Spatial Variability and the Fate of Cesium in Coastal Sediments near Fukushima, Japan

Quantifying the amount of cesium incorporated into marine sediments as a result of the Fukushima Dai-ichi Nuclear Power Plant (FDNPP) accident has proven challenging due to the limited multi-core sampling from within the 30 km zone around the facility; the inherent spatial heterogeneities in ocean sediments; and the potential for inventory fluctuations due to physical, biological, and chemical processes. Using 210 Pb, 234 Th, 137 Cs, and 134 Cs profiles from 20 sediment cores, coastal sediment inventories were reevaluated. A 137 Cs sediment inventory of 100 ± 50 TBq was found for an area of 55 000 km 2 using cores from this study and a total of 130 ± 60 TBq using an additional 181 samples. These inventories represent less than 1% of the estimated 15–30 PBq of cesium released during the FDNPP disaster. The time needed for surface sediment activities (0 to 3 cm) at the 20 locations to be reduced by 50% via sediment mixing was estimated to range from 0.4 to 26 yr. Due to the observed variability in mixing rates, grain size, and inventories, additional cores are needed to improve these estimates and capture the full extent of cesium penetration into the shallow coastal sediments, which was deeper than 14 cm for all cores retrieved from water depths less than 150 m.


Introduction
The Tohoku earthquake and tsunami of 11 March 2011 led to multiple system failures at the Fukushima Dai-ichi Nuclear Power Plant (FDNPP).Over the next month, cooling water releases and hydrogen explosions resulted in the largest nuclear disaster since Chernobyl.Oceanic inputs included direct cooling discharge, runoff, riverine flow, and an estimated 70 to 80 % of the total atmospheric radionuclide release (Aoyama et al., 2012).While later reports indicate additional releases to the ocean (TEPCO, 2014;Kanda, 2013), the initial 137 Cs activities measured in the FDNPP discharge channel from March to May of 2011 represent the most significant oceanic contribution to date, peaking at 68 million Bq m −3 on 6 April 2011 (Buesseler et al., 2011). 137Cs emerged as the major oceanic contaminant, warranting long-term study due to its ∼ 30 yr half-life and large release, generally estimated to be around 15-30 PBq (Buesseler, 2014;Charette et al., 2013;Povinec et al., 2013;Chino et al., 2011;Stohl et al., 2011). 137Cs and 134 Cs (t 1/2 ∼ 2 years) have been found with comparable activities near the FDNPP and were released with a ratio of ∼ 1 (Buesseler et al., 2011).Although the isotopes are highly soluble in the coastal ocean (Buesseler et al., 2012), a small percentage of the release will have been incorporated into the marine sediments and may remain associated with the seafloor, even after currents transport much of the dissolved phase away from the coast.Initial studies of the sediments near the FDNPP have shown the widespread incorporation of 137 Cs well above the relatively small remaining 1960s fallout signal of 1-2 Bq kg −1 (dry) (NRA, 2014b; Kusakabe et al., 2013;Otosaka and Kato, 2014;Otosaka and Kobayashi, 2013).
The most spatially and temporally extensive sediment data sets have been provided by the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT) and the Tokyo Electric Power Company (TEPCO).TEPCO initially reported some of the highest 134 Cs sediment activities in July of 2011 (TEPCO station 1 at 9600 Bq kg −1 wet) in grab samples collected within the 30 km zone around the FDNPP (NRA, 2014a).However, lacking bulk density measurements, inventory estimates have not been published for this region.MEXT, on the other hand, reported monthly Published by Copernicus Publications on behalf of the European Geosciences Union.E. Black and K. O. Buesseler: Spatial variability and the fate of cesium in coastal sediments multi-corer, dry activities for homogenized 0 to 3 cm core tops from outside the 30 km zone (NRA, 2014b).Kusakabe et al. (2013) used 30 different MEXT locations ranging from 45 to 675 m water depth to estimate a 137 Cs sediment inventory of 38 TBq for the top 3 cm.As both data sets are limited to the uppermost layers of the sediment column, calculations using only these measurements will underestimate the total incorporation of cesium into the sediments.In a more recent effort, Otosaka and Kato (2014) paired MEXT cores with 10 cm Smith-McIntyre grab samples and sectioned cores ranging from 3 to 10 cm and estimated a 134 Cs coastal inventory of 200 ± 60 TBq, decay-corrected to March 2011.
Larger data sets are essential for more accurate sediment inventory calculations due to the spatial variability and local heterogeneity observed in previous coastal sediment studies from this region (Kusakabe et al., 2013;Thornton et al., 2013).Kusakabe et al. (2013) found a 137 Cs activity range of 170 to 580 Bq kg −1 with six successive multicorer casts at nominally the same sampling site (mean and standard deviation of 330 ± 160 Bq kg −1 dry).In addition, Thornton et al. (2013) reported activity fluctuations of < 10 to 1500 Bq kg −1 at sites located between 5 and 10 km of the FDNPP and from 500 to 40 000 Bq kg −1 within 3 km of the FDNPP.These local anomalies were observed over distances of less than 1 km using a continuous (1 hertz) towed gamma ray spectrometer and their magnitude illustrates the importance of high-spatial-resolution sampling.
The evolution and ultimate fate of cesium isotopes in the coastal ocean must be better constrained to assess both the short-term implications of the FDNPP accident and the potential for lasting effects.Reports that bottom-water fish contain higher cesium concentrations than pelagic fish suggest that the sediments could be a continued source to bottomdwelling biota (Buesseler, 2012).A model study of the coastal food chain near Fukushima indicated that an additional contamination source beyond ocean water, mostly likely associated with the sediments, would be necessary to sustain the cesium levels observed in higher-trophic organisms (Tateda et al., 2013).
210 Pb (t 1/2 ∼ 22.2 years) and 234 Th (t 1/2 ∼ 24 days), naturally occurring daughter products in the 238 U decay series, can aid in evaluating the rates of sedimentation and bioturbation occurring in sediments from months to decades (Yang et al., 1985). 234Th is particle reactive and will be scavenged easily in the upper ocean, leading to a potential excess in sediments ( 234 Th ex ).Because of its short half-life, measurable 234 Th ex is generally only observed in the top few cm of the sediment column in areas of rapid and recent mixing.Excess 210 Pb ( 210 Pb ex ) is supplied via atmospheric deposition, from the decay of 222 Rn gas, and scavenging in the water column.In sediments 210 Pb ex represents the divergence from secular equilibrium with 226 Ra (supported 210 Pb).If conditions are relatively stable, the 210 Pb ex inventory in a given area will represent the flux to this location averaged over the last century (∼ 5 half-lives).

