Heat transfer in the seabed boundary layer

Abstract We present a theoretical model of the temperature distribution in the boundary layer region close to the seabed. Using a perturbation expansion, multiple scales and similarity variables, we show how free-surface waves enhance heat transfer between seawater and a seabed with a solid, horizontal, smooth surface. Maximum heat exchange occurs at a fixed frequency depending on ocean depth, and does not increase monotonically with the length and phase speed of propagating free-surface waves. Close agreement is found between predictions by the analytical model and a finite-difference scheme. It is found that free-surface waves can substantially affect the spatial evolution of temperature in the seabed boundary layer. This suggests a need to extend existing models that neglect the effects of a wave field, especially in view of practical applications in engineering and oceanography.


Introduction
The present manuscript investigates how surface water waves impact temperature transport at the seabed. The subject of temperature conduction in fluids is incredibly varied, with a vast literature (see e.g. Landau & Lifshitz (1989) and White (1991) for useful background information). In the most general setting, the temperature distribution and flow field mutually influence one another, which makes for a rich and complex problem. However, it is often possible to adopt the simplifying assumption that the flow is independent of temperature.
When it comes to the oceanic environment, temperature transport has been studied at large as well as small scales. At the largest scales, wind-driven gyres and overturning circulation are responsible for oceanic heat transport from the tropics to higher latitudes (Ferrari & Ferreira 2011). Turning to the small scales, research investigations have been devoted to studying temperature in the surface boundary layer, which forms the interface between ocean and atmosphere. Early work by O'Brien (1967) considered † Email address for correspondence: simone.michele@plymouth.ac.uk heat transfer at wavy surfaces specified in Lagrangian coordinates, whereas Witting (1971) later investigated the role of surface gravity and capillary waves on surface heat flux. Experimental work by Veron, Melville & Lenain (2008) has also underscored the importance of surface gravity waves in driving air-sea heat transfer in the ocean.
To the best of our knowledge, the effect of surface waves on heat transfer in the seabed boundary layer has not hitherto been studied. Whereas the motion associated with surface waves decreases with depth, in intermediate depths (relative to the length scale of the waves) this motion extends all the way down the water column. This is an important driver of seabed mass transport (Mei & Chian 1994), as well as the transport of solutes and contaminants (Winckler, Liu & Mei 2013). In the present paper, we show that this motion is responsible for temperature convection.
For a suitable choice of scales, and an idealised flat and smooth bed, we are able to treat the problem analytically by means of perturbation theory. We assume the fluid is incompressible and viscous, enabling the fluid density to be independent of fluid pressure, and the hydrodynamic problem to be decoupled from its thermodynamic counterpart. The second-order wave-induced motion is determined by an approach analogous to that used in investigating mass transport by Mei, Stiassnie & Yue (2005, Chapter 10). This motion drives flow in the boundary layer, and results in a convection-diffusion equation for the temperature. This approach is mathematically similar to earlier studies by Lighthill (1950Lighthill ( , 1954.
By introducing a slow time scale and similarity variables, we are able to solve the equation for the seabed temperature transport to yield explicit expressions for the heat transfer. We also obtain a theoretical criterion by which to estimate the frequency of free-surface waves that maximises heat exchange, and validate the analytical approximation by comparison with a full numerical solution based on a finite-difference scheme. Our results suggest that the effect of free-surface waves should be included in existing models to ensure proper estimation of the temperature field near the seabed, especially when such models are applied in practice.

