Вы находитесь на странице: 1из 14

PUBLICATIONS

Journal of Geophysical Research: Oceans


RESEARCH ARTICLE
10.1002/2014JC010383
Key Points:
 A new objective algorithm for
estimating annual maximum MLD is
proposed
 The two-scale MLD pattern is related
to wind/wave climate and eddy
shedding

Correspondence to:
G. Chen,
gechen@ouc.edu.cn
Citation:
Chen, G., and F. Yu (2015), An objective
algorithm for estimating maximum
oceanic mixed layer depth using
seasonality indices derived from Argo
temperature/salinity proles, J.
Geophys. Res. Oceans, 120, 582595,
doi:10.1002/2014JC010383.
Received 12 AUG 2014
Accepted 29 DEC 2014
Accepted article online 7 JAN 2015
Published online 30 JAN 2015

An objective algorithm for estimating maximum oceanic mixed


layer depth using seasonality indices derived from Argo
temperature/salinity profiles
Ge Chen1 and Fangjie Yu1
1

Qingdao Collaborative Innovation Center of Marine Science and Technology, College of Information Science and
Engineering, Ocean University of China, Qingdao, China

Abstract In this study, we propose a new algorithm for estimating the annual maximum mixed layer
depth (M2LD) analogous to a full range of local ventilation depth, and corresponding to the deepest surface to which atmospheric inuence can be felt. Two seasonality indices are dened, respectively, for
temperature and salinity through Fourier analysis of their time series using Argo data, on the basis of which
a signicant local minimum of the index corresponding to a maximum penetration depth can be identied.
A nal M2LD is then determined by maximizing the thermal and haline effects. Unlike most of the previous
schemes which use arbitrary thresholds or subjective criteria, the new algorithm is objective, robust, and
property adaptive provided a signicant periodic geophysical forcing such as annual cycle is available. The
validity of our methodology is conrmed by the spatial correlation of the tropical dominance of saline effect
(mainly related to rainfall cycle) and the extratropical dominance of thermal effect (mainly related to solar
cycle). It is also recognized that the M2LD distribution is characterized by the coexistence of basin-scale
zonal structures and eddy-scale local patches. In addition to the fundamental buoyancy forcing caused
mainly by latitude-dependent solar radiation, the impressive two-scale pattern is found to be primarily
attributable to (1) large-wave climate due to extreme winds (large scale) and (2) systematic eddy shedding
as a result of persistent winds (mesoscale). Moreover, a general geographical consistency and a good quantitative agreement are found between the new algorithm and those published in the literature. However, a
major discrepancy in our result is the existence of a constantly deeper M2LD band compared with other
results in the midlatitude oceans of both hemispheres. Given the better correspondence of our M2LDs with
the depth of the oxygen saturation limit, it is argued that there might be a systematic underestimation with
existing criteria in these regions. Our results demonstrate that the M2LD may serve as an integrated proxy
for studying the coherent multidisciplinary variabilities of the coupled oceanatmosphere system.

1. Introduction

This is an open access article under the


terms of the Creative Commons Attribution-NonCommercial-NoDerivs
License, which permits use and distribution in any medium, provided the
original work is properly cited, the use
is non-commercial and no modications or adaptations are made.

CHEN AND YU

As a counterpart of the mixed layer in the lower atmosphere, the oceanic mixed layer is a global feature of
turbulence-induced quasi-homogeneous zone in terms of density in the upper ocean. The two coupled
mixed layers are home to major hydrodynamic and thermodynamic activities of the oceanatmosphere system which determine the fundamental patterns of the marine environment and climate change. Better
understanding of their formation, variation, and interaction are of critical importance to both oceanographers and meteorologists. In contrast to the atmospheric mixing height of tens to hundreds of kilometers,
the oceanic mixed layer depth (MLD, specied as MLDT, MLDS, or MLDD when dened with temperature,
salinity, or density) varies only from tens to hundreds of meters, but its dynamic complexity is no less than
the former, including buoyancy uxes induced by radiative heating/cooling, surface precipitation/evaporation, ice formation/melting, and mechanical mixing due to wind/wave stirring, eddy/current shear, as well
as Rossby and internal wave disturbance. As a result of immense diurnal, seasonal, and interannual ventilations, the consequences of oceanic mixed layer are wide spread and far reaching, ranging from climatic
~o-Southern Oscillation, global warming and its hiatus, chemical effects such as nitrogen
effects such as El Nin
and carbon cycles in an oxygen depleted environment, to biological effects such as phytoplankton blooms
rtner and Knust, 2007; Ravichandran et al., 2012].
and shery population shift [Po
Tremendous efforts have been made in the past few decades to develop schemes and optimize algorithms
for identifying oceanic mixed layers over global and regional oceans, leading to numerous criteria for the

C 2015. The Authors.


