1. Introduction
The subsurface bubble clouds produced by breaking waves are a powerful and persuasive indicator of the presence of processes of transport from the atmosphere to the upper ocean. They are important in near-surface optics (Stramski and Tegowski 2001; Terrill et al. 2001), acoustic propagation and ambient noise in the sea (Farmer and Lemon 1984) and particularly in air–sea gas flux. It is known from observations—for example, those reported by Thorpe and Hall (1983), Thorpe (1984c), Farmer and Li (1995), and Thorpe et al. (2003)—that bubble bands form within the downwelling regions produced by Langmuir circulation (hereinafter referred to as Lc). Zedel and Farmer (1991) demonstrate that the most intense and deepest-going bubble clouds appear within these bands, while Thorpe et al. (2003) show that the rate of dissipation of turbulent kinetic energy per unit mass ε is, on average, enhanced within the bands. There has, however, been no simultaneous measurements reported of ε and bubble void fraction or size distribution. Such measurements made using an autonomous underwater vehicle (AUV), Autosub, are described in section 2.
Bubbles form an important route for air–sea gas transfer but, as explained in section 3, models are presently required to assess the transfer rates. Several assumptions concerning bubbles and their dynamics are commonly made in devising these models. Their validity is tested in section 4 using the data described in section 2. As explained in section 5, it is usual for the models to represent the dispersive effects of small-scale turbulence by a constant eddy diffusivity coefficient and to adopt a simple “one cell size” representation of Lc. These aspects of the models are examined at length in sections 7–9, in particular the effects of Lc in promoting vertical diffusion and the effects of including a representation of the commonly observed hierarchy of cell sizes.
2. Observations of bubbles and Lc
Concurrent measurements of the rate of dissipation of turbulent kinetic energy per unit mass ε, void fractions υf, and bubble size distributions have been made using the AUV Autosub running along constant depth “legs” around a 5 km square at depths of 1.5–15 times the significant wave height Hs. Measurements described below are in fairly steady winds of about 11.5 m s−1 with a fetch of 20 km and wave ages of 11.7–17.2. The location, the environmental conditions, and the methods used to determine ε are described by Thorpe et al. (2003, mission1). Turbulence is measured from shear probes mounted on the front of the vehicle, and void fraction υf and bubble size distribution are measured from a bubble resonator (Farmer et al. 1998) mounted on top of the AUV. Estimates of ε and υf are generally made at 1-s intervals. The void fraction υf is measured 0.6 m above and 5.84 m behind the turbulence sensors. In Figs. 1–3, the horizontal offset is corrected from the time record using the mean AUV speed through the water of 1.25 m s−1 and the turbulence probe depth is used to specify depth z. Individual estimates of void fractions and bubble size distributions have a random uncertainty of a factor of about 2 but a mean calibration uncertainty of 10%–20%. Owing to interference between the AUV's ADCP and the resonator, the threshold of reliable estimates of υf is 10−8 and of bubble radii is about 25 μm.
It is found in laboratory experiments that as much as 40% of the energy lost by breakers may go into the production of bubble clouds (Lamarre and Melville 1991, 1994). Some 90% of the turbulent energy generated is dissipated within four wave periods after breaking whilst it penetrates to depths of (1.25 ± 0.25)H for spilling breakers or (1.75 ± 0.45)H for plunging breakers. Here H is the height of the breaking wave (Rapp and Melville 1990). (Rotors formed in wave breaking and which may trap bubbles, reach only to depths of about H, Melville et al. 2002). Breaker heights observed at sea are typically 30% less than the significant wave height Hs (Gemmrich and Farmer 1999b) so that the direct injection of turbulence by even plunging breakers reaches depths of no more than about 1.5Hs. Although in the laboratory experiments waves break in a static fluid without the preexisting turbulence that is present in the upper ocean, these results suggest that the Autosub observations are beyond the direct influence of the breaking waves and that bubbles are carried to the operational depth by turbulent diffusion or by advection in Lc rather than by the downward jetlike flow induced by the falling crest of plunging breakers.
This conclusion is supported by the finding that the mean rate of turbulent dissipation measured from Autosub is consistent with the law of the wall relation, ε =
The largest values of ε at this depth are εmax ≈ 3.2 × 10−5 m2 s−3 and the greatest value of υf is υfmax ≈ 2.5 × 10−6. The pdf of υf measured at W10 = 11.6 m s−1 decreases steadily as υf increases from its threshold value of 10−8. At 2.2-m depth, 5% of the full record of 28 204 values exceed 8.1 × 10−8 and 1% exceed 2.8 × 10−7. Corresponding 5% and 1% values for a record with 28 084 values at 4.2 m are 2.4 × 10−8 and 6.2 × 10−8, respectively.
Much larger instantaneous values of υf are found at shallower depths shortly after waves have broken (Lamarre and Melville 1991; Farmer and Gemmrich 1996). The mean υf decreases approximately exponentially with depth with decay scales of order 1 m (Farmer et al. 1998). Mean values of υf measured at z/Hs = 2.24 at the location of bubble clouds left by breaking waves show a gradual increase in time of about 14% between 30 and 70 s after wave breaking, presumably as a consequence of vertical dispersion and advection of bubbles in Lc. This rise in υf accompanies the 50% increase in ε between about 30 and 50 s after breaking, followed by a reduction over the next 20 s, reported by Thorpe et al. (2003).
The related mean horizontal structure of ε and υf are shown in Fig. 2, conditionally sampled average sections of ε and log(mean υf) across (Fig. 2a) bubble clouds (regions where υf > 2 × 10−8) at a depth of 4.04 m (z/Hs = 4.21) and (Fig. 2b) Lc bubble bands at two depths, 2.1 m (z/Hs = 1.92) and (Fig. 2c) 3.9 m (z/Hs = 4.35).1 The wind speeds in Figs. 2a–c are 11.7, 10.6, and 11.9 m s−1, respectively. Higher than average values of ε and υf are observed near the centre of the bubble clouds or the bands which are located at time t = 0. Vertical displacements of the Autosub indicate mean downward motions of typically 0.5 cm s−1 in bubble clouds and 2–6 cm s−1 in Lc bands. It appears that, as argued by Zedel and Farmer (1991), most, if not all, the bubble clouds at the observed depths are a consequence of downwelling in Lc. Because of the variable nature of Lc and the broken cloud structure within the bands (resulting partly from the intermittent supply of bubbles from the random field of breakers), it is probable that many of the clouds contributing to Fig. 2a are not recognized as Lc bands even by the rigorous method of objective identification.
Figure 3 shows bubble size distributions at the center of 30 Lc bands at (top) 2.1 m and (bottom) 3.9 m. Figure 3a shows the bubble number concentration, n(a)da (the number of bubbles per cubic meter with radii between a − 0.5 μm and a + 0.5 μm) averaged at the centers of all the bands at (top) 2.1- and (bottom) 3.9-m depth. For comparison, the dashed line shows the much smaller bubble concentrations at ±10 m from the center of the bands. Concentrations are smaller at greater depth. At the shallower depth there appears to be a peak at a radius of about 30 μm with corresponding concentrations of 1.8 × 104 bubbles/1-μm radius band/cubic meter. Figure 3b shows log(a3n(a)da), where a3n(a)da is proportional to the void fraction per unit radius band averaged over the Lc bands, with the dashed lines showing values at ±10 m for reference. Bubbles contributing most to the void fraction are of small radii, mostly less than about 160 μm. It is found by plotting separate curves of log[a3n(a)da] versus a at the center of Lc bands that the radius at which the void fraction is greatest decreases as the overall void fraction decreases. This is consistent with Lc cells that have larger vertical downwelling speeds carrying more large bubbles and a consequently producing a greater void fraction.