Mathematical model
Let us consider a two-dimensional fluid domain Ω (x, z, t), where x is horizontal distance, z is vertical distance upwards from the horizontal seabed and t is time. We assume that the fluid is viscous and incompressible, in which case the equation of heat convection and diffusion in a laminar boundary layer assumes the simplified form (Landau & Lifshitz 1989) ∂T ∂t where T is temperature, χ is thermometric conductivity and (u, w) are the horizontal and vertical components of the velocity field. Diffusion effects are expected to be greatest close to a heat source because of the low thermometric conductivity of water. For this reason, we introduce the following non-dimensional quantities (Mei et al. 2005 and ν the kinematic viscosity coefficient. We also define the small parameter = Ak 1, which represents the steepness of the propagating waves. Substitution of (2.2a-f ) into (2.1) yields the governing equation in non-dimensional form: where the small parameter μ is defined by The assumption that μ is comparable with is justified for fluids characterised by small thermometric conductivity, such as water, as will be shown in § 3. Given that δk ∼ O( 4 ) is very small, (2.3) can be simplified as follows: in which we neglect third-order contributions. Therefore, convective effects driven by the moving fluid appear at second order O( ), whereas diffusion is significant at third order O( 2 ). Having obtained the governing equation for fluid temperature, we now derive expressions for the components of the fluid velocity field in the boundary layer at the seabed. The dynamic problem is decoupled from its thermodynamic counterpart, and so the water-particle velocity components (u, w) are determined by analogy to the analysis of mass-transfer phenomena by Mei et al. (2005, chap. 10).

Flow field in the laminar boundary layer
The mass continuity and Navier-Stokes equations in the boundary-layer region can be approximated by where ρ denotes fluid density, g acceleration due to gravity and P the total pressure. Here, the dynamic component of the pressure p = P − ρgz does not depend on vertical elevation and matches the value in the inviscid flow region outside the boundary layer. Hence, where U is the horizontal component of the inviscid flow velocity at the seabed, and the corresponding vertical component is zero. By assuming the following perturbation expansion in terms of = Ak 1, , we obtain the following leading-order problem O(1): with U 0 a function of x, we obtain (2.14) By integrating the continuity equation (2.6), the vertical component of water-particle velocity at the leading order may be expressed as ( 2.15) The governing equation at second order is given by which admits second (2ω) and zeroth harmonic solutions. The second-harmonic component contributes only a small oscillatory correction to the first harmonic obtained at leading order, which can be eliminated by time averaging over the period 2π/ω. More significant is the drift associated with the zeroth harmonic, for which we obtain where the overbar represents the averaged value of the relevant variables. The boundary conditions at second order reduce to and the horizontal drift velocity component is given by in which and () * denotes the complex conjugate of the relevant variable. Let us consider a progressive wave propagating in the positive direction described by the following velocity potential (Mei et al. 2005, Chapter 1): in which A is the wave amplitude and h is the undisturbed water depth. The inviscid horizontal water-particle velocity component at z = 0 is thence given by (2.23) Substitution of (2.23) into expressions (2.14), (2.15) and (2.20) yields the following horizontal and vertical water-particle velocity components in the boundary layer: ( 2.25) These expressions are used in the next section to solve the temperature distribution in the boundary-layer region.

