May 09, 2022 Anthropogenic 236U and 233U in the Baltic Sea: distributions, source

The Baltic Sea receives substantial amounts of hazardous substances and nutrients, which accumulate for decades and persistently impair the Baltic ecosystems. With long half-lives and high solubility, anthropogenic uranium isotopes ( 236 U and 233 U) are ideal tracers to depict the ocean dynamics in the Baltic Sea and the associated impacts on the fates of contaminants. However, their applications in the Baltic Sea are hampered by the inad-equate source-term information. This study reports the first three-dimensional distributions of 236 U and 233 U in the Baltic Sea (2018 – 2019) and the first long-term hindcast simulation for reprocessing-derived 236 U dispersion in the North-Baltic Sea (1971 – 2018). Using 233 U/ 236 U fingerprints, we distinguish 236 U from the nuclear weapon testing and civil nuclear industries, which have comparable contributions (142 ± 13 and 174 ± 40 g) to the 236 U inventory in modern Baltic seawater. Budget calculations for 236 U inputs since the 1950s indicate that, the major 236 U sources in the Baltic Sea are the atmospheric fallouts (~1.35 kg) and discharges from nuclear reprocessing plants ( > 211 g), and there is a continuous sink of 236 U to the anoxic sediments (589 ± 43 g). Our findings also indicate that the limited water renewal endows the Baltic Sea a strong "memory effect" retaining aged 236 U signals, and the previously unknown 236 U in the Baltic Sea is likely attributed to the retention of the mid-1990s ’ discharges from the nuclear reprocessing plants. Our preliminary results demonstrate the power of 236 U- 129 I dual-tracer in investigating water-mass mixing and estimating water age in the Baltic Sea, and this work provides fundamental knowledge for future 236 U tracer studies in the Baltic Sea.


