U.S. Dept. of Commerce / NOAA / OAR / PMEL / Publications


On the Variability of Winds, Sea Surface Temperature, and Surface Layer Heat Content in the Western Equatorial Pacific

Michael J. McPhaden and Stanley P. Hayes

NOAA/Pacific Marine Environmental Laboratory, Seattle, Washington

Journal of Geophysical Research, 96, supplement, 3331-3342 (1991)
This paper is not subject to U.S. copyright. Published in 1991 by the American Geophysical Union.

In this study we examine the surface layer heat balance using wind, current, and temperature data from equatorial moorings along 165E. The analysis focuses primarily on daily to monthly time scale variations during the 1986-1987 El Nio/Southern Oscillation event. The period is one of high mean sea surface temperatures (29C) and frequent outbreaks of westerly winds. We infer that evaporative cooling related to wind speed variations accounts for a significant fraction of the observed sea surface temperature (SST) and upper ocean heat content variability. This evaporative heat flux converges nonlinearly in the surface layer, giving rise to larger temperature variations in the upper 10 m than below. Other processes examined (wind-forced vertical advection and entrainment, lateral advection) were negligible or of secondary importance relative to evaporative cooling. A large fraction of the SST and surface layer heat content variance could not be directly related to wind fluctuations; this unexplained variance is probably related to shortwave radiative fluxes at the air-sea interface.

1. Introduction

The western equatorial Pacific is characterized by mean sea surface temperatures that are among the warmest in the world ocean. This region is also characterized by vigorous air-sea interaction, because turbulent heat exchange between the ocean and the atmosphere is a highly nonlinear function of sea surface temperature (SST). In extreme cases, air-sea interactions in the western Pacific may lead to El Nio/Southern Oscillation (ENSO) events, which are coupled ocean-atmosphere phenomena occurring every 4-7 years [Rasmusson and Wallace, 1983]. ENSO has global climatic impacts, some of which can be traced to western Pacific SST anomalies of 1C [e.g., Nicholls, 1985; Palmer and Mansfield, 1984]. These impacts may be predictable months to years in advance, provided that the processes involved in SST change can be understood and reliably modeled.

While it is generally conceded that understanding SST variability is important for understanding and predicting ENSO, little is known quantitatively about the atmospheric and oceanic processes that control this variability. The purpose of this study therefore is to examine the relationship between wind forcing, SST, and surface layer heat content, using data from a near-equatorial moored array along 165E. The analysis is a continuation of that presented by McPhaden [1989] and focuses mainly on time scales of days to months. The period of study coincides with the 1986-1987 ENSO, during which pronounced interannual variations occurred throughout the tropical Pacific (Figure 1) (see also Kousky and Leetmaa [1989]; McPhaden et al., [1990]).

Fig. 1. (a) Average SST and (b) SST anomaly in the western tropical Pacific for December 1986 to October 1987, based on data from the National Meteorological Center [Reynolds, 1988]. Contour interval is 1C in Figure 1a, except that the 29.5C isotherm is shown as dashed curve. Contour interval in Figure 1b is 0.5C, with positive values shown as solid contours and negative values shown as dashed contours. Hatching in Figure 1b indicates regions of negative SST anomaly. The current meter mooring (solid square) and ATLAS moorings (solid diamonds) are also indicated.

The paper is outlined as follows. Section 2 briefly describes the data used in this study. This is followed in section 3 by discussion of the mechanisms that are potentially important in controlling SST and surface layer heat content. In section 4 we describe the spatial structure of temperature and wind variations observed from the moored array. Then in section 5 we discuss how these variations may be related to turbulent heat fluxes across the air-sea interface, vertical advection, and entrainment from the thermocline. Lateral advection in the surface layer is discussed in section 6. Major results and conclusions are summarized in section 7.

2. Data

The data used in this study were collected from three surface moorings located at 2N, 0, and 2S, 165E in the western Pacific warm pool (Figure 1). Winds were measured on all three moorings at a height 4 m above mean sea level from wind recorders mounted on the surface toroid (Figure 2). Temperatures were measured at 15 depths on the equatorial mooring and 11 depths on ATLAS thermistor chain moorings at 2N and 2S (Figure 3). In each case the shallowest measurement was at 1 m (nominally termed SST) and the deepest was at 500 m. Currents were measured on the equatorial mooring using EG&G Model 610 Vector Averaging Current Meters (VACMS) at six depths, although for this study, only the 50-m record will be considered (Figure 14). All variables were processed to daily averages; however, hourly averaged wind data were first used to estimate wind speed, wind stress, and wind work in (2), (3), and (5) (given below) before computing daily means. Record lengths for winds, temperatures, and currents are explicit in Figures 2, 3 and 14, respectively. Longer records are available from these moorings (see McPhaden et al. [1990]), although the period from December 13, 1986, to October 14, 1987, has the most complete vertical temperature coverage at the equatorial site coincident with reliable wind information during 1986-1987. Further details concerning instrumental accuracies, sampling characteristics, and data processing are given by Freitag et al. [1987] and McPhaden et al. [1990].

Fig. 2. (a) Zonal and (b) meridional winds for December 13, 1986, to October 14, 1987, at 2N, 0 and 2S along 165E.

Fig. 3. Contours of daily averaged temperatures at 2S, 0, and 2N along 165E for December 13, 1986, to October 14, 1987. Contour interval is 2C, except that the 29C contour is shown as a dashed curve. Sensor depths are indicated on the left axis; 7 days of missing data are indicated by crosses.

3.Wind-Related Processes Affecting SST and Surface Layer Heat Content

In this section we review the processes related to wind forcing that can affect SST and surface layer heat content. Arguments are cast in terms of simple mixed layer models [e.g., Niiler and Kraus, 1977; McPhaden, 1982], where it is assumed that the temperature and velocity are uniform over some depth h. This is done for illustrative purposes only, since our measurements will show that the assumption of vertical uniformity of temperature variations is invalid in the western equatorial Pacific on the time scales we consider. Nonetheless, the model is conceptually useful for guiding the analysis.

