Assessment of MITgcm tidal currents in the context of the ODYSEA mission

1 Introduction

Tidal currents play a fundamental role in ocean dynamics by contributing to energy dissipation, vertical mixing, sediment transport, and exchanges between coastal and open ocean waters (Wunsch, 1975; Egbert and Ray, 2001; Stammer et al., 2014; Simpson and Sharples, 2012). Accurate representation of tidal currents is therefore essential for a wide range of applications, including coastal circulation studies, climate modelling, operational oceanography, and the interpretation of satellite observations (Arbic et al., 2010; Egbert and Ray, 2017).

Despite significant progress in global tidal modelling, simulating tidal currents remains particularly challenging in shallow coastal environments, narrow straits, and high-latitude regions. In these areas, complex bathymetry, steep topographic gradients, sea-ice cover, and sparse observational data can significantly affect model accuracy (Egbert et al., 2004; Arbic et al., 2012; Lyard et al., 2021). While global tidal models such as TPXO and FES reproduce tidal elevations with high accuracy through assimilation of satellite altimetry and tide gauge observations (Egbert et al., 1994; Lyard et al., 2006; Carrère et al., 2013; Lyard et al., 2026), the validation of tidal currents remains more limited, particularly in coastal and shelf regions where tidal flows strongly influence ocean circulation (Arbic et al., 2010; Ray and Zaron, 2016). Independent validation using in-situ current measurements is therefore required to assess the reliability of tidal current simulations in such environments.

High-resolution ocean general circulation models provide an alternative approach to studying tidal dynamics because they simulate tides together with the full ocean circulation. The Massachusetts Institute of Technology General Circulation Model (MITgcm) (Marshall et al., 1997; Adcroft et al., 2018) solves the hydrostatic primitive equations and allows the explicit representation of tides, stratification, and mesoscale to submesoscale processes. High-resolution configurations such as the LLC2160 setup have been widely used to investigate ocean dynamics at kilometre-scale resolution, including tidal processes and internal wave generation (Rocha et al., 2016; Arbic et al., 2018; Savage et al., 2017). Because it represents both barotropic tides and background circulation, MITgcm provides dynamically consistent surface current fields that are particularly useful for satellite mission simulations.

This capability is especially relevant in the context of the upcoming ODYSEA mission (Ocean Dynamics and Surface Exchange with the Atmosphere), a joint CNES–JPL satellite mission currently planned for launch around 2030. ODYSEA aims to provide the first global, high-resolution and simultaneous observations of ocean surface winds and surface currents, enabling new investigations of air–sea interactions and upper-ocean dynamics (Rodríguez et al., 2019; Torres et al., 2022).

The ODYSEA satellite will measure ocean surface currents using the Air-Sea Velocity Radar Instrument (AVRI), a wide-swath rotating Doppler scatterometer operating in the Ka-band (35 GHz). The mission is designed to provide 5 by 5 km spatial sampling and near-global coverage, enabling simultaneous observations of ocean surface currents and winds (Lee et al., 2025; CNES, 2026). This sampling strategy allows the observation of surface current variability at spatial scales comparable to those resolved by high-resolution ocean circulation models such as MITgcm LLC2160.

Because ODYSEA measures ocean surface currents directly, tidal currents may contribute substantially to the observed surface-current variability in coastal and high-latitude regions, where tidal currents can reach several cm s⁻¹ or more (Poulain et al., 2015). Evaluating the realism of tidal currents in the MITgcm simulations used as a nature run is therefore important for assessing the capability of ODYSEA observations to capture tidal variability.

Simulations of ODYSEA Level 2 data conducted using MITgcm outputs were provided by OceanDataLab (2024). These outputs were produced to represent realistic ocean surface currents and serve as a testbed for data processing and mission preparation. In this study, we use these existing model outputs to evaluate tidal current representation against in-situ observations.

