We use the Parallel Ocean Program version 2 (POP2), the ocean component of CESM2, to simulate the global transport and transformation of Hg under present-day conditions. Tracer advection is computed using a third-order upwind-biased scheme, with an optional monotonic flux limiter to minimize numerical dispersion and ensure non-negative concentrations. Vertical mixing is represented using the K-profile parameterization (KPP) scheme, which computes vertical diffusivity and viscosity in the surface boundary layer based on local shear, stratification, and wind forcing. The model is configured on a tripolar curvilinear B-grid, with the North Pole displaced over Greenland to avoid coordinate singularities. It employs 60 vertical z-levels, with enhanced resolution in the upper ocean (∼ 10 m layer thickness) to resolve the mixed layer and thermocline structure. Layer thickness increases with depth to approximately 250 m in the abyss, allowing realistic simulation of both upper-ocean processes and deep-ocean tracer transport (Smith et al., 2010). POP2 has previously been used to simulate inert ocean tracers such as CFC-11, CFC-12, and SF₆, which have been widely applied in evaluating large-scale ventilation and circulation patterns (Dutay et al., 2002). Building on this established framework, we extend POP2 to simulate reactive mercury species by incorporating air–sea exchange, redox reactions, and internal oceanic transformations associated with particulate interactions.

For the development of the mercury module, we introduce five passive tracers, Hg⁰, Hg^II, two forms of MeHg, monomethylmercury (MMHg) and dimethylmercury (DMHg), and particulate-bound mercury (Hg^P) (Fig. 1). We also consider the various processes of its biogeochemical cycle (see Sect. 2.3). For initial development and spin-up, we use the G “Compset” with the T62_g37 grid configuration, which couples an active ocean and sea ice model with prescribed atmosphere and land components. This setup provides a nominal horizontal resolution of approximately 3°. Marine biogeochemistry is enabled through the default MARBL configuration included in CESM2.1.3. All build settings and source files for this configuration are archived in the Zenodo repository associated with this paper.

Marine biogeochemical processes in POP2/Hg v1.0 are simulated using the Marine Biogeochemistry Library (MARBL), a modular component of CESM2 that resolves coupled cycles of carbon, nutrients, plankton ecology, and oxygen dynamics. In the configuration used here, the ecosystem is represented using three phytoplankton functional types, which are small phytoplankton, diatoms, and diazotrophs, and one zooplankton grazer. Each phytoplankton type exhibits distinct physiological traits and nutrient requirements. Small phytoplankton dominate in oligotrophic gyres, diatoms require silicate and are prevalent in upwelling and high-latitude regions, while diazotrophs fix atmospheric nitrogen and are restricted to warm, low-nitrate waters. Zooplankton graze on all three phytoplankton types and recycle nutrients through excretion and respiration. MARBL explicitly simulates cycles of macronutrients (nitrate, phosphate, silicate), iron, oxygen, dissolved inorganic carbon (DIC), alkalinity, and both dissolved and particulate organic matter (DOC, POC). Biogeochemical rates such as photosynthesis, nutrient uptake, remineralization, and particle sinking are computed prognostically at every time step, accounting for ambient light, temperature, nutrient concentrations, and ecological stoichiometry. Remineralization of organic matter releases nutrients and dissolved constituents (e.g., Hg^II) back into the water column, strongly modulating vertical tracer profiles (Moore et al., 2001, 2004; Long et al., 2021).

As shown in Fig. 2, at 5 m depth, the carbon pool of small phytoplankton and zooplankton are widely distributed in low-latitude regions, particularly in oligotrophic gyres and equatorial zones. Diatoms are concentrated in high-latitude areas and eastern boundary upwelling systems (e.g., off Peru and California), reflecting their preference for nutrient-rich conditions. Diazotrophs are abundant in tropical and subtropical regions (Long et al., 2021). At 95 m depth, the concentrations of all biological components decline significantly, consistent with light and nutrient limitation below the euphotic zone. However, POC maintains higher concentrations at 95 m in coastal upwelling zones, indicating active export of organic material and a strong biological pump (Long et al., 2021).

