1234567
Intraseasonal latent heat flux based on satellite observations89
10Semyon A. Grodsky1, Abderrahim Bentamy2, James A. Carton1, and Rachel T. Pinker111
1213141516171819
Submitted to Journal of Climate20October 22, 200821
222324252627282930
1Department of Atmospheric and Oceanic Science University of Maryland, College Park, 31MD 2074232
332 Institut Francais pour la Recherche et l’Exploitation de la Mer (IFREMER), Plouzane, 34France35
3637
Corresponding author: [email protected]
40
1
Abstract40
Weekly average satellite based estimates of latent heat flux (LHTFL, positive if the ocean 41
losses heat) are used to characterize spatial patterns and temporal variability in the 42
intraseasonal band (periods shorter than 3 months). The strongest zonally averaged 43
intraseasonal variability of LHTFL in excess of 30 Wm-2 is observed at midlatitudes44
between 400 S to 100 S and 100 N to 450 N. Intrasesonal variability of LHTFL is locally 45
stronger in the regions of major SST fronts (like the Gulf Stream, Agulhas) where the 46
standard deviation of intraseasonal LHTFL is up to 50 Wm-2. The amplitude of the 47
intraseasonal LHTFL decreases at high latitudes and in the regions of equatorial 48
upwelling, reflecting the effect of decreased SST. In midlatitudes the intraseasonal 49
variability of LHTFL is forced by passing storms and is locally amplified by unstable air 50
stratification over warm SSTs. Although weaker in amplitude, but still significant,51
intraseasonal variability is observed in the tropical Indian and Pacific Oceans due to the 52
eastward propagation of Madden-Julian Oscillations. In this tropical region the 53
intraseasonal LHTFL and incoming solar radiation are out-of-phase, namely, evaporation 54
increases just below the convective clusters. Over much of the global Ocean anomalous 55
LHTFL provides a negative feedback on the underlying intraseasonal SST anomaly, 56
although there are considerable geographical variations. The feedback exceeds 20 Wm-57
2/oC in regions around 20oS and 20oN, but decreases at high latitudes and in the eastern 58
tropical Pacific and Atlantic where the time average LHTFL is weak. 59
60
2
1. Introduction60
Latent heat flux (LHTFL) links the air-sea heat exchange with the hydrological cycle. 61
This evaporative heat loss makes up a significant portion of the ocean net surface flux 62
and compensates in part for the ocean heat gain by solar radiation (da Silva et al., 1994). 63
It has been demonstrated that satellite sensors can measure sea surface temperature 64
(SST), near-surface winds, and humidity, and thus provide data for estimating sea surface 65
evaporation. Currently, several satellite-based global ocean latent heat flux products are 66
available (e.g. Chou et al., 2003 and references therein). 67
68
Most observational examinations of LHTFL focus on its behavior on monthly and longer 69
timescales (e.g., da Silva et al., 1994; Yu et al., 2006). Recent studies of midlatitudes 70
(Qiu et al. 2004) and tropics (Zhang and McPhaden, 2000) have shown that intraseasonal 71
variations of LHTFL associated with synoptic meteorological disturbances can alter SST 72
by up to 1oC. Modeling studies (Maloney and Sobel, 2004) suggest that these SST 73
variations may in turn organize intraseasonal atmospheric convection and thus provide an 74
air-sea interaction mechanism for phenomena such as the 30-60 day Madden-Julian 75
Oscillations and may contribute to the year-to-year variability. Since LHTFL is also 76
proportional to evaporation its intraseasonal variations contribute to variations of surface 77
salinity, thus increasing LHTFL impact on surface density. In this study we exploit the 78
availability of a global 16-year (1992 – 2007) record of weekly turbulent fluxes of 79
Bentamy et al. (2008) to examine the observed geographic distribution of intraseasonal 80
LHTFL and its role in air-sea interactions.81
82
3
The tropical atmosphere is subject to a variety of synoptic-scale disturbances including 83
30-60 day fluctuations known as the Madden-Julian Oscillations (MJO) (Madden and84
Julian, 1994) that are driven in part by the evaporation-wind feedback that involves 85
impacts of the zonal wind perturbations on evaporation (Neelin et al., 1987). MJO is a 86
feature of all tropical sectors, although it is most pronounced over the eastern Indian 87
Ocean and western Pacific Ocean in boreal winter and is strongly modulated by ENSO. 88
MJO is characterized by strong fluctuations of surface winds of 2-4m/s and precipitation89
(Araligidad and Maloney, 2008). As a result it produces correlated fluctuations of both 90
LHTFL and shortwave radiation (SWR) with amplitudes of 30-50 Wm-2 and is observed 91
to result in 0.5oC fluctuations of SST (Krishnamurti et al. 1988; Shinoda and Hendon, 92
1998; Zhang and McPhaden, 2000). Wind-induced surface heat exchange, in which 93
increases in wind-induced LHTFL cause reductions in SST, which further increase heat 94
loss, has been implicated in the maintenance of the MJO (Maloney and Sobel, 2004; Han 95
et al., 2007). Moreover, recent research suggests that intraseasonal fluctuations may 96
actively interact with lower frequency climate variations, just as in the Pacific, where the 97
westerly wind bursts trigger the evolution of the El Niño/Southern Oscillation (ENSO) 98
cycle (McPhaden, 2004). Foltz and McPhaden (2004) examine the intraseasonal (30-70 99
day) oscillations in the tropical and subtropical Atlantic and similarly find a close 100
relationship (correlation of 0.75) between SST change and LHTFL variability in this 101
basin. 102
103
The subtropics and midlatitudes are subject to additional synoptic meteorological forcing 104
originating in the mid-latitude storm systems. This additional variability has a strong 105
4
seasonal component and varies from year-to-year. In the Kuroshio extension region106
Bond and Cronin (2008) find that in late fall through early spring cold air outbreaks 107
associated with synoptic events lead to intense episodes of LHTFL and sensible heat loss. 108
In summer and fall cloud shading effects accompanying synoptic disturbances become 109
important sources of intraseasonal flux variations. Based on experiments with a mixed 110
layer model Qiu et al. (2004) suggest that these summertime intraseasonal flux variations 111
can induce SST variations with climatologically significant ±1oC amplitudes. This and 112
other observational evidence suggest significant contributions by LHTFL variability in 113
the intraseasonal band to the state of the climate system. In this study we focus on 114
geographical patterns of LHTFL, consistency with SST, interplay with incoming solar 115
radiation, as well as modulation by longer period processes.116
117
This study is made possible due to several improvements to the climate observing system. 118
Beginning in the early 1990s a succession of three satellite scatterometers provides high 119
resolution surface winds. Brightness temperature estimates from the Special Sensor 120
Microwave Imager provide an estimate of relative humidity. When combined with 121
estimates of surface temperature it is possible to estimate LHTFL at weekly resolution 122
(Bentamy et al., 2008). Clouds and aerosols, the main factors affecting SWR, are 123
available from a variety of sensors flying in both geostationary and polar orbits (Rossow124
and Schiffer, 1999; Pinker and Laszlo, 1992). Finally, an array of more than 90 moorings 125
distributed across all three tropical oceans (McPhaden et al., 1998) provides ground truth 126
