ORIGINAL RESEARCH article

Front. Mar. Sci., 26 October 2026

Sec. Physical Oceanography

Volume 13 - 2026 | https://doi.org/10.3389/fmars.2026.1949368

Assessing tidal sensitivity to bottom friction: analytical insights for model optimization

  • 1. Tianjin Key Laboratory of Oceanic Meteorology, Tianjin, China

  • 2. Tianjin Meteorological Observatory, Tianjin, China

  • 3. Tianjin Institute of Meteorological Science, Tianjin, China

  • 4. School of Marine Science and Technology, Tianjin University, Tianjin, China

Abstract

Bottom friction coefficients (BFCs) are critical parameters in tidal modeling, with significant effects on hydrodynamic conditions and sediment transport. A comprehensive understanding of BFC variability and its influence on tidal models is essential for model optimization. In this study, we incorporate BFC perturbations through a Dirac delta function into a modified, depth-uniform idealized shallow-water model. By transforming the governing equations into polar coordinates, the non-homogeneous system is recast into a homogeneous form with dynamically consistent dipole sources. Analytical solutions constructed from Bessel function combinations, reveal a series of rotating Poincaré waves that propagate around the perturbation site at frequencies corresponding to odd multiples of the base tidal frequency. The weights of the Bessel functions, which depend non-linearly on local water depths, tidal current amplitudes, and phase lags, determine the rotational directions (clockwise, counterclockwise, or stationary) and provide a theoretical indicator of the tidal system’s sensitivity to BFC perturbations. The analytical solution captures the leading spatial pattern and relative response characteristics in the tested cases. The idealized M2 experiment shows the best quantitative agreement with the analytical solution. Bohai Sea simulations retain qualitative consistency with the theory, although quantitative agreement decreases near the coast. Leveraging this framework, we theoretically assess BFC sensitivity across the Bohai Sea, the Yellow Sea, and the East China Sea. Our analysis indicates comparatively high sensitivity in shallow, strong-current regions, including some tidal-wave entrance and exit zones. This finding may provide a physically based prior for targeted model calibration.

Highlights

  • An analytical solution of the linear shallow-water equations is derived to assess tidal sensitivity to localized bottom-friction perturbations.

  • The idealized and Bohai Sea experiments reproduce the leading spatial patterns predicted by the analytical solution, with lower quantitative skill near the coast.

  • Shallow regions with strong tidal currents tend to show higher sensitivity to BFC variations.

1 Introduction

The use of numerical models is essential for simulating hydrodynamic processes in oceanic and coastal regions. Once initialization and open boundary conditions are specified, the key challenge of these models lies in the accuracy of the parameters, which incorporate information through model variables. In simulations of tides, storm surges, circulations, and sediment transport, the bottom friction coefficients (BFCs) play a critical role in energy dissipation (Davies and Kwong, 2000) and vertical transport of momentum and materials (Wu et al., 2011). It is claimed that BFCs depend on multiple factors such as topography, bottom roughness, and current velocity, making direct observations or measurements impractical. Consequently, how to obtain reasonable BFCs is of great concern in such studies.

Direct estimation of BFCs relies on parameterization schemes guided by in situ observations within geophysical frameworks. One common approach is to link the bottom grain roughness to bottom drag coefficients (Li and Amos, 2001; Werner et al., 2003; Warner et al., 2008; Wu et al., 2011; Xu et al., 2017; Fan et al., 2019). Since observations are usually limited in spatial coverage, establishing the spatial structures of BFCs across an ocean field domain requires extensive observational data, which entails expensive resources. To address these limitations, data assimilation methods treat estimation as inverse problems. By minimizing a cost function that integrates dynamic models with observational innovations, these methods provide a suite of schemes to estimate BFCs over a large region under the principles of optimization theory. Multiple data assimilation methods have proven effective in optimizing the initial conditions and model parameters under dynamic constraints (Bannister, 2017; Qian et al., 2021; Wang et al., 2021; Simon et al., 2022).

In practice, the spatial and temporal variability of BFCs leads to a high-dimensional optimization problem, which not only is computationally costly but also suffers from the risk of convergence to a local minimum. An acceptable way to improve efficiency and mitigate the ill-posedness of data assimilation is to incorporate interpolation into adjoint models. Similar to a one-step multigrid procedure, a series of studies, including those by Lv and Zhang (2006); Zhang et al. (2011), and Guo et al. (2017), optimized BFCs at selected key points directly, whereas the remaining values were interpolated. With advances in data assimilation methods, Liang et al. (2023) followed this approach and optimized the BFC in the Bohai Sea through an adjoint-free ensemble data assimilation scheme. There is no doubt that the locations of these key points are essential for the success of these studies. However, how to determine them is not well discussed or studied.

