a. Data

There are ∼55 moored TAO buoys in the tropical Pacific Ocean (9°N–8°S, 165°E–95°W) located approximately every 2°–3° of latitude and 15° of longitude. The spacing of the buoys was chosen to be commensurate with the narrow meridional scale of oceanic processes at the equator, such as equatorially trapped waves and upwelling motions that determine sea surface temperature (SST) anomalies during ENSO. This arrangement of narrow meridional spacing of the buoys is not optimal for the study of large-scale atmospheric circulation. Nevertheless, the TAO array is a valuable tool for monitoring daily variations of surface winds across the tropical Pacific because of its unprecedented spatial and temporal coverage. This study examines 12 yr (1 January 1993–31 December 2004) of hourly wind data from the 51 buoys that had at least 50% of data coverage (Table 1). For the selected buoys, the average number of days per year containing data is 278, ranging from a minimum of 193 at 5°S, 165°E to a maximum of 365 at 5°S, 140°W. These numbers account for the time span of data collection at each buoy as well as the frequent gaps in the record due to instrument failure.

The TAO array samples winds at 3.8 m above sea level. Wind speed measurements are made with a propeller and have a resolution of 0.2 m s⁻¹ and an accuracy of ±0.3 m s⁻¹. A vane or a fluxgate compass is used to measure wind direction at 1.4° resolution with an accuracy of 5°–7.8°. Wind data were recorded for a 6-min interval at the beginning of each hour until 1996, after which they are recorded every 10 min.

In addition to the wind data, hourly measurements of barometric pressure, SST, and air temperature from the TAO buoys are also examined for further analysis. Pressure data became available after 1 January 2000 when pressure sensors were added to the 10 easternmost TAO buoys along 110° and 95°W (Table 1) as part of the East Pacific Investigation of Climate (EPIC) study. Owing to this data limitation, all available pressure data between 1 January 2000 and 31 December 2004 are analyzed regardless of the low temporal coverage at some buoys (Table 1). Data coverage for SST and air temperature is high (Table 1).

Barometric pressure is measured using a pressure transducer at 3 m above sea level with 0.1-hPa resolution and an accuracy of ±0.01%. SST is measured at 1 m below the sea surface using a thermistor with 0.001°C resolution and an accuracy of ±0.02°–0.03°C. Air temperature is measured at 3 m above sea level using a resistance temperature recorder with 0.01°C resolution and ±0.2°C accuracy. Further details of the TAO array datasets can be found on the TAO Project Office Web site of the National Oceanic and Atmospheric Administration (NOAA)/Pacific Marine Environmental Laboratory (available online at http://www.pmel.noaa.gov/tao/index.shtml).

b. Methods

Seasonal (December–February, March–May, June–August, and September–November) mean daily wind marches (i.e., mean variation of wind from hour to hour over the course of a day) were computed for each buoy and year by averaging all available measurements of the two wind components at every hour within the given 3-month period. The seasonally averaged daily wind marches were then averaged over all 12 yr for each buoy. Annual and semiannual (December–May and June–November) mean marches were constructed by averaging the seasonal mean marches. Finally, the daily means of the annual and semiannual mean hourly wind marches were removed.

All annual and semiannual mean hourly wind marches constructed in this manner were decomposed into diurnal and semidiurnal harmonics following

View Expanded

where X denotes the zonal or meridional wind anomalies at the tth hour of the day (t in local time), with k referring to the diurnal harmonic (k = 1) or semidiurnal harmonic (k = 2). Here A and B are values determined by the method of least squares (Draper and Smith 1966). To portray the spatial patterns of the diurnal and semidiurnal wind variations, the amplitudes and phases of the harmonics at each buoy are plotted on a map in vector format.

a. Mean wind field

As background to the analysis of diurnal and semidiurnal variability, we first present the annual and semiannual (December–May and June–November) mean wind fields from the TAO buoys during 1993–2004 (Fig. 1). December–May and June–November correspond to the warm and cold seasons, respectively, in the tropical Pacific (Wallace et al. 1989; Hartten and Datulayta 2004, hereafter HD). Surface winds are predominantly easterly across the basin throughout the year, converging into the region of deep convection (as inferred from outgoing longwave radiation values less than 240 W m⁻²) over the western Pacific. Surface convergence associated with the meridional component of the wind is evident within the areas of deep convection along the intertropical convergence zone (ITCZ) and the South Pacific convergence zone (SPCZ). The semiannual variation of the surface wind field is qualitatively consistent with that of deep convection, with increased meridional flow into the enhanced convective region along the ITCZ from June through November (Fig. 1c) and along the SPCZ from December through May (Fig. 1b).