at high temporal resolution which can be used to explore the accuracy of the remotely 127
sensed estimates. 128
5
129
2. Data and method130
This research is based on the recent update of weekly satellite-based turbulent fluxes of 131
Bentamy et al. (2003, 2008). The three turbulent fluxes, wind stress ( τ ), LHTFL ( EQ ), 132
and sensible heat flux ( HQ ) are estimated using the following bulk aerodynamic 133
parameterizations (Liu et al., 1979): 134
135
)( sasaDC uuuuτ−−=
ρ136
)( sasaEE qqC
LQ
−−−= uuρ
(1)137
)( sasaHp
H TTCC
Q−−= uu
ρ,138
139
where ρ is the air density, L =2.45*106 J/kg is the latent heat of evaporation, pC =1005 140
J/kg is the specific heat of air at constant pressure. The turbulent fluxes in (1) are 141
parameterized using wind speed ( sa uu − ) relative to the ocean surface current, the 142
difference of specific air humidity and specific humidity at the air-sea interface ( sa qq − ), 143
and the difference of air temperature and SST ( sa TT − ). The lower indices (a) and (s) 144
indicate atmosphere at the reference level (normally 10m) and at the sea surface, 145
respectively. The bulk transfer coefficients for wind stress ( DC , drag coefficient), latent 146
heat flux ( EC , Dalton number), and sensible heat flux ( HC , Staton number) are estimated 147
from wind speed, air temperature, and SST using the Fairall et al. (2003) algorithm 148
6
(COARE3 version). LHTFL is positive if the ocean loses heat, while HQ is positive if the 149
ocean gains heat.150
151
The variables needed for the evaluation of (1) are obtained from satellite measurements. 152
Wind speed relative to the ocean surface current sa uu − is measured by scatterometers 153
onboard the European Research Satellites ERS-1 (1992-1996), ERS-2 (1996-2001), and 154
QuikSCAT (1999-2007) (e.g. Liu, 2002). The humidity ( aq ) is derived from the Special 155
Sensor Microwave Imager multi channel brightness temperatures using the Bentamy et al.156
(2003) method, while the specific surface humidity ( sq ) is estimated from daily averaged 157
SST. This version of LHTFL uses the new Reynolds et al. (2007) daily SST while the 158
previous version of LHTFL (Bentamy et al., 2003) is based on the Reynolds and Smith 159
OIv2 weekly SST.160
161
The air temperature is determined from remotely sensed data based on the Bowen ratio 162
( EH QQB /−= ). This method has been suggested by Konda et al. (1996) and validated 163
by comparisons to buoy data in the tropics and midlatitudes. For the gradient-based K-164
parameterization of fluxes and equal eddy diffusivity for heat and water vapor, the 165
Bowen ratio reads166
167
)()(
saE
saHp
a
ap
qqCTTC
LC
qT
LC
B−−
−=∂∂
−= , (2)168
169
7
where the right hand side term is the Bowen ration from (1). Using the Clausius-170
Clapeyron law to relate the saturated humidity ( satq ) and air temperature and neglecting 171
dependence of the relative humidity ( r ) on air temperature, 172
aTTsat Tq =∂∂ /)ln( >>aTTTr =∂∂ /)ln( , (2) takes the form:173
174
)()(
)( saH
saE
TaT
sat
asat
a
TTCqqC
Tq
Tqq
−−
=∂
∂
=
(3)175
176
and is solved for aT .177
178
The turbulent fluxes are calculated using the COARE3.0 algorithm from daily averaged 179
values binned onto a 1° global grid over satellite swaths. Due to differences in sampling 180
by different satellite radars and radiometers, the final flux estimate is further averaged 181
weekly and spatially interpolated on a regular 1° grid between 80° S and 80° N using the 182
kriging method described in Bentamy et al. (2003). The accuracy of the resulting weekly 183
fluxes is assessed by comparisons with in-situ measurements from moored buoys in the 184
tropical Atlantic and Pacific (PIRATA and TAO/TRITON), the northeastern Atlantic and 185
northwestern Mediterranean (UK Met Office and Météo-France), and the National Data 186
Buoy Center (NDBC) network off the U.S. coast in the Atlantic and Pacific Oceans1. 187
Quite high correlations (ranging from 0.8 to 0.92) are found between satellite and in-situ188
LHTFL, while biases and standard deviations are generally low. Standard deviations of 189
satellite and in-situ LHTFL vary from 18 Wm-2 and 25 Wm-2. The highest bias is found 190
1 See ‘New Release of Satellite Turbulent Fluxes 1992 – 2007’ at ftp://ftp.ifremer.fr/ifremer/cersat/products/gridded/flux-merged/documentation/flux.pdf
8
in comparisons with the NDBC buoys in the Gulf Stream region where the time mean 191
satellite LHTFL is 10 Wm-2 below in-situ values (or 7% of the NDBC regional LHTFL 192
mean). In the tropics satellite LHTFL overestimates in-situ LHTFL by 8 Wm-2. These 193
comparisons indicate significant improvements of the new LHTFL product over the 194
previous release described in Bentamy et al. (2003).195
196
Intraseasonal signal is evaluated in few steps. First, the annual cycle is calculated from 197
the weekly data as a sum of the first three harmonics (Mestas-Nuñez et al., 2006). Next, 198
the anomaly is calculated by subtracting the annual cycle from the original signal. 199
Finally, the intraseasonal signal is calculated as the difference between the anomaly and 200
its 13 week running mean. This procedure retains periods shorter than 3 months that are 201
referred to as intraseasonal in this study. The variability of intraseasonal fluxes is 202
characterized by the running standard deviation that mimics the upper envelope of the 203
intraseasonal signal. Running standard deviation of intraseasonal signal is calculated 204
using the same 13 week running window. Comparisons of the satellite intraseasonal 205
LHTFL with in-situ data from the TAO/TRITON moorings in the tropical Pacific, the 206
PIRATA moorings in the tropical Atlantic, and the RAMA moorings in the tropical 207
Indian Ocean are presented in the Appendix.208
209
The LHTFL from this study is compared with LHTFL provided by the National Center 210
for Climate Prediction/ National Center for Atmospheric Research (NCEP/NCAR) 211
reanalysis (Kalnay et al., 1996), the Woods Hole Oceanographic Institution objectively 212
analyzed air-sea fluxes of Yu et al. (2004), and with ship borne estimates collected by the 213
9
International Comprehensive Ocean-Atmosphere Data Set (ICOADS) of Worley et al.214
(2005). Mean sea level pressure for this study is provided by the NCEP/NCAR 215
reanalysis. In-situ measurements from the TAO/TRITON moorings in the tropical Pacific 216
Ocean (McPhaden et al., 1998), the PIRATA moorings in the tropical Atlantic (Bourles 217
et al., 2008), and the RAMA moorings in the tropical Indian Ocean (McPhaden et al., 218
2008) are also used for comparisons.219
220
For several years now, uniform, long-term data from observations made from numerous221
satellites relevant for inferring surface shortwave radiation (SWR) have been prepared 222
into homogeneous time series. The satellites that are being used for SWR retrieval 223
usually have between two to five channels in spectral intervals that are relevant both for 224
inferring SWR (visible) and for detecting clouds. Cloud data are provided by the 225
International Satellite Cloud Climatology Project (version D1) at a nominal resolution of 226
2.5◦ at 3hr time intervals (Rossow and Schiffer, 1999). The original version of the SWR 227
retrieval scheme is described in Pinker and Laszlo (1992) and has been used at 228
NASA/Langley for generating the GEWEX/SRB product2. Since, several modifications 229
have been introduced as related to aerosols (e.g. Liu and Pinker, 2008), data merging 230
