Global and Regional Emissions of Radiocarbon from Nuclear Power Plants from 1972 to 2016

ABSTRACT CH4 and CO2 emissions from geologic sources, which are devoid of radiocarbon (14C), dilute the atmospheric 14C/C ratio. Observations of 14C/C can be used to estimate fossil fuel-derived CH4 and CO2. However, the atmospheric 14C/C ratio is perturbed by emissions of 14C from nuclear power plants (NPPs) and fuel reprocessing sites, which may affect such 14C/C-based estimation if they are not correctly quantified. We calculate NPP 14C emissions for CO2 and CH4 from 1972–2016 using standard emission factors (14C emitted per unit of power produced) and analyze trends in global and regional emissions. We use available observations of 14C emissions and power generation in Europe to assess emission factors for different reactor types, as well as potential differences related to the age or manufacturer of the NPPs. Globally, nuclear 14C emissions increase until 2005 and then decrease, mostly because of the closure of gas-cooled reactors in the United Kindom and the shutdown of light water reactors after the Fukushima nuclear accident in March 2011. Observed emission factors in Europe show strong variability, spanning values from 0.003 to 2.521 TBq/GWa for PWR and from 0.007 to 1.732 TBq/GWa for BWR reactors, suggesting more information and more sophisticated models are needed to improve estimates of 14C emissions.


INTRODUCTION
Carbon dioxide (CO 2 ) and methane (CH 4 ) are the two most important anthropogenic greenhouse gases contributing to climate change (IPCC 2013). In order to mitigate climate change, the 2015 Paris climate agreement stresses the need for reaching a "balance" between anthropogenic emissions and removal of greenhouse gases by the second half of this century (Rogelj et al. 2016), which will require strong reductions in emissions (Shindell et al. 2012;IPCC 2013). Anthropogenic emissions of CO 2 and CH 4 come from both biogenic and fossil sources, and understanding the relative contribution of sources to the global CO 2 and CH 4 emissions is fundamental to underpin mitigation policies. However, the source apportionment between biogenic and fossil sources of CO 2 and CH 4 , and between natural and anthropogenic sources and sinks, is still highly uncertain, particularly for CH 4 (Nisbet et al. 2016;Schwietzke et al. 2016). The short lifetime and large warming potential of CH 4 have led to a strong emphasis on CH 4 emissions mitigation in the Paris Agreement targets for the next 10-15 years (Shindell et al. 2012), amplifying the need for improved understanding of CH 4 emissions.
Atmospheric 14 C measurements have been widely used for estimation of the fossil-fuel fraction of CO 2 and CH 4 . When emitted to the atmosphere, carbon from fossil fuels, which is devoid of 14 C, causes a dilution of the atmospheric 14 C/C ratio (Δ 14 C, Stuiver and Polach 1977) that can be measured. Δ 14 CO 2 observations at "clean-air" and polluted atmospheric monitoring sites have been used to estimate fossil fuel-derived CO 2 over urban and continental regions (e.g. Levin et al. 2008;Turnbull et al. 2009;Miller et al. 2012), and long-term Δ 14 CH 4 observations currently provide the main constraint on the global fossil fraction of CH 4 emissions (Lassey et al. 2007a(Lassey et al. , 2007bKirschke et al. 2013). However, anthropogenic emissions of 14 C from nuclear power plants (NPPs) may affect 14 C/C-based estimation of fossil-derived CO 2 and CH 4 emissions if NPP emissions are not correctly quantified (Lassey et al. 2007a(Lassey et al. , 2007bGraven and Gruber 2011).
The production and release of 14 C from NPPs depends on the type of reactor used. The two main reactor designs are both "Light Water" reactors, which use water both as neutron moderator and coolant. The two designs are pressurized water reactors (PWRs) and boiling water reactors (BWRs). PWRs consist of two circuits, the primary cooling/heat transfer circuit with pressurized water and a secondary circuit where steam is generated, whereas BWRs have one circuit for both (EPRI 2010). Other reactor designs include British Magnox gas-cooled reactors (GCRs), using graphite as the neutron moderator and carbon dioxide as the primary coolant, advanced gas-cooled reactors (AGRs), the second generation of British gas-cooled reactors, pressurized heavy water reactors (HWRs), using heavy water as moderator, and light water graphite reactors (LWGRs), of Russian design.