b. Daily wind harmonics: Annual mean

Figures 2 and 3 show the annual mean amplitudes and phases of the diurnal and semidiurnal harmonics of the zonal and meridional winds, respectively. The length of the vector represents the amplitude of the harmonic at each buoy, and its direction denotes the phase in LT. Arrowheads on the bars of the diurnal harmonics of the zonal (meridional) wind point to the time of maximum westerly (southerly) wind anomaly (the term “anomaly” is used with respect to the daily mean wind). Due north indicates a maximum at 0600 LT, with time increasing in the counterclockwise direction (e.g., due west is 1200 LT, due south is 1800 LT, due east is 0000 LT). Arrowheads are not present on the bars of the semidiurnal harmonics because maximum winds occur twice a day, 12 h apart. For example, a vertical bar in the semidiurnal zonal (meridional) wind plot indicates maximum westerly (southerly) wind anomalies at 0600 and 1800 LT.

The phases and amplitudes of the semidiurnal harmonics of the zonal wind are strikingly uniform; maximum westerly wind anomalies are observed consistently ∼0325/1525 LT with amplitudes ∼0.14 m s⁻¹ (Fig. 2b). The semidiurnal signal is weak along 125° and 110°W at and south of the equator. Diurnal variability has generally smaller amplitudes and less coherent phases than semidiurnal variability (Fig. 2a).

The daily meridional wind cycle is dominated by diurnal variability whose amplitude is approximately 3 times the amplitude of the semidiurnal variability at most buoys (Fig. 3). Diurnal variability of the meridional wind in the western Pacific (165°E–155°W) exhibits an approximately 12-h phase difference between buoys along 2°–5°N and those along 5°–8°S; maximum southerly wind anomalies are observed between 0600 and 0900 LT in the Northern Hemisphere, and between 1800 and 2100 LT in the Southern Hemisphere. In the eastern Pacific (125°–95°W), the phase propagates more gradually from north to south, with maximum southerly wind anomalies ∼1000 LT at 5°N progressing to ∼2000 LT at 8°S. Semidiurnal variations of the meridional wind, despite their small amplitudes, exhibit coherent phases with maximum southerly wind anomalies ∼0000/1200 LT (∼0600/1800 LT) in the Northern (Southern) Hemisphere (Fig. 3b). This pattern of cross-equatorial asymmetry of the semidiurnal meridional wind anomalies is in quadrature to the semidiurnal cycle of the zonal wind (Fig. 2b) and is likely a result of the Coriolis force, as discussed in DS.

Figure 4 quantifies the relative amplitudes of the diurnal and semidiurnal harmonics of each wind component averaged across the buoy array. The mean amplitude of the semidiurnal harmonics of the zonal wind is twice as large as that of the diurnal harmonics (0.14 versus 0.07 m s⁻¹), while the mean diurnal amplitude of the meridional wind is more than 3 times as large as its mean semidiurnal amplitude (0.19 versus 0.06 m s⁻¹).

c. Daily wind harmonics: Semiannual means

Figures 5 and 6 show semiannual (December–May and June–November) mean amplitudes and phases of the diurnal and semidiurnal harmonics of the zonal and meridional winds, respectively. In addition to the general patterns of their annual mean counterparts, the semiannual harmonics exhibit some noteworthy differences. For example, the suppression of the semidiurnal zonal wind harmonic along 125° and 110°W at, and south of, the equator, is even more apparent in June–November than in the annual mean and December–May (Figs. 2b and 5c,d). The diurnal meridional wind harmonics are also seasonally dependent, with larger amplitudes in June–November than in December–May (Figs. 6a,b). The enhancement is particularly evident over the eastern basin.

