g, Proportion of soil nitrogen emissions to denitrification (f_denit) versus total nitrogen emissions (that is, denitrification plus nitrogen leaching) from isotopic modelling (a), CLM 4.0 (d) and CLM 4.5 (g). b,e,h, Frequency distribution of f_denit for the land surface for isotopic modelling (b; N = 20,975), CLM 4.0 (e; N = 14,505) and CLM 4.5 (h; N = 15,337). c,f,i, Soil δ¹⁵N as based on globally projected observations corrected for ammonia volatilization (c; see text), CLM 4.0 (f) and CLM 4.5 (i). Global and spatial inconsistencies between nitrogen isotopic modelling (a–c) and CLM 4.0 (d–f) reveal a high level of discordance in the nitrogen-based forecasts used in aspects of the Fifth Assessment Report of the IPCC. Marginal improvement is seen for CLM 4.5 (g–i). The explicit inclusion of natural isotopic benchmarks into global nitrogen models will allow for more accurate projections of nitrogen-based effects on climate change in future IPCC assessments.

We observe a high level of discordance between CLM's N cycle and empirically projected patterns of soil δ¹⁵N, both globally and spatially within the terrestrial biosphere (compare Fig. 1c with 1f,i). The globally integrated δ¹⁵N of soil predicted by CLM 4.0 is ~13‰, for example, and the newest version of the model, CLM 4.5, provides a slightly lower estimate for soil δ¹⁵N (~11‰). Both of these results greatly exceed empirical estimates of the global mean δ¹⁵N of soil equal to 5.5‰ (ref. 11), thus revealing unrealistically high isotope-fractionating N losses from the land in CLM models. Although ε_denit affects the magnitude of this comparison, we note that the 13‰ isotope effect for denitrification used in our analysis falls at the lower end of laboratory observations¹¹, and so the global disagreement between CLM and empirically projected δ¹⁵N should be taken as a conservative assessment of the model's performance. Using a higher isotope effect for denitrification would only increase the disagreement between CLM-estimated δ¹⁵N and the globally integrated value.

Perhaps more important, CLM predicts invariance in soil δ¹⁵N across the land surface, implying a 'flat-Earth' characterization of N loss pathways from diverse terrestrial ecosystems and conditions (Fig. 1 d,f,g,i). Rather than demonstrating the strong latitudinal gradient in soil δ¹⁵N equal to ~10‰ (refs 6,8; Fig. 1a,c), for example, CLM simulates little to no N isotopic differentiation among Earth's major biomes (Fig. 1f,i). Whereas CLM 4.0 shows no spatial variation at all, CLM 4.5 predicts the lowest δ¹⁵N for desert ecosystems (compare Fig. 1i with Fig. 1c) and exceedingly enriched soil δ¹⁵N for all other terrestrial biomes. This binary pattern and lack of spatial variation is in opposition to thousands of empirical observations of soil δ¹⁵N within the terrestrial biosphere^{10, 15}.

Underlying the lack of conformity to soil δ¹⁵N benchmarks are the exceedingly high denitrification fluxes simulated by the CLM models. In CLM 4.0, for instance, gaseous N losses from the soil account for nearly 100% of terrestrial N outputs (Fig. 1d,e). CLM 4.5 scarcely improves upon this prediction, with dissolved pathways still only accounting for <2% of total N losses from the soil (that is, f_denit = 98%; Fig. 1g,h). CLM thereby over-represents highly fractionating N losses (via denitrification) and in so doing predicts unrealistically high soil δ¹⁵N for the great majority of terrestrial ecosystems. Moreover, CLM 4.5 simulates the highest N leaching loss proportions for desert ecosystems where aridity and negative soil water potentials would greatly preclude such high N leaching losses. In contrast, according to this model, N leaching losses are lowest where water budgets are positive and rivers and streams are perennial in many tropical/subtropical environments.

These results highlight a substantial disconnect between CLM's predictions and our understanding of N emissions from the terrestrial biosphere. The globally integrated isotope model suggests that ~2/3 of the terrestrial N balance can be explained by hydrological N leaching losses¹¹. This agrees with an extensive and profoundly important literature which demonstrates that N leaching losses either dominate or contribute substantially to N balances in many terrestrial ecosystems. Such dissolved N losses are seen in the nitrate concentrations of natural streamwater in the tropics¹⁶ and long-term studies at the Hubbard Brook Experimental Forest¹⁷; dissolved organic N compounds in pristine temperate forest watersheds in South America¹⁸; and spillover of N fertilizers in the hydrosphere, which lead to downstream eutrophication of estuaries and contaminate drinking water⁶, to name but a few. That CLM does not allow for meaningful quantities of hydrological N transport greatly limits its capacity to simulate key processes in the N cycle and critical connections among Earth's land- and water-systems.