With the above assumption, one can write

       (1)

where T/t is the local time rate of temperature change, u is horizontal velocity, w, is an entrainment velocity through the base of the mixed layer at depth z = -h, and Q is surface heat flux. The horizontal temperature gradient is denoted by T, and T is the change in temperature at the base of the mixed layer. Water density () and the heat capacity of water at constant pressure (C) are assumed constant at 10 kg m and 3.94 10 J kgC, respectively. Only terms that can be readily related to the wind field (as discussed below) and can be examined with our data sets have been retained in (1); other terms, namely, lateral turbulent diffusion, vertical turbulent diffusion through the base of the mixed layer, and penetrative solar radiation, have been neglected. These neglected terms may also depend on wind forcing, though it is not obvious how, with out limited data set, we can critically examine their relationship to SST and surface layer heat content.

3.1. Turbulent Heat Exchange at the Surface

Surface heat exchange Q is made up of longwave and shortwave radiative fluxes plus sensible and latent turbulent heat fluxes. Bulk parameterizations of the turbulent heat fluxes are given by

       (2a)        (2b)

where |U| is wind speed, q is an air-sea specific humidity difference, T is an air-sea temperature difference, C is a turbulent exchange coefficient, (= 1.2 kg m) is air density, and L (= 2440 J g) is the latent heat of evaporation. Wind speed dependence is explicit in these expressions; in addition, exchange coefficients may depend on wind speed [e.g., Large and Pond, 1982]. Latent heat fluxes with monthly mean values up to 100 W m are generally the larger of the two turbulent fluxes in the western equatorial Pacific [e.g., Reed, 1985; Liu, 1988], so that most discussions of turbulent exchange focus on this component alone.

Meyers et al. [1986], using shipboard data, and Liu [1988], using satellite data, have shown that during the 1982-1983 ENSO, latent heat flux variations on monthly time scales were highly correlated with wind speed variations in the western equatorial Pacific. However, these studies differed in their conclusion about the relative importance of latent heat fluxes in the surface temperature balance. Liu found little correlation between these fluxes and SST variability along the equator; in contrast, Meyers et al. suggested that a net 1C cooling of the western Pacific warm pool during 1982-1983 resulted from large latent heat fluxes associated with two episodes of strong equatorward meridional winds during mid-1982 (southern winter) and early 1983 (northern winter). The reason for these conflicting conclusions is unclear.

In a study that utilized many of the same time series data presented here, McPhaden et al. [1990] noted that high wind speed variations on weekly to monthly time scales during the 1986-1987 ENSO were associated with cool SST near the equator. This led to a suggestion that there existed a relationship between variations in latent heat flux and surface cooling on these time scales. However, McPhaden et al. did not quantify the suggested relationship between SST and wind speed in support of this hypothesis.

3.2. Vertical Advection and Entrainment

Vertical advection and entrainment are related but distinct processes involved in changing surface layer heat content and SST. Their combined effect during periods of easterly winds is often referred to as equatorial upwelling. Entrainment velocity w can be written as

       (3)

where wis vertical velocity at the base of the surface layer and h/t is the local time rate of change of the surface layer thickness. The inequality implies that in a stably stratified ocean, entrainment can only cool the surface. The velocity w is sometimes inferred from the vertical displacement of isotherms in the thermocline (assuming the effects of turbulent diffusion and lateral advection can be ignored). Processes contributing to w are tides; small-scale, high-frequency internal waves; equatorial waves; and Ekman pumping due to local wind-driven variations in horizontal mass convergence and divergence. When Ekman dynamics are operative, w on the equator will be proportional to the local zonal wind stress

       (4)

where C is a drag coefficient and U is zonal wind speed.

In the limit where h is a constant, entrainment velocity is just the vertical velocity at the base of the mixed layer. Under these circumstances, thermocline displacements (given the assumptions above) and SST would be in phase and highly correlated. Conversely, in the limit where the base of the mixed layer is a material surface (that is, w= 0), the mixed layer depth would change coherently with changes in thermocline depth and there would be no effect of entrainment on SST.

Entrainment can also be parameterized in terms of the surface layer turbulent energy balance [Niiler and Kraus, 1977; McPhaden, 1982]. Two sources of turbulent energy production directly related to the wind field are wind work (proportional to wind speed cubed) and free convection due to heat loss by evaporation and conduction (proportional to wind speed). The role of free convection cannot be evaluated without taking into account surface buoyancy fluxes caused by radiative heating and precipitation. On the other hand, the effect of wind work, which is always a source of turbulent energy production, can be expressed by

       (5)

where is the density jump across the base of the mixed layer, g is gravity, and m is an empirically determined efficiency factor [Niiler and Kraus, 1977]. Note that the efficiency of entrainment decreases as layer depth increases because, for a given wind speed, more of the water column has to be energized with turbulence. Other sources of turbulent energy production may enhance this entrainment (for example, production by vertical shear instability at the base of the mixed layer), but they can not be reliably estimated from our data set.

The role of entrainment mixing in the heat balance of the western equatorial Pacific was discussed by Meyers et al. [1986], who concluded that its effect on SST during the 1982-1983 ENSO was relatively small. McPhaden et al. [1990] found evidence for equatorial upwelling (that is, upward advection and entrainment) during a period of strong easterly winds at 165E in early 1988. However, based on the incoherence of SST and thermocline depth variations (as measured by the depth of the 20C isotherm over a period of 2.5 years during 1986-1988), they inferred that, in general entrainment of cold waters from the thermocline was not important in determining SST variability in the western equatorial Pacific. They suggested that this was owing to the relatively thick (50-100 m) surface layer and the fact that it was not always completely mixed with regard to temperature or salinity. Lukas and Lindstrom [this issue] and Godfrey and Lindstrom [1989] have hypothesized that the existence of a salt-stratified "barrier layer" in the temperature mixed layer inhibits entrainment in the western equatorial Pacific except during infrequent, large-amplitude westerly wind bursts. However, in none of these studies was the relationship between SST, surface layer heat content, and wind work according to (5) examined.

