Professional Documents
Culture Documents
Review
(Received 6 November 2006; in revised form 2 April 2007; accepted 14 April 2007)
The importance of the diurnal variability of sea surface temperature (SST) on air-sea Keywords:
interaction is now being increasingly recognized. This review synthesizes knowledge Sea surface
of the diurnal SST variation, mainly paying attention to its impact on the atmosphere temperature,
or the ocean. Diurnal SST warming becomes evident when the surface wind is weak diurnal variation,
intraseasonal
and insolation is strong. Recent observations using satellite data and advanced in-
variation,
struments have revealed that a large diurnal SST rise occurs over wide areas in a
air-sea interaction,
specific season, and in an extreme case the diurnal amplitude of SST exceeds 5 K. The surface flux.
large diurnal SST rise can lead to an increase in net surface heat flux from the ocean
of 5060 Wm2 in the daytime. The temporal mean of the increase exceeds 10 Wm2,
which will be non-negligible for the atmosphere. A few numerical experiments have
indicated that the diurnal SST variation can modify atmospheric properties over the
Pacific warm pool or a coastal sea, but the processes underlying the modification
have not yet been investigated in detail. Furthermore, it has been shown that the
diurnal change of ocean mixing process near the surface must be considered cor-
rectly in order to reproduce SST variations on an intraseasonal scale in a numerical
model. The variation of mixed-layer properties on the daily scale is nonlinearly re-
lated to the intraseasonal variability. The mixed-layer deepening/shoaling process on
the daily scale will also be related to biological and material circulation processes.
721
Diurnal variation, which is caused by solar radiation called the warm layer (Fairall et al., 1996). The
and the earths rotation, is one of the dominant variations thermocline near the surface that develops only in the
in SST. The existence of the diurnal variation in SST was daytime is referred to as diurnal thermocline. Further-
known at least one century ago (cf. Sverdrup et al., 1942; more, at the top of the ocean, a very thin cool layer, which
Roll, 1965). Sverdrup et al. (1942) indicated that the is usually called the thermal skin layer, cool skin layer
diurnal variation of sea temperature in general is so small or simply skin layer, almost always exists (e.g.,
that it is of little importance to physical and biological Saunders, 1967; Katsaros, 1980; Robinson et al., 1984;
processes in the sea, but knowledge of the small varia- Ward and Donelan, 2006). The thickness of this layer is
tions is essential to the study of the diurnal exchange of usually the order of 0.11 mm, and the temperature at the
heat between the atmosphere and the sea. The diurnal top of the skin layer is generally several tenths of a de-
amplitude of SST is O (0.1 K) on average, but often gree colder than the temperature below it because eddy
reaches a few degrees and can exceed 5 K in extreme diffusion becomes less than molecular diffusion just close
cases (e.g., Flament et al., 1994; Yokoyama et al., 1995). to the surface. The phenomenon of the temperature drop
The satellite remote sensing community has clearly rec- occurring in this thin layer is called the skin effect.
ognized that the diurnal SST variability has to be ad- While the diurnal thermocline vanishes by sunrise next
equately considered to provide better accuracy of satel- morning, the skin layer usually exists in both the daytime
lite-derived SST (e.g., Hepplewhite, 1989; Wick et al., and nighttime, even in windy conditions (Donlon et al.,
2002; Donlon and the GHRSST-PP Science Team, 2005; 2002). In the daytime, the temperature difference across
Notarstefano et al., 2006). However, the diurnal varia- the skin layer becomes smaller due to the absorption of
tion has often been neglected in analytical SST datasets shortwave radiation in this layer. The temperature at the
or numerical modeling, and the effects of air-sea interac- top can theoretically even become higher than that at the
tion on the daily scale have not yet been widely recog- bottom when the air temperature is very much higher than
nized nor adequately revealed. that of the water, as has been confirmed by a tank experi-
Many researchers have recently become interested ment (Ward and Donelan, 2006). The skin layer thick-
in the diurnal SST variation, and increasing numbers of ness and the diurnal thermocline depth vary with surface
studies pay attention to air-sea interaction on the daily heat and momentum fluxes (e.g., Fairall et al., 1996; Ward
scale. This review synthesizes knowledge of the diurnal and Donelan, 2006). A better knowledge of SST demands
SST variation, focusing on why and how the diurnal vari- consideration of both the diurnal warming near the sur-
ation is important for the atmosphere and the ocean. face and the effect of the skin layer.