Organic carbon is partitioned into a “soft” POC pool – composed of rapidly sinking detritus that remineralizes over a few hundred meters – and a “hard” or ballasted POC pool with much longer remineralization length scales (∼ 40 000 m), allowing for deep ocean export. These two POC classes are critical for linking ecological production to elemental cycling and are used in our model to drive particle-associated mercury transport (see Sect. 2.3.4). The total POC flux is highest in regions of elevated primary productivity and intense upwelling, such as the eastern equatorial Pacific, northern Indian Ocean, and Arabian Sea, where values exceed 0.1 mol C m⁻² yr⁻¹.

Below 2000 m, the behaviors of soft and hard POC diverge markedly. Soft POC flux becomes nearly negligible across the global ocean at 2889 m (Fig. 4a), due to rapid remineralization in the upper mesopelagic. In contrast, hard POC retains fluxes even at 2889 m (Fig. 4b) and in the deepest layer (Fig. 4d). At the sea floor, hard POC becomes the main source of total POC flux (up to 99.96 % maximum and an average of 79.21 % in fraction).

The vertical distribution of the organic carbon remineralization rate shows distinct patterns (Fig. 5). At the surface (5 m), Fremin is generally low, as organic matter production exceeds immediate degradation. At 95 m, remineralization peaks in the North Pacific, North Atlantic, Arabian Sea, and eastern equatorial Pacific, coinciding with high POC export. These zones represent areas of high productivity and efficient particle flux attenuation. In contrast, Fremin values in the deeper zone are nearly zero, indicating that most organic matter is remineralized in the upper 200–300 m.

For the simulations in this study, the five ocean Hg tracers were initialized using externally prescribed three-dimensional fields prepared for the MITgcm-Hg configuration described by Zhang et al. (2020). These fields were interpolated onto the POP2 grid and used to initialize the corresponding POP2/Hg tracers. Starting from the prescribed initial fields, POP2/Hg was integrated for 10 years. To assess the influence of integration length, an otherwise identical baseline simulation was extended to 20 years. The vertically integrated Hg^T inventories within the upper 500 m differed by less than 1 % between the 10- and 20-year simulations, indicating that extending the integration had only a minor effect on the upper-ocean Hg^T results examined in this study. As the present study focuses primarily on the upper ocean, the 10-year simulation was used for model evaluation and process analysis. Longer integrations would be required to investigate changes in the deeper ocean in response to ocean circulation and biogeochemical processes.

The model uses atmospheric Hg⁰ concentration and Hg^II deposition prescribed from the CAM6-Chem/Hg model output (Zhang and Zhang, 2022) as atmospheric forcing input (Fig. 6). Hg⁰ forcing is strongest in the western Pacific, broadly reflecting the influence of nearby anthropogenic source regions. Hg^II deposition dominates in the intertropical convergence zone (ITCZ), North Indian Ocean, and eastern equatorial Pacific, consistent with wet deposition patterns driven by deep convection. The current model does not yet include riverine mercury input and focuses on the development and evaluation of the ocean module. The coupling with riverine inputs and the land component will be incorporated in future work. For DMHg, the atmospheric concentration is held spatially and temporally constant at 4 pg m⁻³, following Zhang et al. (2020).

The model evaluation utilized a compiled dataset of ocean Hg concentrations provided by Zhang et al. (2020), and air–sea exchange flux collected by Kuss et al. (2011), Soerensen et al. (2013, 2014), Wang et al. (2017), Kalinchuk et al. (2021), and Nerentorp Mastromonaco et al. (2017). These data were used to evaluate model outputs.