(Zhang et al., 2007), and elevation correction (Ma and Pinker, 2008). 231
232
3. Results233
Mean LHTFL and seasonal variations234
First, presented are global patterns of the LHTFL and its annual and semiannual 235
harmonics. These components form the annual cycle that is used as a reference for 236
2 http://gewex-srb.larc.nasa.gov
10
evaluating anomalies and intraseasonal signal. Spatial patterns of magnitude and phase of 237
these harmonics are similar to Mestas-Nuñez et al. (2006) analysis that is based on the 238
three year long record (1996-1998) from the previous release of the LHTFL archive of 239
Bentamy et al. (2003). Comparison of the time mean LHTFL from this study with the 240
time mean LHTFL provided by alternative analyses (NCEP/NCAR Reanalysis, WHOI 241
air-sea fluxes, and ICOADS) indicates reasonable correspondence of spatial patterns242
(Figs. 1 a-d). The time mean latent heat flux is dominated by evaporation in the trade 243
wind regions and resembles the time average wind speed in the 30o S to 30o N belt (Fig.244
2a). SST impacts are evident across the subtropical fronts where temperature decreases 245
with latitude. Poleward decrease in SST is accompanied by decrease in aT . Hence, the246
humidity contrast, as qqq −=∆ , also decreases sharply poleward of 30o S and 30o N (Fig.247
2b). These meridional changes of q∆ explain weak LHTFL in the extratropical oceans in 248
spite of rather strong winds in the northern and especially southern hemisphere storm 249
track corridors. SST impacts are also noticeable in the equatorial eastern Pacific and 250
Atlantic where the mean LHTFL is weak due to the presence of cold tongues of SST 251
maintained by the equatorial upwelling. Local minimum of evaporation over the cold 252
tongue regions is explained by direct impact of cool SST on the air humidity as well as 253
by indirect impact of SST on the near surface atmospheric boundary layer that tends to 254
decelerate over cold water and vice-versa, accelerate over warm water (Wallace et al., 255
1989; Beal et al., 1997). 256
257
Regardless of the good correspondence of the geographical distribution of the time mean 258
LHTFL, the four analyses are somewhat different in magnitude. In current analysis 259
11
LHTFL (Fig. 1a) has higher values in the trade wind regions in comparison to the other 260
three analyses. This analysis is closer to in-situ ship observations from the ICOADS (Fig. 261
1d) and NCEP/NCAR reanalysis (Fig. 1b), but exceeds the WHOI estimates by 20 to 40 262
Wm-2 (Fig. 1c). Quantification of the LHTFL discrepancy may be achieved only through 263
a poleward extension of the tropical ocean buoy network and better sampling of 264
evaporation in the trade wind regions.265
266
Strong time mean latent heat loss (exceeding 80 Wm-2) is drawn from the warm Gulf 267
Stream waters off the east coast of the US. Similarly strong time mean LHTFL is 268
observed near Japan over the warm Kuroshio (Fig. 1 a-d). In both these regions the 269
LHTFL experiences the strongest annual variation peaking during the winter, when cold 270
dry continental air off-shore of North America and Japan crosses the Gulf Stream north 271
wall in the Atlantic or the Kuroshio SST front in the Pacific, respectively (Fig. 1e, 1f). 272
Semiannual LHTFL variations are prominent in the Arabian Sea and Bay of Bengal due 273
to annual reversal of winds forced by South Asian Monsoon (Figs. 1g, 1h). The monsoon 274
flow in the Arabian Sea low level westerly jet intensifies in boreal summer while 275
northeasterly winds spread over the region in boreal winter when the monsoon ceases. 276
Weaker semiannual variability is observed in the Caribbean low level jet where the 277
easterly winds also intensify twice a year in February and again in July (Munoz et al.,278
2008).279
280
Magnitude of intraseasonal variation281
12
To characterize the intraseasonal variations of LHTFL we consider the intraseasonal 282
variation of qw∆ (Fig.2e). This variable closely corresponds to the LHTFL as the Dalton283
number, EC , has weak dependence on wind speed (for winds ranging from 4 ms-1 to 14 284
ms-1, Large and Pond, 1982) though it depends on the stratification of the near surface 285
atmospheric layer. The spatial patterns of the intraseasonal variability of qw∆ bear only 286
partial correspondence to intraseasonal winds (Fig. 2c). In particular, the decrease in 287
variance of intraseasonal qw∆ towards the equator reflects relatively weak variability of 288
intraseasonal wind at low latitudes. In contrast to low latitudes, the intraseasonal 289
variability of LHTFL decreases at high latitudes despite stronger wind variability there. 290
This behavior is explained by low humidity over cold SSTs (and cold aT ) poleward of 291
40o - 50o in each hemisphere (Fig. 2b). As it is expected from the bulk formulation (1), 292
the regions of strong intraseasonal variability of LHTFL are spatially collocated with the 293
regions of strong intraseasonal variability of winds (Fig. 2c) and of air humidity (Fig. 2d). 294
Although linear decomposition of the intraseasonal variance of qw∆ suggests that wind 295
component ( 'qw∆ , Fig. 2f) accounts for a major portion of variability of the intraseasonal 296
LHTFL, neither components dominates globally. In particular, the air humidity 297
variability term ( 'awq , Fig. 2g) peaks along the major SST fronts and reflects an impact 298
of moisture transport across the ocean SST fronts by synoptic weather systems. SST itself 299
( 'swq , Fig. 2h) also impacts the intraseasonal LHTFL along the major western boundary 300
current fronts and in the Agulhas current area. Both, the mean LHTFL (Fig.1a) and its 301
variability (Fig.2e) weaken over cold SSTs where mean values of LHTFL are also low. 302
13
This is particularly evident in the cold sector of the Gulf Stream, in the Brazil-Malvina 303
confluence region, in the subpolar north Pacific, and in the Southern Ocean. 304
305Standard deviation of LHTFL in the intraseasonal band ( 1σ , Fig. 3a) exceeds the 306
standard deviation in the low frequency band ( 2σ , Fig. 3b) over much of the global ocean 307
except in the equatorial Pacific, where the low frequency variability (due to the 308
interannual ENSO) slightly exceeds the intraseasonal variability. Although intraseasonal 309
variation of LTHFL is stronger than low frequency variation of LHTFL, the relative310
impact of intraseasonal LHTFL on the mixed layer temperature is mitigated by the 311
difference in characteristic periods. Anomalous LHTFL defines the rate of change of 312
anomalous SST, tSSTH ∂∂ /)(δ ~ )(LHTFLδ , where H is the mixed layer depth. Hence, 313
the ratio of SST responses to intraseasonal and low frequency variations of LHTFL is 314
defined by the standard deviation of LHTFL in these two bands and is scaled by the 315
ration of characteristic periods that are 1τ =1 month and 2τ =1 year, respectively. Relative 316
impact of intraseasonal and low frequency variations of LTHFL on SST is roughly 317
evaluated in Fig. 3c as )/()( 2211 τστσ . SST response to intraseasonal LHTFL variation 318
does not exceed ~30% of the SST response to low frequency LHTFL variation because of 319
the difference in the characteristic periods. This ratio has local minimum of ~ 10% in the 320
equatorial regions and increases poleward of 20o S - 20o N band reflecting stronger 321
transient (synoptic) variability at mid-latitudes.322
323
Intraseasonal LHTFL in midlatitudes324
14
The strongest variability of the intraseasonal LHTFL occurs in midlatitudes where the 325
