Analyses of ocean–atmosphere data from Tropical Oceans Global Atmosphere Coupled Ocean–Atmosphere Response Experiment indicate that short-term (weekly to monthly) fluctuations of SST in the western Pacific warm pool are closely linked to the alternation of wet and dry spells driven by the Madden–Julian oscillation (MJO). The dry phase is characterized by increased convection over the Indian Ocean, a prolonged period of atmospheric subsidence, and surface easterlies over the western Pacific warm pool. During this phase, increased surface shortwave radiation and reduced evaporation contribute about equally to the warming of the warm pool. Pronounced diurnal variations in SST observed during the dry phase may be instrumental in leading to the prolonged warming. The dry phase is followed by the wet phase, in which the SST warming trend is arrested and a cooling trend initiated by a reduction in surface shortwave radiation accompanying the buildup of organized convection. Subsequently, the continued cooling of the upper ocean is accelerated by increased westerly surface wind leading to enhanced surface evaporation and increased entrainment of cold water from below the thermocline. At this stage, the increased surface shortwave radiation due to the diminished cloud cover from reduced convection opposes the cooling by evaporation. The cooling trend is reversed as soon as the westerly phase terminates and the dry phase is reinitiated by the establishment of new organized convection over the Indian Ocean.
The authors’ results suggest that short-term SST variability in the western Pacific warm pool is closely linked to surface fluxes, which are strongly modulated by atmospheric low-frequency variability associated with the MJO. The implications of the present results on the dynamics of the MJO and the possible role of coupled SST in influencing the MJO variability are also discussed.
It has long been recognized that large-scale tropical convection tends to organize over sea surface temperature (SST) exceeding an apparent threshold of 27°–28°C and that tropical SST seldom exceeds 30°–31°C. Recently, Waliser and Graham (1993) noticed that while there is marked increase in deep convection for SST >27.5°C, a reduction of convection occurs for SST >29.5°C. Waliser (1996) suggested that the reduction in convection is related to the formation of “hot spots” (defined as limited regions in the Pacific warm pool where SST >30°C), which are associated with atmospheric subsidence. More recently, Lau et al. (1997, submitted to J. Climate) demonstrated that the aforementioned SST thresholds have no fundamental microphysical nor thermodynamic significance, except that they represent different equilibrium circulation/convective states in a coupled tropical ocean–atmosphere system. They point out that the threshold SSTs of the western Pacific warm pool are strong functions of the large-scale circulation generated by differential heating, including that from regions outside of the warm pool (cf. Webster 1994). Clearly, on synoptic and diurnal timescales, SSTs >30°C do occur, presumably in the so-called hot spots (Waliser 1996). If these hot spots persist for a longer period, they may affect the statistics of the long-term mean and, therefore, may have important climatic consequences. In this paper, we use the Tropical Oceans Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) data, which include SST, convection, and surface fluxes with high temporal resolutions, to study the mechanisms of short-term (weekly to monthly) SST regulation and explore its possible climatic impacts.
2. Data description
This study is based on observations taken during the COARE Intensive Observation Period (1 November 1992–28 February 1993). These include
6-h total columnar water vapor integrated from 1000 to 100 mb obtained from upper-air soundings from four Integrated Sounding System (ISS) sites: R/V Shiyan #3, R/V Kexue #1, Kapingamarangi, and Kavieng;
hourly surface meteorological measurements (wind, air temperature, and humidity) and oceanographic measurements (temperature at 0.45 m below sea surface) from the Improved Meteorological Surface Mooring (IMET) at the center of the Intensive Flux Area (IFA,) centered at 1.75°S, 156°E (Weller and Anderson 1996);
hourly subsurface temperature measurements by, TOGA Tropical Ocean–Atmosphere (TOA) mooring at 2°S, 156°E and special SEACAT (conductivity and temperature recorders) measurements, which are merged with the TOA data to obtain mixed layer-depth information; and
net surface radiative fluxes at the IMET site, courtesy of R. Weller and S. Anderson of Woods Hole Oceanographic Institution, and net surface latent and sensible heat fluxes computed by Sui et al. (1997a).