Although several strategies are presented in numerical experiments, evenly distributed grids remain the most commonly used approach for selecting key-point locations. This preference persists because the underlying mechanisms governing tidal system responses to bottom friction coefficient (BFC) perturbations are not yet fully understood. In the absence of prior knowledge, adopting uniformly spaced grids is a practical and acceptable compromise. However, to the best of our knowledge, few studies have systematically explored the sensitivity of tidal systems to BFC variations. From the perspective of idealized models to gain insight into physical mechanisms, an analytical theory is even more urgent. Early work by Rienecker and Teubner (1980) explored the effect of BFCs using Galerkin and collocation methods. Roos and Schuttelaars (2009) extended Taylor’s classical problem to account for horizontally viscous effects and subsequently considered the bottom friction and topographic steps simultaneously in Roos and De Swart (2022).

The majority of the studies summarized above address how BFCs can be estimated or optimized. The present work instead focuses on a forward sensitivity question: how does the tidal system respond to a prescribed local change in BFC? Addressing this question may also provide a physical basis for selecting key points in subsequent model optimizations. This distinction motivated the analytical framework developed below. Based on the linear shallow-water wave equations, a theoretical analytical solution was systematically derived in polar coordinates to describe the wave field generated by localized bottom friction perturbations at an arbitrary point within the study domain. The solution was identified as a suite of rotational Poincaré waves. The resulting spatial structure was subsequently evaluated against numerical experiments. Using this analytical framework, we were able to identify candidate regions expected to exhibit comparatively stronger local responses to BFC perturbations under the assumptions of the analytical model, thereby providing a physically based sensitivity diagnostic that may complement numerical BFC optimization.

This article is organized as follows. In Section 2, the models, methods, and formulas are derived. Section 3 presents the idealized numerical results first in an idealized rectangular domain and subsequently in a realistic setting for the Bohai Sea. Section 4 presents the sensitivity analysis of the tides in relation to BFCs in the Bohai Sea, the Yellow Sea, and the East China Sea. Section 5 discusses the established theory and concludes the study.

2 Models and methods

2.1 The linear shallow-water model

In ocean modeling, if the research focus is solely on tidal dynamics, two-dimensional shallow-water equations generally provide a sufficient and efficient framework. By adopting a depth-integrated (2D) approach, model complexity and computational cost are significantly reduced while still resolving key tidal features, including wave propagation, amplification, resonance, and dissipation due to bottom friction. Processes such as internal tides or tidal straining, which involve vertical structure and density gradients, can be safely neglected when studying barotropic (surface) tides at basin or shelf scales. Thus, for tidal simulation purposes, the 2D shallow-water equations offer an appropriate and widely validated balance between physical completeness and practical feasibility, making them the standard choice in both idealized studies and operational tidal forecasting with one or more tidal constituents. The shallow-water equations adopted here are widely used in both regional (Matsumoto et al., 2000; Qian et al., 2021) and global tide models (Egbert and Erofeeva, 2002; Lefèvre et al., 2002), and they have been well assessed (Zaron and Elipot, 2021). The linear shallow-water equations provide a powerful tool for analyzing tidal processes.

The classical linear shallow-water equations consist of the momentum and continuity equations without advection terms (see Appendix A), as given in Equations (1a–c). With the effects of bottom friction and the Coriolis force, they are given by

where U and V represent the depth-averaged velocity components in the x and y directions, respectively, and η denotes the free surface elevation. The water depth H is spatially uniform here, while the Coriolis parameter f and gravitational acceleration g are taken as constants for simplification.

Here, represents the bottom friction term, which can take either a linear or non-linear formulation. When the effects of bottom friction vanish, and the analytical solution of the shallow-water equations reduces to the well-known Kelvin waves and Poincaré waves. When a linear formulation is adopted with non-zero constant BFCs, the wave structures are distorted. Analyzing this distortion under a non-linear friction law is more challenging, as the term couples with the flow velocity in a non-linear manner, as expressed by Equations (2a) and (2b):

where Cd is spatially and/or temporally varying and dimensionless.