V

582

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

estimation of MLD based on temperature, salinity, and density measurements. An early review on MLDrelated works carried out between 1970s and 1990s is presented by Kara et al. [2000] who separate them into
two general categories. The rst is to dene an isothermal layer depth from a temperature prole and assume
this to be the MLDT, and the second is to dene a MLDD from a density prole with either a property difference or a gradient criterion. As they point out, however, none of these studies (see their Table 1) present a
quantitative analysis justifying a particular property difference value as the most appropriate criterion. A few
years later, an updated summary of criteria used to dene the MLDT,D is compiled by de Boyer Montegut et al.
[2004]. Given the fact of criteria diversity as indicated in their Table 1 (e.g., the temperature threshold varies
from 0.1 to 1.0 C with a changing reference depth of 010 m), the authors emphasize that most often both
the concept and the choice of thresholds are rather arbitrary. This view is shared by Lorbacher et al. [2006]
who argue that the estimated MLD is sensitive not only to choices of threshold and reference value, but also
to the vertical resolution and gradient at the base of the mixed layer. Dong et al. [2008], among others [e.g.,
Holte and Talley, 2009; Ohno et al., 2009], also divide the criteria of MLD estimation into two groups: property
difference-based criteria and gradient-based criteria. They further quantify that in the rst group, values of
potential temperature difference from 0.01 C to 1.0 C and values of potential density difference from 0.005
to 0.125 kg m23 are commonly used. A temperature gradient of 0.025 C m21 and density gradient from
0.0005 to 0.05 kg m24 are applied to determine MLD in the gradient criteria. In addition to these straightforward schemes, more sophisticated approaches have also been proposed as attempts to minimize the uncertainties associated with MLD estimates. For example, Lavender et al. [2002] use the intersection between a
straight-line t to the upper layer and an exponential plus second-order-polynomial t to the deep layer to
estimate the MLDT of individual temperature proles in the Labrador Sea. Thomson and Fine [2003] introduce
a split and merge method, which ts a variable number of linear segments to a prole. Lorbacher et al.
[2006] calculate the MLDT on the basis of the shallowest extreme curvature of temperature proles. Holte and
Talley [2009] develop a hybrid method which nds the MLD of individual ocean proles, models the general
shape of each prole, searches for physical features in the prole, and calculates threshold and gradient to
assemble a suite of possible values before selecting a nal MLD.
As clearly evident from the brief review above, and given the largely arbitrary and subjective nature of
existing criteria, there seems to be no single, optimal, and universal methodology for determining the
MLDs over global oceans. Kara et al. [2000] point out that differences in the criteria can lead to considerable differences in the MLDs, which in turn, could inuence the study ndings. Lorbacher et al. [2006]
summarize three major uncertainties about the threshold method. First, if assuming that a DT 5 0.2 C (a
threshold of temperature difference between the sea surface and a given depth, same as Dq below for
density) is representative of an adequate criterion, the resulting deviations of the estimated MLDT can
be sometimes in the order of MLDT itself. Second, the complicated dependence of MLD on the vertical
resolution is not desirable, especially when comparison studies are made with the low vertical resolution
output of numerical models. Third, because MLD depends on the sea surface temperature (SST) or a reference value and there is no rational choice for this value, leading us to be more than skeptical that
the turbulent region of the upper ocean is captured well by the threshold criterion. Dong et al. [2008]
perform a sensitivity test using alternate net difference values of temperature DT 5 0.1, 0.2, 0.5 C and
density Dq 5 0.01, 0.03, 0.125 kg m23, and nd that the resulting MLDs differ signicantly for the Southern Ocean. In addition, a value of criterion chosen subjectively for one region or season might not be
applicable to another region or season [Lorbacher et al., 2006]. For example, the intersection method
developed by Lavender et al. [2002] works nicely in the North Atlantic, but fails to produce realistic
MLDs in the Southern Ocean [Holte and Talley, 2009].
In this study, we develop a fundamentally different approach for nding the annual maximum mixed layer
depth (M2LD) from Argo oat measurements based on the intrinsic concept of seasonality. Instead of using
the threshold or gradient-based criteria which have a straightforward geophysical background in theory
but a subjective and arbitrary nature in practice, we propose a new algorithm for estimating the M2LD,
which is totally independent of specic temperature, salinity, or density thresholds. The idea behind the
methodology is that, instead of using directly measured property prole data, pseudo seasonality index
proles of an annual property amplitude are rst derived, on the basis of which a local minimum of the
index corresponding to a maximum penetration depth can be determined. As will be demonstrated in the
following sections, this novel approach will eventually benet from the basic fact that seasonal cycle, rather

CHEN AND YU

C 2015. The Authors.


V

583

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

than the dependencies on various complex local assumptions, is ultimately the decisive forcing for the
annual M2LD. The rest of the paper is organized as follows: a brief description of the Argo oat data used in
this investigation along with the proposed M2LD derivation algorithm is provided in section 2. The results
are presented, analyzed, and compared in section 3. The mechanisms of M2LD formation are discussed in
section 4, and some nal remarks are given in section 5.

2. Data and Method


Ten years of Argo (Array for Real-time Geostrophic Oceanography) data spanning from 2004 to 2013 are
used in this study. Argo oats are designed to observe large-scale (seasonal and longer, thousand kilometers and larger) subsurface ocean variability globally [Roemmich et al., 2009]. As the rst observation system of global subsurface ocean in history and one of the main sources of in situ temperature/salinity (T/S)
measurements, the Argo project has an unprecedented spatialtemporal sampling and coverage, the aim
of which is to provide simultaneous T/S observations of the 022000 m upper ocean in near real time. By
January 2014, there are as many as 3613 active oats disseminated around the global ocean spacing nominally at every 3 of longitude and latitude. In this analysis, gridded Argo T/S data are obtained from the
China Argo Real-Time Data Center (http://www.argo.org.cn/). The original observations used to generate
the product are the so-called D-mode Argo data with pressure offsets corrected [Barker et al., 2011]. There
are 48 vertical layers in our data set ranging from 5 to 1950 m with an interpolated spatial resolution of
1 3 1 and temporal resolution of 1 month, respectively [Chen et al., 2014].
The accumulation of continuous time series from available Argo oats has exceeded one decade for the
rst time. The annual harmonic of sea temperature and salinity with a periodicity of P 5 12 months for a
given layer can be derived through Fourier analysis of the time series T(x, y, z, t) and S(x, y, z, t) at each grid
point (x, y) of depth z,
Tx; y; z; t5AT x; y; z3cos2pt=P1uT x; y; z

(1a)

Sx; y; z; t5AS x; y; z3cos2pt=P1uS x; y; z

(1b)

where AT and AS are the amplitudes of the annual component, uT and uS are the phase angles which determine the time when the maximum of the annual harmonic occurs, while t varies from 0 to 120 months for
the present Argo data set. The fact that the Argo array density is considered nonoptimal until 2007 will not
signicantly affect the result of this analysis since we are focusing on the annual and semiannual cycles,
and according to the Nyquist criterion, the time series of over 7 year densely distributed data are long
enough for resolving the T/S variability on seasonal time scales. Given a complete spatiotemporal data set
T(x, y, z, t) for temperature or S(x, y, z, t) for salinity, AT(x, y, z) and uT(x, y, z) or AS(x, y, z) and uS(x, y, z) can be
simultaneously retrieved and are supposed to carry full information of annual T/S variability in terms of
amplitude and phase for the global upper ocean.
Following the procedures described above, pseudoproles of AT and AS can be derived at each grid point
as indices for measuring the strength of T/S seasonality. Starting from the surface layer, our algorithm
searches progressively at each grid deeper layers until it nds a depth where AT or AS reaches its rst signicant local minimum or its level of practical accuracy of 0.02 C for temperature and 0.01 psu for salinity
[Barker et al., 2011]:


AT xi ; yj ; zTmin 5min AT xi ; yj ; zk
k5 1; 2; . . . ; 48
(2a)


k5 1; 2; . . . ; 48
AS xi ; yj; zSmin 5min AS xi ; yj ; zk

(2b)

where k is the index of depth layer, zTmin (zSmin) corresponds to the depth of the rst minimum AT (AS). The
process is repeated for each (xi, yj) with i 5 1 , 2 ,. . ., 360 , and j 5 260 , 259 ,. . ., 60 . The nal M2LD is
chosen as the deeper layer out of zTmin and zSmin:
M2 LDxi ; yj 5maxfzTmin jxi; yj ; zSmin jxi; yj g

(3)

The geophysical validity of this simple approach lies in three aspects. First, since mixed layer depth can
vary by tens of meters over a diurnal cycle [Thomson and Fine, 2003] and given the temporal resolution

CHEN AND YU

C 2015. The Authors.