regional maxima are linked to areas of major SST fronts (Fig. 3a). In particular, in the 326
Atlantic sector the highest intraseasonal variance is observed along the Gulf Stream front. 327
Similarly high intraseasonal variability is observed in the Agulhas Current and in the 328
Brazil-Malvina confluence region. This suggests important roles the stratified 329
atmospheric boundary layer plays in amplifying intraseasonal air-sea interactions. The 330
intraseasonal LHTFL variance changes seasonally and peaks in winter (not shown) 331
suggesting an association with midlatitude storms which also intensify in cold season. 332
We next identify weather systems that are responsible for strong intraseasonal variability 333
of LHTFL in these regional maxima areas by projecting the intraseasonal LHTFL time 334
series spatially averaged over particular index area on atmospheric parameters elsewhere 335
(Fig. 4a). This regression analysis reveals correspondence between strengthening of 336
intraseasonal LHTFL in the Gulf Stream region and midlatitude storm systems. Increase 337
in LHTFL drawn from the region is associated with the area of mean sea level pressure 338
low and corresponding cyclonic anomalous winds centered east of the region. The air 339
pressure pattern is similar to that deducted by Foltz and McPhaden (2004) in their 340
analysis of the intraseasonal variability of the Atlantic trade winds. In fact, the anomalous 341
wind in Fig. 4a decelerates the northern flank of the northeasterly trades (where 342
anomalous LHTFL is somewhat weaker) and significantly accelerates off-shore winds 343
over the Gulf Stream (Fig. 4b). Maximum increase of wind speed is observed over warm 344
sector of the Gulf Stream where winds further accelerate due to the atmospheric 345
boundary layer adjustment. In addition to intensification of mean winds, the anomalous 346
northwesterly wind outbreaks cold and dry continental air toward the sea. Spreading of 347
15
dry continental air lowers air humidity thus increasing the air-sea humidity contrast (Fig. 348
4c). This, in turn, compliments the LHTFL increase due to stronger winds. The ocean 349
responds to continental air outbreak by cooling SST north of the Gulf Stream northern 350
wall that is seen in decreasing values of sq (Fig. 4c). Intraseasonal winds have a weak 351
impact on SST south of the Gulf Stream temperature front where the ocean mixed layer is 352
deep and its thermal inertia is relatively strong.353
354
Similar response is seen in the Agulhas current south of the Cape of Good Hope (Fig. 5). 355
Like in the Gulf Stream area, increase of LHTFL over the warm Agulhas Current is 356
linked to a passing storm. When the storm center locates to the east of the index area the 357
anomalous southerly winds bring cold and dry sub-Antarctic air northward. This 358
amplifies the latent heat loss due to increasing wind speed and increasing air-sea surface 359
humidity contrast. Although storm systems are similarly strong as they propagate around 360
the globe in the South Atlantic and the Southern Oceans, the intraseasonal LHTFL is 361
stronger in the Agulhas region and in the Brazil-Malvina confluence (Fig. 3a) in 362
comparison to values observed at similar latitude in the ocean interior. Both these areas 363
host sharp SST fronts that promotes higher q∆ and stronger LHTFL. It is interesting to 364
note that the regression analysis in Fig. 5 reveals a sequence of propagating storms over 365
open spaces of the South Atlantic and South Oceans. The mean sea level pressure troughs366
in the regression pattern are separated by approximately 90o in longitude suggesting the367
zonal wavenumber of 4.368
369
Intraseasonal surface fluxes and SST 370
16
Variability of LHTFL and SST are related. LHTFL affects SST by affecting the net ocean 371
surface heat balance. But, SST also affects LHTFL directly through sq and indirectly by 372
affecting near-surface winds that accelerate over warmer SSTs. We next characterize 373
interplay between these intraseasonal variations. As expected, the LHTFL feedback to 374
underlying anomalous SST is generally negative (Fig. 6a), i.e. LHTFL increases in 375
response to increased SST. This suggests a damping of the underlying SST anomalies,376
although there are considerable geographical variations (Park et al., 2005). The feedback 377
exceeds 20Wm-2 in the regions around 20oS and 20oN, but decreases at high latitudes and 378
in the eastern tropical Pacific and Atlantic where the time average LHTFL is also weak. 379
In contrast to the LHTFL response to underlying SST that is positive, the SST response to 380
intraseasonal variation of LHTFL is negative over much of the ocean (Fig. 6b). But, in 381
several regions SST warms up in response to LHTFL increase. In particular, this behavior 382
occurs in the cold tongue regions of the eastern tropical Pacific and Atlantic Oceans. The 383
relationship between intraseasonal LHTFL and SST depends on the relative role the 384
LHTFL plays in the mixed layer heat balance. If this balance is local and governed by the 385
surface flux, the SST cools down in response to increasing latent heat loss (negative 386
correlation when LHTFL leads). This negative relationship dominates away from the cold 387
tongue regions and strong currents. In contrast, in the cold tongue regions the mixed layer 388
temperature balance is governed primarily by the vertical (upwelling) or horizontal 389
(Tropical Instability Waves, e.g. Grodsky et al., 2005) heat transports. Here the positive 390
correlation between LHTFL and SST is explained by the stratified atmospheric boundary 391
layer adjustment and associated wind acceleration over warm SSTs. Therefore, in the 392
17
cold tongue regions the LHTFL increases in response to increasing wind and SST rather 393
than SST responses to change in LHTFL.394
395
Over the regions where the surface heat flux dominates the mixed layer heat budget, the 396
variations of LHTFL force variations of the mixed layer temperature and, thus, should be 397
correlated with the SST rate of change, tT ∂∂ / , as seen in Fig. 7a. The time correlation of 398
intraseasonal LHTFL and tT ∂∂ / is statistically significant over much of the ocean3. It 399
decreases at high latitudes where the upper ocean stratification is weak, the mixed layer is 400
deep, and SST response is weak. The time correlation is also weak in the tropical Pacific 401
and Atlantic Oceans in the regions where vertical and horizontal heat transports dominate 402
the mixed layer heat budget. Similar but weaker correlation is found for the short wave 403
radiation (Fig. 7b). If the two components (LHTFL and SWR) of the surface flux are 404
combined, the correlation increases (Fig. 7c) suggesting reasonable correspondence of 405
intraseasonal flux variations with intraseasonal SST.406
407
Intraseasonal LHTFL and SWR408
Both, the intraseasonal LHTFL and SWR agree reasonably with independent 409
measurements of the rate of change of intraseasonal SST. We next explore the global 410
correspondence between intraseasonal variations of the two surface flux components. 411
They are weakly correlated over much of the global ocean with an exception of the 412
tropical Indian Ocean and the western tropical Pacific where the intraseasonal LHFL and 413
SWR are negatively correlated (Fig. 8). Lagged correlation indicates that LHTFL 414
increases in phase with decrease in SWR, suggesting stronger latent heat loss just below 415
