Nitrous oxide emissions are enhanced in a warmer and wetter world

Significance

N₂O has 300 times the global warming potential of CO₂ on a 100-y timescale, and is of major importance for stratospheric ozone depletion. The climate sensitivity of N₂O emissions is poorly known, which makes it difficult to project how changing fertilizer use and climate will impact radiative forcing and the ozone layer. Here, atmospheric inverse analyses reveal that direct and indirect N₂O emissions from the US Corn Belt are highly sensitive to perturbations in temperature and precipitation. We combine top-down constraints on these emissions with a land surface model to evaluate the climate feedback on N₂O emissions. Our results show that, as the world becomes warmer and wetter, such feedbacks will pose a major challenge to N₂O mitigation efforts.

Abstract

Nitrous oxide (N₂O) has a global warming potential that is 300 times that of carbon dioxide on a 100-y timescale, and is of major importance for stratospheric ozone depletion. The climate sensitivity of N₂O emissions is poorly known, which makes it difficult to project how changing fertilizer use and climate will impact radiative forcing and the ozone layer. Analysis of 6 y of hourly N₂O mixing ratios from a very tall tower within the US Corn Belt—one of the most intensive agricultural regions of the world—combined with inverse modeling, shows large interannual variability in N₂O emissions (316 Gg N₂O-N⋅y⁻¹ to 585 Gg N₂O-N⋅y⁻¹). This implies that the regional emission factor is highly sensitive to climate. In the warmest year and spring (2012) of the observational period, the emission factor was 7.5%, nearly double that of previous reports. Indirect emissions associated with runoff and leaching dominated the interannual variability of total emissions. Under current trends in climate and anthropogenic N use, we project a strong positive feedback to warmer and wetter conditions and unabated growth of regional N₂O emissions that will exceed 600 Gg N₂O-N⋅y⁻¹, on average, by 2050. This increasing emission trend in the US Corn Belt may represent a harbinger of intensifying N₂O emissions from other agricultural regions. Such feedbacks will pose a major challenge to the Paris Agreement, which requires large N₂O emission mitigation efforts to achieve its goals.

Global demand for synthetic nitrogen (N) fertilizer is projected to exceed 119 Tg (1 Tg = 10¹² g) N in 2018—a 29% increase since 2008 (1). The current trajectory of N fertilizer demand is exceeding some of the most aggressive forecasts (2). Globally, about 4% of the anthropogenic N in agricultural systems is returned to the atmosphere as nitrous oxide (N₂O) (3)—a potent greenhouse gas (4) and the predominant stratospheric ozone-depleting emission (5). Growing population and demand for food, fiber, and biofuel are accelerating the use of synthetic N and N₂O emissions (6, 7).

In recent decades (1987–2012), N fertilizer use within the US Corn Belt has averaged (mean ± 1 SD) 6.2 ± 0.9 Tg N (8), with a significant increasing trend (+0.08 Tg N⋅y⁻¹, P < 0.01) (SI Appendix, Fig. S1). With each passing decade, an additional 1 Tg of new synthetic N is added to this intensively managed region. Biological nitrogen fixation has also increased over the past three decades as a result of increasing soybean production (SI Appendix, Fig. S1). This implies that N₂O emissions should be increasing by at least 4 Gg N₂O-N⋅y⁻¹. The production and application of organic N (as manure) in the Corn Belt has been relatively stable at 2.0 ± 0.15 Tg N over the same period, but has been increasing globally along with animal protein consumption, thereby enhancing N₂O emissions (9). However, the interannual variability of N₂O emissions from the major agricultural regions of the world and the extent to which climate variations modulate emissions remains poorly known (10, 11).

Climate models and observations indicate that convective precipitation is enhanced in a warmer world, with wet areas becoming wetter (12). These warmer and wetter conditions are expected to enhance N₂O production via (de)nitrification (13). Within the US Corn Belt, air temperature and precipitation have increased over the period 1895–2015 by 0.67 °C per century and 75 mm per century, respectively (14). We posit that increasing trends in anthropogenic N and warmer and wetter conditions in agricultural regions are enhancing N₂O emissions and that these trends will be amplified via positive feedback to climate change.

