4 Comparison of Sea Surface and Mixed Layer Temperatures
7 Semyon A. Grodsky, James A. Carton, and Hailong Liu
11 April 16, 2008
12 Submitted to the Journal of Geophysical Research, Oceans
17 Department of Atmospheric and Oceanic Science
18 University of Maryland, College Park, MD 20742
20 Corresponding author:
25 Mixed layer temperature, TML , and SST are frequently used interchangeably or assumed to be
26 proportional in climate studies. This study examines the historical observational record 1960-
27 2007 for systematic differences between these variables. The results show that globally and time
28 averaged TML is lower than SST by approximately 0.1 oC. TML minus SST is even lower in
29 upwelling zones where abundant net surface warming is compensated for by cooling across the
30 base of the mixed layer. In the upwelling zone of the Equatorial East Pacific this negative TML -
31 SST difference varies out of phase with seasonal SST, reaching a negative extreme in boreal
32 spring when SST is warm, solar radiation is high, and winds are weak. In contrast, on interannual
33 timescales TML -SST varies in phase with SST with small differences during El Niños as a result
34 of low solar heating and enhanced rainfall. On shorter diurnal timescales during El Niños TML -
35 SST differences associated with temperature inversions occur in response to nocturnal cooling in
36 presence of nearsurface freshening. Near surface freshening produces persistent shallow (a few
37 meters depth) warm layers in the northwestern Pacific during boreal summer when solar heating
38 is strong. In contrast, shallow cool layers occur in the Gulf Stream area of the Northwest Atlantic
39 in boreal winter when fresh surface layers developed due to lateral interactions are cooled down
40 by abundant turbulent heat loss. The different impacts of shallow barrier layers on near surface
41 temperature gradients are explored with a one dimensional mixed layer model.
43 1. Introduction
44 Although instantaneous thermodynamic fluxes across the ocean-atmosphere interface are
45 affected by the temperature of the near surface ocean (< 1 m), many climate studies identify the
46 vertically average temperature of the ocean mixed layer to be the most relevant parameter or
47 even use a slab mixed layer as a proxy for the ocean [see e.g. Manabe and Stouffer, 1996]. In
48 general we may expect monthly depth average mixed layer temperature to be lower than SST by
49 a few tenths of a degree. This difference reflects the time average effect of the nearsurface
50 suppression of turbulence by daytime warming. In this study we compare historical analyses of
51 SST with contemporaneous temperature and salinity profile observations to identify the
52 conditions giving rise to systematic differences between mixed layer temperature and SST and to
53 identify the regions where this difference is essential.
55 SST is a difficult parameter to define exactly because the upper 10 m of the ocean has such
56 complex and variable vertical temperature stratification1. This variation in stratification occurs
57 more frequently under conditions in which the ocean surface fluxes cause gains or loses of heat
58 or freshwater or in situations of strong horizontal exchange. Surface fluxes are responsible for a
59 distinct diurnal cycle in the temperature in the uppermost few meters over wide areas of the
60 ocean when winds are weak and solar heating is strong. [Stuart-Menteth et al., 2003; Gentemann
61 et al., 2003; Clayson and Weitlich, 2007; Kawai and Wada, 2007]. This diurnal cycle is
62 particularly prominent in upwelling areas such as the eastern equatorial Pacific where vertical
63 advection of cool water leads to shallow stratification and thus shallow mixed layers (Deser and
64 Smith, 1998; Cronin and Kessler, 2002). In the warm pool region of the western equatorial
65 Pacific diurnal warming arises because the excess rainfall forms a nearsurface barrier layer of
See the GODAE Global High Resolution SST Pilot Project at http://www.ghrsst-pp.org/SST-Definitions.html
66 low salinity water even though the seasonal thermocline is rather deep [Soloviev and Lukas,
69 In order to reduce the impact of diurnal effects the UK Met Office HadISST1 utilizes only the
70 night satellite SSTs (available beginning in 1981) and adjusts them to match in-situ
71 measurements collected by voluntary observing ships, drifters, and buoys (Rayner et al. 2003).
72 The NOAA National Climatic Data Center SST extended analysis uses both day and night
73 satellite SSTs only to evaluate the spatial structure of analysis SST while rely on the same in situ
74 observations to adjust their SST analysis to reflect water temperature at an effective depth at ~1-
75 5 m (Smith and Reynolds 2003). A more precise definition of this analysis depth is impractical
76 for either product because of the variety of depths at which the in situ observations are available.
77 Most recently the Global Ocean Data Assimilation Experiment High Resolution SST Project has
78 introduced the concept of ‘Foundation SST’, defined as the temperature at a depth (10m) that is
79 below the depth of the diurnal cycle. But this 10m depth temperature, which generally lies
80 within the mixed layer, has not been measured frequently enough to calibrate the analyses.
