1. Introduction
Diurnal warm layers (DWLs) form when strong solar radiation and weak-to-moderate winds allow near-surface stratification to develop. In the tropics DWLs appear around 0800 local time (LT), which is 1–2 h after sunrise, as the surface heat flux changes from net ocean cooling to net warming (Martin 1985; Fairall et al. 1996; Moulin et al. 2018). Heat and momentum trapped in this stratified layer cause sea surface temperature (SST) anomalies of O(0.1–1°C) and near-surface velocity anomalies of 0.1–0.2 m s−1 (e.g., Price et al. 1986; Kudryavtsev and Soloviev 1990). The previous night’s mixed layer evolves into a DWL lying above, and decoupled from, the remnant mixed layer below.
Per the optical properties of seawater, more than half of the sun’s heat impinging on the surface is absorbed in the top two meters of the ocean. The subsequent fate of this heat depends on turbulence. Consider two extremes: strong turbulent mixing and no turbulent mixing. In the former case, the sun’s heat is spread throughout the mixed layer and warms each parcel of water by O(0.1°C) by midafternoon. In the latter case, warming is concentrated in the top 2 m and, consequently, more of this heat is likely to be transferred from the ocean back to the atmosphere over a short time scale. In between these extremes heat transport is more complicated. Warming of the lower half of the mixed layer, for example, lags the surface solar forcing by several hours because it depends on the descent of the DWL. Using a large-eddy simulation under 6 m s−1 winds, Sarkar and Pham (2019) show that the lower half of the mixed layer warmed over their week-long simulation (net mean surface heat input to the ocean was 95 W m−2). Yet the turbulent heat transport causing this warming occurred over only four hours each day during the late afternoon to evening.
The timing and magnitude of subsurface turbulent heat flux governs SST and the consequent latent, sensible, and longwave radiative fluxes at the surface. In the tropics, the atmosphere is sensitive to small SST variations (Webster and Lukas 1992). Increases in SST during the day can lead to air–sea heat flux anomalies of 50 W m−2 relative to what they would be if the sea surface remained at its presunrise temperature (Fairall et al. 1996). Understanding these increases can improve numerical weather forecasts (e.g., Pimentel et al. 2008) and inform operational procedures such as corrections to SST based on bulk temperature measurements below the surface from ships (Fairall et al. 1996). The daily changes also add up over longer time scales. Development of the widely used K-profile parameterization (KPP) mixing scheme was motivated in part by a need to ensure physically reasonable mixing on diurnal time scales given its influence over climate time scales (Large et al. 1994).
A necessary step in predicting or parameterizing the evolution of the near-surface temperature field is an observational depiction of the relevant coevolving parameters (stratification, turbulence dissipation, and turbulent heat fluxes) and their responses to atmospheric forcing. Many studies have described the daytime reduction of turbulence dissipation ϵ in the remnant mixed layer immediately below the DWL (e.g., Dillon and Powell 1979; Moum and Caldwell 1985; Caldwell et al. 1997; Pujiana et al. 2018). The DWL, however, behaves differently. Two recent studies measuring microstructure shear from either a freely rising profiler or a glider demonstrate enhanced near-surface turbulence during daytime (Sutherland et al. 2016; St. Laurent and Merrifield 2017). The enhancement is attributed to shear and internal waves induced by the diurnal thermocline. It is this turbulence that drives the daytime descent of the DWL and hence controls the mixed layer heat budget.
Here we characterize near-surface turbulence dissipation ϵ and turbulent heat fluxes using recent measurements from a surface-following platform. Such platforms negate many of the challenges faced by other types of profilers. Vertical, free-falling profilers are unreliable in the top 5–10 m (e.g., Moulin et al. 2018). Gliders and freely rising vertical profilers can measure closer to the surface, but are limited in profile sampling rate (e.g., Greenan et al. 2001; Sutherland et al. 2013). Moreover, ϵ from shear at microscales depends on body speed to the fourth power making profiles sensitive to the orbital velocities of surface waves (e.g., Lueck 2016).
Recent examples of surface-following platforms include a sailboard adapted to measure salinity profiles in the top meter of the ocean (Asher et al. 2014), a trimaran adapted to measure atmospheric turbulence just above the sea surface (Bourras et al. 2014), and “SWIFT” drifters to measure near-surface turbulence and shear (Thomson 2012; Thomson et al. 2019). Like Asher et al. (2014), our platform is towed so as to sample undisturbed water outside the ship’s wake. Although temperature is measured at only discrete depths, the resulting horizontal profiles capture a broader bandwidth of turbulence compared to vertical profiles (Goodman et al. 2006; Klymak and Moum 2007).
Our platform houses several fast thermistors and thereby allows temperature spectra to be measured at several fixed depths (section 2). Turbulence dissipation rates inferred from the spectra differ by more than two orders of magnitude depending on the position relative to the diurnal thermocline (section 3). Combining the inferred diffusivity with vertical temperature gradients, we directly evaluate subsurface turbulent heat fluxes, which play a first-order role in the near-surface heat budget except in calm conditions (section 4). Overall, we aim to diagnose the feedbacks between turbulence and stratification and the relative roles of solar and turbulent heat fluxes in the development of diurnal warm layers.
2. Instrumentation and deployments
SurfOtter is a platform specially designed for measurements of turbulence, temperature, salinity, and velocity in the near-surface ocean (Fig. 1). It was first introduced in Hughes et al. (2020) to observationally assess vertical current shear induced by DWLs. It is a 3-m-long aluminum tube with a 2-m-deep fin that hosts a suite of oceanographic sensors. A cable attached to the bottom of the fin can extend the depth range to 8 m. SurfOtter is towed outside the ship’s wake at speeds of approximately 1 m s−1 in the manner shown in Fig. 1c. The overall design ensures that the water sampled is undisturbed and that the influence of surface waves is minimized relative to ship-mounted sensors.
While SurfOtter was deployed, we continually profiled with the microstructure profiler Chameleon off the stern of the ship, with profiles every 6–12 min (Fig. 1c). For this paper, this additional dataset provides measurements of temperature and salinity at depths of 8 m and below. Above this depth, Chameleon measurements are influenced by ship wake and prop wash as shown by Moulin et al. (2018).
In this paper, we use data from a 2-month field campaign in 2018 at 12°–18°N, 134°E (time zone: UTC + 9). During this period, SurfOtter cumulatively sampled 24 days. Of these, 12 were full-day records (Table 1) with surface warming of >0.05°C and without a confounding salinity influence (i.e., a transient rain layer). On some days we transited around a point within a radius of 10 km. On others we transited in a straight line and moved >100 km in 24 h. By comparison, for the location and season of our campaign, DWLs with an equivalent radius of O(1000) km are not uncommon (Bellenger and Duvel 2009). To minimize horizontal variability, we discuss temperatures as anomalies relative to their values at a depth near or below the base of the DWL.
Summary of the DWLs observed during the 2018 field campaign analyzed in this paper.
We primarily consider three representative days that differ by wind forcing. We designate these with the relative labels “calm,” “moderate,” and “windy.” The respective wind speeds and wind stresses were 2, 4, and 8 m s−1 and 0.005, 0.02, and 0.08 N m−2. The calm day (4 October) had the lowest wind speed of all days during which SurfOtter was deployed. The cable was not present during this deployment (Fig. 1b). The windy day (27 October) is near the wind limit at which DWLs occur (Kudryavtsev and Soloviev 1990; Thompson et al. 2019). The moderate day (29 October) is a point between the two extremes. Of the 12 days in the dataset, there are two low-, six moderate-, and four high-wind-speed days (Fig. 2). Similar three-category classifications were used by Kondo et al. (1979) and Soloviev and Lukas (1997). The role of surface buoyancy flux in dictating whether near-surface stratification can persist under certain winds is detailed by Thompson et al. (2019).