In addition, we use the Japanese Geostationary Meteorological Satellite (GMS) infrared-equivalent blackbody temperature (Tbb) as a proxy for deep convection in the region 20°N–20°S, 80°E–160°W. The temporal and spatial resolution of the Tbb data is hourly and 0.1° × 0.1° latitude–longitude, respectively.
a. The large-scale environment
During TOGA COARE, the organization of large-scale convection over the warm pool was modulated by a wide spectrum of atmospheric circulation conditions. The most dominant of these included the diurnal cycle, the two-day waves, the quasi-biweekly cycles, and the Madden–Julian (30–60 day) oscillation (MJO). These features have been reported in numerous previous papers (e.g., Takayabu et al. 1996; Chen et al. 1996). They are evident in Fig. 1, which shows the time–longitude section of the GMS Tbb data for the eastern Indian Ocean–western Pacific region along 2°S. The diurnal variations can be seen as stationary pulsations of Tbb anchored to the large islands of New Guinea (125°E), southern Borneo (120°E), and Sumatra (115°E). The mechanism of diurnal variations has been studied in Sui et al. (1997b). The two-day waves were made manifest as westward propagating features embedded within an eastward propagating envelope of supercloud clusters associated with the MJO. Two MJOs propagating from the Indian Ocean to the central Pacific were identified from December 1992 to the end of February 1993 (see Lau et al. 1996; Gutzler et al. 1994; McBride et al. 1995). The propagation characteristics are consistent with the dispersion relationship of equatorially trapped Kelvin waves and mixed Rossby–gravity waves (Takayabu et al. 1996; Lau et al. 1989). Notice that while the large-scale SST over the region remained above 28°C at all times, deep convection over a specific region did not occur continuously. In some regions within the warm pool, such as the IFA, deep convection occurred less than 30% of the time, while that in the Indian Ocean and in the vicinity of the date line occurred more than 50% of the time. For the entire COARE period, more than 50% of the regions shown in Fig. 1 were without deep convection. This is because the establishment of deep convection in one region will suppress convection in other regions through the setting up of a large-scale mass recirculation over the warm pool itself.
The coupling between organized convection and the large-scale circulation can be seen in Fig. 2, which shows the time series of the Tbb and the zonal wind component in the lower troposphere (averaged from 1000 to 700 mb), and the total columnar water vapor content obtained from the ISSs at the IFA. High correlations exist among the three variables. While the correlation of the above indexes with rainfall is high, it is difficult to define the boundaries of the wet and dry phases without some uncertainties. In the following analyses, the wet and dry phases are defined with respect to the low-frequency evolution of the total water vapor within the IFA (see upper panel of Fig. 2 and Sui et al. 1997a). Because of the large variability of tropical convection, it can be seen from Fig. 2 that short-period convective outbursts may occur within the dry period and dry lulls may occur within the wet period. Figure 2 shows that the modulation of convection by the two-day waves, the quasi-biweekly cycle, and the 30–60-day oscillation is quite obvious. The two-day waves were much more pronounced during the wet phase. During the main MJO events (1 December 1992 through 28 February 1993), strong surface westerly winds appeared to lag the wet phase by about a quarter of a cycle. This is dynamically consistent with the eastward propagation of the supercloud cluster envelopes and the low-level westerly inflow into the convective region (Lau et al. 1989). The genesis of the westerly wind over the IFA during TOGA COARE has been attributed to multiple factors, including the propagation of the MJO from the Indian Ocean into the region, the onset of the austral summer monsoon, lateral meridional flux from the east Asian continent, and the influence of anomalous convective activity over the central Pacific due to lingering warm SST from the 1991–92 ENSO. For details the readers should refer to documentation of the large-scale circulation during TOGA COARE (e.g., Lau et al. 1996).