V

584

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

of present Argo data [Roemmich et al., 2009], it is actually difcult to track the evolution of MLD on
daily or instantaneous basis in a meaningful way. On the other hand, it is understood that the mixed
layer contains the oceans memory of air-sea exchange out to periods of a year or longer [Sprintall
and Roemmich, 1999], we therefore decide to focus on annual maximum MLD, which provides an overall dynamic connection (both hydro and thermal) between the atmosphere and the ocean and thus
plays a central role in climate variability. Second, seasonality indices of temperature and salinity at a
given location decrease monotonically with depth as a result of a normal ventilation until the rst
abnormal bump, i.e., local minimum is encountered, which signies the termination of vertical mixing
dominance due to possible occurrence of other intervening processes such as the intrusion of a front
or a jet of sinking colder water which distorts the vertical structure of the homogeneous temperature
or salinity layer [Kara et al., 2003]. This seasonality minimum coincides with the M2LD since annual cycle
of MLD is usually an order of magnitude larger over most of the worlds oceans compared with other
variabilities including diurnal cycle [e.g., Price et al., 1986] or interannual oscillation [e.g., Keerthi et al.,
2013]. Alternatively, mixed layer is considered to be vanishing or insignicant when AT (AS) reaches the
Argo measurement accuracy. Third, as pointed out by de Boyer Montegut et al. [2007], the base of the
mixed layer is usually dened as the top of the pycnocline, being the depth where the density has
increased by a certain threshold from its surface value. Presently, the available density proles are about
an order of magnitude less than the temperature proles [Lorbacher et al., 2006]. Given also the lack of
salinity data, it is often concentrated on the temperature stratication, assuming that the top of the
thermocline and halocline have the same depth and thus together dene that of the pycnocline. However, this view is oversimplied for the real ocean. Temperature and salinity can be stratied in distinctly different ways as a result of specic thermal and haline forcings [Lorbacher et al., 2006; Mignot
et al., 2007]. As a combined result, our algorithm of equations (1)2(3) can be justied in search of an
annual maximum MLD especially in the tropical oceans where the saline effect is supposed to
dominate.
Practically, we start by looking at the global distribution of recovered annual amplitude of sea temperature
and salinity variabilities at selected depths, as shown in Figure 1. Focusing on the left column, it is obvious
that the geographical pattern of sea temperature seasonality is highly inhomogeneous in each layer and is
largely dissimilar for different layers. For example, the near-surface temperature seasonality is most prominent around the Japan islands in the northwest Pacic, and off the east coasts of the U.S. and Canada in the
northwest Atlantic (Figures 1a and 1b). Large annual variability is also apparent in the circumpolar belt
between 25 S and 40 S of the Southern Ocean. The least dynamic region coincides with the western Pacic
warm pool (WPWP) [Chen et al., 2004b]. In contrast, the most dynamic areas for seasonal sea temperature
~o-3/4 and the WPWP regions at 100 and 200 m depths, respectively (Figures
changes are located in the Nin
1c and 1d). Localized strong seasonalities are also observed in deeper oceans such as the Gulf of Mexico at
500 m, and the Indonesia waters at 1950 m (Figures 1e and 1f). A similar degree of diversity can be
observed for salinity seasonality as well (see right column of Figure 1), particularly in the Asian monsoon
region (Figures 1g1i) and the intertropical convergence zone.
Browsing through the color scales of the left and right columns in Figure 1, one nds a common characteristic shared by temperature and salinity seasonalities that they both drop down rapidly as going deeper. This
can be quantitatively illustrated by plotting the spatially averaged seasonality of temperature and salinity
with respect to depth as shown in Figure 2 (also overlaid with dashed lines are the semiannual amplitudes
of sea temperature and salinity). As expected, the two indices decline monotonically with the dramatic
reduction of solar penetration into the deeper ocean. Globally, it is estimated that only 1.1% of the surface
strength of seasonal variability is left for temperature and 2.5% for salinity at about 2000 m depth. Individually, however, the vertical reduction behaviors of seasonal signals are tremendously diverse. Three typical
examples of temperature seasonality proles are shown in Figure 3a for (330 E, 25 S), (232 E, 1 S), and
(160 E, 30 N), where a local minimum can be rst identied at approximately layers 8, 14, and 23 (corresponding to 60, 120, and 220 m depths), respectively. Plots with similar features are also generated for salinity (not shown). The existence of these well-dened seasonality minima justies the validity of our
proposed scheme and enables us to practically determine the M2LD by maximizing the concurrent temperature and salinity-derived depths, i.e., nding a deeper value out of zTmin and zSmin, as described in equation
(3). It should be pointed out that, since our scheme searches for the rst local minimum (rather than the

CHEN AND YU

C 2015. The Authors.


V

585

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

Figure 1. Global distribution of recovered annual amplitude (seasonality index) of (left column) sea temperature and (right column) salinity variabilities at selected depths: (a and g) 5 m;
(b and h) 50 m; (c and i) 100 m; (d and j) 200 m; (e and k) 500 m; and (f and l) 1950 m.

global minimum in the whole prole) of the seasonality index corresponding to the initial termination of
turbulence, it is unlikely to be biased toward seasonal thermocline features as described in Brainerd and
Gregg [1995].
By denition, density is the most relevant parameter in constructing a MLD climatology. Variations in temperature and salinity combine to control the density of ocean surface water. The density increases with a decreasing temperature and an increasing salinity. As evidenced in Lorbacher et al. [2006] and Mignot et al. [2007],

CHEN AND YU

C 2015. The Authors.