82 In this study we focus on the difference between the mixed layer temperature and SST provided
83 by historical analyses of Rayner et al.  and Smith and Reynolds . The mixed layer is
84 defined as the near-surface layer of uniform properties such as temperature and salinity. The
85 presence of weak stratification and the nearness to atmospheric momentum sources give rise to
86 values of the Richardson number consistent with flow instabilities and thus a high potential for
87 turbulent motion. Under conditions where density is primarily determined by temperature de
88 Boyer Montégut et al.  (with a generalization introduced by Kara et al., 2000a) define the
89 base of the seasonal mixed layer to be the depth at which temperature changes by 0.2C from its
90 value at 10m Foundation depth. From this we can define a seasonal mixed layer temperature,
91 TML , as the vertical average temperature of the mixed layer, which when multiplied by the depth
92 of the mixed layer and the specific heat of seawater gives the heat capacity of the layer of ocean
93 in direct contact with the atmosphere on seasonal timescales.
95 The near surface processes that affect the monthly TML minus SST difference, dT= TML -SST, are
96 dominated by the integrated effect of diurnal warming. But, a variety of processes including
97 rain, river discharge, or lateral interactions may produce fresh barrier layers that trap the heat
98 near the surface by shoaling the penetration depth of wind stirring and nocturnal convection
99 [Lukas and Lindstrom, 1991; Soloviev and Lukas, 1997]. Moreover stable salinity profiles may
100 permit nocturnal temperature inversions due to radiative cooling [Anderson et al., 1996; Cronin
101 and Kessler, 2002] with magnitudes comparable to those of diurnal warming. Barrier layers are
102 observed over wide ocean areas, in particular, they are produced by abundant rainfall and river
103 discharge in the tropics, by an excess precipitation over the North Pacific, and by lateral
104 exchanges across the western boundary currents [de Boyer Montégut et al., 2007]. In all these
105 areas we also expect significant stratification of near surface layers that affect the difference
106 between TML and SST.
108 2. Data and Methods
109 The mixed layer properties for this study are estimated from individual temperature profiles
110 provided by World Ocean Database 2005, WOD05, [Boyer et al., 2006] for the period 1960
111 through 2004. We use data from the mechanical bathythermographs (MBT), expendable
112 bathythermographs (XBT), conductivity-temperature-depth casts (CTD), as well ocean station
113 data (OSD), moored buoys (MRB), and drifting buoys (DRB). The final four years of the
114 database contain an increasing number of profiles from the new Argo system (PFL). The Argo
115 profiles for the period through 2007 are obtained from the Argo Project web site. For better
116 characterization of the tropical Pacific region the data provided by the TAO/TRITON moorings
117 [McPhaden et al., 1998] are also used.
119 The mixed layer depth (MLD) may be defined in a number of different ways. In this study we
120 use the concept of the isothermal mixed layer depth (MLD) that is evaluated from individual
121 vertical profiles based on the temperature difference from the temperature at a reference depth of
122 10 m [de Boyer Montégut et al., 2004]. This reference depth was shown to be sufficiently deep to
123 avoid aliasing by the diurnal signal, but shallow enough to give a reasonable approximation of
124 monthly TML . Here the isothermal MLD is defined as the depth at which temperature changes by
125 | T | = 0.2oC relative to its value at 10m depth. Following Kara et al. [2000a], the isothermal
126 MLD is defined by the absolute difference of temperature, | T |, rather than only the negative
127 difference of temperature to account for mixed layers with temperature inversions in salt-
128 stratified situations (most common at high latitudes). The mixed layer temperature is evaluated
129 as the temperature vertically averaged above the base of the mixed layer assuming uniform
130 temperature above the reference depth, T ( z 10m) T ( z 10m) .
132 An alternative definition of the mixed layer depth (based on the dynamical stability criterion)
133 defines it as the depth of a density uniform layer. Vertically average temperature of temperature
134 uniform layer is the same as vertically average temperature of density uniform layer if the latter
135 layer is not deeper than the former (barrier layer). If a density uniform layer is deeper than a
136 temperature uniform layer (density compensation) their average temperatures may be different.
137 Here we follow de Boyer Montégut et al.  and define the mixed layer as a layer vertically
138 uniform in both, temperature and salinity. Hence, the mean mixed layer temperature is the same
139 as the mean temperature of isothermal layer. The latter is referred in this study as the mixed layer
140 temperature, T ML .
142 The mixed layer archive and the seasonal and interannual variability of mixed layer properties
143 are described by Carton et al. . They show that the temperature difference criterion works
144 reasonably well even at high latitudes in the North Atlantic and provide further details on data
145 quality control procedures. After estimating TML at each profile location we then apply
146 subjective quality control to remove ‘bulls eyes’ and bin the data into 2ox2ox1mo bins with no
147 attempt to fill in empty bins.