b. SST distribution
The distribution of hourly and daily SST during the dry and wet phases of the MJO are shown in Figs. 3a and b, respectively. The former is dominated by the diurnal cycle, and the latter is more representative of persistent conditions. During the wet phase, regardless of whether daily or hourly values are used, the mean SST is sharply peaked at 29°C, with a Gaussian-like distribution. During the dry phase, there is a broadening of both distributions, with a noticeable increase in occurrence of hot spots with SST >29.5°–30°C. In the hourly data, there is also a secondary increase in SST below 28.4°C in the dry phase, indicating that the diurnal cycle can lead to both warming and cooling. The increased occurrences of cooler surface water may also be due to increased surface evaporation induced by the strong westerly wind arising from the phase lag between the SST response and the wet and dry phases. Examination of the individual events (not shown) indicates that the diurnal SST cycle is highly asymmetric, with the warming phase occurring much faster and reaching higher amplitude than the cooling phase. As a result, the diurnal cycle may produce a residual warming that leads to a persistent warming trend in the upper ocean. This is evident in the daily distribution (Fig. 3b), which clearly shows the warming trend in the dry phase. In the daily distribution, the secondary cooling peak at 28.4°C found in the hourly distribution is not present. As noted before, the slightly increased occurrence of lower SST (<28°C) during the dry phase is probably due to the phase lag between SST response and the surface forcings that are operative during the wet and dry phases.
c. Modulation of surface fluxes
The modulation of atmospheric convection by the MJO is responsible for the alternation of disturbed (wet) and undisturbed (dry) phases for tropical large-scale convection. Similar modulations can be seen in the surface fluxes, SST, and mixed layer depth. Since the sensible heat and longwave radiation fluxes are small compared to the shortwave (SW) and latent heat (LE) fluxes, the following discussion on fluxes is focused only on the latter two quantities. The changes in daily surface fluxes and mixed layer depth as functions of SST are shown in Figs. 4a–c. The scatterplots are constructed using data for the entire period. Here, the mixed layer depth is defined as the depth at which the sea water temperature is 0.2°C lower than the surface temperature.
A quasi-linear relationship between SW and SST is quite apparent in Fig. 4a, with SW reduced to about 150–180 W m−2 for SST <29°C but exceeding 220 W m−2 for hot spots (SST >29.5°C). Note that although SW is strongly tied to organized convection that is modulated by the MJO, a better separation of the SST–SW relationship is found when it is categorized according to the wind regime than when it is categorized according to the dry/wet phases. This is also true for the SST–LE relationship. The easterly and westerly phases of the MJO are identified by different symbols in Fig. 4. It is evident that the westerly (easterly) phase is associated with reduced (enhanced) shortwave radiation and strong (weak) evaporation (Figs. 4b,c). This is because of the phase lag between the SST response and the daily mean surface fluxes, which causes the SST–SW and SST–LE relationships to be more directly related to the variation of zonal winds than to the wet/dry phases.
The formation of hot spots in SST is accelerated by the formation of a very shallow mixed layer (less than 5 m) on top of the main thermocline (Fig. 4c). On the other hand, lower SST is associated with a much deeper mixed layer caused by strong surface wind, which occurs mostly in the westerly phase. The deep mixed layer tends to slow down the cooling by surface processes. The surface fluxes and the mixed layer depth all show much larger variability in the westerly phase than in the easterly.
The tight coupling between the tropical atmosphere and the ocean through the surface fluxes can be seen in Fig. 5, which shows the evolution of the SST, surface shortwave, latent heat, and the total fluxes, as recorded by the IMET buoy for the period 11 November 1992 to 20 January 1993. The numbers shown in Fig. 5 are the mean values for that period. The timing of dry/wet phases and westerly wind regimes is also identified for reference. The most conspicuous signals in the SST time series are the diurnal cycles, which appear as sawtooth features during the warming periods from 11 November to 5 December and 5 to 15 January. The amplitude of the diurnal cycle was approximately 1°–2°C on average but can be as high as 3°C on individual days. The SST rise coincided very well with the dry phase and SST >29.5°C being sandwiched between the two westerly phases. In addition, the increased total net downward flux at the surface coincided very well with the increased amplitude of the diurnal cycle during the warming phase of the SST. During the warming phase, the shortwave radiation appeared to dominate the surface energy input. The SST rise appeared to be arrested by the onset of the wet phase associated with the arrival of the rising branch of the MJO and accompanying supercloud clusters. These gave rise to a large reduction in surface shortwave radiation from cloud shielding and a concomitant increase in surface evaporation from increased surface wind. During the cooling phase, the diurnal variation in SST was very small and the shortwave flux was dominated by synoptic-scale variability. The cooling of the upper ocean was maintained by a continued increase in surface evaporation and a reduction of shortwave radiation reaching the surface. The cessation of the cooling was timed to the termination of the westerly phase and the reemerging of the dry conditions as the convection associated with the MJO dissipated over the central western Pacific, while new convection was being built up over the Indian Ocean (see discussion for Fig. 1).