3. Models of gas flux and Lc
Bubbles transfer their constituent gases into the water by dissolution. For some gases at high enough wind speeds (typically those exceeding the global ocean average surface wind speed of about 7.8 m s−1), bubbles appear to provide an efficient and possibly dominant mechanism of air–sea gas transfer (Woolf and Thorpe 1991; Schudlich and Emerson 1996) although, largely because of the existing uncertainty in quantifying their effect and relating it to forcing factors (e.g., wind speed), this route of gas transfer is not always represented in basinwide or global estimates of air–sea gas flux (e.g., see Garcia and Keeling 2001). While it is now possible to measure the bubble size distribution in breakers (Deane and Stokes 2002) and as a function of depth below the sea surface (Farmer et al. 1998, 1999), this information about bubbles is insufficient to allow the direct estimation of their gas transfer into the surrounding water, even when supplemented by measures of the gas saturation in the bubble-containing water. In being diffused downward from the sea surface by turbulent motions, including Lc, the gases within bubbles pass into solution at rates that depend on the individual gases, their partial pressures in bubbles and their saturation within the surrounding water, as well as on the bubble radii and on the degree to which the surface of bubbles may be covered by surfactants. Consequently, at depth, the relative concentration of gases within bubbles is no longer that in the air above the ocean from which they originated. The mean partial pressures of the gases within bubbles are unknown and it is not yet possible to measure them at sea. The gas flux cannot therefore be estimated directly from observations. Unless or until measurements of gas composition of bubbles in the sea becomes feasible or laboratory experiments can be conducted that fully represent the wave breaking and the bubble and turbulence production that occurs at sea, resort must be made to the use of models of bubble transport and dissolution, constrained by available measurements, to estimate the in situ gas composition of bubbles and hence the bubble contribution to air–sea gas flux.
A variety of dynamical models have been devised to estimate the gas flux from bubbles created by breaking waves. Models include the compression of bubbles, the dissolution of their component gases as they are subjected to increased hydrostatic pressure, and their buoyant rise through the water, factors identified by Garrettson (1973) as being important in describing bubble dynamics. Where models vary and where there remains considerable uncertainty are in the representation of the field of motion that disperses bubbles from their sources at the water surface (see section 5). Other areas of great uncertainty are the bubble size distribution generated by breakers and the frequency and energetics of breaking across the range of sizes of breaking waves. Several basic assumptions about bubbles and their surrounding turbulence [including some identified by Thorpe (1986)] underpin the validity of many models of bubble dynamics and their contribution to air–sea gas flux. These are reviewed in the following section in the light of the observations described in section 2.
4. The validity of assumptions in existing models of bubbles
The models referred to in section 3 generally assume that the following effects are negligible.
a. Fragmentation of bubbles
Models assume that bubbles are not shattered by turbulence and can be tracked as single particles. The radius of the smallest bubble, which may be broken up by turbulence, is given by the Hinze scale, aH = c(γ/ρ)3/5ε−2/5, where c is a constant, γ is the surface tension, ρ is the density of the sea water, and ε is the rate of dissipation of turbulent kinetic energy per unit mass. Garrett et al. (2000) estimate c as 0.5, but Deane and Stokes (2002) suggest that this value is applicable in a steady state, such as a jet from a hosepipe entering the water, and that a value of about 0.83 (given by 2−8/5 times a critical Weber number of about 4.7 to the power 3/5) is more appropriate in the transient conditions of breakers. Extreme values, εmax, at z/Hs = 2.22 of about 3.2 × 10−5 m2 s−3 (see Fig. 1) are 5.3 times the law of the wall value at this depth at the observed wind speed of 12.4 m s−1. Taking γ = 7.2 × 10−2 N m−1 for “clean” and 3.6 × 10−2 N m−1 for a “dirty” bubble, one covered by an immovable surfactant film, respectively, ε = εmax, and ρ = 1.03 × 103 kg m−3, we find aH is at least 0.1 m (clean bubbles) or 0.07 m (dirty bubbles), using the lower value c = 0.5 suggested by Garrett et al. These values of aH are much larger than the vast majority of bubble radii observed or inferred at z/Hs = 2.22, even in the regions of bubble concentration in Langmuir bands. Turbulent fragmentation of large volumes of air trapped within spilling or plunging breakers is observed by Deane and Stokes within the high ε breaker regions at shallower depths but is not significant in the disruption of bubbles at the depths considered here.
b. Coalescence of bubbles
c. Distortion of bubbles from spherical shape by turbulence
This affects the rise speeds of the spherical bubbles assumed in the models. Turbulence will induce bubble oscillations that may affect their rise speeds as well as leading to sound generation (e.g., see Lamarre and Melville 1994). Significant distortion will occur if bubble radii are much greater than the smallest characteristic length scale of the turbulence, the Kolmogorov length scale, lk = (ν3/ε)1/4. Taking again the extreme value εmax at z/Hs = 2.22, lk = 420 μm, greater than the majority of bubble radii (Fig. 3a), and implying that at these depths the bubbles remain approximately spherical.
d. Effect of a cloud of bubbles on bubble rise speed
The mean speed of ascent of a layer of bubbles in a cloud differs from that of a single bubble. The mean rise speed is wb(1 − 6.55υf) (Batchelor 1972) where wb is the rise speed of a single spherical bubble in the absence of others. The correction, 6.55υf, is negligible for even the highest void fractions, υfmax, observed at z/Hs = 2.22.
e. Interaction of bubbles and mean flow
These interactions are generally ignored in models. Although Smith (1998) has suggested that vacillations in the strength of Lc may be associated with the reduction in its downwelling by the buoyancy of high bubble concentrations, the maximum mean υf observed in Lc is only about 1.6 × 10−7 at z/Hs = 1.92 and winds of about 10.6 m s−1 (Fig. 2b). While this void fraction is equivalent to the density change caused by a temperature variation of about 1 mK and variations of this size are typical of those found in Lc (Thorpe and Hall 1982; Thorpe et al. 2003), it leads to a drag per unit mass or a fluid acceleration, gυf, of 1.6 × 10−6 m s−2, very much less than that typically encountered by fluid downwelling in Lc cells of order
f. Effects of turbulence on bubble rise speeds
g. Contribution of bubbles to the generation of turbulence
Models assume that the bubbles react passively to turbulence and that turbulence is not generated by rising bubbles. Bubbles may contribute substantially to the motion field if the following situations occur:
- They shed eddies with a scale and with vorticity comparable to those within the ambient turbulence, thereby contributing to the turbulent motion. Taking a = 250 μm and wb = 4.2 cm s−2 for bubbles covered with surfactants which immobilize their surface, to characterize the largest (and few) observed bubbles and to give extreme values, the bubbles' Reynolds number Re is about 10 and the Weber number, We = wb(ρa/γ)1/2, is 0.11. These values are much less than the critical values, Re = 202 and We = 1.26 at which bubbles begin to oscillate as they rise (Hartunian and Sears 1957). While flow may separate and form vortex rings which are stationary relative to the rising bubbles, they are unlikely to be shed or detached from the bubbles except possibly in the presence of impurities (e.g., microscopic particles) in the water. The scale of attached wake eddies is comparable to that of the bubble radii, a, and by section 4c, is consequently generally less than the Kolmogorov scale. Although Re is small, there may however be interactions between rising bubbles which may promote the detachment of bubble wake eddies or lead to random motions, which may contribute to turbulence. The vorticity of shed eddies is of order of the bubbles' rise speed, wb, divided by a. This is about 170 s−1 for the scales considered, exceeding the vorticity in turbulent eddies of order (ε/ν)1/2 ≈ 5.7 s−1. In view of the relatively rare close approach of bubbles (see section 4b), interactions are infrequent and the contribution of such eddies to turbulence appears negligible at the depths considered here.