149 SST is provided by Met Office Hadley Centre sea ice and sea surface temperature (HadISST1) of
150 Rayner et al.  and by extended analysis (version 2) of Smith and Reynolds . Both
151 products are globally complete monthly averaged grids spanning time period beginning the late
152 19-th century onward. HadISST1 combines a suite of historical and modern in situ near surface
153 water temperature observations from ships and buoys with the recent satellite SST retrievals,
154 while the Smith and Reynolds  data is mostly based on in-situ measurements. Neither of
155 these products uses the vertical temperature profiles from WOD05. The SST provided by these
156 archives is based to a large extend on measurements collected by voluntary observing ships and
157 is referred here as bulk SST or simply SST. Data adjustment to measurements taken from a few
158 meters depth (where the diurnal signal is relatively weak) effectively attenuates but doesn’t
159 eliminate completely impacts of transient near surface processes on bulk SST.
161 The local response of the mixed layer to the forcing from the atmosphere is simulated using the
162 one dimensional hybrid mixed layer model of Chen et al. . This model is based on the
163 Kraus-Turner-type bulk mixed layer physics in which the depth of the mixed layer is determined
164 by a turbulent energy balance equation, while the temperature and salinity of the mixed layer is
165 determined by budget equations forced by surface fluxes and entrainment. These balances are
166 augmented in the Chen et al. formulation by the addition of convection and Richardson Number-
167 dependent mixing. The model is forced by 6-hour surface fluxes provided by the National Center
168 for Environmental Predictions/Department of Energy (NCEP/DOE) Reanalysis-2 of Kanamitsu
169 et al. .
171 3. Results
172 We begin by examining the average dT based on the 1960-2004 WOD05 data set (Figs. 1a,b).
173 Because of the distribution of observations only the Northern Hemisphere is well sampled. On
174 average, T ML is colder than SST by about -0.1oC, with large <-0.4C negative anomalies in the
175 region north of the Kuroshio-Oyashio extension and along the equator in the eastern Pacific, and
176 large >0.4C positive anomalies (temperature inversions) in the Gulf Stream region (the results
177 are similar for the two SST analyses). The equatorial Atlantic shows negative anomalies as well,
178 but not as large as the equatorial Pacific.
180 To illustrate this relationship in the Southern Hemisphere we examine average dT using the
181 spatially more homogeneous Argo profile data set which, however, is mainly restricted to the
182 years 2004-onward (Fig. 1c). The Argo results in the Northern Hemisphere show only a few
183 differences from the distribution of dT based on the WOD05 data set. In the Labrador Sea
184 positive values of dT (indicating nearsurface temperature inversions) are now more evident. In
185 contrast, the subtropical North Atlantic and North Pacific both show negative values in the
186 regions of weak winds where diurnal warming of the nearsurface is a frequent occurrence. In the
187 Southern Hemisphere dT based on Argo shows large negative anomalies at several longitudes,
188 in the South Pacific west of Chile as well as southwest of Australia and South of Cape of Good
189 Hope. We next focus on the Northern Hemisphere patterns because they are evaluated from
190 longer time records then the southern counterparts. To explore the causes of the largest
191 anomalies of dT we next examine in detail the time changes in the three regions in the Northern
192 Hemisphere identified in Fig. 1.
194 These three regions are distinguished by persistent shallow near surface stratification due to
195 either upwelling or impact of the barrier layers (nearsurface freshening) that trap warming
196 (cooling) in the near surface. On the other hand, the air-sea interactions are particularly strong
197 over these regions. It is illustrated by climatological maps of the net surface heat gain by the
198 ocean. During the northern winter (Fig. 2a) the turbulent heat loss in excess of 200 Wm-2 occurs
199 over the warm western boundary currents due to strong air-sea temperature contrast and
200 enhanced evaporation over warm SSTs. In northern summer (Fig. 2b) the ocean gains heat in
201 excess of 150 Wm-2 in the northwestern Pacific and over the shelf waters north of the Gulf
202 Stream. In both these areas the local increase of the ocean heat gain is due to a decreased
203 evaporation over cool SSTs. The ocean also gains heat at a rate exceeding 100 Wm-2 in the
204 eastern equatorial Pacific cold tongue (Fig. 2b) due to abundant solar radiation and relatively
205 weak latent heat loss. In the cold tongue the heat gain is compensated for by entrainment cooling.
206 In the near surface it produces remarkable magnitudes of diurnal warming. We shall next analyze
207 the origins of persistent shallow stratifications in these three regions.