4. Summary discussions
The role of the large-scale circulation, and local and remote processes in the atmosphere and ocean that regulated the local SST over the warm pool during TOGA COARE is illustrated in the schematic shown in Fig. 6. The estimates for the fluxes are based on a composite of the dry, wet, and westerly phases only for the period shown in Fig. 5. The dry and wet phases represent the two basic states of the warm pool. The westerly phase is a transitory phase from the wet to the dry period. The dry phases are initiated from convection over the Indian Ocean when the prevailing surface wind over the warm pool is easterly. As indicated in Fig. 6 (upper panel), the dry phase over the warm pool is connected by Walker-type circulation, with rising motion over the Indian Ocean and sinking motion over the warm pool region, as a part of the large-scale circulation associated with the MJO. The sinking motion over the IFA is likely to increase the static stability and inhibits the formation of deep convection. During this period, the total SW (= 214 W m−2) dominates the surface energy input to the ocean surface and leads to a warming trend in the upper ocean. The shortwave anomaly ΔSW is 22 W m−2 above the mean for the period, and the latent heat flux anomaly ΔLE is 20 W m−2 below the mean for the period. The two processes reinforce each other in producing a large net warming effect on the upper ocean. The rate of warming is also accelerated by the shoaling of the mixed layer. A prolonged dry phase will likely lead to the formation of hot spots over the warm pool (Waliser 1996). The diurnal cycle in surface temperature is very pronounced during this period and, as discussed earlier, may be instrumental in producing the prolonged warming.
Typically, the dry spells do not last more than about a month before the rising branch of the MJO, heralded by supercloud clusters, moves in from the Indian Ocean and thereby establishes the wet phase. The warming trend in SST and the shoaling trend in ocean mixed layer depth are reversed, as a result of a reduction in shortwave radiation from cloud shielding and an increase in the evaporative fluxes. The latter is due to enhanced surface wind associated with the enhanced atmospheric turbulence accompanying the buildup of convection. Comparing the magnitudes of anomaly (ΔSW = −14 W m−2 and ΔLE = 4 W m−2), it may be inferred that the cloud shielding of shortwave radiation dominates the cooling process during the wet phase. In addition, ocean mixed layer processes are likely to play a key role in determining the rate of upper-ocean cooling during this period. Recent studies indicated that the increase in precipitation during TOGA COARE can lead to the formation of a freshwater lens overlying the oceanic thermal mixed layer, which may delay the cooling of the ocean surface by restricting vertical mixing of cold water from below the thermocline (e.g., Anderson et al. 1996).
As the MJO moves further eastward into the central Pacific, the surface westerly wind prevails over the warm pool, where deep convection is reduced substantially due to subsidence generated by the departing supercluster. The warm pool enters the transition phase. As long as the convection remains over the central Pacific, the westerly inflow into the convective region remains strong. This phase corresponds to large variability in atmospheric circulation, including the development of westerly wind bursts and tropical cyclones over the warm pool (Lau et al. 1996). Here, surface evaporation contributes predominantly to the cooling of the upper ocean. Relative to the mean condition, the increase in surface evaporation ΔLE is 62 W m−2. On the other hand, shortwave radiation contributes to the warming, with ΔSW = 20 W m−2, because of the increased occurrence of clear-sky conditions associated with reduced convection. At this stage, the surface shortwave radiation opposes the cooling. There is little diurnal variability associated with the cooling of the upper ocean. The cooling is likely to be accelerated by the increased entrainment of cold water from below the thermocline due to increased wind stirring. The concomitant deepening of the oceanic mixed layer will slow down the cooling by surface fluxes. As soon as the westerly wind regime terminates, the cooling trend is arrested. This is generally accompanied by the dissipation of convection near the date line and the reappearance of convection over the Indian Ocean—that is, a return to the dry and surface easterly wind conditions. During the TOGA COARE period, the above cycle was repeated twice for the entire atmosphere–ocean system over the Indian Ocean–western Pacific in association of the two major MJOs that formed and propagated across the above region.