3.3. Lateral Advection

Another process that can cause SST to change is lateral advection. This is represented by the term u T in (1), which assumes slablike flow over the layer depth h. Whereas this is not strictly valid in the warm-pool region of the western Pacific, a significant component of flow is depth coherent in the surface layer [e.g., McPhaden et al., 1988]. Moreover, for a purely advective balance,

       (6)

which can be tested with data from a single depth.

4. Structure of Temperature and Wind Variations

Fig. 4. (a) Mean and (b) standard deviation of temperature between 2N and 2S, 165E for December 13, 1986 to April 16, 1987. Contour interval in Figure 4a is 2C (except for the 29C isotherm shown as a dashed line); contour interval in Figure 4b is 0.2C, with contours less than 1C dashed.

Figure 4a shows the mean meridional temperature section between 2N and 2S along 165E for the period from December 1986 to April 1987, which is the longest period of overlapping data at the three mooring sites. Mean SST is near to or greater than 29C during this time. The surface layer is nearly isothermal, with a mean vertical temperature difference in the upper 50-75 m at each of three locations of about 0.2C. The thermocline is centered between 100-200 m, with equatorial weakening indicative of the presence of the Equatorial Undercurrent. Note that during the second half of the record the upper thermocline is weaker and the surface layer is shallower because of an abrupt shoaling of the 24-28C isotherms in late May 1987 (Figure 3).

Figure 4b shows that the largest temperature variations (standard deviation 2C) are found in the thermocline on the equator between 100 and 150 m. Conversely, temperature variations in the surface layer are typically only 0.1-0.3C. Notably, SST variations tend to be larger than those at 50 m, which is a point that will be discussed in detail in section 5.

Figure 5 shows the first two empirical orthogonal eigenfunctions (EOFs) for temperature data summarized in Figure 4. Variations associated with these eigenfunctions are determined by multiplication of the EOFs with their corresponding time series. EOFs are defined to be orthogonal and the time series uncorrelated [Davis, 1976]; as such, they are an efficient means of isolating coherent modes of variability in large data sets.

Fig. 5. (a) Temperature empirical orthogonal eigenfunctions (EOF) mode 1 and (b) mode 2 for the period December 13, 1986, to April 16, 1987. Percent variance explained by each EOF is indicated. Hatching indicates negative eigenvector values.

The first EOF (Figure 5a) accounts for 56% of the temperature variance and is similar in structure to the total variability shown in Figure 4b, that is, the largest variance is on the equator in the thermocline. Variations are in phase in latitude by out of phase between the surface layer and the thermocline. The second EOF (Figure 5b), which accounts for 12% of the variance, shows a latitudinally antisymmetric mode of variability in the thermocline. In combination with meridional wind and velocity measurements from the 165E moorings (including those shown in Figures 2b and 14), Hayes et al. [1988] interpreted this structure in terms of a first baroclinic mode mixed Rossby-gravity wave, with a dominant period near 10 days. Hence in this analysis we will focus mainly on the first EOF, which accounts for most of the temperature variability during the period from December 1986 to April 1987.

We also performed an EOF decomposition of the overlapping wind record from December 1986 to April 1987 in Figure 2. Only the first EOF is shown, since it accounts for 77% (55%) of the variance in the zonal (meridional) directions (Figure 6). The first zonal wind EOF indicates in-phase variations of comparable amplitude across 2N-2S and a low-frequency variation that is similar to that shown in Figure 5a for the first temperature EOF. The cross-correlation coefficient for these two time series (0.54) is significantly nonzero, with 95% confidence. In contrast, the first meridional wind EOF has most of its variance concentrated at 2S, and its correlation with the first temperature EOF is not significant at the 95% level of confidence.

Fig. 6. Mode 1 wind EOFs for the period December 13, 1986, to April 16, 1987, in (a) the zonal direction and (b) the meridional direction. Percent variance explained by each EOF is indicated.

These results indicate that westerly winds (relative to the mean) are associated with warm thermocline temperatures and that easterly winds (relative to the mean) are associated with cold thermocline temperatures, a relationship similar to that expected for Ekman pumping. However, the out-of-phase relationship between surface layer temperature and thermocline temperature shown in Figure 5a suggests that wind-driven vertical advection of the thermocline may not be a significant factor in changing SST. This issue is examined in greater detail in section 5, where the analysis focuses on the longer equatorial mooring record with its finer vertical resolution in the surface layer.

Fig. 7. Time series of daily averaged wind speed (|U|), zonal wind pseudostress (|U|U), wind pseudowork (|U|), SST, and temperature at 50 m (dashed curve) from the current meter mooring at 0, 165E for the period December 13, 1986, to October 14, 1987. Note that 50-m temperatures are off-scale frequently after August 1987 because the upper thermocline has risen to this depth by then (see Figure 3).

5. Surface Turbulent Fluxes, Vertical Advection, and Entrainment

Figure 7 shows time series of wind speed (|U|), zonal wind pseudostress (|U|U), and pseudowork (|U|) from the mooring at 0, 165E for December 1986 to October 1987. In each case, estimates were first computed from hourly data, then averaged to daily values. Ignoring the weak wind speed dependence of C and C for the range of speeds shown in Figure 7, turbulent air-sea heat exchange will be proportional to |U|, Ekman pumping proportional to |U|U, and wind work proportional to |U| . The SST and 50-m temperature from the equatorial mooring are also shown in Figure 7.

If your browser cannot view the following table correctly, click this link for a GIF image of Table 1.
TABLE 1.  Summary of Means and Standard Deviations for Wind Speed, Zonal Wind
Pseudostress, Pseudowork, and Sea Surface Temperature From the Current Meter
Mooring at 0, 165E for the Period From December 13, 1986, to October 14, 1987

Wind Zonal
Speed Pseudostress, Pseudowork, SST, T(50 m),
m s m s m s C C

Mean 4.5   9.7 184 29.2 (29.2) (29.0)
Standard deviation 2.0 24.8 227     0.31 (0.22)     (0.11)

   Numbers in parentheses for SST and T(50 m) are statistics for December 13, 1986 to May 31, 1987,
when the surface layer extended to at least 50-m depth.