209 3.1 Eastern Equatorial Pacific
210 The equatorial Pacific thermocline shoals eastward in response to annual mean easterly winds
211 that along with entrainment cooling form a tongue of cool water in the east. Here, in the cold
212 tongue, the ocean gains heat from the atmosphere in excess of 100 Wm-2 (Fig.2b) that is
213 compensated for by entrainment cooling. In response to this surface heat flux the nearsurface
214 ocean develops substantial diurnal warming of SST, in excess of 0.2C in time average [Deser
215 and Smith, 1998]. Here dT averages approximately -0.4oC (Fig. 3a) with more negative values
216 ( T ML <SST) in March when SST reaches its monthly maximum and diurnal warming is large
217 (Fig. 3b) [Cronin and Kessler, 2002]. In contrast, on interannual timescales dT is weak
218 ( TML SST) when El Nino warms SST, the mixed layer deepens, solar radiation decreases and
219 freshwater input increases, and dT has negative extreme during the La-Niñas when the mixed
220 layer shoals and atmospheric convection shifts westward [Cronin and Kessler, 2002; Clayson
221 and Weitlich, 2005]. This relationship is most clear after the early 1980s, as the data coverage
224 In order to understand the causes of the seasonal and interannual relationships we examine
225 conditions at the Tropical Ocean Atmosphere/TRITON mooring at 0N, 140W for the seven years
226 1995-2001 encompassing the 1997-98 event (Fig. 4a). We focus on 0N, 140W location where
227 the records are continuous during the event. At this location 1m temperature, a proxy for SST,
228 increases by 5C during 1997 and then decreases by nearly 7C in mid-19982. Coinciding with the
229 drop in 1m temperature is a substantial development of negative dT meaning that the mixed
230 layer has developed some nearsurface temperature stratification. The negative values of dT are
231 even more striking in 1999 and 2000 when SST increases during January-March as part of the
232 climatological seasonal cycle at this location phase with interannual variation of dT .
234 To identify the mechanisms giving rise to differences in seasonal and ENSO changes in dT we
235 examine a one dimensional mixed layer model simulation beginning with homogeneous initial
236 conditions (Fig. 4b). The model responds seasonally to weakened winds in boreal spring (Fig.
237 4d) with increased near-surface stratification ( dT <0) as observed. The conditions arising during
238 the onset of El Nino such as occurred during the first half of 1997 are somewhat different.
239 During those months the winds also weakened, but solar heating decreased (Fig. 4c) and
240 freshwater input increased (Fig. 4d) as a result of the eastward shift of convection. The decrease
241 in the ocean heat gain due to decreased solar heating is accompanied by increased latent heat loss
242 due to warmer SST (Fig. 4c). The result is weakening values of dT followed in the summer and
243 fall by occasional temperature inversions. In mid-1998 through early 1999 as El Nino
244 transitioned into cooler La Nina conditions, the nearsurface again becomes strongly stratified due
245 to enhanced solar heating and weaker latent heat loss and resulting diurnal warming of the
TAO/TRITON moorings measure SST at z=1m. Time mean difference of T1m from HadISST1 at 0N, 140W is -
0.3C while time correlation is 0.96.
248 Intermittent temperature inversions (0.2-0.5C cooler SSTs) are evident in observations (Fig. 4a)
249 and simulations (Fig. 4b). They are associated with nocturnal cooling of shallow freshwater
250 lenses produced by enhanced rainfall (Fig. 4d). Stable salinity stratification (barrier layer)
251 produced by local rainfall captures the nocturnal convection in the near surface layer until the
252 cooling or wind stirring is strong enough. If the freshwater surface flux is set to zero, the one
253 dimensional model doesn’t simulate temperature inversions [see also Anderson et al., 1996].
255 As we have seen the stable salinity stratification produced by local rainfall may impact
256 significantly the near surface temperature stratification. An alternative mechanism of barrier
257 layers formation is associated with the lateral interactions. In particular, in the equatorial Pacific
258 near the dateline, salt and warm water can be subducted under the western Pacific warm fresh
259 water to form barrier layers [Lukas and Lindstrom, 1991]. This advection mechanism (that is not
260 in a one dimensional model physics) may be effective near the frontal interfaces and contribute
261 to temperature inversions during the seasons when the ocean loses heat.
263 3.2 Gulf Stream
264 We next consider dT in the western North Atlantic where time mean positive values
265 (temperature inversions) lie along the Gulf Stream path (Fig. 1). This regional anomaly may
266 result from differences in spatial interpolation of T ML and bulk SST. However, we’ll next see in
267 Fig. 3d that dT in the Gulf Stream region displays noticeable seasonal variations suggesting that
268 physical processes play a role as well.
270 Spatial patterns of dT during the two contrasting seasons are illustrated in Fig.5 using CTD and
271 Argo profiles. In distinction from Fig. 1 where T ML is compared with bulk SSTs from climate
272 archives, the SST in Fig. 5 is taken from the same profiles as TML . This eliminates contribution
273 due to the difference in spatial interpolation of SST and T ML . In summer dT is less than zero
274 (Fig. 5a) due to the formation of a shallow a few meter deep diurnal warming trapped in a 10-
275 30m shallow fresh layer (Fig. 6a). This shallow barrier layer limits the depth of nocturnal
276 convection and mechanical stirring above the base of halocline and thus separates the shallow
277 near surface warm layer from the seasonal mixed layer. In contrast a near-surface temperature
278 inversion is largest in the cold seasons (Fig. 5b) when heat loss is very large (Fig.2a, also see
279 Dong and Kelly, 2004). An examination of the spatial structure of dT during the winter months
280 (Fig. 5b), indeed, shows large inversions, frequently exceeding 1C along the path of the Gulf
281 Stream, while SST is close to T ML in this area in summer. Variations of dT are similar if an
282 alternative, gradient-based MLD is used (Fig.2c).