Averaged over the buoy array, the daily zonal and meridional wind cycles are dominated by semidiurnal and diurnal variability, respectively, for both semiannual means (Fig. 7). The basin-averaged amplitude of the semidiurnal zonal wind harmonics does not vary significantly from season to season, while that of the diurnal meridional wind harmonics is clearly larger during June–November compared with December–May. The semidiurnal variability of the zonal wind component and the diurnal variability of the meridional wind component are analyzed further, as they dominate the climatological daily mean wind cycle over the tropical Pacific Ocean.

a. Semidiurnal variability

In the absence of friction and nonlinear effects, the momentum balance for the zonal wind component may be expressed as

View Expanded

where the zonal wind acceleration (δu/δt) is associated with the pressure gradient in the east–west direction (δp/δx), the density of air (ρ = 1.2 kg m⁻³), and the Coriolis force ( fυ, where f is the Coriolis parameter and υ is the meridional wind). We use Eq. (2) to assess whether the observed semidiurnal zonal wind variations over the tropical Pacific are consistent with those derived from the local pressure gradient force associated with atmospheric thermal tides.

Classical atmospheric thermal tidal theory (e.g., Chapman and Lindzen 1970) predicts a zonally uniform semidiurnal pressure wave in the tropics. Haurwitz and Cowley (1973, hereafter HC) show that the amplitude and phase of the semidiurnal pressure variability are relatively constant within 20° of the equator, with values ∼1 hPa and ∼0945/2145 LT, respectively. Their formula for the annual mean semidiurnal surface pressure variation based on zonally averaged land station records is

View Expanded

where θ is latitude and t is local time in degrees (i.e., one day is divided into 360 equal parts). Since HC approximated the global distribution of the semidiurnal pressure wave based only on land station data, the accuracy of their results over the ocean is questionable. Nevertheless, ship measurements of sea level pressure over the tropical Pacific (9°N–9°S, 162.5°E–92.5°W) from the COADS as analyzed by Dai and Wang (1999, hereafter DW) indicate semidiurnal pressure variations of similar spatially uniform amplitude and phase (∼1 hPa and ∼0950/2150 LT, respectively).

We use the pressure data analyzed by DW, available as seasonal (3 month) and annual mean amplitudes and phases of the diurnal and semidiurnal harmonics on a 2° latitude × 5° longitude grid, and we use the semidiurnal surface pressure formula from HC [Eq. (3)] to compute the semidiurnal variation of the zonal wind component from our Eq. (2). The pressure harmonics analyzed by DW were first smoothed along the latitude and longitude directions using a 4-point boxcar average. The smoothed pressure harmonics were then used to reconstruct the daily pressure marches at each grid point and the zonal pressure gradients centered at the TAO buoy locations were computed (e.g., the zonal pressure gradient for the buoys at 165°E was approximated by the pressure difference between DW pressure data at 167.5° and 162.5°E). Two neighboring pressure gradients in the longitude direction were averaged for buoys situated at the midpoints of the DW pressure grid (e.g., pressure gradients computed at 1°N and 1°S were averaged to obtain the pressure gradient for the buoys along the equator). All available meridional wind data, including those of buoys with <50% data coverage (Table 1), were used to compute the Coriolis term in Eq. (2).

The annual mean semidiurnal variability of the observed zonal wind averaged over all buoys in the tropical Pacific agrees well with those derived from the two pressure data sources, although the observed amplitude is smaller than the pressure-derived amplitudes (0.14 versus 0.20 m s⁻¹; Fig. 8a). Differences in the amplitudes between the observed and pressure-derived zonal winds may be due to neglected effects of surface friction and nonlinear terms in the momentum balance as well as measurement errors. Comparisons of the seasonal averages of the observed and pressure-derived winds are similar to the annual mean results (not shown). We note that the lack of seasonal variation in the semidiurnal zonal wind and pressure harmonics is consistent with atmospheric thermal tidal theory (Chapman and Lindzen 1970; HC).

Next we examine whether the reduction in amplitude of the semidiurnal zonal wind harmonics along 125° and 110°W at, and south of, the equator is reflected in the pressure field. The semidiurnal zonal wind amplitude derived from the pressure data of DW in the region (1°N–5°S, 127.5°–107.5°W) is the same as the basinwide average (0.20 m s⁻¹) and is thus considerably larger than observed (0.07 m s⁻¹; Fig. 8b). Note that regional pressure information is not available from HC. The spatial coverage of the TAO buoy pressure data (2°N–2°S along 110°W and 8°N–8°S along 95°W) is too limited to allow a meaningful assessment of the semidiurnal zonal pressure gradient force. However, the observed semidiurnal pressure harmonic field (Fig. 9) shows no evidence of a reduction in amplitude at the locations where the semidiurnal zonal wind harmonic is suppressed, except for the buoy at 2°S, 110°W in December–May (however, this estimate is based on only 16% data coverage during 2000–04; Table 1). Indeed, the semidiurnal pressure harmonics from the TAO buoys are highly uniform in both amplitude and phase (∼1.4 hPa and ∼1010/2210 LT, respectively) for annual and semiannual means (Fig. 9).