For the ODYSEA mission, synthetic observations generated from MITgcm simulations provide a framework for testing retrieval algorithms, evaluating sampling strategies, and assessing potential scientific applications of the mission. In this context, high-resolution numerical simulations are commonly used as nature runs to generate synthetic observations representative of the expected measurement conditions. The objective of a nature run is not necessarily to reproduce the exact ocean state, but rather to realistically represent the dominant physical processes governing ocean circulation. Ensuring realistic tidal dynamics, including tidal energy levels, phase propagation, and bathymetry-driven tidal steering, is therefore important for producing physically consistent synthetic observations. Although tidal signals may be filtered or accounted for during subsequent processing, their accurate representation in the nature run remains important.

Beyond mission preparation, ODYSEA observations may also provide new opportunities for the study of surface tidal currents in regions where in-situ observations remain sparse, particularly at high latitudes and over continental shelves. Evaluating the realism of tidal currents simulated by MITgcm is therefore an important step in assessing the reliability of the synthetic datasets used for ODYSEA mission preparation and the potential of future satellite observations for tidal studies.

The objective of this study is therefore to quantitatively assess the accuracy of tidal current magnitudes simulated by MITgcm. Modelled tidal ellipses derived from the LLC2160 configuration are compared with in-situ current meter observations distributed across eight coastal and high-latitude regions characterized by strong tidal dynamics. Model results are also compared with the FES2022 global tidal solution (Lyard et al., 2026), which provides a reference barotropic tidal model. By focusing on tidal ellipse parameters and semi-major axis amplitudes, this study provides a detailed evaluation of MITgcm tidal current performance and discusses its relevance for the preparation of the ODYSEA mission and future oceanographic applications.

The remainder of the paper is organized as follows. Section 2 describes the observational datasets and the MITgcm model outputs used in this study, as well as the tidal analysis methodology. Section 3 presents the comparison between modelled and observed tidal currents across the different regions. Section 4 discusses the main sources of discrepancies and the implications of these results for ODYSEA mission preparation.

2 Data and methods

2.1 In situ current measurements

The in-situ observations used in this study were obtained from the Global Multi Archive Current Meter Database (GMACMD) (Scott et al., 2010; Scott and Furnival, 2013), which compiles current measurements from multiple international oceanographic programs. The dataset includes measurements collected using acoustic Doppler current profilers (ADCPs) as well as mechanical or electromagnetic current meters, depending on the deployment period and location.

Eight study regions were selected in order to sample geographically diverse environments characterized by strong tidal currents and where long current-meter records are available. These regions include high-latitude seas (Greenland, Svalbard, Weddell Sea), energetic continental shelves (Alaska, eastern United States), and complex coastal systems influenced by bathymetric steering (Faroe Islands and Australia). Such environments are particularly relevant for ODYSEA because tidal currents represent a significant component of the total surface current variability in coastal and shelf regions.

The selection criteria (record duration greater than 300 days, M2 tidal current amplitude larger than 10 cm s⁻¹, and sampling interval shorter than 60 minutes) were applied to all available current-meter records within these regions in the GMACMD database in order to ensure sufficient data quality and temporal resolution for tidal analysis. A total of 62 moorings satisfied these criteria across the eight regions considered in this study (Figure 1). To increase spatial coverage in a tidally energetic region of particular relevance for ODYSEA sampling, 44 additional current-meter records from the Australian Integrated Marine Observing System (IMOS; https://imos.org.au) were also included.

Each time series underwent several diagnostic checks, including spike removal, verification of temporal continuity, exclusion of records with data gaps exceeding 10%, and consistency checks between the zonal (u) and meridional (v) velocity components. For moorings equipped with multiple current meters at different depths, depth-averaged velocities were computed in order to isolate the barotropic component of the tidal current.

2.2 Model data

The MITgcm simulations used in this study were not produced by the authors but correspond to publicly available outputs generated as part of the ODYSEA mission preparation activities. These datasets were produced using the LLC2160 configuration with a high resolution, varying from 1.7 km at high latitudes to 4.6 km at low latitude (Arbic et al., 2018) and distributed by OceanDataLab (OceanDataLab, 2024).

The present study therefore evaluates the tidal currents contained in these existing simulations rather than performing new numerical experiments.