Production of 14 C occurs through nuclear reactions involving the parent isotopes 14 N, 17 O, and 13 C in the nuclear fuel, coolant and structural material of the reactor (Yim and Caron 2006). Heavy water reactors (HWRs), predominantly in Canada, and gas-cooled reactors (Magnox GCRs and AGCRs) in the UK emit much more 14 C for the same amount of electrical power generation than other reactor types. Most of the 14 C produced in HWRs results from the large amount of 17 O in the heavy-water moderator, while 14 C emissions in the gas-cooled reactors come from the purification of the CO 2 circuits used to cool the reactor and from the isotopic exchange between the graphite and the CO 2 circuit (Dubourg 1998). In PWRs, the most common type of NPP in use today, gaseous 14 C effluents are mostly in the form 14 CH 4 (70-95%) (Kunz 1985), due to the reducing chemistry of the reactor coolant, whereas in all other reactor types the gaseous 14 C effluent is almost entirely 14 CO 2 .
Estimates of 14 C emissions are conducted by measurements at some facilities ( Van der Stricht andJanssens 2001, 2005) or estimated with information about the reactor (EPRI 2010). The European Commission reports yearly measurements of radioactive airborne and liquid discharges for each European NPP in the European Commission RAdioactive Discharges Database (RADD Database 2017). In the United States, reported 14 C emissions are typically not measured but only estimated according to recommendations by EPRI (EPRI 2010), which use a theoretical model for emissions of 14 C from light water reactors on the basis of the design, release pathways and unit-specific reactor core physics, including parameters such as the neutron flux profile, the mass of coolant in the active core and the concentration of nitrogen.
To estimate global 14 C emissions, emission factors based on electrical power production are typically used, due to the sparseness of 14 C atmospheric measurements from nuclear power facilities (Lassey et al. 2007a(Lassey et al. , 2007bGraven and Gruber 2011). Globally, 14 CO 2 production by NPPs is relatively small compared to natural production, about 11% (Turnbull et al. 2009, Graven et al. 2012. However, regions where nuclear facilities are concentrated may contribute to an atmospheric 14 C enrichment both at local (Levin et al. 2003) and continental scale (Graven and Gruber 2011) that may offset the 14 C dilution by fossil fuel emissions. Vogel et al. (2013) investigated local 14 CO 2 emissions from the nuclear industry in a hotspot region of Canada with multiple power plants, and highlighted how an underestimation of 14 CO 2 emissions from NPPs may cause errors in the calculated fossil fuel derived CO 2 . The effect can also extend several hundred kilometres from NPP emissions, counteracting the regional decreases in Δ 14 CO 2 caused by fossil fuel combustion (Graven and Gruber 2011).
In contrast to CO 2 , NPPs have a much stronger effect on the global inventory of 14 CH 4 and uncertainty in NPP emissions is a primary limitation on the use of Δ 14 CH 4 as a fossil fuel tracer (Lassey et al. 2007a(Lassey et al. , 2007b. 14 C releases from PWRs increased the 14 C/C ratio in CH 4 in the period between 1987 and 1995, representing 20-40% of the overall 14 CH 4 budget (Quay et al. 1999). By using a time series of atmospheric Δ 14 CH 4 from 1986 to 2000 guided by a mass balance approach, Lassey et al. (2007aLassey et al. ( , 2007b) determined a fraction of global CH 4 emissions of fossil origin of 30.0 ± 2.3% (1 σ) and an emission factor for 14 CH 4 produced during nuclear power generation by PWRs of 0.286 ± 0.026 TBq GWe -1 yr -1 . Eisma et al (1995) estimated an emission factor of 0.260 ± 0.050 TBq GWe -1 yr -1 for European PWRs using atmospheric Δ 14 CH 4 measurements from a sampling station in the Netherlands.