284 Seasonal variations of dT in the Gulf Stream region occur in accord with the seasonal variations
285 of the net surface flux that displays a peak in heat loss over the warm Gulf Stream in winter (Fig.
286 2a) and strong warming over the cool shelf water in summer (Fig. 2b). In distinction from the
287 equatorial Pacific where interannual dT significantly correlates with local SST, these values are
288 weakly correlated in the Gulf Stream area (Fig. 3c). This weak correlation reflects probably an
289 impact of interannual migration of the Gulf Stream front that affects both the box averaged SST
290 and dT .
292 Winter mixed layer is occasionally warmer than SST in the Gulf Stream sector (Fig. 5b).
293 Associated temperature inversions are aligned along the Gulf Stream northern wall suggesting
294 that lateral cross-frontal interactions between water masses may play a role. Subduction of warm
295 and salt water carried by the Gulf Stream below the cold and fresh shelf water produces a
296 shallow cold layer (Fig. 6b) that is further cooled down as the ocean losses heat. The temperature
297 inversion doesn’t overturn until the stable salinity stratification is overcome by instability
298 introduced by near surface cooling or wind stirring. In distinction from numerous occurrences of
299 the near surface warm layers in summer (Fig. 5a), the near surface cold layers are observed only
300 sporadically in winter (Fig. 5b). In fact, they are destroyed by transient storms that stir these
301 shallow density compensated layers. Despite their intermittence, the near surface temperature
302 inversions might impact wintertime infrared SSTs because passing storms that eventually destroy
303 the inversions are normally associated with the cold air outbreaks and significant convection
306 Discussions above emphasize impacts of salinity on the nearsurface temperature stratification.
307 Next the temperature response to the presence of the near surface salinity gradients (occurring in
308 the Gulf Stream area) is explored with one-dimensional mixed layer model (Fig. 7). To contrast
309 the impact of salinity, the twin runs are compared. Each pair of model runs is forced by the same
310 fluxes but differs in initial conditions. The first (control) run starts from the vertically
311 homogeneous temperature and salinity while the initial salinity profile for the second run has
312 salinity decreasing toward the surface in the upper 20 m at a rate of 0.1 psu m-1 (in accord with
313 observations in Fig. 6).
315 Fig. 7b illustrates simulations during the warm season. It displays the difference in temperature
316 between the two runs that evidents an impact of the near surface freshening. In the presence of
317 stabilizing salinity gradient the diurnal warming is stronger during the first day of simulations
318 (Fig. 7b), but is surprisingly similar during the second day when it is limited by the shear
319 instability of diurnal currents. Relative warming in the upper 20 m is even stronger as wind
320 strengthens. This is explained by slower deepening of the mixed layer and weaker entrainment
321 cooling in salinity stratified case. Although the one-dimensional mixed layer model simulates
322 warmer near surface temperature in salinity stratified case, the simulated temperature
323 stratification in the upper 10 m column doesn’t exceed a few tenth of degree in contrast with
324 observations (Fig. 6a). This is explained in part by relatively short (only a few days long) run as
325 well as by limitations of the model. If a strong ( ~ 1 day) relaxation of salinity to its initial
326 conditions is introduced (to account indirectly for mechanisms producing shallow halocline) the
327 temperature gradient in the upper 10m amplifies up to 1 oC but never reaches values shown in
328 Fig. 6a.
330 In winter the mixed layer model simulates 1oC colder mixed layer in salt stratified case than in
331 control run (Fig. 7d). The difference is due to the stably stratified halocline that limits
332 penetration depth of wind stirring. In turn, the shallower mixed layer is the faster it cools down
333 due to net surface heat loss. Although the anomalous cooling of 1oC compares well with
334 observations (Fig. 6b), the simulated mixed layer is relatively deep. Therefore, the stratification
335 is weak in the upper 10 m in distinction from observations. This suggests again that lateral
336 interactions (missing by one-dimensional model) are important for establishing winter
337 temperature inversions in the region, while the net surface heat loss further amplifies existing
340 3.3 Northwestern Pacific
341 Salinity in the Northwestern Pacific decreases towards the surface. This stable halocline is
342 produced by annual mean excess of precipitation over evaporation north of 30N and is
343 maintained by upward vertical pumping driven by cyclonic wind curl [Kara et al., 2000b].
