Sediment cores ranging from 20 to 32 cm in length were taken at the Compass Station (44° 41’ 30” N, 63° 38’ 30” W, Figure 1 ) on February 12^th, April 12^th, May 7^th, and August 2^nd of 2019 using a DC Denmark four-core multicorer. The cores were extruded and sectioned in 1-4 cm intervals, with greater resolution in the upper 5 cm. In August, two sets of paired cores (four cores total) were homogenized in 1 cm increments to provide more mass for analysis in the upper 5 cm and to test for spatial heterogeneity. Methods follow the general processing and counting procedures in Black and Buesseler (2014). Cores were analyzed for ²³⁴Th (63.3 keV) and ancillary isotopes at the Radionuclide Facility at Dartmouth College using Canberra Intrinsic High Purity Germanium Detectors. A visual inspection indicated that the sediments were fine-grained, organic-rich, and relatively homogenous. Porosity (i.e. water content) was determined for the upper 5 cm using the mass difference between wet and dry sediment subsamples, correcting for salt content, and assuming a sediment density of 2.65 g cm^-3 (Burdige, 2006). Sediment uncertainties for ²³⁴Th activities are the propagation of gamma counting errors and an assessment of potential variability in porosity and density estimates. Carbon content of sediments was analyzed in the CERC.OCEAN facility at Dalhousie University using a Costech Instruments Elemental Combustion System 4010. The average carbon content (± s.d.) did not differ between acid-fumigated samples (n = 29, 5.5% ± 0.5%) and non-fumigated samples (n = 39, 5.7% ± 0.8%), suggesting that most of the total carbon is organic in nature. Therefore, it is assumed that the reported ‘total’ carbon content here can be used interchangeably with ‘organic’ carbon content.

Niskin bottles (2L to 10L) were used to collect seawater samples on April 11^th, May 9^th, and August 2^nd and temperature, salinity, fluorescence, and oxygen profile data were collected with a Seabird SBE 19plus CTD. For size-fractionated particulate ²³⁴Th measurements (>51 µm and 1-51 µm), the ~42 liters were collected at three water depths. The ~42 liters collected at each depth were temporarily stored in cleaned plastic barrels prior to filtration on ship using a custom-built, 142 mm vertical filter apparatus. Elevated barrel water was vacuum pumped through a 51 µm pore size nitex filter and a pre-combusted 1 µm pore size quartz microfiber filter (QMA) in sequence. Filters were visually examined for swimmers and a 400 µm pre-filter test found none (August sampling). Particulate material from the nitex screens was rinsed onto pre-combusted 25 mm QMA using filtered seawater (<1 µm) and dried overnight. The 142 mm QMA were dried similarly and then a 25 mm punch was collected for successive ²³⁴Th and carbon measurements. The particulate samples were counted within 6 to 12 days of filtration using low-level beta GM multicounters at Woods Hole Oceanographic Institution (WHOI) and then again at least 5 months after filtration. Final counting was used to determine the residual particulate β activity derived from beta counting after all unsupported ²³⁴Th has decayed away (RA_P234). RA_P234 was first described in Lin et al. (2016) as a resuspension proxy. After beta counting, particle filters were analyzed for total carbon content at the CERC.OCEAN facility.

For total ²³⁴Th analyses, 4L seawater samples were collected with Niskins and prepared for counting with the precipitation method detailed in Pike et al. (2005) and amended in Black et al. (2018). In summary, the bottles were acidified the day of sampling with nitric acid and then spiked with 1 mL of 2.8 dpm g^{-1 230}Th yield monitor. After ~8 hours of equilibration, solutions of KMn0₄ and MnCl₂ were added to the 4L bottles. The pH was raised above 8 to allow for precipitation. After ~8 hours custom filterheads were added to the bottles and the total ²³⁴Th was collected (as a precipitate) on 25 mm QMAs. The initial counting of particulate and total ²³⁴Th samples was performed using low-level beta GM multicounters at WHOI within 4 to 12 days of filtration. After all beta counting was complete, a ²²⁹Th recovery yield monitor of 7.7 × 10³ dpm g^-1 was added for the filter digestion and ICP-MS analysis. The average recovery was 83% with a s.d. of 13.5% and median of 85%. The uncertainties on total ²³⁴Th are derived from counting statistics and the extrapolation of errors associated with sample processing (i.e., mass and volume measurements, ICPMS recovery analysis).