Sample collection
Twenty sediment cores ranging in length from 6 to 20 cm were collected during cruise campaigns in May 2012 (R/V Tansei Maru), June and July 2012 (R/V Mirai), May 2013 (R/V Umitaka Maru), and September 2013 (R/V Daisan Kaiyo Maru).Stations were located 2 to 1865 km from the FDNPP (Supplement S1).Individual core tubes were retrieved from a multi-corer and cross-sectioned at sea into 0.5 to 2 cm layers.A sample (plug) of 1 or 7 cm 3 was taken from each layer as the cores were cross-sectioned for density calculations.Sediment layers were preserved in sealed bags and the plug samples in capped vials.Eight samples (0 to 5 cm) from separate core tubes (R1 to R8) were retrieved from a single cast at the location of core 13 for an analysis of local variability.

Grain size analysis
Grain size analysis was performed on a subset of ∼ 3 cm 3 core samples using a Beckman Coulter LS13320 particle size analyzer with capabilities of 0.4 µm to 2 mm.Grain size results are reported as percent clay and percent fines (silt plus clay) averaged over the entire depth of each core.Percent clay and percent fines were calculated by summing the frequency outputs from 0 to 3.86 µm and from 0 to 63.41 µm, respectively.D50 values were determined, which signify the grain size at which 50 % of the sample is smaller or larger by particle count.All grain average size results are reported with standard deviations (SDs).Samples from cores 15 and 16 consisted of up to 42 % of grains over 1 mm by mass and were processed differently due to the counter limitations.To ensure that no prolate grains with a nearly 2 mm axis passed to the counter, a 1 mm sieve was placed over the sample delivery system.The total mass of the sample used, ranging from 5 to 45 grams, was determined by the counter's optimal obscuration range or percentage of light blockage by grains (15 to 25 %).

Isotope measurements
Samples were dried at 40-60 • C for a minimum of 1 day and analyzed using Canberra GCW4030S germanium gamma well detectors for the following energy peaks: 46.5 keV ( 210 Pb), 63.3 keV ( 234 Th), 352 keV ( 214 Pb), 661 keV ( 137 Cs), and 795 keV ( 134 Cs).Samples were counted for 7 to 24 h depending on the time to achieve counting uncertainties of less than 5 % on the primary peaks.Detectors were calibrated using a dilute pitchblende ore standard (US EPA Environmental Monitoring Systems Lab) and river sediment standard (NBS 4350 B).Minimum detectable activities (MDAs) were calculated using 24 h background spectra and efficiencies based on the average sample mass of 16.75 grams and sample volume of 14.5 mL (Currie, 1968).The calculated MDAs in Bq kg −1 were 4.2 ( 210 Pb), 3.2 ( 234 Th), 0.7 ( 214 Pb), 0.4 ( 137 Cs), and 0.8 ( 134 Cs).Activities under the MDAs and those with counting uncertainties over 50 % were reported as not detectable (ND).Total uncertainties for a given sample and isotope (in Bq kg −1 dry) represent the higher of either the counting uncertainty or 7 %.The minimum uncertainty (7 %) is the average percent difference between sample activity results when duplicate measurements were made using the same and different detectors.The total uncertainty is propagated through all activity and inventory calculations for individual sections and full cores. 210Pb activities were decay-corrected to the collection date and adjusted for supported values from 214 Pb (assumed to be at equilibrium with parent 226 Ra) to determine 210 Pb ex .Since an excess of 234 Th generally exists only in the first few centimeters, an estimation of equilibrium 234 Th values at depth were used to determine 234 Th supported , the activity supported by the decay of its 238 U parent, to calculate 234 Th ex (decay-corrected to collection).All 134 Cs and 137 Cs data were decay-corrected to the date of maximum concentrations at the FDNPP (6 April, 2011;Buesseler et al., 2011).By decay correcting, the changes we observe can be attributed to physical and biological processes and not radioactive decay.All final 210 Pb ex , 234 Th ex , and cesium activities for an individual layer are reported as Bq kg −1 dry, and surface activities were calculated as the weighted (relative to layer thickness) average of layer activities for the top 3 cm of each core.
Layer inventories in Bq m −2 were calculated for 134 Cs, 137 Cs, 210 Pb ex , and 234 Th ex using the following relationship: where layer thickness was in meters, activities were in Bq kg −1 dry, and bulk density (kg m −3 ) was equal to the amount of dry mass of sample in the 1 or 7 cm 3 plug divided by the total volume.All layer inventories for each isotope were summed to calculate total inventories for each individual core except for 234 Th ex , which was only reported for the exponentially decreasing surface activities.Because core lengths ranged from 6 to 20 cm, some core inventories would not be representative of the complete isotope profile at that location if 210 Pb ex values remained elevated at depth.We estimated the expected sediment inventories of 210 Pb ex in the coastal sediments near Fukushima, Japan, using an average 210 Pb ex atmospheric delivery flux of approximately 200 Bq m −2 × yr −1 to obtain an expected inventory from atmospheric input of 6400 Bq m −2 (200 × 32 years, mean life of 210 Pb).The atmospheric delivery flux estimate was derived from average monthly 210 Pb deposition measurements from 1993 to 2001 for Tokai-mura (Ueno et al., 2003) and from 2000 to 2001 for Tokyo and Sendai (Yamamoto et al., 2006).Since 210 Pb is particle reactive, in situ scavenging provides an additional source, the strength of which is a function of the 210 Pb : 226 Ra disequilibrium and water depth over which the deficit is found.Since we do not have 210 Pb : 226 Ra profiles, we can only estimate a rough scavenging flux for 210 Pb using a Pacific 210 Pb / 226 Ra ratio of 0.75 (Tsunogai and Harada, 1980) with a deep sea 226 Ra of approximately 33 dpm per 100 kg activity (28.5 • N and 145 • E; Nozaki and Tsunogai, 1976).This flux would be a linear function of water depth (z) over which the disequilibrium applies, and it ranges from 70 to 7000 Bq m −2 for depths of 50 to 5000 m (1-D scavenging supply equal to 226 Ra-210 Pb ×λ × z× 32 years).Thus, in most coastal settings, the dominant supply of 210 Pb is atmospheric deposition, whereas inventories are expected to increase with depth in the shelf and slope due to scavenging processes in the water column above.In addition, with a relatively long residence time for scavenging, ocean margins in general are sites of enhanced boundary scavenging of 210 Pb (Cochran et al., 1990).