Introduction
The Baltic Sea is a semi-enclosed marginal sea in Northern Europe comprised of a series of shallow sub-basins divided by sills. The inputs of freshwater from the vast catchment and saline water from the North Sea result in a density-stratified water structure (Burchard et al., 2005;Meier, 2007;Mohrholz et al., 2015). Limited water exchange and ~30-year water residence time (Franck et al., 1987) make the Baltic Sea vulnerable to anthropogenic disturbances. Via river runoff and atmospheric deposition, substantial amounts of hazardous substances and nutrients enter the Baltic Sea and accumulate for decades (HEL-COM, 2018a) through the discharges from the highly populated (~85 million residents) catchment areas. Interactions between natural variability, anthropogenic pressures, and climate change lead to stronger stratification, less O 2 -rich saline inflows, and increasing eutrophication and deoxygenation of bottom waters since the 1980s (Carstensen et al., 2014;HELCOM, 2021).
Slow water circulation plays a significant role in determining the fates of nutrients/pollutants in the Baltic Sea (HELCOM, 2018b). Consequently, efforts have been dedicated to using the radionuclides ( 137 Cs, 99 Tc, 129 I, and 236 U), namely, those discharged from the two European nuclear reprocessing plants at Sellafield and La Hague, as oceanic tracers to track the formation, transport, and mixing of water masses in the Baltic Sea Qiao et al., 2017Qiao et al., , 2020aQiao et al., , 2021. Compared with the commonly used tracers in the Baltic Sea including salinity, temperature, nutrients, and dissolved organic matters Højerslev et al., 1996;Aure, 1998;Meier, 2005Meier, , 2007Stedmon et al., 2010;Kristiansen and Aas, 2015), the reprocessing-derived radionuclides are characterized by extremely low natural backgrounds, nonequilibrium dispersion states in the ocean, and accessible discharge histories. In particular, their well-constrained point-source releases are favorable for the hindcast simulations with reliable input functions to depict the large-scale and long-term hydrodynamics in the Baltic Sea. Among the reprocessing-derived radionuclides, 129 I (t ½ =15.7 Ma) and 236 U (t ½ =23.4 Ma) are promising candidates with long radioactive half-lives (i.e. no decay correction needed) and high solubility. In addition, the reverse temporal discharging trends from the reprocessing plants between 129 I and 236 U lead to a monotonic increase of reprocessing-derived 129 I/ 236 U atomic ratios in the North Sea since the 1970s (Christl et al., 2015c), making 129 I/ 236 U atomic ratio an ideal chronologic tracer. The combination of 129 I and 236 U has been used to quantify the circulation timescale of Atlantic waters in the Arctic Ocean (Casacuberta et al., 2016;Wefing et al., 2021) and can be potentially applied for estimating the ages of the saline inflowing waters from the North Sea to the Baltic Sea, which were only calculated by three-dimensional (3D) ocean models (Meier, 2007(Meier, , 2005 but not verified by any observation data. Up to now, there have been numerous works about the sources, distributions, speciation, transport, and tracer applications of 129 I in the Baltic Sea (Hou et al., 2002(Hou et al., , 2007Hansen et al., 2011;Yi et al., 2013aYi et al., , 2013bYi et al., , 2011, but little information is known about 236 U. During the Atomic Era, significant amounts of 236 U have been released into the environment by human nuclear activities. 236 U is mainly produced by thermal neutron capture reaction 235 U (n, γ) 236 U in nuclear reactors (NRs) and fast neutron reaction 238 U (n, 3n) 236 U in nuclear explosions (Steier et al., 2008;Sakaguchi et al., 2009). As a result, the two primary sources of anthropogenic 236 U in the environment are global fallout (GF) from atmospheric nuclear weapons testing (900 -2100 kg; Sakaguchi et al., 2009;Christl et al., 2012;Winkler et al., 2012;Casacuberta et al., 2014) and the authorized discharges from the reprocessing plants (> 260 kg; Castrillejo et al., 2020). However, the ubiquitous GF-derived 236 U interferes with the point-source-released reprocessing-derived 236 U in observations (Christl et al., 2015a). Until recently the accessible measurement of another anthropogenic uranium isotope, 233 U (t U-233 = 0.16 Ma), in environmental samples by accelerator mass spectrometry provided a new possibility to discriminate 236 U sources between GF and NR (Hain et al., 2020). 233 U is primarily produced by thermal neutron capture reaction 232 Th (n, γ) 233 U in some thorium-based prototype reactors and fast neutron reaction 235 U (n, 3n) 233 U in nuclear explosions (Hain et al., 2020). In contrast to 236 U, almost no 233 U is produced in commercial NRs and released by reprocessing plants, and the only substantial release of 233 U (7 -15 kg; Lin et al., 2021b) is from the tests of 233 U-fueled devices and thermonuclear weapons using enriched uranium as tamper material (Hain et al., 2020). This difference makes the 233 U/ 236 U atomic ratio a source-specific fingerprint allowing possible distinction between different emission sources for 236 U (Hain et al., 2020;Qiao et al., 2020bQiao et al., , 2021Lin et al., 2021bLin et al., , 2021c. The representative 233 U/ 236 U atomic ratio for the integrated GF signal was suggested to be (1.40 ± 0.15) × 10 − 2 by recent research on a peat core (Black Forest, Germany) and a coral core (Kume Island, Japan) (Hain et al., 2020). The reactor-modeling result and documented La Hague discharge data indicate that the 233 U/ 236 U atomic ratio in the NR-related releases from normal operations and accidents should be at the level of 10 − 8 -10 − 6 (Naegeli, 2004;Hain et al., 2020;HELCOM, 2020). This is supported by relatively low 233 U/ 236 U atomic ratios (avg. (0.12 ± 0.01) × 10 − 2 ) measured in the seawater and sediments close to Sellafield (Hain et al., 2020). Using a binary end-member mixing algorithm, the GF-derived and NR-derived 236 U signals can be quantified in a multiple-source environment (e.g. the Baltic Sea and the Greenland coast; Hain et al., 2020;Qiao et al., 2020bQiao et al., , 2021Lin et al., 2021b), and thus facilitating their tracer applications in the ocean.
Due to the limitations of measurement techniques in the past, observations of 236 U and 233 U in the Baltic Sea are scarce. Qiao et al. (2017) and Lin et al. (2021b) investigated the levels and distribution of 236 U in the Danish coastal waters and temporal evolution of 236 U and 233 U in the central Baltic Sea, respectively. They demonstrated that the Baltic Sea receives 236 U from both GF and reprocessing plants. A follow-up study by Qiao et al. (2021) reported the distributions of 236 U and 233 U in the surface waters of the Baltic Sea and suggested an "unknown" local NR-derived 236 U source, e.g. the undocumented discharges from local nuclear facilities or accidental leakages from nuclear waste dumping. However, there are still many knowledge gaps of the 236 U and 233 U in the Baltic Sea, e.g. their 3D distributions, major (input/output) fluxes, and budgets in seawater. In addition, pinpointing the source of the "unknown" NR-derived 236 U in the Baltic Sea requires further investigation to ease the public concern on radiological safety. Hence, systematic observations of 236 U and 233 U become the prerequisite of their tracer applications in the Baltic Sea.
In this work, we present the first dataset of 3D distribution of 236 U and 233 U in the modern Baltic Sea (2018-2019). Using the novel 233 U/ 236 U signature, we aim to locate the source terms of 236 U and quantify their contributions to the total 236 U budget in the Baltic seawater. By combining observation and numerical modeling, we also seek to uncover the transport and mixing processes related to the dispersion of NR-derived 236 U from the North Sea to the Baltic Sea, thus providing a better understanding of the dynamics and fates of contaminants in the Baltic Sea.