Table 1 summarizes the means and standard deviations of the time series in Figure 7. The mean wind speed is only 4.5 m s and the mean zonal pseudostress is eastward at 9.7 m s (for a constant C of 1.2 10, this pseudostress translates into a mean zonal stress of about 0.014 N m). Standard deviations in both of these variables are also small: 2.0 m s for |U| and 24.8 m s for |U|U (equivalent to 0.036 N m). SST is high on average (29.2C), with a standard deviation of 0.31C. Temperature at 50 m is 29.0C on average, with a standard deviation two times smaller than the SST standard deviation during the first part of the record, when 50 m is located in the surface layer.

Fig. 8. Cross correlations between SST and wind speed, zonal wind pseudostress, and wind pseudowork. Cross correlation extrema (r) and the lag (in days) at which they occur are shown. In each case, the lag indicates that cold SST follows high winds by 1 day. The 95% confidence limits for the null hypothesis of uncorrelated variability are indicated by horizontal bars on the abscissa.

Figure 8 indicates the cross correlations of the three wind constructs with SST are significantly nonzero at the 95% level of confidence and range between -0.43 (for zonal pseudostress) and -0.52 (for wind speed). The negative correlation indicates that high wind speed, high wind work, and westerly wind stress (relative to the mean) are associated with cool SST; low wind speed, low wind work, and easterly wind stress (relative to the mean) are associated with warm SST. In each case the maximum cross correlation occurs with SST lagging the winds by 1 day, consistent with the atmosphere forcing the ocean on these time scales. Although not presented, the magnitude of the cross correlation between meridional wind pseudostress and SST is generally 0.2 over the range of lags shown in Figure 8.

Fig. 9. Coherence and phase spectra for SST and wind speed, zonal wind pseudostress, and wind pseudowork at 0, 165E for the period December 13, 1986, to October 14, 1987. Positive phase indicates that high winds lead high SST, and negative phase indicates that high winds lead low SST. Horizontal lines superimposed on coherence estimates indicate 95% confidence limits for the null hypothesis of incoherent variability, based on five (15) frequency band averages at periods longer (shorter) than about 8 days.

Another way to view these relationships is in the frequency domain. Figure 9 shows the coherence and phase as a function of frequency between the three wind constructs and SST. Two bands of high and statistically significant coherence are evident. These are at periods between approximately 3-8 days and 30-60 days; to a lesser extent, variations at 10-20 days are also coherent. Consistent with the cross-correlation statistics, the phase in most frequency bands is between -90 and -180, implying that maxima in the winds are followed by cold SST and minima are followed by warm SST. Coherence at periods shorter than 20 days may be related to westerly wind burst activity. Coherence at periods of 30-60 days could be related to atmospheric fluctuations associated with Madden and Julian [1972] waves.

Note that although there are differences between the coherence and phase diagrams in Figure 9 and the cross-correlation analysis summarized in Figure 8, the statistical relationships of the three wind constructs with SST share many similarities; indeed, wind speed, zonal wind pseudostress, and wind work are highly correlated among themselves (0.61-0.90). Hence without further analysis one cannot unambiguously identify which of the processes related to the wind field are most important in affecting SST.

Fig. 10. Coherence amplitude as a function of depth averaged over periods of approximately 3-8 days (encompassing 36-65 frequency bands depending on depth) for SST and wind speed, zonal wind pseudostress, and wind pseudowork at 0, 165E. The 95% confidence limits for the null hypothesis of incoherent variability are indicated by the dashed lines. Note that the temperature record lengths for this analysis were chosen such that the statistics were stationary; for example, the period June-October 1987 was excluded at 50 m and 75 m. Shown on the right are the mean temperature profile (solid curve) and the standard deviation of temperature variations at periods of approximately 3-8 days (dashed curve) for December 13, 1986, to October 14, 1987.

Fig. 11. Coherence amplitude as a function of depth averaged over periods of approximately 30-60 days (that is 4-6 frequency bands, depending on depth) for SST and wind speed, zonal wind pseudostress, and wind pseudowork at 0, 165E. The 95% confidence limits for the null hypothesis of incoherent variability are indicated by the dashed lines. Shown on the right are the mean temperature profile (solid curve) and the standard deviation of temperature variations at periods of approximately 30-60 days (dashed curve) for December 13, 1986, to October 14, 1987.

Figure 10 shows coherence amplitude as a function of depth in the 3- to 8-day period band. Also shown in Figure 10 is the mean temperature profile for December 1986 to October 1987. Wind speed, wind stress, and wind work are all significantly coherent with temperature variations between 1 and 10 m. Coherence drops rapidly below these depths, however, and is statistically insignificant at the 95% level in the lower part of the surface layer between 50 and 75 m. In the thermocline, wind speed and wind work are generally incoherent with temperature fluctuations; on the other hand, zonal pseudostress shows a clear pattern of coherence between 125 and 300 m. The coherence profiles for the 10- to 20-day and 30- to 60-day period bands are generally similar to those shown in Figure 10. Figure 11, for example, shows the coherence as a function of depth for wind speed, zonal pseudostress, and wind work in the 30- to 60-day band. In each case the coherence is high at the surface, then falls off rapidly to relatively low and/or insignificant values. Only in the case of zonal pseudostress does significantly nonzero coherence appear consistently in the thermocline.

Fig. 12. Coherence amplitude and phase as a function of depth averaged over periods of approximately 3-8 days (encompassing 36-65 frequency bands, depending on depth) for SST and zonal wind pseudostress at 0, 165E. The 95% confidence limits for the null hypothesis of incoherent variability are indicated by the dashed curve.

Figure 12 shows that the phase of zonal pseudostress coherence in the 3- to 8-day band is such that warm thermocline temperatures lag relative westerly winds by about 90 and cold thermocline temperatures lag relative easterly winds by the same amount. This is consistent with the Ekman pumping hypothesis for thermocline variations discussed in section 4. However, phase changes by 180 across the base of the surface layer, indicating that temperature variations are of opposite sign in the surface layer and thermocline. A similar phase change occurs at lower frequencies as well. Vertical advection from the thermocline is therefore not likely to be the principal mechanism controlling SST on these time scales, as was earlier suggested in the discussion of Figure 5a. Likewise, the rapid decrease in coherence with depth below 10 m in Figures 10 and 11 suggests that entrainment from the thermocline is not, in general, important in controlling SST. One would expect coherence between temperature and |U| to be much higher at 50-75 m if the thermocline were the source of cold water for the surface layer. This leaves turbulent surface fluxes proportional to |U| as the most likely candidate for affecting SST and surface layer heat content.