3 The 99% confidence level of zero correlation is 0.1.
18
convective systems. This phase relationship is similar for satellite flux data and in-situ416
TAO/TRITON mooring data (Fig. 8, inlay). It is consistent with Zhang and McPhaden 417
(2000) analysis of the TAO/TRITON surface fluxes who also have found near in-phase 418
relationship among maxima in latent heat flux and minima in solar radiation during 419
passage of the MJO events. From one side, the out-of-phase variations of intraseasonal 420
LHTFL and SWR support a hypothesis that the evaporation affects the humidity and 421
therefore the cloudiness and thus solar radiation at the sea surface. But theoretical 422
considerations (see Zhang and McPhaden, 2000 for a summary of existing approaches)423
suggest a lagged relationship between intraseasonal LHTFL and SWR forced by MJO. In 424
particular, in the Neelin et al. (1987) model the maximum LHTFL is shifted to the east of 425
the convective center, if mean wind is easterly. Explanation of the phase relationship 426
between LHTFL and SWR variations on the intraseasonal timescales is not clear, 427
although a qualitative description based on the observed variations of air humidity is 428
briefly discussed below.429
430
Coherent variations of the intraseasonal LHTFL and SWR are apparent in the time-431
longitude diagrams in Figs. 9e, 9f. These accorded intraseasonal variations propagate 432
eastward between 600 E and the dateline at an average speed of 7.5 ms-1 typical of the 433
MJO. East of the dateline the correlation between LHTFL and SWR is weak (Fig. 8). 434
This zonal change of correlation is explained by the lack of cloudiness east of the dateline 435
that is the only major source of SWR variability. Intraseasonal LHTFL variations are 436
mostly driven by the intraseasonal variations of zonal wind. In fact, this is illustrated in 437
Fig. 9a that shows substantial correlation of LHTFL and zonal wind along the whole 438
19
equatorial belt. The sign of correlation switches depending on the zonal wind direction. 439
In the equatorial Indian Ocean and western equatorial Pacific the time mean zonal wind is 440
westerly. In this region a superposition of the mean eastward wind with an eastward 441
anomalous wind enhances the wind speed and LHTFL resulting in a positive correlation 442
of the zonal wind velocity, U , and LHTFL as seen in Fig. 9a. In contrast, a negative zero 443
lag correlation is observed east of the dateline where the mean zonal wind is easterly.444
445
In distinction from the correlation of U and LHTFL that is observed along the whole 446
equatorial belt, coherent variations of intraseasonal zonal wind and SWR occur only west 447
of the dateline (Fig. 9b) with a gap in correlation over the maritime subcontinent. In the 448
Indian Ocean where the time mean westerly wind is stronger the anomalous eastward 449
wind lags by ~1 week the anomalously low SWR (see negative correlation at positive 450
lags in Fig. 9b). In the western Pacific where the mean zonal wind is weak the anomalous 451
zonal wind is almost out-of-phase with anomalous insolation. In spite of zonally varying 452
phase relationship of intraseasonal U and SWR, the intraseasonal SWR and LHTFL are 453
firmly out-of-phase west of the dateline (Fig. 9c). Possible explanation for this out-of-454
phase behavior may include an impact of air humidity that varies in phase with 455
intraseasonal SWR (Fig. 9d). Specific air humidity decreases below convective systems 456
in response to cooling of the near-surface atmosphere while the sea surface saturated 457
humidity doesn’t change much because of the thermal inertia of the ocean mixed layer. 458
Difference in responses of aq and sq leads to an increase in the vertical gradient of 459
specific air humidity below convective cloud clusters that, in turn, enhances evaporation 460
and LHTFL.461
20
462
Intraseasonal and longer period variability of LHTFL463
The interannual evolution of the ocean surface fluxes has been extensively studied. But, it 464
appears that amplitude of intraseasonal fluxes is not stationary and experiences465
significant modulation by longer period variability. Noting that our dataset is only 16 466
years long, the consideration is limited to the tropical Pacific Ocean that hosts the ENSO 467
and, thus, displays significant interannual variability that is resolved by relatively short 468
records. Interannual SWR anomaly is modulated by ENSO through zonal displacements469
of convection. These interannual displacements of convection between the western 470
tropical Pacific and the central tropical Pacific produce SWR anomalies that are well 471
detected by satellite techniques (Rodriguez-Puebla et al., 2008). Because clouds are the 472
only physical mechanism driving the intraseasonal SWR, the amplitude of intraseasonal 473
SWR also shifts zonally following anomalously low SWR. In the central equatorial 474
Pacific the magnitude of intraseasonal SWR increases in-phase with warming of the475
Nino3 SST (Fig. 10a, inlay). Here, the standard deviation of intraseasonal SWR increases 476
by up to 5 Wm-2 in response to a 1 oC rise of SST in the Nino3 region (Fig. 10a). As such, 477
interannual variation of the amplitude of intraseasonal SWR reaches 15 Wm-2 during a 478
mature phase of El Niño when anomalous Nino3 SST warms up by 3 oC. This interannual 479
modulation of amplitude of the intraseasonal SWR is comparable to the characteristic 480
amplitude of SWR variation by MJO (Shinoda et al., 1998).481
482
In distinction to the amplitude of intraseasonal SWR that varies in-phase with El Niño, 483
the magnitude of intraseasonal LHTFL doesn’t have similarly significant in-phase 484
21
variation. Impact of El Niño on the intraseasonal LHTFL differs from the impact on the 485
total anomalous LHTFL that enhances in the eastern tropical Pacific, around the 486
Maritime Continent, and the equatorial Indian Ocean (Mestas-Nunez et al., 2006). In 487
contrast, the magnitude of intraseasonal LHTFL amplifies over the western tropical 488
Pacific approximately 8 months in advance of the mature phase of El Niño (Fig. 10b and 489
inlay). This amplification reflects impacts of the westerly wind bursts that precede the 490
onset of El-Nino, which were evident in advance of the 2002/03 El Niño and particularly 491
noticeable in advance of the 1997-98 event (McPhaden, 2004).492
493
4. Conclusions494
Although the major portion of the intraseasonal variability of LHTFL is accounted for by 495
