Continental-scale enrichment of atmospheric 14 CO 2 from the nuclear power industry : potential impact on the estimation of fossil fuel-derived CO 2

The 14C-free fossil carbon added to atmospheric CO2 by combustion dilutes the atmospheric 14C/C ratio (114C), potentially providing a means to verify fossil CO 2 emissions calculated using economic inventories. However, sources of14C from nuclear power generation and spent fuel reprocessing can counteract this dilution and may bias 14C/C-based estimates of fossil fuel-derived CO 2 if these nuclear influences are not correctly accounted for. Previous studies have examined nuclear influences on local scales, but the potential for continental-scale influences on 114C has not yet been explored. We estimate annual 14C emissions from each nuclear site in the world and conduct an Eulerian transport modeling study to investigate the continental-scale, steady-state gradients of 114C caused by nuclear activities and fossil fuel combustion. Over large regions of Europe, North America and East Asia, nuclear enrichment may offset at least 20 % of the fossil fuel dilution in 114C, corresponding to potential biases of more than −0.25 ppm in the CO2 attributed to fossil fuel emissions, larger than the bias from plant and soil respiration in some areas. Model grid cells including high14C-release reactors or fuel reprocessing sites showed much larger nuclear enrichment, despite the coarse model resolution of 1 .8×1.8. The recent growth of nuclear14C emissions increased the potential nuclear bias over 1985–2005, suggesting that changing nuclear activities may complicate the use of 114C observations to identify trends in fossil fuel emissions. The magnitude of the potential nuCorrespondence to: H. D. Graven (hgraven@ucsd.edu) clear bias is largely independent of the choice of reference station in the context of continental-scale Eulerian transport and inversion studies, but could potentially be reduced by an appropriate choice of reference station in the context of localscale assessments.


Introduction
Since radiocarbon ( 14 C) is absent in highly aged fossil fuels, fossil fuel combustion strongly dilutes the ratio of 14 C/C in atmospheric CO 2 , reported as 14 C including corrections for age and fractionation.Atmospheric observations can be used to quantify the dilution of 14 C and thereby provide an estimate of the amount of CO 2 added by fossil fuel combustion, relative to a clean air reference site (e.g., Levin et al., 2003).Thus, 14 C observations and atmospheric transport modeling may provide a means for independently validating CO 2 emissions calculated from economic data (Pacala et al., 2010).
One way that 14 C observations may be employed to estimate CO 2 emissions from fossil fuel combustion is through the joint inversion of atmospheric CO 2 and its 14 C/C ratio on continental scales (Peters et al., 2007;Pacala et al., 2010).Previous studies using observations and models have shown that fossil fuel emissions cause discernable continental-scale 14 C gradients (Randerson et al., 2002;Hsueh et al., 2007;Turnbull et al., 2009).Implementing observation sites along such 14 C gradients may allow continental-scale fossil fuel emissions to be estimated with inversion schemes such as CarbonTracker (Peters et al., 2007).However, in order to Published by Copernicus Publications on behalf of the European Geosciences Union.12340 H. D. Graven and N. Gruber: Atmospheric 14 CO 2 gradients from nuclear industry make such inversion-based estimates of fossil fuel fluxes on the basis of 14 C observations, all other influences on 14 C gradients must be known and corrected for.
One such influence is caused by activities of the nuclear power industry.Nuclear power and spent fuel reprocessing sites release 14 C in gaseous and liquid effluents, enriching 14 C of CO 2 in air and carbon in plant material and water surrounding nuclear sites by 4-20 000 ‰ (Levin et al., 1988(Levin et al., , 2003;;Dias et al., 2008).
Most prior studies of the nuclear influence on 14 C have focused on the impact of these emissions on CO 2 in the local areas surrounding nuclear sites, i.e., on scales of less than a hundred kilometers.For example, (Levin et al., 2003) calculated the influence of a nearby reactor on 14 C measured at the Heidelberg atmospheric sampling site in Germany using dispersion modeling of 14 C emissions observed at that reactor (Levin et al., 2003).Nuclear 14 C emissions may also contribute to 14 C gradients at larger, i.e. continental, scales extending to several hundred or thousand kilometers; however, the potential for nuclear 14 C emissions to influence continental-scale gradients of 14 C has not yet been explored.A previous modeling study found that the 14 C enrichment caused by the nuclear industry was negligible, but this study unrealistically applied 14 C emissions homogeneously across northern continental regions without considering the spatial distribution of individual nuclear sites (Turnbull et al., 2009).