Modeling
While the mixing of marine sediments often occurs via spatially and temporally variable bioturbation processes, such as non-local transport, the cumulative result of sediment accumulation, mixing, and isotope decay is often modeled as a diffusive process and can be written where ρ represents density, A activity of radionuclide, t time, w sedimentation rate, and λ represents the decay constant (Cochran, 1985).To generate mixing rates (D B ) using this relationship for each core, steady state, constant porosity (φ), and constant D B with depth (x) were assumed.Equation (2) can be simplified using these assumptions to where A o is the activity at x = 0 (Anderson et al., 1988).Sedimentation can be ignored if Historical 210 Pb ex -derived mixing rates from locations in the western North Pacific for depths greater than 4000 m range from 0.1 to 1 cm 2 yr −1 (Moon et al., 2003;Yang et al., 1985).Sedimentation rates for these locations were found to be 0.0001 to 0.002 cm yr −1 .Using the offshore mixing rates as a minimum for the coastal region near Fukushima, w would need to approach 0.12 cm yr −1 to be considered significant.Therefore, sedimentation was considered negligible over the decades following the FDNPP release and Eq. ( 2) can be further simplified to (Cochran, 1985;Yang et al., 1985) Activity Lower right: core locations 1 and 2. Some of the reported values may underestimate total inventories at the locations where 210 Pb ex continues to the core bottom (Fig. 2).
A dynamic, exponential regression analysis was performed on 234 Th ex and 210 Pb ex activity (Bq kg −1 dry) profile sections of each core using Eq. ( 5).In practice, mixing rates are typically higher in surface sediments, and therefore 234 Th ex profiles were used to derive D B for upper layers, when present, and 210 Pb ex was used below.Bioturbation was assumed to be the dominate process controlling isotope distributions in the sediments, although D B is a diffusive approximation of the combined effect of bioturbation and potential physical mixing processes.
To model the expected changes over time in 137 Cs profiles, we used a pulse input model and the 234 Th ex -and 210 Pb exderived D B s (Cochran, 1985;Yang et al., 1985): where A is the activity at any given depth x, A o is the shallowest surface activity, D B is the mixing rate, and t is the time since the maximum input (6 April 2011; Buesseler et al., 2011).Keeping all other variables constant, we modeled the cesium activity profiles to determine the time needed for average cesium activities from the top 3 cm to decrease by 50 % due to sediment mixing and, hence, the transport of the cesium deeper into the core.Therefore, we have assumed that most of the cesium input to the sediments would track cesium delivery to the coastal waters.Sediment trap data in the North Pacific show rapid transport of cesium to depth, even in deep waters, within days of the input of cesium to the surface ocean (Honda et al., 2013).

Variability between multi-core tubes
From a single multi-corer cast, we retrieved eight separate sample tubes (R1 to R8) and analyzed each for 210 Pb ex , 137 Cs, and 134 Cs.Core 13 was retrieved from the same location during an additional mutli-corer cast (Fig. 1). 137Cs activities in the top 0 to 5 cm of these subcores varied from 45 ± 3 to 220 ± 20 Bq kg −1 and 210 Pb ex from 140 ± 10 to 220 ± 20 Bq kg −1 (Supplement S2).Inventories for 137 Cs in the top 5 cm were 1600 ± 100 to 6500 ± 500 Bq m −2 and 5500 ± 400 to 7800 ± 600 Bq m −2 for 210 Pb ex .Inventories for core 13 for the same 0 to 5 cm layer were 8000 ± 200 and 6400 ± 200 Bq m −2 for 137 Cs and 210 Pb ex , respectively.The large inventory differences for both 137 Cs and 210 Pb ex indicate a high degree of variability between casts and between individual tubes from a single cast.We could not obtain a coefficient of determination (R 2 ) greater than 0.1 for an exponential, logarithmic, or linear regression of 134 Cs and 210 Pb ex inventories, which confirms that the factors controlling inventories at the local scale vary for the two isotopes (Supplement S3).
Although we observed large tube-to-tube differences in the cesium activities, the grain size results for each of the 8 subcores were relatively similar in magnitude (Supplement S4).D50 values ranged from 19 to 41 µm, percent clay from 7.3 to 12, and percent fines from 66 to 85.Although a study of the Irish Sea found elevated 137 Cs activities with increasing percentages of higher surface area particles (Poole et al., 1997), such as clays, we could not find any relationship (exponential, linear, or logarithmic) with an R 2 value of greater than 0.1 between the three grain size parameters and cesium activities or inventories at this single site (Supplement S5). 210Pb ex activities showed some relationship with percent clay in the replicate cores (R 2 values ranging from 0.34 to 0.36).While Kusakabe et al. (2013) posited that low bulk densities could indicate finer grain sizes and abundant organic matter content, we did not see evidence of this in the replicate cores.Although particle size characteristics do not appear to control local differences in radionuclide activities, where variations in average particle size and abundances of particle types (clays, etc.) are relatively small, they may be important over larger regional scales where mineralogical differences are greater (see Sect. 3.5).