As mentioned above, to understand the effects of BFCs, it is worthwhile examining the impact of BFC perturbation on the tides. A fundamental and simple approach is to consider a perturbation confined to an infinitesimally small spatial point, which can be conveniently represented using the Dirac function . Let the difference between the solutions with and without the BFC perturbations be . Given that these perturbations are small relative to the original background flow, the system can be linearized, yielding Equations (3a–c):

where the value of the constant c is the product of the BFC perturbations and the grid area in numerical models rather than a small value relative to the BFC Cd itself.

Assumptions of locally uniform depth and Coriolis parameter, a small fixed perturbation, and linearized barotropic dynamics are adopted to isolate the leading local response. The point-source strength is matched to a finite numerical grid-cell perturbation through . The ΔA denotes the grid-cell area. Accordingly, the analytical approximation is expected to be most reliable where bathymetry and background tidal currents vary slowly over the response scale; its limitations are discussed in Section 5.3.

Eliminating either u or v from (3b) or (3a), then taking the divergence and applying the continuity Equation 3c leads to a single equation for ξ only. That is

where

to simplify the notation (in Equation 5).

Equation 4 shows that the effects of tidal currents are coupled with the Dirac delta function. A reasonable assumption is that the amplitudes of tidal currents remain approximately uniform over a limited region. Let the tidal currents be described by amplitude and angular frequency σ* and phase θ* in U and V directions, respectively. This gives the harmonic representation in Equations (6a) and (6b):

which are typical tidal behaviors in many coastal seas and oceans. Thus, one would have

It is easy to see that the right-hand sides of Equation 7 provide dipole sources due to the presence of and .

Physically, a local increase in bottom drag decelerates the incoming tidal flow, producing an upstream water-level accumulation and a downstream deficit. As the tidal current reverses and rotates, the orientation of this dipole forcing changes over time, providing the physical origin of the rotating sea-level pattern derived below.

In fact, the harmonic expansion of the sources shows that

where Fx,n and Fy,n are expressed with coefficients such as ax,n, bx,n and ay,n and by,n (see Appendix C).

2.2 Equations in polar coordinates and Poincaré waves

Because the Dirac point source exists only at a fixed point, the original equation is already homogeneous everywhere else. We therefore solve the homogeneous equation in the region away from the source, while allowing the solution to have a controlled singularity at the Dirac point. This singularity does not affect the homogeneous equation, but when the solution is interpreted in the distributional sense over the whole domain, applying the differential operator to that singularity generates the Dirac delta function, or its derivative, at the source point. In this way, the point source is recovered exactly. To this end, the polar coordinate system with the polar radius r and angle ϕ is introduced. We set the point (x0, y0) to (0, 0) for simplicity.

The homogeneous wave equation is

Applying the method of separation of variables, the solution of Equation 9 is decomposed into three parts: the time variable, the polar radius, and the angle coordinates, as shown in Equation 10.

Thus, the solutions satisfy

where λ and m are eigenvalues of the equations. The solutions of Equations 11 are given in Equations (12a–c):

where the and are weighting coefficients, while the and are Bessel functions of the first and second kind of order m. Combining all these separated variable solutions leads to the general solution to the original problem.

In addition, the velocities are expressed in terms of their components along the radial and azimuthal axes , respectively. They are expressed in Equations (13a) and (13b), respectively:

What is natural and important is that the full current velocity in Equation (14)

depends only on the variable t.

Comparisons between the expressions for and the weight of variable ϕ in the dipole sources give the eigenvalue m. In fact, on the right-hand side of Equation 11, it is shown that only terms Vϕ and Ur involving ϕ as their variables and, in addition, the weights of ϕ are all unity. To ensure the equations hold, m must be one. Similar but more complicated situations apply to the eigenvalue λ by taking the weight coefficients of the variable t in sine and cosine functions into account. It is indicated that the resulting eigenvalue is given by Equation (15):

Combining all solutions in Equation 16, the expressions of the dipole sources and that of the eigenvalues, the general solution is a suite of combinations of the first and second kinds of Bessel functions, i.e.,

with where the temporal and angular basis functions are defined in Equation (17):

and the units of the coefficients and are both meters (m). The wave number of the induced wave above is given by Equation (18):

which satisfies the frequency dispersion relation given in Equation (19): of Poincaré waves, i.e.,

2.3 Solutions considering dipole sources

First, one should note that only the weights of N1 can be determined at this stage. This is because the Bessel functions of the first kind, J1, vanish at the source point. Consequently, any linear combination of these does not affect the expressions of the dipole sources.