In this study, we consider the influence of 14 C emissions from nuclear sites on continental scales.When nuclear 14 C emissions are mixed into the larger atmosphere, will a high density of nuclear sources or even one large nuclear source create large-scale regions of high 14 C, relative to areas on the same continent without nuclear sites or relative to the free troposphere?An analogy could be made to SO 2 emissions from coal-fired power plants causing acid rain deposition over a large-scale region that extends several hundred kilometers downwind of the power plants.
In order to investigate the potential for 14 C emissions from the nuclear energy industry to cause continental-scale gradients in 14 C, we estimate 14 C emissions from individual nuclear sites and conduct Eulerian atmospheric transport simulations of spatially-resolved nuclear 14 CO 2 and fossil fuel CO 2 sources.We assess the potential for 14 C gradients from nuclear 14 C emissions to cause biases in fossil fuel CO 2 at continental scales and compare the pattern and magnitude of the potential nuclear biases to those arising from 14 C exchange with the ocean and terrestrial biosphere (Turnbull et al., 2009).By compiling observed 14 C emission rates, we also consider variability and uncertainty in nuclear 14 C emissions.
Unlike previous work examining the dispersion of temporary, severe radioactive sources using Lagrangian approaches (e.g., Klug et al., 1992;Draxler and Hess, 1998), our study focuses on 14 C emissions from multiple nuclear sites that occur continually within continental regions of the North-ern Hemisphere.These 14 C emissions are part of the normal operating procedures of the nuclear sites and are within government-imposed limits.We use an Eulerian framework, rather than a Lagrangian framework, to estimate steady-state gradients over large scales.This Eulerian framework is similar to that used in global and regional inversions of CO 2 that exploit gradients between observation stations located 200-10 000 km from one another (e.g., Gurney et al., 2002;Peters et al., 2007), as well as in other studies of continental 14 C gradients (Hsueh et al., 2007;Turnbull et al., 2009).Our results therefore have specific relevance for applications utilizing steady-state, continental-scale 14 C gradients, while they do not address the small-scale gradients that exist in the local vicinity of individual nuclear sites and may also influence 14 C at some observation sites.

14 CO 2 emissions from individual nuclear power plant sites
Radiocarbon is produced mainly through reactions of nitrogen impurities and oxygen in uranium oxide fuel or coolant water of nuclear reactors, but also in structural material, in the graphite of graphite-moderated reactors and the cooling gas of gas-cooled reactors (Yim and Caron, 2006).Nearly all 14 C is released in the form of 14 CO 2 , except in Pressurized Water Reactors (PWRs) where 14 C is mainly released as 14 CH 4 (Kunz, 1985;Uchrin et al., 1998;Van der Stricht andJanssens, 2001, 2005).We assume the lifetime of 14 CH 4 (approx.10 yr; Prather, 1994) to be too long to contribute to continental-scale gradients in 14 C of CO 2 , permitting us to neglect 14 CH 4 emissions.
The 14 C emission factors are associated with substantial uncertainties as they vary, for example, due to episodic venting, replacement of resin columns and other maintenance (Kunz, 1985;Stenström et al., 1995;Sohn et al., 2004).To examine temporal and site-to-site variability, we compiled available observations of gaseous 14 C emissions and compared them to electrical energy output at several individual PWRs, BWRs, HWRs and GCRs (Fig. 1).Observations at LWGRs (Konstantinov et al., 1989) were consistent with UNSCEAR (2000).No observations from FBRs were found.
Substantial variability spanning 300-1000 % was found in the observations for different reactors and for individual reactors over several years, particularly in PWRs, HWRs and Magnox GCRs.No consistent differences between reactors in different countries were apparent.We calculated the 15 and 85 % limits of the lognormal cumulative distribution of the observations for each reactor type in Fig. 1 to define a 70 % confidence interval for the emission factors, similar to a 1-sigma uncertainty in a normal distribution.We apply the observed confidence intervals to estimate uncertainty in 14 C emissions and uncertainty in the resulting enrichment in atmospheric 14 CO 2 (Sects.3 and 4).