Zonal divisions
Previous studies have delineated various coastal regions to estimate the total cesium delivery to the sediments near the FDNPP.Kusakabe et al. (2013) used a boundary from approximately 35.5 to 38.5 • N, while Otosaka and Kato (2014) used incremental isobaths running from 35.67 to 38.50 • N to create eight separate zones (0 to 1500 m).For consistency, we used 35.5 to 38.5 • N for our northernmost and Table 1.Grain size results and isotope inventories for sediment cores by zone.A general size was assigned to cores 1, 2, 3, and 5 after visual inspection, as grain size analysis was not possible for these cores.All other cores are described according to the Udden-Wentworth scale with sediment classifications C (clay), M (silt), and S (sand), and size qualifiers vf (very fine), f (fine), m (medium), c (coarse), and vc (very coarse).The uncertainty on all inventory values represents the propagation of the counting uncertainty for each section (minimum 7 %) as discussed in Sect.2.3.D50, % clay, and % silt and clay represent the average value for all layers within a given core and are reported with standard deviations.

Zone
No. Water southernmost zonal boundaries and isobaths for eastern and western divisions (Fig. 1 and Table 1).We divided the sediment reservoir into five zones based on inventories, grain size, and mixing rate estimates from our 20 sediment cores (see Sect. 3.3 to 3.7).The northern coastal zone (NCZ, n = 6) and southern coastal zone (SCZ, n = 3) are bound by the 150 m isobath to the east.The northern and southern boundaries of the NCZ and SCZ were shortened to 38.20 and 36.25 • N, respectively, after consideration of additional surface cores from MEXT (Kusakabe et al., 2013) (see Sect. 3.9).The mid-coastal zone (MCZ, n = 5) is bound by the 800 m isobath to the east and the Japanese coast, NCZ, and SCZ to the west.The offshore zone (OZ, n = 4) is bound to the west and east by the 800 and 4000 m isobaths, respectively.We included core 3 (4066 m) in the OZ because of its proximity to core 4 and an underrepresentation of deeper cores in this zone.The remaining cores (n = 2) are located in the abyssal zone (AZ).All result and calculations in the following sections will be discussed relative to these zones.

Grain size analysis
Two cores from the OZ and all cores from the NCZ, SCZ, and MZ were analyzed by layer with a Beckman Coulter counter, and the results were averaged by core (Table 1).
The remaining four cores were visually assessed with a hand lens and assigned a general size class relative to the Udden-Wentworth classification scheme.The NCZ cores contained the lowest percent clay (0.2 to 0.8 %) and yielded the highest D50 values, ranging from 160 to greater than 690 µm.These cores were composed of relatively well sorted sands, and D50 standard deviations were generally less than 10 % of the D50 value.The SCZ cores contained more poorly sorted grain assemblages and could be characterized as mostly siltdominated with some fine sands.D50 values for these cores ranged from 33 to 80 µm, an order of magnitude lower than those observed in the NCZ, and the percent clay ranged from 5 to 9 %, an order of magnitude higher than those in the NCZ.The MCZ cores were fairly similar in characteristics to the SCZ cores and ranged from 5 to 10 % clay and in D50 from 40 to 110 µm.The two OZ cores were very fine grained and had relatively similar D50 and percent clay values of approximately 18.5 µm and 12.5 %, respectively.The average values for D50, percent fines, and percent clay reflected wide variations in grain size from the coastal to offshore regions, although for a given core the changes in grain size between individual layers from top to bottom were generally small.We saw that, as water depth increased to 5900 m, average grain sizes generally transitioned from coarser sands and granules (0.5 to 4 mm) to very fine silts and clays (< 63 µm).This broad relationship did not always hold true in the nearshore, however, as the NCZ and SCZ cores differed in average grain size characteristics but not in water depth.Although a few cores showed D50 layer variations up to 220 µm over 2 cm increments, no observable pattern ex-isted when these fluctuations were compared with changes in isotope activities.Conversely, many large layer-to-layer cesium fluctuations, like those seen in core 17, could not be explained by changes in grain size.Despite the likelihood that some or all of the cores might contain tsunami deposits, there was no grain size evidence of fining-upward sequences in these cores, sharp layer contacts with largely differing grain sizes, or definitive indicators that would allow us to separate potential tsunami layers from sections impacted by bioirrigation, bioturbation, or local physical processes (Sakuna et al., 2012).

Cesium activity profiles
Average surface activities for the top 3 cm in the cores ranged from 2.1 ± 0.1 (core 2, C2) to 630 ± 30 (C20) Bq kg −1 dry for 137 Cs and from 0 (C2) to 550 ± 30 (C20) Bq kg −1 dry for 134 Cs (Supplement S6).The AZ cores contained no detectable 134 Cs.The average surface activities in the OZ ranged from 0 (C3) to 12 ± 1 (C6) Bq kg −1 dry for 134 Cs, while the MCZ activities ranged from 3 ± 0.5 (C7) to 57 ± 2 (C11) Bq kg −1 dry for 134 Cs.NCZ cores 15 through 19 had a similar range of average surface activities from 19 ± 1 (C16) to 50 ± 2 (C18) Bq kg −1 dry.While most surface 134 Cs activities in the NCZ were unexpectedly low, despite their proximity to the FDNPP, core 20, located within 3 km of the FD-NPP, had the highest average surface activity overall.With the exception of this core, the cores in the SCZ had the greatest average surface activities of all zones, with 134 Cs values ranging from 170 ± 10 (C13) to 230 ± 10 (C14) Bq kg −1 .
Cesium profiles varied strongly in shape and penetration depth between zones (Fig. 2).Cores in the AZ and OZ had either fallout-only profiles with no 134 Cs and very low 137 Cs or rapidly decreasing 134 Cs and 137 Cs activities with depth.FDNPP-derived cesium penetration was limited to the top 3 cm in these zones.The MCZ cores generally showed similar exponentially decreasing cesium activities with depth with the exception of core 11, which showed a pronounced peak between 1 and 4 cm.Generally, the penetration depth of 134 Cs was deeper here than in the AZ and OZ cores.The SCZ cores maintained substantially higher cesium activities from 0 to 8 cm relative to most other cores, and we observed both exponentially decreasing activities and vertical profiles below 8 cm. 134Cs was found to a depth of 18 cm in core 12 and over the full length of cores 13 (19 cm) and 14 (16 cm) in the SCZ.While the NCZ cores had similar surface activities to those in the MCZ, most cores in this zone showed distinctive semi-vertical profiles from the surface to the coring depths.Core 17 was the only exception and the only core to exhibit large fluctuations in activity with depth.Because we observed 134 Cs at all core bottoms and without decreasing trends in most of the NCZ cores, we could not estimate maximum penetration depths.