In this study, tidal analysis of the model output was performed using hourly surface current velocities provided in the available MITgcm dataset. Depth-averaged barotropic velocities could in principle be derived from the full three-dimensional model output, but only surface currents were available in the processed dataset distributed for ODYSEA mission preparation. Using surface currents is also consistent with the observational perspective of the ODYSEA mission, which will measure ocean surface currents.

In many continental shelf environments, tidal currents are largely barotropic and vertical shear is generally weak, so that surface velocities provide a reasonable proxy for the dominant tidal signal. However, surface currents may also include contributions from baroclinic tides or background circulation, which may partly contribute to the amplitude differences observed between model results and in-situ measurements.

The FES2022 global tidal model (Lyard et al., 2026; Carrere et al., 2022) was used as a reference barotropic solution. FES2022 assimilates satellite altimetry and tide gauge data and provides tidal elevations and barotropic currents with enhanced resolution in coastal and shallow-water areas In this study, FES2022 tidal currents were analysed using the same harmonic constituents as MITgcm and the in-situ data.

Shallow waters regions are defined here as areas with water depth shallower than 200 m, and high-latitude regions as latitudes poleward of 60°.

2.3 Analysis approach

In each of the eight geographical regions, tidal ellipses derived from FES2022 and MITgcm were compared with those computed from observational current meter data. Instead of analyzing the traditional Cartesian velocity components (u, v), tidal ellipse parameters were used, as they provide a compact representation of the amplitude and directional characteristics of tidal currents.

Tidal currents were extracted from the velocity time series using classical harmonic analysis (Pawlowicz et al., 2002). For each tidal constituent, the horizontal velocity components were expressed as:

u(t)=∑k=1NAkcos(ωkt+∅k), v(t)=∑k=1NBkcos(ωkt+ψk),

where A_k and B_k are the amplitudes, ω_k the angular frequencies, and ∅k, ψk the phases of tidal constituents k.

From these coefficients, the complex amplitudes were computed and converted into tidal ellipse parameters: semi-major axis, semi-minor axis, inclination, and phase. These parameters summarize the amplitude and directional variation of tidal currents over a full cycle, making it easier to compare different tidal models and observations The semi-major axis, which represents the maximum tidal current amplitude, was used as the primary metric for model evaluation.

Due to the complexity of retrieving full three-dimensional (3D) velocity fields from the tidal MITgcm outputs, only surface current velocities were analysed. This approach is justified in the present study because the selected regions are predominantly coastal and continental shelves environments, where barotropic tidal processes dominate and vertical shear associated with baroclinic tides is generally weaker (Arbic et al., 2004). Surface velocities therefore provide a representative estimate of the dominant tidal signal in these areas.

Once the tidal ellipses were computed for observations and both models, differences in semi-major axis amplitudes were quantified using the root mean square error (RMSE), defined as:

RMSE= 1N∑i=1N(Aimodel−Aiobs)2,

where Aimodel and Aiobs denote the modelled and observed semi-major axis amplitudes for each station i, respectively and N is the total number of stations within a region.

In addition to the absolute RMSE values, a normalized metric was computed to assess the relative magnitude of model errors with respect to the observed tidal signal. For each region, the relative error was defined as:

Relative error(%)=RMSEAobs,max×100

with RMSE is the root mean square error of the semi-major axis amplitudes across all moorings within a region, and A_obs,_max is the maximum observed semi-major axis amplitude in that region. This normalization provides a dimensionless measure of model performance, allowing inter-regional comparison despite differences in tidal energy levels. The corresponding RMSE values, observed maximum and mean amplitudes, and computed relative errors are reported in Table 1.

Table 1. Regional comparison of model performance based on observed amplitudes.

To account for the heterogeneous number of moorings per region, uncertainty in the regional RMSE estimates was quantified using bootstrap resampling (10,000 realizations), providing confidence intervals for each regional mean, derived from the 2.5th and 97.5th percentiles of the resulting RMSE distribution. This method provides uncertainty bounds for areas with few stations, ensuring that results from regions with small sample sizes (e.g., Weddell Sea, California) are interpreted cautiously.”