The modelled dissolved elemental mercury (Hg⁰) concentrations in surface waters typically range from 0.04 to 0.08 pM, accounting for approximately 10 % of Hg^T in the upper ocean (Fig. 7a). Localized maxima reaches approximately 0.12 pM, with high concentrations (> 0.10 pM) occurring in the western Pacific warm pool and eastern equatorial Pacific. In contrast, localized low Hg⁰ levels (< 30 fM) occur in the North Indian Ocean and Arabian Sea. These patterns are driven by sea surface temperature, solar irradiance, and wind speed: warmer, low-latitude regions with high light availability and low wind speeds favor photoreduction of Hg^II and reduced evasion, allowing accumulation of Hg⁰ in surface waters.

Figure 7a presents the spatial distribution of modelled surface Hg⁰ along with pointwise observations for comparison. The model qualitatively captures the broad enhancement of Hg⁰ concentrations in tropical and subtropical regions well, and reproduces the observed order of magnitude at many open-ocean locations, with simulated concentrations generally ranging from 0.04 to 0.08 pM. However, significant discrepancies occur in coastal and estuarine zones such as Southeast Asia and the eastern seaboard of North America, where some of the observed values exceed 0.2 pM, while simulated Hg⁰ concentrations generally remain around 0.07 pM. The collocated comparison shows a significant positive spatial association (r=0.35, p<0.001, n=381). Correlations are stronger in the northern and southern subtropics (r=0.39 and 0.53, respectively; both p<0.001), with NMBs of -47.1 % and +62.5 %, respectively. These differences likely result from the lack of riverine Hg input and limited resolution nearshore, which are not yet incorporated in the model. In polar regions, additional discrepancies may arise from the omission of Hg uptake by and release from sea ice during ice formation and melt.

At 95 m depth, the modelled Hg⁰ concentrations typically range from 0.02 to 0.08 pM, and in some regions exceed surface values, locally reaching up to 0.18 pM in upwelling zones. This increase is particularly evident in upwelling and remineralization zones such as the eastern equatorial Pacific, North Indian Ocean, and the subtropical boundary currents. At this depth, light attenuation limits photoreduction, while remineralization-driven Hg^II reduction increases the fraction of Hg⁰, combined with weaker air–sea exchange rates that allow greater retention of Hg⁰. At 985 m depth, the simulated Hg⁰ distribution remains strongly influenced by the prescribed initial field, with a spatial correlation of 0.93 between the simulated and initial fields.

The simulated distribution of Hg^II reflects the combined influence of atmospheric deposition, redox transformations, and internal cycling. At the sea surface (Fig. 8a), Hg^II concentrations range from approximately 0.1 to 0.8 pM, with elevated values in the North Indian Ocean, tropical western Pacific, and parts of the South China Sea. These regions coincide with intense wet and dry deposition of atmospheric Hg^II and zones of active photochemical cycling. Extreme values occur in biologically productive or upwelling regions, such as the Bay of Bengal and Southeast Asian marginal seas, where concentrations locally reach up to 2.0 pM.

At 95 m depth (Fig. 8b), Hg^II concentrations are markedly enhanced in high-productivity or upwelling regions, including the eastern equatorial Pacific and the Arabian Sea, typically ranging from 0.3 to 0.8 pM. This enhancement is primarily due to the downward export and remineralization of Hg^P, combined with reduced photoreduction efficiency that limits the conversion to Hg⁰. In addition, oxygen-deficient zones at this depth promote the persistence of divalent mercury, making them key sources of microbial methylation. At 985 m depth (Fig. 8c), Hg^II becomes the largest dissolved mercury species. High concentrations are observed in the North Pacific and equatorial Atlantic, approaching 1.5 pM. Similar to Hg⁰, this deep distribution remains partly influenced by the initial field after the 10-year integration.

The simulated Hg^P reflects the combined influences of particulate POC and Hg^II in both horizontal and vertical dimensions. At the ocean surface (Fig. 9a), Hg^P concentrations exhibit distinct regional maxima in the northwestern Indian Ocean, equatorial Atlantic, and northeastern Pacific, particularly in the Bay of Bengal, where values exceed 0.3 pM. Elevated Hg^P concentrations are co-located with regions where both POC levels and Hg^II concentrations are relatively high (see Figs. 3a and 8a), promoting enhanced sorption of Hg^II onto organic particles.