Seawater sampling
To investigate the spatial distributions of 236 U and 233 U in the Baltic Sea, 125 seawater samples were collected from 35 stations (depth files) in the Baltic Sea during 2018-2019 (Table 1). The scientific expeditions were carried out by the Swedish Meteorological and Hydrological Institute (SMHI), the National Institute of Aquatic Resources at Technical University of Denmark (DTU Aqua), the Finnish Environment Institute and Radiation and Nuclear Safety Authority (SYKE and STUK), and the Leibniz Institute for Baltic Sea Research Warnemünde (IOW), Germany. Based on the sampling periods and geographical distribution, these samples could be grouped into three transects: Transect #1 from the Skagerrak to Western Gotland Basin in the winter of 2018; Transect #2 from the Bothnian Bay to the Gulf of Finland in the summer of 2019; and Transect #3 from the Arkona Basin to the Western Gotland Basin in the winter of 2019 (Fig. 1). The sampling area covers all major regions of the Baltic Sea, including the transition zone between the North Sea and Baltic Sea (Skagerrak, Kattegat, and Danish Straits), the southern Baltic Sea region (Arkona Basin, Bornholm Basin, and South Baltic Proper), the middle Baltic Sea region (Northern Baltic Proper, Western Gotland Basin, and Eastern Gotland Basin), and the northern Baltic Sea region (Archipelago Sea, Åland Sea, Bothnian Sea, Bothnian Bay, and Gulf of Finland).
To facilitate the presentation of results and discussions, we identified six water masses in the Baltic Sea based on the salinity and geographical locations. In the transition zone, the Atlantic water (salinity > 34.5), Jutland coastal water (salinity = 32 -34.5), and Baltic inflowing/outflowing water (salinity = 15 -32) are three dominating water masses . The Atlantic water dominates the bottom of Skagerrak and Kattegat and originates from a southeastward branch of the Norwegian Atlantic Current expanding along the slope of the Norwegian Trench (Andersson, 2007). The Jutland coastal water is comprised of the central/southern North Sea waters and the runoff-influenced English Channel water, which transports waters from the southern North Sea to the transition zone by the Jutland Coastal Current (Aure, 1998). The main body of the Baltic Sea is separated by a permanent halocline into the Baltic bottom water (salinity = 9 -15) and Baltic surface water (salinity = 5 -9), which are influenced by the saline inflows from the transition zone, less saline brackish water from the northern Baltic region, and large freshwater supply from the rivers. The Baltic freshwater (salinity < 5) is mainly distributed in the Bothnian Bay and the estuary of Neva River in the Gulf of Finland.

238
U concentrations in seawater samples were directly measured by inductively coupled plasma mass spectrometry (Agilent 8800 ICP-QQQ) after 20 -50 times dilution. For the determination of ultra-trace levels of 236 U and 233 U, an optimized radiochemical separation procedure was utilized to extract uranium isotopes from 4.5 -9 L seawater samples (Qiao et al., 2015;Lin et al., 2021c), which is detailed in Supporting Information. Procedure blanks (in a total of nineteen) were prepared in every batch (typically seven samples) of analysis using ultrapure water following the same analytical procedure. Special measures were implemented to reduce the backgrounds of 236 U and 233 U, including purification of chemicals, utilization of laminar flow bench, and acid boiling for glassware (Lin et al., 2021c). Less than 1.95 ng of 238 U, 2.79 × 10 6 atoms of 236 U, and 8.94 × 10 3 atoms of 233 U were detected in the procedure blanks, which were negligible compared to the uranium isotopes measured in our seawater samples. The accelerator mass spectrometry measurements of the atomic ratios of 236 U/ 238 U and 233 U/ 238 U were carried out at the Vienna Environmental Research Accelerator (VERA) facility in the University of Vienna using the established setup and method for actinides (Steier et al., 2019;Hain et al., 2020;). 238 U was also used as an intrinsic tracer to obtain the chemical yield of uranium, which was on an average of ~70% in this work. The robustness of the analytical method was verified by repeatedly analyzing a quality control sample (2018184, Baltic seawater), provided by the Federal Maritime and Hydrographic Agency of Germany, together with the seawater samples in this work (Lin et al., 2021c). The background subtraction and uncertainty calculation for the actual uranium concentrations and isotope ratios are detailed in Supporting Information.

Hindcast simulation
In this work, a 3D ocean circulation model HIROMB-BOOS Model (HBM) was applied to simulate the dispersion of reprocessing-derived 236 U in the North-Baltic Sea. More details of the HBM model, setup, simulation, and validation are given in Supporting Information.
HBM is a well-calibrated operational model for storm surge forecast, oil spill forecast warning service in Denmark and Germany (Berg and Poulsen, 2012;She and Murawski, 2018;Murawski et al., 2021). The model has three nested layers: (1) a 2D barotropic model covering the North Atlantic Shelf Sea in 6-nautical-mile resolution to provide wind-induced sea level for the 3D regional model; (2) a 3D North-Baltic Sea model in 3-nautical-mile resolution, which is two-way nested with two high-resolution sub-domains; (3) two 0.5-nautical-mile model sub-domains in the Wadden Sea and the transition zone between the North Sea and the Baltic Sea. The dynamic two-way nesting with high resolution is essential for resolving the narrow Danish Straits and simulating the water exchange between the North Sea and the Baltic Sea (She et al., 2007). Two point sources were set for the two European reprocessing plants at La Hague and Sellafield. The annual discharges of 236 U from two point sources were based on the reconstructed data from shell records by Castrillejo et al. (2020), and a constant discharge rate was assumed for each point source within a specific year. As Sellafield is located out of the North-Baltic Sea domain, the corresponding point source was moved to the upper left corner of the domain. A loss rate of 50% and a lag time of 1 year were assumed for the transport of 236 U from Sellafield to its new location as suggested by Christl et al. (2015b). The model was simulated from 1971 to 2018 and validated with the observed NR-derived 236 U concentrations in Transect #1.