Convection and diffusion of temperature in the boundary layer
We introduce the following expansion for the temperature T : where τ = 2 t denotes the slow time scale of diffusion effects. Substitution of (2.26) in (2.5) yields a sequence of boundary-value problems at different orders. The leading-order problem yields indicating that the temperature at O(1) is a function of spatial coordinates and slow time scale only, i.e. T 1 (x , z , τ ). The governing equation at second order O( ) expressed in physical variables is given by representing the convective effect of the oscillatory flow, where the terms (u 1 , w 1 ) are given by (2.14)-(2.15). Note that the term reappears here because of our use of physical variables. The water-particle velocity components are harmonic in time, and so the response T 2 can be written as in which U 0 is given by (2.23). Hence, the component T 2 oscillates periodically in time with the frequency of the propagating waves ω. Note that the corresponding slowly varying amplitude T 1 is an unknown yet to be determined. Moving to third order O( 2 ), we obtain the following non-dimensional governing equation: Averaging over a wave period yields an equation for the slow evolution of the leading-order temperature T 1 which, in physical variables, takes the simplified form where T w is the ambient water temperature at a large distance from the seabed. We then consider the steady-state configuration ∂T R /∂t = 0 and apply the following boundary conditions: which indicate that the temperature matches the bed temperature T b > 0 at the seabed z = 0, and the ambient water temperature T w > 0 far from the boundary layer, respectively. For simplicity we will denote by T = T b − T w the temperature difference between bed and ambient water. The solution to the boundary-value problem can be found numerically by applying a standard finite-difference scheme. However, the properties of the steady solution of (2.31) are first investigated analytically by considering the behaviour of the water-particle velocity field (2.32) close to the seabed. By Taylor-expandingũ about z → 0 and retaining terms up to first order, we obtain The form of (2.35a) implies the following similarity solution (White 1991;Landau & Lifshitz 1989): where Θ(γ ) is a normalised temperature. Substitution of (2.36a,b) in (2.35a) yields the following boundary-value problem in the new variable γ : (2.39) The ordinary differential equation (2.37) is simpler than the parabolic equation given by (2.35a). The corresponding solution is are the exponential integral function and the gamma function, respectively. Substitution of the similarity variables (2.36a,b) into (2.40) gives the final solution for the temperature field: (2.42) Solution (2.42) is similar to that found for heat transfer from a flat plate by (White 1991, chap. 4-3.2). Once the fluid temperature is known, the seabed heat transfer can be evaluated from Fourier's law as follows: where κ = χρc p is thermal conductivity and c p is specific heat at constant pressure. By integrating (2.43) along x from 0 up to a finite distance L, we obtain the total heat flux exchanged between seabed of length L and the overlying seawater as (2.44) Equation (2.44) elucidates the influence of the length of interest L and wave-field characteristics inherently expressed byũ 0 . For a fixed value of L, maximum heat exchange occurs whenũ 0 is maximised. To investigate the behaviour ofũ 0 with wave frequency ω and hence find an approximate location for its maximum, we first Taylor-expand about k → 0 both the denominator ofũ 0 in (2.35a,b) and the dispersion relation (2.22b) to obtaiñ This procedure allows us to obtain an explicit location of the maximum without resorting to numerical methods. Differentiating (2.45), and equating the result to zero, we find the maximum as Thus, as the water depth increases the maximum heat flux is observed for longer waves. It is surprising that this maximum does not increase monotonically with wavelength, but rather is obtained for a fixed value of the wave frequency ω. This unexpected result is validated in the next section, which is devoted to practical application of the theoretical model.