At 95 m depth (Fig. 9b), Hg^P concentrations get lower, with the highest values (exceeding 0.07 pM) persisting in subsurface convergence and export regions such as the western Pacific and eastern equatorial Atlantic. The influence of POC dominates over that of Hg^II, as high Hg^P concentrations closely correspond to regions of elevated POC. At 985 m depth (Fig. 9c), where biological productivity declines sharply, Hg^P concentrations nearly vanish. At this depth, the dominant process becomes remineralization, and nearly all Hg^P is converted back to the dissolved form, with notable residual Hg^P concentrations only persisting in localized regions where POC remains. This vertical attenuation of Hg^P aligns with previous modeling studies (e.g., Sunderland and Mason, 2007; Zhang et al., 2014b).

The simulated distribution of MeHg, including MMHg and DMHg, exhibits pronounced vertical and regional gradients shaped by subsurface methylation, biological productivity, and redox conditions. Across the three evaluated depths, the collocated MeHg comparison gives r=0.47 (p<0.001, n=259), RMSE =0.156 pM, and NMB =-11.7 %, indicating that the model captures part of the overall spatial variability with a moderate negative bias.

At the surface (Fig. 10a), MeHg concentrations remain generally low across most of the global ocean, typically below 0.05 pM. However, due to strong photo-demethylation and rapid air–sea exchange of volatile DMHg, surface MeHg remains significantly depleted relative to deeper layers. This indicates that surface MeHg is tightly regulated by photochemical loss and air–sea fluxes (Soerensen et al., 2010; Zhang et al., 2020). These open ocean surface patterns correspond with available observations generally (r=0.49, p<0.001, n=86).

At 95 m depth (Fig. 10b), peak MeHg concentrations increase and occur in regions of high productivity or upwelling, such as the eastern tropical Pacific, equatorial Atlantic, and the northern Indian Ocean, with simulated values exceeding 0.12 pM. The model captures part of the observed spatial variability (r=0.37, p<0.001, n=78), matching observed hotspots along the eastern boundary currents (e.g., Peru and California coasts) and lower in open waters, although regional biases remain. In the North Pacific and tropical Atlantic, the available observations and modelled values are generally of the same ranges, but the regional sample sizes are small.

At 985 m (Fig. 10c), the model simulates widespread accumulation of MeHg, with background values around 0.1–0.3 pM, and local maxima exceeding 0.6–0.8 pM in the North Pacific interior. These elevated values are consistent with observations from deep sampling campaigns, especially in the North Pacific Subarctic Gyre and Equatorial Pacific, where deep MeHg concentrations often exceed 0.5 pM. The persistence of MeHg at these depths is likely to be sustained by slow degradation rates and limited transport, allowing accumulation over longer timescales. This pattern is consistent with previous global ocean mercury models (e.g., Wang et al., 2023). When compared with observation, association remains positive (r=0.42, p<0.001, n=95) although the model generally overestimates MeHg. The distribution also remains influenced by the prescribed initial field.

At the surface (Fig. 11a), simulated Hg^T concentrations generally range from 0.1 to 1.0 pM, with higher values (> 1.2 pM) found in the equatorial Pacific, western boundary currents (e.g., Gulf Stream, Kuroshio), and eastern boundary upwelling zones (e.g., off Peru and southwest Africa). The patterns capture part of the broad surface spatial variability, but local agreement remains modest (r=0.29, p<0.001, NMB =-64.2 %, n=140), indicating a substantial overall underestimate. For example, in the North Atlantic Ocean (excluding coastal regions), observed Hg^T concentrations at the surface range from 0.6 ± 0.2 pM, while the model simulates values around 0.7 ± 0.1 pM. In contrast, the model underestimates Hg^T in some coastal and marginal seas, including Southeast Asia, the Gulf of Mexico, and the western North Atlantic shelf. These mismatches likely arise from the exclusion of riverine Hg inputs and insufficient resolution of shelf and margin dynamics in the current model configuration.