Even though emission factors are commonly used to estimate 14 C emissions from NPPs, they may differ greatly both among sites and over time within the same site. Graven and Gruber (2011) gathered observations available from 45 reactor sites and reported emission factors spanning about a factor of two for different reactors of the same type, and for the same reactor in different years. Vogel et al. (2013) similarly found that mean emission factors and some yearto-year variations for individual HWRs in Ontario, Canada, varied by a factor of two. Additional analyses of potential causes for the variation in 14 C emission factors may help to refine the emission factors used and resulting emission factor-based 14 C emissions estimates.
In this paper we update the global NPP 14 C emission database of Graven and Gruber (2011) to cover the period 1972-2016 and use the emission factor-based estimates to analyze trends in NPP 14 C emissions for CO 2 and CH 4 globally and by region. We also include data on 14 C emitted from spent nuclear fuel reprocessing sites. We find that global 14 C emissions from NPPs and fuel reprocessing sites decreased over 2006-2016 due to the decommissioning of nuclear reactors mainly in the UK, Germany and Japan. Then we use available observations of 14 C emissions and power generation for each European nuclear site to assess emission factors for PWRs and BWRs and for the fraction of 14 C emitted as CO 2 , as well as potential differences in emission factors related to the age or manufacturer for PWRs. The emission factors for European PWRs show large variability, which is not explained by the reactor age or manufacturer, demonstrating large uncertainties in the emission factor approach to estimating 14 C emissions.

METHODOLOGY
We update the nuclear power plant electricity production database and associated 14 C emissions from Graven and Gruber (2011) to cover the period 1972-2016. Annual energy output for each reactor was compiled from the International Atomic Energy Agency's Power Reactor Information System (IAEA PRIS 2017). The emission factors from Graven and Gruber (2011) ( Table 1), which are based on averages for 1990-1995 reported by UNSCEAR (2000), are used to calculate annual 14 C emissions from individual NPPs. Estimated annual 14 CO 2 and 14 CH 4 discharges are given for each nuclear site in TBq yr -1 in Table S1, where we assume PWRs emit 72% of the 14 C as CH 4 and 28% of the 14 C as CO 2 and all other reactor types emit all 14 C as CO 2 . Measured emissions from the Sellafield, La Hague (RADD Database 2017) and Tokai reprocessing Table 1 14 C emission factors for different reactor types with 70% confidence intervals from Graven and Gruber (2011). The confidence interval for the LWGR emission factor was not given by Graven and Gruber (2011) Nakada et al. 2008) are also included in Table S1, with 14 C emitted as CO 2 . Emissions from Tokai are available only until 2008 and we assumed no emissions after this. We do not include emissions from other fuel reprocessing sites, or 14 C produced by other activities such as medical applications for isotopes.
Emission factors were analyzed by comparing measured 14 C emissions (GBq yr -1 ) for 71 European nuclear plants from the RADD database (RADD 2017), to the IAEA PRIS energy output data (IAEA PRIS 2017), on an annual basis. We bin data by country for 1995-2005 and 2006-2015 to examine differences in emission factors in different countries over time. Then, focusing on the 48 European PWRs with the most data available, we examine difference in emission factors for reactors of different ages. Three age intervals were chosen: 0-25, 25-35, and more than 35 years. Finally, we assess differences in emission factors for the 48 European PWRs from the four main manufacturers: VVER, Siemens, Areva, and Westinghouse.
For most NPPs, only total 14 C emissions are reported in RADD. However, emissions of 14 CO 2 and 14 CH 4 are reported separately for Spanish, Hungarian, and German reactors, and these data were used to calculate the 14 CH 4 fraction for BWR and PWR reactors. The percentage of 14 C from non-methane hydrocarbons is not available, and therefore excluded from the analysis.

Global and Regional 14 C Emission Trends
By estimating 14 C emissions using the emission factors in Table 1 with electricity production data, and including observed 14 C emissions from reprocessing plants, we find global total nuclear power plant 14 C emissions increased from 1972 to 2005 as the number of nuclear facilities expanded, but this trend recently reversed as a result of changes in nuclear energy production ( Figure 1). The apparent emissions reduction after 2005 is mostly due to the closure of GCR-type reactors sited in the UK and the closure or temporary shutdown of PWR and BWR reactors in Germany and Japan after the Fukushima nuclear accident in March 2011 ( Figure 1a). However, an uncertainty of 14-35% on the annual global 14 C estimate, based on the emission factors uncertainties in Table 1 and a Monte Carlo analysis, must be taken into account (see below).