The ²³⁸U-²³⁴Th method is an established technique for quantifying the particulate flux of chemical constituents in the open ocean on time scales from weeks to months (e.g. Buesseler et al., 1992; Benitez-Nelson et al., 2001; Benitez-Nelson et al., 2007). ²³⁴Th is particle reactive and originates from the decay of ²³⁸U, which is conservative with respect to salinity in typical open ocean environments (e.g. Owens et al., 2011; Not et al., 2012). A deficit of ²³⁴Th with respect to ²³⁸U is created when ²³⁴Th is transported out of surface waters via sinking particles more rapidly than the rate at which ²³⁴Th and ²³⁸U come into secular equilibrium with one another (i.e., when the ²³⁴Th present is supported by ²³⁸U). The particulate flux of ²³⁴Th in dpm m^-2 d^-1 (P_Th ) can be determined using the activity balance:

where Th is the total thorium-234 activity (dpm L^-1), U is the total uranium-238 activity (dpm L^-1), λ_Th is the decay constant of ²³⁴Th (0.0288 day^-1), and V represents the sum of the advective and mixing-diffusive terms. In steady state conditions, when it can be determined or reasonably assumed that no significant change in thorium activities occurred over the time period of interest, the calculation for P_Th at any depth z becomes:

The physical processes potentially impacting a local ²³⁴Th water column budget can be further described by:

where u is the zonal velocity, v is the meridional velocity, w is the upwelling-downwelling velocity, K are dispersion coefficients, and the gradients (∂ terms) are W-E, S-N, and vertical, respectively. The signs of the physical factor terms could change in time or space. The physical processes are accounted for with Equations (2) and (3) or, if 1-D steady state assumptions are valid for the region and time period, Equation (2) can be simplified to:

These determinations and assumptions are examined further in the Discussion.

The deficit of ²³⁴Th in the water column with respect to parent ²³⁸U reflects the removal of ²³⁴Th via particle settling. Therefore, in a 1-D system (i.e. no lateral losses and no significant temporal changes in ²³⁴Th activities), and assuming large particle settling rates of 10–150 m d^-1 (McDonnell and Buesseler, 2010), we anticipate that the excess of ²³⁴Th in surface sediments is roughly equal to the deficit of ²³⁴Th in water column in the shallow Bedford Basin (~70 m water depth). Similar to Equation (4), the sediment accumulation flux of ²³⁴Th (P_seds ) can be expressed as the integrated difference in sediment inventories of the parent ²³⁸U and the daughter ²³⁴Th multiplied by the decay constant for ²³⁴Th:

Here Th and U activities are in units of dpm m^-3 and found by gamma counting dry sediments (dpm kg^-1) and multiplying by the dry bulk density. Dry bulk density is assumed to be 0.080 g cm^-3 (97% porosity, sediment grain density = 2.65 g cm^-3). Excess ²³⁴Th (Th_excess ) is the amount that is not in secular equilibrium with the ²³⁸U present in the sediments, presumably because it arrived via the supply of settling particles. Excess ²³⁴Th is typically observed in the top 3-5 cm, where recent sedimentation and bioturbation has occurred. The depth of equilibrium or integration (EQ) is where Th_excess is no longer measurable.

The residence time (τ) of a chemical constituent can be calculated by dividing the integrated inventory by a source or removal flux. Residence times provide insight into how the particle dynamics of a region impact the rate at which ²³⁴Th is scavenged from the dissolved pool and how quickly ²³⁴Th on smaller particles is incorporated into larger, sinking particle phases.

The residence time of scavenging or the transfer of dissolved ²³⁴Th (Th_diss ) into the particulate phase can be calculated with:

where τ_scav is expressed in days, Th_diss is the dissolved activity, and U represents the total ²³⁸U activity (Baskaran et al., 1996). Likewise, the residence time of particulate ²³⁴Th (Th_part ) with respect to removal via sinking particles can be determined as:

where the denominator corresponds to the flux as noted in Equation (4).

While the ²³⁴Th method for the determination of particulate carbon fluxes has been less frequently applied in coastal regions (Charette et al., 2001; Waples et al., 2006; Le Moigne et al., 2013; Waples and Klump, 2013), the same principles used in the open ocean apply here as well. The flux of ²³⁴Th at a given depth z in the water column is combined with measurements of the POC:²³⁴Th ratio on sinking particles at that depth to determine POC flux:

Particles collected on 51 µm filters (‘large particles’) are often considered to be sinking and those from 1 to 51 µm (‘small particles’) to be suspended (Bishop et al., 1977). However, slowly settling, small particles have been shown to contribute to carbon fluxes (e.g. Durkin et al., 2015; Puigcorbé et al., 2015) and therefore the POC:²³⁴Th ratios were assessed for both size fractions in this study.

Here the sediment accumulation fluxes of ²³⁴Th with small and large particle POC:²³⁴Th ratios are also used to assess sedimentary carbon accumulation fluxes. POC:²³⁴Th ratios were measured at ~60-65m, which represent particles that are about to be deposited on the seafloor. These ratios were combined with the sediment accumulation fluxes of ²³⁴Th in Equation (8) to calculate the accumulation flux of POC in the surface sediments. For comparison, POC:²³⁴Th ratios in the uppermost surface sediments were also measured, which represent deposited material that has likely experienced some degree of diagenetic processing.