The calm day is representative of large parts of the tropical Pacific and Indian Ocean where winds are weak and daily SST increases of 1 ± 0.5°C are common (Webster et al. 1996; Stuart-Menteth et al. 2003). For the tropics in general (23.5°S–23.5°N), daily-mean wind speed over the ocean has a mean of 6.5 m s−1 (based on the NCEP–DOE reanalysis for 2018; see Kanamitsu et al. 2002); 11% of values are 0–3 m s−1, 33% are 3–6 m s−1, and 39% are 6–9 m s−1. These values, however, overestimate the prevalence of DWLs, which are weaker if clouds occur. If we recalculate with an additional criterion that cloud cover must be <50%, the respective percentages become 5%, 15%, and 15%.
Across the three representative days, differences in the temperature fields are immediately evident (Figs. 3d–f; note the varying color scales). For example, the surface warms by 1°C relative to 8-m depth on the calm day but only 0.05°C on the windy day. Because temperature is well resolved vertically, we consider its vertical gradient Tz (Figs. 3g–i) a continuous quantity. From Tz, we can discern a descent rate of the DWL using
Temperature microstructure from fast thermistors (100 Hz) was measured at discrete depths. Figures 3g–i include the measured temperature rate of change Tt as a qualitative depiction of turbulence that is partly dependent on Tz. Knowing both Tt and Tz, we can infer ϵ. SurfOtter housed two different instruments with fast thermistors: GusTs (Becherer et al. 2020) and TPods. Both are built in-house at Oregon State University and sample temperature and accelerations at 100 Hz. They are almost the same for our purposes here. On the calm day, there were six fast thermistors: two GusTs and four TPods. The latter lack associated pressure sensors, so we interpolate from GusTs when needed (appendix B). On the moderate and windy days there were only four reliable thermistors, all GusTs. Overall, turbulence measurements during the calm day encompassed a smaller depth range, but were better resolved vertically. Much of our analysis focuses on the GusTs, which were at depths of 0.4, 1.7, 4.2, and 7.7 m.
Fast thermistor measurements are continuous for whole deployments. For turbulence calculations, we segment these into 10-min blocks. With each block, we first convert Tt to an equivalent horizontal gradient Tx = Tt/U where U is the 10-min mean flow speed past the sensor. We then calculate the dissipation of thermal variance χ by fitting Tx spectra to a theoretical form for the inertial–convective (IC) subrange (Batchelor 1959) over the frequency range 0.067–2 Hz, corresponding to periods of 15–0.5 s. We finally calculate turbulence dissipation ϵ whenever there is measurable stratification (Tz > 0.002°C m−1) by assuming a steady-state turbulent kinetic energy balance (Osborn 1980). Hereinafter, we denote ϵ as ϵχ to recognize the intermediate calculation of χ. Although ϵχ is a commonly derived quantity (e.g., Moum and Nash 2009; Becherer and Moum 2017; Moulin et al. 2018), its derivation can include study-specific details. Here, these include aspects specific to SurfOtter and the proximity to the sea surface as detailed in the appendices.
3. Diurnal evolution of turbulence dissipation
The changing stratification throughout the day is the predominant factor explaining the daytime evolution of ϵχ. Moulin et al. (2018) describe the evolution at a given depth in three stages: (i) an initial decay of turbulence 1–2 h after sunrise as the near-surface stratifies, (ii) a rapid increase in turbulence once the DWL descends to the depth in question, and (iii) a near-steady state with a three-term balance for turbulent kinetic energy involving shear production, buoyancy flux, and dissipation. In this section, we extend Moulin et al.’s analysis with new observations that clarify the change in microstructure above and below the diurnal thermocline, the role of wind forcing, and the vertical scales of turbulence.
a. Temperature variance above and below the diurnal thermocline
A clear example of the influence of stratification on temperature microstructure occurs at 1530 LT on the moderate day. The top two GusTs at 0.4 and 1.7 m show turbulent temperature fluctuations of 0.01°–0.02°C (Figs. 4a,b). The third GusT at 4.2 m shows fluctuations that are 5 times as large (Fig. 4c). The deepest GusT at 7.7 m shows 0.01°C fluctuations that are not predominantly turbulent. Instead, they result from the background temperature gradient being advected by wave orbital velocities (see appendix B). Overall, these changes in temperature microstructure across the four GusTs, and the associated values of ϵχ, are explained by the depths relative to
For the top three GusTs (0.4–4.2 m), temperature gradient spectra display +1/3 slopes characteristic of the IC subrange (Figs. 4e–g; see also Fig. D2). In the top 2 m, ϵχ is nearly the same magnitude (7–8 × 10−7 W kg−1). Note that ϵχ does not account for the possibility of turbulence generated by wave breaking. In appendix C, we infer that ϵχ underestimates the true ϵ in the top 1 m in moderate winds. (On the calm and windy days, ϵχ underestimates ϵ in the top 0.5 and 2 m, respectively.) Perhaps surprisingly, ϵχ is larger at the third GusT at 4.2 m despite the stratification being 3–10 times stronger than at the two GusTs above. Presumably there is stronger vertical current shear here induced by the stratification (Sutherland et al. 2016).
Beneath
b. Depth and time variability of turbulence
To see the influence of DWLs on upper-ocean turbulence, we compare ϵχ with the value we expect without diurnal stratification. The latter is often predicted based on a wind-dependent scaling:
where κ = 0.4 is the von Kármán constant and z is measured relative to the free surface. The expression follows from wall layer scaling based on the friction velocity [u* = (τ/ρw)1/2 with τ being the wind stress and ρw being the density of water]. A second term to account for convectively driven turbulence is often included in Eq. (1) (see, e.g., Lombardo and Gregg 1989). In our case, this term is zero between approximately 0700 and 1700 LT when there is a net surface ocean heat input. Further, tests with the extra term showed that it made little difference to our results between 0500 and 2000 LT, although this would not hold if our analysis extended later into the night or to deeper depths.
The stabilizing influence of a DWL limits the vertical transport of stress. Whereas the ϵ* scaling follows from assuming a turbulent length scale of κz, the true length scale in the presence of stratification is likely smaller. More generally, to illustrate the influence of stratification, we compare ϵχ and ϵ*. On the calm and moderate days, we find that below the DWL ϵχ is one–two orders of magnitude smaller than ϵ* (Figs. 5b,d). We also see ϵχ ≪ ϵ* near the surface on the windy day (Fig. 5f), but this may reflect ϵχ being an underestimate as noted in section 3a.
Above
Although there is variability in ϵχ/ϵ* on 10-min scales, the dominant signal at most depths is a change from values of <1 to values of >1 as the DWL descends. This indicates that the diurnal time scale is dominant, consistent with past studies (e.g., Caldwell et al. 1997; Sutherland et al. 2016). At the low latitudes in this study, the Coriolis period is too long to influence the deepening rate (cf. Pollard et al. 1972; Pham and Sarkar 2017).
c. Inferring anisotropy of diurnal warm layer turbulence
A second depiction of DWL turbulence that complements Fig. 5 is the evolution in log(N)–log(ϵχ) space (e.g., Moum 1997). This space maps the progression of the Ozmidov scale: LO = (ϵ/N3)1/2, a measure of the vertical scale of turbulent overturns (Fig. 6). On the calm and moderate days, LO varies by more than two orders of magnitude with minima less than 0.1 m. During morning and early afternoon, this is the dominant vertical turbulent length scale, whereas later in the day LO increases to values comparable to the distance to the free surface. On the windy day, LO fluctuates around 1–10 m, so both stratification and proximity to the surface limit the vertical extent of turbulent eddies. [See Sutherland et al. (2016) for a comparison of other various length scales relevant to DWLs.]