Levels and distributions of 238 U, 236 U, 233 U concentrations and 236 U/ 238 U, 233 U/ 238 U atomic ratios
An overview of the concentrations of 238 U, 236 U, and 233 U, as well as the atomic ratios of 236 U/ 238 U, 233 U/ 238 U, and 233 U/ 236 U in the Baltic Sea in 2018-2019 is given in Table S1. The 238 U concentrations ranged between 0.3 and 3.8 μg/kg and showed a strong correlation with salinity (R 2 =0.999, Fig. 2A) in the surface waters (depth < 10 m) of the Baltic Sea. The intercept of linear fitting suggested an average riverine 238 U concentration of 0.06 μg/kg. However, compared to the calculated 238 U concentrations from the surface 238 U-salinity correlation, negative offsets were apparent (avg. − 15%) for measured 238 U concentrations in the sulfidic deep waters of the central Baltic Sea, and the offsets showed a strong connection with sulfide concentrations ( Fig. 2B and C).
The measured 236 U ((1.99 -8.75) × 10 7 atoms/kg) and 233 U ((1.31 -3.73) × 10 5 atoms/kg) concentrations in this study were also similar to the previously determined levels in the surface waters of the Baltic Sea in 2011-2016 ((3.31 -12.5) × 10 7 atoms/kg and (0.46 -5.63) × 10 5 atoms/kg, respectively; Qiao et al., 2017Qiao et al., , 2021. A comparison of currently available 236 U datasets (Fig. 3) indicated that the North Sea and the Baltic Sea have higher 236 U concentrations than the inflowing eastern North Atlantic Central waters to the North Sea. As the eastern North Atlantic Central waters solely carry GF signal (Castrillejo et al., 2018), the elevated 236 U concentrations in the North Sea and the Baltic Sea should be attributed to the regional sources (i.e. reprocessing plants). The 236 U concentrations in the Baltic Sea were slightly lower than those in the North Sea, and relatively high 236 U concentrations (> 5 × 10 7 atoms/kg) were observed in the Jutland coastal water, Baltic inflowing/outflowing water, and part of Baltic bottom water along the passage from the Danish Straits to the Gotland Basin (Fig. 4), suggesting the transport of 236 U from the North Sea to the Baltic Sea. The 233 U concentrations were distributed within a narrow range ((2.15 -3.50) × 10 5 atoms/kg) throughout the Baltic Sea with slightly lower levels in the Atlantic water and Baltic freshwater.

Levels and distributions of 233 U/ 236 U atomic ratios, GF-derived and NR-derived 236 U concentrations
The atomic ratios of 233 U/ 236 U in the Baltic Sea were within (0.18 -0.97) × 10 − 2 in 2018-2019, reflecting contributions of 236 U from both GF and NR sources. Higher 233 U/ 236 U atomic ratios as observed in the Bothnian Bay indicated higher relative contributions of GF-derived 236 U, Fig. 2. The mixing diagram of 238 U concentration vs salinity in the Baltic Sea in 2018-2019 (A) and the distribution of uranium depletion (B) and sulfide concentrations (C) in Transect #3. In plot A, the mixing line function for the surface waters (depth < 10 m) is C 238 = 0.105 × S + 0.0587 (R 2 = 0.999), where C 238 and S represent the 238 U concentration (μg/kg) and salinity (PSU), respectively. In plot B, uranium depletion is defined as the difference between the salinity-based 238 U concentration calculated by the aforementioned function and the measured 238 U concentration in the seawater.
whereas lower 233 U/ 236 U atomic ratios in the Jutland coastal water, Baltic inflow water, and Baltic bottom water reflected the relatively strong influences of NR-derived 236 U in the transition zone and southern Baltic region (Fig. 5).

Inventories of 236 U and 233 U in the modern baltic sea
Based on the observation data in 2018-2019, we extrapolated the  spatial distributions of 236 U and 233 U concentrations, as well as GFderived and NR-derived 236 U concentrations, throughout the Baltic Sea using a 3D estimation tool in Ocean Data View (5.3.0). The 3D estimation was implemented as a fast weighted averaging procedure and used a specified high-resolution topography dataset of the Baltic Sea (latitude: 1 ′ ; longitude: 2 ′ ; depth: 5 m; Seifert et al., 2001). The total dissolved 236 U and 233 U inventories in the modern Baltic seawater were thereby estimated to be 316 ± 25 g and 1.96 ± 0.25 g, respectively. The GF-derived and NR-derived 236 U accounted for 44.9 ± 4.2% (142 ± 13 g) and 55.1 ± 13.0% (174 ± 40 g) of the total 236 U in the Baltic Sea, respectively.

Behaviors of U in the Baltic Sea
The mixing diagram of 238 U concentration vs salinity ( Fig. 2A) indicates that uranium has conservative behaviors in the oxic waters (e.g. surface waters) of the Baltic Sea, where uranium is present in +VI oxidation state as highly soluble uranyl carbonate complexes (Djogić et al., 1986). However, in the middle and northern Baltic Sea regions, 10 -20% of 238 U was scavenged from the sulfidic deep waters in 2018-2019, and this fraction could be up to 40% reported by Anderson et al. (1989) and Dellwig et al. (2021). The Baltic bottom water is permanently hypoxic and periodically anoxic, as its renewal only relies on the irregularly strong barotropic inflow events, or so-called Major Baltic Inflows (Carstensen et al., 2014;Meier et al., 2019). The recent investigations on the water column and sediments of the central Baltic Sea (Dellwig et al. 2021;Lin et al., 2021b) suggested that the uranium depletion in sulfidic waters arises from the redox-driven processes in the water column and diagenesis-driven removal in the sediment: (1) the soluble U(VI) in the water column can be reduced to insoluble U(IV) by sulfides and deposited on the seabed as particulate forms (Anderson et al., 1989); and (2) the dissolved uranium can be removed from the sulfidic porewaters into sediments, leading to a diffusion of dissolved uranium from bottom waters to the porewaters (Barnes and Cochran, 1990;Klinkhammer and Palmer, 1991;Lovley et al., 1991). The Major Baltic Inflows can temporally oxygenate the euxinic Baltic bottom water, suppress the removal of uranium from the water column, and remobilize a part of the scavenged uranium in the surface sediments (20 -50%), leading to a release of uranium from the uppermost porewaters to the water column (Dellwig et al. 2021). However, the remobilization of sedimentary uranium is quite limited and temporal, which slightly elevates the 238 U concentrations in the near-seabed waters (Dellwig et al. 2021). The uranium depletion recurs after the reestablishment of the bottom euxinia in the central Baltic Sea region (Dellwig et al. 2021).
The Baltic sediments in the anoxic sub-basins generally act as a net sink for natural and anthropogenic uranium isotopes. Our previous investigation on the Baltic sediments indicated that 589 ± 43 g of 236 U (GF-derived: 266 ± 5 g; NR-derived: 323 ± 43 g) were scavenged from the Baltic seawater to the sediments (Lin et al., 2021b). The comparable atomic ratios of 236 U/ 238 U, 233 U/ 238 U, and 233 U/ 236 U (Section 3.1) between the bottom waters and surface sediments in the central Baltic Sea revealed no significant isotopic fractionation for uranium during the scavenging processes. The 233 U/ 236 U atomic ratio is not affected by the scavenging processes and thus can still be used as a robust signature for tracing the water-mass movement and identifying emission sources of anthropogenic uranium (Hain et al., 2020;Qiao et al., 2020b).