A range in daily net primary production (NPP) for the Bedford Basin was also found using available in-situ measurements and global database products ( Supplementary Text 1.1 ). Benthic carbon fluxes (i.e., flux from the water column to the sediments) were then determined using the modeled NPP results. The model-based estimates are used here for broad comparison with the measurement-based, sediment trap and ²³⁴Th-derived carbon fluxes.

The base of the euphotic zone in Bedford Basin typically ranges from 15 to 25 m ( Figure 2C ). March 2019 marked the onset of increased Chlorophyll-a (Chl-a) concentrations in the euphotic zone ( Figure 2C ). Beginning in April-May, the water column slowly stratified and around July a persistent and significant temperature and density difference was observed between the upper ~20 m and the deeper water column. Due to persistent overcast skies in early 2019, the spring ‘bloom’ in the basin did not reach the peak levels of the two previous years ( Figure 2D ). Nevertheless, integrated Chl-a concentrations in the euphotic zone were noticeably higher in April compared to those observed in the months before or after April (e.g., February and June, Figure 2D ).

Salinities (S) ranged from 29.7 to 31.3 and thus, salinity-derived ²³⁸U activities ranged from 2.02 dpm L^-1 to 2.14 dpm L^-1 ( Figure 3A ). ²³⁸U activities were derived using the Owens et al. (2011) relationship of ²³⁸U = 0.0786 × S – 0.315 and its associated uncertainties. At all sampling dates, total ²³⁴Th activities are lower than ²³⁸U activities at all depths, revealing a Th deficit in the entire water column. The lowest ²³⁴Th activities, however, were observed in the upper 5 to 10 meters in all months and especially in August. Total ²³⁴Th activities for all dates averaged (± s.d.) 0.9 ± 0.2 dpm L^-1 and ratios of ²³⁴Th:²³⁸U ranged from 0.18 to 0.61 ( Figure 3A ).

The particulate ²³⁴Th results were also relatively consistent between depths and sampling dates ( Figure 3B ). Large particle ²³⁴Th activities averaged 0.008 ± 0.006 dpm L^-1 with a range of 0.003 to 0.02 dpm L^-1. The small particle ²³⁴Th average was significantly higher and averaged 0.6 ± 0.2 dpm L^-1 with a range of 0.29 to 1.0 dpm L^-1. The large particle fraction was only 0.9% ± 0.6% of the total ²³⁴Th present and the small particle fraction was 63% ± 22% of the total. Total particulate RA_P234 activities, the sum of the small and large particle RA_P234, were generally below the method detection limit (~92 dpm m^-3, Figure 4 , left top). The detection limit for Bedford Basin is equivalent to two times the s.d. of the dipped blank mean. The highest RA_P234 activity by far was measured at 40 m in August (170 dpm m^-3) and all other RA_P234 activities were below 120 dpm m^-3. The small particle data dominated the RA_P234 signal and the graph for small particle RA_P234 is visually indistinguishable from the total particulate graph in Figure 4 (left top).

The large and small particle POC concentrations averaged 3.0 ± 0.9 µM and 14 ± 13 µM, respectively ( Figure 3C ). The relative standard deviation for the large particle ²³⁴Th activities (78%) was much greater than that of the corresponding POC concentrations (30%). The reverse was true for the small particles, with the ²³⁴Th activity relative standard deviation at 39% and the POC concentration relative standard deviation at 90%. The large particle POC:²³⁴Th ratios averaged 485 ± 262 µmol:dpm and ranged from 132 to 923 µmol:dpm ( Figure 3D ). The small particle POC:²³⁴Th ratios were significantly lower, averaging 32 ± 39 µmol:dpm and ranging from 7 to 122 µmol:dpm. Like the ²³⁴Th partitioning, most of the total particulate carbon, averaging 17 µM, was in the small particle fraction (~75% on average).

Excess ²³⁴Th was concentrated in the upper 2 cm of the cores ( Figure 5A ). The 0-2 cm layer had a much higher ²³⁴Th activity in April (~30 dpm g^-1) compared to the other dates (9 to 18 dpm g^-1). The equilibrium depth was consistent and a lack of excess ²³⁴Th was always observed by ~5 cm core depth, with supported ²³⁴Th activities between 5 cm and 7 cm averaging 2.6 ± 0.2 dpm g^-1 (s.d.). Sedimentary POC concentrations, averaging 4.8 ± 0.7 mmol C g^-1 ( Figure 5B ), and %POC, averaging 5.8% ± 0.9%, were remarkably consistent with core depth between 0-5 cm. Including all the sampled sediment layers (0-32 cm), the average %POC was statistically the same at 5.6% ± 0.7% for all dates. The %POC found in this study matched well with historical coring efforts in the basin that found 4.6% at the Compass Station and between 4.3% and 5.8% nearby (Buckley et al., 1995). The POC:²³⁴Th ratios ( Figure 5C ) of interest are those in the 0-2 cm layers, because these layers are the most recently deposited. The POC:²³⁴Th in these layers averaged 355 ± 187 µmol dpm^-1 and increased below 2 cm due to the drop in excess ²³⁴Th activities.