Since it appears that the zonal pressure gradient force cannot account for the reduction in semidiurnal zonal wind amplitude over the cold surface waters in the eastern Pacific, as an alternative mechanism we explore a possible role of enhanced boundary layer stability that may attenuate the surface wind amplitude by inhibiting the vertical mixing of momentum. Boundary layer stability has been shown to affect the strength of near-surface winds in the eastern equatorial Pacific on a variety of time scales from annual to subseasonal (Wallace et al. 1989; Chelton et al. 2001; Xie 2004). Clearly, a full investigation of the role of boundary layer stability and associated boundary layer depth requires observations of the vertical structure of the semidiurnal zonal wind cycle, information not currently available over the tropical Pacific Ocean. However, as a first step, we analyzed the spatial structure of atmospheric boundary layer stability as estimated from the difference between long-term daily mean SST and air temperature from the TAO buoys.

SSTs exceed near-surface air temperatures over the entire domain in the annual and semiannual means, reflecting that the tropical marine boundary layer is generally unstable (Fig. 10). However, the magnitude of the SST–air temperature difference is greatly diminished over the cold surface waters of the eastern equatorial upwelling zone and southeast tropical Pacific, approaching near-zero values in the cold season (June–November; Fig. 10c). This reduction of boundary layer instability (or relative enhancement of boundary layer stability) is consistent with the prevalence of stratiform low clouds at, and south of, the equator in the eastern Pacific (Deser et al. 1993; Norris 1998). The annual and semiannual distributions of mean boundary layer stability (Fig. 10) compare well with those of semidiurnal zonal wind amplitude (Figs. 2b and 5c,d): their pattern correlation coefficients are in the range of 0.7–0.8. In particular, the region of enhanced stability over the eastern Pacific is associated with a reduction in semidiurnal amplitude. This correspondence is confirmed by a scatterplot between seasonal (3 month) mean boundary layer stability and semidiurnal zonal wind amplitude, which reveals a strong positive relationship for SST–air temperature differences <0.6°C (e.g., buoys in the eastern Pacific; Fig. 11). Whether the similarity between the spatial patterns of mean boundary layer stability and semidiurnal zonal wind harmonic amplitude is causal or incidental remains to be seen.

One source of information on the vertical structure of the daily wind cycle is the lower-tropospheric wind profiler at Isla San Cristóbal in the Galápagos Islands (0.90°S, 89.61°W), which provides half-hourly wind data between 300 and 2600 m. Using these data, HD found that the daily cycle of winds below ∼500 m is decoupled from that aloft during periods of low (<23°C) SST. However, contrary to our findings, higher-amplitude semidiurnal zonal wind signals were observed near the surface during periods of increased boundary layer stability, and these were speculated to be due to water vapor tidal forcing in the lowest 500 m (HD). Further work is needed to reconcile the different dependencies of the semidiurnal zonal wind amplitudes on boundary layer stability in the Galápagos profiler and TAO buoy data.

b. Diurnal variability

The spatial pattern of the diurnal meridional wind harmonics is more complex than that of the semidiurnal zonal wind harmonics. Diurnal meridional wind variations tend to be out of phase north and south of the equator (southerly wind anomaly maxima ∼0700 LT at 5°N and ∼1900 LT at 5°S), while on the equator in the eastern Pacific, southerly wind anomaly maxima occur around noon, particularly during the cold season (June–November; Fig. 6b). We note that the noon phase and seasonal dependence of the amplitude of the diurnal meridional wind variations in the eastern equatorial Pacific are consistent with the Galápagos wind profiler results of HD.

Following the same method used to analyze the semidiurnal zonal wind variability, we compare the observed diurnal meridional wind variability with that derived from the diurnal pressure harmonics of DW (smoothed as described in section 4a) based on the momentum balance for the meridional wind component:

View Expanded

The annual mean spatial patterns of the observed and pressure-derived diurnal meridional wind harmonics are qualitatively similar, with opposite phases in the Northern and Southern Hemispheres of the western Pacific (∼0600–0900 and 1800–2100 LT, respectively) and approximately noon phases at the equator in the eastern Pacific (Figs. 3a and 12). The annual mean amplitudes tend to be 15% larger for the pressure-derived harmonics than the observed harmonics. Semiannual mean patterns of the pressure-derived diurnal meridional wind harmonics are also largely consistent with those of the TAO winds (not shown). The good agreement between the observed and pressure-derived wind harmonics indicate that the pressure gradient force plays an important, if not dominant, role in determining the diurnal cycle of the meridional winds over the tropical Pacific, while other factors such as boundary layer stability may be of secondary importance.