Now for a fixed single wave number κn, set

It is easy to see applying the wave operator to Equation (20) gives Equation (21):

Because, in the sense of two-dimensional distributions, the identity in Equation (22) holds:

So, we have substituting Equation (22) into Equation (21) gives Equation (23):

Then, taking the time derivative of both sides of the above equation and comparing its coefficients with Equation 8 yields the coefficients in Equation (24):

Substituting the obtained expression into Equation 20 gives Equation (25):

Similarly, for the y-direction, we have Equation (26):

The (partial) solutions of Equation 8 are then readily obtained with the coefficients summarized in Equation (27):

The determination of the weights of J1, i.e., , can be determined by fitting the numerical results using the least-squares method. Nevertheless, the numerical experiments presented in the following section demonstrate that the Bessel function of the second kind alone can capture the perturbation pattern with acceptable accuracy. This is because a localized BFC perturbation generates a distinct bulge in the water surface along the upstream direction, with the maximum elevation occurring near the perturbation point, a spatial structure that is well described by the second-kind Bessel function rather than the first.

The analytical solution can be interpreted as the dynamical adjustment of the tidal system to a localized momentum sink. A positive perturbation in the bottom friction coefficient increases the quadratic drag acting on the background tidal current. This locally decelerates the flow, producing an upstream free surface increase and a downstream decrease, i.e., a dipole-like pressure anomaly. As the background tidal current changes direction during a tidal cycle, the orientation of this dipole forcing also changes. Away from the perturbation point, the response satisfies the homogeneous rotating shallow-water equations and therefore projects onto order-one Poincaré-wave modes satisfying . The angular factors cos φ and sin φ represent the dipolar structure, while the Bessel functions describe its radial adjustment around the perturbation site. Thus, the Fourier–Bessel solution should not be viewed merely as a mathematical representation, but as the rotating barotropic-wave response generated by a localized perturbation of tidal momentum loss.

3 Validation in numerical experiments

In this section, numerical experiments are performed for an idealized case and realistic cases. The simulated wavefields (hereafter termed “simulated waves”) are reconstructed from analytical weight solutions (hereafter termed “analytical waves”) and/or from the estimated weights using a least-squares method based on the initial conditions (estimated waves) for comparison to validate the theoretical formulation.

3.1 An idealized case in a rectangular domain

3.1.1 Model setup and tide simulations

The idealized model is performed using the external mode of the Princeton Ocean Model (POM). POM is a simple-to-run yet powerful ocean modeling code that can simulate a wide range of problems, from small-scale coastal processes to global ocean climate change. Although its external mode is a two-dimensional, non-linear, shallow-water model with quadratic bottom friction terms, it meets the validation needs here.

A spatially uniform rectangular study domain spans from 100°E to 180°E in the east-west direction and 20°N to 50°N in the south–north direction, with regular grids every 0.5 degrees. The depth is set to 1000 m, representative of the typical scale of the western Pacific Ocean. The model boundaries are closed, and forcing is applied only through the M2 tidal constituent’s generating forces.

Background constant BFCs are set to 0.0025 s. Model spin-up is performed, and the first 4 days of simulation are discarded to ensure a dynamically stable tidal system. Figure 1 shows the cotidal charts for the idealized case. It is clear that the classical tidal wave propagates along the boundaries on its right-hand side. Five amphidromic points are evident, with the largest tidal amplitude (approximately 0.5 m) located near 130°E along the southern boundary. The tidal ellipses are consistent with the tidal wave structure. The amplitudes in the western or southern portions are generally larger than those in the eastern or northern portions. The maximum current amplitude occurs near (130°E, 30°N) and is approximately 4 cm/s.

Figure 1

To directly examine the effect of representing a point-source perturbation with a finite numerical grid cell, two additional idealized experiments were performed at horizontal resolutions of 0.25° and 0.125°, in addition to the original 0.5° experiment. The physical model configuration, water depth, M2 forcing, background BFC, and BFC increment () were retained. In each experiment, the perturbation was applied to one grid cell, and the corresponding analytical source strength was area-matched as . The M2 current harmonic constants used by the analytical solution were recalculated from the corresponding background run at each resolution.

3.1.2 Detection of rotating Poincaré waves

A perturbation of 0.0001 is applied at the grid point corresponding to 140°E, 35°N, which is the center of the model domain. Again, the first 4-day spin-up results are discarded. The direct difference in simulated sea-level heights reveals an intuitive image of the rotating Poincaré wave (see Figure 2). The wave pattern exhibits clockwise rotation with the maximum amplitude on the order of 5×10−8 m. The period of rotation is approximately 13 h, which is the same as that of the M2 tidal constituent, as stated in our theory.