The cumulative flux of ²³⁴Th in the water column linearly increased with water depth on all sampling dates ( Figure 6 ) and always with a coefficient of determination (R²) > 0.99. If the data from all dates are combined, the resulting linear regression produces a rate of 31 dpm m^-2 d^-1 per meter (y-intercept = 135) or 34 dpm m^-2 d^-1 per meter assuming no ²³⁴Th flux at the sea surface. The ²³⁴Th fluxes at the deepest sampling depths for each date (60-65 m) are reported in Table 1 with linearly extrapolated values for the predicted flux at the sediment-water interface of ~70 m. The water column export of ²³⁴Th averaged 2000 dpm m^-2 d^-1 at 60 m to 65 m while the extrapolated flux of ²³⁴Th at 70 m is predicted to be 2300 dpm m^-2 d^-1 averaging all sampling dates. More variability was observed with the sediment accumulation fluxes of ²³⁴Th ( Figure 6 , Table 1 ), which averaged 2000 dpm m^-2 d^-1 but ranged from ~1100 dpm m^-2 d^-1 (February) to 3600 dpm m^-2 d^-1 (April). The ²³⁴Th fluxes from the August replicate cores were similar within the calculated uncertainties ( Table 1 ).

The residence time of dissolved ²³⁴Th (τ_scav ) at the Compass Station was 9 ± 8 days, while the residence time of particulate ²³⁴Th (τ_part ) was 21 ± 9 days. With most of the total ²³⁴Th present in the small particle (63%) and large particle (0.9%) phases, it is unsurprising that τ_scav <τ_part was always observed. The mid-water column had the longest observed τ_scav in August (τ_scav = 25 days). The surface layers in April and August had the shortest (τ_scav = 1 to 2 days). This August surface layer also had the lowest total ²³⁴Th observed ( Figure 3A ).

The POC flux at 70 m, determined using a large particle ratio for Equation (4), reached a minimum in August (710 mmol C m^-2 d^-1) and a maximum in May (2070 mmol C m^-2 d^-1, Table 1 ). The average large particle POC flux was 1200 mmol C m^-2 d^-1. The POC fluxes calculated using the small particle POC:²³⁴Th ratios were much lower, averaging 70 mmol C m^-2 d^-1. In contrast to the large particle-based fluxes, the small particle-based carbon export shifted from 20-25 mmol C m^-2 d^-1 in April and May to a high of 150 mmol C m^-2 d^-1 in August. To determine the flux of POC to the sediments (i.e., the sediment accumulation flux of carbon), the large and small POC:²³⁴Th ratios from the 60-65 m water column samples were multiplied by the sediment accumulation flux of ²³⁴Th determined from the sediment cores. The average POC sediment accumulation fluxes ( Table 1 ) were 1100 mmol C m^-2 d^-1 (large particle ratio) and 60 mmol C m^-2 d^-1 (small particle ratio). The average POC flux estimates determined from the water column and sediment measurements were therefore comparable for the respective POC:²³⁴Th ratios used, which is related to the similarity in observed ²³⁴Th fluxes ( Table 1 ).

To determine a ²³⁴Th budget for the water column at the Compass Station using Equation (4), to compare water column fluxes of ²³⁴Th to the sediment accumulation fluxes found using Equation (5), and to be able to evaluate the factors influencing the ²³⁴Th budget in Bedford Basin relative to those influencing other types of marine regimes, there are five main assumptions and sources of uncertainty to assess: steady state, ²³⁴Th measurements, ²³⁸U estimates, the significance of physical factors (i.e. advection), and the potential impact of sediment resuspension. A summary table of these assumptions and uncertainties is included Supplementary Table 1 . Each element is explored in this section with respect to its potential influence on export calculations, except for ²³⁴Th measurements. In this study, ²³⁴Th measurement accuracy and precision were carefully evaluated by using process blanks, yield monitors, replicates, and beta counting standards.

As with other coastal systems, particle dynamics highly influence the ²³⁴Th distributions in the Bedford Basin. The substantial particulate fraction of ²³⁴Th, relative to the total ²³⁴Th activity measured, is typical (e.g. Wei and Murray, 1992; Charette et al., 2001). The particulate ²³⁴Th in this study averages ~63% of the total activity for all depths and dates, and in a prior study of the basin it was ~74% for surface waters (Niven et al., 1995). Similarly, the scavenging (9 ± 8 days) and particulate (21 ± 9 days) residence times calculated here fit well within the ranges observed in other coastal locations (Wei and Murray, 1992; Forster et al., 2009).