- They are substantially distorted by the turbulent motion, consequently being a sink of turbulent energy (section 4c indicates this effect is insignificant).
h. Effect of turbulence on gas transfer rates from single bubbles
A further assumption made in many models is that a steady state exists with a uniform distribution of sources. This is a more difficult assumption to justify, although waves appear to break randomly and not preferentially at a position of earlier breaking (Thorpe et al. 2003) and the location of breaking does not appear to be affected by windrows (Thorpe 1992). The assumption might be better avoided by devising models in which bubbles are injected from breakers that conform to observations (e.g., that have the same recurrence period at a point, that inject bubbles randomly into an underlying Lc pattern, and that produce bubbles at a rate representing those in real breakers at specified wind speeds).
The dependence of the calculations made above on the rate of turbulent dissipation emphasizes the need to obtain further measurements of dissipation rates and bubble size distribution, especially within the breaker zone closer than 1 significant wave height from the sea surface. Without such observations we are left with the option of constructing a model in which bubbles are injected at some specified level near the sea surface with atmospheric gas composition, but without representation of the uncertain factors described above.
5. Kz and Lc in models
Thorpe (1982) used the measured variation of acoustic scattering cross section with depth and wind speed to derive estimates of the flux of oxygen and its variation with gas saturation in the water. The estimates were revised by Thorpe (1984d). While this method does account, in principle, for the wide variations in bubble clouds caused by breakers by using real data to represent their natural variability, it cannot account for the variation of the gas composition within bubbles as gases pass at different rates from bubbles during their lifetimes. It is consequently flawed. Thorpe (1984a) therefore devised a turbulent diffusion model based on a Monte Carlo simulation of dispersion from a near-surface source of bubbles containing two gases, O2 and N2, injected into the sea at atmospheric concentrations. In the absence of reliable data to guide in the selection of better alternatives, the effective vertical diffusion coefficient of momentum, the eddy viscosity, was chosen to be constant. Available optical measurements of bubble size distribution (Johnson and Cooke 1979) were used to constrain the models. It was assumed in some models (e.g., Thorpe 1982, 1984b, 1986), that the eddy diffusivity of bubbles was the same as that of momentum, that is, equal to eddy viscosity. As demonstrated in sections 8 and 9, this assumption is erroneous. (In the atmospheric literature it is usual to introduce a correction factor expressing the ratio of the two eddy diffusion coefficients.)
Models of a wide range of suspensions, including sediment over the seabed or blowing snow (e.g., Taylor et al. 2003), use an eddy-diffusion-based representation of turbulence. Results are in fairly good agreement with observations without a requirement for any specific representation of vortices or eddies in the flow. It may consequently appear sufficient to represent the dispersion of bubbles in the upper ocean using a bubble dispersion model based on a Monte Carlo method, perhaps constructed to give a law of the wall eddy viscosity, Kz = ku∗wz. This, however, fails to represent the presence of Lc and its known tendency to trap bubbles (Stommel 1949). The agreement between eddy diffusion models and observations for other suspensions may be because the flow field at the distances from the boundary, where comparison is made is not dominated by large eddies that contain a substantial concentration of suspended matter, or because the turbulent flow differs in its basic structure from that in the upper ocean. Langmuir circulation is generally thought to be generated and driven by the CL2 mechanism (e.g., see Leibovich 1983; Li and Garrett 1995), which depends on a vortex force resulting from the presence of shear flow and the Stokes drift induced by wind waves. This large eddy structure of the upper ocean may be absent in other boundary layer flows where the particular mechanisms forcing Lc are absent and where eddy diffusion based models are therefore more successful. It follows that, in modeling the dispersion of bubbles, some realistic representation is required of the nature of the vortical motions, the “large eddies” below the water surface that may contribute to bubble trapping or maintenance within the water column, particularly of coherent motions in which the largest downward flow is greater than the bubble rise speed, typically about 0.5–4 cm s−1.
The advective effect of a single-cell-sized representation of Lc was included in a bubble diffusion model by Thorpe (1984c). This model was later used to investigate the effects of temperature, gas saturation, and particulate concentrations on bubble size distribution (Thorpe et al. 1992), and developed to include four gases and to investigate the effect of varying the velocity in Langmuir cells on gas transfer (Woolf and Thorpe 1991). It was subsequently used by Farmer et al. (1999) with a modified representation of the flow within the Langmuir cell to predict bubble size distributions for comparison with those measured using a bubble resonator. Gemmrich and Farmer (1999a) devised a model including a one-size-cell representation of Lc to explain their observations of near-surface temperature fluctuations in winds of 12–16 m s−1 and significant wave heights of about 4.5 m.
Simple examples of how the effective vertical diffusivity of bubbles is affected by Lc are given in section 8. Consideration is given to the combined effect of cells of different sizes in section 9.
In addition to the problem of appropriately representing Lc in models of gas transfer, there is a major problem of parameterizing turbulence and bubbles close to the sea surface where breaking waves are a critical component, both in injecting bubbles and in generating turbulence. Agrawal et al. (1992) and Terray et al. (1996) find that at depths, z, less than the significant wave height, Hs, the rates of dissipation of turbulent kinetic energy per unit mass, ε, exceed the law of the wall relationship, ε =
6. The effect of Langmuir circulation on submerged bubbles
Langmuir circulation will be particularly effective in bubble transport when the depth to which bubbles are driven by breaking waves exceeds the depth at which the downwelling motions in Lc first exceed the rise speed typical of bubbles, wb. In these conditions bubbles are advected downwards by the circulation and may be captured in a subsurface recirculating region. The discussion in section 2 suggests that the maximum depth to which turbulence and consequently bubbles are carried through the direct action of wave breaking is about 1.5Hs. The largest downwelling motions observed by Weller et al. (1985) are typically 10–20 cm s−1 and occur near the middle of the mixed layer. Scaling of Lc speeds with wind or other velocity scales is poorly known. Smith (1999) finds a scaling of surface currents in Lc with Stokes drift, but with other presently unknown factors affecting the constant of proportionality. D'Asaro (2001) finds that the rms vertical speeds in the mixing layer (excluding those directly induced by waves) are some 1.75–2.0 times greater than those found in turbulent flow near a rigid wall, with maximum values of 1.41u∗w at depth of about 0.2 times the mixed layer depth. This appears to differ from Weller et al.'s finding of largest motions at middepth but may be a consequence of (i) vacillations in the strength or presence of Lc and processes associated with cell amalgamation or break down or (ii) a broadband scaling of Langmuir cells, leading to enhanced motions caused by the relatively small cells that affect mainly the shallower layers. We consider the effects of the latter in section 9.