Here, we analyze high-precision hourly N₂O mixing ratios measured at a tall tower in Minnesota (100 m above ground level) from April 2010 to April 2016 (SI Appendix, Figs. S2 and S3) to examine the sensitivity of N₂O emissions to weather and climate phenomena within the US Corn Belt. We use Bayesian inverse modeling to constrain the total regional N₂O emissions for each year and to partition the emissions into direct and indirect components (15), defined here as emissions from cropped and grazed lands and emissions occurring off-site, respectively. Indirect emissions result from leaching and runoff of fertilizer and manure N as well as deposition of reactive N (16) and have been shown to be significantly larger than previously reported for this region (15, 17). These improved constraints on N₂O emissions were then used to optimize the Community Land Model to evaluate climate feedbacks on N₂O emissions from present to the year 2050.

Results and Discussion

Atmospheric inversions (15) reveal that the interannual variability of N₂O emissions is large, varying from 316 Gg N₂O-N⋅y⁻¹ to 585 Gg N₂O-N⋅y⁻¹ (mean = 411 ± 113 Gg N₂O-N⋅y⁻¹) (Fig. 1). Further, direct (192 ± 39 Gg N₂O-N⋅y⁻¹) and indirect (220 ± 75 Gg N₂O-N⋅y⁻¹) emissions are similar on average (15, 17), but indirect emissions contribute more to the interannual variability of the regional N₂O budget. Indeed, the interannual variability of nitrate export from the Mississippi River has also been shown to be large, varying from 0.02 T nitrate N to 0.08 Tg nitrate N, and is highly sensitive to variations in winter and spring precipitation (18). The export of nitrate from runoff and leaching is known to be an important control on dissolved N₂O concentrations and fluxes from riverine systems within the region (17, 19), and we expect increases in winter and spring precipitation to amplify the indirect emissions.

Fig. 1.

Interannual variability of N₂O emissions from the US Corn Belt based on Bayesian atmospheric inverse analyses. (A) Total emissions from 2010 to 2015. (B) Direct emissions. (C) Indirect emissions. The error bars in A represent the sensitivity of the total Bayesian inversion estimate to model assumptions related to background and tall tower concentration uncertainty, data quality control, and model transport processes.

The large interannual variability in the N₂O budget, despite relatively consistent rates of synthetic N input, implies that emission factors have high sensitivity to climate. The largest N₂O emissions were measured in 2012 (585 Gg N₂O-N⋅y⁻¹), which was characterized by the warmest mean annual air temperature (11.4 °C, +2.4 °C warmer than the climate normal) during our observations, and the warmest spring (March through May, 13.9 °C, +4.5 °C warmer than normal) in the last 122 y for the Corn Belt (14). Although 2012 was also one of the driest years on record (648 mm average, rank 10 of 122 y), the antecedent precipitation in the previous 2 y and during the winter months of 2011–2012 were above normal (32 mm to 130 mm above normal) (14), resulting in significant spring stream flow (SI Appendix, Fig. S8). Nitrous oxide emissions were remarkably consistent from 2013 to 2015, likely from compensating effects among environmental drivers. For instance, indirect emissions in June 2014 were significantly enhanced by precipitation, whereas direct emissions in June 2013 increased because of a combination of antecedent precipitation and relatively warm temperatures (SI Appendix, Fig. S9). Consideration of the N inputs to the system indicates a mean regional emission factor of 5.3 ± 1.2%, which is in good agreement with the global emission factor of 4% (3) that is based on top-down analyses. However, the emission factor reached a maximum of 7.5% in 2012, nearly double the previous estimates for this region (16, 20).