Another indicator of the state of turbulence is the buoyancy Reynolds number Reb = ϵ/νN2. As Reb increases, turbulence likely tends from anisotropy to isotropy. Although this is a continuum, we consider three representative values: 25, 300, and 1500. When Reb is below 25, turbulent buoyancy fluxes become negligible (Stillinger et al. 1983). The larger two values follow from Gargett et al. (1984). For values of Reb ≲ 300, they describe the turbulence as chaotic and “two dimensional,” with the vertical motions being small due to attenuation by the stratification. At larger, but still moderate Reb (≲1500), the turbulence remains anisotropic. Note that we are being necessarily qualitative about anisotropy. Reb can be interpreted in more than one way depending on physical constraints or assumptions about the generation and evolution of stratified turbulence (Gargett 1988). Here we are assuming that the aforementioned relationship holds in our situation. That is, measurements in geophysical-scale flows demonstrate increasing anisotropy, particularly at lower wavenumbers, with decreasing Reb (Gargett et al. 1984; Yamazaki 1990). The value of 25 comes from laboratory experiments but is subject to similar ambiguity.
With the approximate bounds added to Fig. 6, we infer anisotropy on the moderate and calm days during morning and early afternoon. During a subset of these times, turbulent buoyancy fluxes become negligible. This occurs on the calm day at 0900–1400 LT at 1.7 m and on the moderate day at 1100–1600 LT at 7.7 m. Although values of ϵχ are uncertain at such times because the thermistor record is near digitization noise (e.g., Fig. 4h), the signal being near noise level is consistent with vanishing turbulent buoyancy fluxes.
In log(N)–log(ϵχ) parameter space, the evolution of DWL turbulence reduces to a counterclockwise path, a result that holds across the entire dataset (1). This pathway neatly illustrates the sequence shown by Moulin et al. (2018). Nighttime convection leaves small N and comparatively large ϵχ early in the day. The building diurnal stratification quenches ϵχ as N increases. At depths below
4. Subsurface heat transport
The evolution of turbulence within the DWL governs if and when heat is transported downward and the consequent evolution of mixed layer temperature. In this section, we use the ϵχ(z, t) fields presented in the previous section to calculate turbulent heat fluxes during our three case study days to show when subsurface turbulent heat fluxes on diurnal time scales are a first-order process.
Here we consider thermodynamic heating between discrete depths in which heat content H changes due to divergences of turbulent and solar heat fluxes:
where Jq is a heat flux in watts per meter squared that is separated into its solar (s) and turbulent (t) components. The balance excludes the air–sea interface, so latent, sensible, and longwave radiative fluxes do not contribute. Hereinafter, the lack of a superscript s or t implies the sum of the two terms. Downward fluxes are negative (z increases upward). Hence, ∂Jq/∂z < 0 indicates a convergence of heat leading to warming. Everywhere,
Parameter
The heat content per unit volume between vertically adjacent GusTs at z1 and z2 is related to the temperature above the mixed layer temperature TML:
One difficulty in this calculation is an appropriate definition and measurement of TML. We use a variable TML from our microstructure profiling to minimize any apparent horizontal advective heat fluxes that can arise due to transiting at ~1 m s−1. However, there is often no clear single value for TML. Such an issue is well illustrated by Fig. 6b of Matthews et al. (2014). Ultimately, we define TML as the median temperature between 15 and 25 m. We also low-pass filter H with a 2-h cutoff before taking the time derivative. Gerbi et al. (2008) also identified a 2-h smoothing as necessary for a similar analysis.
We evaluate Eq. (2) at both fixed depths and fixed times. Fixed depths are preferable for demonstrating similarity between the left- and right-hand sides, whereas vertical profiles at fixed time better demonstrate the predominantly vertical nature of the heat transport. In Fig. 7, we present the left- and right-hand sides of Eq. (2) and also isolate the turbulent component to indicate its comparative magnitude. In Fig. 8, we show vertical profiles.
a. Time series of subsurface thermodynamic heating
We first consider the heating throughout the day between vertically adjacent GusTs (Fig. 7). For the calm day, we consider the top two GusTs (0.4 and 1.7 m); for the moderate day, the middle two (1.7 and 4.2 m); and for the windy day, the bottom two (4.2 and 7.7 m). Each range best encompasses the diurnal thermocline on the respective day (Fig. 3).
For much of the calm day,
At 1500 LT on the calm day, ∂H/∂t switched from warming to cooling (feature C; note that −∂H/∂t is the quantity plotted). At the same time, the magnitudes of
In contrast to the calm day, the dominant balance on the moderate day is between
On the windy day, we approach the limits of our ability to derive
b. Vertical heating profiles
Turbulent heat flux convergence near
In section 4a, we discussed the convergence of heat on the windy day between 4.2 and 7.7 m. Figures 8i–l further elucidate this convergence. In the top 2 m, Tz has a local maximum due to the continual input and absorption of solar radiation near the surface.
c. Evaluation and consequences of subsurface heat fluxes
The primary consequence of subsurface turbulent mixing in DWLs is to move the sun’s heat downward away from the surface and thereby modulate diurnal variation of SST and the associated surface heat fluxes. Taking into account climatological ocean and atmospheric processes over the tropical Pacific, Seager et al. (1988) estimated that a 1 W m−2 change in air–sea heat flux is associated with a 0.04°–0.08°C change in SST. A change of, say, 5 W m−1 equates to ~0.3°C. In our dataset, on the calm day, the temperature at 0.2 m (the shallowest temperature logger) exceeded the mixed layer temperature TML by 0.3°C for 8 h (not shown). On the moderate and windy days, this decreased to 5 and 0 h, respectively. Matthews et al. (2014) demonstrate similar relationships in a 3-month-long dataset from the Indian Ocean (see their Figs. 2 and 8).
Although there is a well-established inverse correlation between wind speed and diurnal SST anomalies (e.g., Price et al. 1986; Lukas 1991; Gentemann et al. 2009), the amount of heat mixed downward is typically an unknown. With SurfOtter, we have rare direct measurements of the near-surface turbulent heat flux into the ocean based on accurate mean temperature profiles.
In moderate conditions (4–5 m s−1), we observe daytime subsurface
From observed temperature rates of change, Price et al. (1986) inferred maximum fluxes of −100 W m−2 at 1400–1500 LT at 5 m deep under 4–6 m s−1 winds. With respect to vertical structure, they highlight the distinct middepth maximum (see their Fig. 3d), which they term a “knee.” Its depth closely follows the deepening surface layer. We also observe such a knee in our
To put
Using our observed heating time series and profiles, and aided by an idealized simulation detailed by Hughes et al. (2020), we schematize the heating within a DWL in Fig. 10. Primarily representative of moderate wind conditions, the figure summarizes cases when the DWL descends by O(10) m over the day. Before sunrise, convection leads to an upward turbulent heat flux throughout the mixed layer. After the net surface heat flux changes sign, the near surface stratifies and turbulent heat fluxes throughout the mixed layer decay. Much of the incoming solar energy is absorbed close to the surface and then transported downward by turbulent mixing. Turbulent heat fluxes are largest in the top half of the diurnal thermocline in the afternoon and evening owing to the optimal combination of turbulent diffusivity and vertical temperature gradient. The decrease in diffusivity in the lower part of the thermocline causes a convergence of turbulent heat flux and hence warming.