Fig. 13. Time series of H/t for depth ranges of (a) 0-10 m for the period December 13, 1986, to July 20, 1987, and (b) 0-75 m for the period December 13, 1986, to May 31, 1987. The standard deviation of heat content variations () and cross correlation with wind speed at zero lag (r) are shown.

Further evidence to support the role of wind-related surface turbulent heat flux variations can be found in an examination of upper ocean heat content (defined as H = pChT) from the 0, 165E mooring. Figure 13 shows time series of H/t estimated using centered differences over two depth ranges: 0-10 m (where coherence with the wind speed is highest) and 0-75 m (the average depth of the surface layer). The calculation for 0-75 m is shown only through the end of May 1987, because of the nonstationarity of the time series during the second half of the record when the upper portion of the thermocline reaches depths of 50-75 m (see Figures 3 and 7). The standard deviation of H/t() and the cross correlation at zero lag (r) of H/t and |U| also appear in Figure 13. For both depth ranges the cross correlations are significantly nonzero (with 95% confidence) and are higher than for either pseudostress or for wind work.

Variations in heat content within the 0- to 75-m layer are 102 W m. Multiplying this number by 0.32, that is, the magnitude of the correlation between H/t and |U|, indicates that heat content changes of approximately 33 W m in the upper 75-m surface layer are due to wind speed variations. This estimate is comparable to what would be expected for latent heat flux variations based on (2a) and typical observed parameter values. Specifically, Q 35 W m for wind speed variations of 2.0 m s (Table 1), C, of 1.2 10 [Large and Pond, 1982] and a typical q of 5 g kg [Reed, 1985; Liu, 1988].

For comparison, the magnitude of variations in entrainment mixing due to wind work can be estimated from (1) and (5), using a scale analysis similar to that for latent heat flux. For |U| of 227 m s (Table 1), m = 0.4 [Davis et al., 1981; McPhaden, 1982], (T/p) = 3C m kg at the base of the surface layer (determined from conductivity-temperature-depth (CTD) casts near 0, 165E), and h = 75 m, we find that CwT =5 W m . Other choices of m, h, and ( T/ p) would result in different estimates of entrainment from the thermocline; however, it is unlikely that a reasonable combination of these parameters would increase our 5 W m estimate by an order of magnitude. Thus we conclude that latent heat flux variations are significantly more important than wind work-related entrainment fluxes from the thermocline in controlling SST and surface layer heat content during the time period of this study.

A comparison of heat content variations over the 0- to 10-m layer and the 0- to 75-m layer indicates that surface cooling associated with Q converges nonlinearly in the upper 75 m. The product of |r| and over the two depth ranges in Figure 13 implies a standard deviation of 14 W m and 33 W m for wind-related heat content variations in the upper 10 m and 75 m, respectively. The 19 W m difference between these two flux estimates represents variability in the depth range 10-75 m. Thus 14 W m (42% of the total) converges in the upper 10 m (13% of the 75-m layer depth) whereas 19 W m (58% of the total) converges between 10 and 75 m (87% of the surface layer depth). Using an integral time scale (t) for wind-speed-related temperature variations of 3-4 days (based on the cross-correlation analysis in Figure 8) and a H/t of 14 W m (0-10 m) and 19 W m (10-75 m), we find the expected magnitude of temperature variations (T ~ tr/Ch) associated with latent heat flux variations in each of these depth ranges to be about 0.1C and 0.02C, respectively. The latter temperature is close to the instrumental accuracy of our temperature sensors [e.g., McPhaden et al., 1990] and probably accounts for the insignificant temperature and wind coherences on a pointwise basis for depths greater than 10 m in the surface layer.

6. Lateral Advection

Another possible candidate contributing to variations in SST and heat content is lateral advection. A 50-m velocity record is the shallowest time series we have in 1987 to test this hypothesis. However, as discussed by McPhaden et al. [1990], daily averaged 10- and 50-m currents are highly correlated (0.87 for 164 days of overlapping 1986 data), so that fluctuations at 50 m can be used as an index of flow in the surface layer.

Fig. 14. Zonal velocity (u), meridional velocity (v), and time rate of change of temperature (T/T) at 50 m from December 13, 1986, to May 31, 1987, at 0, 165E. Daily averaged data are shown by thin curves and 15-day Hanning filtered data are shown by thick curves.

Velocity at 50 m is shown in Figure 14 for that time period when the 50-m depth is located in the surface layer (see Figure 3). Flow in the zonal direction is strongly eastward at speeds greater than 100 cm s in December 1986, westward at speeds up to 70 cm s from late January to early April 1987, then eastward again beginning in mid-April. The range of zonal speeds in these month-to-month measurements is about 200 cm s. Meridional velocities vary at periods near 10 days with peak-to-peak amplitudes of 50-100 cm s ; these fluctuations are associated with the second temperature EOF (Figure 5b) and are related to mixed Rossby-gravity wave dynamics [Hayes et al., 1988]. Velocity variations in both directions are in part forced by local winds, as discussed by McPhaden et al. [1990] and Hayes et al. [1988].