There are three modes of seasonal behavior (Fig. 2): spring thaw (DOY 50 to 120; February 19 to April 30); the priming of agricultural systems with spring N fertilizer and early season crop growth (DOY 135 to 200; May 15 to July 19); and episodic behavior in late fall and winter (DOY 300 to 365; October 27 to December 31) driven by fall application of N and freeze−thaw activity (21). Spring thaw has been recognized as a dominant control on the annual N₂O budget (21, 22) and is shown here to be a regional phenomenon concentrated within a 70-d window, typically initiated in mid-March. Considering the mean fluxes over the 6-y period (Fig. 1), spring thaw accounted for 30% of the annual budget, while the early growing season (May to July) accounted for about 53% of the annual emissions. Thus, over 80% of the annual N₂O emissions occurred within a period of about 150 d. The seasonality of the direct and indirect emissions was not in perfect phase. Indirect emissions have a more pronounced seasonal peak in May and June, which reflects the temporal dynamics and interaction of synthetic N application with runoff and stream flow within the region (15, 17). In the Corn Belt, most synthetic N fertilizer is applied in April (23), with peak tile flow and stream discharge occurring approximately April through June (24). Thus, the large total emission in 2012 was driven by large direct emissions in an exceptionally warm April followed by relatively large indirect emissions in May.

Fig. 2.

Seasonal emission patterns and dominant modes of temporal variability in the tall tower N₂O mixing ratio observations extracted using wavelet techniques. (A) Ensemble (2010–2015) hourly average N₂O mixing ratios observed at the tall tower 100-m level. (B) Wavelet analysis of N₂O mixing ratios showing power in time-frequency space.

The sensitivity of N₂O emissions to climate drivers, including soil temperature and soil water content, has been studied under laboratory and field conditions (25–27), and these functions have been incorporated into land surface models such as the Community Land Model (CLM45-BGC-CROP). These empirical functions have been shown to have low statistical power under field conditions, a consequence of episodic behavior related to hot spots and hot moments (28). Tall tower mixing ratios and soil temperatures display moments of high correlation where the signals are in phase (SI Appendix, Fig. S10). We extracted periods of high wavelet coherence (R² > 0.8) from the multiyear time series to estimate the sensitivity of boundary layer N₂O mixing ratio to changes in air temperature and water-filled pore space (WFPS) (SI Appendix, Fig. S11 and Table S5). The linear sensitivity to air temperature during snow melt, growing season, and annually was 0.0488 parts per billion (ppb) per degree Celsius, 0.1710 ppb per degree Celsius, and 0.0404 ppb per degree Celsius , respectively. A nonlinear response in N₂O mixing ratios was observed at approximately −1.0 °C to 10.0 °C during spring thaw, corresponding to strong N₂O production near the 0 °C threshold and rapidly decreasing above 10 °C, presumably from increasing activity of N₂O reductase (29). Using this technique, the sensitivity of boundary layer N₂O mixing ratios to soil temperature was examined for conditions when WFPS limited N₂O production (26) (i.e., WFPS < 45%) and when it was near-optimal (i.e., 60% < WFPS <80%). These wavelet analyses revealed large differences in the planetary boundary layer N₂O mixing ratios, yielding temperature sensitivities of 0.0189 ppb per degree Celsius and 0.0583 ppb per degree Celsius, respectively. Similar wavelet analyses indicated weak but positive relationships between boundary layer N₂O mixing ratios and daily precipitation amount, indicating enhanced N₂O emissions for warmer and wetter climate conditions.

These sensitivity factors were applied to an equilibrium boundary layer budget analysis (16) to assess the potential impact of climate on the regional N₂O budget (Fig. 3). The 2010 equilibrium boundary layer budget of 420 Gg N₂O-N⋅y⁻¹ (16) was similar to the mean emissions (411 Gg N₂O-N⋅y⁻¹) derived using the atmospheric inversions from 2010 to 2015 (Fig. 1) and is used here as a baseline value. The mean annual and spring temperature anomalies observed in 2012 demonstrate a change in the 2012 budget ranging from 20 Gg N₂O-N⋅y⁻¹ to 120 Gg N₂O-N⋅y⁻¹, for the nonlimiting WFPS cases, giving an annual emission estimate ranging from 431 Gg N₂O-N⋅y⁻¹ to 531 Gg N₂O-N y⁻¹. These equilibrium boundary layer budget analyses provide independent support for the Bayesian atmospheric inversion budget estimate of 585 Gg N₂O-N⋅y⁻¹. Further, these equilibrium boundary layer budget analyses indicate that the temperature sensitivity for the optimal WFPS function and the mean ensemble of the temperature sensitivity factors (Fig. 3A) yielded a Q₁₀ temperature coefficient of 2.0. These regionally derived values are at the low end of reported Q₁₀ values (2.1 to 8.1) based on flux chamber observations made in agricultural systems and laboratory microcosm experiments using agricultural soils (25–27).