winds, neither components (wind, air humidity, or sea surface humidity) dominates the 496
variability globally. In particular, contributions of aq and sq are significant along major 497
SST fronts due to moisture transport across the ocean SST fronts by synoptic weather 498
systems. Both the mean LHTFL and its intraseasonal variability weaken over cold SSTs 499
due to low air-sea humidity contrast. In contrast, the strongest intraseasonal LHTFL is 500
observed over the warm sectors of SST fronts.501
502
The strongest variability of the intraseasonal LHTFL (in excess of 50 Wm-2) occurs at 503
middle latitudes where the regional maxima are linked to areas of major SST fronts. In 504
particular, in the Atlantic sector the highest intraseasonal variance is observed along the 505
Gulf Stream. Similarly high variability is observed in the Agulhas Current and in the 506
Brazil-Malvina confluence. Coincidence of the regional maxima of intraseasonal LHTFL 507
22
with SST fronts suggests important roles the stratified atmospheric boundary layer plays 508
in amplifying intraseasonal air-sea interactions. Temporal variation of the intraseasonal 509
LHTFL in these regional maxima is linked to passing midlatitude storms. The 510
intraseasonal variability of LHTFL forced by these passing storms is locally amplified by 511
unstable atmospheric stratification over warm SSTs.512
513
Although weaker in amplitude but still significant intraseasonal variability of LHTFL 514
(standard deviation of 20 to 30 Wm-2) is observed in the tropical Indian and Pacific 515
Oceans. This variability is linked to the eastward propagating Madden-Julian 516
Oscillations. In this tropical region the intraseasonal LHTFL and incoming solar radiation 517
vary out-of-phase, i.e. evaporation enhances just below the convective clusters while the 518
SWR low leads by approximately a week the eastward wind anomaly. The out-of-phase 519
relationship is observed west of the dateline, while east of the dateline both, intraseasonal 520
LHTFL and SWR, are weak and their relationship is not significant. Intraseasonal 521
variation of aq that varies in phase with intraseasonal SWR contributes to this out-of-522
phase relationship. Specific air humidity decreases below convective systems following 523
cooling of the near-surface atmosphere while sq doesn’t change much because of the 524
ocean thermal inertia. Difference in responses of aq and sq increases the vertical 525
gradient of air humidity below convective cloud clusters and thus enhances evaporation 526
and LHTFL.527
528
Amplitude of intraseasonal LHTFL and SWR displays significant interannual variations 529
in the tropical Pacific Ocean. Amplitude of intraseasonal SWR increases in the central 530
23
equatorial Pacific by 15 Wm-2 during mature phase of El Niño following the eastward 531
shift of convection. In distinction to the amplitude of intraseasonal SWR that varies in-532
phase with El Niño, the amplitude of intraseasonal LHTFL doesn’t have similarly 533
significant in-phase variation. In contrast, the intraseasonal LHTFL amplifies over the 534
western tropical Pacific approximately 8 months in advance of the mature phase of El535
Niño. This amplification reflects impacts of the westerly wind bursts that normally 536
precede the onset of El-Nino.537
538
Over much of the global ocean anomalous LHTFL provides negative feedback on the 539
underlying intraseasonal SST anomaly, although there are considerable geographical 540
variations. The feedback exceeds 20 Wm-2/oC in the regions around 20o S and 20o N, but 541
decreases at high latitudes and in the eastern tropical Pacific and Atlantic where the time 542
average LHTFL is weak.543
544
Appendix545
Comparisons of in-situ LHTFL with satellite-derived LHTFL in the intraseasonal band is 546
shown in Fig. 11. This comparison is based on in-situ buoy measurements in the tropics 547
including 68 TAO/TRITON buoys in the Pacific, 21 PIRATA buoys in the Atlantic, and 548
10 RAMA buoys in the Indian Ocean. During the 1992-2007 the data set has 30592549
concurrent buoy-satellite weekly measurements in the Pacific, 3044 weeks of data in the 550
Atlantic, and 318 weeks of concurrent buoy and satellite data in the Indian Ocean. The551
aggregate time series of buoy and satellite LHTFL have statistically significant 552
correlation around 0.6. The 99% confidence level of zero correlation is %99corr <0.1 for 553
24
the Pacific and Atlantic while it is slightly higher %99corr =0.14 for the Indian Ocean due 554
to shorter time series. Time series of intraseasonal LHTFL at each buoy location also 555
indicate significant correlation (Figs. 11a, 11c). Time correlation (TCORR) exceeds 0.6 556
over much of the tropical Pacific where average length of the LHTFL time series at557
particular buoy is around 450 weeks ( %99corr =0.12). TCORR increases towards the west 558
following the westward increase of the mean LHTFL in the tropical Pacific (Fig. 1a). In 559
contrast, somewhat weaker TCORR is observed along 50 N where LHTFL is weaker due 560
to weaker winds and higher specific humidity in the ITCZ. The impact of the ITCZ is 561
better seen in the Atlantic where TCORR decreases below 0.5 in the ITCZ area (Fig. 562
11c). These comparisons suggest that the LHTFL retrieval should be rectified in the 563
ITCZ area. The air relative humidity has regional maximum in the ITCZ area. Therefore, 564
the meridional displacement of the ITCZ could produce variations of the relative 565
humidity strong enough that need to be accounted for in the Kondo et al. (1996) Bowen 566
ration approach.567
568
Satellite intraseasonal LHTFL compares well with in-situ LHTFL to within the scatter of 569
the data (Figs. 11b, 11d, 11f). However, the magnitude of satellite intraseasonal LHTFL 570
is weaker than in-situ data. This bias is more evident in the Pacific where the 571
intraseasonal variations of satellite LHTFL are 15% to 20% weaker than those from 572
buoys, while this bias is less than 10% in the Atlantic and is not evident in the Indian 573
Ocean. We attribute this bias to the spatial and temporal smoothing of satellite data that 574
inevitably results in loosing of a portion of variance observed at fixed location and high 575
temporal resolution.576
25
577
Acknowledgements. SAG and JAC acknowledge support from the NASA Ocean Wind 578Vector Science Team (OWVST). AB is grateful for the technical and financial supports 579provided by CERSAT/IFREMER and MERCATOR. RTP gratefully acknowledges 580support by NASA grant NNG04GD65G from the Radiation Sciences Program and NSF 581grant ATM0631685 to the University of Maryland. The ISCCP data were obtained from 582the NASA Langley Research Center EOSDIS Distributed Active Archive Center.583