Most 14 C emissions come from PWRs, even though the PWR emission factor is the lowest (Table 1), as PWRs are by far the most common type of reactor. On the other hand, because of the high emission factor, HWR type reactors produce 28% of total 14 C emissions, even though they represent only about 5% of the generating capacity of all current operating reactors today ( Figure 1a). Emissions from HWRs showed the strongest increase over 2005-2015. Spent fuel reprocessing (SFR) contributes 0-32 TBq.
There are clear differences in 14 C emissions trends based on energy production by region in recent years (Figure 1b). We find emissions decreased in Europe, increased in Asia (except Japan) and remained approximately steady in the US and Canada. Emission factors for European PWRs and BWRs calculated using observed 14 C emissions and power produced overlap the emission factors from Graven and Gruber (2011), but exhibit large variability (Tables 2 and 3; Figure 2). Emission factors are generally higher for Other European reactors than for German, Spanish, and British reactors. In comparing reported 14 C emissions and electricity production in France, we found that the resulting emission factors were very consistent, suggesting the data reported for French reactors appear to be estimated on the basis of the power production rather than on measurements. The apparent emission factor in the French PWRs is 0.209 [0.208-0.210] and 0.209 [0.208-0.210] TBq/GWa before and after 2005 respectively. We therefore omit the French data from our analysis of observed emission factors.
The median value of emission factors based on measurements taken after 2005 is slightly higher than the previous period for Other European PWR and BWRs and for German PWRs, whereas we might expect a decrease in the emission factors due to increasing efficiency of power plants. For Other European PWRs, this might be explained by the availability of more 14 C measurements from nuclear facilities with higher emission factors after 2005 than before 2005 (e.g. in Czech Republic, Sweden; Figure 3). Measurements from only one Spanish PWR were available before 2005.
From a closer look into the outliers in Figure 2, we found that the highest values, greater than 0.75 TBq/GWa, are associated with very low energy production-i.e. maintenance periods or shutdown of the reactor-for German PWR reactors Biblis A and B (Figures 3 and S2). High emissions were found for some PWRs in certain years: Bohunice in Slovakia, Temelin in Czech Republic and Asco 1-2 and Vandellos in Spain. The emission factors for the Swedish reactor Ringhals 2, the oldest of the four reactors within the Ringhals nuclear plant, are consistently higher throughout the whole period of study   (Figures 3 and S2). 14 C releases from the Swedish NPP have been investigated in the study of Stenström et al. (1995), but measurements were taken only from reactors 1 and 4, which are characterized by a lower emission factor than Ringhals 2 in our analysis. For the PWR Ringhals 4 they observed a substantial increase of emissions during the venting of the reactor containment and gas decay tanks, where the cover gas from the primary system is compressed and stored before release, to allow for the decay of short-lived radionuclides. Venting operations were concurrent with provisional reactor outages, together with the total replacement of the cover gas in the primary coolant, which explains the persistence of 14 C releases in PWR reactors during temporary shutdown periods. A strong 14 C release during NPP shut-down events has also been found in other  Graven and Gruber (2011). Outliers shown with circles are calculated as less than Q1 -1.5*IQR and greater than Q3 + 1.5*IQR.
The occurrence of 14 C emissions in some NPPs during temporary shutdown periods (e.g. Slovakia), is highlighted in Figure 3. The strong variability in emission factors (R 2 of regression line is 0.09) for the European PWR reactors might be explained by the variety of operating procedures (e.g. power production, shutting down, maintenance, testing, refueling) adopted by each nuclear plant. Different techniques employed for the emission measurements and related uncertainties may affect the emission factor variance as well. However, neither the measurement techniques used nor the data precision were reported in the RADD database and therefore could not be assessed. Standardized information on the operating procedures adopted by each plant or on specific shutdown periods were also not provided. Nevertheless, the variation in annual 14 C emission factors is narrowed down when the mean of the total 14 C emission values over the whole time period for each reactor is plotted against the averaged power production (Figure 4; R 2 of 0.59). This suggests that higher capacity reactors do produce more 14 C emissions, consistent with the emission factor model, even though year-to-year variations may not be well-explained by power production (Figure 3). An alternate model of 14 C emissions using the emission factors in Table 1 with the reactor capacity, neglecting year-to-year variations in power production, may provide (time-invariant) estimates of 14 C emissions with similar skill as the emission factor model applied to annual power production data.  (Table 1). Figure 5 shows the emission factors for three PWR manufacturers: Siemens, VVER, and Westinghouse. A fourth manufacturer, Areva, produced all of the reactors in France. Since we found that the 14 C emissions from France were likely to be based on a standard emission factor of 0.209 TBq/GWa, we exclude the French Areva reactors from the analysis here.