Figure 14 also shows the centered difference equivalent of T/t, variations of which would be highly correlated with one or both of the velocity components if lateral advection were dominant in the temperature balance. In addition, according to (6), one would be able to estimate the time-averaged horizontal temperature gradient by regressing velocity against T/t. However, cross correlations of daily averaged zonal velocity (u) and meridional velocity (v) with T/t are only 0.05 and 0.17, respectively, neither of which exceeds the 95% confidence limits for the null hypothesis of 0.11 in the zonal direction and 0.26 in the meridional direction. We repeated these calculations for data smoothed with a 15-day Hanning filter (approximately equivalent to computing weekly averages). As shown in Figure 14, there are times such as early March to mid-April when the low-pass T/t and meridional velocity track reasonably closely. Overall however, the cross correlations for the 15-day Hanning (u, T/t) and (v, T/t) data are 0.14 and 0.23, neither of which exceeds the 95% confidence limits of 0.34 and 0.38 for the low-pass time series. We repeated these calculations for T/t computed from SST and from temperature averaged over 0-10 m and 0-75 m, assuming that the 50-m currents were representative of depth-averaged flow in the surface layer. No clear pattern of significant cross correlation emerged, implying that, in general, lateral advection has a negligible effect on surface layer temperature variability on daily to monthly time scales for the period of this study. This is consistent with a mean temperature field characterized by weak horizontal gradients in the vicinity of 0, 165E during much of 1987 (for example, Figure 1).

7. Summary and Discussion

The foregoing analysis has focused on a diagnosis of the surface layer heat balance using equatorial mooring data along 165E during 1986 and 1987. Several processes related to wind forcing were examined, using statistical relationships between winds, temperature, and currents. We found that among the processes considered, evaporative cooling was the most important, accounting for heat content changes of typically 35 W m in the upper 75 m. Moreover, these evaporative heat fluxes converged nonuniformly in the surface layer, giving rise to temperature variations of about 0.1C in the upper 10 m, but only about 0.02C between 10 and 75 m. Entrainment from the thermocline due to wind work was an order of magnitude smaller. In addition, the affects of lateral advection were not significant, in spite of large (50-100 cm s) velocity fluctuations in the surface layer. This is because the horizontal temperature gradients were weak near 0, 165E during the time period of our analysis.

The inferred nonuniform convergence of evaporative heat flux in the surface layer, the weak temperature variations at 50 m relative to those at the surface (for example, Figure 7), and the weakness of wind work entrainment from the thermocline indicate that some process or combination of processes significantly limits the vertical extent to which heat can be mixed in the western Pacific warm pool. One possibility is the formation of a salt-stratified barrier layer, as suggested by Lukas and Lindstrom [this issue] and Godfrey and Lindstrom [1989]. Although we do not have salinity time series from the period of our analysis to test this hypothesis quantitatively, CTD data along 165E show the existence of shallow, 0- to 10-m thick lenses of low-salinity water, with surface values up to 1 psu (practical salinity unit) lower than normal. These may be related to enhanced precipitation associated with an eastward displacement of deep convection across 165E during the 1986-1987 ENSO. It has yet to be determined, however, whether the haloclines associated with these buoyant lenses are sufficiently strong to trap heat in a shallow layer near the surface.

Another possibility is that formation of the diurnal thermocline may be involved in limiting the depth penetration of heat on longer time scales. Daytime radiative heating leads to typical diurnal temperature variations of O(0.1C) in the upper 10 m of the western equatorial Pacific [Taft and McPhaden, 1988]. The amplitude of the diurnal cycle is related to wind speed in qualitatively the same manner in which daily averaged SST is related to wind speed, that is, the diurnal cycle is weak when wind speed is high (>6 m s) and strong (>1C) during periods of calm winds. Wind speed was greater than 6 m s for 22% of the time during the period from December 1986 to October 1987; however, most of the time the winds were not strong enough to erode the diurnal thermocline at the mooring site. One-dimensional mixed layer models indicate that inclusion of the diurnal heating cycle quenches convection during the day, which leads to shallower mixed layer depths than would be otherwise expected without the diurnal cycle because of nonlinearities in the turbulent energy balance [Woods and Barkmann, 1986]. Moreover, in the eastern equatorial Pacific Imawaki et al. [1988] have shown that the diurnal cycle modulates the turbulent heat flux out of the surface layer on longer time scales. Most of the turbulence and mixing there occurs at night in association with free convection [Peters et al., 1988]. Hence if formation of the diurnal thermocline were inhibited, nighttime convection could start sooner, last longer, be more intense, and penetrate to greater depths, resulting in a larger net downward heat flux. Conversely, if the winds were weak during the day and the diurnal thermocline strong, nighttime convection could be delayed, of shorter duration, weaker, and less vertically penetrative.

Lukas and Lindstrom [this issue] have suggested that repeated westerly wind burst forcing associated with ENSO may lead to long-period, interannual cooling of the western Pacific warm pool through wind work-generated entrainment mixing with the thermocline. As noted above, entrainment does not appear to be important in general along 165E during 1987 at the height of the ENSO. Moreover, visually and by least squares analysis, there is no significant SST cooling trend over the record length shown in Figure 7, even though several 5-10 m s westerly wind events occur during this time period. In addition, the mean SST for December 1986 to October 1987 was not significantly colder than climatology (Figure 1); specifically, SST was 29.2C (plus or minus one standard error of 0.1C) at 0, 165E compared to a climatological mean over the same time period of 29.0C [Reynolds, 1988]. Thus while we have shown that short-period westerly wind fluctuations lead to local SST variability on the same time scale, there is no evidence from our data that they lead to local low-frequency SST cooling associated with ENSO. Whether this result applies to other locations in the western equatorial Pacific as well is not clear.

Our conclusions with regard to advection and evaporation rely primarily on inference, because we do not have specific humidity to calculate latent heat flux, nor do we have accurate enough information on horizontal temperature gradients to estimate advection explicitly. In addition, our estimate of 5 m as a typical entrainment heat flux based on wind work is only approximate, because we do not have the data to incorporate into our estimate the affects of other sources and sinks of turbulent energy (for example, buoyancy forcing due to insolation and precipitation; vertical shear instability, etc.). Because of these and other data limitations, we have been able to account for only about 10-25% of the observed SST and heat content variance in terms of evaporative cooling, based on correlations with wind speed of 0.3-0.5 (for example, Figures 8 and 13). Shortwave radiation is likely to account for a large percentage of the unexplained temperature variance in the surface layer. Insolation at the equator under clear-sky conditions ranges between about 300 and 350 W m. Using Reed's [1977] flux formula, changes in cloudiness of only 10% would lead to changes in the net shortwave radiation at the sea surface of about 20 W m. Changes in cloudiness of this magnitude are likely to occur on daily to monthly time scales, given that normally the skies are 60-70% cloud-covered owing to very active deep convection in the western equatorial Pacific.