V

586

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

the commonly used assumption of temperature dominance


is found to be systematically
violated under certain circumstances such as intense precipitation, river runoff, and even El
~o events. They show examNin
ples of proles where salinity
controls the depth of the
mixed layer due to the existence of the so-called barrier
layer. Accordingly, we are
motivated to assess the specic contribution of thermal
effect and haline effect in our
M2LD derivation scheme. To do
so, the spatial distribution of
the differences between
temperature-based and
Figure 2. Globally averaged annual (solid lines) and semiannual (dashed lines) amplitudes
salinity-based M2LDs is pre(seasonality index) of sea temperature (red) and salinity (blue) as a function of depth layer
sented in Figure 4, and the
corresponding to 02000 m.
zonally averaged value of the
difference is shown in Figure 5 (note that positive values are used for zTmin and zSmin in the calculation). It is
found that, as far as the seasonal convective circulation is concerned, 78.8% of the global ocean is thermal
driven, while 21.2% is haline driven. Combining Figures 4 and 5, one recognizes that the salinity-controlled
area and temperature-controlled area are basically latitudinally divided: the former covers a majority of the
tropical oceans while the latter dominates the extratropical oceans of the two hemispheres. Such a pattern
is in good agreement with Mignot et al. [2007] who locate their quasipermanent barrier layers in the equatorial and western tropical Pacic and Atlantic, in the Bay of Bengal and the eastern equatorial Indian Ocean,
in the Labrador Sea, and parts of the Arctic and Southern Ocean, thus suggesting that our M2LD estimation
algorithm is a robust one, effectively triggering the salinity criterion in deep tropics while activating the
temperature criterion in the rest of the worlds oceans.
Also, it is interesting to examine the effect of other seasonal harmonics on M2LD determination in addition
to the annual component. In doing so, two seasonality indices based on semiannual temperature and salinity amplitudes are derived using the same algorithm described above, given the fact that semiannual cycle
is usually the second largest regime in seasonal variabilities (being roughly 1/41/2 of the annual amplitude,
see the dashed lines in Figure 2) and may even exceed the annual cycle in some areas of the ocean. It turns
out that semiannual dominance is signicant in about 9.6% of the oceanic regions (Figure 6), in which 4.5%
of the cases are triggered by the temperature criterion and 5.1% by the salinity criterion. A majority of these
areas are located in the tropical oceans of the three basins, as well as the Southern Ocean along the Antarctic Circumpolar Current (ACC). No particular geographic preference is found for the temperature or salinity
index, and the resulting deepening of the M2LD is estimated to be approximately 510 m at the affected latitudes (e.g., 15 S20 N and 50 S60 S, not shown). Therefore, the semiannual indices provide slight but signicant regional corrections to the annual harmonic-based algorithms without changing the basic pattern
of the global M2LD.

3. Results and Comparisons


Following the procedures described in section 2, a global map of M2LD is created using the combined temperature/salinity algorithm (Figure 7). The overall pattern of the M2LD climatology is characterized by a
zonally oriented three-band structure: (1) a band with largely deep M2LDs in the Southern Ocean between
30 S and 60 S; (2) a band with generally shallow M2LDs in the tropical oceans between 30 S and 30 N; (3)
two basin-wide zones with relatively deeper M2LDs between 30 N and 60 N in the North Pacic and North
Atlantic. In Figure 7, the deepest M2LDs in the Southern Ocean are found in the northern side of the

CHEN AND YU

C 2015. The Authors.


V

587

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

Figure 3. (a) Annual amplitude (seasonality index) of sea temperature as a function of depth layer corresponding to 02000 m at selected grid points: red for (160 E, 30 N), green for
(330 E, 25 S), and blue for (232 E, 1 S); overlaid plot of seasonality index (red), winter temperature (pink), and salinity (blue) proles (corresponding to maximum MLD) for (b) (160 E,
30 N) and (c) (232 E, 1 S).

subantarctic front (SAF) resulting from the Antarctic intermediate water (AAIW) with relatively low salinity,
high oxygen, and low potential vorticity. It is the densest, deepest, and freshest portion of the subantarctic
mode water and is thought to form partly in the Southeast Pacic just before the ACC enters the Drake Passage [Dong et al., 2008; Holte and Talley, 2009]. In the North Pacic and North Atlantic, the two deep M2LD
bands are located in the frontal zones between the subtropical and subarctic gyres in the western basin
due to the formation of a density compensated layer [Oka et al., 2007].
The observed M2LD pattern is naturally related to prevailing wind and wave actions. But a careful comparison between Figure 7 and the global wind speed climatology [see, e.g., Chen et al., 2003a, Figure 1] indicates that the two distributions differ signicantly: the trade wind belts and horse latitudes of oceanic
winds are all absent in the M2LD map. Instead, Figure 7 is rather similar in its general pattern to the global
signicant wave height (SWH) climatology (a measure of dynamic degree of wave-induced total turbulence)
derived from satellite data [see, e.g., Chen et al., 2002, Figure 3b], except that the zone of deepest M2LDs
appears in the North Atlantic while that of the maximum SWH is found in the southern Indian Ocean. The

CHEN AND YU

C 2015. The Authors.


V

588

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

Figure 4. Global distribution of the differences between (positive) temperature-based and salinity-based annual M2LDs. Positive values are
depicted in color while negative ones in black and white.

two comparisons imply that wave stirring seems to be a direct forcing for the mixed layer generation.
Because SWH consists of both wind sea and swell sea (nonwind) contributions, whereas the latter has a
totally different spatial pattern with respect to wind climatology [see Chen et al., 2002, Figure 3a], the resulting zonal pattern of mixed waves thus reduces to a three-band structure consistent with the M2LDs rather
than following the seven-band structure of the global wind climatology. Having demonstrated that ocean
wave action might be the main dynamics in mixed layer formation, we continue to examine the relationship
between M2LD (Figure 7) and the 10 year return wind speeds derived from satellite data using a methodology developed by Chen et al. [2004a] (Figure 8). Surprisingly, the two maps resemble each other not only in
overall pattern but also in relative strength of major features in mid-to-high latitudes, which further suggests that it is ultimately the extreme winds and the corresponding large waves (SWH) that play a critical
role in shaping the extratropical M2LD distribution.
Next, in comparison of our M2LDs with those reported in the literature, we nd a general similarity in their
global patterns (see our Figure 7, Figure 14b of de Boyer Montegut et al. [2004], and Figure 6c of Ohlmann
et al. [1996]). In particular, the result of this study is in quantitative agreement with that of de Boyer Montegut et al. [2004]: the M2LD varies from 10 m to approximately 150 m in the tropical oceans, while reaches
a belt of maximum exceeding 300 m in the midlatitudes of the two hemispheres. A close scrutiny of available maps reveals that our result contains numerous well-dened eddy-like patches with contrasted local
M2LD highs and lows mostly in the westerlies of the two hemispheres (Figure 7), which is also evidenced to
some extent in Ohno et al. [2004] and predicted by Keerthi et al. [2013] using an eddy permitting numerical
model. This argument is supported by the fact that the deepest localized M2LDs in the Southern Ocean
exceed 550 m in our Figure 7, in contrast to the nding of 400 m on a regional scale by Dong et al. [2008],

Figure 5. Zonally averaged differences between (positive) temperature-based and salinity-based annual M2LDs. The horizontal dashed
line indicates a zero value.