The Compass Station has considerable particle removal rates, illustrated by the 2:1 activity ratio of ²³⁸U to ²³⁴Th from May to August 2019 ( Figure 3A ). The extent of ²³⁴Th removal by particles can be seen in Figure 7 , which combines August data from this study with ²³⁴Th:²³⁸U data from September 1997 transect off the Scotian Shelf (Charette et al., 2001). Open ocean locations typically have an active surface layer with a ²³⁴Th deficit and secular equilibrium throughout most of the remaining water column (far right, Figure 7 ), while the Compass Station does not reach equilibrium. The deficit is slightly larger in the euphotic zone ( Figures 2 , 3A ) compared to the rest of the water column, but this study suggests that a ²³⁴Th:²³⁸U ratio of ≤0.6 is present at all depths for much of the year. In a prior weekly sampling study of the basin (0-5 m surface only), ²³⁴Th:²³⁸U ranged from 0.33 to 0.48 and total ²³⁴Th from 0.67 dpm L^-1 (May 17) to 0.92 dpm L^-1 (March 24, Niven et al., 1995). Similarly, most surface ²³⁴Th:²³⁸U ratios measured from February to April in 1998 fell between 0.3 and 0.7 (0-5 m, Azetsu-Scott and Niven, 2005). The general agreement between this study’s results ( Figure 3A ) and those from earlier studies (Niven et al., 1995; Azetsu-Scott and Niven, 2005), indicates that the particle dynamics in the surface of the basin have likely not been significantly changed by any natural or anthropogenic forcings over the last 30 years.

A particle tracking and connectivity model from the basin indicates that there is little particle loss from Bedford Basin to the rest of Halifax Harbor during normal flow times (i.e. in the absence of severe storms or intrusions, Shan and Sheng, 2012). This suggests that water column particles created in the basin stay in the basin and that export and accumulation budgets, in theory, should balance over annual timescales. However, particles in the water column likely have a multi-stage history. The sensitivity analysis of the ²³⁴Th budget assumptions (Section 4.1) suggests that complex particle dynamics and biogeochemical changes (e.g., shifts in community structure related to seasonality and bloom timing) must play a role in maintaining the observed ²³⁴Th deficit at the Compass Station in the warmer months of 2019. Slowly settling particles in the deeper layers of the Bedford Basin might help explain some of the ²³⁴Th deficits in the deeper water column. There is some evidence for slow settling in the Bedford Basin (Azetsu-Scott and Johnson, 1994). In March (peak of spring bloom) there were higher particle loads in the 0-8m surface mixed layer, while the rest of the water column had a smaller and uniform particle load. Then in mid-April to July, the presence of a benthic nepheloid layer was identified at a depth range of 45-65m using backscatter data. In late July-August the nepheloid layer was much thinner. While the cause or creation of the nepheloid layer is unknown, Azetsu-Scott and Johnson (1994) hypothesize that bloom material can travel rapidly (days) from the surface layer to depths below 50 m in the basin and then reside there for weeks before completely settling and accumulation on the sea floor. Slow settling material would not chemically resemble the sediment layers below and would still be part of the ‘fresh particle’ water column. However, these slow settling could scavenge additional dissolved ²³⁴Th from the bottom 10-20m of the water column before they fully become part of the seafloor sediment layers. The potential presence of ‘biogenic’ nepheloid layers and the possibility for true resuspension events (i.e., reentry of processed sediments into the water column) provide ample questions for future studies at the Compass Station and in other coastal regimes.

The similarity in water column ²³⁴Th fluxes between the sampling dates is striking ( Figure 6 ). Despite seasonal changes in Chl-a ( Figure 2C ) and a greater partitioning of ²³⁴Th onto small particles in April and August, total ²³⁴Th does not vary substantially between the sampling periods ( Figure 3A ). The few studies of coastal regions where full profiles of total ²³⁴Th were collected also showed a linear increase in ²³⁴Th flux with depth. This study produced per meter fluxes of 31 dpm m^-2 d^{-1 234}Th ( Figure 6 ), while those from Dabob Bay ranged from 33-43 dpm m^-2 d^-1 (Washington, USA, Wei and Murray, 1992), those from the Beaufort Sea Shelf in the Arctic ranged from 23-29 dpm m^-2 d^-1 (Baskaran et al., 2003), and those from Mecklenburg Bay were 28 dpm m^-2 d^-1 (southwestern Baltic Sea, Forster et al., 2009). This uniformity in fluxes is intriguing. It was not seen throughout the open and deeper regions (water depths >300 m) of the Gulf of Maine and the Scotian Shelf ( Figure 7 ), nor was it observed on the Bering Sea Shelf (Baumann et al., 2013). These less-restricted coastal regions are likely impacted by bathymetric changes ( Figure 7 ), more intense currents, and other physical factors, such as advection. Indeed, Charette et al. (2001) employed a 3-D model of the Scotian Shelf to evaluate local ²³⁴Th fluxes.