Theoretical estimates of 14 C emission factors (Fig. 1; Yim and Caron, 2006) were similar to observations for PWR and BWRs, but quite different for HWRs and GCRs.This is likely a result of theoretical estimates not accounting for 14 C capture at some HWRs and GCRs or the poorly-known release of 14 C produced in the moderators of GCRs.
Fossil fuel-derived CO 2 and Δ 14 C gradients Nuclear Δ 14 C gradients and potential bias in fossil fuel-derived CO 2 Absolute ratio of nuclear bias to fossil fuel-derived CO 2 Our estimates of total 14 C emissions do not include some additional anthropogenic 14 C sources, despite the fact that they could also contribute to 14 C enrichment at continental scales.These sources include emissions from experimental research reactors, reactors that were recently shutdown, radiochemical production facilities, military operations, and disposal or incineration sites for medical or research waste.We omitted these sources due to lack of data on emission rates and chemical forms of 14 C.However, observations from research reactors in Germany (BMU, 2002(BMU, -2008) ) and a radiochemical production facility in the UK (UKEA, 1996(UKEA, -2008) ) showed 14 C emissions that were similar to mediumto large-sized BWRs.Emissions from newly shutdown reactors can be as large as 300 % of the average release during active periods (BMU, 2002(BMU, -2008;;UKEA, 1996UKEA, -2008)), but are neglected here by our use of emission factors that are tied to electrical production.As a result, our estimated 14 C emission from the nuclear power industry does not comprise the total anthropogenic emission of 14 C.

Transport modeling
Surface fluxes of 14 C from nuclear sites and CO 2 from fossil fuel combustion were used as boundary conditions in simulations of the global TM3 atmospheric transport model with 1.8 • ×1.8 • resolution and 28 vertical levels (Heimann and Korner, 2003).Annual mean emissions of CO 2 from fossil fuel combustion were given by the Emissions Database for Global Atmospheric Research version 4.0 (EDGAR, available at http://edgar.jrc.ec.europa.eu/index.php)for individual years 1985-2005, aggregated from 0.1 • to 1.8 • resolution.
We computed 4-yr simulations with constant fluxes corresponding to each year 1985-2005, similar to the specifications of the Transcom 3 Experiment (Gurney et al., 2000), and averaged the simulated concentrations over the 4th year.Meteorological forcing was given by 6-h NCEP reanalysis fields specific to each year 1985-2005(Kalnay et al., 1996)).
We examine gradients in 14 C over three continental regions in the Northern Hemisphere, relative to a regional reference site: Niwot Ridge, USA (NWR, 3.75 km a.s.l.) for North America, Jungfraujoch, Switzerland (JFJ, 3.45 km a.s.l.) for Europe and Mt.Waliguan, China (WLG, 3.81 km a.s.l.) for Asia (Fig. 2a-c).Spatial maps of gradients in 14 C in the lowest model level are presented for 2005 in Sect.3, while temporal changes at selected sites are presented in Sect. 4.
Gradients were calculated using the simulated enhancement in CO 2 (δC ff ) or 14 CO 2 (δA nuc ) relative to the regional reference sites, i.e. δC ff = C ff −C R ff and δA nuc = A nuc −A R nuc , where R indicates the reference site.The dilution in 14 C caused by fossil fuel emissions, δ ff , and the enhancement in 14 C caused by nuclear emissions, δ nuc , were calculated by: These equations were derived by approximate mass balance of carbon and 14 C. R s is 1.176×10 −12 , the 14 C/C ratio in the Modern Standard.The change in 14 C also depends on the background air CO 2 mixing ratio and 14 C (C R and R ), which was assigned to be the global average for each year (Table S1).We use global average values at each regional reference site since observations are not available for all sites in all years.Though annual mean 14 C and CO 2 in Northern Hemisphere background air can vary by ±5 ‰ and ±1.6 ppm from the estimated global average (Levin et al., 2010;Graven et al., 2011;Keeling and Whorf, 2005), the potential error in δ ff caused by using global average values at the regional reference sites is less than 0.8 %.