Nine major tidal constituents were considered: M2, O1, P1, K1, Q1, N2, S2, K2 and M4. FES2022 was used as a reference barotropic tidal solution, providing a benchmark against which the MITgcm tidal current magnitudes could be evaluated. Lower RMSE values indicate better agreement with observations, while larger discrepancies may reflect limitations, that arise due to model errors as lack of resolution, and likely related to bathymetric representation, or tidal forcing parameterization within the circulation model.

3 Results

This section presents a quantitative comparison of tidal currents simulated by MITgcm and FES2022 against in-situ current meter observations across eight geographically diverse areas. Results are first described in terms of spatial patterns of the dominant M2 tidal constituent, and then analysed quantitatively using tidal ellipse parameters and root mean square error (RMSE) statistics.

3.1 Spatial distribution of M2 tidal current amplitudes

Figure 2 shows the spatial distribution of the M2 tidal current amplitudes from both MITgcm and FES2022 across four representative macro- and meso-tidal regions. Overall, both models exhibit strong spatial consistency, capturing the primary hydrodynamic features, particularly the dramatic tidal amplification over shallow continental shelves, within narrow constrictions and along steep bathymetric slopes (> 0.6-0.8 m s^-1).

In the Gulf of Alaska and Bering Sea, both models capture the intense tidal currents (> 0.8 m s^-1) in Bristol Bay and along the Aleutian Passes, driven by coastal shoaling and geometric constrictions. However, MITgcm predicts a broader offshore extent of elevated amplitudes in Bristol Bay compared to FES2022. This discrepancy is likely attributable to differences in effective bathymetric resolution and bottom friction parameterizations, as FES2022 benefits from data assimilation and localized unstructured grid refinement that restricts maximum currents closer to the coast.

In Greenland (Davis Strait and Hudson Strait), both models exhibit near-perfect spatial alignment, accurately highlighting extreme current velocities (> 0.8 m s^-1) at the mouth of the Hudson Strait. This phenomenon reflects the well-documented resonant response of the Hudson Bay–Strait system to the semidiurnal M2 tide. In the North-East Atlantic (Faroe Islands), both models capture the strong tidal steering around the Faroe Shelf (> 1 m s^-1). Nevertheless, MITgcm depicts a localized southward expansion of high velocities relative to FES2022, which remains tightly confined to the steep insular slope. In Australia, the known regional hotspots—the North-West Shelf, the Gulf of Carpentaria, and Bass Strait—are clearly depicted. Over the North-West Shelf slope and open ocean areas, MITgcm displays fine-scale, high-frequency spatial patterns. Because surface total currents from MITgcm were used, these features reflect localized signatures of internal tide generation along the continental slope, whereas the barotropic hydrodynamic model FES2022 yields a smoother offshore field. Finally, while regional current speeds routinely reach 0.5 to > 0.8m s^-1 (50–80 cm s^-1) in shallow coastal constrictions, the in-situ mooring validation (Figure 2) typically displays smaller ellipse magnitudes (2–20 cm s^-1). This distinction arises because observational moorings are preferentially deployed offshore or along continental slopes in deeper waters where barotropic tidal currents are naturally moderated, unlike the nearshore maxima highlighted in the spatial domain maps. These features are not the focus of the present analysis and are largely absent from the in-situ comparison, which is restricted to coastal and shelf areas dominated by barotropic tidal dynamics.

3.2 Comparison with in-situ global tidal ellipses by area

Figure 3 compares observed total tidal current ellipses (in situ) with those simulated by MITgcm and FES2022 across eight representative oceanographic regions. Statistical performance metrics for each region, including Root Mean Square Errors (RMSE) with 95% confidence interval (CI), relative errors, and observed current magnitudes, are summarized in Table 1. Overall, both models capture the orientation, inclination, and rotation sense of the observed tidal ellipses, indicating a realistic representation of tidal flow dynamics and topographic steering.