7. Modeling the effects of Lc
Two eddy diffusion coefficients are defined in the models that represent dispersion of bubbles or, as below, of particles with local concentration, C(x, z, t), which has bubblelike properties. The first is Kz, the eddy viscosity, ascribed to momentum transport by small-scale turbulent motions, motions that are of scale much less than those of Lc and are uncorrelated to Lc. The second is the effective diffusivity Kp produced by the combined effects of small-scale turbulence and Lc. We assume here that turbulence is of much smaller scale than the Lc cells and that the two are independent, but stress that our observations showing enhanced turbulence in particular locations within Lc cells (e.g., Fig. 2) suggest that this assumption is not robust. Attention is focused on two questions:
- How much does Lc change the effective vertical eddy diffusivity Kp?
- What is the effect of multiple cells?
The horizontally averaged concentration of particles, C(z, t), is determined at wind speeds of 6–14 m s−1, after 96 particles of unit concentration have been added at each 1 s time step randomly across the domain width of depth 1.5Hs representing that within which bubbles are injected by breaking waves. The model was run until a steady state was reached, judged by steady concentration, C, values and equality (to better than 10%) between the numbers of particles injected and those reaching the surface per time step. This required 800–2400 time steps, the latter in the higher winds. Values of uLc, Hs, and l at the wind speeds of 6–14 m s−1 of model runs are given in Table 1.
8. Model results: Single cells
Figure 4 shows results when Kz is given by (14). The full lines show steady-state values of logC as a function of z/l at various wind speeds with (full line) uLc set to zero (i.e., without Lc) and (crosses) with uLc given by (5). The bottom of the particle injection region is marked by an arrow. Below this the solutions for C without Lc decay exponentially. A similar decay is seen at depths below the bottom of the Langmuir cells at z/l = 1. The main effect of the circulation is to increase the concentration below the injection region and to produce a region near the center of the cell where the concentration is uniform or increases with depth leading to negative particle diffusivity, Kp. In consequence, values of r in this region are large or negative. (Negative values where the concentration gradient becomes positive are not shown in Fig. 4.) For comparison, Fig. 5 shows similar plots when Kz is given by (15). The concentration is again enhanced by the circulation but decays gradually in runs with or without Lc; as in observations and in contrast to Fig. 4, no uniform concentration layer is found. The location of the foot of the Langmuir cell is not evident in the concentration profiles, and the maximum values of r are typically about 2, much lower than in Fig. 4.
Tables 2 and 3 give parameter values and model estimates of the mean times for which injected particles take to reach the surface (the particle lifetimes), the standard deviation of these times Σt, and the mean concentrations of particles reaching the surface, once a steady state is established. Corresponding values with no eddy diffusivity and no Lc, only particle rise, are given in Table 1. Lifetimes are increased and concentrations decreased with the addition of Lc. Lifetimes increase with increasing wind speed, with a commensurate decrease in the concentrations. The relative values with and without Lc provide an indication of the effect of Lc in retaining submerged bubbles and so providing longer periods of time for gas diffusion to occur. When W10 ≥ 8 m s1, values of Σt exceed the mean bubble lifetimes, indicative of a pdf that is positively skewed, many particles surfacing rapidly after introduction but a number being submerged for relatively long periods by Lc. The ratio of Σt to the mean bubble lifetimes is greatest (about 1.5) at W10 ≈ 8 m s−1, close to the value where Lc begins to subduct injected bubbles (section 6). There is no substantial difference in the variation of the ratio, s, for the two forms of Kz examined, nor significant variation with W10. In each case s exceeds unity over the major extent of the cell depth, with values of typically 2.5 at middepth, indicating the strong influence, but not overwhelming domination, of the circulation in driving the vertical transport of particles. Values of θ in Table 2 (Kz uniform in depth) are large, inferring that Lc dominates over eddy diffusivity, while φ and q decrease, but are of order unity, implying the particle rise and decay are important in comparison with eddy diffusion (but still less so than Lc). Parameters θ (proportional to uLc/u∗w) and q shown in Table 3 are calculated for values of Kz given by (15) at the center of the cell, Kz = ku∗wl/2. Values of θ are of order unity, so here eddy diffusion, Lc, and particle decay are of comparable importance, but φ (proportional to wb/u∗w and a measure of particle rise) is relatively small.
Model runs in which the depth of injection of particles was halved show similar trends but with concentrations at depths l/2 reduced by a factor slightly more than 2, and with a reduction in the mean particle lifetimes.
9. Model results: Multiple cells
So far, effects of the so-called observed hierarchy of cell sizes have been ignored in modeling or only included, by inference, within the effects of eddy diffusivity. Estimation of the effect of smaller cells requires quantitative information about how their vertical velocities and penetration depths relate to their horizontal size. These sizes appear to have a lognormal distribution, symptomatic of turbulence (Csanady 1994). The smaller cells may
- enhance the diffusivity in the near-surface layer;
- augment the downward velocities in larger cells if they are coincident in position, so reducing the depths at which the downward speed equals the rise speeds of bubbles in the larger cells, or advect bubbles to depths at which the vertical velocities in large cells exceeds bubble rise speeds, so effectively increasing the parameter R characterizing the effectiveness of cells in vertical transport; or
- reduce the vertical rise speeds of bubbles as they, and the smaller cells, are advected toward downwelling regions in larger cells, so enhancing the number of bubbles supplied to the deeper downwelling region of larger cells.
It is also conceivable that the small cells or vortices generated by breakers are bodily subducted by much larger cells, so contributing to subsurface turbulence, but this is not investigated here.
The effect of multiple cell sizes has been examined by including in the model
- nonoverlapping cells with scales l and l/2 and speeds uLc and uLc/2, respectively, but with the same total number of bubbles as in the single cell case (this is equivalent to there being twice as many cells of half scale);
- superimposed cells of scales l and l/2 with speeds uLc and uLc/2, respectively [in this case the downward flows of both cells are located at the same position, so enhancing the local downward flow (this is equivalent to conditions in which the smaller cells are advected into the convergence regions of the larger cells)]; or
- nonoverlapping cells with scales 5l/6, 3l/4, and l/2 with speeds 5uLc/6, 3uLc/4, and uLc/2, respectively, close to those of the primary cell, l and uLc, but with the same total number of bubbles as in the single cell case [model runs (i) and (ii) were also made with third sets of nonoverlapping or superimposed cells of size and speed equal to one-third of the largest].
Results for wind speeds of 10 m s−1 are shown in Figs. 6 and 7 for Kz given by (14) and (15), and with supporting data given in Tables 4 and 5, respectively. The addition of more nonoverlapping cells, as in (i) above (Figs. 6a, 7a), leads to relatively fewer particles being carried by Lc to depth, reducing the concentration and ratio, r. The superposition of cells, as in (b) above (Figs. 6b, 7b), strengthens the downwelling and more particles are transported downwards, enhancing the concentration and r, particularly at depth z < l/2, where ratios approaching 10 are found. Addition of the third cells makes relatively little difference. For the depth-uniform eddy diffusion coefficients (Fig. 6), the thickness of the region of near-uniform or inverted concentration is increased, but the “law of the wall” profiles of Kz (Fig. 7) have concentrations that decay more nearly exponentially with depth. Nonoverlapping cells slightly reduce bubble lifetimes and their standard deviation Σt (Tables 4 and 5), while superimposed cells increase lifetimes and Σt, and reduce the concentration of surfacing particles, implying an enhancement of the gas flux.