Problems with CLM's N-cycle has been raised before^{19, 20}, although not in a spatial context or side-by-side comparison against natural N isotope benchmarks. The lack of an empirical ground-truth scheme has hitherto limited our ability to quantitatively appraise the model's spatial and global performance. A central motivation behind our emphasis on δ¹⁵N is that this natural tracer provides quantitative constraints on gaseous and N leaching losses simultaneously. Natural N isotope composition of soil is simple to measure and it integrates over the time-frame of ecosystem N turnover; thus, models can use soil δ¹⁵N patterns to ensure proper baseline conditions for simulation-forecasting. Building towards a truly predictive understanding of natural and human influences on N emissions from the soil, and their competing radiative forcing effects on the climate⁴, will require many different N-cycle models. We envisage δ¹⁵N as a common standard through which models of various degrees of complexity can communicate with one another.

Hence, we offer a solution to the challenge of improving N in global climate forecasts via natural isotopic benchmarking. As we have demonstrated with our offline simulations, the direct inclusion of natural N isotope benchmarks into Earth system models is relatively straightforward and provides both global and regional constraints that can be incorporated into online simulations of CLM as well (that is, equations (1), (2); see also ref. 12). Moreover, past work has shown that N loss predictions from the widely used DAYCENT model can be validated at smaller watershed-scales using our δ¹⁵N benchmarking technique²¹. The δ¹⁵N approach assumes negligible net N accumulation in plant and soil pools, an assumption that will not hold for every terrestrial ecosystem, particularly those exposed to new disturbance regimes. Further research and more data on soil δ¹⁵N will help to reduce the uncertainty in the N isotope model itself, and an improved understanding of isotope effect expression of denitrification across scales and ecosystems will lead to more accurate estimates of terrestrial N balances via the isotopic benchmarking approach¹¹.

Developing a more thoughtful and accurate forecasting scheme for terrestrial N cycling has implications for climate science and policy development. Nitrogen is a key limiting nutrient that controls CO₂ sequestration in the marine and terrestrial biospheres²². Past work has shown that N limitation of terrestrial CO₂ uptake could result in up to 2 °C of additional warming by 2100 (ref. 3). Uncertainty surrounding the effect of N limitation is large⁸, however, and will ultimately come down to the balance of N in terrestrial ecosystems⁸. In addition, N₂O is the third most important greenhouse gas behind CO₂ and CH₄ (ref. 2). Isotope-based models have provided a baseline for natural N inputs via fixation, constraints on gaseous N emissions (including N₂O, NO, NH₃ and N₂) and hydrological N leaching to downstream ecosystems^{12, 23}, and insights into 'unexplained' atmospheric NO₂ concentrations observed in space-borne satellites over the Sahel region of Africa¹². We suggest that multi-model frameworks that consider human N inputs to agricultural soils can benefit from isotopic benchmarks too, allowing the fate of N to be traced from human sources into natural terrestrial ecosystems, the air we breathe and the water we drink.

1. Sutton, M. A. et al. (eds) The European Nitrogen Assessment (Cambridge Univ. Press, 2011).
2. Ciais, P. et al. in Climate Change 2013: The Physical Science Basis (eds Stocker, T. et al.) 465–570 (IPCC, Cambridge Univ. Press, 2013).
3. Wang, Y. P. & Houlton, B. Z. Geophys. Res. Lett. 36, L24403 (2009).
   • CAS
   • Article
4. Pinder R. et al. Biogeochemistry 114, 25–40 (2013).
   • CAS
   • ISI
   • Article
5. Galloway, J. N. et al. Biogeochemistry 70, 153–226 (2004).
   • CAS
   • ISI
   • Article
6. Vitousek, P. M. et al. Ecol. Appl. 7, 737–751 (1997).
   • ISI
   • Article
7. Brink, C. et al. The European Nitrogen Assessment (eds Sutton, M. A. et al.) 513–540 (Cambridge Univ. Press, 2011).
8. Hungate, B., Dukes, J., Shaw, M., Luo, Y. & Field, C. Science 302, 1512–1513 (2003).
   • CAS
   • ISI
   • PubMed
   • Article
9. Thornton, P. E., Lamarque, J-F., Rosenbloom, N. A. & Mahowald, N. M. Global Biogeochem. Cycles 21, GB4018 (2007).
   • CAS
   • Article