584References:585Araligidad, N. M., and E. D. Maloney, 2008: Wind-driven latent heat flux and the 586
intraseasonal oscillation. Geophys. Res. Lett., 35, L04815, 587doi:10.1029/2007GL032746.588
589Beal, R. C., V. N. Kudryavtsev, D. R. Thompson, S. A. Grodsky, D. G. Tilley, V.A.590
Dulov, and H. C. Graber, 1997: The influence of the marine atmospheric boundary 591layer on ERS 1 synthetic aperture radar imagery of the Gulf Stream. J. Geophys. 592Res., 102, 5799-5814.593
594Bentamy, A., K.B. Katsaros, A.M. Mestas-Nuñez, W.M. Drennan, E.B. Forde, and H. 595
Roquet, 2003: Satellite Estimates of Wind Speed and Latent Heat Flux over the 596Global Oceans. J. Climate, 16, 637–656.597
598Bentamy, A., L-H. Ayina, W. Drennan, K. Katsaros, A. M. Mestas-Nuñez, R. T. Pinker, 599
2008: 15 Years of Ocean Surface Momentum and heat Fluxes from Remotely 600Sensed Observations, FLUXNEWS, 5, 14-16. Available online at 601http://sail.msk.ru/newsletter/fluxnews_5_final.pdf602
603Bond, N.A., and M.F. Cronin, 2008: Regional Weather Patterns during Anomalous Air–604
Sea Fluxes at the Kuroshio Extension Observatory (KEO). J. Climate, 21, 1680–6051697.606
607Bourlès, B., R. Lumpkin, M.J. McPhaden, F. Hernandez, P. Nobre, E. Campos, L. Yu, S. 608
Planton, A. Busalacchi, A.D. Moura, J. Servain, and J. Trotte, 2008: The Pirata 609Program: History, Accomplishments, and Future Directions. Bull. Amer. Meteor. 610Soc., 89, 1111–1125. 611
612Chou, S.H., E. Nelkin, J. Ardizzone, R.M. Atlas, and C.L. Shie, 2003: Surface Turbulent 613
Heat and Momentum Fluxes over Global Oceans Based on the Goddard Satellite 614Retrievals, Version 2 (GSSTF2). J. Climate, 16, 3256–3273.615
616da Silva, A., C. C. Young, and S. Levitus, 1994: Algorithms and Procedures. Vol. 1, 617
Atlas of Surface Marine Data 1994, NOAA Atlas NESDIS 6, 83 pp.618619
26
Fairall, C.W., E.F. Bradley, J.E. Hare, A.A. Grachev, and J.B. Edson, 2003: Bulk 620Parameterization of Air–Sea Fluxes: Updates and Verification for the COARE 621Algorithm. J. Climate, 16, 571–591.622
623Foltz, G. R., and M. J. McPhaden, 2004: The 30–70 day oscillations in the tropical 624
Atlantic. Geophys. Res. Lett., 31, L15205, doi:10.1029/ 2004GL020023.625626
Grodsky, S. A., J. A. Carton, C. Provost, J. Servain, J. A. Lorenzzetti, and M. J. 627McPhaden, 2005: Tropical instability waves at 0N, 23W in the Atlantic: A case 628study using Pilot Research Moored Array in the Tropical Atlantic (PIRATA) 629mooring data. J. Geophys. Res., 110, C08010, doi:10.1029/2005JC002941.630
631Han W., D. Yuan, W. T. Liu, D. J. Halkides, 2007: Intraseasonal variability of Indian 632
Ocean sea surface temperature during boreal winter: Madden-Julian Oscillation 633versus submonthly forcing and processes. J. Geophys. Res., 112, C04001, 634doi:10.1029/2006JC003791.635
636Jones, C., and B. C. Weare, 1996: The role of low-level moisture convergence and ocean 637
latent heat fluxes in the Madden-Julian oscillation: An observational analysis using 638ISCCP data and ECMWF analyses. J. Climate, 11, 1057– 1072.639
640Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project, Bull. 641
Amer. Meteorol. Soc., 77, 437-471.642643
Konda, M., N. Imasato, and A. Shibata, 1996: A new method to determine near-sea 644surface air temperature by using satellite data, J. Geophys. Res., 101(C6), 14349-64514360.646
647Krishnamurti, T., D. Oosterhof, and A. Mehta, 1988: Air–Sea Interaction on the Time 648
Scale of 30 to 50 Days. J. Atmos. Sci., 45, 1304–1322. 649650
Large, W., and S. Pond, 1982: Sensible and Latent Heat Flux Measurements over the 651Ocean. J. Phys. Oceanogr., 12, 464–482.652
653Liu, H. and R. T. Pinker, 2008: Radiative fluxes from satellites: Focus on aerosols. J. 654
Geophys. Res., 113, D08208, doi:10.1029/2007JD008736.655656
Liu, W. T., K. B. Katsaros, and J. A. Businger, 1979: Bulk parameterization of air-sea657exchanges of heat and water vapor including the molecular constraints at the sea 658surface. J. Atmos. Sci., 36, 1722-1735.659
660Liu, W.T., 2002: Progress in scatterometer application, J. Oceanogr., 58(1), 121-136. 661
662Madden, R.A., and P.R. Julian, 1972: Description of Global-Scale Circulation Cells in 663
the Tropics with a 40–50 Day Period. J. Atmos. Sci., 29, 1109–1123.664665
27
Maloney, E. D., and A. H. Sobel, 2004: Surface fluxes and ocean coupling in the tropical 666intraseasonal oscillation, J. Climate., 17, 4368–4386.667
668McPhaden, M.J. , and Coauthors, 1998: The Tropical Ocean-Global Atmosphere observing 669
system: A decade of progress, J. Geophys. Res., 103(C7), 14,169-14,240. 670671
McPhaden, M. J., G. Meyers, K. Ando,Y. Masumoto, V. S. N. Murty, M. Ravichandran, 672F. Syamsudin, J. Vialard, L. Yu, W. Yu, 2008: RAMA: The Research Moored 673Array for African-Asian-Australian Monsoon Analysis and Prediction, subm. Bul. 674Amer. Meteor. Soc., http://www.pmel.noaa.gov/tao/oceansites/RAMA_BAMS.pdf675
676McPhaden, M.J., 2004: Evolution of the 2002/03 El Niño. Bull. Amer. Meteor. Soc., 85, 677
677–695.678679
Mestas-Nuñez, A.M., A. Bentamy, and K.B. Katsaros, 2006: Seasonal and El Niño 680Variability in Weekly Satellite Evaporation over the Global Ocean during 1996–98.681J. Climate, 19, 2025–2035.682
683Muñoz, E., A.J. Busalacchi, S. Nigam, and A. Ruiz-Barradas, 2008: Winter and Summer 684
Structure of the Caribbean Low-Level Jet. J. Climate, 21, 1260–1276.685686
Neelin, J.D., I.M. Held, and K.H. Cook, 1987: Evaporation-Wind Feedback and Low-687Frequency Variability in the Tropical Atmosphere. J. Atmos. Sci., 44, 2341–2348.688
689Park, S., C. Deser, and M.A. Alexander, 2005: Estimation of the Surface Heat Flux 690
Response to Sea Surface Temperature Anomalies over the Global Oceans. J. 691Climate, 18, 4582–4599.692
693Pinker, R. T., and I. Laszlo, 1992. Modeling surface solar irradiance for satellite 694
applications on a global scale. J. Appl. Meteor. Clim., 31 (2), 194-211.695696
Qiu, B., S. Chen, and P. Hacker, 2004: Synoptic-Scale Air–Sea Flux Forcing in the 697Western North Pacific: Observations and Their Impact on SST and the Mixed 698Layer. J. Phys. Oceanogr., 34, 2148–2159.699
700Reynolds, R. W., T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey and M. G. Schlax, 701
2007: Daily High-resolution Blended Analyses for sea surface temperature. J. 702Climate, 20, 5473-5496.703