For comparison, Fig. 8 shows the results of including nonoverlapping cells of similar sizes as in (iii) in winds of 10 m s−1, with corresponding lifetimes and concentrations given in Table 6. The effect is to removed the depth-range of near-uniform mean concentration and of very high r when Kz is uniform in depth and to produce a concentration profile similar to that of a law of the wall eddy diffusivity. Values of s reach 1.5 for Kz given by (15), lower than those of 2.3 when Kz is given by (14).
10. Discussion
a. Model assumptions tested against observations
Concurrent observations of ε and υf near the sea surface are presented in section 2 and are used in section 4 to test the assumptions commonly made in models of subsurface bubbles. In the wind speeds and wave conditions of the observations, these appear to be valid at depths greater than about 1.5Hs. It is however important to note that, even at this depth some of the assumptions are only weakly satisfied and that all these assumptions may, and some clearly will, be violated in a more intense turbulence region described by Agrawal et al. (1992) and Terray et al. (1996) at depths less than 1.5Hs. This is especially so in the proximity of breaking waves where enhanced turbulence will fragment bubbles, lead to collisions and perhaps coalescence, distort the bubbles leading to changes in rise speeds through the water and affect the rates of dissolution. Models designed to describe bubble gas flux in the immediate vicinity of breakers will have to account more carefully for the effects of turbulence on bubble dispersion, of large rising bubbles in producing turbulence, as well as other properties mentioned above, than does the present generation of models.
b. Langmuir circulation and algae
As well as leading to accumulation of bubbles, Lc causes increased concentrations of mobile or buoyant algae (e.g., see Bainbridge 1957; Smayda and Reynolds 2001). Turbulence affects the supply of nutrients to algae, may increase predator–prey contact and causes damage to flagellates (for review see Estrada and Berdalet 1998). For example, the growth of the red-tide dinoflagellate,
c. The effective diffusivity and mushroom clouds
Substantial increase in C occurs at mid–cell depth when the wind reaches about 8 m s−1 (see Figs. 4b, 5b). As seen by comparing C at z/l = 0.5 in Figs. 4c–e and 5c–e at winds of 10–14 m s−1, the concentrations level off as the wind increases from 10 to 14 m s−1, perhaps reaching “saturation” as the effects of Lc dominate the process of vertical transfer The largest increase in concentration occurs near the wind speeds that are in general accord with the condition, R = 1, in section 6 at which bubbles are injected to levels where vertical advection overcomes their tendency to rise. This simplistic estimate is however strictly valid only when Kz is negligible and takes no account of the downward vertical diffusion of particles as they are being advected by Lc towards the center of the downwelling region. The concentration profiles with no Lc (full lines in Figs. 4 and 5) show the effect of this diffusion may be substantial.
The concentration profiles with Kz given by (14) show increasing levels below the middepth of the cells (Figs. 4 and 6). This is a consequence of the vertical acceleration and deceleration of particles in Lc, a “log jam” effect occurring in horizontally averaged concentration profiles below middepth and resulting in negative Kp values in (22). The deceleration in vertical motion occurs where the Lc streamlines diverge to produce a horizontal spread of relatively young advected particles, resulting in an “inverted mushroom” shape of concentration contours. Such inverted mushroom-shaped bubble clouds are sometimes seen in sonograph observations using narrow vertically pointing acoustic beams, notably in conditions when convection is favored (e.g., see Thorpe 1982, Figs. 11c and 13b,c). The greater local eddy diffusivity when Kz has a law-of-the-wall behavior spreads the downwelling jet and the particle concentration contours do not show the mushroom shapes.
The estimated values of the ratio, r = Kp/Kz, is not unity, as sometimes supposed, but depends on depth, z, and on the form selected for Kz. When Kz is independent of depth, values of r exceeding 10 or which are negative are found in the center of the cells where the vertical gradient in concentration becomes small or positive. In contrast, when the law of the wall variation of Kz is used, r is typically between 1 and 2, similar to values reported in the literature for sedimenting particles (e.g., see Taylor et al. 2003), and the mean concentration profiles have distributions that do not differ greatly from exponential, in general agreement with observations. The lifetimes of particles (Tables 2 and 3) also depends on the choice of Kz. The simple model shows that prediction of bubble distributions and of gas transfer (which is related to bubble lifetime) is sensitive to the selection of Kz and that a constant value will not give realistic results if cells of only one size are represented.
d. The hierarchy of cell sizes
One assumption common in models of the role of bubble in air–sea gas flux is that Lc can be represented by a regular array of cells of a single specified size (e.g., Thorpe 1982, 1984c; Woolf and Thorpe 1991; Farmer et al. 1999). Neither the hierarchy of cells that are known to occur simultaneously nor the random unstable nature of the circulation are included. The sensitivity of concentration profiles and hence gas transfer to the representation of Lc, in particular to the histogram of cell size and flow speeds, is demonstrated in section 9. The model runs (Figs. 6 and 7) show that the concentration profiles and diffusion ratios depend on the way in which cells coexist, whether small and large cells occur separately on the water surface, or whether they are superimposed. If small cells are generated by breaking waves within a preexisting larger cells [e.g., through the mechanism suggested by Csanady (1994)], then their effect may be to make the bubble concentration gradients more uniform (Fig. 7b), to enhance the eddy diffusion of bubbles, and to increase the bubble lifetimes (Table 4) and consequently the air–sea gas flux. The results are however sensitive to the range of cell sizes which are included. Figure 8 suggests that in a suitable combination of cells and flow speeds, even with Kz constant (or even zero), it may be possible to reproduce results similar to those with a law-of-the-wall Kz dependence.
e. Requirements for future models
LES models (Skyllingstad and Denbo 1995; McWilliams et al. 1997; McWilliams and Sullivan 2000) provide realistic representations of many aspects of Lc, allowing for three-dimensional effects, the unsteady turbulent nature of the circulation, and the development of cells of different size, and they may well prove suitable for future studies of the effect of bubble on air–sea gas flux. They do not at present, however, include a representation of breaking waves, or of their consequent bubble generation and production of vorticity and turbulence (McWilliams and Sullivan 2000). Recent developments of large wave simulations (LWS: see Dimas and Fialkowski 2000) may make the modeling of wave breaking and turbulence and bubble generation feasible in the future. Even if this becomes possible, it will be necessary to have more quantitative information about waves (their frequency of breaking and the bubbles and turbulence which they generate) and about Langmuir cells (particularly of their flow speeds, growth, persistence, and demise and of how, in reality, a hierarchy of cells of different scales is constituted) to tune and test dynamical models to a stage at which they can be regarded as being sufficiently reliable for accurate gas flux estimates or algal modeling. This is a challenging, multiple-parameter problem demanding multisensor observations in a variety of sea and wind states in which AUVs can play a useful part.
11. Summary
Concurrent observations of the rate of dissipation of turbulent kinetic energy and of the void fraction and size distribution of bubbles produced by breaking wind waves have been made using an AUV, Autosub, operating along constant depth “legs” near the sea surface. Regions of relatively high dissipation rates and high void fraction are observed in bubble bands produced by Lc. Data are used to test eight of the basic assumptions underpinning existing models of the subsurface distribution of bubbles and their effect on air–sea gas transfer. Where sufficient data is available, the assumptions appear to be generally valid at depths exceeding 1.5Hs and in wind speeds of about 11.5 m s−1, but may break down at shallow depths.