Source terms of 236 U in the Baltic Sea
In addition to a negligible natural background, the majority of 236 U in the Baltic Sea originates from GF-related and NR-related sources and enters the Baltic Sea through various pathways. The GF-derived 236 U in the modern Baltic seawater is potentially from: (1) the dissolution of the directly deposited GF-derived 236 U on the Baltic Sea; (2) the riverine inputs of GF-derived 236 U deposited on the catchment; and (3) the oceanic transport of dissolved GF-derived 236 U in the North Sea via Baltic inflows. The NR-derived 236 U in the modern Baltic seawater may be attributed to: (1) the marine transport of reprocessing-derived 236 U (2) the dissolved 236 U fallouts from the Chernobyl accident, both in Baltic seawater and river waters from the catchment; and (3) unreported discharges from local nuclear facilities or accidental leakages from dumped nuclear wastes. In the following subsections, we will further discuss the specific sources/inputs and sinks/outputs of GFderived and NR-derived 236 U, investigate their related transport and mixing processes, and perform the budget calculations for them respectively.

Major inputs of NR-derived 236 U
Our recent study on the Baltic sediments indicated that the 236 U fallouts from the Chernobyl accident were very limited in the Baltic region, and little Chernobyl-derived 236 U was dissolved in the Baltic seawater (Lin et al., 2021b). Even though our earlier findings suggested a potential unaccounted NR-derived 236 U source in the Baltic Sea (Qiao et al., 2021), the observations in 2018-2019 did not demonstrate distinctly high NR-derived 236 U concentrations in the Baltic bottom water or freshwater, indicating no significant releases of NR-derived 236 U from the local nuclear facilities or dumped nuclear wastes to the modern Baltic Sea. Both the spatial distribution of NR-derived 236 U (Fig. 5) and the mixing diagram of NR-derived 236 U vs salinity (Fig. 6A) suggested that the NR-derived 236 U in the modern Baltic seawater should be mainly from the North Sea (more specifically, two reprocessing plants) rather than other riverine or benthonic sources.

