Data and discussion

VI.2.5 Data and discussion 99

assumption that a 50:50 binary mixing is also valid for the ²³⁶U data suggests that pristine AABW does not contain any anthropogenic ²³⁶U. This result is consistent with measured and modeled radiocarbon ages of > 300 yr for AABW in the equatorial Atlantic Ocean (Broecker and Peng 1982; Campin et al. 1999). We conclude that the observed ²³⁶U/²³⁸U ratio at 4,250 m depth results from a simple mixing of pristine (no anthropogenic ²³⁶U) AABW with NADW. Therefore, the explanation of the local AABW signal is tightly connected with the explanation of the local NADW signal at 2,500 m depth (see Section 4.2).

Assuming that the ²³⁶U concentration decreases exponentially with depth (r² = 0.95, not shown) a ²³⁶U inventory of 1.1 × 10⁹ at cm^-2 is calculated. Extrapolating from these five data points to a global scale an oceanic input of about 1.5 t ²³⁶U would be necessary to explain the observed inventory. If no other source of ²³⁶U was involved, this number translates into an alternative assessment for the total amount of ²³⁶U from global fallout, resulting in a value of about 2,100 kg. This is more than a factor of two higher than previous estimates (Sakaguchi et al. 2009). However, the uncertainty of this estimation is large because the calculated ²³⁶U inventory strongly depends on the quality of one single surface data point as a representative for the global SML and because the calculated inventory also depends on the applied fit function. Additionally, the ²³⁶U inventory might be systematically biased by ²³⁶U transported by southward flowing NADW, which in turn might contain significant amounts of ²³⁶U from nuclear reprocessing (see Section 4.2).

The measured ²³⁸U concentrations at stations 39 and 40 (Figure VI.2-3, right axis) are in good agreement with the generally observed relation between U-concentration and salinity (Pates and Muir 2007; Owens et al. 2011) in the open ocean (black lines in Figure VI.2-3). The ²³⁶U-concentration also exhibits a significant (r² = 0.997) but much steeper correlation with salinity (Figure VI.2-3, left axis, red dashed line). The tight correlation, however, is dominated by one (high saline) surface data point. On the base of the current data set it is not clear if the observed correlation is accidental. An apparent ²³⁶U-salinity correlation could simply be caused by the fact that a two component mixing of a pre-anthropogenic and a single anthropogenic ²³⁶U signal is observed.

VI.2.5 Data and discussion 101

Figure VI.2-3. ²³⁶U and ²³⁸U concentration.

Measured ²³⁶U concentration (filled squares) and ²³⁸U concentration (open circles) vs. salinity at stations 39 (black symbols) and 40 (red symbols). Different U-salinity relations for the open ocean are shown: black straight line (Pates and Muir 2007), black dashed line (Owens et al. 2011). The red dashed line represents a linear fit to the ²³⁶U data (r² = 0.997). The measured surface ²³⁶U concentration at station 40 (indicated by the red dotted arrow) is regarded as an outlier and therefore neglected in the fit.

236U in the western equatorial Atlantic Ocean – modeling results. A main conclusion from the ²³⁶U data presented above is that the whole examined water column in the western equatorial Atlantic Ocean is influenced by anthropogenic ²³⁶U.

A major question therefore is: how has the anthropogenic ²³⁶U signal propagated into the deep ocean over the past 50 – 60 yr? There are two different transport processes that may be involved. First, sinking particles may transport U from the biologically active surface layer into the deep ocean where it is released due to remineralization.

Second, deep water production in the North Atlantic Ocean supplies fresh surface waters to the deep Atlantic Ocean via NADW formation. In the following, three simple conceptual models (Figure VI.2-4) are used to simulate particle flux and NADW production in order to identify the transport mechanism for ²³⁶U into the deep ocean.

Figure VI.2-4. Box models.

Box models used for the different simulations; (a) three box model to investigate the effect of vertical transport of U with sinking particles (SML: surface mixed layer). The deep ocean box is divided at 3,000 m depth to separate eqNADW from AABW. In the simulation 90% of the particulate U is remineralized in the eqNADW box, with the remainder remineralizing in the AABW box; (b) three box model for the simulation of NADW formation and export, nNADW: northern NADW, eqNADW:

equatorial NADW. The arrows indicate the flow direction, the labels give the volume rate; (c) the same model as in (b) with an additional and delayed input of 1% North Sea Water (NSW).

The input function: In all three box models ²³⁶U is introduced into the surface mixed layer (SML) box (Figure VI.2-4). To estimate the ²³⁶U/²³⁸U level in the SML box including its temporal evolution it is assumed that 70% of the 900 kg ²³⁶U, which were supposedly released during atmospheric bomb explosions, were at once deposited into the surface ocean in the year 1957 (roughly representing the median of the atmospheric bomb tests). In addition, the spatially heterogeneous deposition pattern of atmospheric fallout (southern vs. northern hemisphere = 1:3.8, (Hardy et al. 1973)) is considered. According to these assumptions more than 700 kg ²³⁶U were deposited on the northern hemisphere, with 432 kg thereof directly entering the SML.