704Rodriguez-Puebla, C., R. T. Pinker, and S. Nigam, 2008: Relationship between 705
downwelling surface shortwave radiative fluxes and sea surface temperature over 706the tropical Pacific: AMIP II models versus satellite estimates, Ann. Geophys., 26, 707785–794.708
709Rossow, W. B., and R. A. Schiffer, 1999: Advances in understanding clouds from 710
ISCCP. Bull. Amer. Meteor. Soc., 80, 2261-2287.711712
28
Shinoda, T., and H.H. Hendon, 1998: Mixed Layer Modeling of Intraseasonal Variability 713in the Tropical Western Pacific and Indian Oceans. J. Climate, 11, 2668–2685.714
715Wallace, J.M., T.P. Mitchell, and C. Deser, 1989: The influence of sea surface716
temperature on surface wind in the eastern equatorial Pacific: seasonal and 717interannual variability. J. Climate, 2, 1492-1499.718
719Worley, S.J. S. D. Woodruff, R. W. Reynolds, S. J. Lubker, and N. Lott, 2005: ICOADS 720
release 2.1 data and products, Int. J. Climate, 25, 823-842, doi:10.1002/joc.1166721722
Yu, L., R.A. Weller, and B. Sun, 2004: Mean and Variability of the WHOI Daily Latent 723and Sensible Heat Fluxes at In Situ Flux Measurement Sites in the Atlantic Ocean.724J. Climate, 17, 2096–2118.725
726Yu, L., X. Jin, and R.A. Weller, 2006: Role of Net Surface Heat Flux in Seasonal 727
Variations of Sea Surface Temperature in the Tropical Atlantic Ocean. J. Climate, 72819, 6153–6169. 729
730Zhang, B. L., R. T. Pinker, and P. W. Stackhouse, 2007: An empirical orthogonal 731
function iteration approach for obtaining homogeneous radiative fluxes from 732satellite observations. J. Appl. Meteor. Clim., 46, 435-444.733
734Zhang, C., and M.J. McPhaden, 2000: Intraseasonal Surface Cooling in the Equatorial 735
Western Pacific. J. Climate, 13, 2261–2276.736737
29
737Figure 1. 1992-2007 mean LHTFL (Wm-2) from (a) this study, (b) NCEP/NCAR 738reanalysis, (c) WHOI objectively analyzed air-sea fluxes, and (d) ICOADS. The means 739are based on whatever part of this time interval is available. Annual harmonics (e) 740magnitude and (f) phase. Semiannual harmonics (g) magnitude and (h) phase. Zero phase 741corresponds to January.742
743
30
743Figure 2. Time mean (a) wind speed, w , (b) sea-air specific humidity difference, 744
as qqq −=∆ . Standard deviation of intraseasonal (c) wind speed, 'w , (d) humidity 745difference, 'q∆ . (e) Standard deviation of intraseasonal )'( qw∆ and contribution to it 746from intraseasonal variation of (f) wind speed, (e) specific humidity, and (h) saturated 747near surface humidity.748
31
749750
Figure 3. Standard deviation of latent heat flux STD(LHTFL) in (a) intraseasonal band 751(periods shorter than 3 months, HF) and (b) low frequency band (periods longer than 3 752months, LF). (c) Relative impact of HF and LF variations of LHTFL on SST. Only areas 753with at least 5 years of data are shown.754
755
32
755756
Figure 4. Time regression of intraseasonal latent heat flux index averaged over the Gulf 757Stream area on (a) mean sea level pressure and winds, (b) latent heat flux elsewhere and 758wind speed, (c) saturated near surface humidity and humidity. The index area is defined 759as the area where STD of intraseasonal LHTFL exceeds 40 Wm-2 (see Fig. 4a) and is 760dotted in panel (a). The index is defined as the index area average intraseasonal LHTFL 761normalized by its standard deviation (25 Wm-2). Only wind arrows exceeding 0.4 ms-1 are 762shown.763
764
33
764765
Figure 5. Time regression of intraseasonal latent heat flux index averaged over the 766Agulhas Current area on intraseasonal mean sea level pressure (contours) and winds 767(arrows). The index is defined as an area average intraseasonal LHTFL normalized by its 768standard deviation (37 Wm-2). The index area is shown by the shaded rectangle. Only 769wind arrows exceeding 0.4 ms-1 are shown.770
771
34
771
772Figure 6. Lagged regression of intraseasonal latent heat flux (LHTFL) and SST. (a) SST 773leads LHTFL by 1 week, (b) LHTFL leads SST by 1 week. Areas where time correlation 774exceeds the 99% confidence level of zero correlation are dotted. Inlay in panel (b) shows 775lagged correlation spatially averaged over the equatorial east Pacific (black) and the 776midlatitude Pacific (red) areas shown in panel (b). Negative lags are LHTFL lead in 777weeks.778
35
779780
Figure 7. Time correlation of the rate of change of intraseasonal SST, tT ∂∂ / , with (a) 781intraseasonal LHTFL, (b) intraseasonal short wave radiation, SWR, (taken with the minus 782sign), and (c) sum of the two, LHTFL-SWR. Correlation exceeding 0.1 is significant at 783the 99% confidence level.784
36
785
786Figure 8. Time correlation of intraseasonal LHTFL and SWR. Inlay shows lagged 787correlation of LHTFL and SWR averaged over the rectangle area (solid with symbols) 788and from the TAO/TRITON mooring at 0N, 165E (dashed). The mooring location is 789shown by the cross. Lags are in weeks. Positive lags imply that SWR leads LHTFL.790
37
791792
Figure 9. Lagged correlation of intraseasonal (a) zonal wind and LHTFL, (b) zonal wind 793and SWR, (c) SWR and LHTFL, and (c) specific air humidity and SWR. 794Positive/negative correlations are shown in solid/dashed. Correlations below 0.2 in 795magnitude are not shown. Positive lags imply that the second variable leads the first. 796Time-longitude diagrams of intraseasonal (e) LHTFL and (f) SWR averaged 5S to 5N in 797Wm-2. Shading in panel (a) is the time mean zonal wind (ms-1).798
799
38
799800
Figure 10. Time regression of anomalous Nino3 SST (210E-270E, 5S-5N) with running 801standard deviation, σ , of intraseasonal (a) SWR and (b) LHTFL. Correlation for SWR is 802instantaneous, while LHTFL is correlated with Nino3 SST that lags it by 35 weeks. 803Inlays show lagged correlation of Nino3 SST with the intraseasonal variance of flux804spatially averaged over the rectangle shown in each panel. Positive lags imply flux 805leading Nino3 SST. Areas where the time regression is significant at the 99% level are 806cross-hatched.807
808
39
808809
Figure 11. Comparison of intraseasonal buoy and satellite-derived latent heat flux. Spatial 810maps of time correlation (left) and scatter diagrams (right) for (a,b) tropical Pacific, (c,d) 811tropical Atlantic, and (e,f) tropical Indian Ocean. Gray shading in (b,d) shows the 812standard deviation of satellite intraseasonal LHTFL in 2 Wm-2 intervals of in-situ data. 813TCORR is the time correlation evaluated from the aggregate satellite/buoy comparisons814for each basin, NUM is the length of the aggregate record (in weeks). Buoy locations are 815shown by closed circles in (a,c,e).816