Cesium inventories
Total 134 Cs and 137 Cs inventories ranged from 0 (C1) to 74 000 ± 2000 (C20) Bq m −2 and 21 ± 1 (C3) to 73 000 ± 2000 (C20) Bq m −2 , respectively (Table 1).Our inventory range for 137 Cs was almost 4 times larger than those reported in Kusakabe et al. (2013; approximately 100 to 18,000 Bq m −2 ) and Otosaka and Kato (2014; approximately 100 to 19 100 Bq m −2 ).This could be attributed to both greater core lengths and larger total area sampled, including the 30 km zone around the FDNPP.Cesium inventories generally decreased with water depth.AZ and OZ inventories were less than 400 Bq m −2 for both isotopes, and the MCZ contained only one core inventory above 1000 Bq m −2 .Excluding core 20, we observed cesium inventories on the order of 1500 to 7000 Bq m −2 in the NCZ and 4000 to 36 500 Bq m −2 in the SCZ.The cesium inventories for core 20 were double that of the next-highest value and reflected a proximity to the FDNPP.Pre-accident cores from the broader western North Pacific have been reported to contain tens to thousands of Bq m −2 of 137 Cs (Moon et al., 2003;Lee et al., 2005).Assuming a weapons-testing delivery of 2500 Bq m −2 for 30°to 40 • N (Bowen et al., 1980) and 1-5 % delivery to the sediments, we expect an average sediment inventory of 137 Cs less than 50 Bq m −2 prior to Fukushima.We observed inventories consistent with the presence of substantial weapons-testing fallout in the AZ and OZ cores with inventory ratios for 134 Cs / 137 Cs of 0 to 0.86.Larger inventories and 134 Cs / 137 Cs ratios of ∼ 1 in most of the MCZ, NCZ, and SCZ cores suggested negligible contributions from weaponstesting 137 Cs, relative to the larger and more recent FDNPP source.
Since many prior studies have focused on the 0 to 3 cm surface layer, we also calculated the percentage of total inventory that lie below 3 cm within each of our cores (Table 1).Percentages generally increased with shallower water depths.We observed 75 ± 6 and 75 ± 5 % for 134 Cs and 73 ± 7 and 75 ± 5 % for 137 Cs (mean ± SD) below 3 cm as a fraction of the total inventory in the NCZ and SCZ, respectively.In the MCZ, the percentage of total inventory below 3 cm was highly variable, ranging from 0 (C7 and C10) to 33 % (C8) for 134 Cs and from 10 (C10) to 36 % (C8) for 137 Cs.The average inventory below 3 cm in the MCZ cores attributed to Fukushima ( 134 Cs) was 15 ± 16%, which agreed closely with the Otosaka and Kato (2014) 134 Cs average of 19 % from core locations within the MCZ.When we combined the two data sets for the MCZ (n = 15) the average inventory below 3 cm was 18 ± 16 %.In the OZ we saw no 134 Cs below 3 cm in any of the cores and a range of 0 (C3) to 32 % (C4) for 137 Cs.We did not observe any detectable amount of either isotope below 3 cm in the AZ.
The core inventories from this study reflect changes in three main factors: distance from the FDNPP, water depth, and grain size.Proximity impacted core 20, located within 3 km of the FDNPP (Fig. 3a). Figure 3a, shows a general de- crease in cesium inventories with increasing distance from the FDNPP and water depth, despite cesium inventories for cores 1 to 19 showing little to no statistical relationship to proximity or water depth (exponential regression R 2 values < 0.25).Grain size was a controlling factor in the nearshore NCZ and SCZ, where cores were sampled from similar water depth ranges and yet contained widely varying inventories (Table 1, Fig. 3a).Although no trend was observed in our replicate core tubes between grain size and cesium activities, the larger differences in grain size between the SCZ and NCZ (D50s from ∼ 30 to 700 µm) did correspond to changes in total 137 Cs inventories.The NCZ and SCZ grain size trends mimic those observed in the Irish Sea, where an estimated 41 PBq of cesium released over 40 yr has been preferentially incorporated into an area known as the Sellafield mud patch (Poole et al., 1997).In summary, proximity dominated within the 3 km zone of FDNPP, despite the high sand content of core 20 (D50 of 160 µm, percent sand ∼ 94 %); grain size was the most important difference between the NCZ and SCZ; and water depth (higher particle fluxes are generally observed over more productive coastal waters) may have contributed to cesium variability in locations such as the MCZ and OZ, where water column activities were relatively low.