Since the spatial gradients in fossil fuel CO 2 are small relative to the absolute concentration of CO 2 in the atmosphere, i.e., δC ff C R , the dilution of 14 C by fossil fuel emissions (δ ff ) relates to δC ff by a roughly constant factor of −2.8 ‰ : 1 ppm in 2005.The bias in δC ff that would occur if nuclear 14 C enrichment was not accounted for (β nuc ) similarly relates to δ nuc by approximately −2.8 ‰ : 1 ppm, since nuclear enrichment reduces apparent δ ff .
We performed sensitivity tests to evaluate the effect of uncertainty in 14 C emission factors and the choice of regional reference site.To test the effect of uncertainty in the emission factors, we performed additional simulations for emissions calculated with emission factors at the lower and upper limits of the 70 % confidence intervals shown in Fig. 1.To test the sensitivity to the choice of reference site, we additionally calculated 14 C gradients relative to free tropospheric air at 2.9 km a.s.l.(the 10th model level).
3 Regional gradients in 14 C of CO 2 The largest simulated δC ff of 11-18 ppm was associated with the most densely populated areas (Fig. 2a-c), while over large regions of North America, Europe, and Asia δC ff exceeded 0.5 ppm (δ ff < − 1.4 ‰).In contrast, nuclear 14 C emissions enhanced 14 C by more than 0.7 ‰ over large regions of North America, Europe and Asia in 2005 (Fig. 2df), offsetting the dilution of 14 C from fossil fuel emissions substantially.
The largest δ nuc and β nuc was simulated over northern France and the UK due to releases from La Hague and Sellafield reprocessing sites and several Gas-Cooled Reactors.Though enhancement of 14 C was largest in grid cells containing large nuclear sources, negative values of β nuc extend far into downwind regions without nuclear sources.Outflow from northern France and the UK contributed to high δ nuc and β nuc over much of Northern Europe (Fig. 2e).The Great Lakes region of North America, central Japan and South Korea also showed substantial δ nuc and β nuc extending >400 km away from nuclear sites.
The simulations clearly show a continental-scale influence of nuclear emissions: significant δ nuc gradients extended more than 700 km (3 grid cells) away from nuclear sites in northeastern North America, Europe and Asia.This spatial scale is sufficiently resolved by the model resolution of TM3, 100-200 km in mid-latitude regions.However, since Eulerian models like TM3 homogenize point sources over the local grid cell, simulated β nuc near nuclear point sources are sensitive to the model resolution and the location of the model grid.For example, the largest simulated β nuc is in the grid cell containing the spent fuel reprocessing site at La Hague, France.β nuc simulated for the area within this grid cell (−8 ppm) is likely to change if a different model resolution or model grid is used, particularly for areas within the grid cell that are particularly near to or distant from the La Hague site.
The relative magnitude of the potential biases in inferred fossil fuel-derived CO 2 , i.e. the absolute of the ratio of β nuc to δC ff , can amount to more than 100 % (Fig. 2g-i).Over the English Channel, β nuc was as large as 260 % of δC ff .In large regions, such as Eastern Canada, Northwestern France, the UK, Ireland, the Baltic Sea, Russia and Japan, the potential bias remained above 20 %.There were also areas with very little potential bias, owing to intense fossil fuel emissions but little to no nuclear activity, such as over the west coast of North America and most of China.
Simulated β nuc for 2005 using emission factors at the 15 % and 85 % limits of the cumulative distribution of observed emission factors are shown in Fig. 3. Increasing emission factors to the upper limit caused β nuc to be 300 % larger, on average.The area of β nuc < − 0.25 ppm spread over the Atlantic Ocean, Eastern Canada, Russia, Scandinavia, Southern Europe, China and Korea.In these areas, β nuc was generally larger than 20 % of δC ff .In simulations with emission factors Nuclear Δ 14 C gradients and potential bias in fossil fuel-derived CO 2 -Upper limit of 70% confidence Absolute ratio of nuclear bias to fossil fuel-derived CO 2 -Upper limit of 70% confidence Absolute ratio of nuclear bias to fossil fuel-derived CO 2 -Lower limit of 70% confidence Nuclear Δ 14 C gradients and potential bias in fossil fuel-derived CO 2 -Lower limit of 70% confidence   (d-f) and lower (j-l) limits of 70 % confidence.
at the lower limit, δ nuc and β nuc became 60 % smaller in North America and Asia and 40 % smaller in Europe, on average.Potential biases were much less important in North America and Asia, but in large regions of Northern Europe β nuc was still comparable in magnitude to δC ff (>20 %).
