The advent and industrialization of the Haber Bosch process in the early twentieth century ushered in a new era of reactive nitrogen distributions on Earth. Since the appearance of the first commercial scale Haber Bosch fertilizer plants, fertilizer application rates have greatly increased in the U.S. While the contributions of fertilizer runoff to eutrophication and anoxic dead zones in coastal regions have been well-documented, the potential influences of increased fertilizer applications on air quality and precipitation chemistry are poorly constrained. Here we combine a 255-year record of precipitation nitrate isotopes preserved in a Greenland ice core, historical reconstructions of fertilizer application rates, and field characterization of the isotopic composition of nitrogen oxides produced biogenically in soils, to provide new constraints on the contributions of biogenic emissions to North American NOx inventories. Our results indicate that increases in twentieth century commercial fertilizer use led to large increases in soil NO, a byproduct released during nitrification and denitrification reactions. These large shifts in soil NO production are evidenced by sharp declines in ice core δ15N-NO3− values. Further, these results suggest that biogenic NOx emissions are underestimated by two to four fold in the U.S. NOx emission inventories used to construct global reactive nitrogen budgets. These results demonstrate that nitrate isotopes in ice cores, coupled with newly constrained δ15N-NOx values for NOx emission sources, provide a novel means for estimating contemporary and historic contributions from individual NOx emission sources to deposition.
 Nitrogen oxides (represented by NO + NO2 for the purposes of this study) emissions can have detrimental effects on the environment and human heath [Galloway et al., 2004]. While the atmospheric lifetime of NOx can be less than 24 h and is mainly lost to HNO3 formation and aerosol uptake [Lamsal et al., 2010], the HNO3 formation product has a lifetime of up to 5 days allowing it to be transported and deposited over large regions [McElroy, 2002]. Thus, HNO3 formation and transport can deposit NOy (NOx oxidation products) far from the point of emission [Elliott et al., 2007]. As a consequence, targeted emission reductions in Europe and the U.S. aim to ameliorate associated environmental and human health impacts. For example, recently the U.S. EPA strengthened the health-based National Ambient Air Quality Standard for NOx [US EPA, 2010] and the European Commission's Thematic Strategy on Air Pollution  has objectives of reducing year 2000 levels of NOx emissions by 60% before 2020. While the primary source of NOx is fossil fuel combustion (vehicles, power plants), there are significant natural NOx sources including lightning, biomass burning, and biogenic emissions produced during nitrification and denitrification reactions in soils. Given the short lifetime of NOx and the heterogeneity of sources in space and time, NOx emission inventories are notoriously difficult to evaluate. As a consequence, emission inventories from area sources, such as lightning and soil emissions, are subject to large uncertainties. For instance, Holland et al.  report a global soil NOx emission range of 4–21 Tg N yr−1 and recent studies have reported a lightning-produced NOx range of 1 to 20 Tg yr−1 [Schumann and Huntrieser, 2007].
 Recent advances in characterizing nitrogen isotope ratios of NOx sources (δ15N-NOx) can be used to help constrain emission inventories. Recent studies report power plant and vehicle emissions δ15N-NOx values of +10 to +20‰ and +3.7 to +5.7‰, respectively [Felix et al., 2012; Ammann et al., 1999; Pearson et al., 2000; Redling et al., submitted]. In comparison, biogenic soil emissions have lower δ15N-NOx values −48.9 to −19.9 [Li and Wang, 2008]. These relatively large differences in δ15N-NOx values allow the use of isotope mixing models to clarify NOx source apportionment in gases, aerosols, and resulting nitrate deposition. For example, δ15N-NO3− in precipitation across the Northeastern U.S. is strongly correlated with NOx emissions from electricity generating units within 400 km of rainwater monitoring sites [Elliott et al., 2007].