Soloviev and Lukas (2006) gave a detailed, comprehen- A sharp temperature gradient sometimes appears
sive account of the structure and dynamics of the near- above 1-m depth in the daytime (e.g., Yokoyama et al.,
surface ocean. Furthermore, previous studies of the skin 1995; Soloviev and Lukas, 1997; Ward, 2006; see Figs.
effect (see Section 2) were reviewed or introduced by 13 in the present paper). The large temperature differ-
Katsaros (1980), Robinson et al. (1984), and Ward et al. ence between the sea surface and about 1-m depth, where
(2004a), for example. In this paper we focus on the diur- ships and buoys usually measure the seawater tempera-
nal sea surface warming and its effect on the atmosphere, ture as SST, has been recognized as one of the major
with less emphasis on the skin effect. We do not take up sources of error in satellite-derived SST. The satellite in-
the diurnal variation caused by the tide here. frared sensor, microwave sensor, and in situ sensors ob-
First, the vertical temperature structure near the sur- serve different sea surfaces, i.e., skin, subskin, and a
face and the definition of SST are explained in Section 2. depth of one or a few meters, respectively. Hence it was
Section 3 summarizes previous observational studies that indispensable to define SST exactly and consider careful
showed large diurnal variations in SST. Section 4 intro- treatment of different SSTs when producing an accurate
duces several models used to simulate the diurnal SST SST dataset. The Global Ocean Data Assimilation Ex-
variation. Sections 5 and 6 discuss how the diurnal vari- periment (GODAE) High-Resolution SST Pilot Project
ability of SST affects the surface flux estimation and the (GHRSST-PP) Science Team has been doing work on
atmosphere. Some main issues on modeling and observa- coordinating a new generation of global, multi-sensor,
tion of the diurnal SST variation are mentioned in Sec- high-resolution SST products for the benefit of the op-
tion 7. The paper is summarized in Section 8. erational and scientific community (Donlon et al., 2007).
This Science Team has defined five kinds of SST:
2. Vertical Thermal Structure near the Sea Surface interface SST (SSTint), skin SST (SSTskin), subskin SST
and SST Terminology (SSTsubskin), sea temperature at depth (SSTdepth), and foun-
Under clear and calm conditions, thermal stratifica- dation SST (SSTfnd) (Donlon and the GHRSST-PP Sci-
tion is formed within the top few meters of the ocean due ence Team, 2005). A schematic picture of the vertical tem-
to strong insolation. This diurnal stratified layer is often perature profile is shown in Fig. 1, and the definitions of
the SSTs are explained in Table 1. In actuality, we cannot SST (Notarstefano et al., 2006). We should note that the
know SSTint even with current technology (Donlon et al., satellite SST tuned against SSTdepth has the above char-
2002), and SSTskin is usually utilized as a substitute for acteristics. For the recent satellite SST products from the
SSTint on the assumption that SSTskin is close enough to Along Track Scanning Radiometer (ATSR) (Mutlow et
the true SSTint. The in situ SST measured at about 1-m al., 1994) and the Tropical Rainfall Measurement Mis-
depth or deeper has been called bulk SST. The Science sion (TRMM) satellites Microwave Imager (TMI)
Team recommends using SSTdepth rather than the con- (Gentemann et al., 2004), the algorithms were developed
ventional term bulk SST referring to an in situ SST based on the radiance simulated by a radiative transfer
measurement made at 1-m depth as SST1m, for example. model in order to derive SSTskin or SST subskin exactly.