Using this input value two different simulations are run assuming two different average SML depths (150 and 50 m; SML150 and SML50 in Figure VI.2-5). The SML

VI.2.5 Data and discussion 103 depths of 50 and 150 m represent typical minimal/maximal annual mean values for the subpolar and equatorial ocean (Monterey and Levitus 1997).

Figure VI.2-5. Simulated ²³⁶U/²³⁸U ratio.

The simulated ²³⁶U/²³⁸U ratio in the eqNADW given for different input scenarios and for different transport mechanisms on a logarithmic scale. Four different input scenarios are realized for the simulation of NADW production and export (upper dashed and dotted lines): (i) a transient ²³⁶U/²³⁸U scenario with an initial SML depth of 50 m (SML50, blue dashed line), (ii) a transient ²³⁶U/²³⁸U scenario with an initial SML depth of 150 m (SML150, red dashed line), (iii) and (iv) input scenarios (i) and (ii) plus an additional contribution of ²³⁶U from nuclear reprocessing. For better visibility only one input scenario (iii): SML50 (blue dotted line) is shown that includes ²³⁶U from reprocessing. The calculated

236U/²³⁸U ratios in the eqNADW-box for the four different input scenarios are indicated by the red and blue straight lines (without reprocessing) and by the red and blue short dotted lines (including reprocessing) in the middle of the plot. The color reflects the respective input scenario. The calculated

236U/²³⁸U ratio in the eqNADW box caused by the vertical transport of particulate U is indicated by the red and blue dashed-dotted lines in the lower part of the diagram. The filled circle and diamond mark the measured ²³⁶U/²³⁸U ratios (sampled in 2010) in the surface layer and in eqNADW, respectively.

The smaller the mixed layer in the model the higher is the calculated initial

236U/²³⁸U ratio. The SML50 scenario represents a maximum estimate for the modeled

236U/²³⁸U ratio in all simulations because in this scenario a higher ²³⁶U/²³⁸U ratio is always exported from the SML into the deep ocean than in the SML150 simulation.

The above assumptions result in initial ²³⁶U/²³⁸U ratios (in the model year 1957) of 5.6 × 10^-9 and 1.7 × 10^-8 for the SML150 and SML50 scenarios, respectively.

In these transient scenarios it is further assumed that from year to year the initial ²³⁶U signal in the SML penetrates deeper into the ocean where it then mixes

with water masses that do not contain anthropogenic ²³⁶U. The average yearly penetration rate of the signal is determined under the assumption that the simulated

236U/²³⁸U ratio in the surface box (dashed lines in Figure VI.2-5) equals the measured

236U/²³⁸U ratio at 25 m depth in the year 2010 (filled circle in Figure VI.2-5). For a SML depth of 150 m (50 m) an average signal penetration rate of 20.7 m yr^-1 (22.5 m yr^-1) is determined. This implies that in both input scenarios the calculated

236U/²³⁸U surface ratios decrease with time and per definition reach a value of 6.8 × 10^-10 in the model year 2010. An average penetration rate of 21 m yr^-1 further implies that in the model year 2010 the ²³⁶U signal has homogenized with the uppermost 1,200 m of the water column. Further, an annual penetration length of L = 21 m corresponds to a vertical eddy diffusivity of D = 0.14 cm² s^-1 (with L = (Dt)^½), which is well within the range of observational data (Gargett 1984).

Particle flux: In the surface waters of the open ocean dissolved U may be fixed to predominantly organic particles and exported to the bathypelagic layer (ca.

1,000 – 4,000 m) where it then remineralizes (Anderson 1982). In general, this process also transports ²³⁶U from the surface layer into the deep ocean. To simulate the vertical transport of particulate U a simple conceptual model is applied (Figure VI.2-4a). In this model U is exported (e.g., by sinking particles) from the surface box and completely remineralized in the two deep ocean boxes below. For the surface layer we apply the two (SML50 and SML150) input scenarios described above.

Measured export rates of particle bound bio-authigenic U range from 0.1 to 5.6 ng cm^-2 yr^-1 (Anderson 1982). A strict additional constraint for the maximum flux of particulate U is given by the U – salinity relation in the open ocean. Since no vertical gradient of the salinity normalized U-concentration is observed in the open ocean (Pates and Muir 2007) the flux of remineralized bio-authigenic U to the deep sea must be insufficient to measurably change the uranium concentration within the mixing time of the ocean (Anderson 1982).