In conclusion, we note that the SST changes observed in our mooring data are only O(0.1C). This is an order of magnitude smaller than those seen in the eastern equatorial Pacific, where O(1C) fluctuations on daily to interannual time scales are common [e.g., McPhaden and Hayes, 1990]. It is encouraging that in spite of these small temperature signals and our limited data set, we have been able to draw inferences about physical processes at work in the surface layer heat balance. We emphasize that these inferences are specific to the time period and location of this study. Entrainment from the thermocline, for example, could be more important during periods of strong upwelling favorable easterly winds. Likewise, lateral advection could be important during periods when horizontal SST gradients are pronounced, or on longer time scales (e.g. interannual) than those considered in this study. Our analysis emphasizes the need, therefore, for longer and more complete data sets in future investigations of air-sea interaction in the western equatorial Pacific.

Acknowledgments. We would like to thank Stuart Godfrey of CSIRO, Hobart, Australia, for helpful comments on an earlier version of this manuscript. We would also like to acknowledge M. McCarty and H.P. Freitag for their assistance in carrying out the analyses. This work was supported by the U.S. Tropical Ocean-Global Atmosphere (TOGA) Project Office. Contribution 1166 from the Pacific Marine Environmental Laboratory, National Oceanic and Atmospheric Administration.

References

Davis, R. E., Predictability of sea surface temperature and sea level pressure anomalies over the North Pacific Ocean, J. Phys. Oceanogr., 6, 249-266, 1976.

Davis, R. E., R. De Szoerke, and P. P. Niiler, Variability in the upper ocean during MILE, II, Modeling the mixed layer response, Deep Sea Res., 28, 1453-1476, 1981.

Freitag, H. P., M. J. McPhaden, and A. J. Shepherd, Equatorial current and temperature data: 108W to 110W; October 1979 to November 1983, NOAA Data Rep., ERL PMEL-17, 99 pp., Environ. Res. Lab., Natl. Oceanic and Atmos. Admin., Boulder, Colo., 1987.

Godfrey, J. S., and E. J. Lindstrom, The heat budget of the equatorial western Pacific surface mixed layer, J. Geophys. Res., 94, 8007-8017, 1989.

Hayes, S. P., L. J. Mangum, M. J. McPhaden, and J. Picaut, Thermal structure variability along 165E, paper presented at Symposium on Western Tropical Pacific Air-Sea Interactions, U.S. TOGA Project Office, Beijing, China, Nov. 15-17, 1988.

Imawaki, S., P. P. Niiler, C. H. Gautier, D. Halpern, R. A. Knox, W. G. Large, D. S. Luther, J. C. McWilliams, J. N. Moum, and C.A. Paulson, A new method for estimating the turbulent heat flux at the bottom of the daily mixed layer, J. Geophys. Res., 93, 14,005-14,012, 1988.

Kousky, V. E., and A. Leetmaa, The 1986-87 Pacific warm episode: Evolution of oceanic and atmospheric anomaly fields, J. Clim., 2, 254-267, 1989.

Large, W. G., and S. Pond, Sensible and latent heat fluxes over the ocean, J. Phys. Oceanogr., 12, 464-482, 1982.

Liu, W. T., Moisture and latent heat flux variabilities in the tropical Pacific derived from satellite data, J. Geophys. Res., 93, 6749-6760, 1988.

Lukas, R., and E. Lindstrom, The mixed layer of the western equatorial Pacific Ocean, J. Geophys. Res., this issue.

Madden, R. A., and P. Julian, Description of global scale circulation cells in the tropics with a 40-50 day period, J. Atmos. Sci., 29, 1109-1123, 1972.

McPhaden, M. J., Variability in the central equatorial Indian Ocean, II, Oceanic heat and turbulent energy balances, J. Mar. Res., 40, 403-419, 1982.

McPhaden, M. J., On the relationship between winds and upper ocean temperature variability in the western equatorial Pacific, in Proceedings of the Western Pacific International Meeting and Workshop on TOGA-COARE, edited by J. Picaut, R. Lukas, and T. Delcroix, pp. 283-290, ORSTOM, Noumea, New Caledonia, 1989.

McPhaden, M. J., H. P. Freitag, S. P. Hayes, B. A. Taft, Z. Chen, and K. Wyrtki, The response of the equatorial Pacific Ocean to a westerly wind burst in May 1986, J. Geophys. Res., 93, 10,589-10,603, 1988.

McPhaden, M. J., S. P. Hayes, L. J. Mangum, and J. Toole, Variability in the western equatorial Pacific during the 1986-87 El Nio/Southern Oscillation event, J. Phys. Oceanogr., 20, 190-208, 1990.

McPhaden, M. J., and S. P. Hayes, Variability in the eastern equatorial Pacific Ocean during 1986-1988, J. Geophys. Res., 95, 13,195-13,208, 1990.

Meyers, G., J.-R. Donguy, and R. K. Reed, Evaporative cooling of the western equatorial Pacific Ocean by anomalous winds, Nature, 323, 523-526, 1986.

Nicholls, N., Impact of the Southern Oscillation on Australian crops, J. Climatol., 5, 553-560, 1985.

Niiler, P. P., and E. B. Kraus, One-dimensional models of the upper ocean, in Modelling and Prediction of the Upper Layers of the Ocean, edited by E. B. Kraus, pp. 143-172, Pergamon, New York, 1977.

Palmer, T. N., and D. A. Mansfield, Response of two atmospheric general circulation models to sea-surface temperature anomalies in the tropical east and west Pacific, Nature, 310, 483-488, 1984.

Peters, H., M. C. Gregg, and J. M. Toole, On the parameterization of equatorial turbulence, J. Geophys. Res., 93, 1199-1218, 1988.

Rasmusson, E. M., and J.M. Wallace, Meteorological aspects of the El Nio/Southern Oscillation, Science, 222, 1195-2002, 1983.

Reed, R. K., On estimating insolation over the ocean, J. Phys. Oceanogr., 7, 482-485, 1977.