210 Pb ex and 234 Th ex activities and inventories
210 Pb ex surface activities (0 to 3 cm average) ranged from 12 ± 3 (C20) to 1900 ± 100 (C1) Bq kg −1 and generally increased with water depth, as expected, due to the increased input from water column scavenging (Supplement S6, Figs. 1  and 2). 210Pb ex activities were lowest in the NCZ and highest in the AZ.Intrazonal 210 Pb ex activity ranges were smaller than for cesium.Similar to the vertical cesium activity profiles, 210 Pb ex activities remained relatively constant with depth in the NCZ.In the other zones, activities generally decreased with core depth but also exhibited a variety of profile shapes.Generally, the termination of 210 Pb ex , the depth where all 210 Pb is supported, was not attained within our sampling depths, with the exception of cores 3 and 6 (Fig. 2).
210 Pb ex inventories, ranging from 2700 ± 200 (C18) to 28 000 ± 1000 (C1) Bq m −2 , reflect changes in grain size, water depth, and local processes -the factors that also impacted cesium inventories (Table 1, Figs. 1 and  3b). 210Pb ex inventories generally increased with water depth as expected from scavenging.The NCZ values were slightly lower than anticipated (average inventory of 4000 ± 1000 Bq m −2 ) based on our expected atmospheric inventory calculation of 6400 Bq m −2 , and the SCZ values were more than twice as high as expected (average inventory 20 000 ± 10 000 Bq m −2 ).The SCZ is therefore a deposition center, receiving additional particles via horizontal transport.These trends are supported by the prevalence of sands in the NCZ cores and higher percentages of fines in the SCZ cores, similar to what has been observed with cesium inventories (He and Walling, 1996).An exponential regression of 210 Pb ex inventories versus percent clay in the NCZ and SCZ indicated a strong relationship (R 2 > 0.9) between grain size and inventories.The MCZ, OZ, and AZ core inventories of 210 Pb ex were generally 1 to 4 times higher than atmospheric delivery.This finding is similar to along the shelf and slope of the northeastern USA (Buesseler et al., 1985(Buesseler et al., /1986) ) and can be attributed to both scavenging within the water column, which is higher with increasing water depth, and horizontal transport and faster scavenging rates over boundary sediments in general.Evidence of boundary scavenging of 210 Pb has been shown for the western North Pacific by Cochran et al. (1990) based upon an analysis of deep sediment cores and the expected inventories from water column scavenging and atmospheric sources.
Of the 17 cores that were analyzed rapidly enough to measure 234 Th ex , 15 contained 234 Th ex activities ranging from 17 ± 3 (C19) to 1300 ± 100 (C12) Bq kg −1 in the top 0 to 0.5 cm.Surface (0 to 3 cm) 234 Th ex inventories peaked at 2400 ± 300 (C14) Bq m −2 in the SCZ.Most cores showed classic exponential 234 Th ex profile shapes with higher values at the surface that dropped to near-zero excess values within 1 to 3 cm.Although the lowest 234 Th ex inventories were found in the OZ and NCZ and the highest were generally located in the SCZ and MCZ, intrazonal variability was high, and more often than not inventories did not decrease with increasing depth.

D B estimates
Since all of the cores showed evidence of sediment mixing, we estimated local mixing rates using 234 Th ex and 210 Pb ex activity profiles and a steady-state diffusive mixing model (Eqs. 2 and 5; see Fig. 4 for example profile).Mixing rates in the MCZ, OZ, and AZ ranged from 0.7 (C9) to 9.6 (C10) cm 2 yr −1 for 234 Th ex -derived estimates and from 0.06 (C1) to 3.7 (C10) cm 2 yr −1 for 210 Pb ex -derived estimates (Table 2).At a given location, 234 Th ex -derived estimates were generally higher as they reflect processes occurring in the uppermost active sediment layers (Aller and Cochran, 1976).Rates in the SCZ and NCZ suggested intense mixing, with full core 210 Pb ex -derived estimates starting at 11 cm 2 yr −1 and the majority of rates being unquantifiable due to the vertical 210 Pb ex profiles.The wide range in rates was not unexpected with cores from diverse sedimentary environments and water depths.While 234 Th ex -derived mixing rates were not correlated with water depth (R 2 < 0.1), we observed an exponentially decreasing trend in 210 Pb ex -derived mixing rates with water depth (R 2 = 0.96), as observed in other coastal regions where biomass, species diversity, and bioturbation decrease with depth (Henderson et al., 1999;Joydas and Damodaran, 2009).
Although 234 Th ex was absent from some of the deepest AZ, OZ, and MCZ cores, we succeeded in obtaining 210 Pb ex and 234 Th ex mixing rates for most cores located in these zones.In general, the activity profiles exponentially decreased with depth (Figs. 2 and 4  210 Pb ex -derived mixing rates agreed well with the observed historical rates from this region of 0.1 to 1 cm 2 yr −1 (Moon et al., 2003;Yang et al., 1985).Cores from the shallowest regions (NCZ and SCZ) were the most difficult to fit with the simplified sediment mixing-only model (Eq.5).The SCZ cores generally contained three layers of mixing: a 1 to 3 cm moderately fast mixing zone at the surface, a 5 to 10 cm rapid mixing layer with an almost vertical profile, and a deep mixing zone with similar mixing rates to the surface layers.Equation ( 5) could not be used at all to derive 210 Pb ex mixing rates for any of the NCZ cores due to the almost vertical activity profiles throughout.In these instances, the observed profile trends could have resulted from extremely rapid mixing or from resuspension and deposition resulting from a tsunami or storm-related event, although we did not see evidence of individual storm deposits in the grain size analysis (see Sect. 3.3).

Cesium modeling of changes in surface activities
To explore how quickly sediment mixing could potentially reduce surface (0 to 3 cm) cesium activities over time, we calculated the change in surface cesium activity, assuming a pulse-like input (Eq.6).The 234 Th ex -derived mixing rate was used for the depths specified in Table 2, and the corresponding 210 Pb ex -derived mixing rate was used for the rest of the core.If no rate existed for one of the isotopes, the mixing rate from the other was applied to all depths.The decay of 137 Cs, although small over 5 to 10 yr, was subtracted from the profile activities after all other calculations to show changes due to mixing only.In general, the measured cesium activity profiles for the top 3 cm matched fairly well with the pulse model output for our sampling dates using the local bioturbation rate.Below 3 cm, the model generally followed the AZ, OZ, and MCZ measured cesium profiles (see Fig. 4 for example profile) and underestimated the NCZ and SCZ profiles.
The time until the surface (0 to 3 cm) activities decreased by 50 % ranged from 0.4 (C14) to 26 (C1 and C3) years for the 20 cores (Table 2).The SCZ times ranged from 0.4 (C14) to 0.9 (C13) years and the MCZ times from 2 (C10) to 15 (C8) years.We could not model any of the cesium profiles in the NCZ using 210 Pb ex mixing rates, and the two with 234 Th ex mixing rates yielded estimates from 12 (C16) to 14 (C15) years, which seemed unrealistically long when considering the water depth and shape of the vertical profiles.
The longest time period for decrease will occur in the slowmixing OZ and AZ regions, whose 50 % reduction time varied from 6 (C4) to 26 (C1 and C3) years.