5. Conclusions
Heat flux convergence in the near-surface tropical ocean includes solar and turbulent components. The latter was measured here using fast thermistors and vertical temperature gradients from a temperature array. The total convergence was corroborated by measuring the temporal rate of change of heat from the array. This agreement between two independent measurements was made possible by a new platform SurfOtter. Its ability to follow the surface, together with its suite of sensors, enabled a new method to derive near-surface turbulence dissipation and turbulent heat fluxes. Temperature spectra were fit across and above frequencies corresponding to surface waves. This large bandwidth minimized uncertainty in the calculation of the turbulent quantities.
The diurnal thermocline clearly suppressed the vertical extent of turbulence. Specifically, the depth of the peak temperature gradient,
There is also evidence that turbulent heat fluxes shut off completely in the remnant mixed layer. This was quantitatively demonstrated through calculation of the buoyancy Reynolds number. It was also apparent in the raw temperature records. Under 4 m s−1 winds at 7.7-m depth, for example, the temperature record showed digitization noise superimposed on the signal due to surface waves. By comparison, at the same time but at 4.2 m, turbulence dominated (Fig. 4). This depth variation in turbulence, and hence turbulent diffusivity, has a perhaps curious consequence for the diurnal temperature gradient. Although we normally regard turbulence as something that helps mix away gradients, it is not so simple when diffusivity is spatially variable. In such a case, the consequent convergence can maintain gradients in the presence of the otherwise diffusing influence of turbulence. Such convergence was evident in vertical profiles throughout our dataset.
Turbulence, or lack thereof, governed whether downward heat fluxes during the afternoon were appreciable and thereby whether diurnal SST anomalies were O(0.1)°C or O(1)°C . On the moderate and windy days, turbulent heat fluxes were hundreds of watts per meter squared. Such values would be rare in much of the ocean. In diurnal warm layers, however, large turbulent heat fluxes are commonplace given that, in the top 2 m of the ocean on a clear-sky day in the tropics, a daytime mean of O(300) W m−2 of solar energy is absorbed and that heat has to be redistributed vertically.
Acknowledgments
This research is part of the PISTON project: Propagation of Intra-Seasonal Oscillations, funded by the Office of Naval Research. Shipboard measurements described in this paper were taken aboard RV Thomas G. Thompson. Craig Van Appledorn, Kerry Latham, and Pavan Vutukur designed and constructed SurfOtter and the turbulence instruments it housed. Sally Warner processed the Chameleon dataset. Meteorological data were collected and provided by NOAA’s Earth System Laboratory (Chris Fairall, Byron Blomquist, Ludovic Bariteau, Simon de Szoeke, and Elizabeth Thompson). Comments from two anonymous reviewers helped to clarify several aspects of this paper.
Data availability statement
Datasets for each of the 12 days that are described in Table 1 are available online (kghughes.com/data and http://dx.doi.org/10.5281/zenodo.3894910). These include the temperature fields, the full thermistor records, the derived turbulent quantities, and many other auxiliary data.
APPENDIX A
Deriving Turbulence Dissipation from Fast Thermistor Records
The spectrum of temperature gradient at small, turbulent scales is separated into three distinct subranges. At the largest turbulence scales, an IC subrange occurs in which the downscale turbulent transfer is unaffected by molecular processes. With decreasing length scale, the effects of viscosity and then diffusivity become leading order, resulting in the viscous–convective and viscous–diffusive subranges. In this paper, we infer ϵχ by fitting measured temperature spectra to an analytical form for the IC subrange. Good agreement overall in ϵ values derived from IC fits and from fits to higher-frequency subranges has been demonstrated by Zhang and Moum (2010) and Becherer and Moum (2017). There are times, however, when the two estimates disagree for reasons we do not yet understand.
With a measurement of the mean flow speed past the sensor U, temporal temperature gradients can be converted to horizontal temperature gradients as Tx = Tt/U. We exclude any influence of wave orbital velocities in this conversion for three reasons. First, we tow SurfOtter through the water at speeds greater than orbital velocities. Second, the variance of Tt in the 15–0.5-s band can span three orders of magnitude in a day. This dominates the variance of Tx relative to the effects of orbital velocities. Third, the 200-m-long cable used to tow SurfOtter takes a catenary shape, providing elasticity at surface wave frequencies. Hence, to some extent, SurfOtter can behave in a Lagrangian manner, which minimizes the orbital velocities relative to the thermistors.
With Tx calculated, we use the following expressions to derive ϵχ:
where
We infer ϵχ over 10-min segments and fit over the frequency range corresponding to periods of 15–0.5 s. Segments of this length are necessary for an accurate estimate of the coherence between measured and wave-induced temperatures (appendix B). Although 10 min is longer than one eddy time scale, we will demonstrate that the resulting spectra display the expected spectral slope (appendix D). Further, by using the IC range here rather than higher-frequency subranges, we avoid having to correct for thermistor response at high frequencies.
In the main text, and in what follows, there are parallels with a recent study by Moulin et al. (2018), who limited their spectral calculations to frequencies below those of surface waves. They used frequencies corresponding to 60–15-s periods, which is similar to the band Gerbi et al. (2008) used to calculate heat and momentum fluxes from cospectra of temperature and velocities. Development of SurfOtter was motivated, in part, to improve on the measurements used by Moulin et al. (2018) with measurements that are less affected by waves. As we will show, these new measurements permit fitting temperature gradient spectra over a much larger frequency range. In addition, they yield a more accurate measurement of Tz. For ϵχ and
APPENDIX B
The Influence of Surface Waves on Temperature Records
By following the surface, SurfOtter minimizes the influence of surface waves on its temperature records, which is the primary obstacle to fitting spectra over the IC subrange. Although not completely eliminated, the remaining influence of waves on measured spectra can be predicted and subsequently removed. Here, we describe the reason for, and the method to predict, the wave-induced component of the measured temperature.
If SurfOtter perfectly follows the surface, a thermistor at a nominal depth d will always be this distance from the sea surface. However, wave orbital velocities induce a depth-dependent vertical advection of isotherms. Consequently, the thermistor traverses isotherms thus recording a variable temperature (Fig. B1). Here we describe two methods to predict the temperature signal introduced by surface waves for a thermistor at a distance d from the sea surface η(t). Both methods are based on a background temperature field advected by the wave orbital velocities.
Assume a monochromatic wave η(t) = A cos(−2πft). In a frame of reference in which the sea surface has a vertical position of z = 0 and is negative below, the position Z(z, t) of fluid parcels beneath the wave is given by
where k is the wavenumber.
The background temperature profile Tb(z) is assumed to be constant in time over several wave periods. It is described by a first-order Taylor expansion about z = −d:
Here, T0 and Tz are determined from a fit using two temperature sensors above z = −d and two below (or only one below for the deepest GusT). A second-order Taylor expansion is possible and was tested but makes minimal difference to the following analysis.
Vertical advection of Tb by the waves is equivalent to evaluating Eq. (B2) at Z from Eq. (B1):
In our frame of reference where z = 0 at the surface, the thermistor depth is constant, z = −d. Therefore, evaluating Eq. (B3) at z = −d gives the temperature that would be recorded by a thermistor beneath the monochromatic wave field:
Because we started by assuming z = 0 at the sea surface, we ensured that the depth decay term e−kd is always less than 1. Had we started by using 0 to define the mean water level, then the depth decay term would have been ek(η−d), which gives decay terms greater than 1 when η < d. Smit et al. (2017) provides a thorough discussion of these issues as well as a method that predicts wave kinematics near the surface that is accurate to second order. Their complete method is beyond the scope of this study, but we note that our approach is equivalent to the deep water limit of their first-order correction [their O(ϵ) solution].