Langmuir circulation is known to lead to the concentration of mobile or buoyant algae, particularly in the downwelling regions, and the observed mean dissipation rates in these regions of the circulation are found to be close to those at which algal dynamics are significantly affected.
A simple model with parameters adjusted to those appropriate in the range 6 ≤ W10 (m s−1) ≤ 14 is devised to examine the qualitative effect of Lc on the vertical diffusion of bubbles and the representation of Lc in models of gas diffusion. The circulation is particularly effective in vertical bubble transfer when bubbles are injected by breaking waves to depths at which they are carried downward by the circulation against their tendency to rise. The estimated values of the ratio r of the eddy diffusivity of particles (resembling bubbles) Kp to the eddy viscosity Kz depends on depth z and on the form selected for Kz. With Langmuir cells of a single size and when Kz is independent of depth, values of r that exceed 10 are estimated as the depth is approached at which the downwelling speed in the cells is greatest and where the mean vertical concentration gradient is small. Negative values of Kp, and therefore of r, occur at mid cell depths where the vertical gradient in particle concentration is positive even though Kz is positive. In contrast, when the law of the wall variation of Kz is used, values of r are typically between 1 and 2, and the mean concentration profiles have distributions which do not differ greatly from an exponential decay in depth, in general agreement with observations. The effect of nonoverlapping or superimposed Langmuir cells of different size is also investigated, and it is shown that their effects may be quite different. Multiple nonoverlapping cells of similar scales with Kz independent of depth can result in concentration profiles that resemble those of a law of the wall Kz. The periods of time for which particles remain submerged is increased by Lc and depends on the choice of Kυ. The simple model therefore suggests that model prediction of bubble distributions and of gas transfer (which is related to bubble submergence time) is sensitive to Kz and to the size distribution of Langmuir cells.
The paper draws attention to several aspects of upper ocean dynamics for which data are not yet sufficient to construct realistic and quantitative models of transfer processes involving Lc. Many of the assumptions made in models of bubble clouds and their contribution to the air–sea transfer of gases depend on the rate of turbulent dissipation, and this emphasizes the need to obtain further measurements of dissipation rates and bubble size distribution, particularly at depths less than Hs.
The authors are grateful for the help provided by Jeff Jackson in programming and in the analysis of sonar and turbulence data and Peter Chandler in the analysis of the bubble resonator data, and thank the Autosub Team and master and crew of the Tershelling, without whose enthusiastic assistance the data would not have been obtained. Part of this research was undertaken with funding from NERC Grant NER/A/S/2000/00359. TRO was supported by the Office of Naval Research (Grant N000149910088). SAT thanks Prof. Alan Elliot (SOS, University College of North Wales, United Kingdom) for the use of a workstation, and Kate Davis for assistance with figures.
REFERENCES
Agrawal, Y. C., , E. A. Terray, , M. A. Donelan, , P. A. Hwang, , A. J. Williams III, , W. M. Drennan, , K. K. Kahma, , and S. A. Kitagorodskii, 1992: Enhanced dissipation of turbulent kinetic energy beneath surface waves. Nature, 359 , 219–220.
Bainbridge, R., 1957: The size, shape and density of marine phytoplankton concentrations. Cambridge Philos. Soc. Biol. Rev., 91–115.
Batchelor, G. K., 1972: Sedimentation in a dilute dispersion of spheres. J. Fluid Mech., 52 , 245–268.
Batchelor, G. K., 1980: Mass transfer from small particles suspended in turbulent flow. J. Fluid Mech., 98 , 609–623.
Csanady, G. T., 1994: Vortex pair model of Langmuir circulation. J. Mar. Res., 52 , 559–581.
Csanady, G. T., 2001: Air–Sea Interaction. Cambridge University Press, 239 pp.
D'Asaro, E. A., 2001: Turbulent vertical kinetic energy in the ocean mixed layer. J. Phys. Oceanogr., 31 , 3530–3537.
Deane, G. B., , and M. D. Stokes, 2002: Scale dependence of bubble creation mechanisms in breaking waves. Nature, 418 , 839–844.
Dimas, A. A., , and L. T. Fialkowski, 2000: Large-wave simulation (LWS) of free-surface flows developing weak spilling breaking waves. J. Comput. Phys., 159 , 172–196.
Estrada, M., , and E. Berdalet, 1998: Effects of turbulence on phytoplankton. Physiological Ecology of Harmful Algal Blooms, D. M. Anderson et al., Eds., NATO ASI Series, Vol. G41, Springer, 601–618.
Farmer, D. M., , and D. D. Lemon, 1984: The influence of bubbles on ambient noise in the ocean at high wind speeds. J. Phys. Oceanogr., 14 , 1762–1778.
Farmer, D. M., , and M. Li, 1995: Patterns of bubble clouds organized by Langmuir circulation. J. Phys. Oceanogr., 25 , 1425–1440.
Farmer, D. M., , and J. R. Gemmrich, 1996: Measurements of temperature fluctuations in breaking surface waves. J. Phys. Oceanogr., 26 , 816–825.
Farmer, D. M., , S. Vagle, , and A. D. Booth, 1998: A free-flooding acoustical resonator for measurement of bubble size distributions. J. Atmos. Oceanic Technol., 15 , 1132–1146.
Farmer, D. M., , S. Vagle, , and M. Li, 1999: Wave breaking, turbulence and bubble distributions in the ocean surface layer. The Wind-Driven Air–Sea Interface, M. Banner, Ed., School of Mathematics, The University of New South Wales, 187–192.
Garcia, H. E., , and R. F. Keeling, 2001: On the global oxygen anomaly and air–sea flux. J. Geophys. Res., 106 , 31155–31166.
Garrett, C., , M. Li, , and D. M. Farmer, 2000: The connection between bubble size and energy dissipation rates in the upper ocean. J. Phys. Oceanogr., 30 , 2163–2171.
Garrettson, G. A., 1973: Bubble transport theory with application to the upper ocean. J. Fluid Mech., 59 , 187–206.
Geernaert, G. L., 1990: Bulk parametrization for the wind speed and heat fluxes. Surface Waves and Fluxes, G. L. Geernaert and W. J. Plant, Eds., Vol. 1, Kluwer Academic, 91–172.
Gemmrich, J. R., , and D. M. Farmer, 1999a: Near-surface turbulence and thermal structure in a wind-driven sea. J. Phys. Oceanogr., 29 , 480–499.
Gemmrich, J. R., , and D. M. Farmer, 1999b: Observations of the scale and occurrence of breaking surface waves. J. Phys. Oceanogr., 29 , 2595–2606.
Gibson, C. H., , and W. H. Thomas, 1995: Effects of turbulence intermittency on growth inhibition of a red tide dinoflagellate Gonyaulax polyedra Stein. J. Geophys. Res., 100 , 24841–24846.
Hartunian, R. A., , and W. R. Sears, 1957: On the instability of small gas bubbles moving uniformly in various liquids. J. Fluid Mech., 3 , 27–47.
Johnson, B. D., , and R. C. Cooke, 1979: Bubble populations and spectra in coastal waters: A photographic approach. J. Geophys. Res., 84 , 3761–3766.
Lamarre, E., , and W. K. Melville, 1991: Air entrainment and dissipation in breaking waves. Nature, 351 , 469–472.