Total seafloor cesium inventory calculations
We estimated that the marine sediments contained approximately 100 ± 50 TBq of 137 Cs for an area of 55 000 km 2 (Tables 3a and b).The final totals for each zone were calculated by finding the mean inventory of all cores within that zone and multiplying by the area of the zone.Zonal and final totals are reported with standard deviations for data set (1), using cores from this study only, in Table 3b.Inventory calculations should represent underestimates of actual reservoir totals because the profiles in the highest inventory zones, the NCZ and SCZ, still showed Fukushima 134 Cs in their deepest sample layers.We did not calculate an inventory value for the AZ, which is extensive but should not significantly contribute to overall inventories due to low cesium activities.Furthermore, the cores from the greatest water depths, 1 and 2, did not contain Fukushima 134 Cs, although it has been detected in sediments as far offshore as the Japan Trench (water depth of > 7000 m; Oguri et al., 2013).
To improve our inventory estimate, we added cores recovered in February 2012 by MEXT (Kusakabe et al., 2013) and from August to November 2011 reported in Otosaka and Kato (2014) (OTKA).These provided an additional 50 locations, referred to as data sets (3) and (2), respectively.The inventories of the shorter sediment cores were increased using the average percent 134 Cs inventory below 3 and 10 cm determined for each zone when applicable (see Sect. 3.5).Cesium activity and inventory values from the MEXT and OTKA data sets were not changed if the depth sampled was sufficient to capture the full extent of cesium penetration.The additional locations did not impact the sediment inventory estimate, now 100 ± 40 TBq, and did not substantially change the distribution of cesium between zones (Tables 3a  and b).
In a final effort to further improve our inventory estimates, we utilized all earlier sampling dates from the 30 MEXT locations in Fig. 5 and calculated a total sediment reservoir estimate of 130 ± 60 TBq (n = 199, Table 3b).These locations were sampled at approximately monthly intervals starting June 2011 and are referred to as data set (4).As we have observed, the range in local variability can be almost as high as that on the zonal or regional scale (Kusakabe et al., 2013;Thornton et al., 2013), so additional sampling at one site should not bias the zonal averages towards these repeated locations.However, because of the much earlier sampling dates for the MEXT cores relative to our last date in September 2013 and the 3 cm sampling length, we considered the sensitivity of our inventory estimate to mixing, which could transport cesium to deeper layers between sampling events.Using Eq. ( 6) and the estimates of mixing rates, we determined the average percent of cesium below 3 cm for each zone in June  (Kusakabe et al., 2013) and OTKA samples from August through November of 2011 (Otosaka and Kato, 2014) were adjusted as necessary by zone to include estimated inventories deeper than 3 and 10 cm, respectively (Sect.3.5).
2011 and used these smaller percentages instead of those in Sect.3.5 to adjust the MEXT data.After this adjustment, the total inventory estimate decreased slightly to 100 ± 60 TBq.
The three approximation methods all resulted in similar total 137 Cs sediment inventories, ranging from 100 to 130 TBq (Tables 3a and b).The OZ comprised more than half the entire area but consistently totaled only 4 to 6 % of the 137 Cs sediment inventory.The MCZ represents 30 % of regional surface area and contained between 15 and 18 % of the total 137 Cs.The NCZ was split into two subzones to account for the extremely high inventory observed within the 3 km radius of the FDNPP.Based on Thornton et al.'s continuous tow data, the area immediately around the nuclear facility (∼ 3 km) showed the largest number of high 137 Cs Bq kg −1 anomalies relative to adjacent sediments (Thornton et al., 2013).Core 20 confirmed that this region was distinctly higher in activity and warranted its own subzone, so as not to skew the NCZ inventory average.The 3 km subzone adjacent to the FDNPP accounted for approximately 1 % of the total inventory.Although this inventory could change with the addition of new and deeper cores, it would not change the total regional inventory more than a few terabecquerels, even assuming a mixed layer depth of 50 cm (cesium penetration to 50 cm; Walbran, 1996), because the 3 km subzone only comprises 0.03 % of the total area.The NCZ and the SCZ, on the other hand, contributed the most cesium by far (over 70 %),  3a and b).
Figures 5 and 6 show the spatial distribution of cesium inventories for all zones surrounding the FDNPP and illustrate both the high intrazonal variability and the overall decrease of inventories with depth.Included are cores from this study (n = 18), from the February 2012 MEXT sampling dates (n = 30; Kusakabe, 2013), and from all OTKA dates (n = 20; Otosaka and Kato, 2014), the latter two containing adjusted cesium inventories when applicable (see Sect. 3.5).Best-fit exponential regressions of these inventories versus water depth are shown in Fig. 6.As observed with the cores from this study, the complied inventories from 0 to 150 m in the NCZ and SCZ did not reflect a significant relationship with water depth (R 2 < 0.01), which we suggest reflects the importance of grain size distribution in these zones (Fig. 3a).