Fig. 3.

Sensitivity of regional N₂O emissions to air temperate and WFPS derived from cross-wavelet coherence and equilibrium boundary layer budget analyses. (A) Sensitivity of the US Corn Belt N₂O budget to air temperature for different time periods and soil water content status. (B) Estimated perturbations in the US Corn Belt N₂O budget based on the long-term air temperature record of the Corn Belt and the ensemble of sensitivity factors shown in A. The shading represents 1 SD of the ensemble of sensitivity factors.

We applied the mean ensemble of the nonlimiting WFPS air temperature sensitivity factors to the annual air temperature record for the US Corn Belt for the synthetic N intensive period (1970–2015) to examine perturbations in the regional N₂O budget and to identify possible trends over the past 46 y (Fig. 3B). Year-to-year fluctuations (peak-to-trough) are on the order of 65 Gg N₂O-N⋅y⁻¹, and there is a significant (P = 0.023) increasing trend of 1.7 Gg N₂O-N per decade (or about 7.6 Gg N₂O-N over the last 46 y) driven by increasing air temperature. These analyses imply that the interannual variability is large and that the change in N₂O emissions due to air temperature over this period is about 8% of that caused by the long-term trend of increasing synthetic N fertilizer use.

Simulations based on the biophysical model, CLM45-BGC-CROP, were used to help diagnose the sensitivity and feedback of the Corn Belt N₂O emissions to climate. Here, CLM45-BGC-CROP was modified to simulate direct and indirect emissions, and was optimized using the 2015 monthly emissions obtained from the inverse Bayesian analyses (Fig. 1) (SI Appendix, Fig. S12 and Tables S6 and S7). We then ran the model for the years 2010–2015 using this optimization. Because land use and synthetic N input were held constant in these simulations, the emission variability can be ascribed to variation in climate. The 2010–2015 simulations for direct (216.5 ± 42.4 Gg N₂O-N⋅y⁻¹), indirect (199.0 ± 57.0 Gg N₂O-N⋅y⁻¹), and total (415.5 ± 90.8 g N₂O-N⋅y⁻¹) emissions showed very good agreement with the Bayesian estimates. Model and observational constraints of the climate sensitivity showed good agreement. The CLM45-BGC-CROP Q₁₀ temperature coefficient ranged from 1.1 to 1.6, depending on the WFPS, and was in good agreement with that derived from the tall tower wavelet analyses. The simulated emissions and observations were highly correlated over the 6-y period (R² = 0.86, P = 0.0062) and support that emissions in 2012 were anomalously large due to the extremely warm spring.

The model was run forward in time to 2050 using the Coupled Model Intercomparison Project Phase 5 (CMIP5) [scenario RCP8.5 (representative concentration pathway of +8.5 watts per square meter radiative forcing by the year 2100)] forcing data, while holding land use and synthetic N fertilizer input constant. For this long-term simulation, CLM45-BGC-CROP was reoptimized using all of the available emission data from the tall tower Bayesian analyses. Fig. 4 shows a general increasing trend in indirect (+0.87 Gg N₂O-N⋅y⁻¹, P = 0.004), direct (+0.13 Gg N₂O-N⋅y⁻¹, P = 0.81), and total (+1.0 Gg N₂O-N⋅y⁻¹, P = 0.17) N₂O emissions driven by climate, with an increasing trend in annual air temperature (0.038 °C⋅y⁻¹, P < 0.001) and precipitation (2.7 mm⋅y⁻¹, P < 0.001) for the US Corn Belt. The dominant driver of the increased annual emissions is the link between increasing precipitation and indirect emissions. The seasonal variability indicates that substantially warmer temperatures in April and wetter antecedent conditions in spring (i.e., increased wintertime precipitation in December and February) are important drivers of the increasing spring runoff (SI Appendix, Fig. S13) and emissions. However, there is also an increase in summer and late fall emissions. Considering the individual trends in synthetic N application (SI Appendix, Fig. S1) and climate suggests that emissions could increase from the Corn Belt by about 200 Gg N₂O-N, with total average emissions exceeding 600 Gg N₂O-N by 2050. These model results indicate that the long-term increase in N₂O emissions due to climate change is about 25%. Thus, using two different constraints (modeling and planetary boundary layer observations and sensitivity analyses), we found that the climate sensitivity of N₂O emissions accounts for about 8 to 25% of the long-term trend.