In Alaska (N=15), MITgcm reproduces observed tidal ellipses with a regional RMSE of 9.1 cm s^-1 (95% CI: 3.1-14.9 cm s^-1, relative error of 15.6%), compared to 2.8 cm s^-1 (4.8%) for FES2022. Although MITgcm RMSE is larger, the relative error remains modest when compared to high observed peak total tidal amplitudes exceeding 58.1 cm s^-1 (mean: 23.6 cm s^-1). The largest discrepancies occur near the shelf break, where complex slope bathymetry induces strong localized velocity shear.

In Greenland including Hudson Bay region (N=15), both models reproduce the overall magnitude of tidal currents well. MITgcm achieves an RMSE of 4.6 cm s^-1 (13.9%), compared to 3.2 cm s^-1 (9.7%) for FES2022. Larger MITgcm errors occur primarily in Hudson Strait and then Baffin Sea, where several stations are located near steep bathymetric gradients (Figure 3). However, the mean error remains small relative to the maximum observed currents amplitude of 33.2 cm s^-1.

In East USA (N =14) and California (N = 2), FES2022 demonstrates high accuracy with relative errors of 6.3% (RMSE = 3 cm s^-1) and 5.2% (RMSE = 0.4 cm s^-1), respectively. MITgcm exhibits larger errors in these areas, 13.6 cm s^-1 (28.5%) off the East USA and 2.9 cm s^-1 (34.8%) in California, driven mainly by local internal tide signals and steep slope interactions in regions where observed total tidal currents range from moderate (6.2 cm s^-1 in California) to energetic (30.9 cm s^-1 mean in East USA).

In the North-East Atlantic, around the Faroe Islands (N=6), MITgcm overestimates tidal current amplitudes at several insular slope stations, resulting in a regional RMSE of 7.5 cm s^-1 (35.4%), compared to 0.9 cm s^-1 (4.2%) for FES2022 model. Despite this offset, MITgcm accurately preserves the geometric shape and principal axis orientations of the global tidal flow around the Faroe Shelf.

High-latitude subpolar regions highlight key differences in model formulation. In Svalbard (N=5), MITgcm outperforms FES2022, recording an RMSE of 1.6 cm s^-1 (relative error of 15.8%) compared to 2.5 cm s^-1 (25.5%) for FES2022. This demonstrates the strength of MITgcm’s fully dynamic formulation in Arctic conditions where satellite altimetry coverage and assimilation in FES2022 are constrained by ice cover. In the Weddell Sea (N = 1), both models closely mirror the observed ellipse geometry (observed mean total speed of 13.4 cm s^-1), though FES2022 achieves a lower RMSE (0.3\ cm s^-1) than MITgcm (4.9cm s^-1).

Finally, Australia (N=48) exhibits significant regional variability due to its vast, complex shelves (e.g., North-West Shelf, Bass Strait). MITgcm yields an RMSE of 11.7 cm s^-1 (28.2%) compared to 8 cm s^-1 (19.2%) for FES2022. Despite higher absolute errors driven by high localised total tidal current speeds (peak observed: 41.6 cm s^-1), both models successfully reproduce the dominant polarization and flow alignment with local bathymetry.

3.3 Tidal constituent-dependent errors

Figure 4 and Table 2 summarize the mean RMSE associated with individual tidal constituents across all regions. In FES2022, mean RMSE values remain below 2 cm s^-1 for almost all constituents, except for the dominant M2 component which reaches approximately 4.3 cm s^-1. In contrast, MITgcm exhibits larger errors that vary significantly by tidal frequency.

Table 2. Mean RMSE (cm/s) of the semi-major axis of tidal ellipses simulated by MITgcm and compared with observations, by region and tidal constituent.

Values are given as mean RMSE (cm/s), with the 95% confidence interval in parentheses. N stations denotes the number of observation stations. M2, K1, O1, S2, N2, P1, Q1, K2, and M4 are the tidal constituents considered.

In MITgcm, the largest mean errors across all regions are associated with the main semidiurnal and diurnal constituents. S2 shows the highest mean RMSE (~ 9.5 cm s^-1), closely followed by M2 (~ 8.8 cm s^-1), O1 (~ 8.6 cm s^-1), and K1 (~ 8.2 cm s^-1). N2 exhibits an intermediate mean RMSE of 7.2 cm s^-1, whereas lower-amplitude or higher-frequency components (P1, Q1, K2, and M4) display smaller errors ranging between 1 and 3.2 cm s^-1.