Transport and mixing of NR-derived 236 U
The NR-derived 236 U vs salinity diagram revealed that there were three endmembers for NR-derived 236 U in the Baltic Sea (Fig. 6A). The first endmember was the saline Atlantic water characterized by low NRderived 236 U concentrations (<2 × 10 7 atoms/kg). The second endmember was the slightly less saline water (Jutland coastal water) with high and variable NR-derived 236 U concentrations ((3.5 -16) × 10 7 atoms/kg) constrained by the upper and lower boundaries for the potential mixing lines (red and blue dot lines in Fig. 6A). The third endmember was the brackish Baltic surface water with relatively constant NR-derived 236 U concentrations ((1.5 -3.5) × 10 7 atoms/kg), likely arising from the long residence time of saline water in the Baltic Sea.
Due to the limitations in sampling frequency and spatial resolution, the mixing diagram lost many details in the transition zone and southern Baltic Sea region. To elaborate further on the transport/mixing processes affecting NR-derived 236 U distribution, the 3D ocean model HBM was used to simulate the dispersion of 236 U discharges from Sellafield and La Hague. It should be noted that there are methodological difficulties in distinguishing NR-derived 236 U from the reprocessing plants and those from other civil nuclear facilities in the observations, but only the former ones were considered in our simulation. For a more accurate description, we prefer to use the term "reprocessing-derived 236 U" to represent "NR-derived 236 U from the reprocessing plants" in the simulation results.
The simulated water-column results for 128 stations in the North-Baltic Sea drew a full picture for the dispersion of reprocessingderived 236 U by oceanic circulation (Fig. S1). In the North Sea, higher 236 U concentrations were restricted in the near-shore areas rather than the central and northern regions by the coastal current transports. 236 U from Sellafield and La Hague was advected to the southern North Sea by the Scottish Coastal Current and English Channel Current, respectively, and further to the Skagerrak along the western European coastline by Jutland Coastal Current, which agreed with the previous observations (Christl et al., 2017(Christl et al., , 2015a. Upon arrival at the Skagerrak, the Jutland coastal water tagged by reprocessing-derived 236 U was sandwiched into the surface brackish outflowing water and the bottom saline Atlantic water due to the density stratification, which was also confirmed by the earlier findings Dahlgaard et al., 1995;Stedmon et al., 2010).
The numerical modeling resolved an open question related to the mixing of NR-derived 236 U: why the Jutland coastal water had a variable endmember of NR-derived 236 U concentration (Fig. 6A)? The Jutland coastal water is a combination of the central/southern North Sea water and the English Channel water (Aure, 1998), thereby its endmember is determined by: (1) the NR-derived 236 U concentrations in the central/southern North Sea water and English Channel water; and (2) the fractions of central/southern North Sea water and the English Channel water. The NR-derived 236 U concentrations in both central/southern North Sea water and English Channel water were kept relatively stable  Table S2, and the large variation in the second endmember is explained in Section 4.3.2). In plot B and plot C: the dark purple and blue rectangles represent the ranges of resolved and simulated GF-derived 236 U concentrations in the Baltic Sea and the North Sea in the 2010s, respectively. The resolved GF-derived 236 U concentrations are estimated from two sediment cores collected at the Gotland Basin and the Landsort Deep (Lin et al., 2021b). Here we assumed that the sedimentary GF-derived 236 U/ 238 U atomic ratios were equal to those in the bottom waters of two sub-basins and the 238 U during 2018-2019, as 236 U discharge rates from two reprocessing plants were generally constant since the 2010s. Our recent investigation on the water-mass compositions in the transition zone indicated that the fraction of the North Sea water has negligible inter-seasonal variability, whereas obvious seasonality is observed in the fraction of the English Channel water (Lin et al., 2021a). Driven by the predominant southwesterly winds in the North Sea, more English Channel water, along with La Hague-derived 236 U, flows into the Skagerrak in winter and vice versa in summer (Aure, 1998). The variations in the simulated reprocessing-derived 236 U concentrations ((6 -19) × 10 7 atoms/kg) of Jutland coastal water in 2018 were consistent with our observations ((3.5 -16) × 10 7 atoms/kg), and higher concentrations were observed in winter (Fig. 7A). In addition, both our observation and numerical modeling results revealed that the vertical mixings between the intermediate, less saline Jutland coastal water carrying relatively higher reprocessing signals and the bottom, more saline Atlantic water /surface, relatively fresh Baltic outflowing water in the transition zone caused the inverted v-shaped mixing lines.

Long-term dynamics of NR-derived 236 U
The long-term exchange of NR-derived 236 U between the North Sea and the Baltic was investigated by the temporal evolution of simulated inventories in both seas. As the 236 U discharges from the two reprocessing plants have decreased since the 1970s ( Fig. 7C; Castrillejo et al., 2020), our hindcast simulation revealed a significant decline of reprocessing-derived 236 U concentrations in the Jutland coastal water (Fig. 7B) and the decreasing inventories of NR-derived 236 U in the North Sea ( Fig. 7D; from the max. 42.7 kg to 1.52 kg) and Baltic Sea (Fig. 7E; from the max. 624 g to 211 g) in recent 40 years. The decrease of NR-derived 236 U was much slower in the Baltic Sea than the North Sea, owing to a much longer water residence time in the Baltic Sea (~30 years) than the North Sea (1 -2 years). The narrow and shallow channels of the Danish Straits restrict the bottom saline inflows driven by either the horizontal density gradients (observed in summer; Feistel et al., 2006) or storms (mainly in winter; Mohrholz, 2018). Limited water renewal endows the Baltic Sea with a strong "memory effect" and favors trapping the "aged" pollutions in the central Baltic Sea. The numerical modeling indicated that by the end of 2018, up to 57% of reprocessing-derived 236 U present in the Baltic Sea originated from the discharges before 1990. In contrast, almost 100% of reprocessing-derived 236 U in the North Sea was released within the recent 30 years. In the HBM model, even though the endmember for the Jutland coastal water varied largely for individual inflows, the monthly imported/exported reprocessing-derived 236 U was marginal (< 10%) compared to the NR-derived 236 U budget of the Baltic Sea since the 1990s, and could not significantly change the endmember for the Baltic surface water. This was also supported by our observations on the central Baltic Sea, where the distribution of NR-derived 236 U was relatively even (Figs. 5 and 6A). Qiao et al. (2021) estimated that the "unknown" local source contributed 200 ± 47 g of NR-derived 236 U to the Baltic Sea, which was at the same level as the total NR-derived 236 U inventory obtained in this work. Considering the notable differences in sampling scales (periods, depths, and locations), endmember settings, and modeling approaches between the two studies ( Fig. S2 and Table S2), the "unknown" source of NR-derived 236 U is likely the "aged" reprocessing-derived 236 U trapped in the central Baltic Sea. The direct evidence is the much lower 129 I/NR-derived 236 U atomic ratios in the surface waters of the middle Baltic Sea region (113 -370) than those of the transition zone (1202 -2608) in 2015 (Qiao et al., 2021). Our preliminary estimation suggested that the saline fraction of the surface waters of the central Baltic Sea carried at least 20 years older reprocessing signals than the saline waters of the transition zone (Fig. S3), which generally agreed with the simulated water-age results by Meier (2007). The endmember of Jutland coastal water for the "aged" reprocessing signal should be much higher than that for the "modern" signal, leading to the large deviation of 236 U/ 238 U atomic ratios in the surface of the central Baltic Sea between the observations and the calculated results by the endmember mixing model adopted in our previous work (Qiao et al., 2021).