To move from monochromatic waves to a broadband wave field, we invoke the Fourier transform F of η and the wavenumber to frequency conversion from linear wave theory:
Products involving η and the depth-decay term are undertaken in frequency space, and the result is converted back with the inverse Fourier transform F−1. This is effectively a linear filtering process (e.g., Wheeler 1970). Equation (B4) becomes
where zd = exp(−4π2f2d/g) is the (nondimensional) depth-decay term.
The second method to predict the temperature variance introduced by surface waves is to make use of the similarities between pressure and temperature measured by SurfOtter. The pressure expected at a distance d from the surface is given by Eq. (A4) of Hughes et al. (2020):
Note the similarity with Eq. (B7). The second and third terms are identical except for the coefficients: −ρg in Eq. (B8) and Tz in Eq. (B7). Therefore, we can predict the wave-induced temperature variability from the measured pressure as
where T0 is now a constant of integration, which goes to zero when we high-pass filter signals to remove frequencies lower than ~1/(30 s).
The quantity −pt/ρg is in some sense a vertical velocity. Indeed, on moorings,
We test the two methods on two 30-s segments during two periods for which surface waves clearly influence the temperature records (Fig. B2). To encompass extremes, the two examples use the shallowest and deepest GusTs. In both examples, the two predictions (the dark blue and black curves) are similar. For the shallow GusT, the predictions agree only somewhat with the measured temperature because the latter includes a nonnegligible turbulent component. Conversely, for the deeper GusT, the measured temperature is predominantly wave-induced variability. This deeper record is also smoother because of the depth decay of the higher frequencies. Of the two methods, the second method using the measured pressure is preferable and used in this paper. It is both easier to implement and advantageous in that is does not need to assume that sensors stay exactly at a fixed distance from the sea surface.
With the prediction of Twave established, its effect can be removed. A seemingly simple approach would be to subtract it from the measured temperature. Such a method, however, is sensitive to slight inaccuracies in the measurement of Tz or small phase differences between the measured p and T. Here, removing Twave spectrally is a superior option. To do so, we first calculate the spectral coherence between Tt and
Following Levine and Lueck (1999) and Zhang and Moum (2010), we define a corrected spectrum in terms of γ:
Results presented in this paper are derived from corrected spectra unless otherwise stated.
APPENDIX C
Uncertainty in ϵχ Values near the Surface
Implicit in the derivation of ϵχ [Eqs. (A1) and (A2)] is the assumption of a steady-state turbulent kinetic energy (TKE) balance between shear production Ps and removal by buoyancy production Jb and dissipation ϵ (Osborn 1980):
This assumes both temporal variability and vertical flux of TKE are negligible. The results in section 3 suggest the latter is the larger issue. Hence, we consider here the effect of a vertical TKE flux.
Divergence of F, the vertical TKE flux, would change the balance to
Parameter F can result from (i) a vertical redistribution or (ii) an input of TKE at the surface.
Based on simulations, we expect vertical redistribution to be small during conditions conducive to DWL formation. A large-eddy simulation by Wang (2001) under 4 m s−1 winds demonstrated a shear–buoyancy–dissipation balance. Turbulent transport was largely canceled by pressure transport (see their Fig. 1b). Klein and Coantic (1981) demonstrated the same balance with a second-order turbulence model (see the daytime panel of their Fig. 19).
Whether there exists an appreciable input of TKE at the surface is a matter of debate, especially in low-to-moderate winds. Brainerd and Gregg (1993) observed that, above the diurnal thermocline, ϵ was much larger than predicted by a wind-speed-dependent scaling (Lombardo and Gregg 1989), implying an input of TKE. Indeed, observations by Gerbi et al. (2009) from a tower in 16-m-deep water under winds averaging 7 m s−1 suggested that a TKE flux is necessary to close the turbulent energy budget. In developing DWL simulations, however, Pimentel et al. (2008) argued to forgo surface TKE fluxes. Similarly, using a free-rising turbulence profiler during 5 m s−1 winds, Callaghan et al. (2014) did not observe any persistently enhanced near-surface ϵ due to wave breaking.
Over the depths on which it acts, a surface TKE flux will cause ϵχ to underestimate the true ϵ. For example, using a second-order turbulence model with a surface TKE flux and a stabilizing buoyancy flux, Noh (1996) found that the shear–buoyancy–dissipation balance holds within the diurnal thermocline, but not above it (see his Fig. 4). In our case, this implies that ϵχ will underestimate ϵ near the surface but will be accurate within the region of strong Tz. Consequently, turbulent heat fluxes will also be accurate in this region, which is where they are strongest (section 4).
Our intent here is not to incorporate ∂F/∂z into the calculation of ϵχ but rather to estimate the depths affected by, and the magnitude of, a surface input of TKE insofar as it guides us on the depths over which to treat ϵχ as questionable. For this, Craig and Banner (1994) provide a convenient framework. Their intent was to account for breaking waves and so their expression may be questionable at low-to-moderate wind speed, but it does scale in a reasonable way
For an unstratified fluid, Craig and Banner (1994) considered two separate analytical models: balances between (i) a TKE flux and ϵ with a surface input and (ii) shear production and ϵ without a surface input. The former predicts a depth profile of ϵ [their Eqs. (25)–(27)] of
where n = 2.4 and z0 is set to 0.1 m. The parameter α defines the TKE flux at the sea surface:
A wide range of estimates for α (5–250) appear in the literature (e.g., Klein and Coantic 1981; Noh et al. 2004; Feddersen et al. 2007). Craig and Banner (1994) used α = 100.
The z−3.4 dependence in Eq. (C3) implies that the effect of a surface TKE flux decays rapidly. Craig and Banner (1994) found that for unstratified water under 9 m s−1 winds, the transition from the near-surface region with a predominantly diffusion–dissipation balance to a predominantly shear production–dissipation balance occurred at 0.6 m. We find similar depths when comparing ϵχ with Eq. (C3) (Fig. C1). The black curves denoting the analytical diffusion–dissipation balance exceed ϵχ in the top 0.5 m during the calm day and the top 1–2 m on the windy day.
Using an LES to simulate diurnal thermocline formation, Noh et al. (2009) showed that wave breaking has a much larger effect on the TKE budget than Langmuir circulation. Based on this study, uncertainty induced by Langmuir circulation, a process not considered here, is presumed to be second order compared to uncertainty associated with F.
Ultimately, our analysis suggests that (i) care is needed in interpreting ϵχ in approximately the top meter and (ii) turbulent diffusion of TKE from the surface can explain why ϵχ profiles do not increase approaching the surface as is observed when ϵ is measured by shear probes.
APPENDIX D
Comparing Spectra over Differing Frequency Ranges
Because SurfOtter reliably follows the surface (Hughes et al. 2020), we can obtain robust results by fitting spectra over a larger band than the 60–15-s band used by Moulin et al. (2018). In theory, a fit to the inertial–convective range could extend to frequencies up to those at which turbulence transitions to a viscous–convective regime. This occurs near a wavenumber k* (cpm) (Dillon and Caldwell 1980):
with values for the constants C* ≈ 0.05 and ν = 1 × 10−6 m2 s−1. To estimate a lower bound for k*, we use typical low values of ϵ = 10−8–10−6W kg−1 and oncoming flow speed of 0.4 m s−1. Together, these give k* = 3–8 cpm, or an equivalent frequency of f* = 1–3 Hz.