Fig. 4.

Simulation of N₂O emissions using CLM45-BGC-CROP for the US Corn Belt from 2011 to 2050. Simulations were performed using the CMIP5 (scenario RCP8.5) forcing data, while holding land use and synthetic N fertilizer input constant. (A) Trend in the annual total, indirect, and direct N₂O emissions. (B) Trend in annual air temperature. (C) Trend in annual precipitation. (D) Monthly trend in total N₂O emissions. (E) Monthly trend in air temperature. (F) Monthly trend in precipitation. In D–F, the shading indicates the 95% confidence intervals.

Although the long-term N₂O emission trends are derived primarily from the increased use of synthetic N (3, 7, 11), the large observed interannual variability of N₂O emissions and the model-simulated trends support that positive feedbacks associated with climate change will be important. These temporal dynamics pose a major challenge to mitigation efforts and emissions accounting. Indeed, Davidson (30) has shown that a 50% reduction in N₂O emission factors and per capita meat consumption would be required to stabilize emissions within representative concentration pathways outlined in the fifth assessment of the Intergovernmental Panel on Climate Change. The positive climate feedback effect implies that even greater reductions in synthetic N application or increased nitrogen use efficiency (6) will be required to balance emission budgets as outlined in the Paris Agreement. Such reductions will be difficult to achieve given the increasing demand for food, fiber, and biofuel (6, 7), and the fact that grain yields are likely to decline with increasing air temperatures (31). Increasing runoff and N leaching due to increased precipitation are likely to exacerbate the situation (17). Finally, evidence is emerging that mitigation strategies related to N fertilizer management, e.g., delaying the timing of fertilizer application (32), or using microbial inhibitors (33, 34), are made less reliable under warmer, wetter, and more variable climate. However, we cannot rule out that warmer and wetter conditions could benefit other mitigation options such as the use of winter cover crops (35) or perennial companion crop systems (36). Here, the growth of cover crops in early spring could reduce the available soil nitrate and reduce N₂O emissions. Results from these cover crop studies, however, have produced inconsistent results. The potential impact of climate change on N₂O mitigation strategies remains speculative, and requires further research.

Materials and Methods

Tall Tower Observations.

Tall tower N₂O mixing ratios were measured at the Minnesota Public Radio communications tower, KCMP (44° 41′ 19′′ N, 93° 4′ 22′′ W; 290 m above sea level), using a tunable diode laser technique (TGA100; Campbell Scientific Inc.) at four heights including 32, 56, 100, and 185 m above the ground (16). These measurements were initiated in April 2010 and are on-going as of September 2017. Calibrations were performed hourly with gases traceable to the National Oceanic and Atmospheric Administration−Earth System Research Laboratory (NOAA-ESRL) 2006A N₂O mole fraction scale. Details regarding calibration and data filtering are provided in SI Appendix.

Inverse Modeling.