Budget calculation for NR-derived 236 U
Our budget calculation indicates a net input of 211 g of reprocessingderived 236 U to the Baltic Sea, and 323 ± 43 g of NR-derived 236 U scavenged from the Baltic seawater to the sediments during 1971-2018 (Table 2). In general, there was a surplus of (174 -(211 -323) =) 286 g for the observed inventory of NR-derived 236 U (174 ± 40 g) in the modern Baltic seawater, which most likely originate from the underestimated inputs from Sellafield due to the lack of its discharge history before 1970 (Castrillejo et al., 2020) and the undocumented discharges from the local nuclear facilities.
The overall 236 U budget scheme in the modern Baltic Sea (2018-2019) is shown in Fig. 8. The annual net output of NR-derived 236 U to the North Sea was estimated as 11 g/year (average value during 1981-2018 in hindcast simulation). Based on the investigation on the Baltic sediments, the scavenged NR-derived 236 U was estimated to be 5.3 g/year, assuming that the scavenging rate of NR-derived 236 U by the anoxic sediments was 2.7 × 10 11 atom/(m 2 •year) in 2018 (Lin et al., 2021b) and an average hypoxic area of 49,000 km 2 in the Baltic Sea (Conley et al., 2009). Therefore, the Baltic Sea lost 5.3 + 11 = 16.3 g of NR-derived 236 U in 2018. Considering that the anoxia in the middle Baltic Sea region will likely persist due to the large legacy of nutrients (HELCOM, 2018a), the loss of NR-derived 236 U in the Baltic Sea water will continue at least in this decade.

Major inputs of GF-derived 236 U
The mixing diagram of GF-derived 236 U vs salinity (Fig. 6B) showed slightly higher GF-derived 236 U concentrations in the brackish waters of the Baltic Sea (Baltic surface water, Baltic bottom water, and Baltic outflow water) than the incoming saline and fresh waters (Atlantic water, Jutland coastal water, and Baltic freshwater). This phenomenon suggested that the present GF-derived 236 U in the main body of the Baltic Sea did not originate from the current inflowing waters but an old GFderived 236 U input. Unfortunately, the historical atmospheric deposition, riverine input, and oceanic transport of GF-derived 236 U to the Baltic Sea are not well constrained due to a lack of observation data, therefore we only made a rough estimation for different inputs of GFderived 236 U.
According to the areal inventory of GF-derived 236 U in the Black Forest peat core (9.0 × 10 12 atoms/m 2 ; Quinto et al., 2013), the total atmospheric depositions of GF-derived 236 U on the Baltic Sea and its catchment were estimated to be 1.4 kg (393,000 km 2 ) and 5.8 kg (1630, 000 km 2 ), respectively (Lin et al., 2021b), with the predominant deposition occurring in the 1950s-1960s (Quinto et al., 2013;Hain et al., 2020;Ohno et al., 2021). For the GF-derived 236 U deposited on the Baltic Sea, a maximum of 47 g of GF-derived 236 U directly reached the seabed (corresponding to the sedimentary inventory of GF-derived 236 U before 1970; Lin et al., 2021b), and the rest (1.35 kg) should be dissolved in the Baltic seawater. For the GF-derived 236 U deposited on the catchment, the mixing diagram suggested a low GF-derived 236 U concentration (~10 7 atoms/kg) in the river runoff of the Baltic Sea, which was close to the extrapolated level in the Elbe River water to the North Sea (< 10 7 atoms/kg; Christl et al., 2017). Assuming an average runoff of 454 km 3 /year (Johansson, 2016), the annual riverine input of GF-derived 236 U to the Baltic Sea was estimated to be 1.8 g/year in 2018, which was a minor contribution to the modern Baltic seawater inventory (266 ± 5 g). For oceanic transport, a net output of GF-derived 236 U from the Baltic Sea to the North Sea was estimated over the past five decades (see details in Section 4.4.2). Hence, we concluded that the dissolution of deposited GF-derived 236 U in the Baltic seawater should be the major input of GF-derived 236 U.