At 95 m depth (Fig. 11b), elevated concentrations persist in eastern boundary regions and extend westward into the tropical Atlantic and Pacific basins. Simulated Hg^T values typically range from 0.4 to 1.4 pM at this depth, showing good agreement with observations (r=0.49, p=0.0013, n=41).

At 985 m, (Fig. 11c), Hg^T shows widespread accumulation. In parts of the Atlantic Ocean and the Southern Ocean, the model tends to overestimate deep Hg^T relative to observations. Because the deep ocean does not fully adjust over the integration period, this mismatch may partly reflect the prescribed initial field.

Across the evaluated Hg species, the collocated comparisons generally show positive model–observation associations and reproduce the observed ranges, although the magnitude and direction of the biases vary among species, regions, and depths. The largest discrepancies occur in coastal regions. The nominal 3° configuration is selected to enable computationally efficient model development, iterative testing, and sensitivity experiments. However, its relatively large grid cells cannot adequately resolve narrow continental shelves, estuarine exchange, coastal currents, or mesoscale eddies. Riverine Hg inputs are also not represented in the present model configuration, partly because their transport and dilution cannot be resolved adequately at this spatial resolution. The resulting underestimation in some coastal and marginal seas is therefore consistent with the current model structure and forcing.

In the open ocean, most Hg observations were collected along ship tracks during individual cruises and represent specific seasons, hydrographic conditions, and short-term meteorological states. In contrast, the model fields represent annual-mean conditions at the global scale. Differences at individual sampling locations can consequently arise from spatial and temporal sampling mismatches. In addition, the model is driven by the atmospheric Hg deposition flux predicted by the CAM6-Chem/Hg model, which simulates an elevated flux in the tropical region due to the inclusion of OH/O3 oxidation of Hg⁰ and the absence of sea salt debromination processes (Zhang and Zhang, 2022). This potential overestimation of atmospheric Hg deposition may bias the simulated oceanic Hg^T concentrations, particularly at lower latitudes. These factors partly explain the greater scatter in comparisons at individual sampling locations. Nevertheless, the simulated fields reproduce the major basin-scale Hg gradients in the available observations. At depth, the continuing influence of the prescribed initial conditions also contributes to the model–observation differences.

The ocean surface acts as the primary interface for the exchange of Hg between the atmosphere and the ocean. Model simulations estimate a global Hg⁰ net evasion flux of approximately 4.52 × 10³ Mg yr⁻¹. Including DMHg, the simulated total net gaseous Hg evasion is 4.61 × 10³ Mg yr⁻¹, comparable to the atmospheric Hg^II deposition flux of 4.20 × 10³ Mg yr⁻¹, indicating a near balance in global air–sea Hg exchange.

In Fig. 12, net air–sea exchange flux is defined as Hg species come into the sea minus emissions, positive values indicate net input, and negative values indicate net emission. Spatially, Hg⁰ evasion is strongest in tropical and subtropical regions, where warm surface waters and high solar radiation enhance photoreduction and lead to Hg⁰ supersaturation. Elevated fluxes are also observed in the South Pacific and Indian Ocean. Along the western boundary of the North Atlantic, enhanced evasion is driven by the Gulf Stream's elevated temperatures and wind speeds. The Southern Ocean shows strong evasion as well, attributed to persistent westerlies and high wind forcing.