The March-August sediment accumulation fluxes for ²³⁴Th vary more than the water column fluxes ( Figure 6 ). Although the two multicores collected simultaneously in August showed little difference ( Table 1 ), in general, sediment-based radioisotope estimates inherently have larger uncertainties (e.g. Black and Buesseler, 2014) due to potential variations in grain size distribution, sediment composition (i.e. mineral type), and local processes (e.g. bioturbation). ²¹⁰Pb profiles of all cores provide evidence of bioturbation in the top 10 cm of the sediment column ( Supplementary Figure 3 ). A small part of the differences in ²³⁴Th accumulation flux between sampling dates could be due to factors that influence spatial variability, however, the trends from February to August cannot likely be explained by these factors alone. The sediment accumulation flux of ²³⁴Th in February is distinctly lower than in April, which could be explained by the onset of the spring bloom in March and April following a period of winter quiescence in phytoplankton activity. Other studies suggest that seasonal swings in sediment accumulation fluxes are possible, if not typical. Sediment fluxes in the Long Island Sound, a larger, but restricted basin averaging ~20 m water depth, have been shown to range from as little as 227 dpm m^-2 d^-1 and to as much as 4953 dpm m^-2 d^-1 across the basin (Aller et al., 1980; Aller and Cochran, 2019). At a single station sampled between May and August of the same year, the sediment fluxes varied by >3000 dpm m^-2 d^-1 (Aller and Cochran, 2019).

Despite the greater variability observed in sediment data, the average sediment accumulation of ²³⁴Th of 2000 dpm m^-2 d^-1 (s.d. ± 1000) is roughly equal to the measured water column average and the extrapolated water column average (2300 dpm m^-2 d^-1 at 70 m, Table 1 ) within the associated uncertainties. Similarly, good agreement was found between predicted water column ²³⁴Th fluxes and results from sediment measurements on the Scotian Shelf for locations with depths of 100 m to 300 m (Charette et al., 2001). However, station to station variation along the Scotian Shelf showed water column fluxes ranging from 1500 to 10000 dpm m^-2 d^-1. Thus, the more open shelf locations had significantly more variability in water column fluxes than was observed in the more restricted Bedford Basin. In the sediments of Bedford Basin, the larger range in accumulation estimates over the short-term (²³⁴Th-based) versus the more stable long-term estimates (²¹⁰Pb, Supplementary Figure 3 ) indicates that seasonal studies of coastal regions would benefit from including multiple sets of replicates (spatial variability) and increased frequency of sampling efforts, including winter sampling (temporal variability). ²³⁴Th budgets may agree over annual timescales, as it appears they do in Bedford Basin in this study, but the stability of the water column ²³⁴Th activities juxtaposed with the fluctuations in the sediment ²³⁴Th inventories indicates that more can be learned about the seasonal controls on ²³⁴Th activities in the basin.

Since the averages of water column- and sediment-based ²³⁴Th fluxes overlap, the major source of variability in carbon flux estimates is related to the POC:²³⁴Th ratios ( Table 1 ). Indeed, assessing POC:²³⁴Th ratios is an ongoing field of inquiry (e.g. Buesseler et al., 2006; Puigcorbé et al., 2020). In open ocean studies, the large particle ratio is most often utilized as representing sinking material. However, the POC:²³⁴Th ratio on large particles observed in this study for May (>900 µmol:dpm) would translate into an unreasonably high carbon flux of ~2100 mmol C m^-2 d^-1 at the Compass Station ( Table 1 ). This POC flux estimate is more than an order of magnitude higher than models predict for coastal regimes globally (Siegel et al., 2014) and is unreasonable considering that all prior coastal export studies utilizing similar methods found fluxes less than 200 mmol C m^-2 d^-1. POC:²³⁴Th ratios ranging from 2 to 534 µmol:dpm have been observed previously on the Scotian Shelf and in the Gulf of Maine and, therefore some of the ratios >100 µmol:dpm in Bedford Basin could be attributed to natural, seasonal variability (Benitez-Nelson et al., 2000; Charette et al., 2001). Nevertheless, the >100 µmol:dpm ratios found in other studies were generally rare occurrences (Buesseler et al., 2006; Puigcorbé et al., 2020), whereas the high POC:²³⁴Th ratios measured on large particles were frequent in this study (for further discussion see Supplementary Text 1.3 ).