Most of the reactors within the same nuclear facility are built by the same manufacturer, which allowed allocating each nuclear site to one manufacturer. Areva and Siemens are based on a licensed Westinghouse design, so these three manufacturers use similar technical specifications, whereas the Soviet-design VVER reactors are substantially different in the components of the primary system and in the safety measures implemented (Cacuci 2010). The use of a larger volume of coolant and nitrogen solutes as chemical regulators in the primary system of VVERs may result in a larger production of 14 C via the 14 N(n,p) 14 C reaction. Newer VVER models have incorporated more features from the western-type reactors; however, more advanced VVER designs are operating mostly in Russia, and are not reported in the RADD database. Only up to generation-2 VVER reactors are operating in Europe and included in the analysis. Siemens-manufactured PWRs were built mostly in Germany, whereas Westinghouse-manufactured reactors were built in Slovenia, Spain, Sweden and the UK. While the mean emission factor for Westinghouse is higher than for Siemens, the median value of 0.193 TBq/GWa is similar. In particular, emission factors from the Swedish Westinghouse reactor Ringhals 2 are notably higher (>1 TBq/GWa, Figure S2), and are apparent in Figure 5 as outliers.
The analysis of 14 C emissions from European PWR reactors of different ages is shown in Figure 6. Younger reactors seem to produce less 14 C per GWa of electricity produced, as would be expected since they use newer technologies and adhere to environmental standards put in place more recently (IAEA PRIS 2017). However, older reactors might have been updated throughout the period of study and their emission factor reduced. Again, median values are more similar than mean values across the three types, therefore emissions factors distributions cannot be consider statistically different.
Emissions of 14 CO 2 and 14 CH 4 Only Germany, Spain, Slovakia, and Hungary differentiate total gaseous 14 C effluents in 14 CO 2 and 14 CH 4 emissions. 14 CH 4 and 14 CO 2 emissions measurements from the PWR reactors within these countries have been compared to the emission factor based estimates, and to the mean 14 CH 4 and 14 CO 2 emissions over all reactors (Figure 7).
Based on the measurements for 1995-2015 from reactors in the aforementioned countries, we calculated an average fraction of 14 CH 4 of 72% for PWR and 0.5% for BWRs. These are similar to other studies that observed a 14 CH 4 fraction in a range within 57 and 93% for PWRs and a fraction of 0.5% for BWRs (Kunz 1985;Uchrin et al. 1997).
Similar to the total 14 C emissions (Figure 3), we find a large range of 14 CH 4 and 14 CO 2 emissions in the measurements and the emission factor-based estimates. However, we do find that the emission factor-based estimates represent the measured emissions better than the simple average of 14 CH 4 and 14 CO 2 emissions over all reactors does (Figure 7). The root mean square error between the measured 14 CO 2 emissions and the emission factor based estimates is 0.149 TBq, compared to a root mean square error of 0.161 TBq for the average emissions of the 14 C measurements from the PWR reactors (0.071 TBq). For 14 CH 4 emissions, the root mean square error is also lower for the emission factor based 14 CH 4 estimates than for the average emissions (0.192 TBq), 0.076 and 0.081 TBq respectively. This suggests that the use of the emission factor approach is more suitable than using a constant 14 C emission per reactor type.