This terminology is introduced to encourage reporting of The atmosphere senses only the exact interface be-
the measurement depth along with the temperature, be- tween the atmosphere and the ocean. Hence we have to
cause, as depicted in Fig. 1, the temperature can change know SSTint (or SSTskin practically) and its diurnal varia-
drastically with depth when the diurnal thermocline is tion for accurate estimation of air-sea heat and gas fluxes
formed. The new concept of foundation SST is intro- (e.g., Sarmiento and Sundquist, 1992; Fairall et al., 1996).
duced as a more precise, well-defined quantity instead of If the temperature at a few meters depth is used as SSTint
previous, loosely-defined bulk SST, which is affected in a flux calculation, the atmosphere will receive incor-
by the diurnal warming. SST fnd will be similar to a rect heat and water vapor from the ocean. This impact is
nighttime minimum or pre-dawn value at depths of ~15 not always negligible, as discussed in Sections 5 and 6.
m, but note that this depth is only a rough estimate and
can deviate from this range in some cases. This paper 3. Observational Facts Concerning Diurnal SST
basically adopts the terminology proposed by Donlon and VariationWhen and How can the Diurnal Vari-
the GHRSST-PP Science Team (2005). When referring ation Become Large?
to the temperature near the surface in a general sense,
loosely, the authors simply use SST. 3.1 In situ observation
While a satellite sensor sees SSTskin or SST subskin, in Sverdrup et al. (1942) and Roll (1965) introduced
situ SST observed from ships and buoys is the tempera- some early observational studies of diurnal SST varia-
ture at about 1-m depth or deeper. Algorithms of satel- tion. These early studies used research vessel SST data,
lite-derived SST are conventionally tuned by using buoy- which correspond to SSTdepth, and the depths of the meas-
observed SST, i.e., SSTdepth , as the truth. Hence the aver- urements were not mentioned. For example, according to
age of the satellite SST agrees with that of SSTdepth. How- them, a 1923 report based on data of German and British
ever, it is expected that the variability of the satellite SST research vessels obtained during 18721903 stated that
reflects that of SSTskin or SST subskin, rather than SST depth mean diurnal amplitude of SST in the low latitude was
(cf. Kearns et al., 2000; Kilpatrick et al., 2001; Wick et about 0.30.4 K, and the amplitude decreased to 0.26 K
al., 2002; Stuart-Menteth et al., 2003; Dong et al., 2006). at 4555S and 0.15 K at 5560S. Koizumi (1956)
Some researchers call such satellite SST pseudo-bulk analyzed ocean station data at Extra (39N, 153E) and
Fig. 5. Seasonal mean day-night difference of the AMSR-E ver. 5 SST produced by Remote Sensing Systems during June 2002
May 2006. Original grid size of the SST is 0.25, and the seasonal mean is calculated in 1 grids. Nominal observation time is
approximately 0130/1330 LST.
SSTskin
1 24 12 i + 1
= f + a( PS) + b( P) + c[ln(U )] + d ( PS) ln(U ) + e(U ), Ti, j = i 1, j i , j 1 1 , (3)
(2 ) i =1
12 12
( ) ( )
i, j = f + a Si, j + c ln Ui, j Table 2. Coefficients for the empirical formula (2). Reproduced
from Webster et al. (1996).
Si. j > 0
+ d ( Si, j ) ln(Ui, j ) + e(Ui, j ), ( 4)
i, j = 0, Si. j 0, Coefficient U > 2 m s1 U 2 m s1
f 0.262 0.328
a 0.00265 0.002
where Si,j and Ui,j is the hourly surface solar radiation b 0.028 0.041
and wind speed at the i-th hour on the j-th day, respec- c 0.838 0.212
tively. The precipitation is ignored here. The time of lo- d 0.00105 0.000185
24 e 0.158 0.329
cal sunset is defined as i = 1, and i , j 1 is the sum of
i =1
on the previous day. They showed that their
parameterization could well reproduce diurnal SST skin
variations. Since these simple schemes introduced here showed an example in the north Atlantic where monthly
use daily mean meteorological values and/or previous- mean SST was higher by about 0.2 K in the case that di-
day information, it may be difficult to utilize these urnal SST variations were included than in the case that
schemes for discussing simultaneous air-sea interaction they were ignored. This difference reduces the net heat
processes in a coupled model. However, they will be use- flux of 5 W m2 entering the ocean.