The remaining assessment of the basin’s carbon budget in the next sections focuses on the small particle ratio-based estimates only. The small particle ratios are emphasized for five reasons. First, the uniquely low large particle ²³⁴Th activities require corroboration with a multi-method sampling effort before export values well above 100 mmol C m^-2 d^-1 should be considered. Second, most other coastal ²³⁴Th-based studies employed total particulate ratios to calculate export, which are governed by small particles. Third, recent work has suggested that for some open ocean locations (e.g. North Atlantic at BATS, Durkin et al., 2015), coast-proximal regions (e.g. Puigcorbé et al., 2015) and shallow seas (Hung and Gong, 2010), small particles are important to the overall local flux of carbon. Fourth, large particle (>51 um) POC:²³⁴Th ratios from the nearby Gulf of Maine (March – Sept 1995, Charette et al., 2001) averaged (± s.d.) 49 ± 68 µmol:dpm, which is similar to the average found in Bedford Basin for small particles (32 ± 39 µmol:dpm). Fifth, using a simple Stokes ideal sphere approximation (particle density = 2 g cm^-3), a ‘suspended’ 10 µm particle can pass through the entire 70 m water column at the Compass Station in ~4 days. More complex particle geometries and lower densities could potentially cause the suspended matter to move downward more slowly, however, with a 24-day half-life for ²³⁴Th, the suspended particles likely play a key role in the total export of carbon to the sediments.

Model-based NPP estimates from Beford Basin show a high degree of seasonal variability, ranging from 0.4 to 184 mmol C m^-2 d^-1 ( Table 2 ). The yearly average NPP from these estimates, which utilize 2019 in-situ Chl-a data, is 66 mmol C m^-2 d^-1 ( Table 2 ). This modeled yearly average matches well with the mean from prior incubation studies in the basin (50 mmol C m^-2 d^-1, Platt, 1975). The modeled NPP estimates also overlap with those reported for the Scotian Shelf and Gulf of Maine at 26-399 mmol C m^-2 d^-1 (August-September 1997, Charette et al., 2001). Broadly speaking, NPP in Bedford Basin is in the 100s of mmol C m^-2 d^-1 at peak times, often in spring, and in the 10s of mmol C m^-2 d^-1 otherwise. As with the export estimates, those for NPP illustrate the importance of multiple sample efforts in these types of dynamic coastal systems. NPP is the main source of carbon driving export and sedimentary accumulation of carbon in Bedford Basin and production will incorporate new fixed carbon from atmospheric CO₂ and recycled carbon as discussed below. Other potential sources of dissolved carbon and POC to the basin, like the discharge from the Sackville River, have been previously constrained (Hargrave and Taguchi, 1978). Daily riverine inputs of POC of ~3 mmol C m^-2 d^-1 are small with respect to NPP and export estimates and it can be assumed that most of the POC created within the basin remains in the basin due to the shallow, restrictive sill at its south end ( Figure 1 ; Shan and Sheng, 2012).

POC export fluxes for the entire water column, derived here from the small particle POC:²³⁴Th ratio and ²³⁴Th fluxes (hereafter referred to as POC export), ranged from ~25 mmol C m^-2 d^-1 in April and May 2019 to a high of 150 mmol C m^-2 d^-1 in August 2019 ( Table 1 ). The mean export estimate for April to August of 70 mmol C m^-2 d^-1 (s.d. = 70, Table 1 ) compares favorably with historical observations from the basin and estimates from nearby coastal shelf locations. Prior estimates of export in Bedford Basin were measured using bottom tethered sediment traps and no preservatives, which can result in lower flux estimates due to particle remineralization and/or feeding by swimmers (Hargrave and Taguchi, 1976). Despite the differing methods, the average export reported in 1976 (17 mmol C m^-2 d^-1 at 60 m) is only slightly lower than the spring export estimates from this study (~25 mmol C m^-2 d^-1, Table 1 ). Full water column POC export from this 2019 study at ~70 m depth also fall in the range for those observed on the Scotian Shelf and in the Gulf of Maine at 50 m water depth (average = 34 ± 44 mmol C m^-2 d^-1, Charette et al., 2001) ( Table 2 ).