Reed, R. K., An estimate of the climatological heat fluxes over the tropical Pacific Ocean, J. Clim. Appl. Meteorol., 24, 833-840, 1985.

Reynolds, R. W., A real-time global sea surface temperature analysis, J. Clim., 1, 75-86, 1988.

Taft, B. A., and M.J. McPhaden, Diurnal cycle of sea surface temperature in the western tropical Pacific, paper presented at Symposium on Western Tropical Pacific Air-Sea Interactions, U.S. TOGA Project Office, Beijing, China, Nov. 15-17, 1988.

Woods, J. D., and W. Barkmann, The response of the upper ocean to solar heating, I, The mixed layer. Q. J. R. Meteorol. Soc., 112, 1-28, 1986.


Figure Captions

Fig. 1. (a) Average SST and (b) SST anomaly in the western tropical Pacific for December 1986 to October 1987, based on data from the National Meteorological Center [Reynolds, 1988]. Contour interval is 1C in Figure 1a, except that the 29.5C isotherm is shown as dashed curve. Contour interval in Figure 1b is 0.5C, with positive values shown as solid contours and negative values shown as dashed contours. Hatching in Figure 1b indicates regions of negative SST anomaly. The current meter mooring (solid square) and ATLAS moorings (solid diamonds) are also indicated.

Fig. 2. (a) Zonal and (b) meridional winds for December 13, 1986, to October 14, 1987, at 2N, 0 and 2S along 165E.

Fig. 3. Contours of daily averaged temperatures at 2S, 0, and 2N along 165E for December 13, 1986, to October 14, 1987. Contour interval is 2C, except that the 29C contour is shown as a dashed curve. Sensor depths are indicated on the left axis; 7 days of missing data are indicated by crosses.

Fig. 4. (a) Mean and (b) standard deviation of temperature between 2N and 2S, 165E for December 13, 1986 to April 16, 1987. Contour interval in Figure 4a is 2C (except for the 29C isotherm shown as a dashed line); contour interval in Figure 4b is 0.2C, with contours less than 1C dashed.

Fig. 5. (a) Temperature empirical orthogonal eigenfunctions (EOF) mode 1 and (b) mode 2 for the period December 13, 1986, to April 16, 1987. Percent variance explained by each EOF is indicated. Hatching indicates negative eigenvector values.

Fig. 6. Mode 1 wind EOFs for the period December 13, 1986, to April 16, 1987, in (a) the zonal direction and (b) the meridional direction. Percent variance explained by each EOF is indicated.

Fig. 7. Time series of daily averaged wind speed (|U|), zonal wind pseudostress (|U|U), wind pseudowork (|U|), SST, and temperature at 50 m (dashed curve) from the current meter mooring at 0, 165E for the period December 13, 1986, to October 14, 1987. Note that 50-m temperatures are off-scale frequently after August 1987 because the upper thermocline has risen to this depth by then (see Figure 3).

Fig. 8. Cross correlations between SST and wind speed, zonal wind pseudostress, and wind pseudowork. Cross correlation extrema (r) and the lag (in days) at which they occur are shown. In each case, the lag indicates that cold SST follows high winds by 1 day. The 95% confidence limits for the null hypothesis of uncorrelated variability are indicated by horizontal bars on the abscissa.

Fig. 9. Coherence and phase spectra for SST and wind speed, zonal wind pseudostress, and wind pseudowork at 0, 165E for the period December 13, 1986, to October 14, 1987. Positive phase indicates that high winds lead high SST, and negative phase indicates that high winds lead low SST. Horizontal lines superimposed on coherence estimates indicate 95% confidence limits for the null hypothesis of incoherent variability, based on five (15) frequency band averages at periods longer (shorter) than about 8 days.

Fig. 10. Coherence amplitude as a function of depth averaged over periods of approximately 3-8 days (encompassing 36-65 frequency bands depending on depth) for SST and wind speed, zonal wind pseudostress, and wind pseudowork at 0, 165E. The 95% confidence limits for the null hypothesis of incoherent variability are indicated by the dashed lines. Note that the temperature record lengths for this analysis were chosen such that the statistics were stationary; for example, the period June-October 1987 was excluded at 50 m and 75 m. Shown on the right are the mean temperature profile (solid curve) and the standard deviation of temperature variations at periods of approximately 3-8 days (dashed curve) for December 13, 1986, to October 14, 1987.

Fig. 11. Coherence amplitude as a function of depth averaged over periods of approximately 30-60 days (that is 4-6 frequency bands, depending on depth) for SST and wind speed, zonal wind pseudostress, and wind pseudowork at 0, 165E. The 95% confidence limits for the null hypothesis of incoherent variability are indicated by the dashed lines. Shown on the right are the mean temperature profile (solid curve) and the standard deviation of temperature variations at periods of approximately 30-60 days (dashed curve) for December 13, 1986, to October 14, 1987.

Fig. 12. Coherence amplitude and phase as a function of depth averaged over periods of approximately 3-8 days (encompassing 36-65 frequency bands, depending on depth) for SST and zonal wind pseudostress at 0, 165E. The 95% confidence limits for the null hypothesis of incoherent variability are indicated by the dashed curve.

Fig. 13. Time series of H/t for depth ranges of (a) 0-10 m for the period December 13, 1986, to July 20, 1987, and (b) 0-75 m for the period December 13, 1986, to May 31, 1987. The standard deviation of heat content variations () and cross correlation with wind speed at zero lag (r) are shown.

Fig. 14. Zonal velocity (u), meridional velocity (v), and time rate of change of temperature (T/T) at 50 m from December 13, 1986, to May 31, 1987, at 0, 165E. Daily averaged data are shown by thin curves and 15-day Hanning filtered data are shown by thick curves.

Table 1. Summary of means and standard deviations for wind speed, zonal wind pseudostress, pseudowork, and sea surface temperature from the current meter mooring at 0, 165E for the period from December 13, 1986, to October 14, 1987.


Return to Abstract

PMEL Outstanding Papers

PMEL Publications Search

PMEL Homepage


ryan.layne.whitney@noaa.gov