Conclusions
Our 137 Cs sediment inventory estimates of 100 to 130 TBq for 55 000 km 2 (0 to 4000 m water depth off Japan) represent the most comprehensive attempt to date for quantifying FDNPP cesium incorporation into marine sediments.The coastal sediments contain the majority of the total TBq delivered, with inventories ranging from 100 to 125 TBq for 24 100 km 2 (0 to 800 m water depth).With expanded spatial coverage inside the 30 km radius of the FDNPP and improved vertical resolution in the NCZ and SCZ, our coastal inventory estimates fall between the 38 TBq of Kusakabe et al. (2013) and the 200 TBq of Otosaka and Kato (2014).Kusakabe et al. (2013) had calculated only surface inventories for the upper 3 cm for a similar area, thus providing a lower limit for total coastal inventories.Otosaka and Kato's (2014) estimate of 200 TBq is higher because it was calculated using exponential water depth versus cesium inventory relationships that were derived without any NCZ cores, which effectively applied the relatively elevated SCZ core inventories to the entire NCZ and SCZ (Fig. 5).
Our zonal analysis of cesium inventories, in conjunction with 210 Pb ex and grain size analyses, has provided an assessment of the total cesium inventory that will inform future studies.We have identified key areas where more cores may improve inventory estimates.First, additional sampling is needed to supplement sparse coverage in two regions: north of the FDNPP along the coast from 37.70 to 38.20 • N in the MCZ, and in locations deeper than 2000 m in the OZ.With the large observed spatial variability from scales of 1 m 2 (between multi-corer tubes) to 10 000 m 2 (among zones), greater numbers of cores are essential to better constrain our estimates at present.Second, while most MCZ and OZ cores are long enough to capture complete cesium inventories, in the NCZ and SCZ cesium had penetrated deeper than 18 to 20 cm.More extensive and deeper coring in both zones is needed to determine ultimate penetration depths and full inventories.
Both the relative inventories of cesium in the coastal environment and estimated zonal 'recovery' timescales have implications for bottom-dwelling biota and those that may consume demersal species.If we assume a total 137 Cs release to the ocean of approximately 15 to 30 PBq from the Fukushima accident (Buesseler, 2014), our calculated sedimentary inventories comprise less than 1.0 % of the total release in 2011.However, cesium concentrations in the water are decreasing and only 15 TBq remain in coastal waters (Buesseler, 2014).Therefore, the sediment reservoir currently represents approximately 86 to 90 % of the total 137 Cs inventory (water and sediment) off the coast of Japan.Rapid mixing in some locations may decrease biological access to these high activity sediments by transporting cesium isotopes deeper and therefore decreasing the overall activity of the more easily resuspended surface layers.While the MCZ and OZ require more time to reduce surface activities by 50 %, they do not pose as much of an environmental hazard because of their relatively low initial activities.The areas of greatest concern are the coastal zones shallower than 150 m water depth.Additional biological studies should take note of these zonal hot spots as cesium levels may remain elevated here much longer despite faster observed mixing rates.
In addition to bioturbation, other processes impacting cesium distributions in the coastal ocean require further study.Yamashiki et al. (2014) found that suspended particle loads in the Abukuma River basin near Fukushima delivered over 5 TBq of cesium to the NCZ in a 10-month period.A typhoon that occurred during this time contributed more than half of this inventory over only 8 days.While 5 TBq would account for only 4 to 5 % of the total sediment reservoir, depending on where and how the particles are transported, this contribution could create local hot spots.Chemical remobilization rates of cesium from sediment pore waters have not been published for this area of Japan but are likely less than 0.1 TBq per month for cesium, based on previous studies of the Irish Sea (Mitchell et al.,1999), and could decrease the total coastal inventory.Even though these source and loss terms are relatively small compared to the total cesium inventory of the sediments, only long-term monitoring within the study region will indicate whether decay alone or other factors are controlling cesium activities and distributions over the next decade.
The Supplement related to this article is available online at doi:10.5194/bg-11-5123-2014-supplement.

Figure 1 .
Figure1.Core locations with 210 Pb ex inventories.Larger map: core locations 3 to 20 and the northern coastal (NCZ), southern coastal (SCZ), mid-coastal (MCZ), offshore (OZ), and abyssal (AZ) zones.Lower right: core locations 1 and 2. Some of the reported values may underestimate total inventories at the locations where 210 Pb ex continues to the core bottom (Fig.2).

Figure 3 .
Figure 3. 137 Cs and 210 Pb ex inventories relative to water depth and D50.Grain size distribution is shown for 16 of the 20 cores, including all of the NCZ, SCZ, and MCZ cores.Isotope inventories for 134 Cs (a) and 210 Pb ex (b) show general depth trends with prominent grain size effects in the NCZ and SCZ.

Figure 4 .
Figure 4. Example model best-fit regressions for 210 Pb ex , 234 Th ex , and 137 Cs for core 9. Mixing rates derived from the 210 Pb ex . 234Th ex profiles in the left panel were used in the cesium pulse model for the right panel.In general, the model similarly fit the profiles from the MCZ, OZ, and AZ.Model fits for the SCZ and NCZ were less successful due to almost vertical profiles for some cores.
), and the AZ and OZ www

Figure 5 .
Figure 5.A compilation of sediment 137 Cs inventories from the coastal region around Fukushima, Japan.Core inventories from this study are shown as measured from May 2012 to September 2013.MEXT cores from February 2012(Kusakabe et al., 2013) and OTKA samples from August through November of 2011(Otosaka and Kato, 2014) were adjusted as necessary by zone to include estimated inventories deeper than 3 and 10 cm, respectively (Sect.3.5).

Figure 6 .
Figure 6.Spatial variations in 137 Cs sediment inventories with water depth.The measured and adjusted inventories from Fig. 5 are plotted against water depth.The upper figure shows the exponential regression for core inventories from 0 to 150 m, excluding core 20.The lower figure shows all inventories and the regression for cores located from 150 to 1500 m.

Table 2 .
Mixing rates and model results for sediment cores.Analysis of samples from cores marked NP was not possible because no 234 Th ex was present or because of insufficient trends in 234 Th ex or 210 Pb ex .Due to the limited number of surface 234 Th ex points, the uncertainties reported represent the higher of the standard uncertainty of the model fit for variable D B or 25 % of D B .The last column contains values representing the time in years since the Fukushima maximum for surface concentrations (0-3 cm) to decrease by 50 % via mixing (modeled).The uncertainties reflect the mixing rate(s) used in the model.

Table 3b .
Kusakabe et al. (2013) results by zone and data sets used.Estimates of total marine sediment 137 Cs inventories were found for this study only (n = 18) and by including data sets from Otasaka and Kato (2014) andKusakabe et al. (2013)that were adjusted by zone based on calculated average percent inventories observed below 3 to 10 cm.See Table3afor group numbers.