CHEN AND YU

C 2015. The Authors.


V

589

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

Figure 6. Geographical locations with semiannual temperature-based or salinity-based M2LDs exceeding annual ones.

conrming the statement that some proles are even different from adjacent ones in the same year [Ohno
et al., 2004].
In fact, using 16 years of sea surface height data during 19922008, Chelton et al. [2011] have identied and
tracked 35,891 mesoscale eddies (more than 3000 per year on average) over the global oceans. These longlived eddies have an average lifetime of 32 weeks and an average propagation distance of 550 km (which
enable them, at least most of the larger ones in middle and high latitudes, to be marginally resolved by the
present Argo oats with nominal intervals of 3 3 3 in space and 10 day in time). Their mean amplitude
and a speed-based radius scale are 8 cm and 90 km, respectively. In addition to these characteristic properties, there is a remarkable overall resemblance between the small circular patches in our M2LD distribution
and the origin and termination locations of the observed eddies (see bottom panel of Figure 1 and Figure 6
of Chelton et al. [2011]). Moreover, the mean eddy amplitudes are found to peak around 40 50 in both
hemispheres (see the upper panels of their Figure 10). These geographically correlated features hint that
mesoscale eddies play a dominant role in modifying the large-scale pattern of M2LD determined jointly by
wind/wave climate and solar radiation (see also next section).
It is understood that the eddy-induced convective process is localized in space. Three phases can be identied in open ocean deep convection: (1) preconditioning on the large scale in the order of 100 km, (2)
deep convection occurring in localized, intense plumes on scales of the order of 1 km, and (3) lateral
exchange between the convection site and the ambient uid through advective processes on a scale of a
few tens of kilometers [Marshall and Schott, 1999]. The last two phases are not necessarily sequential and
often occur concurrently. Specically, cooling events may initiate deep convection in which a substantial
part of the uid column overturns in numerous plumes that distribute the dense surface water in the vertical. The plumes have a horizontal scale of the order of their lateral scale (<1 km) with vertical velocities of
up to 10 cm/s [Schott et al., 1996]. In concert, the plumes are thought to rapidly mix properties over the preconditioned site, forming a deep mixed patch ranging in scale from several tens of kilometers to over
100 km in diameter. With the cessation of strong forcing, or if the cooling continues for many days, the

Figure 7. Global distribution of combined annual/semiannual M2LD derived from the proposed new algorithm using combined temperature/salinity data.

CHEN AND YU

C 2015. The Authors.


V

590

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

Figure 8. Global distribution of 10 year return wind speed derived from merged TOPEX/Jason-1/2 altimeter data.

predominantly vertical heat transfer on the convective scale gives way to horizontal transfer associated
with eddying on geostrophic scales [Gascard, 1978] as the mixed patch laterally exchanges uid with its surroundings. Individual eddies tend to organize the convected water into coherent lenses in geostrophic balance, forming the cell-rich pattern in our M2LD map (Figure 7).
It is desirable to make a quantitative comparison of similar results obtained in this study and those by previous
investigators. Before doing so, we try to rst compare the zonally averaged M2LDs for different oceans using the
proposed scheme (Figure 9a). It shows that there is a high consistency between 40 S and 35 N for the derived
M2LDs, meaning that the mixed layers at these latitudes are basin independent to a large extent. M2LDs are considerably deeper in the North Atlantic (nearly 700 m at 57 N) than in the North Pacic (around 200 m) and the
difference enlarges with latitude, being consistent with a similar result of Hosoda et al. [2010]. In the Southern
Ocean, however, the Pacic sector usually has deeper M2LDs than the two other basins. We nd that the deepest average M2LDs can reach 300 m north of the mean SAF in the southeastern Pacic Ocean. Also within the
Pacic sector, a minimum M2LD occur at 53 S, which is slightly shifted from the 57 S value obtained by Holte
and Talley [2009], and is concurrent with the subsurface salinity minimum, a signature of AAIW.

Figure 9. Zonally averaged M2LDs for (a) the Pacic (red), Atlantic (blue), Indian
(green), and global (black) oceans using our scheme; and (b) the global ocean using
JAMSTEC (red), IFREMER (green), SIO (blue), and our (black) schemes.

CHEN AND YU

C 2015. The Authors.


V

As some of the investigators of previous methodologies have released


and updated a corresponding MLD
climatology product based on Argo
oat data, it is possible for us to further compare the zonally averaged
M2LDs computed using our scheme
and those from the JAMSTEC (Japan
Agency for Marine-Earth Science and
Technology [Hosoda et al., 2010];
original MLD data at http://www.
jamstec.go.jp/ARGO/argo_web/
MILAGPV/index_e.html), IFREMER
(French Research Institute for the
Exploitation of the Sea [de Boyer
Montegut et al., 2004]; original MLD
data at http://www.ifremer.fr/cerweb/deboyer/mld/Surface_Mixed_
Layer_Depth.php), and SIO (Scripps
Institution of Oceanography [Holte
and Talley, 2009]; original MLD data
at http://mixedlayer.ucsd.edu/)
schemes as shown in Figure 9b. The
four lines agree generally well within
620 of the tropical oceans, but

591

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

Figure 10. Zonal distributions of M2LD along with annual amplitude of SST variability and 10 year return extreme wind speed as well as
the bathymetry for the global ocean.

diverge signicantly beyond those regions. Further poleward, the JAMSTEC M2LDs become the deepest
with the largest discrepancies exceeding 300 m in zonal mean. It should be noticed that, as proposed by
Reid [1982], the 95% oxygen saturation limit from CTD data gives a proxy for M2LD which is also called a
bowl by Guilyardi et al. [2001]. That is, the 95% oxygen saturation corresponds to the oxygen dissolved
during ventilation at the surface at the time of the deepest convective mixing during the year. Such a nding appears to support our result of deeper M2LDs in midlatitude oceans global wide. It may serve as observational evidence for an improved accuracy of our proposed algorithm, and suggest its better performance
compared with other schemes in areas of strong seasonality (see Figures 1a and 1b) due to a potentially
higher signal-to-noise ratio.