Patterns were largely the same when we used free tropospheric air as the background instead of the continental reference sites, and δC ff changed by less than ±0.1 ppm and β nuc changed by less than ±0.01 ppm in more than 85 % of grid cells shown in Fig. 2.

Temporal changes in δC ff and β nuc
Concurrent changes to the patterns and magnitudes of fossil fuel and nuclear emissions could cause spurious trends in δC ff inferred from 14 C observations.To estimate the potential for such an effect, we examine modeled annual mean δC ff and β nuc , relative to the continental reference sites, over 1985-2005 at 6 sites where 14 C in CO 2 is currently measured or may be initiated in the future: Cape May, USA (CMA) and Sable Island, Canada (SBL) in North America; Lutjewad, Netherlands (LUT) and Schauinsland, Germany (SCH) in Europe; and Gosan, South Korea (GSN) and Ryori, Japan (RYO) in Asia.
Modeled δC ff was between 1 and 7 ppm at the 6 sites over 1985-2005 (Fig. 4a-c).At each site, δC ff spanned ±0.2 to ±1.0 ppm from the mean value due to an overall trend and/or to variations in emission and atmospheric transport.β nuc was −0.1 to −0.8 ppm, with the largest negative potential biases at Cape May, Lutjewad and Ryori (Fig. 4d-f).At all sites, β nuc grew in proportion to δC ff (Fig. 4g-i) as the number and activity of nuclear reactors expanded between 1985-2005 and, at the European sites, as δC ff decreased.A strong increase in β nuc is apparent at Gosan, caused by the implementation of 3 Heavy Water Reactors at Wolsong, South Korea in the 1990s.To assess the impact of growth in β nuc on the apparent trend in δC ff , we compare 5-yr means of δC ff and δC ff + β nuc for 1985-1989 and 2001-2005 (Table 1).Simulated δC ff increased at the North American and Asian sites and decreased at the European sites between 1985-1989 and 2001-2005.Including the simultaneous change in β nuc caused δC ff to appear to have increased 6-7 % less at Cape May and Gosan, to have decreased 2-3 % more at Schauinsland and Lutjewad, and to have decreased by 4-5 % instead of increased by 1-2 % at Sable Island and Ryori.The largest effects were at Cape May and Ryori, significantly larger in magnitude than uncertainties in the fractional change in local δC ff or δC ff + β nuc due to variations in emission and atmospheric transport.Our results indicate that concurrent trends in β nuc can bias and change the sign of 14 C-based observations of δC ff trends.
δC ff calculated in comparison to free tropospheric air was 3-40 % smaller than δC ff calculated using the continental reference sites, except at Schauinsland where it was slightly larger.However, in comparison to free tropospheric air, β nuc was simultaneously reduced by a comparable amount (1-44 %) so that the ratio of β nuc to δC ff changed very little.
Simulations using emission factors at the limits of 70 % confidence demonstrate very large uncertainties that are skewed toward stronger β nuc (Fig. 4).At the upper limit, β nuc compensated 15-50 % of the dilution from δC ff at the sites.At the lower limit, β nuc compensated 5-10 % of δC ff .These uncertainties further complicate the identification of trends in δC ff using 14 C observations.While we have set emission factors to either the lower or upper limit at all sites, 12346 H. D. Graven and N. Gruber: Atmospheric 14 CO 2 gradients from nuclear industry Table 1.Change in simulated δC ff and δC ff + β nuc between 5-yr means for 1985-1989 and 2001-2005 at the sites shown in Fig. 4. Uncertainties were calculated using the standard error in simulated δC ff and δC ff + β nuc over the 5-yr periods, which comprise only variations in emissions and atmospheric transport over the 5-yr periods.Uncertainties in 14 C emission factors are not included.