In choosing our upper frequency bound, in addition to our estimates of f*, we consider a previous study focused on temperature spectra beneath surface waves. Specifically, Stevens and Smith (2004) demonstrate −5/3 temperature spectra indicative of inertial–convective turbulence from a profiler rising vertically at 0.1 m s−1 beneath waves. Their frequency range of 1–10 Hz corresponds to a wavenumber range of 10–100 cpm. From this, it appears that the expression for k* above underestimates where the transition occurs. Ultimately, we do not expect our measured spectra to include any of the viscous–convective subrange if we apply an upper frequency bound of 2 Hz.
We choose 1/(15 s) as a low-frequency cutoff, rather than using 1/(60 s) as Moulin et al. (2018) did. There are two benefits to excluding the 60–15-s band here. First, we exclude length scales of several tens of meters, which are often outside the turbulent cascade. [Moulin et al. (2018) faced a trade-off between including these larger, potentially nonturbulent scales while still retaining a reasonable bandwidth over which to fit spectra.] Second, with separate bands we independently test the validity of the 60–15-s band. We will show this lower-frequency band to be adequate for capturing patterns of variability, albeit with ϵχ generally underestimated by a factor of 3–5, assuming our values from the higher-frequency band are the benchmark.
There is clear correlation between log10(ϵχ)calculated over the two different frequency bands. For all values from all thermistors for all 12 days in Table 1, R2 = 0.72 (Fig. D1a). At higher wind speeds (>6 m s−1; Fig. D1b), the magnitudes of the two ϵχ values are similar, with a median ratio of 3.0 (the 60–15-s band underestimates relative to the 15–0.5-s band). With moderate winds (3–6 m s−1; Fig. D1c), the median ratio increases to 5.0. This increase occurs because turbulent length scales decrease with decreasing wind, so the 60–15-s band now includes a larger portion of nonturbulent scales, which have an associated lower spectral level. For example, Figs. 4e–h show that the measured spectrum starts to fall off relative to the +1/3 slope with decreasing frequency below ~1/(10 s). The median does not continue to increase moving to lower winds (<3 m s−1; Fig. D1d). Rather, the distribution of the ratios widens indicative of a weaker correlation between ϵχ from the two bands.
For 10-min segments, the 15–0.5-s band encompasses 1160 spectral coefficients. Because coefficients are linearly spaced in frequency, 600 of the 1160 are within 1–2 Hz. Conversely, the 60–15-s band encompasses only 30. For the higher-frequency band, the distribution of slopes from linear fits to
The tight distribution of slopes in the higher-frequency band makes it possible, if desired, to use noticeable deviations from this slope as a flag for periods when the fit is questionable. For example, between 1000 and 1600 LT on the calm day, the best-fit slope oscillates noticeably (Fig. D2a). During this time, the GusT at 1.7 m deep is below
REFERENCES
Anis, A., and J. N. Moum, 1995: Surface wave–turbulence interactions. Scaling ε(z) near the sea surface. J. Phys. Oceanogr., 25, 2025–2045, https://doi.org/10.1175/1520-0485(1995)025<2025:SWISNT>2.0.CO;2.
Asher, W. E., A. T. Jessup, R. Branch, and D. Clark, 2014: Observations of rain-induced near-surface salinity anomalies. J. Geophys. Res. Oceans, 119, 5483–5500, https://doi.org/10.1002/2014JC009954.
Batchelor, G. K., 1959: Small-scale variation of convected quantities like temperature in turbulent fluid. Part I. General discussion and the case of small conductivity. J. Fluid Mech., 5, 113–133, https://doi.org/10.1017/S002211205900009X.
Becherer, J., and J. N. Moum, 2017: An efficient scheme for onboard reduction of moored χpod data. J. Atmos. Oceanic Technol., 34, 2533–2546, https://doi.org/10.1175/JTECH-D-17-0118.1.
Becherer, J., J. N. Moum, J. A. Colosi, J. A. Lerczak, and J. M. McSweeney, 2020: Turbulence asymmetries in bottom boundary layer velocity pulses associated with onshore-propagating nonlinear internal waves. J. Phys. Oceanogr., 50, 2373–2391, https://doi.org/10.1175/JPO-D-19-0178.1.
Bellenger, H., and J.-P. Duvel, 2009: An analysis of tropical ocean diurnal warm layers. J. Climate, 22, 3629–3646, https://doi.org/10.1175/2008JCLI2598.1.
Bourras, D., and Coauthors, 2014: A new platform for the determination of air–sea fluxes (OCARINA): Overview and first results. J. Atmos. Oceanic Technol., 31, 1043–1062, https://doi.org/10.1175/JTECH-D-13-00055.1.
Brainerd, K. E., and M. C. Gregg, 1993: Diurnal restratification and turbulence in the oceanic surface mixed layer: 1. Observations. J. Geophys. Res., 98, 22 645–22 656, https://doi.org/10.1029/93JC02297.
Caldwell, D. R., R.-C. Lien, J. N. Moum, and M. C. Gregg, 1997: Turbulence decay and restratification in the equatorial ocean surface layer following nighttime convection. J. Phys. Oceanogr., 27, 1120–1132, https://doi.org/10.1175/1520-0485(1997)027<1120:TDARIT>2.0.CO;2.
Callaghan, A. H., B. Ward, and J. Vialard, 2014: Influence of surface forcing on near-surface and mixing layer turbulence in the tropical Indian Ocean. Deep-Sea Res. I, 94, 107–123, https://doi.org/10.1016/j.dsr.2014.08.009.
Craig, P. D., and M. L. Banner, 1994: Modeling wave-enhanced turbulence in the ocean surface layer. J. Phys. Oceanogr., 24, 2546–2559, https://doi.org/10.1175/1520-0485(1994)024<2546:MWETIT>2.0.CO;2.
Deardorff, J. W., 1966: The counter-gradient heat flux in the lower atmosphere and in the laboratory. J. Atmos. Sci., 23, 503–506, https://doi.org/10.1175/1520-0469(1966)023<0503:TCGHFI>2.0.CO;2.
Dillon, T. M., and T. M. Powell, 1979: Observations of a surface mixed layer. Deep-Sea Res., 26A, 915–932, https://doi.org/10.1016/0198-0149(79)90105-5.
Dillon, T. M., and D. R. Caldwell, 1980: The Batchelor spectrum and dissipation in the upper ocean. J. Geophys. Res., 85, 1910–1916, https://doi.org/10.1029/JC085iC04p01910.
Fairall, C. W., E. F. Bradley, J. S. Godfrey, G. A. Wick, J. B. Edson, and G. S. Young, 1996: Cool-skin and warm-layer effects on sea surface temperature. J. Geophys. Res., 101, 1295–1308, https://doi.org/10.1029/95JC03190.
Feddersen, F., J. H. Trowbridge, and A. J. Williams, 2007: Vertical structure of dissipation in the nearshore. J. Phys. Oceanogr., 37, 1764–1777, https://doi.org/10.1175/JPO3098.1.
Gargett, A. E., 1988: The scaling of turbulence in the presence of stable stratification. J. Geophys. Res., 93, 5021–5036, https://doi.org/10.1029/JC093iC05p05021.
Gargett, A. E., T. R. Osborn, and P. W. Nasmyth, 1984: Local isotropy and the decay of turbulence in a stratified fluid. J. Fluid Mech., 144, 231–280, https://doi.org/10.1017/S0022112084001592.