4. Discussions
It is necessary to explore the mechanisms of annual M2LD formation and understand their consequences.
As is commonly known, oceanic MLD is ultimately determined by solar heating and wind forcing from the
atmosphere [Sallee et al., 2010]. Generally, the MLD is proportional to the intensity of combined forcings of
these two factors as illustrated in Figure 10, where the annual SST amplitude is considered as a proxy of
buoyancy forcing, and the 10 year return extreme wind speed is regarded as a proxy of turbulent forcing.
Specically, as far as solar radiation is concerned, it is well known to have a dominant annual cycle. Spatially,
the combined effects of atmospheric path length and inclination of earths axis cause earth to receive more
solar heat in the tropics, less at the temperate latitudes, and the least at the poles. Temporally, however, taking into account the seasonal north-south migration of the sun, the annual cycle of average daily solar radiation is most pronounced at the middle and higher latitudes where the angle at which the suns rays strike
earth and the length of daylight change dramatically from summer to winter. As a combined effect, the
annual ux of solar radiation is greatest at the midlatitudes, smallest around the equator, with polar regions
in between. Consequently, heat is transferred to and from the ocean at the sea surface in hemispheric
summer and winter, respectively. The net effect of these processes is a varying annual change in zonal SSTs
(see Figure 10): 1 2 C in the tropics and 2 6 C at the middle latitudes. The smaller 1 4 C change in temperature at polar latitudes results from the heat transferred locally in the formation and melting of sea ice.
Particularly, the solar forcing is intensied in the northern hemisphere as a result of the signicant asymmetry of landocean distribution. A good correspondence between the solar forcing and the M2LD response
appeared as a phase reversal is evident for the global oceans between 45 S45 N (especially the welldened regional feature within 35 N45 N, see the red and black lines in Figure 10). For the subpolar latitudes of the two hemispheres, solar effect becomes less obvious because of other important forcings such
as the annual cycle of large-scale sea ice variation and the hostile marine environment known as furious fties [e.g., Sverdrup and Armbrust, 2008].
Furthermore, heat that is absorbed at the ocean surface in hemispheric summer is transferred downward
by winds, waves, and currents. While in hemispheric winter, heat is transferred upward toward the cooling
surface. As a result, an obvious correspondence between the basin-scale M2LD pattern and the global

CHEN AND YU

C 2015. The Authors.


V

592

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

distribution of 10 year return extreme wind speed as well as the SWH climatology is found, and an overall
similarity between the maps of mesoscale M2LD features and eddy evolution behaviors is observed. Note
that a direct correlation between zonal wind belt and M2LD bands does not appear as might be expected.
Instead, our analysis suggests that it is the large waves generated by extreme winds in tandem with nonwind swells (as jointly represented by SWH, the mean value of the rst 1/3 largest ocean waves) which are
virtually responsible for the observed large-scale M2LD pattern. Also, it is worth noting that extreme winds
seem to replace solar heating to be the primary forcing in higher latitudes beyond 645 given their clear
phase relationship with zonal M2LDs. In addition, the much shallower bathymetry within the 50 N60 N
band (1000 m) compared with its southern hemisphere counterpart (4000 m) is thought to considerably
accelerate the convection process by enhanced tidal mixing across the water column over the North Atlantic which is known to be an area of deep water formation due to substantial subduction [Nakamura et al.,
2006]. As for the mesoscale features, contributing factors other than eddy shedding may include violent
storms, heavy rain cells, and so on.
It is worth to further discuss the possible reasons behind the largest discrepancy between our result and
those published previously, i.e., the constantly deeper M2LDs in midlatitude oceans (Figure 9b). It should be
kept in mind that the regions of 6(20 45 ) are homes to all subtropical gyres including the worlds most
energetic western boundary currents such as the Kuroshio and the Gulf Stream. As reported by Guilyardi
et al. [2001], the oxygen-based estimate of the MLD is larger than the temperature-based one in all subtropical gyres, which they ascribe to the persistent downwelling Ekman pumping. Moreover, these regions coincide with the notorious roaring forties where extremely high sea state prevails throughout the year [see,
e.g., Chen et al., 2002, Figure 3b]. The coupled effect of strong currents and waves makes the mixing processes of these zonal belts quite unique in that mechanical stirring (both lateral and vertical) is competing
with buoyancy forcing to form an insufcient state of mixing at a given location: the turbulence or disturbance may have reached a certain depth while the thermohaline convection is still underway. As such, the
commonly used temperature-based MLD criteria are unlikely to work properly, while our approach appears
to be more effective in dealing with these specic cases from a density point of view. A joint look at the seasonality index in combination with T/S proles conrms that our algorithm is able to produce reasonable
M2LD estimates under complex thermohaline situations for both temperate and equatorial oceans (see Figures 3b and 3c). The fundamental difference to account for the observed discrepancy is that previous methodologies mostly rely on individual or monthly averaged T/S proles, while our seasonality-index-based
methodology takes into account the overall effect of the annual mixing progression. Since reaching the
M2LD is a cumulative process which may last for nearly half a year especially in the midlatitudes where the
annual variability of SST has a maximum, our result is, therefore, expected to be closer to the reality, as also
supported by the oxygen-based estimation.
It is also interesting to relate the observed discrepancy to the concept of mixing layer and mixed layer
depths as distinguished by Brainerd and Gregg [1995]. According to their denitions, the mixing layer is the
depth zone being actively mixed from the surface at a given time and generally corresponds to the depth
zone in which there is strong turbulence directly driven by surface forcing. The mixed layer is the envelope
of maximum depths reached by the mixing layer on longer time scales that has been mixed in the recent
past. It generally corresponds to the zone above the top of the seasonal pycnocline. Obviously, our
approach basically deals with the mixed layer while many others might be more sensitive to the mixing
layer. The resulting difference thus corresponds to the so-called remnant layer [Brainerd and Gregg, 1995],
which is maximized in midlatitudes where the wave and current-induced turbulences as well as the solar
forcing are peaked (see Figure 10).