344 Although the regional precipitation peaks in winter, the near surface freshening persists year
345 around. In summer when the ocean heating is particularly strong (Fig. 2b), the shallow stably
346 stratified halocline localizes the ocean heat uptake in the nearsurface layer (Fig.1) by limiting the
347 penetration depth of wind stirring and nocturnal convection. In distinction from the Gulf Stream
348 region where shallow warm layers develop mostly in the cold sector of the front, the shallow
349 warm layers are observed randomly in the Northwestern Pacific (Fig.5c). They are not destroyed
350 by nocturnal convection (see sample profile taken at 20:30 local time, Fig. 8a). Meridional
351 variations of dT follow the meridional variations of net surface heating and are similar if
352 different a gradient-based definition of the mixed layer depth is used (Fig. 2d). Occasional SST
353 inversions seeing in Fig. 5c are associated with nocturnal cooling of freshwater lenses (Fig. 8b).
355 Shallow stratified layers observed in the Northwestern Pacific in summer are destroyed by winter
356 storms (Fig. 5d and Fig. 3f). Despite similarly strong heat loss over the warm western boundary
357 currents (Fig. 2a), the winter SST inversions are not observed in the Kuroshio region in
358 distinction from the Gulf Stream region (Fig.1). This may be linked to the differences in spatial
359 patterns of salinity. In fact, the spatial gradients of salinity (that are vital for producing the
360 temperature anomalies) are significantly weaker in the Northwestern Pacific in comparison with
361 the Northwestern Atlantic [see e.g. Antonov et al., 2006].
363 4. Summary
364 This study compares the magnitudes of two ocean temperature variables frequently used in
365 climate studies, mixed layer temperature and bulk SST as represented by the widely used
366 analyses of Rayner et al.  and Smith and Reynolds . Mixed layer temperature is
367 defined as the vertical average temperature above the mixed layer base, and the depth of the base
368 here is defined following Kara et al. [2000a] and de Boyer Montégut et al.  as a function
369 of the temperature difference relative to 10m temperature. Our analysis shows that areas with
370 shallow temperature stratification such as upwelling zones frequently have significant
371 differences between mixed layer temperature and SST. Shallow temperature stratification also
372 occurs in regions of near surface freshening (barrier layers) which limits the depth of convection
373 and wind stirring. In both cases shallow stratification occurs in zones of strong air-sea heat
374 exchange. In the northern hemisphere the local peaks of heat gain by the ocean are observed in
375 local summer over the areas of equatorial cold tongues and over the areas of cold SSTs to the
376 north of the Kuroshio extension front and the Gulf Stream north wall. While in winter the ocean
377 loses much heat over the warm SSTs of the western boundary currents.
379 We examine the temporal relationship between SST and T ML in the Equatorial East Pacific
380 where abundant net surface warming is compensated for by cooling across the base of the mixed
381 layer. Here T ML is persistently cooler than SST by approximately -0.4oC. On seasonal time
382 scales, it has a negative extreme during the boreal spring warm season when winds are weak. In
383 contrast, on interannual timescales the magnitude of dT = TML -SST increases during La Ninas
384 and weakens during El Niños as a result of increases/decreases in solar radiation and
385 decreases/increases in precipitation. Increased precipitation during El Niños produces freshwater
386 stratified barrier layers leading to nocturnal cooling.
388 In the subtropics negative values of dT are found in the Gulf Stream area of the western North
389 Atlantic. In summer the shallow warming in excess of 1 oC develops above the cool shelf waters
390 to the west and north of the Gulf Stream where the ocean gains heat at a rate of 150 Wm-2. The
391 presence of nearsurface freshening prevents the nighttime destruction of this shallow warm layer.
392 In contrast, during winter the near surface layer within the Gulf Stream itself has an inverted
393 temperature structure (the time averaged dT=0.6C) as the result of strong surface cooling in the
394 presence of a nearsurface barrier layer.
396 Another region where the salinity stratified barrier layers are present is the Kuroshio extension
397 region of the Northwest Pacific. Here the barrier layer is produced due to excess of precipitation
398 accompanied by upward Ekman pumping preventing the vertical exchange of this freshwater. As
399 in the case of the Gulf Stream region, the ocean gains heat in the summer at a rate of 150 Wm-2
400 producing a warm surface layer during the day which has the time averaged dT =-0.5 oC. In
401 winter, T ML and SST match in this region.
403 One of the persistent issues in coupled atmosphere-ocean general circulation models is tendency
404 to develop cold biases in the eastern equatorial Pacific [Davey et al. 2002]. However the surface
405 temperature of such models is actually more analogous to mixed layer temperature since the
406 uppermost ocean gridpoint is well below the ocean surface and diurnal processes are generally
407 neglected. Thus, any systematic differences in SST and T ML are likely to be reflected in the
408 evaluation of model SST bias. Indeed, Danabasoglu et al.  have shown that adding the
409 diurnal cycle to the daily mean incoming solar radiation does warm the model eastern equatorial
410 Pacific SST and shoals the ocean boundary layer in better agreement with SST observations.
411 Even greater improvements to model SST estimates seem possible if the nearsurface
412 stratification of temperature and salinity can be more accurately represented.