ful to test the sensitivity of the atmosphere to the diurnal Fairall et al. (1996) showed that the cool skin de-
variation of SST. creased the net heat flux from the ocean by about 11
W m2 and the warm layer increased it by about 4 W m2
5. Effect of Diurnal Thermocline on Air-Sea Flux on average over 70 days during COARE. The effect of
Estimations the warm layer may seem to be fairly small, but this was
the mean value including the cases when the warm layer
5.1 Air-sea heat flux estimation did not develop. Fairall et al. (1996) and Ward (2006)
Sverdrup et al. (1942) suggested the importance of indicated that the net heat flux from the ocean can in-
the diurnal SST variability on air-sea heat exchange. Since crease 5060 W m2 in the daytime under calm and clear
the atmosphere contacts the sea skin, not the water at a conditions due to the effect of the warm layer (Fig. 3).
few meters depth, evaluating the temperature difference Clayson and Curry (1996) estimated surface turbulent heat
across the warm layer and skin layer is indispensable for fluxes during COARE from satellite data and compared
accurate air-sea heat flux estimation. An error of 1 K in them with in situ measurements. They showed that deter-
SSTskin can lead to an error of 27 W m 2 in net surface mining the fluxes every 3 h from interpolated satellite-
heat flux in the tropical western Pacific (Webster et al., derived input variables, i.e., including diurnal cycles in
1996). Furthermore, Cornillon and Stramma (1985) SSTskin and atmospheric variables, improved the estimate
of daily-mean surface fluxes. However, they did not evalu- this scheme increased the latent heat flux during the day-
ate the effect of the diurnal variation of SSTskin on the time by 1020 W m2 and reduced it in the nighttime by
fluxes. 05 W m2. The increase in time-mean net heat loss of
Zeng and Dickinson (1998) investigated the impact the ocean was about 10 W m2. Zeng and Beljaars (2005)
of the diurnal variation of SSTskin on surface fluxes over reported that incorporating their SSTskin scheme (see Sub-
the equatorial Pacific using TOGA Tropical Atmosphere section 4.2) into an operational forecasting model changed
Ocean (TAO) buoy hourly data from 95W to 137E in ensemble annual mean surface latent heat flux by more
19901996. The surface latent and sensible heat fluxes than 10 W m2 over several regions in the north Atlantic.
showed clear diurnal variability, and the average diurnal The authors also checked the impact of the warm
amplitudes of the latent and sensible heat fluxes were 19.7 layer on the air-sea heat transfer by a simple numerical
and 5.6 W m2, respectively. They also calculated the heat experiment. The test data used here were obtained with a
fluxes by replacing hourly SSTskin with daily or monthly moored buoy of the Triangle Trans-Ocean buoy Network
mean SSTskin. Figure 9 shows the examples of the differ- (TRITON) in the western tropical Pacific at 2.07N,
ences between the heat fluxes with the hourly SSTskin and 138.06E during 313 March 2004. The authors simu-
those with the daily or monthly SSTskin. Evidently the lated near-surface temperature using the buoy-observed
diurnal cycle in SSTskin is the main cause of the diurnal meteorological data and Kawai and Kawamuras (2000)
variability of latent and sensible heat fluxes. They sug- one-dimensional model. The details of the observations
gested that numerical modeling may require the inclu- and the model simulation are reported in Kawai et al.