Lamarre, E., , and W. K. Melville, 1994: Void-fraction measurements and sound-speed fields in bubble plumes generated by breaking waves. J. Acoust. Soc. Amer., 95 , 1317–1329.
Leibovich, S., 1983: The form and dynamics of Langmuir circulation. Annu. Rev. Fluid Mech., 87 , 119–123.
Li, M., , and C. Garrett, 1995: Is Langmuir circulation driven by surface waves or surface cooling? J. Phys. Oceanogr., 25 , 64–76.
McWilliams, J. C., , and P. P. Sullivan, 2000: Vertical mixing by Langmuir circulations. Spill Sci. Technol. Bull., 6 , 225–238.
McWilliams, J. C., , P. P. Sullivan, , and C-H. Moeng, 1997: Langmuir turbulence in the ocean. J. Fluid Mech., 334 , 31–58.
Melville, W. K., , F. Veron, , and C. J. White, 2002: The velocity field under breaking waves: Coherent structures and turbulence. J. Fluid Mech., 454 , 203–234.
Moisander, P. H., , J. L. Henck, , K. Kononen, , and H. W. Paerl, 2002: Small-scale shear effects on heterocystous cyanobacteria. Limnol. Oceanogr., 47 , 108–119.
Rapp, R. J., , and W. K. Melville, 1990: Laboratory measurement of deep-water breaking waves. Philos. Trans. Roy. Soc. London, 331A , 735–800.
Saffman, P. G., , and J. S. Turner, 1956: On the collision of drops in turbulent clouds. J. Fluid Mech., 1 , 16–30.
Schudlich, R., , and S. Emerson, 1996: Gas supersaturation in the surface ocean: The roles of heat, gas exchange and bubbles. Deep-Sea Res., 43 , 569–589.
Skyllingstad, E. D., , and D. W. Denbo, 1995: An ocean large-eddy simulation of Langmuir circulations and convection in the surface mixed layer. J. Geophys. Res., 100 , 8501–8522.
Smayda, T. J., , and C. S. Reynolds, 2001: Community assembly in marine phytoplankton: applications of recent models to harmful dinoflagellate blooms. J. Plankton Res., 23 , 447–461.
Smith, J. A., 1998: Evolution of Langmuir circulation during a storm. J. Geophys. Res., 103 , 12649–12668.
Smith, J. A., 1999: Observations of wind, waves and the mixed layer: The scaling of surface motion. The Wind-Driven Air–Sea Interface, M. Banner, Ed., School of Mathematics, The University of New South Wales, 231–238.
Spelt, P. D. M., , and A. Biesheuvel, 1997: On the motion of gas bubbles in homogeneous isotropic turbulence. J. Fluid Mech., 336 , 221–244.
Stommel, H., 1949: Trajectories of small bodies sinking slowly through convection cells. J. Mar. Res., 8 , 24–29.
Stramski, D., , and J. Tegowski, 2001: Effects of intermittent entrainment of air bubbles by breaking wind waves on ocean reflectance and underwater light field. J. Geophys. Res., 106 , 31345–31360.
Taylor, P. A., , P. Y. Li, , and J. D. Wilson, 2003: Lagrangian simulation of suspended particles in the neutrally stratified surface boundary layer. J. Geophys Res., in press.
Terray, E. A., , M. A. Donelan, , Y. C. Agrawal, , W. M. Drennen, , K. K. Kahma, , A. J. Williams III, , P. A. Hwang, , and S. A. Kitagorodskii, 1996: Estimates of kinetic energy dissipation under breaking waves. J. Phys. Oceanogr., 26 , 792–807.
Terrill, E. J., , W. K. Melville, , and D. Stramski, 2001: Bubble entrainment by breaking waves and their influence on optical scattering in the upper ocean. J. Geophys. Res., 106 , 16815–16824.
Thomas, W. H., , and C. H. Gibson, 1990: Quantified small-scale turbulence inhibits a red tide dinoflagellate Gonyaulax polyedra Stein. Deep-Sea Res., 37 , 1583–1593.
Thorpe, S. A., 1982: On the clouds of bubbles formed by breaking wind-waves in deep water, and their role in air–sea gas transfer. Philos. Trans. Roy. Soc. London, 304A , 155–210.
Thorpe, S. A., 1984a: A model of the turbulent diffusion of bubbles below the sea surface. J. Phys. Oceanogr., 14 , 841–854.
Thorpe, S. A., 1984b: On the determination of Kυ in the near-surface ocean from acoustic measurements of bubbles. J. Phys. Oceanogr., 14 , 855–863.
Thorpe, S. A., 1984c: The effect of Langmuir circulation on the distribution of submerged bubbles caused by breaking waves. J. Fluid Mech., 142 , 151–170.
Thorpe, S. A., 1984d: The role of bubbles produced by breaking waves in super-saturating the near-surface ocean mixing layer with oxygen. Ann. Geophys., 2 , 53–56.
Thorpe, S. A., 1986: Bubble clouds: A review of their detection by sonar, of related models, and of how Kυ may be determined. Oceanic Whitecaps, E. C. Monahan and G. MacNioaill, Eds., D. Reidel, 57–68.
Thorpe, S. A., 1992: Bubble clouds and the dynamics of the upper ocean. Quart. J. Roy. Meteor. Soc., 118 , 1–22.
Thorpe, S. A., , and A. J. Hall, 1982: Observations of the thermal structure of Langmuir circulation. J. Fluid Mech., 114 , 237–250.
Thorpe, S. A., , and A. J. Hall, 1983: The characteristics of breaking waves, bubble clouds and near-surface currents observed using side-scan sonar. Cont. Shelf Res., 1 , 353–384.
Thorpe, S. A., , P. Bowyer, , and D. K. Woolf, 1992: Some factors affecting the size distributions of oceanic bubbles. J. Phys. Oceanogr., 22 , 382–389.
Thorpe, S. A., , M. S. Curé, , A. Graham, , and A. J. Hall, 1994: Sonar observations of Langmuir circulation, and estimation of dispersion of floating bodies. J. Atmos. Oceanic Technol., 11 , 1273–1294.
Thorpe, S. A., , W. A. M. Nimmo Smith, , A. Graham, , and A. M. Thurnherr, 1999: Patterns in foam and shallow tidal flows. The Wind-Driven Air–Sea Interface. M. Banner, Ed., School of Mathematics, The University of New South Wales, 257–264.
Thorpe, S. A., , T. R. Osborn, , J. E. F. Jackson, , A. J. Hall, , and R. G. Lueck, 2003: Measurements of turbulence in the upper-ocean mixing layer using Autosub. J. Phys. Oceanogr., 33 , 122–145.
Weller, R. A., , J. P. Dean, , J. Marra, , J. F. Price, , E. A. Francis, , and D. C. Broadman, 1985: Three-dimensional flow in the upper ocean. Science, 227 , 1552–1556.
Woolf, D. K., , and S. A. Thorpe, 1991: Bubbles and the air–sea exchange of gases in near-saturation conditions. J. Mar. Res., 49 , 435–466.
Zedel, L., , and D. M. Farmer, 1991: Organized structures in subsurface bubble clouds: Langmuir circulation in the open ocean. J. Geophys. Res., 96 , 8889–8900.