415 Acknowledgements. We gratefully acknowledge the Ocean Climate Laboratory of the National
416 Oceanographic Data Center/NOAA, under the direction of Sydney Levitus for providing the
417 database upon which this work is based. Mixed layer temperature estimate based on the
418 Lorbacher et al.  approach has been provided by Dietmar Dommenget, IFM-GEOMAR.
419 Support for this research has been provided by the National Science Foundation (OCE0351319)
420 and the NASA Physical Oceanography Programs.
423 Antonov, J. I., R. A. Locarnini, T. P. Boyer, A. V. Mishonov, and H. E. Garcia, 2006: World
424 Ocean Atlas 2005, Volume 2: Salinity. S. Levitus, Ed. NOAA Atlas NESDIS 62, U.S.
425 Government Printing Office, Washington, D.C., 182 pp.
426 Anderson, S.P., R.A. Weller, and R.B. Lukas, 1996: Surface Buoyancy Forcing and the Mixed
427 Layer of the Western Pacific Warm Pool: Observations and 1D Model Results. J.
428 Climate, 9, 3056–3085.
429 Boyer, T.P., J.I. Antonov, H.E. Garcia, D.R. Johnson, R.A. LocarNiñi, A.V. Mishonov, M.T.
430 Pitcher, O.K. Baranova, and I.V. Smolyar, 2006: World Ocean Database 2005. S.
431 Levitus, Ed., NOAA Atlas NESDIS 60, U.S. Government Printing Office, Washington,
432 D.C., 190 pp., DVDs.
433 de Boyer Montégut, C., G. Madec, A. S. Fischer, A. Lazar, and D. Iudicone, 2004: MLD over
434 the global ocean: An examination of profile data and a profile-based climatology, J.
435 Geophys. Res., 109, C12003, doi:10.1029/2004JC002378.
436 de Boyer Montégut, C., J. Mignot, A. Lazar, and S. Cravatte (2007), Control of salinity on the
437 mixed layer depth in the world ocean: 1. General description, J. Geophys. Res., 112,
438 C06011, doi:10.1029/2006JC003953.
439 Carton, J.A., S. A. Grodsky, and H. Liu (2008), Variability of the Oceanic Mixed Layer 1960-
440 2004, J. Climate, 21, 1029–1047.
441 Chen, D., L.M. Rothstein, and A.J. Busalacchi (1994), A hybrid vertical mixing scheme and its
442 application to tropical ocean model, J. Phys. Oceanogr., 24, 2156-2179.
443 Clayson, C.A., and D. Weitlich, 2007: Variability of Tropical Diurnal Sea Surface Temperature.
444 J. Climate, 20, 334–352.
445 Cronin, M.F., and W.S. Kessler, 2002: Seasonal and interannual modulation of mixed layer
446 variability at 0N, 110W, Deep-Sea Research I, 49, 1–17.
447 Danabasoglu, G., W.G. Large, J.J. Tribbia, P.R. Gent, B.P. Briegleb, and J.C. McWilliams,
448 2006: Diurnal Coupling in the Tropical Oceans of CCSM3. J. Climate, 19, 2347–2365.
449 Davey, M., M. Huddleston, K.R. Sperber et al. (2002), STOIC: A study of coupled model
450 climatology and variability in tropical ocean regions. Clim. Dyn., 18, 403-420.
451 Deser, C. and C. A. Smith (1998): Diurnal and semidiurnal variations of the surface wind field
452 over the tropical Pacific Ocean. J. Climate, 11, 1730–1748.
453 Dong, S., and K.A. Kelly (2004): Heat Budget in the Gulf Stream Region: The Importance of
454 Heat Storage and Advection. J. Phys. Oceanogr., 34, 1214–1231.
455 Gentemann, C. L., C. J. Donlon, A. Stuart-Menteth, and F. Wentz (2003), Diurnal signals in
456 satellite sea surface temperature measurements, Geophys. Res. Lett., 30(3), 1140,
458 Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.K. Yang, J.J. Hnilo, M. Fiorino, and G.L. Potter,
459 2002: NCEP–DOE AMIP-II Reanalysis (R-2), Bull. Amer. Meteor. Soc., 83, 1631–1643.
460 Kara, A. B., P. A. Rochford, and H. E. Hulburt (2000a), Mixed layer depth variability and barrier
461 layer formation over the North Pacific Ocean, J. Geophys. Res., 105, 16,803–16,821.
462 Kara, A. B., P. A. Rochford, and H. E. Hulburt (2000b), Mixed layer depth variability and barrier
463 layer formation over the North Pacific Ocean, J. Geophys. Res., 105, 16,783-16,801.
464 Kawai, Y., and A. Wada (2007), Diurnal Sea Surface Temperature Variation and Its Impact on
465 the Atmosphere and Ocean: A Review, J. Oceanogr., 63, 721-744.