Gentemann, C. L., P. J. Minnett, and B. Ward, 2009: Profiles of Ocean Surface Heating (POSH): A new model of upper ocean diurnal warming. J. Geophys. Res., 114, C07017, https://doi.org/10.1029/2008JC004825.
Gerbi, G. P., J. H. Trowbridge, J. B. Edson, A. J. Plueddemann, E. A. Terray, and J. J. Fredericks, 2008: Measurements of momentum and heat transfer across the air–sea interface. J. Phys. Oceanogr., 38, 1054–1072, https://doi.org/10.1175/2007JPO3739.1.
Gerbi, G. P., J. H. Trowbridge, E. A. Terray, A. J. Plueddemann, and T. Kukulka, 2009: Observations of turbulence in the ocean surface boundary layer: Energetics and transport. J. Phys. Oceanogr., 39, 1077–1096, https://doi.org/10.1175/2008JPO4044.1.
Goodman, L., E. R. Levine, and R. G. Lueck, 2006: On measuring the terms of the turbulent kinetic energy budget from an AUV. J. Atmos. Oceanic Technol., 23, 977–990, https://doi.org/10.1175/JTECH1889.1.
Greenan, B. J. W., N. S. Oakey, and F. W. Dobson, 2001: Estimates of dissipation in the ocean mixed layer using a quasi-horizontal microstructure profiler. J. Phys. Oceanogr., 31, 992–1004, https://doi.org/10.1175/1520-0485(2001)031<0992:EODITO>2.0.CO;2.
Halpern, D., and R. K. Reed, 1976: Heat budget of the upper ocean under light winds. J. Phys. Oceanogr., 6, 972–975, https://doi.org/10.1175/1520-0485(1976)006<0972:HBOTUO>2.0.CO;2.
Hodges, B. A., and D. M. Fratantoni, 2014: AUV observations of the diurnal surface layer in the North Atlantic salinity maximum. J. Phys. Oceanogr., 44, 1595–1604, https://doi.org/10.1175/JPO-D-13-0140.1.
Hughes, K. G., J. N. Moum, and E. L. Shroyer, 2020: Evolution of the velocity structure in the diurnal warm layer. J. Phys. Oceanogr., 50, 615–631, https://doi.org/10.1175/JPO-D-19-0207.1.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1644, https://doi.org/10.1175/BAMS-83-11-1631.
Kantha, L. H., and C. A. Clayson, 1994: An improved mixed layer model for geophysical applications. J. Geophys. Res., 99, 25 235–25 266, https://doi.org/10.1029/94JC02257.
Klein, P., and M. Coantic, 1981: A numerical study of turbulent processes in the marine upper layers. J. Phys. Oceanogr., 11, 849–863, https://doi.org/10.1175/1520-0485(1981)011<0849:ANSOTP>2.0.CO;2.
Klymak, J. M., and J. N. Moum, 2007: Oceanic isopycnal slope spectra. Part II: Turbulence. J. Phys. Oceanogr., 37, 1232–1245, https://doi.org/10.1175/JPO3074.1.
Kondo, J., Y. Sasano, and T. Ishii, 1979: On wind-driven current and temperature profiles with diurnal period in the oceanic planetary boundary layer. J. Phys. Oceanogr., 9, 360–372, https://doi.org/10.1175/1520-0485(1979)009<0360:OWDCAT>2.0.CO;2.
Kudryavtsev, V. N., and A. V. Soloviev, 1990: Slippery near-surface layer of the ocean arising due to daytime solar heating. J. Phys. Oceanogr., 20, 617–628, https://doi.org/10.1175/1520-0485(1990)020<0617:SNSLOT>2.0.CO;2.
Large, W. G., J. C. McWilliams, and S. C. Doney, 1994: Oceanic vertical mixing: A review and a model with a nonlocal boundary layer parameterization. Rev. Geophys., 32, 363–403, https://doi.org/10.1029/94RG01872.
Levine, E. R., and R. G. Lueck, 1999: Turbulence measurement from an autonomous underwater vehicle. J. Atmos. Oceanic Technol., 16, 1533–1544, https://doi.org/10.1175/1520-0426(1999)016<1533:TMFAAU>2.0.CO;2.
Lombardo, C. P., and M. C. Gregg, 1989: Similarity scaling of viscous and thermal dissipation in a convecting surface boundary layer. J. Geophys. Res., 94, 6273–6284, https://doi.org/10.1029/JC094iC05p06273.
Lueck, R., 2016: Calculating the rate of dissipation of turbulent kinetic energy. Tech. Note TN-028, 18 pp., http://rocklandscientific.com/?wpdmdl=1034.
Lukas, R., 1991: The diurnal cycle of sea surface temperature in the western equatorial Pacific. TOGA Notes, No. 2, Nova Southeastern University, Davie, FL, 1–5.
Martin, P. J., 1985: Simulation of the mixed layer at OWS November and Papa with several models. J. Geophys. Res., 90, 903–916, https://doi.org/10.1029/JC090iC01p00903.
Matthews, A. J., D. B. Baranowski, K. J. Heywood, P. J. Flatau, and S. Schmidtko, 2014: The surface diurnal warm layer in the Indian Ocean during CINDY/DYNAMO. J. Climate, 27, 9101–9122, https://doi.org/10.1175/JCLI-D-14-00222.1.
Moulin, A. J., J. N. Moum, and E. L. Shroyer, 2018: Evolution of turbulence in the diurnal warm layer. J. Phys. Oceanogr., 48, 383–396, https://doi.org/10.1175/JPO-D-17-0170.1.
Moum, J. N., 1997: Quantifying vertical fluxes from turbulence in the ocean. Oceanography, 10, 111–115, https://doi.org/10.5670/oceanog.1997.02.
Moum, J. N., and D. R. Caldwell, 1985: Local influences on shear-flow turbulence in the equatorial ocean. Science, 230, 315–316, https://doi.org/10.1126/science.230.4723.315.
Moum, J. N., and J. D. Nash, 2009: Mixing measurements on an equatorial ocean mooring. J. Atmos. Oceanic Technol., 26, 317–336, https://doi.org/10.1175/2008JTECHO617.1.
Noh, Y., 1996: Dynamics of diurnal thermocline formation in the oceanic mixed layer. J. Phys. Oceanogr., 26, 2183–2195, https://doi.org/10.1175/1520-0485(1996)026<2183:DODTFI>2.0.CO;2.
Noh, Y., H. S. Min, and S. Raasch, 2004: Large eddy simulation of the ocean mixed layer: The effects of wave breaking and Langmuir circulation. J. Phys. Oceanogr., 34, 720–735, https://doi.org/10.1175/1520-0485(2004)034<0720:LESOTO>2.0.CO;2.
Noh, Y., G. Goh, S. Raasch, and M. Gryschka, 2009: Formation of a diurnal thermocline in the ocean mixed layer simulated by LES. J. Phys. Oceanogr., 39, 1244–1257, https://doi.org/10.1175/2008JPO4032.1.
Osborn, T. R., 1980: Estimates of the local rate of vertical diffusion from dissipation measurements. J. Phys. Oceanogr., 10, 83–89, https://doi.org/10.1175/1520-0485(1980)010<0083:EOTLRO>2.0.CO;2.
Paulson, C. A., and J. J. Simpson, 1977: Irradiance measurements in the upper ocean. J. Phys. Oceanogr., 7, 952–956, https://doi.org/10.1175/1520-0485(1977)007<0952:IMITUO>2.0.CO;2.