The present results may have important implications regarding the mechanism for the formation and evolution of the MJO. The eastward propagating organized convection and its embedding westward propagating features, as shown in Fig. 1, are consistent with wave-CISK (conditional instability of the second kind)-driven coupled Kelvin–Rossby modes (Lau and Peng 1987; Wang 1988). The preferred spatial and temporal scales of westward propagating features also lend support to recent theories that suggest that high-frequency inertial gravity waves may be needed to account for the hierarchical structures and asymmetric east–west motions associated with the MJO (e.g., Takayabu et al. 1996; Chao and Lin 1994). However, the present findings do not support theories of evaporative wind feedback for the MJO, which predict maximum enhanced evaporation to the east of convection (e.g., Neelin et al. 1987; Emanuel 1987). As shown in Fig. 6, it is found that evaporation is reduced to the east of convection and that the increase in evaporation occurs mostly to the west of the convection after passage of the ascending branch of the MJO.
The present results also suggest a possible role of short-term SST variations in amplifying the MJO. The rising SST in the western Pacific caused by shortwave radiation during the dry phase, in advance of the rising branch of the MJO (see Fig. 6, upper panel), provides a condition that increases available moisture energy in the lower troposphere. This is so because the relative humidity within the boundary layer is likely to remain approximately constant during both the wet and dry phases, due to strong thermodynamic coupling between the boundary layer and the sea surface. However, while the moist available energy keeps increasing, actual convection is suppressed by the subsidence associated with convection to the west, creating a bottom-up and top-down situation akin to that of a “pressure cooker” over the western Pacific. As soon as the MJO arrives, the subsidence motion is replaced by rising motion in a way similar to breaking off of the mechanism that keeps the lid of the pressure cooker in place. The release will result in explosive growth of deep convection, “blowing off the lid.” As convection grows, the cooling phase begins in the manner described previously. The cooling phase, similar to the warming phase, will then precondition the ocean–atmosphere system, but in an opposite sense, preventing the further development of convection. According to the above scenario, the MJO fluctuations will be amplified in the presence of SST coupling.
5. Concluding remarks
Based on our results, we conclude that fluctuations of local SST on timescales of weeks to months over the warm pool are part of a large-scale dynamic and thermodynamic adjustment to inherent atmospheric low-frequency variability associated with the MJO. On these timescales, both surface processes (shortwave radiation and evaporation) and upper-ocean processes contribute to the regulation of SST of the warm pool. Additionally, our results show that the diurnal SST cycle may play an important role during the warming phase of SST but not during the cooling phases of SST. This is consistent with recent numerical experiments indicating that the diurnal forcing is essential in simulating the low-frequency oceanic response during TOGA COARE (Sui et al. 1997a).
While our results emphasize the importance of the MJO in regulating short-term SST through induced surface fluxes, we need to point out that the organization of tropical convection, including that associated with the MJO, is controlled by the climatological distribution of SST, as well as its interannual variations. As discussed in the previous section, our results also suggest a possible role that short-term SST fluctuations may play in influencing the MJO. Hence the MJO–SST link is part of a complex multiscale coupling among convection and surface fluxes, as well as ocean and atmospheric dynamics. The role of local air–sea interaction in modifying the MJO over the warm pool is unclear and needs to be further investigated.
The authors are indebted to Dr. R. Lukas for providing the SEACAT data, and Drs. R. Weller and S. Anderson for providing the IMET data. This work is supported under the EOS Hydrologic Processes and Climate Interdisciplinary Investigations and the Tropical Rainfall Measuring Mission of NASA’s Mission to Planet Earth Office.
Corresponding author address: Dr. William K.-M. Lau, Climate and Radiation Branch/Code 913, NASA Goddard Space Flight Center, Greenbelt, MD 20771.