As with the ²³⁴Th-based export estimates (and the model-based NPP results), large seasonal variations in export were observed in both Hargrave and Taguchi (1976) and Charette et al. (2001). Hargrave and Taguchi (1976) found six-fold seasonal swings from a maximum of 36 mmol C m^-2 d^-1 (April) to 2 mmol C m^-2 d^-1 (July). The Gulf of Maine had a peak spring bloom of ~120 mmol C m^-2 d^-1 and dramatically smaller summer-fall fluxes at<10 mmol C m^-2 d^-1. The 2019 Bedford Basin export fluxes showed the opposite trend in 2019, with higher fluxes in August ( Tables 1 , 2 ). One explanation for the differing seasonal trends is that 2019 was an uncharacteristic year in Bedford Basin, having a more prolonged spring bloom without a definitive peak ( Figure 2D ). Another reason for the differences could be the use of the large particle ratios in the Charette et al. (2001) study to determine export, which might reflect the prevalence of different communities and particle-formers at different times of year. While the export estimates from all three studies are overall comparable, the collective variability shows the importance of repeated sampling in coastal regions when attempting to decrease uncertainties on coastal carbon budget estimates. Furthermore, the broad similarity in average carbon export estimates from each study is curious considering that each method likely captures different particle types: Bedford Basin traps should collect truly sinking particles over the period of sampling (with noted the remineralization and swimmer caveats), the Gulf of Maine study used >51 µm large particles to find carbon flux, and this 2019 Bedford Basin study utilized 1-51 µm small particles to find carbon flux. Additional carbon flux studies utilizing size fractionated sampling methods could provide valuable insight into the potential biological and chemical similarities between large and small particles in coastal zones and the relative importance of small particles in export estimates.

Comparing the daily model-based and ²³⁴Th-derived estimates, the benthic carbon flux estimates from modeled NPP (0.8 to 61 mmol C m^-2 d^-1) appear to be lower, on average, than the ²³⁴Th-based carbon export estimates from Bedford Basin (20 to 150 mmol C m^-2 d^-1), and from the Gulf of Maine (<10 to ~120 mmol C m^-2, Table 2 ). One reason for the difference in the three sets of results are that the model-based estimates incorporate Chl-a data from the entire year, including winter months, and so the lowest export times have yet to be quantified via ²³⁴Th measurements. The model-based estimates suggest that export is typically in the 10s of mmol C m^-2 d^-1,throughout the spring, summer, and fall, with higher export during the spring bloom. ²³⁴Th-based export estimates from Bedford Basin in August ( Table 1 ) were much higher than modeled benthic carbon flux at that time ( Table 2 ), but as noted earlier, export events of the same magnitude (>100 mmol C m^-2 d^-1) were also observed in Charette et al. (2001).

To calculate and compare annual estimates, it is assumed that the ²³⁴Th-derived Compass Station estimates are representative of the carbon flux during the most productive months in the basin (i.e., 9 months from March to November) and that relatively little flux occurs in the remaining winter months. This is supported by the model-based NPP and export trends. The average ²³⁴Th-based export of 70 mmol C m^-2 d^-1 for a 9-month period translates to an annual carbon flux estimate of ~19 mol C m^-2 yr^-1. The modeled daily benthic flux of 19 mmol C m^-2 d^-1 equates to an annual export of ~7 mol C m^-2 yr^-1 ( Table 2 ), which agrees within a factor of two to three with the ²³⁴Th -based estimates. If only the April and May measurement-based export estimates are extrapolated over the 9-months (at 20-25 mmol C m^-2 d^-1), the ²³⁴Th-based annual estimate is the same as the modeled estimate at ~7 mol C m^-2 yr^-1. Considering the overall variability in the three measurement-based carbon fluxes and model assumptions, the general agreement between the two methods observed here is very encouraging.

Like the water column and sedimentary ²³⁴Th budgets ( Figure 6 , Tables 1 , 2 ), the carbon export flux and the sediment accumulation flux of carbon also appear to be quasi-balanced if multi-timescale methods are utilized to create a sediment burial budget for the Compass Station. The sedimentary carbon content from this study (average weight % = 4.5%, upper 5 cm) and estimates of the sedimentation rate (0.2 cm yr^-1 from this study via ²¹⁰Pb, Supplementary Figure 3 ; Buckley et al., 1995) can be used to determine an apparent burial rate of organic carbon of 6 mmol C m^-2 d^-1 (dry bulk density ~275 kg m^-3, Table 2 ). The average ²³⁴Th-based sediment accumulation flux is 60 mmol C m^-2 d^-1 ( Table 2 ), suggesting that ~10% of exported carbon is buried. Consistent with these estimates, the amount of dissolved inorganic carbon lost from the sediments as a result of carbon respiration (36 to 56 mmol C m^-2 d^-1, Burt et al., 2013), implies that up to ~90% of the carbon export reaching the seafloor is re-released into the water column as dissolved inorganic carbon prior to burial. Caution should be used when comparing results derived from varying methodologies, however, as shown here, the broad patterns in the Bedford Basin emerge when multiple metrics for the same process are used (e.g., sediment rate) and repeated sampling is performed. The paired organic carbon export from the water column and its burial in sediments (this study) and the prior measures of inorganic carbon loss from the sediments via respiration (Burt et al., 2013) combine to close the internal, annual carbon cycling budget of Bedford Basin quite well ( Tables 1 , 2 ).