 Prior investigations suggest that nitrate isotopes in ice cores also record temporal changes in NOx source contributions. Hastings et al.  report a clear change in δ15N-NO3− in a Greenland ice core over the last 255 years and suggest changes NOx sources that contribute to HNO3 formation over time. Here building on these results, we hypothesize that twentieth century increases in fertilizer use led to higher fluxes of biogenic NOx from soils; these changes are consequently recorded as negative δ15N-NO3− excursions in a Summit, Greenland, ice core (72.5°N, 38.4°W). We examine this hypothesis by reconstructing historical rates of fertilizer application and biomass burning, isotopically characterizing NOx emitted from conventionally fertilized agricultural fields and developing a multiple source mixing model that constrains modern and historic NOx emission fluxes.
 We compare the historical record of δ15N-NO3− values in a Summit, Greenland, ice core with historic land use records, biomass burning data, fertilizer application data, and δ15N-NOx source characterization data to investigate sources of NOx to a remote location in Greenland. Approximate δ15N-NOx source values were derived from literature values with the addition of the biogenic NOx emissions characterized herein and described below [Felix et al., 2012; Ammann et al., 1999; Pearson et al., 2000; Redling et al., submitted; Li and Wang, 2008].
2.1 Reconstruction of Agricultural History and Air Mass Trajectories
 Fertilizer consumption data (1850 to 1890 decadal, 1891 to 2005 annual) and farmland acreage data (1850 to 1910 decadal, 1911 to 2005 annual) was obtained from the U.S. Department of Commerce reports [US Department of Commerce, 1975; US Department of Commerce Census 1970–2010]. Biomass burning data (billion btu produced commercial and residential sector) were obtained from the U.S. EIA . Evidence for biomass burning NOx contributing to δ15N-NO3− was obtained by comparing δ15N values with a record of black carbon and vanillic acid concentrations in a Central Greenland ice core [McConnell et al., 2007]. North American CO2 emissions [Marland et al., 2008] were correlated with δ15N-NO3− to infer contributions from fossil fuel combustion NOx.
 Given patterns in air mass trajectories, fertilizer consumption data for the European Union [Fertilizers Europe, 2010] was not considered explicitly in this interpretation of the δ15N-NO3− ice core data. Specifically, the primary air masses (85%) traveling over Greenland during the spring fertilizer application period originate in the American Midwest [Kahl et al., 1997]. This is the case for 700 hPa back trajectory, the closest to the actual altitude of Summit, Greenland.
2.2 Assumptions Made to Reconstruct Emissions Contributing to NO3− Deposition in Greenland.
 A δ15N-NOx mixing model equation was created using the δ15N value of combined biomass burning/fossil fuel NOx emissions and the δ15N value of biogenic NOx, such that δ15N-NOx ice core = (1 − ƒbiogenic) * (δbiomass burn/fossil fuel) + (ƒbiogenic * δbiogenic). From 1750 to 1850, we assume that the NOx source contributing to ice core δ15N-NO3− values is biomass burning. From 1850 to 1920, we assume that the main sources of NOx contributing to ice core δ15N-NO3− values are biomass burning and fossil fuel. This results in an average δ15N-NO3− value of +9.5‰ and a δ15N-NOx value of +12.4‰ (where δ15N-NO3− values are from Hastings et al.  and δ15NOx values are calculated using δ15N-NO2 = (δ15N-HNO3 + (1000 * (1 – 0.9971)))) [Freyer, 1991].
 The biogenic δ15N-NOx value was obtained from Li and Wang  who reported a biogenic δ15N-NO range from urea fertilized soil of −49 to −20‰ (urea fertilizer δ15N = 1.3‰). This laboratory experiment sampled NO emissions from fertilized soil daily over 13 days, and the range in δ15N is attributable to lower δ15N-NO released after initial fertilizer application because the lighter 14N of NO will be reacted first by the soil bacteria. The heavier 15N will only be more readily used as the residual fertilizer pool becomes more enriched in 15N. This range also suggests that the biogenic δ15N-NOx value may be dependent on the δ15N value of the fertilizer applied. Since fertilizer is made from air (N2 = 0‰), commercial fertilizer δ15N values tend deviate slightly around 0 ± 2‰ [Bateman and Kelly, 2007]. NOx produced from lightning is not considered separately, as annual contributions are assumed to be consistent over the ice core record.