The atmospheric inverse modeling approach used in this study is based on Chen et al. (15), who partitioned regional N₂O emissions from the US Corn Belt into direct and indirect emissions. A scale factor Bayesian inverse (SFBI) method was applied for each month (April to October from 2010 through to 2015). Briefly, this methodology consists of the following key steps: (i) The tall tower concentration footprints for the 100-m air sample level (receptor) were estimated for each hour using the Stochastic Time-Inverted Lagrangian Transport (STILT) model. The Weather Research and Forecasting model (version 3.5.1) provided input to STILT, including wind conditions, atmospheric stability, and planetary boundary layer height. (ii) The tall tower concentration footprints were multiplied by the a priori emission estimates to obtain an initial guess of the N₂O mixing ratio enhancement at the tall tower receptor. The a priori emissions were obtained from the Emission Database for Global Atmospheric Research (EDGAR) 4.2 and EDGAR2.0 databases; CLM45-BGC-CROP model, Global Fire Emissions Database (version 3, 2011, www.globalfiredata.org), and the upscaling of observations of indirect N₂O emissions from rivers and streams (17). (iii) The estimate of the tall tower N₂O mixing ratio enhancement was then added to the background N₂O mixing ratio. Background N₂O mixing ratios were estimated based on observations from the NOAA Carbon Cycle and Greenhouse Gases program (37) and were obtained at the outer edge of the concentration source footprint. (iv) Finally, the SFBI was used to optimize the a priori emissions along with the relative contributions of the direct and indirect emissions. A detailed description of the SFBI is provided in SI Appendix, and the sensitivity of the inversion results to a priori source types, tall tower data quality, tall tower data filtering, a priori emission errors, and a priori indirect emission source estimates are provided in SI Appendix, Tables S1–S4. Note that the partitioning of total N₂O emissions into indirect and direct emissions is sensitive to the prescribed uncertainties of the a priori emissions. For instance, the average (for years 2010–2015) indirect emissions varied from 154.5 Gg N₂O-N⋅y⁻¹ to 200.2 Gg N₂O-N⋅y⁻¹ for a 200 to 400% change in the a priori uncertainty when the direct emission a priori uncertainty was fixed at a nominal value of 66% (SI Appendix, Table S3).

Wavelet Analyses.

Nitrous oxide mixing ratios and associated environmental variables were analyzed using techniques based on the continuous wavelet transform (CWT). All CWTs were calculated on the fluctuating component of the signal using the complex Morlet wavelet basis with the nondimensional frequency (ω₀) set to 6 (38). Scales were set to have a minimum of 2 h (i.e., twice the hourly averaging interval), and to have 12 suboctaves per octave. Calculating the CWT of the signal yields a set of wavelet coefficients, W_n(s), spanning all times (n) and scales. We examined the relation between N₂O mixing ratios in the planetary boundary layer (PBL) and environmental drivers by applying the multivariate technique known as wavelet coherence analysis to identify correlation and phase relationships among variables (39). Here, we applied a diurnal filter (24-h running mean) to the N₂O mixing ratio data to reduce the effects of diurnal PBL growth dynamics on the variations of N₂O. We used a minimum coherence threshold (R² = 0.80) to select relationships among environmental factors. Linear regression was then used to quantify these relationships and estimate climate sensitivity factors. See SI Appendix for further details.

Boundary Layer Budget Sensitivity.

We used the equilibrium boundary layer budget theory as a framework to help diagnose how observed changes in temperature and WFPS influence the N₂O budget as

$$F_{\text{EBL}}=\underset{0}{\overset{h}{{\displaystyle \int }}}\rho \frac{dc}{dt}dh+S\left(C_t-C_m\right),$$ [1]

where F_EBL is the surface flux, S represents the subsidence of air from the free troposphere into the boundary layer (units of moles per square meter per second and positive toward the surface), and C_t and C_m represent the mixing ratios of N₂O in the free atmosphere and mixed layer, respectively. For the sensitivity analyses shown in Fig. 3, we use the values of S and C_t for 2010 (16) with C_m = C_t + C_m(T). Here, C_m(T) represents the air temperature sensitivity factors shown in SI Appendix, Fig. S11.

Land Surface Model Projections.

The CLM45-BGC-CROP model was modified to simulate direct N₂O emissions based on previously published algorithms (40, 41), with indirect emissions constrained using model estimates of nitrate runoff/leaching and applying the most recent estimates of the runoff/leaching emission factors for the Corn Belt region (15). The model was then optimized using the 2015 monthly Bayesian N₂O emissions by adjusting the timing and amount of fertilizer applied to croplands and by adjusting the model’s nitrification parameters (i.e., the maximum fraction of NH₄⁺ nitrified per day) (SI Appendix, Fig. S12 and Tables S6 and S7). CLM45-BGC-CROP was run for 40 y (2011–2050), forced with atmospheric data from CMIP5 (42). The forcing data were simulated using a global climate model developed by the Meteorological Research Institute (MRI), MRI-CGCM3 (43). The CMIP5 simulations represent the RCP8.5 scenario run for years 2006–2080 with a spatial resolution of 1.125° × 1.125°. In the model, 3-hourly air temperature, wind, incident solar radiation, precipitation, and specific humidity were used to drive CLM45-BGC-CROP.