466 Lorbacher, K., D. Dommenget, P. P. Niiler, and A. Köhl, 2006: Ocean MLD: A subsurface
467 proxy of ocean atmosphere variability, J. Geophys. Res., 111, C07010,
469 Lukas, R., and E. Lindstrom (1991), The mixed layer of the western equatorial Pacific Ocean, J.
470 Geophys. Res., 96, 3343– 3357.
471 Manabe, S., and R.J. Stouffer (1996), Low-frequency variability of surface air temperature in a
472 1000-year integration of a coupled atmosphere-ocean-land surface model, J. Climate, 9
473 (2), 376-393.
474 McPhaden , M.J. , and Coauthors (1998), The Tropical Ocean-Global Atmosphere observing
475 system: A decade of progress, J. Geophys. Res., 103(C7), 14,169-14,240.
476 Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C.
477 Kent, and A. Kaplan (2003), Global analyses of sea surface temperature, sea ice, and
478 night marine air temperature since the late nineteenth century, J. Geophys. Res.,
479 108(D14), 4407, doi:10.1029/2002JD002670.
480 Smith, T. M., and R. W. Reynolds (2003), Extended reconstruction of global sea surface
481 temperatures based on COADS data (1854 –1997), J. Clim., 16, 1495– 1510.
482 Soloviev, A., and R. Lukas (1997), Observations of large diurnal warming events in the near-
483 surface layer of the western equatorial Pacific warm pool, Deep Sea Res., 44, 1055–
485 Stuart-Menteth, A. C., I. S. Robinson, and P. G. Challenor (2003), A global study of diurnal
486 warming using satellite-derived sea surface temperature, J. Geophys. Res., 108(C5),
487 3155, doi:10.1029/2002JC001534.
490 Figure 1. Time mean difference, dT , of mixed layer averaged temperature, TML , and bulk SST.
491 (a) T ML from WOD05 and bulk SST from HadISST1, (b) T ML from WOD05 and SST
492 from Smith and Reynolds extended v.2. In (a) and (b) grid points with less than one year
493 of data aren’t shown. (c) T ML from Argo floats and bulk SST from HadISST1 at grid
494 points with at least 6 months of data. dT is the global and time mean difference.
498 Figure 2. (Left) Climatological net surface heat flux for (a) January, (c) August (solid contours
499 indicate heat gain by the ocean). Boxes show the same areas as shown in Fig.1a. (Right)
500 Mixed layer minus SST temperature difference, dT , zonally averaged over longitude
501 belts shown in the left panels. Solid lines show result of this study (against bottom x-
502 axis) while dashed lines show results based on depth estimates of Lorbacher et al. 
503 (against top x-axis) that is based on the gradient-based definition of the mixed layer
510 Figure 3. (a), (c) ,(e) Time series of annual running mean box-averaged dT , standard deviation
511 of dT (shading), and anomalous SST for the three boxes, equatorial east Pacific, Gulf
512 Stream, and northwestern Pacific. (b), (d), (f) Seasonal cycle of box-averaged dT and
513 SST based on HadISST1 data. Time series combine dT evaluated from WOD05 data
514 through 2004 and Argo data afterwards.
517 Figure 4. (a) Time series of 1m temperature, T1m , and mixed layer temperature gradient, dT ,
518 from (a) TAO/TRITON mooring at 0N, 140W, (b) mixed layer model (MLM). (c)
519 monthly running mean short wave radiation (SWR) and latent heat loss (LHTFL), (d) 6-
520 hour precipitation (PRECIP) and monthly zonal wind stress (TAUX).
523 Figure 5. dT during June-August (JJA) and October-March (ONDJFM) evaluated from
524 individual CTD and Argo profiles in (a,b) Gulf Stream area, (c,d) northwestern Pacific.
525 Circles mark locations of vertical profiles shown in Figs. 6 and 8.
529 Figure 6. Sample vertical profiles of temperature and salinity with (a) negative and (b) positive
530 dT . Profiles are taken in the northwestern Atlantic in (a) boreal summer and (b) boreal
531 winter at locations shown in Figs. 5a and 5b, respectively. ‘LT’ indicates the local sun
536 Figure 7. Mixed layer model response to sample winds and net surface flux in the Gulf Stream
537 area in (a,b) summer (c,d) winter as a function of salinity stratification. (b,d) Difference
538 of temperature between two simulations with the same surface forcing and with the same
539 vertically uniform temperature initial conditions but with different salinity initial
540 conditions. In the first experiment initial salinity is vertically uniform while in the second
541 initial salinity has a uniform vertical gradient, S / z =0.1 psu m-1.
545 Figure 8 Sample vertical profiles of temperature and salinity with (a) negative and (b) positive
546 dT . Profiles are taken in the northwestern Pacific in boreal summer at locations shown
547 in Fig. 5c. ‘LT’ indicates the local sun time.