2.3 Characterization of the δ15N-NOx Value of Biogenic NOx
 To constrain the large range of biogenic δ15N-NOx reported by the Li and Wang laboratory study, we collected biogenic NO2 at a heavily instrumented conventionally managed cornfield at the Beltsville Agricultural Research Center in Beltsville, MD. The cornfield is managed to represent conventionally managed agriculture in the U.S. fertilized with Urea/Ammonium/Nitrate (UAN). For this study, after a 120 lb N/acre UAN application, an Ogawa passive NO2 sampler containing a glass fiber filter coated with triethanolamine was placed in a Teflon flux chamber deployed into the fertilized soil. The sampler was deployed in the chamber for 1 month. The resulting δ15N-NO2 of −27‰ is within the range reported by Li and Wang but rather represents an integrated δ15N value of biogenic NO2 emissions over 1 month in an actual field setting. Given that this δ15N-NO2 end-member represents an average δ15N value from fertilized soil over time in a field setting, this value was used in the mixing model as a biogenic NOx end-member.
3 Results and Discussion
 Historic U.S. fertilizer application rates increase dramatically after the advent of the Haber-Bosch method in the early twentieth century [US Department of Commerce, 1975; US Department of Commerce Census 1970–2010]. Fertilizer application rates from 1890 to 2005 are strongly, negatively correlated with ice core δ15N-NO3− for equivalent years (R2 = 0.87, p < 0.0001) (Figure S1a). This strong correlation, coupled with our observed δ15N-NOx values of biogenic soil emissions reported herein, indicates that the negative trends observed in ice core δ15N may be attributable to the intensification of soil NOx emissions stemming from widespread application of industrial fertilizers starting after 1920. This hypothesis is further constrained by long-term studies of air mass trajectories to Greenland that conclude that 85% of springtime (when biogenic emissions are likely the greatest) air masses over Greenland originate from North America [Kahl et al., 1997]. Together, these findings suggest that temporal changes in ice core δ15N-NO3 values reflect shifting NOx sources in the U.S. largely driven by the advent of industrial fertilizer application (Figure 1). In the following discussion, we outline relative changes in potential contributions from three major NOx sources (biomass burning, fossil fuel combustion, and biogenic soil emissions) for periods pre-Industrial Revolution and pre– and post–Haber-Bosch process. This is followed by quantitative prediction of contributions from these individual sources over time using a mixing model.
 From 1750 until the onset of the Industrial Revolution in 1850, biomass burning is the primary NOx source contributing to δ15N-NO3− values. This is supported by a strong correlation between black carbon concentrations and vanillic acid data prior to 1850 in a prior Greenland ice core study [McConnell et al., 2007]. This suggests that conifer burning is the primary source of black carbon to Greenland [McConnell et al., 2007]. If so, this indicates a biomass burning δ15N-NO3− end-member value of +11.5‰ and δ15N-NOx value of +14.4‰. The δ15N values represent a mixture of biomass burning and natural background NOx sources (i.e., biogenic soil and lightning), but biomass burning is assumed to be the primary source represented. The representative biomass burning δ15N values fall within the range (+10.6 to +25.7‰) of δ15N values of total nitrogen on aerosol particles collected from biomass burning plumes [Hastings, 2010].
 From 1850 to 1920, the main NOx sources contributing to δ15N-NO3 ice core values are biomass burning and fossil fuel combustion, corresponding to the onset of the Industrial Revolution in 1850. Accordingly, correlations between North American CO2 emissions [Marland et al., 2008] and ice core δ15N-NO3− values are significant between 1850 and 1920 (R2 = 0.23, p < 0.05) but not prior to this time (1750 to 1850; R2 = 0.02, p = 0.6). This indicates that fossil fuel NOx emission contributions to ice core δ15N-NO3− values become important beginning in 1850. Data from this time period result in an average biomass/fossil fuel δ15N-NO3− end-member value of +9.5‰ and a δ15NOx value of +12.4‰. With increased fossil fuel burning, there is an overall decrease in δ15N-NOx of ~2‰ between 1850 and 1920. This decrease is corroborated by a recent study of δ15N-NOx values from power plants lacking selective catalytic reduction emission controls (+10‰) [Felix et al., 2012] and recent δ15N-NOx values reported for roadside vehicle emissions (+3.7 to +5.7) [Ammann et al., 1999; Pearson et al., 2000; Redling et al., in review]. Combined, contributions from these δ15N-NOx source values would lower the pre-1850 δ15N-NOx value of +14.4‰. It should be noted that the fossil fuel combustion δ15N-NOx values may have changed over time with varying combustion efficiency and varying emission reduction technology as reported in Felix et al. .