The regional contribution of individual constituents to the overall error budget varies markedly (Table 2). In Alaska errors are mainly driven by the semidiurnal tides M2 (8.63 cm s^-1) and S2 (8.14 cm s^-1), alongside O1 (5.44 cm s^-1). In East USA, M2 is by far the largest contributor to the error budget (11.66 cm s^-1), followed by K2 (7.66 cm s^-1) and N2 (4.45 cm s^-1). In California, errors remain generally small across all components, peaking slightly for N2 (3.56 cm s^-1), M2 (3.26 cm s^-1), and S2 (3.14 cm s^-1). In Greenland and Svalbard, discrepancies are heavily dominated by the diurnal components O1 and S2 (7.87 cm s^-1, 6.08 cm s^-1) and (1.81 cm s^-1, 2.80 cm s^-1), respectively. In the Weddell Sea, errors are primarily concentrated in M2 (3.09 cm s^-1) and S2 (2.30 cm s^-1). Finally, in complex dynamic regions such as the Faroe Islands and Australia, no single tidal constituent dominates, instead, high RMSE values (8 -11.6 cm s^-1) are uniformly distributed across all five primary tides (M2, K1, O1, S2 and N2).

4 Discussion

This study provides a quantitative evaluation of tidal surface current magnitudes simulated by the MITgcm LLC2160 configuration across a range of coastal and high-latitude environments. The comparison with in-situ current meter observations shows that MITgcm generally reproduces the main spatial patterns and orientations of tidal currents, although a systematic tendency to overestimate current amplitudes is observed in several regions.

The selected regions should therefore be considered as representative test cases of challenging tidal environments rather than as a statistically representative global sample. They include coastal and high-latitude areas where complex bathymetry, strong tidal currents, and sparse in-situ observations provide relevant conditions for assessing the potential contribution of satellite observations such as those from ODYSEA.

The magnitude of the model errors varies geographically, with regional RMSE values ranging from 1.6 cm s⁻¹ in the Svalbard region to approximately 11.7 cm s⁻¹ and 13.6 cm s⁻¹ in regions characterized by complex bathymetry such as Australia and the East Coast of the United States, respectively (Table 1). When normalized by the maximum observed tidal amplitudes, these errors correspond to relative differences between approximately 15.8% (Svalbard) and 28.2% (Australia) or 28.5% (East USA), respectively. These values are broadly consistent with previous evaluations of high-resolution ocean circulation models in coastal environments, where unresolved bathymetric features and steep topographic gradients can introduce uncertainties in simulated tidal currents (Arbic et al., 2012). The 95% confidence intervals reported in Tables 1 and 2 also highlight the larger uncertainty associated with regions containing fewer moorings, and regional differences should therefore be interpreted with caution.

The main discrepancies in MITgcm simulations are dominated by the primary semidiurnal and diurnal tidal constituents (Figure 4). Although M2 typically dominates the barotropic tidal energy budget, S2 and diurnal components display comparable or larger mean errors in several subpolar and coastal shelf environments (Table 2). This behavior is expected, as major tidal components are highly sensitive to bathymetric representation, grid resolution, and coastal boundary conditions. Previous studies have shown that even small bathymetric errors can locally modify tidal current amplitudes and phase propagation, particularly on continental shelves and near steep slopes (Egbert et al., 2004; Arbic et al., 2018).

Compared with the barotropic tidal model FES2022, MITgcm generally exhibits larger RMSE values. This difference is not unexpected, as FES2022 benefits from the assimilation of satellite altimetry and tide gauge observations, which strongly constrain the barotropic tide at the global scale (Lyard et al., 2026). In contrast, MITgcm simulations rely on hydrodynamic modelling without direct tidal data assimilation. Nevertheless, the MITgcm errors remain within the same order of magnitude as the observed tidal amplitudes in all regions analysed here, indicating that the model captures the main characteristics of tidal current variability.