Paulson, C. A., and J. J. Simpson, 1981: The temperature difference across the cool skin of the ocean. J. Geophys. Res., 86, 11 044–11 054, https://doi.org/10.1029/JC086iC11p11044.
Pham, H. T., and S. Sarkar, 2017: Turbulent entrainment in a strongly stratified barrier layer. J. Geophys. Res. Oceans, 122, 5075–5087, https://doi.org/10.1002/2016JC012357.
Pimentel, S., K. Haines, and N. K. Nichols, 2008: Modeling the diurnal variability of sea surface temperatures. J. Geophys. Res., 113, C11004, https://doi.org/10.1029/2007JC004607.
Pollard, R. T., P. B. Rhines, and R. O. R. Y. Thompson, 1972: The deepening of the wind-mixed layer. Geophys. Astrophys. Fluid Dyn., 4, 381–404, https://doi.org/10.1080/03091927208236105.
Price, J. F., R. A. Weller, and R. Pinkel, 1986: Diurnal cycling: Observations and models of the upper ocean response to diurnal heating, cooling, and wind mixing. J. Geophys. Res., 91, 8411–8427, https://doi.org/10.1029/JC091iC07p08411.
Pujiana, K., J. N. Moum, and W. D. Smyth, 2018: The role of turbulence in redistributing upper-ocean heat, freshwater, and momentum in response to the MJO in the equatorial Indian Ocean. J. Phys. Oceanogr., 48, 197–220, https://doi.org/10.1175/JPO-D-17-0146.1.
Sarkar, S., and H. Pham, 2019: Turbulence and thermal structure in the upper ocean: Turbulence-resolving simulations. Flow Turbul. Combust., 103, 985–1009, https://doi.org/10.1007/s10494-019-00065-5.
Seager, R., S. E. Zebiak, and M. A. Cane, 1988: A model of the tropical Pacific sea surface temperature climatology. J. Geophys. Res., 93, 1265–1280, https://doi.org/10.1029/JC093iC02p01265.
Smit, P. B., T. T. Janssen, and T. H. C. Herbers, 2017: Nonlinear wave kinematics near the ocean surface. J. Phys. Oceanogr., 47, 1657–1673, https://doi.org/10.1175/JPO-D-16-0281.1.
Smyth, W. D., P. O. Zavialov, and J. N. Moum, 1997: Decay of turbulence in the upper ocean following sudden isolation from surface forcing. J. Phys. Oceanogr., 27, 810–822, https://doi.org/10.1175/1520-0485(1997)027<0810:DOTITU>2.0.CO;2.
Soloviev, A., and R. Lukas, 1997: Observation of large diurnal warming events in the near-surface layer of the western equatorial Pacific warm pool. Deep-Sea Res. I, 44, 1055–1076, https://doi.org/10.1016/S0967-0637(96)00124-0.
Stevens, C. L., and M. J. Smith, 2004: Temperature microstructure beneath surface gravity waves. J. Atmos. Oceanic Technol., 21, 1747–1757, https://doi.org/10.1175/JTECH1659.1.
Stillinger, D. C., K. N. Helland, and C. W. Van Atta, 1983: Experiments on the transition of homogeneous turbulence to internal waves in a stratified fluid. J. Fluid Mech., 131, 91–122, https://doi.org/10.1017/S0022112083001251.
St. Laurent, L., and S. Merrifield, 2017: Measurements of near-surface turbulence and mixing from autonomous ocean gliders. Oceanography, 30, 116–125, https://doi.org/10.5670/oceanog.2017.231.
Stramma, L., P. Cornillon, R. A. Weller, J. F. Price, and M. G. Briscoe, 1986: Large diurnal sea surface temperature variability: Satellite and in situ measurements. J. Phys. Oceanogr., 16, 827–837, https://doi.org/10.1175/1520-0485(1986)016<0827:LDSSTV>2.0.CO;2.
Stuart-Menteth, A. C., I. S. Robinson, and P. G. Challenor, 2003: A global study of diurnal warming using satellite-derived sea surface temperature. J. Geophys. Res., 108, 3155, https://doi.org/10.1029/2002JC001534.
Sutherland, G., B. Ward, and K. H. Christensen, 2013: Wave-turbulence scaling in the ocean mixed layer. Ocean Sci., 9, 597–608, https://doi.org/10.5194/os-9-597-2013.
Sutherland, G., L. Marié, G. Reverdin, K. H. Christensen, G. Broström, and B. Ward, 2016: Enhanced turbulence associated with the diurnal jet in the ocean surface boundary layer. J. Phys. Oceanogr., 46, 3051–3067, https://doi.org/10.1175/JPO-D-15-0172.1.
Thompson, E. J., J. N. Moum, C. W. Fairall, and S. A. Rutledge, 2019: Wind limits on rain layers and diurnal warm layers. J. Geophys. Res. Oceans, 124, 897–924, https://doi.org/10.1029/2018JC014130.
Thomson, J., 2012: Wave breaking dissipation observed with “SWIFT” drifters. J. Atmos. Oceanic Technol., 29, 1866–1882, https://doi.org/10.1175/JTECH-D-12-00018.1.
Thomson, J., and Coauthors, 2019: A new version of the SWIFT platform for waves, currents, and turbulence in the ocean surface layer. 2019 IEEE/OES Twelfth Current, Waves and Turbulence Measurement and Application Workshop, San Diego, CA, IEEE, 1–7, https://doi.org/10.1109/CWTM43797.2019.8955299.
Van Roekel, L., and Coauthors, 2018: The KPP boundary layer scheme for the ocean: Revisiting its formulation and benchmarking one-dimensional simulations relative to LES. J. Adv. Model. Earth Syst., 10, 2647–2685, https://doi.org/10.1029/2018MS001336.
Wade, M., G. Caniaux, Y. duPenhoat, M. Dengler, H. Giordani, and R. Hummels, 2011: A one-dimensional modeling study of the diurnal cycle in the equatorial Atlantic at the PIRATA buoys during the EGEE-3 campaign. Ocean Dyn., 61, 1–20, https://doi.org/10.1007/s10236-010-0337-8.
Wang, D., 2001: Large-eddy simulation of the diurnal cycle of oceanic boundary layer: Sensitivity to domain size and spatial resolution. J. Geophys. Res., 106, 13 959–13 974, https://doi.org/10.1029/2001JC000896.
Webster, P. J., and R. Lukas, 1992: TOGA COARE: The coupled ocean–atmosphere response experiment. Bull. Amer. Meteor. Soc., 73, 1377–1416, https://doi.org/10.1175/1520-0477(1992)073<1377:TCTCOR>2.0.CO;2.
Webster, P. J., C. A. Clayson, and J. A. Curry, 1996: Clouds, radiation, and the diurnal cycle of sea surface temperature in the tropical western Pacific. J. Climate, 9, 1712–1730, https://doi.org/10.1175/1520-0442(1996)009<1712:CRATDC>2.0.CO;2.
Wheeler, J. D., 1970: Method for calculating forces produced by irregular waves. J. Pet. Technol., 22, 359–367, https://doi.org/10.2118/2712-PA.
Yamazaki, H., 1990: Stratified turbulence near a critical dissipation rate. J. Phys. Oceanogr., 20, 1583–1598, https://doi.org/10.1175/1520-0485(1990)020<1583:STNACD>2.0.CO;2.
Zhang, Y., and J. N. Moum, 2010: Inertial-convective subrange estimates of thermal variance dissipation rate from moored temperature measurements. J. Atmos. Oceanic Technol., 27, 1950–1959, https://doi.org/10.1175/2010JTECHO746.1.