Table 2 compares selected ocean Hg inventories and net air–sea exchange in POP2/Hg with corresponding estimates for MITgcm-Hg and ICON-MERCY obtained from MCHgMAP model outputs. The MCHgMAP model-comparison framework is described by Dastoor et al. (2025). Net air–sea exchange is defined as positive into the ocean, and negative values therefore indicate net Hg evasion. For POP2/Hg, annual-mean concentrations are integrated over native wet-cell volumes, using the model interfaces at 50.0 and 1041.4 m as the boundaries of the nominal surface, intermediate, and deep-ocean reservoirs. All three models produce comparable surface-ocean inventories and simulate net oceanic Hg evasion. The intermediate-, deep-, and total-ocean POP2/Hg inventories are closer to the MITgcm-Hg estimates, whereas ICON-MERCY produces substantially larger intermediate and deep-ocean inventories. The agreement with MITgcm-Hg is consistent with the use of its Hg initial fields and chemical parameterizations in POP2/Hg and provides a useful consistency check that their implementation within the POP2 circulation and MARBL biogeochemistry preserves the first-order vertical inventory structure. The remaining differences reflect variations in vertical grids, forcing, circulation, ecosystem representation, and diagnostic definitions, and the comparison therefore provides a broad model-to-model benchmark rather than a controlled budget intercomparison.

Inorganic mercury transformations in the ocean, including redox and methylation reactions, vary with depth, region, and season (Figs. 13–15). Integrated over the surface ocean (0–50 m), gross reduction slightly exceeds gross oxidation in the surface layer, giving a weak net tendency toward Hg⁰ formation that is consistent with the simulated Hg⁰ evasion. Photoreduction of Hg^II is particularly strong at low-latitude regions, with maxima in the equatorial Pacific, North Atlantic gyres, and the Indian Ocean (> 40 µg m⁻³ yr⁻¹), consistent with intense solar irradiance and shallow mixed layers.

At intermediate depths (50–1000 m), the vertically integrated redox budget remains net reducing, with an annual total flux of 1.74 × 10³ Mg yr⁻¹. Both oxidation and reduction rates at 95 m (Fig. 13c–d) are roughly one order of magnitude lower than surface values, but oxidation becomes dominant in certain mid-depth waters, notably in the North Atlantic and eastern equatorial Pacific upwelling zones. At 985 m (Fig. 13e–f), redox fluxes further decrease by roughly three orders of magnitude, and a net oxidation trend emerges. Integrated over the deep ocean below 1000 m, the redox budget shows net oxidation of 3.74 × 10³ Mg yr⁻¹, suggesting slow but sustained chemical transformation of Hg⁰ to Hg^II in deep waters.

The spatial distributions of methylation and demethylation rates are closely linked to the modelled Hg^II concentrations and organic matter remineralization patterns (Figs. 3 and 5). Regions of enhanced methylation tend to coincide with strong remineralization zones, particularly at 95 m depth in the eastern tropical Pacific, the northwest Pacific, and the Arabian Sea. Besides, the regeneration of Hg^II from sinking organic matter plays a key role in promoting MeHg production. While in surface waters (5 m depth), demethylation dominates, reflecting efficient photochemical and microbial degradation processes under high light conditions.

The corresponding layer-integrated net budgets indicate net demethylation of 1.17 × 10³ Mg yr⁻¹ in surface waters and 2.96 × 10³ Mg yr⁻¹ in the deep ocean. In intermediate waters, annual methylation fluxes are 84.99 Mg yr⁻¹, and methylation hotspots align with remineralization zones, especially in the eastern tropical Pacific, the northwest Pacific, and the Arabian Sea. These processes are supported by regenerated Hg^II from sinking organic matter. However, the relatively coarse horizontal resolution of the model may lead to an underestimation of localized methylation or demethylation hotspots in these regions.

Seasonal variation is pronounced in surface redox fluxes, with higher rates near the beginning and end of the year and a minimum around June (Fig. 15). Methylation and demethylation show less seasonality and remain active throughout the year, especially at depths > 1000 m, where conditions are more stable.

As shown in Fig. 16, the downward flux of Hg^P associated with soft and hard POC exhibits distinct spatial and vertical patterns, coinciding with the contribution of soft and hard POC fluxes (Figs. 3 and 4). Across the 50 m interface, the total downward Hg^P flux is 5.55 × 10² Mg yr⁻¹, of which soft- and hard-POC-associated pathways account for 5.49 × 10² and 6.41 Mg yr⁻¹, respectively. Thus, soft POC dominates upper-ocean Hg^P transport, whereas hard POC contributes only 1.15 % of the total.