the observations show that emission factors at each site vary from year to year (Fig. 1), which may cause different patterns and larger variability than our simulations.and 3).This is a consequence of us emitting the nuclear 14 C from point sources rather than spreading the emissions homogenously over the northern continents.Our result for Cape May could also be overestimated by the presence of two nuclear reactors in the local model grid cell, both located near the western edge of the grid cell while Cape May is located near the eastern edge.But the potential biases in the 5th model level (900 m) above Cape May and in the adjacent grid cell to the east are −0.4 ppm, also substantially larger than Turnbull et al. (2009).Similarly, there are two nuclear sites within the local grid cell of Orleans but the potential bias in the 5th model level above Orleans is also substantially larger than Turnbull et al. (2009), −0.5 ppm.No nuclear sites are present in the grid cells containing Sable Island, Lutjewad, Schauinsland, Ryori and Gosan.Simulated continental-scale effects can be compared with local-scale effects at Heidelberg, Germany (49.4 • N, • E), estimated by Levin et al. (2003).Our simulated continentalscale influence of nuclear emissions at Heidelberg is half as large as the estimated local-scale influence from the nearby Philippsburg nuclear site.Levin et al. (2003) used observed 14 C emissions at Philippsburg with a dispersion model to estimate local 14 C enrichment of 0.2-10 ‰ over 1986-2002, averaging 4.8 ±2.0 ‰.Our simulated nuclear enrichment in the grid cell containing Heidelberg is 2. 1 [1.1, 3.7] ‰ in 2005 (Figs. 2 and 3), which equates to a potential bias of −0.7 [−0.4,−1.3] ppm.Our estimate is lower than Levin et al. (2003) mainly because, as described above, the coarse resolution model underestimates 14 C near to nuclear point sources.However, another fundamental difference from Levin et al. (2003) is that we consider the influence from all nuclear sites, not only from Phillipsburg.Our results indicate that long-range transport from more distant nuclear sites is likely to be significant in Heidelberg, in addition to local transport from the Phillipsburg site.

Discussion and conclusions
Accounting for the spatial distribution of nuclear sites reveals several regions with a high density of 14 C sources that are important to consider in determining continental-scale influences on 14 C. Simulation of spatially-resolved 14 C emissions from individual nuclear sites in the Northern Hemisphere shows that these 14 C emissions contribute to a 14 C enrichment at continental scales that is substantial enough to partially counteract the fossil fuel dilution effect.Simulated potential nuclear biases of more than −0.25 ppm to δC ff extend over spatial scales on the order of 1000 km in populated regions of the Northern Hemisphere.This spatial scale is sufficiently resolved by the coarse Eulerian model we used, 100-200 km in mid-latitude regions, so this result is not limited by our model or model resolution.
Potential nuclear biases of −0.25 ppm or more make a substantial contribution to the total uncertainty in fossil fuelderived CO 2 determined by 14 C measurements, which is comprised of a component from measurement uncertainty and a component from uncertainty in non-fossil influences on 14 C.The measurement uncertainty is presently >0.5 ppm for an individual sample but can be as low as 0.3 ppm for an annual mean, calculated by averaging many samples (Levin and Rödenbeck, 2008).In continental studies, respiration of 14 C-enriched carbon from terrestrial ecosystems has been regarded to be the only substantial non-fossil influence on 14 C (Hsueh et al., 2007;Turnbull et al., 2009).However, by accounting for the location of nuclear point sources, rather than spreading the emissions homogenously over the northern continents as in Turnbull al. (2009), our results suggest that nuclear influences on 14 C are likely to be larger than those previously estimated (Sect.5; Turnbull et al., 2009).
Our work suggests that the potential bias in δC ff caused by nuclear 14 C releases may be as large or larger than the potential bias caused by exchanges with the terrestrial biosphere over some areas.Turnbull et al. (2009) simulated biases caused by respiration in recent years to be −0.2 ppm above northern continents, on average, and as large as −1 ppm, consistent with model results of 14 C enrichment of 0-2 ‰ above North America by Hsueh et al. (2007).Our simulated potential biases are more than −0.25 ppm over large regions, and up to several ppm near to nuclear sites.Additionally, β nuc tends to show stronger gradients than those resulting from relatively homogeneous biospheric sources (Turnbull et al., 2009).Together, nuclear and respiratory influences on 14 C likely cause potential negative biases in δC ff larger than 0.5 ppm over large regions of the Northern Hemisphere, similar to uncertainty contributed by measurement precision.