Results and discussion
The foregoing theory will now be applied to a practical example of waves in intermediate depth. Special care must be taken in choosing the wave parameters for application to laminar boundary layers. Several theories, numerical methods and experiments (Jonsson 1966;Blondeaux & Vittori 1994;Verzicco & Vittori 1996;Vittori & Verzicco 1998) have derived different criteria to establish the development of turbulence. For the Reynolds number in the boundary layer defined as R δ = Aωδ/(ν sinh kh) and a smooth seabed, the aforementioned works indicate the onset of turbulence around R δ ∼ O(10 3 ).
To this end, let us consider a regular field of free-surface waves of amplitude A = 0.3 m and frequency ω = 1 rad s −1 , travelling in the positive x-direction over water of intermediate depth, h = 5 m. The idealised seabed is assumed flat and without roughness, which is of fundamental importance in triggering turbulence in the Stokes boundary layer. The Reynolds number in this case is R δ ∼ 490, hence the assumption of laminar flow is justified.
We now explore the effect of free-surface waves on temperature transport near the seabed. We assume for simplicity that water is pure and at an ambient temperature of T w = 10 • C, whereas the seabed temperature is T b = 20 • C. In the boundary layer we take fixed values of thermometric conductivity χ = 1.4 × 10 −7 m 2 s, specific heat at constant pressure c p = 4.18 × 10 3 J kg −1 • C and kinematic viscosity ν = 10 −6 m 2 s −1 . Figure 1(a) depicts the depth profile of the horizontal velocity component in the boundary layer, which constitutes the crucial convective term (2.32) in (2.31) for the leading-order temperature. Figure 1(b) shows the temperature field relative to the ambient water temperature, obtained by applying a finite difference scheme to the governing equation (2.31) and associated boundary conditions (2.33)-(2.34). Details of the numerical scheme are given in Appendix A. Figure 1(b) shows that the thermal boundary-layer thickness -defined as the point at which the fluid temperature is 99 % of the ambient water temperature T w -grows to approximately 2 cm after x ∼ 10 m. The growth of the boundary-layer thickness δ T is initially very rapid, but slows as x 0. This behaviour is also predicted by the analytical expression (2.42) which additionally provides several physical insights. For example, let us now consider the heat flux at the seabed. Figure 2(a) shows the behaviour of q with horizontal distance x. The solid line represents the heat flux predicted by the numerical model based on the full governing equation (2.31), whereas the dashed line represents the analytical approximation (2.43). The latter successfully captures the behaviour of the heat flux, even though the analysis solely considers the first term in the Taylor expansion for the velocityũ. The heat flux tends to infinity like x −1/3 as x approaches 0 + . This is due to the instantaneous increment of the temperature field with z at x = 0 + , as also shown in figure 1(b). We now compare the numerical solution of (2.31) shown in figure 1(b) to a reference case of interest. Let us consider the flat-plate heat transfer process with velocity field described by the Blasius solution as in White (1991), with velocity at infinity equal to |U|. numerical solver of the full convection-diffusion equation, whereas figure 3(b) shows the corresponding results from the analytical approximation (2.44). The two sets of results agree with each other qualitatively, indicating that the analytical model can be adopted for practical evaluations. Even so, it should be noted that the analytical model underpredicts heat flux, primarily because it underestimates the magnitude of the forcing termũ. It is interesting to observe the presence of a maximum heat flux in each case. As the water depth h increases, this maximum heat flux occurs at progressively lower wave frequencies and longer wavelengths. This behaviour is captured explicitly by (2.46), which highlights the dependence on h −1/2 . Note also that all the curves tend to Q = 0 at large values of wave frequency. In such cases, the termũ 0 becomes very small and the governing equation (2.35a) can be approximated by ∂ 2 T 1 /∂z 2 ∼ 0. In other words, we require ∂T 1 /∂z ∼ 0 to satisfy the condition at z → ∞, so that heat transfer is absent. Therefore, short waves over moderate sea depths do not contribute to heat transfer at the seabed boundary layer.

Conclusions
We have investigated the mechanism of heat transfer in the boundary-layer region at the seabed. Using multiple-scale analysis and a perturbation approach, we first find the velocity field close to the seabed, then solve the governing convection-diffusion equation for fluid temperature. Given the small thermometric conductivity of water, large temperature gradients occur in the region close to the ocean bed. This simplifies the problem, enabling us to elucidate the effects of convection and diffusion at different orders, and to find the corresponding evolution equation for temperature.
We have also found an analytical expression for the temperature field based on an approximation of the streaming current in the bed boundary layer. The theoretical results shed light on the effect of surface waves on near-bed thermal transport processes. Specifically, they suggest that, rather than increasing monotonically with wavelength and phase speed, maximum heat transfer occurs at a finite value of wavelength which depends on the water depth. Given the spatial complexity of the velocity field, good agreement is achieved between the approximated analytical solution and predictions from a full numerical model based on a finite-difference scheme.
Our results provide a theoretical foundation for further applications to topics of considerable current interest, including biofouling (Vinagre et al. 2020), coral bleaching (Monismith 2007), cooling of underwater data centres (Cutler et al. 2017), calibration of satellite data (Donlon et al. 2002) and the emerging area of seawater air-conditioning solutions (Hunt, Byers & Sánchez 2019). For many of these applications, effects arising from turbulence, variable topography, more complex free-surface wave fields, Coriolis acceleration, internal waves, seabed mobility (e.g. ripples and dunes), etc. must be taken into account. These phenomena considerably complicate ocean heat-transfer mechanisms near the seabed, and would provide fruitful opportunities to extend the results discussed herein.