Figure 2

Figure 3 displays the waves at the end of the first hour by fitting the numerical experimental results via the least-squares method (the estimated wave) and the analytical coefficients derived in the last section (the analytical wave). The estimated wave shows good agreement with the direct numerical results. In contrast, the analytical wave exhibits considerably larger amplitudes and slight phase offsets, because the pure analytical reconstruction contains the source-determined N1 particular solution but not the regular J1 component associated with finite-domain boundary conditions.

Figure 3

In idealized experiments, the theoretical analytical particular solution N1 alone is already able to reproduce the basic spatial structure of the numerical perturbation field, with a correlation coefficient , a Willmott agreement index of 0.499, and an NSE of 0.207. However, the relative L2 error remains as high as 0.890, indicating that the pure point-source particular solution can describe the main propagation direction and part of the phase structure of the disturbance waves, but is insufficient to fully represent the wave field within a finite domain. After keeping the theoretical N1 coefficients unchanged and determining only the regular J1 component via least squares (not shown), the fit improves markedly: the correlation coefficient increases to 0.880, the RMSE decreases from to , the relative L2 error drops to 0.478, and the NSE and Willmott agreement index rise to 0.771 and 0.935, respectively. This suggests that, in a uniformly idealized environment closest to the theoretical assumptions, the singular N1 component from the point-source theory possesses good physical plausibility, and the main discrepancy between the numerical field and the pure analytical particular solution can be largely attributed to the regular homogeneous solution corresponding to the finite-domain boundaries. Further fitting both N1 and J1 simultaneously raises the correlation coefficient only to 0.931, indicating that, in this ideal experiment, the theoretically prescribed N1 coefficients are already quite close to the optimal coefficients required by the numerical model.

3.1.3 Grid-resolution sensitivity of the finite-area perturbation

The analytical model represents the localized BFC perturbation as a two-dimensional point source, whereas the numerical model applies the same ΔCd over a finite grid cell. To test whether this difference contributes to the remaining discrepancy, the idealized experiment was repeated at 0.5°, 0.25°, and 0.125° resolutions. Since , grid refinement reduces the integrated strength of a one-cell perturbation together with the cell area. Therefore, absolute RMSE is not used as the primary convergence measure; instead, we compare the correlation and relative L2 error.

For the pure analytical N1 solution, the correlation increases systematically from 0.514 at 0.5° to 0.528 at 0.25° and 0.546 at 0.125°, while the relative L2 error decreases from 0.859 to 0.849 and 0.838, respectively (Table 1). The improvement is modest but monotonic, indicating that the finite area of the numerical perturbation contributes to the point-source/finite-cell discrepancy. However, the improvement is limited. When the theoretical N1 coefficients are retained, and only the regular J1 coefficients are fitted, the correlations remain high (0.937, 0.932, and 0.925) but show no corresponding increase with resolution. Thus, the finite perturbation area is only one component of the analytical–numerical mismatch; finite-domain adjustment and other resolution-dependent numerical effects remain relevant.

Table 1

ResolutionΔA (m²)c (m²)Pure Y1 RPure Y1 RelL2TC RTC RelL2
0.5°2.5531 × 1092.5531 × 1050.5140.8590.9370.349
0.25°6.3829 × 1086.3829 × 1040.5280.8490.9320.362
0.125°1.5957 × 1081.5957 × 1040.5460.8380.9250.381

Grid-resolution sensitivity of the idealized experiment.

“Pure N1” denotes the source-matched analytical particular solution without fitted regular coefficients. “TC” denotes the theory-constrained reconstruction, in which the analytical N1 coefficients are fixed and only the regular J1 coefficients are fitted.

3.2 A realistic case in the Bohai Sea

Beyond the ideal model, it is worth studying the conditions in a real ocean to validate the theory. Here, the experiment is conducted in the Bohai Sea. The Bohai Sea (also known as the Bohai Gulf) is a semi-enclosed inland sea located in northeastern China (Figure 4), serving as the northwestern extension of the Yellow Sea. With an average depth of 18 m and a maximum depth of approximately 80 m, the region exhibits mixed semidiurnal tides. Tidal amplitudes are largest in the Bohai Bay due to resonance, and the area experiences powerful tidal currents near the strait, alongside frequent storm surges from winter winds and summer typhoons. The region’s shallow bathymetry, river-dominated hydrology, and complex tidal dynamics make it a unique region for coastal studies, particularly for the estimation of BFCs (Qian et al., 2021).