5. Final Remarks
Although enough have been said on the importance of MLD to the hydrodynamics and thermodynamics of
the coupled ocean-atmosphere system, it is believed that the special geophysical and biogeochemical
implications of the annual maximum MLDs have not yet been fully recognized. For example, using maximum monthly MLD ensures that thermal energy lost to the mixed layer in form of solar penetration is lost
on annual time scales and is not merely trapped within seasonal pycnocline or barrier layer waters that are
subsequently entrained into the mixed layer on seasonal time scales. Several studies suggest that changes

CHEN AND YU

C 2015. The Authors.


V

593

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

in the maximum depth of the mixed layer from one winter to the next may explain the reemergence of sea
surface temperature anomalies and thus persistence of wintertime SST patterns [Alexander et al., 2001; Carton et al., 2008]. Prakash et al. [2012] argue that the existence of perennial oxygen minimum zones is the
result of restricted ventilation which limits the recharges of the water column with oxygen. Ravichandran
et al. [2012] report a persistent occurrence of a subsurface chlorophyll a maximum just above the top of
permanent thermocline and euphotic depth (to which almost all biological activities are restricted). The
accumulation of multidisciplinary evidences all points to the need for an effective methodology which can
produce realistic estimates of the M2LDs over the global oceans.
With the above scientic backgrounds in mind, and considering the fact that all existing schemes for MLD
estimation are arbitrary and subjective to some degree, we are motivated to develop a truly objective algorithm for deriving M2LD using the best available Argo data. In addition to its pure objectiveness, our criteria
have at least another unique characteristic: the introduction of two independent seasonality indices for both
temperature and salinity based on a decade-long concurrent T/S data set. The combined use of these two
indices is found to effectively optimize the contributions of thermal and haline convections with nearly 80%
of the cases from the former and over 20% from the latter. The proposed strategy is supposed to work globally, regionally or locally as long as signicant periodic forcings such as annual and/or semiannual cycles are
available for one or more of the geophysical properties, but will not be applicable to nonperiodic forcings at
interannual or longer time scales. The validity of our methodology is further conrmed by the spatial correlation of the tropical dominance of the saline effect (mainly related to rainfall cycle [see, e.g., Chen et al., 2003b,
Figure 4]) and the extratropical dominance of the thermal effect (mainly related to solar cycle [see, e.g., Chen
and Li, 2008, Figure 5c]). As far as M2LDs are concerned, a general geographical consistency and a good quantitative agreement are found between our new algorithm and those published in the literature by American,
French, and Japanese scientic groups (Figure 9b). Meanwhile, however, it is particularly worth noting that a
major discrepancy of our result is the occurrence of a constantly deeper zonal band compared with other
results of its kind in the midlatitude oceans of both hemispheres. This suggests that our algorithm produces
similar results like other schemes under fully mixed seas, but appears to be more effective under semimixed
(or mixing) seas (i.e., turbulent mixing precedes convective overturning, in which mechanical stirring occurred
without fully homogenizing the temperature). Given the better agreement of our result with the depth of the
oxygen saturation limit (a proxy of the maximum depth reached by the oceanic mixed layer every year), it is
argued that the M2LDs are likely underestimated by some of the existing criteria in these regions. Also, it is
recognized that our M2LD distribution is characterized by the coexistence of basin-scale zonal structure and
eddy-scale local structure (Figure 7). The impressive two-scale pattern reveals that our new algorithm could
be more efcient in resolving ner structures associated with realistic M2LDs. As such, it is clear that the M2LD
may serve as an integrated proxy for studying the coherent multidisciplinary variabilities in the coupled
ocean-atmosphere system, whose roles will be further amplied by long-term deployment of physical, chemical, and biological sensors onboard future Argo oats.

Acknowledgments
This research was jointly supported by
the Natural Science Foundation of
China under grants 41331172,
U1406404, and 61361136001 and the
Global Change Research Program of
China under grant 2012CB955603. We
would like to thank J. Buck from the
British Oceanographic Data Center and
an anonymous Reviewer for their
thorough and helpful reviews on the
previous version of this manuscript.
Special thanks go to the China Argo
Real-Time Data Center for providing us
with the gridded Argo data product
used in this study (http://www.argo.
org.cn/).

CHEN AND YU

References
Alexander, M. A., M. S. Timlin, and J. D. Scott (2001), Winter-to-winter recurrence of sea surface temperature, salinity and mixed layer depth
anomalies, Prog. Oceanogr., 49, 4161.
Barker, P. M., J. R. Dunn, C. M. Domingues, and S. E. Wijffels (2011), Pressure sensor drifts in Argo and their impacts, J. Atmos. Oceanic Technol., 28, 10361049.
Brainerd, K. E., and M. C. Gregg (1995), Surface mixed and mixing layer depths, Deep Sea Res., Part I, 42, 15211543.
Carton, J. A., S. A. Grodsky, and H. Liu (2008), Variability of the oceanic mixed layer, 19602004, J. Clim., 21, 10291047.
Chelton, D. B., M. G. Schlax, and R. M. Samelson (2011), Global observations of nonlinear mesoscale eddies, Prog. Oceanogr., 91, 167216.
Chen, G., and H. Li (2008), Fine pattern of natural modes in sea surface temperature variability: 19852003, J. Phys. Oceanogr., 38, 314336.
Chen, G., B. Chapron, R. Ezraty, and D. Vandemark (2002), A global view of swell and wind sea climate in the ocean by satellite altimeter
and scatterometer, J. Atmos. Oceanic Technol., 19, 18491859.
Chen, G., S. W. Bi, and J. Ma (2003a), Global structure of marine wind variability derived from TOPEX altimeter data, Int. J. Remote Sens., 24,
51195133.
Chen, G., J. Ma, C. Fang, and Y. Han (2003b), Global oceanic precipitation derived from TOPEX and TMR: Climatology and variability, J. Clim.,
16, 38883904.
Chen, G., S. W. Bi, and R. Ezraty (2004a), Global structure of extreme wind and wave climate derived from TOPEX altimeter data, Int. J.
Remote Sens., 25, 10051018.
Chen, G., C. Fang, C. Zhang, and Y. Chen (2004b), Observing the coupling effect between warm pool and rain pool in the tropical Pacic,
Remote Sens. Environ., 91, 153159.
Chen, G., H. Zhang, and X. Wang (2014), Annual amphidomic columns of sea temperature in global oceans from Argo data, Geophys. Res.
Lett., 41, 20562062, doi:10.1002/2014GL059430.