 Another possible explanation for the post-1850 decrease in δ15N-NO3− is contribution from soil NOx produced from cleared land from 1850 to 1900 when farmland increased from 293,561,000 to 841,202,000 acres in the U.S. However, from 1850 to 2002, the farmland acreage and ice core δ15N-NO3− are weakly but significantly correlated (R2 = 0.31, p < 0.01) (Figure S1b) [US Department of Commerce, 1975; US Department of Commerce Census 1970–2010]. This weak correlation suggests that the application of fertilizer, rather than the acreage of cleared land, is driving the increase in soil NOx emissions.
 After 1920, the main emission sources of NOx are biogenic soil NOx, biomass burning, and fossil fuel emissions. After 1920, widespread industrial fertilizer application began in the U.S. following the advent of the Haber-Bosch method, which reacts nitrogen and hydrogen gas in the presence of a catalyst to create ammonia (NH3) fertilizer. Soil nutrient enrichment via fertilizer application can contribute to large pulses of biogenic soil NOx [Veldkamp and Keller, 1997]. For example, Hudman et al.  report a 50% increase in soil NOx over the agricultural Great Plain in June 2006 due to rainwater induced pulsing. Jaeglé et al.  suggest that during the summer in the northern mid-latitudes, soil NOx emissions can reach half that of fossil fuel combustion sources.
 The strong temporal constraints on ice core history allow for exploration of temporal linkages between agricultural history and NOy deposition recorded in the ice core. For example, higher δ15N-NO3− values during the early 1930s can be linked to the Great Depression that was coincident with a period of decreasing fertilizer use in the U.S. (Figure S2) [US Department of Commerce, 1975]. As further example, the greatest rate of negative change in δ15N-NO3− values is from 1950 to 1980 (slope of −0.25‰/yr)—a period coincident with the Green Revolution when farmers worldwide nearly tripled grain production [Mann, 1997]. Lastly, higher δ15N-NO3− values in the 1980s (1983 to 1990) can be attributed to the U.S. recession and heavy farm debt leading to less fertilizer application [US Department of Commerce Census 1970–2010]. While an increase in biomass burning could have contributed to higher δ15N-NO3− values during this period, a weak correlation between δ15N-NO3− ice core values and biomass burning is observed between 1949 and 2005 (R2 = 0.13; p < 0.05) (Figure S1c) [U.S. Energy Information Administration, 2009]. The relatively minor influence of biomass burning is further supported by prior studies of Greenland ice cores that report weakening correlations between black carbon and vanilic acid (a biomass burning indicator) after 1951 [McConnell et al., 2007].
 Contemporary and historical NOx source inventories are difficult to constrain, particularly for estimating biogenic NOx source emissions. Large discrepancies exist between modeled estimates of soil biogenic NOx emissions and estimates derived from remote sensing observations of tropospheric NO2 column concentrations [Hudman et al., 2010; Jaeglé et al., 2005]. In the following, we use ice core δ15N-NO3− to constrain U.S. emission inventories of historic and contemporary soil NOx. In this analysis, we couple approximately yearly δ15N-NO3− ice core measurements with an isotope mixing model that incorporates contributions from biomass burning/fossil fuels and soil biogenic NOx (Figure 2).