Figure 4

3.2.1 Model setup and tide simulations

In this experiment, the external mode of the POM model is employed at a horizontal resolution of 1/25°. The bathymetry is derived from the GEBCO dataset (https://www.gebco.net). Open boundary conditions along the east boundary are prescribed using the FES global tide model dataset (https://www.aviso.altimetry.fr) to drive the model. Consistent with the idealized model, only the M2 tidal constituent is considered.

Figure 5 shows the simulated cotidal chart with a constant BFC set to 0.0013. Due to the comprehensive influences of topography, bathymetry, and sediments, higher tidal amplitudes are found along the offshore regions, with the highest level of 0.8 m occurring in the northern bay. Two amphidromic points emerge in the tidal system near the coast. Moreover, the cotidal charts of the M2 tidal constituent in the Bohai Sea (Figure 5) display the same patterns as those in other studies (not shown).

Figure 5

3.2.2 Detection of rotating Poincaré waves

With a value of 0.0001, the first BFC perturbation is also placed near the center of the domain (38.56°N, 119.86°E). The results are similar to those shown in the preceding section. The Poincaré waves exhibit clockwise rotation with a period of approximately 13 hours. Figure 6 (top) shows the results after 5 days of spin-up. It should be noted that the amplitudes of the waves are up to 5×10−6 m, which is significantly larger than that in the idealized models.

Figure 6

In contrast, as shown in Figure 6 (bottom), the second BFC perturbation is located near the shoreline at (39.6°N, 119.1°E). The highest wave amplitude is only up to 1×10−7 m, which is much lower than that in the centrally perturbed case. A visually striking difference is the markedly altered wave structure, which now includes shorter wavelengths. This shorter-scale component is interpreted as an M4 overtide, generated by non-linear shallow-water dynamics in the POM simulation; it is not a direct harmonic of the idealized pure M2-point-source forcing.

To estimate or reconstruct the waves, if the BFC perturbation is in the middle of the Bohai Sea, the M2 tidal constituent alone is sufficient. However, if the perturbation is located near the shoreline, an M4 component must also be included in the empirical reconstruction. Figures 6b, d show these reconstructions. Reconstructions based on the analytical coefficients are presented in Figure 7. As expected, some detailed wave features are not captured because the pure analytical reconstruction retains the source-determined N1 particular solution and omits the regular J1 component associated with finite-domain boundary conditions. Consistent with earlier results, the analytically derived wave amplitudes exceed those obtained numerically. Although differences exist between the theoretical analysis and the numerical experiments, the simulations retain the contrast between the interior and nearshore responses at the two tested perturbation locations.

Figure 7

In the central Bohai Sea, the purely analytical N1 solution for the disturbance point still exhibits a certain degree of consistency with the numerical results, with a correlation coefficient , which is higher than that in the idealized experiment, and a Willmott , indicating that the analytical solution can reasonably identify the large-scale spatial distribution characteristics of the numerical perturbation field. However, the RMSE is and the relative L2 error is 0.897, suggesting that there remain notable errors in amplitude and local waveform. Notably, when both the N1 and J1 coefficients are allowed to be empirically fitted, the correlation coefficient reaches 0.920, the RMSE drops to , and the NSE and Willmott d attain 0.845 and 0.957, respectively. Therefore, the Fourier–Bessel spatial structure itself can still well represent the disturbance waves in the central Bohai Sea, but factors such as realistic water depth variations, spatial inhomogeneity of the background tidal currents, and basin boundaries cause the amplitude and phase of N1 given by the idealized local point-source theory to deviate from the optimal effective coefficients in the numerical model.

The coastal experiment exhibits characteristics distinctly different from those of the previous two groups. When using only the theoretical N1 particular solution, the correlation coefficient is merely 0.181, the relative L2 error reaches 1.218, the RMSE is , and the NSE is −0.484, indicating that the purely analytical point-source solution can no longer effectively reproduce the numerical perturbation field in this coastal environment. In contrast, when both N1 and J1 are allowed to be empirically fitted, the correlation coefficient significantly increases to 0.852, the RMSE drops to , the relative error decreases to 0.523, and the NSE and Willmott d reach 0.726 and 0.915, respectively. This demonstrates that the Bessel-type spatial basis still possesses strong descriptive capabilities, but the non-linearity, coastal boundary constraints, strong water-depth gradients, and higher-order shallow-water tidal constituents generated by the numerical model in the shallow-water coastal environment exceed the applicability of the current uniform depth, local linear point source theory. The coastal case therefore serves primarily to illustrate the boundary of the theory’s applicability, rather than providing a quantitative validation comparable to the idealized experiment.