An important limitation of the present comparison is that MITgcm surface currents were used rather than depth-averaged barotropic velocities. While the observational records were vertically averaged when multiple instruments were available, the model analysis relied on surface velocities because the publicly available LLC2160 outputs used for ODYSEA simulations primarily provide only surface current fields. In many continental shelf environments, tidal currents are largely barotropic and vertical shear is generally weak (Arbic et al., 2010), which supports the use of surface velocities as a reasonable proxy for tidal current magnitude. Nevertheless, surface currents may include contributions from baroclinic tides or background circulation, which may partly explain the slight amplitude overestimation observed in some regions.

Interestingly, MITgcm performs comparably to or slightly better than FES2022 in some high-latitude regions, such as Svalbard (RMSE = 1.6 cm s^-1 vs. 2.5 cm s^-1 for FES2022). The Svalbard region is characterized by steep bathymetric gradient and strongly stratified water masses, which provide favourable conditions for then generation and propagation of baroclinic tides. These processes can lead to substantial vertical variability in tidal currents and may contribute to differences between surface and depth-averaged velocities (Sundfjord et al., 2007; Alford et al., 2015; Pickart et al., 2016), this difference may partly reflect the different observational constraints and model formulations. FES2022 is primarily constrained by satellite altimetry and tide-gauge observations, which may be less effective in regions affected by sea ice. In contrast, MITgcm explicitly represents three-dimensional ocean dynamics. However, the present analysis does not separately quantify barotropic and baroclinic contributions, and therefore the mechanisms responsible for the lower MITgcm error in Svalbard cannot be established directly from the present results.

Previous studies have demonstrated that the representation of bathymetry and grid resolution can strongly influence tidal propagation and dissipation, particularly in narrow straits and shallow shelf regions (Egbert et al., 2004; Arbic et al., 2018). The spatial correspondence between larger model–observation differences and regions of steep bathymetric gradients observed in our dataset is consistent with these previous findings. However, because bathymetric sensitivity experiments were not performed in the present study, we cannot directly attribute the observed discrepancies to bathymetric representation.

In some locations characterized by strong bathymetric steering of tidal currents, the observed tidal ellipses are highly elongated. MITgcm generally reproduces the orientation of these ellipses but sometimes simulates wider ellipses than observed. This behaviour may be related to differences in the representation of small-scale bathymetric features and grid resolution, although this interpretation cannot be directly tested with the present analysis.

This limitation may also have implications when using MITgcm simulations as a nature run for synthetic ODYSEA observations. Wider tidal ellipses may in principle produce a tidal signal that is easier to detect or characterize under a given satellite sampling strategy, potentially leading to an overly optimistic assessment of ODYSEA’s ability to detect or reconstruct strongly bathymetry-driven tidal currents. Nevertheless, MITgcm generally preserves the overall tidal energy levels and ellipse orientations, which are important characteristics for evaluating the potential of ODYSEA sampling. The present results therefore support the use of MITgcm as a physically consistent nature run, while highlighting the need to account for its limitations in representing highly elongated tidal ellipses.

Despite these limitations, the results indicate that MITgcm provides a dynamically consistent representation of the main characteristics of tidal surface currents across the range of coastal and high-latitude environments examined here. This capability is particularly relevant for the preparation of the ODYSEA satellite mission because, unlike tide-only models, MITgcm includes the full ocean circulation and therefore provides realistic total surface current fields for generating synthetic observations.

Future work could further improve the validation of tidal currents by separating barotropic and baroclinic tidal components in the model output and by extending the analysis period to multi-year simulations. In addition, investigating the sensitivity of tidal simulations to bathymetric resolution would help better quantify the sources of model error in complex coastal environments. Such developments would further enhance the value of high-resolution numerical simulations for the preparation of upcoming satellite missions and for broader oceanographic applications.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Marco Bajo, National Research Council (CNR), Italy

Reviewed by: Kenneth Andrew Peterson, Environment and Climate Change Canada (ECCC), Canada

Luca Possenti, Wageningen Marine Research IJmuiden, Netherlands