sion of the diurnal SSTskin variation. Parsons et al. (2000) (2006b). The diurnal temperature variation was large on
also showed that the average diurnal amplitudes of sur- 3, 7 and 9 March, and the model could reproduce the vari-
face sensible and latent heat fluxes were about 4 and 35 ation of SST0.3m approximately (Kawai et al., 2006b). This
W m2, respectively, within the inner intensive flux array simulation result of SSTskin is shown in Fig. 10(a) with a
of the COARE experiment in mid-November 1992. solid line, and is called the control run. The layer be-
Schiller and Godfrey (2005) examined the effect of tween the sea surface and 1.5-m depth was then forcibly
the diurnal variability of SST on surface latent heat flux mixed by setting the minimum of the eddy diffusion co-
during COARE using a one-dimensional coupled model efficients to 5.0 10 4 m2s1 above 1.5-m depth. This
with their sublayer scheme (see Subsection 4.2). Use of simulation is referred to as the no-warm-layer run (bro-
diurnal cycle of insolation tends to be higher than with- tropics cannot be reproduced if an inappropriate
out it. parameterization for shortwave radiation is used.
Figure 16 shows the result of a numerical experi- An approximate formula of a polynomial exponen-
ment performed by Bernie et al. (2005), who simulated tial function (e.g., Paulson and Simpson, 1977) consider-
mixed layer depth and temperature with hourly surface ing only the simple classification of water type (e.g.,
fluxes and with daily mean fluxes, respectively. The am- Jerlov, 1968) is convenient and is often used to calculate
plitude of the intraseasonal SST variation clearly de- the downward shortwave radiation flux in seawater. How-
creases if the diurnal variations of the surface fluxes are ever, this parameterization is not always satisfactory (e.g.,
neglected. The mixed-layer depth simulated with the Sui et al., 1997b), and cannot consider the interaction
hourly fluxes becomes a little deeper than that with the between physical and biological processes. Recently a
daily fluxes. Sui et al. (1997b) insisted that, because of significantly improved parameterization that depends on
the asymmetric heating rate of the diurnal cycle, the vari- upper-ocean chlorophyll-a concentration, cloud amount,
ation of mixed-layer properties on the diurnal timescale and solar zenith angle has been proposed (Ohlmann and
is nonlinearly related to the intraseasonal variability. The Siegel, 2000). Fairall et al. (1996) originally adopted
diurnal change of the ocean mixed layer will affect the shortwave-radiation parameterizations that depended on
phase of intraseasonal atmospheric variation through the only depth for their skin-layer and warm-layer models.
amplification of the intraseasonal SST variation (cf. Li et Ohlmann and Siegel (2000) and Wick et al. (2005) re-
al., 2001; Woolnough et al., 2007). The inclusion of the placed them with Ohlmann and Siegels parameterization,
mixed-layer deepening/shoaling process on the diurnal and compared the temperature and surface fluxes simu-
scale cannot be neglected in biological modeling lated by this modified version of Fairall et al.s model
(McCreary et al., 2001). with those by the original model. If Ohlmann and Siegels
b. Solar extinction modeling parameterization is adopted, due to the reduction of the
Unlike longwave radiation, sensible and latent heat absorption of insolation, the skin effect is strengthened
fluxes, shortwave radiation can penetrate the ocean. De- and the temperature difference across the warm layer is
termination of vertical distribution of warming by decreased. As a result, SSTskin and the net heat flux from
shortwave radiation in the upper ocean is critical to re- the ocean can be reduced by about 0.2 K and 5 Wm1
producing accurate diurnal and intraseasonal variations under calm and clear conditions, respectively.
of SST (e.g., Kantha and Clayson, 1994; Sui et al., 1997b; As reviewed here, use of an improved
Shinoda, 2005; Wick et al., 2005; Ward, 2006; Clayson parameterization for shortwave radiation in the ocean is
and Weitlich, 2007). Sui et al. (1997b) pointed out that important to simulate SST and surface fluxes in a numeri-
realistic diurnal and intraseasonal SST variations in the cal model accurately. It is known that biological proc-