Long-term dynamics of GF-derived 236 U
Different from the reprocessing-derived 236 U, there are still some knowledge gaps in the historical inputs of GF-derived 236 U to the Baltic Sea, therefore it was not feasible to perform a long-term hindcast simulation for GF-derived 236 U without reliable input functions. Instead, we used sedimentary records to uncover the temporal evolution of GFderived 236 U in the Baltic Sea.
The reported GF-derived 236 U records resolved from coral and sediment cores suggested a general exponential decrease of GF-derived 236 U in global marine systems since the 1980s (Winkler et al., 2012;Sakaguchi et al., 2016;Nomura et al., 2017;Lin et al., 2021b). A comparison ( Fig. 6B and C) between the simulated GF-derived 236 U concentrations in the surface layer of the North Sea (55 ֯ N, 5 ֯ E; data from Christl et al., 2015b) and the reconstructed GF-derived 236 U concentrations in the bottom waters of the Gotland Basin and the Landsort Deep (Lin et al., 2021b) revealed a longer effective half-life of GF-derived 236 U in the Baltic Sea (21 -33 years) than the North Sea (11 years). This should also be related to the strong "memory effect" of the Baltic Sea. Similar to NR-derived 236 U, higher GF-derived 236 U concentrations in the Baltic brackish waters might represent the "aged" GF-derived 236 U signals trapped in the central Baltic Sea due to the limited water renewal.
Based on the differences between the simulated level in the North Sea and the resolved level in the Baltic Sea, we could roughly estimate the exchange of GF-derived 236 U between the two seas. The annual import of GF-derived 236 U to the Baltic Sea was estimated by multiplying the simulated GF-derived 236 U concentrations in the North Sea with the mean Baltic inflow (634 km 3 /year; Dahlgaard, 2002). Due to the Table 2 The budget calculation for NR-derived 236 U in the Baltic Sea.
homogenous distribution of GF-derived 236 U in the main body of the Baltic Sea ( Fig. 5 and 6B) arising from its slow water circulation, we assumed that the Baltic outflowing water had the same GF-derived 236 U concentrations as the Baltic bottom waters. Thereby the annual export of GF-derived 236 U from the Baltic Sea was calculated by multiplying the resolved concentrations in the central Baltic Sea with the mean Baltic outflow (1107 km 3 /year; Dahlgaard, 2002). The net decadal flux of GF-derived 236 U between the two seas was estimated to be 73 ± 13 g, -23 ± 15 g, -63 ± 18 g, -52 ± 2 g, and -44 ± 3 g for the past five decades, respectively (positive/negative values represent input to/output from the Baltic Sea). In total, a net output of GF-derived 236 U (108 ± 41 g) from the Baltic Sea was obtained during 1971-2018.

Budget calculation for GF-derived 236 U
Regarding the GF-derived 236 U budget of the Baltic Sea during 1971-2018, we found a deficit of ((1353 -108 -266) -142 =) 837 g for the observed inventory of the modern Baltic seawater (142 ± 13 g) when taking all available inputs (dissolution of the atmospheric deposition, 1353 g) and outputs (oceanic transport, 108 ± 41; scavenging process, 266 ± 5) into account. It should be noted that this deficit could be even larger if there were significant historical riverine inputs. An overestimated atmospheric deposition and/or an underestimated sedimentary inventory of GF-derived 236 U in the Baltic seawater might be the major contributors to this deficit.
In 2018-2019, the annual riverine input of GF-derived 236 U to the Baltic Sea was estimated to be 1.8 g (Fig. 8). The annual net output of GF-derived 236 U from the Baltic Sea was estimated as 5.5 ± 0.3 g based on the aforementioned calculations. The annual deposition rate of GFderived 236 U to the Baltic sediments was estimated to be 3.5 ± 1.5 g (Lin et al., 2021b). In summary, the Baltic Sea lost 5.5 + 3.5 -1.8 = 7.2 ± 1.5 g of GF-derived 236 U in the year of 2018. Similar to the case of NR-derived 236 U, the loss of GF-derived 236 U in the Baltic Sea water will likely continue in this decade.

Conclusions and perspectives
In this work, based on the large-scale observation and long-term 3D numerical simulation, we obtained in-depth knowledge on the source terms, transport, mixing, and fate of anthropogenic U in the Baltic Sea. The major findings of this work include: (1) the budget calculations for anthropogenic 236 U in the modern Baltic seawater indicate that the nuclear weapon testing and civil nuclear industries have comparable contributions (142 ± 13 g and 174 ± 40 g), primarily via the global atmospheric deposition in the 1950s-1960s and liquid discharges from the European reprocessing plants since the 1970s, respectively; (2) limited water renewal results in a strong "memory effect" of the Baltic Sea retaining the aged 236 U signals (as well as other pollutants/nutrients) for decades, and the previously unknown NR-derived 236 U in the Baltic Sea should be the reprocessing-derived 236 U discharged in the mid-1990s; (3) there is a notable sink of uranium from the water column to anoxic sediments in the Baltic Sea, which is the major contributor to the loss of 236 U in the Baltic seawater. Our preliminary results also demonstrate the potentials of reprocessing-derived 236 U and 129 I to trace the water-mass movement/mixing and estimate the water ages in the Baltic Sea. Combining field observations and 3D ocean modeling will enable us to further investigate the transit passages and circulation timescales of saline water in different sub-basins of the Baltic Sea, providing fundamental knowledge of the pollutant dynamics in the Baltic Sea to the decision-makers for the environmental management in the Baltic region.
Our observations reveal that uranium has unconservative behaviors in the sulfidic waters of the Baltic Sea. However, this shortcoming will not affect the robustness of using 233 U/ 236 U atomic ratios for tracing the water-mass movement and identifying emission sources of anthropogenic uranium in the Baltic Sea, as no significant isotopic fractionation occurs in the uranium scavenging processes. Besides, although the uranium depletion in the central Baltic Sea may increase the 129 I/NRderived 236 U atomic ratio by up to 66% (=100% / (100% -40%) -100%) resulting in underestimated water age, considering the rapid increase (two orders of magnitude) of the reprocessing-derived 129 I/ 236 U atomic ratios in the inflowing waters since the 1970s, the introduced uncertainties for the water-age estimation is negligible (1-3 years). Nevertheless, for a more accurate prediction of the spatiotemporal distribution of 236 U in the Baltic Sea, coupling sedimentation model with 3D ocean model should be considered in the future numerical simulations, where parameterization and verification for the uranium scavenging kinetics need to be investigated by laboratory experiments and field observations.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.