C 2015. The Authors.


V

594

Journal of Geophysical Research: Oceans

10.1002/2014JC010383

de Boyer Mont
egut, C., G. Madec, A. S. Fischer, A. Lazar, and D. Iudicone (2004), Mixed layer depth over the global ocean: An examination
of prole data and a prole-based climatology, J. Geophys. Res., 109, C12003, doi:10.1029/2004JC002378.
de Boyer Mont
egut, C., J. Mignot, A. Lazar, and S. Cravatte (2007), Control of salinity on the mixed layer depth in the world ocean: 1. General description, J. Geophys. Res., 112, C06011, doi:10.1029/2006JC003953.
Dong, S., J. Sprintall, S. T. Gille, and L. Talley (2008), Southern Ocean mixed-layer depth from Argo oat proles, J. Geophys. Res., 113,
C06013, doi:10.1029/2006JC004051.
Gascard, J.-C. (1978), Mediterranean deep water formation, baroclinic eddies and ocean eddies, Oceanol. Acta, 1, 313315.
Guilyardi, E., G. Madec, and L. Terray (2001), The role of lateral ocean physics in the upper ocean thermal balance of a coupled oceanatmosphere GCM, Clim. Dyn., 17, 14231452.
Holte, J., and L. Talley (2009), A new algorithm for nding mixed layer depths with applications to Argo data and subantarctic mode water
formation, J. Atmos. Oceanic Technol., 26, 19201939.
Hosoda, S., T. Ohira, K. Sato, and T. Suga (2010), Improved description of global mixed-layer depth using Argo proling oats, J. Oceanogr.,
66, 773787.
Kara, A. B., P. A. Rochford, and H. E. Hurlburt (2000), An optimal denition for ocean mixed layer depth, J. Geophys. Res., 105, 16,80316,821.
Kara, A. B., P. A. Rochford, and H. E. Hurlburt (2003), Mixed layer depth variability over the global ocean, J. Geophys. Res., 108(C3), 3079, doi:
10.1029/2000JC000736.
Keerthi, M. G., M. Lengaigne, J. Vialard, C. de Boyer Mont
egut, and P. M. Muraleedharan (2013), Interannual variability of the Tropical Indian
Ocean mixed layer depth, Clim. Dyn., 40, 743759.
Lavender, K. L., R. E. Davis, and W. B. Owens (2002), Observations of open-ocean deep convection in the Labrador Sea from subsurface
oats, J. Phys. Oceanogr., 32, 511526.
Lorbacher, K., D. Dommenget, P. P. Niiler, and A. K
ohl (2006), Ocean mixed layer depth: A subsurface proxy of ocean-atmosphere variability,
J. Geophys. Res., 111, C07010, doi:10.1029/2003JC002157.
Marshall, J., and F. Schott (1999), Open-ocean convection: Observations, theory, and models, Rev. Geophys., 37, 164.
Mignot, J., C. de Boyer Mont
egut, A. Lazar, and S. Cravatte (2007), Control of salinity on the mixed layer depth in the world ocean: 2. Tropical areas, J. Geophys. Res., 112, C10010, doi:10.1029/2006JC003954.
Nakamura, T., T. Toyoda, Y. Ishikawa, and T. Awaji (2006), Effects of tidal mixing at the Kuril Straits on North Pacic ventilation: Adjustment
of the intermediate layer revealed from numerical experiments, J. Geophys. Res., 111, C04003, doi:10.1029/2005JC003142.
Ohlmann, J. C., D. A. Siegel, and C. Gautier (1996), Ocean mixed layer radiant heating and solar penetration: A global analysis, J. Clim., 9,
22652280.
Ohno, Y., T. Kobayashi, N. Iwasaka, and T. Suga (2004), The mixed layer depth in the North Pacic as detected by the Argo oats, Geophys.
Res. Lett., 31, L11306, doi:10.1029/2004GL019576.
Ohno, Y., N. Iwasaka, F. Kobashi, and Y. Sato (2009), Mixed layer depth climatology of the North Pacic based on Argo observations, J. Oceanogr., 65, 116.
Oka, E., L. D. Talley, and T. Suga (2007), Temporal variability of winter mixed layer in the mid- to high-latitude North Pacic, J. Oceanogr.,
63, 293307.
P
ortner, H. O., and R. Knust (2007), Climate change affects marine shes through the oxygen limitation of thermal tolerance, Science, 315,
9597.
Prakash, S., T. M. Balakrishnan Nair, T. V. S. Udaya Bhaskar, P. Prakash, and D. Gilbert (2012), Oxycline variability in the central Arabian Sea:
An Argo-oxygen study, J. Sea Res., 71, 18.
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, 84118427.
Ravichandran, M., M. S. Girishkumar, and S. Riser (2012), Observed variability of chlorophyll-a using Argo proling oats in the southeastern
Arabian Sea, Deep Sea Res., Part I, 65, 1525.
Reid, J. L. (1982), On the use of dissolved oxygen concentration as an indicator of winter convection, Nav. Res. Rev., 34, 2839.
Roemmich, D., G. C. Johnson, S. Riser, R. Davis, J. Gilson, W. B. Owens, S. L. Garzoli, C. Schmid, and M. Ignaszewski (2009), The Argo program
observing the global ocean with proling oats, Oceanography, 22, 3443.
Sall
ee, J. B., K. G. Speer, and S. R. Rintoul (2010), Zonally asymmetric response of the Southern Ocean mixed-layer depth to the Southern
Annular Mode, Nat. Geosci., 3, 273279.
Schott, F., M. Visbeck, U. Send, J. Fischer, L. Stramma, and Y. Desaubies (1996), Observations of deep convection in the Gulf of Lions, northern Mediterranean, during the winter of 1991/92, J. Phys. Oceanogr., 26, 505524.
Sprintall, J., and D. Roemmich (1999), Characterizing the structure of the surface layer in the Pacic Ocean, J. Geophys. Res., 104, 23,297
23,311.
Sverdrup, K. A., and V. Armbrust (2008), An Introduction to the Worlds Oceans, 10th ed., 528 pp., McGraw-Hill, Columbus, Ohio.
Thomson, R. E., and I. V. Fine (2003), Estimating mixed layer depth from oceanic prole data, J. Atmos. Oceanic Technol., 20, 319329.

CHEN AND YU

C 2015. The Authors.


V

595

Вам также может понравиться