To realize a maximum estimate of the vertical ²³⁶U flux in the model it is assumed that 90% of the particulate U remineralizes in the eqNADW box. Under the restriction that the U concentration in this box is allowed to change by not more than 2% per 1000 yr a maximum U export flux of 20 ng cm^-2 yr^-1 is calculated. If particulate U export from the surface is the sole mechanism transporting ²³⁶U into the deep ocean the above constraint implies that (assuming a constant ²³⁶U/²³⁸U ratio in the

VI.2.5 Data and discussion 105 SML) only 1.1‰ (i.e., 2% × 53 yr/1000 yr) of the ²³⁶U/²³⁸U signal at the surface reaches the deep ocean (the resulting ²³⁶U/²³⁸U ratio in eqNADW would not even be visible on the logarithmic scale of Figure VI.2-5). The non-steady state simulations use a transient ²³⁶U/²³⁸U signal as input function (SML50 and SML150). The results (blue and red dashed dotted lines in Figure VI.2-5) clearly indicate that the export and the subsequent remineralization of particulate U is not a significant source of

236U in the deep ocean, neither in eqNADW nor in AABW.

We note that the ²³⁶U/²³⁸U signal in the deep ocean reacts much more sensitively to vertical transport processes than, for example, ²³⁴U/²³⁸U since the ²³⁶U signal is not at steady state. It may therefore provide a powerful tool to quantify the particulate transport of U in oceanic regions where deep water formation does not play a significant role (e.g., the Pacific Ocean).

NADW formation: To simulate the temporal evolution of the ²³⁶U/²³⁸U ratio in the deep equatorial Atlantic Ocean a simple three box model was used (Figure VI.2-4b). In this model pristine (northern) nNADW is produced from the surface layer at a volume rate of 17 Sv, which is in accordance with estimates from hydrographic section data at 24° N (Roemmich and Wunsch 1985). Simulating the advective southward transport via the Deep Western Boundary Current (DWBC) the nNADW is transported into the equatorial NADW box (eqNADW) assuming the same volume rate (corresponding to a renewal time of 150 yr for each box). Since the modeled time period (53 yr) is much shorter than the oceanic mixing time (about 1000 yr) the box model does not conserve water (i.e., there is no return flux of water leaving the eqNADW box).

The results of the box model simulations (straight lines in Figure VI.2-5) show that only the SML50 simulation is able to explain the observed ²³⁶U/²³⁸U ratio in eqNADW quantitatively (filled diamond in Figure VI.2-5). The SML150 simulation produces ²³⁶U/²³⁸U ratios that are almost a factor of 2 lower than the measured data in eqNADW. Regarding all the assumptions made and the simplicity of the model the SML50 simulation is in reasonable agreement with the observations.

We are aware of the fact that the application of a box model to simulate the spatial propagation of a tracer signal a priori represents a maximum approach since no constraints for the maximal mixing lengths are given. To transport the

anthropogenic ²³⁶U signal within about 50 yr from the region of deep water formation (Labrador and Nordic Seas) to the sampling location (6,000 – 8,000 km further south) a minimum average current velocity of 0.4 – 0.5 cm s^-1 would be necessary. For the DWBC very variable velocities reaching values of up to 20 cm s^-1 have been reported (Joyce et al. 1986).

Further, a rapid spreading of ²³⁶U into the deep equatorial Atlantic Ocean is consistent with tritium and chlorofluorocarbon (CFC) data. For example, Jenkins and Rhines (1980) found a southward flowing tritium jet flow in the DWBC at 3,500 m depth (originating from nuclear weapons testing in the atmosphere). More recently, attempt has been made to access the transit time for the NADW pathway from the Labrador and Nordic Seas to the tropics using CFCs (Andrié et al. 2002). In this study a mean transit time of 25 yr (27 yr) was determined for upper (lower) NADW to reach the equatorial Atlantic Ocean.

All these results support the interpretation that, on a time scale of several decades, it is generally possible to transport significant amounts of anthropogenic

236U from the SML into the deep western equatorial Atlantic Ocean via NADW formation.

NADW formation and nuclear reprocessing: A contribution of anthropogenic ²³⁶U from the nuclear reprocessing plants in northwestern Europe might represent an additional source of ²³⁶U in the eqNADW. Although the releases started more than 50 yr ago the spatial distance from the North Sea to the western equatorial Atlantic Ocean is even larger. Therefore it seems unlikely that much of the

236U from nuclear reprocessing has reached the deep equatorial Atlantic Ocean. Yet on the other hand, even small contributions of North Sea Water (NSW) to the North Atlantic regions would significantly increase the ²³⁶U/²³⁸U ratio of the NADW forming surface waters. Furthermore, there is evidence for the presence of anthropogenic ¹²⁹I from nuclear reprocessing in the North East Atlantic Deep Water (NEADW) (Edmonds et al. 2001) and in the Deep Western Boundary Current (DWBC) (Santschi et al. 1996; Orre et al. 2010). Taking into account the different input functions (¹²⁹I releases peak around the year 2000 while the ²³⁶U release peak was about 20 yr earlier) it might become possible that ²³⁶U from nuclear reprocessing is already present in the eqNADW.

VI.2.6 Conclusion 107