 Model results illustrate the ability of the δ15N-NO3− values to aid in constraining contributions of biogenic NOx emissions to deposition (Table S1). For instance, in 1996 using the −27‰ δ15N-NOx value reported herein, biogenic NOx emissions are estimated to contribute approximately 25% to total NO3− deposited to the ice core from North America. This contribution is 4 times larger than the 6% of the NOx inventory reported by the U.S. EPA for U.S. biogenic NOx emissions during this time period [US EPA, 1998]. Further, a recent bottom-up NOx emission inventory for 2005–2006, produced using a global model of tropospheric chemistry (GEOS-Chem), predicts a biogenic soil NOx contribution of ~10% [Lamsal et al., 2010] to total U.S. NOx emissions. Our mixing model, on the other hand, estimates greater biogenic NOx contributions to Greenland during this period at 21% for 2005. Thus, these mixing model results complement recent remote sensing studies and indicate historic biogenic NOx emission contributions and associated NOy deposition are underestimated, with important implications for contemporary estimates of biogenic NOx emissions to global reactive nitrogen budgets. The offset between the modeled fraction of biogenic NOx peaks and actual U.S. fertilizer consumption may result from ice core data that represent several years of deposition, rather than strictly annual increments. This may be resolved by analyzing annual or seasonal sections of the ice core, although post-depositional process may mask seasonal trends in the ice core δ15N-NO3− values. For instance, post-depositional processes in the surface snow such as photolysis of NO3− or evaporative loss of HNO3 [Röthlisberger et al., 2002] would lead to the snow NO3− being enriched in 15N and the resulting NOx and HNO3 released to the atmosphere being depleted in 15N. The enrichment in the snow would be more prevalent during warmer temperatures and increased sunlight of spring and summer. This enrichment due to post-depositional processes may contribute to the seasonal trend of δ15N-NO3− [Hastings et al., 2004] found in Greenland surface snow that is opposite to that observed in precipitation and dry deposition across the U.S. [Elliott et al., 2007; Elliott et al., 2009; Kendall et al., 2007] and Julich, Germany [Freyer, 1978] with higher 15N in winter and lower in spring and summer. Diffusion of NO3− through the ice, especially during warmer temperatures [Thibert and Domin, 1998], may also confound the ability to investigate seasonal trends in δ15N-NO3− values. Thus, while the offsets we observe between peaks in modeled biogenic NOx and actual U.S. fertilizer consumption may be influenced by multiple years of deposition, the use of annual sub-annual or seasonal ice core increments warrants further investigation.
4 Conclusions and Implications
 The primary finding of this work, that increases in the twentieth century commercial fertilizer use have led to an increase in biogenic soil NOx emissions that drive the steep negative δ15N-NO3− trend recorded in the ice core, may also indicate that biogenic soil NOx emissions contribute to decreasing δ15N trends recently reported in sediment records from 25 remote Northern Hemisphere lakes [Holtgrieve et al., 2011]. Moreover, a similar decreasing 15N trend of N2O in firn air samples from Antarctica has been observed and is attributed to increased microbial activity due to increasing commercial fertilizer use [Park et al., 2012]. While it has been established that soil microbial activity is a major source of N2O, our work indicates that soil microbial activity is a historically underestimated source of NOx. Our results demonstrate that while biogenic sources may be relatively minor part of the overall NOx budget (relative to fossil fuels), their contribution to the isotope mass balance is substantial. Future work should aim to characterize isotopic end-members and may prove to influence some of these mass balance constraints. As global population growth and N fertilizer use continue to increase, it will become increasingly important to more accurately determine the amount of biogenic, as well as other NOx source emission contributions to the environment. Given that excess inputs of reactive nitrogen cause global air, water quality, and ecosystem impacts with important implications for human health, it is imperative that efforts to curb reactive nitrogen loading to the environment are based on accurate emission inventories. These results demonstrate that nitrate isotopes in ice cores, coupled with newly constrained δ15N-NOx values for NOx emission sources, provide a novel means for estimating contemporary and historic contributions from individual NOx emission sources to deposition. In regions with incomplete empirical records of biomass burning, fossil fuel combustion, and fertilizer consumption, δ15N-NO3− ice core data can provide essential records for understanding the role of these evolving human activities on air quality, water quality, and ecosystem health.
 JDF and EME acknowledge support from AFRI (Grant no. 05-13204-6800-00000-404178) and EPRI. This manuscript benefitted from significant contributions by Meredith G. Hastings and two anonymous reviewers.