Histograms of dissipation. On the right are the normalized histograms (pdfs) of logε at mean depths z of (top) 2.22 m (z/Hs = 2.2, W10 = 10.8 m s−1) and (bottom) 4.16 m (z/Hs = 4.16, W10 = 11.7 m s−1). The regions in black show the proportion of dissipation rates that are found in regions of bubble clouds where the void fraction υf > 2 × 10−8. The corresponding ratios of logε values in bubble clouds to those in the full record are shown on the left
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Conditional sample plots made by taking averages across (a) 196 bubble clouds (regions where υf > 2 × 10−8) at a depth of 4.04 m (z/Hs = 4.21, W10 = 11.8 m s−1), (b) 31 Langmuir bubble bands at a depth of 2.1 m (z/Hs = 1.92, W10 = 10.6 m s−1), and (c) 154 bands at a depth of 3.9 m (z/Hs = 4.35, W10 = 11.9 m s−1), showing the variation of dissipation ε (m2 s−3) and void fraction log(υf). The bands were crossed at approximately 45° (see Thorpe et al. 2003) and a time of 10 s corresponds to a distance of about 12.5 m. Dashed horizontal lines show leg-mean values
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Bubble size distributions at the center of 30 Langmuir bands at mean depths of (top) 2.1 and (bottom) 3.9 m and in wind speeds of 10.6 and 11.9 m s−1, respectively. (a) The bubble number concentration, n(a)da [the number of bubbles per cubic meter with radius a between (a − 0.5) and (a + 0.5) μm], averaged over the 30 Lc bands. The dashed line shows the mean bubble concentrations at 8 s before and after the center of the bands is reached. (b) log[a3n(a)da]. Here a3n(a)da is proportional to the void fraction per unit radius band and is averaged over values at the center of all 30 bands.
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Model estimates of particle concentration and diffusivity ratio as a function of depth. The log of the mean particle concentration, C(z), is plotted against z/l, with Lc (crosses) and without Lc (full line) in wind speeds of (a) 6, (b) 8, (c) 10, (d) 12, and (e) 14 m s−1, for depth-independent values of Kz given by (14). Also shown is log(r), with the ratio of particle diffusivity to eddy viscosity r given by (20): Lc is modeled by (13) using (5) and (6). The particles are introduced randomly is in the region 0 < z < 1.5Hs [with Hs given by (6)], corresponding to the depth range in which bubbles are injected by breakers. The bottom of the injection region is marked by an arrow. The particle decay rate is 0.006 25 s−1. Negative values of log(r) where the downward gradient of mean concentration is positive are omitted. Input and output measures are given in Table 2
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
As in Fig. 4 but with Kz given by (15). Input and output measures are given in Table 3.
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Particle concentration and diffusivity ratio r vs z/l with Kz given by (14) for (circles) 1 cell, (crosses) 2 cells, and (squares) 3 cells. The wind speed is 10 m s−1 and ancillary values are given in Table 4. Negative values of log(r) where the downward gradient of mean concentration is positive are omitted. (a) Nonoverlapping cells; (b) superimposed cells
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Particle concentration and diffusivity ratio r vs z/l as in Fig. 6, but with Kz given by (15) for (circles) 1 cell, (crosses) 2 cells, and (squares) 3 cells. The wind speed is 10 m s−1 and ancillary values are given in Table 5
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Particle concentration and diffusivity ratio r vs z/l with Kz given by (14) (crosses) or (15) (dots) for the set of four nonoverlapping cells described by (iii) in text. The wind speed is 10 m s−1 and ancillary values are given in Table 6
Citation: Journal of Physical Oceanography 33, 9; 10.1175/1520-0485(2003)033<2013:BCALCO>2.0.CO;2
Values of scales and parameters corresponding to the different wind speeds, W10 , examined in the numerical model. Five values of wind speed W10 are given, with corresponding Lc speeds u Lc [see (5)], significant wave height H s [see (8)], and cell depth l [see (6)]. The lifetime and concentrations are those of bubbles reaching the surface after release at random positions in the depth range 0 < z < 1.5 H s when there is no eddy diffusivity and no Lc
Values corresponding to the profiles in Fig. 4 with Kz given by (14), independent of z. Values u Lc , H s and l [see (6)] are as given in Table 1. Here Σ t is the std dev of the bubble lifetimes. The two values of lifetime Σ t and concentration are those of particles reaching the surface with no Lc/with Lc. Values of θ, φ, and q are found from (10)–(12) and of R from (9)
Values corresponding to the profiles in Fig. 5 with K z given by (15) and calculated at z = 1.5H s . Values u Lc , H s , and l [see (6)] are as given in Table 1. Here Σ t is the std dev of the bubble lifetimes. The two values of lifetime, Σ t and concentration are mean values for particles reaching the surface with no Lc/with Lc. Values of θ, φ, and q found from (10)–(12) are calculated using the value of K z at mid–cell depth (z = l /2). Values of u Lc , H s , l, and R are as given in Table 1
Values corresponding to the profiles for multiple cells in Fig. 6 with K z given by (14) at W10 = 10 m s −1 ; Σ t is the std dev of the bubble lifetimes. The mean lifetimes may be compared with that of 60.8 s with no eddy diffusivity or Lc (Table 1). Values of u Lc , H s and l are as given in Table 1 and R as in Table 3 at 10 m s −1
Values at W10 = 10 m s −1 corresponding to the profiles for multiple cells in Fig. 7 with K z given by (15) at z = 1.5H s ; Σ t is the std dev of the bubble lifetimes. The mean lifetimes may be compared with that of 60.8 s with no eddy diffusivity or Lc (Table 1). Values of u Lc , H s , and l are as given in Table 1 and R as in Table 3 at W10 = 10 m s −1
Values corresponding to the profiles for a combination of four nonoverlapping cells in Fig. 8 at W10 = 10 m s −1 ; Σ t is the std dev of the bubble lifetimes. Values of u Lc , H s and l are as given in Table 1
Details of conditional sampling and of the variation of other measured quantities are given by Thorpe et al. (2003). In brief, the conditional plots consist of sets of values which are averaged together so as to provide a representation of the mean variation across a time location identified by some criterion, here the time of crossing the centre of bubble clouds or bands of bubbles associated with Lc. Bands are identified using a forward and a sideways pointing side-scan sonar carried on Autosub. The forward pointing sonar identifies bands as they are approached, and the sideways pointing sonar views bands as the AUV converges on them. Crossing times are determined by objective analysis.
If the overall eddy viscosity within the near-surface region is given by a law of the wall formulation as suggested by the Autosub measurements of ε, it appears inappropriate to include both a formulation of eddy diffusivity in the form, Kz = ku∗wz, and a representation of momentum diffusion by Lc, within the same model. If however (as here) the motion within Lc is as represented by (13), steady and purely two-dimensional in the (x, z)-plane normal to the wind direction y, with no mean shear flow in the x direction, then the spatially averaged product of the two components of velocity induced by Lc, ∂ψ/∂z and −∂ψ/∂x, at a given depth will be zero. So too will products of the circulation velocity components with those of small-scale turbulent motions provided the two are uncorrelated. The contribution of Lc to the x-directed Reynolds stress momentum transfer is therefore zero unless the turbulent fluctuations are correlated to the circulatory flow field. (The high ε in the downwelling motion of Lc found in section 2 suggests this assumption of decorrelated Lc and turbulent motions is not valid.) It is usual to define an eddy diffusion coefficient as a stress divided by a mean velocity gradient, but in this case both stress and velocity shear are zero. There is no immediate route to estimating an overall eddy viscosity (other than Kz) in the present two-dimensional model.