Across the 1041 m interface, the downward Hg^P flux decreases by more than three orders of magnitude, reflecting extensive particle remineralization within the intermediate ocean. Although soft POC remains the dominant carrier, the hard-POC-associated contribution increases to approximately 10.0 %.

In contrast, the contribution pattern changes with increasing depth, and hard POC becomes relatively more important than in the upper ocean. Over the upper 1000 m, the vertically integrated diagnostic attributes 98.55 % of the combined soft- and hard-POC-associated Hg^P flux to the soft-POC pathway. Across all seafloor grid cells, the modeled Hg^P export is 16.86 Mg yr⁻¹. For seafloor cells deeper than 1000 m, the flux-weighted hard-POC-associated contribution increases to 43.94 % and locally exceeds 99 %. The persistence of hard POC at depth supports its role as an important carrier of Hg^P below 1000 m. However, its impact on deep-ocean Hg^P inventories remains limited due to both the small Hg load and the continued degradation of POC. Overall, the biological pump redistributes mercury vertically via particle-associated transport, with pronounced stratification.

To assess the model's sensitivity to redox rates, we compared a 20-year baseline simulation with a 20-year reduced-rate simulation following Bieser et al. (2023): the light-normalized Hg⁰ oxidation coefficient was lowered from 4.7 × 10⁻⁶ to 2.4 × 10⁻⁷ m² W⁻¹ s⁻¹, and the Hg^II photoreduction from 1.6 × 10⁻⁶ to 1.0 × 10⁻⁸ m² W⁻¹ s⁻¹, while all other parameters remained unchanged.

These changes lead to weaker net reduction at the surface (0–50 m), lowering net redox fluxes from 1.67 × 10⁵ to 8.22 × 10³ Mg yr⁻¹. Hg⁰ evasion slightly increased from 4.43 × 10³ to 4.86 × 10³ Mg yr⁻¹. The most pronounced changes occur in the surface ocean. At 5 m, Hg^II concentrations decrease substantially, while Hg⁰ concentrations approximately double. This leads to a 10 % increase in the air–sea evasion flux of elemental mercury. Meanwhile, total mercury concentrations in the surface ocean decline markedly, although deep-ocean total mercury remains largely consistent with the baseline simulation. In the mid-depth ocean, net reduction increased from 1.53 × 10³ to 3.62 × 10³ Mg yr⁻¹. However, in the deep ocean (> 1000 m), the redox budget remained net oxidizing, but the magnitude of net oxidation decreased from 3.69 × 10³ to 1.11 × 10³ Mg yr⁻¹, consistent with the reduced oxidation rates.

The monthly variability of net transformations (Fig. 18) further illustrates this vertical redistribution. In surface waters, the seasonal cycle of net reduction decreases, while demethylation maintains a weak late-summer peak. In the intermediate ocean, the decrease of redox rates increases net reduction, whereas net demethylation changes more modestly. In the deep-ocean, the redox budget remains net oxidizing, but its magnitude is substantially reduced. Methylation and demethylation exhibit weaker seasonal variability than the inorganic redox reactions.

Overall, the sensitivity experiment shows that uncertainty in the prescribed redox rates propagates beyond the direct oxidation and reduction fluxes to upper-ocean Hg speciation, methylation, seasonal variability, and air–sea exchange. The strongest responses occur in the surface ocean, where the changes in Hg0 and HgII concentrations are accompanied by enhanced Hg0 evasion. Comparisons with the currently available observations do not consistently favor either parameter set, identifying the redox-rate parameterization as an important source of uncertainty in the simulated upper-ocean Hg cycle. Further in situ and laboratory measurements under environmentally relevant Hg concentrations are needed to constrain these rates more accurately.