The broad, continental-scale patterns we simulated using an Eulerian transport modeling approach are caused by the aggregate influence on 14 C from all nuclear sites in the region, which cannot be accounted for by dispersion modeling of nearby reactors only.Our results show that the comparison of observed 14 C to a reference site >100-200 km away may therefore include a substantial continental-scale effect, in addition to any local-scale effects from nearby reactors.Observational studies at finer (urban) scales may be effective in reducing the continental-scale β nuc , however, by using local sites to define background air composition, particularly in areas that are far from nuclear sources.
While our objective was not to resolve local-scale dispersion and transport, simulated continental-scale β nuc is still highly dependent on model resolution such that stronger gradients exist within the 100-200 km grid used in the rather coarse TM3 simulations.Higher resolution regional models are likely to provide better estimates of continental-scale β nuc at particular sites.Higher resolution models may also represent transport to high altitude sites more accurately.Our results are also sensitive to errors in model transport, particularly in the vertical transport out of the boundary layer, though TM3 shows realistic vertical profiles of CO 2 on an annual mean basis (Stephens et al., 2007).
The simulated 14 C gradients include substantial uncertainties due to the large uncertainty associated with estimated 14 C emissions.The observed variability in emission factors (Fig. 1) suggests that β nuc could be much stronger (+300 %) or weaker (−60 %) in magnitude (Fig. 3).Moreover, 14 C emissions can vary strongly between different reactors or years (Fig. 1; Sect.2.1) and can occur in discrete periods when the reactor effluent is vented to the atmosphere.
In the coming decades, nuclear 14 C release is likely to grow in Asia and decline in Europe.58 nuclear power reactors are currently under construction in Asia, with the largest share (23) in China.Nearly all the reactors will be of the Pressurized Water Reactor type that has the lowest emission factor.However, a high density of low-14 C release reactors caused simulated biases of up to −1.5 ppm in Germany (Fig. 2e).At the same time, several reactors are being shut down due to age, including high-14 C release Magnox-type gas-cooled reactors in the UK, or due to the nuclear accident at Fukushima Daiichi Nuclear Power Plant in Japan in March 2011.More than half of Japan's nuclear reactors were immediately shut down for at least several months after the acci-dent, while Germany immediately shut down several older reactors and pledged to phase out all nuclear reactors within a decade.Several other countries delayed or canceled plans to build new reactors and the fuel reprocessing site at Sellafield, UK was shut down.
Whether 14 C releases grow or decline, trends in β nuc can bias the apparent change in δC ff over time and complicate the use of atmospheric 14 C to identify growth or reduction in CO 2 emissions.Trends in β nuc caused potential biases of 2-7 % in δC ff trends in our simulations, comparable to the emissions reductions agreed upon in the Kyoto Protocol.
Our results suggest that the influence of nuclear activities on atmospheric 14 C must be correctly accounted for in large regions of North America, Europe and Asia to estimate δC ff accurately using observations of 14 C in CO 2 .High resolution 14 C release data from each nuclear reactor site would improve estimates of 14 C enrichment by transport modeling.Alternatively, measures to reduce or eliminate 14 C release would improve accuracy in observation-based estimates of δC ff , though such measures would cause temporal changes to β nuc that would influence apparent trends in δC ff .

Fig. 2 .
Fig. 2. (a-c) Maps of fossil fuel-derived CO 2 (δC ff ) and 14 C dilution (δ ff ) in continental regions of the Northern Hemisphere.Regional reference sites are indicated by triangles and observation sites by squares.(d-f) Nuclear 14 C enhancement (δ nuc ) and potential nuclear bias to fossil fuel-derived CO 2 (β nuc ).Locations of low-14 C release reactors are indicated by crosses, high-14 C release reactors by triangles, and spent fuel reprocessing sites are labeled.(g-i) The ratio |β nuc : δC ff |, in percent, shown only in grid cells where β nuc was less than −0.25 ppm.

Fig. 3 .
Fig. 3. Results from transport model simulations of 14 C emissions for 2005 estimated using emission factors at the upper and lower limits of the 70 % confidence intervals as shown in Fig. 1.Nuclear 14 C enhancement (δ nuc ) and potential nuclear bias to fossil fuel-derived CO 2 (β nuc ) for emissions at the upper (a-c) and lower (g-i) limits of 70 % confidence.The ratio |β nuc : δC ff |, in percent, shown only in grid cells where β nuc was less than −0.25 ppm for emissions at the upper (d-f) and lower (j-l) limits of 70 % confidence.