4 Sensitivity of tides to BFCs

The numerical experiments in Section 3 show that the Fourier–Bessel structure captures important features of the perturbation field, although quantitative skill depends on environmental complexity. The coefficients of the second-kind N1 particular solution are determined directly from local tidal currents, water depth, and perturbation strength, whereas the regular first-kind J1 component requires boundary or initial-condition information and may be estimated from numerical results. Here, we use the analytically determined N1 coefficients as a comparative indicator of local BFC sensitivity.

From this perspective, we explore the sensitivity of the main tide (M2) to BFCs in the East China Sea, the Yellow Sea, and the Bohai Sea. The tidal currents’ harmonic constants are taken from the FES global tide model dataset, and the bathymetry is from the GEBCO dataset. Because the induced waves rotate and propagate, the sign of the weight coefficients is not meaningful for assessing sensitivity; only their magnitudes matter. Only the mean absolute values (MAVs) of the analytical weights are considered.

As shown in Figure 8, in the study domain, within the present M2 analysis, elevated values of the analytical indicator occur mainly along shallow coastal regions with strong tidal currents. Relatively large values are found in parts of the Yellow Sea coast, including regions associated with strong tidal exchange.

Figure 8

The analytical formulation suggests that the sensitivity of tides to localized perturbations in bottom friction is primarily controlled by the combined effects of tidal current magnitude and water depth. To leading order, the amplitude of the source-induced N1 mode scales approximately with , with additional modulation by latitude and tidal-ellipse geometry. Consequently, enhanced sensitivity is expected mainly over shallow regions with strong tidal currents, such as continental shelves, straits, tidal channels, and bay entrances, whereas deep-ocean regions are generally less sensitive. The Coriolis effect further modifies this pattern through the factor , particularly for diurnal constituents near their critical latitudes. Within the present M2 analysis, this scaling provides a physical interpretation of the mapped sensitivity pattern; extension to other constituents or regions requires further validation.

5 Discussion and conclusions

5.1 Discussion on the geophysical perspective

Referring to Equation 16, variations in the weights induce waves that rotate clockwise, counterclockwise, or remain stationary. The origin of the wave structure with alternating positive and negative phases lies in the angular-dependent function within the driving term, which ensures that the eigenvalue remains non-zero. A pertinent question arises: Can the angular dependence in the driving term be eliminated? This is indeed feasible under specific conditions. For instance, in the case of a steady flow where the driving velocity is time-invariant, the eigenvalue would reduce to zero. Nevertheless, the Coriolis parameter ensures that the driving term retains an angular dependence. Only under the additional assumption of a non-rotating Earth would the driving term vanish entirely. Consequently, if one examines the perturbation effects of bottom friction on a steady flow in the absence of planetary rotation, such wave phenomena would not manifest.

In essence, a highly intriguing wave mode and the emergence of the rotating wave structure are nearly inevitable. This phenomenon can be rationalized by considering tidal flow dynamics near an anomaly point: flow deceleration leads to water accumulation and a consequent rise in surface elevation on the upstream side, while the downstream side experiences a decline due to insufficient replenishment. This persistent driving mechanism, modulated by the direction and magnitude of the incoming flow, naturally gives rise to the observed pattern.

5.2 Relation to existing studies and our contributions

In this article, the influence of bottom friction coefficient perturbations on tidal simulations is discussed in detail. As mentioned above, an increase in the bottom friction coefficient leads to the blocking of tidal flow in the incoming direction, resulting in an upstream bulge and a downstream depression. Combined with the rotational effect of the Coriolis force, this leads to the formation of outward-radiating waves centered on the perturbation point.

The comparison with studies of isolated seamounts and islands is intended as a mathematical analogy rather than a claim of physical equivalence. Internal-tide generation over seamounts (Baines, 2007; Le Dizes et al., 2025) involves stratification and barotropic-to-baroclinic energy conversion, which are absent from the present depth-uniform model, while near-inertial motions around islands (Siegelman et al., 2023) involve a different forcing mechanism. Nevertheless, localized forcing in a rotating system can produce qualitatively similar dipole-like or radiating spatial patterns.