Based on this analysis, we assume that 72% of 14 C released from PWRs is 14 CH 4 , with the rest of the emissions from all reactor types as 14 CO 2 , and we calculate global emissions of 14 CH 4 and 14 CO 2 (Figure 8). To calculate uncertainty we conducted 600 Monte Carlo simulations to estimate emissions from each reactor using the estimated log-normal distributions of emission factors (Figure 1, Graven and Gruber 2011). The number of simulations has been chosen in order to obtain a converged standard deviation (Rochman et al. 2014). The filled area in Figure 8 shows the interquartile range of the estimates for each year.
Global estimates of 14 CO 2 show a reduction after 2005 concomitant with the closure of the UK GCR-type reactors ( Figure 1)  1078 G Zazzeri et al.

CONCLUSIONS
Global 14 C emissions appear to have peaked in the mid-2000s and then decreased slightly as a result of some European and Japanese NPPs shutting down. The shutdown of UK Magnox GCR-type reactors after 2005 played a key role in the 14 CO 2 emission reduction, as they have the highest emission factor. As global PWR energy output decreased after 2010, 14 CH 4 emissions are also expected to have decreased in the past 7 years. While our estimates have large uncertainties, it is unlikely that global 14 CH 4 or 14 CO 2 emissions increased substantially over the past 10 years. Therefore, atmospheric studies of 14 CH 4 or 14 CO 2 should account for this inflection point in the growth of 14 C emissions over previous decades. At the same time, regional shifts in emissions have occurred, with emissions increasing in China and India but remaining steady or decreasing elsewhere. It is not clear how 14 CH 4 or 14 CO 2 emissions will change in the future. There are 57 NPPs currently in construction, primarily in Asia (IAEA PRIS 2017), which are nearly all PWR types. Future projections in nuclear energy production in the shared socioeconomic pathways (Riahi et al. 2017) show both increases and decreases in different scenarios. Some strong greenhouse gas mitigation scenarios show nuclear energy production increasing more than 10-fold by the end of the century (SSP Public Database 2017).
Our analysis shows strong variability in observed 14 C emission factors for European NPPs, spanning values from 0.003 to 2.521 TBq/GWa for PWR and from 0.007 to 1.732 TBq/GWa for BWR reactors for the period 1995-2015. The values used in Graven and Gruber (2011) of 0.24 and 0.51 TBq/GWa for PWR and BWR reactors, respectively, based on observations from 1990-1995, and theoretical 14 C production rates of 0.3 for PWRs and 0.6 for BWRs TBq/GWa (Yim and Caron 2006) are within these ranges. The value of 0.286 ± 0.026 TBq GWe -1 yr -1 for PWR reactors calculated by Lassey et al. (2007b) is also consistent with our range. Emission factors for American PWR reactors (US NRC 2016) calculated based on the EPRI recommendations (EPRI 2010) are also within these ranges, but have a somewhat higher median value of 0.4 TBq/GWa.
Average 14 C emissions observed at the European sites show a correlation with their average power production over 1995-2015, but year-to-year variations in observed 14 C emissions do not show a strong correlation with year-to-year variations in energy production. Previous studies have shown that emissions can be elevated during outage periods, effectively showing the opposite relationship as assumed in the emission factor approach. We found that intrinsic characteristics of nuclear reactors such as age and manufacturer did not explain the observed variability, except for VVER-type PWRs that have higher emissions. Mean emission factors from older reactors (35-50 years old) are slightly higher (0.36 TBq/GWa) than younger reactors, but median values are nearly the same (0.21 TBq/GWa).
We have applied the power-based emission factor approach to estimate 14 C emissions from individual power plants because observations of 14 C emissions are only available for a small number of sites and the emission factor approach is the only model available. While the use of power-based emission factors is likely to provide a reasonable estimate of global 14 C emissions on interannual timescales, power-based emission factors might lead to spurious estimates in some regions over shorter timescales. A better understanding of variations in 14 C emissions and a model that can predict such variability are needed. The model from EPRI (2010), which is based on theoretical calculations and limited data not including the European data we use here, could be implemented for the European reactors and refined by comparing estimated 14 C emissions with recent measurements reported in RADD Database (2017). Information on temporary shutdown and venting periods and more finely resolved electricity production and 14 C emissions data would help to identify patterns of 14 C emissions.
Radiocarbon Emissions from Nuclear Power Plants 1079