Hourly wind vector snapshots representing departures from the daily mean winds, reconstructed from the amplitudes and phases of the diurnal zonal and meridional wind harmonics, are shown at 3-h intervals from 0100 to 1300 LT for the June–November season in Fig. 13 (results based on annual and December–May means exhibit similar patterns with generally smaller amplitudes; not shown). These snapshots confirm the remarkable asymmetry of the diurnal meridional wind component about the equator, especially in the western Pacific. This pattern results in enhanced equatorial divergence ∼0700 LT (convergence ∼1900 LT). There is also an indication of a diurnal cycle in surface wind divergence along the southern border of the ITCZ that is out of phase with the one on the equator, although the spatial coverage of the TAO array is not sufficient to fully delineate this feature.

The diurnal cycles of wind divergence relative to the daily mean along the near-ITCZ (8°–5°N) and near-equatorial (0°–2°S; similar results are obtained for 2°N–2°S) bands for December–May and June–November are shown in Fig. 14. Solid lines denote zonal averages across the tropical Pacific (165°E–95°W) and dashed lines denote averages for the western basin (165°E–155°W). All available wind data within the domains, including those of buoys with <50% data coverage (Table 1), were used to compute the divergence. The diurnal cycles in the near-ITCZ and near-equatorial bands are approximately out of phase in both seasons and longitudinal domains, with maximum equatorial divergence (ITCZ convergence) at 0700–0900 LT, consistent with Fig. 13. In addition, they are slightly larger in amplitude and earlier in phase over the western basin than the full tropical Pacific, also consistent with the wind vectors in Fig. 13. In the western Pacific, the diurnal amplitude of the near-equatorial divergence is 7.1 × 10⁻⁷ s⁻¹ in December–May, similar to the value for June–November (8.0 × 10⁻⁷ s⁻¹), while that of the near-ITCZ divergence is 6.2 × 10⁻⁷ s⁻¹ in December–May, larger than the value for June–November (3.6 × 10⁻⁷ s⁻¹). We speculate that the smaller diurnal amplitude of the near-ITCZ divergence in June–November may be due to the more northerly position of the mean ITCZ in that season such that the TAO buoy domain does not fully capture the associated low-level convergence (recall Fig. 1).

The out-of-phase relationship between the diurnal cycles of surface wind divergence for the near-ITCZ and near-equatorial regions suggests that the two may be linked through a diurnally oscillating meridional overturning circulation, at least in the western Pacific where the diurnal wind harmonics exhibit a relatively simple pattern (see also DS). Although it remains to be seen how this diurnal surface wind divergence pattern relates to diurnal precipitation changes over the tropical Pacific, the early morning maximum in rainfall reported in numerous studies (Gray and Jacobson 1977; Hendon and Woodberry 1993; Janowiak et al. 1994; Serra and McPhaden 2004; Bowman et al. 2005; Yang and Smith 2006) appears to be broadly consistent with the morning maximum in surface wind convergence into the western Pacific ITCZ. Furthermore, a simple estimate of the magnitude of the diurnal rainfall change expected from the amplitude of the diurnal surface wind divergence compares favorably with the observed amplitude of diurnal rainfall variations (DS). However, a joint analysis of diurnal precipitation and wind variability is needed to link the two processes more definitively.

As we have shown, the pattern of diurnal meridional wind variations from the TAO buoys is generally consistent with the diurnal pressure gradient force. The zonally symmetric portion of the pressure gradient force that results in out-of-phase meridional wind variations north and south of the equator appears to be consistent with the structure of the diurnal atmospheric thermal tide (Braswell and Lindzen 1998; DW). The diurnal component of sea level pressure variability over the eastern Pacific cold tongue that drives the noon equatorial southerly wind maximum may result from the diurnal cycle of SST and associated atmospheric boundary layer specific humidity. We note that this mechanism would account for the confinement of the noon southerly wind maximum below ∼500 m during periods of low SST at the Galápagos Islands as documented by HD. Further work is needed to explore these mechanisms more fully.