While these studies adopt comparable methodologies and formulations, the motivation of our work differs substantially from theirs. Beyond these similarities, the primary aim of our study is to examine BFC sensitivity from a forward analytical perspective, which has received comparatively less direct treatment. A practical contribution of our work is the construction of an indicator, i.e., one that makes it possible to calculate indicators analytically and directly from tidal harmonic constants without any numerical simulations that may be computationally costly. Specifically, whereas those investigations are primarily dedicated to elucidating the excitation mechanisms of internal waves and internal tides in realistic ocean environments, our study aims to inform BFC sensitivity analysis and potentially support subsequent model optimization. This distinction underscores the adaptability and utility of similar theoretical frameworks when applied to different research questions. With these distinctions in mind, it is equally important to recognize the assumptions underlying our analytical framework and the conditions under which it can be reliably applied.

5.3 Applicability and limitations

The analytical framework is intentionally local and first order. It assumes locally uniform water depth and Coriolis parameter, a small area-matched point perturbation, a prescribed harmonic background tide, and linear barotropic dynamics. It is therefore expected to perform best where bathymetry and background currents vary slowly over the response scale and where boundary and non-linear overtide effects are limited; the nearshore Bohai experiment illustrates the breakdown of these assumptions. Extensions to finite-area sources, spatially varying bathymetry, multiple tidal constituents, non-linear interactions, and operational forecasting would require further theoretical development and independent numerical or observational validation.

The grid-resolution experiment further shows that reducing the finite numerical perturbation area produces a modest but systematic improvement in agreement with the pure point-source solution, supporting finite-area regularization as one contributor to the discrepancy. The limited magnitude of this improvement, however, indicates that finite-area representation is not the dominant source of error, and a formal analytical treatment of finite-area sources remains a topic for future work.

5.4 Conclusions

A theoretical shallow-water equation framework is developed to describe the response of tidal systems to perturbations in the BFC values. These equations are expressed in polar coordinates and solved using the method of separation of variables. The resulting general solutions take the form of wave expressions involving Bessel functions that correspond to a suite of rotating Poincaré waves.

The numerical experiments in the idealized domain and the Bohai Sea show that the analytical solution captures the leading disturbance pattern, while quantitative skill decreases as environmental complexity increases, especially near the coast. Accordingly, the analytically determined N1 coefficients are used here as a comparative indicator of BFC sensitivity rather than as a universally validated measure of calibration importance. For the M2 application in the Bohai Sea, Yellow Sea, and East China Sea, larger indicator values are generally associated with shallow regions of strong tidal currents, including several tidal-wave entrance and exit zones. The framework may therefore provide a physically based prior for sensitivity analysis and calibration design for tide and ocean models.

Statements

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions

KL: Conceptualization, Funding acquisition, Software, Visualization, Writing – original draft, Writing – review & editing. WL: Conceptualization, Formal Analysis, Funding acquisition, Methodology, Supervision, Writing – original draft, Writing – review & editing. XQ: Resources, Writing – original draft. LL: Resources, Writing – original draft. QD: Resources, Writing – original draft.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This research was co-sponsored by grants from the National Natural Science Foundation (42406194) of China, the National Natural Science Foundation (42376190) of China, the Science and Technology Key Program of Tianjin (25ZXSFSN00060), and the Open Project of the Tianjin Key Laboratory of Oceanic Meteorology (2024TKLOM02).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

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.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2026.1949368/full#supplementary-material

References

Summary

Keywords

analytical solution, bottom friction coefficient, Poincaré waves, shallow-water equations (SWE), tidal sensitivity

Citation

Liang K, Li W, Qiu X, Liu L and Dong Q (2026) Assessing tidal sensitivity to bottom friction: analytical insights for model optimization. Front. Mar. Sci. 13:1949368. doi: 10.3389/fmars.2026.1949368

Received

27 July 2026

Revised

20 September 2026

Accepted

24 September 2026

Published

26 October 2026

Volume

13 - 2026

Edited by

Chunyan Li, Louisiana State University, United States

Reviewed by

Wenang Xie, Seikei University, Japan

Muhammad Zainuddin Lubis, Shanghai Ocean University, China

Updates

Copyright

*Correspondence: Wei Li,

Disclaimer

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.

Outline

Figures

Cite article

Copy to clipboard


Export citation file


Share article

Article metrics