Journal of Geophysical Research: Oceans

Seasonal heat budget in the mixed layer of the southeastern tropical Indian Ocean in a high-resolution ocean general circulation model

Authors


Abstract

[1] Using results from a high-resolution ocean general circulation model, this study identifies the impact of ocean dynamics on the mean seasonal cycle of sea surface temperature (SST) in the southeastern tropical Indian Ocean. An important question about the mechanisms of regional SST variation arises in a quick look at satellite-observed SST and ocean color. Although ocean color indicates a distinctive upwelling season from July to September, the SST depression off Java and Sumatra is small, forming a sharp contrast with other eastern boundary regions. The model results suggest an explanation of the process responsible for the small SST depression. Our analysis indicates that there are three dynamically different regimes within the region. First, near the coast of northwest Australia, surface heat flux controls the seasonal variation of SST, while horizontal advection and vertical entrainment are relatively weak. This result is consistent with previous studies. Second, south of Java and farther to the east, warm horizontal advection of the Indonesian throughflow (ITF) neutralizes the cold upwelling. The transport of the ITF, especially the outflow from the Lombok Strait, reaches its seasonal maximum in July–September, at the same time that the maximum upwelling occurs. Third, west of Sumatra, a thick barrier layer exists, which impedes the cold thermocline water from entering the mixed layer (ML). Although upwelling occurs, it has no significant effect on the SST as a result of the small vertical temperature gradient at the bottom of the ML.

1. Introduction

[2] The southeastern tropical Indian Ocean (STIO) has been highlighted as an area of ocean-atmosphere interaction by the discovery of the dipole/zonal mode [Webster et al., 1999; Saji et al., 1999]. The variability of regional sea surface temperature (SST), especially that in the region off Java and Sumatra, has a strong influence on both regional and global climate [Clark et al., 2003; Saji and Yamagata, 2003]. The emergence of this region as a factor in global climate variation is the motivation for a study of the regional heat budget of the surface layer.

[3] In contrast to other eastern boundary regions, we do not find an SST depression off Java and Sumatra during the season of upwelling favorable winds: the southeast monsoon from June to September [Wyrtki, 1962]. The high-resolution climatology of AVHRR satellite SST, for example, does not show a depression of any more than a few tenths degree off Java in July–September (Figure 1a), which forms a sharp contrast with the situation off Peru in the Pacific (Figure 1b), where the SST depression within 500 km of the coast exceeds 5°C and the advective effect of upwelling extends thousands of kilometer offshore.

Figure 1.

Advanced Very High Resolution Radiometer (AVHRR) Oceans Pathfinder sea surface temperature (SST) (°C) and Special Sensor Microwave Imager (SSM/I) 10 m wind velocity (m/s) off (a) Indonesia and (b) South America in July–September. AVHRR SST is averaged from 1985 to 1999 with 1/12° resolution. SSM/I wind is averaged from 1987 to 1999 with 1° resolution.

[4] The small depression of SST off Java and Sumatra is also evident in the World Ocean Atlas 2001. Nevertheless, there is good reason to believe that upwelling occurs in this region. Satellite ocean color (Figure 2) [Hendiartia et al., 2004] indicates high values of chlorophyll concentration within 200–500 km of the coast. Below the surface, the depression of temperature toward the coast becomes much larger within the depth range of the thermocline. Thus it seems that the southeast Monsoon of southern winter (Figure 1a) generates an offshore Ekman transport near the sea surface, which produces an upwelling off Java and Sumatra [Wyrtki, 1962; Susanto et al., 2001]. The upward movement brings high-nutrient water from the thermocline into the euphotic layer where phytoplankton develops under increasing light penetration. In the mean, upwelling off Java and Sumatra starts in June and continues to exist until September. As part of the westward Ekman current, high Chlorophyll concentration water permeates from the coast to about 98°E. However, why does the upwelling not appear in the pattern of SST, as it does in other eastern boundary regions?

Figure 2.

Monthly Sea-viewing Wide Field-of-view Sensor (SeaWiFs) chlorophyll concentration in July–September averaged from 1997 to 2003 with 0.25° resolution image

[5] To some extent, this question has been addressed in earlier studies. Both internal and external factors can be attributed to the small SST depression. The region is unique, partly because of the existence of the Indonesian Throughflow (ITF) [Gordon, 1986; Wyrtki, 1987]. The ITF transfers a large amount of warm, fresh water from the Pacific to the Indian Ocean, and affects both the regional circulation and thermal structure. Estimates of the annual mean transport of ITF have ranged widely in the past (1.7 Sv to >20 Sv) (see the review by Godfrey [1996]), but recently have settled to a value near 10–12 Sv [Gordon, 2001]. Seasonal and interannual variation of ITF has also been observed [Meyers et al., 1995; Meyers, 1996]. The uniquely warm upper layers of the southeastern tropical Indian Ocean are generally ascribed to the transport of ITF.

[6] A quantitative analysis of the upper layer heat budget using results from an ocean model [Qu et al., 1994] found that the seasonal warming advection by the ITF cancels the cooling by upwelling. Annamalai et al. [2003] pointed out that an interannual reduction in ITF transport helps to trigger an enhanced SST depression in some years. The previous modeling studies were limited by coarse model resolution that was not good enough to resolve narrow passages or straits in the Indonesian archipelago. For example, the outflow from Lombok Strait has been known to be important to the dynamics and thermodynamics in the region [Birol and Morrow, 2001], but it is not well resolved in the previous models [Périgaud and Delecluse, 1992; Qu et al., 1994; Masumoto and Meyers, 1998; Annamalai et al., 2003; Masson et al., 2002, 2004; Song and Gordon, 2004].

[7] In addition to ITF, the barrier layer (BL) is another unique factor influencing the regional SST variability. This intermediate layer, between the bottom of the mixed layer (ML) and the top of the thermocline, is an impediment to the exchange of heat between the ML and the thermocline. In the region off Java and Sumatra, large rainfall and runoff persists all year-round, and thus BL seems to be a permanent phenomenon [Sprintall and Tomczak, 1992; Masson et al., 2002; T. Qu and G. Meyers, Seasonal variation of the barrier layer in the southeastern tropical Indian Ocean, submitted to Journal of Geophysical Research, 2005, hereinafter referred to as Qu and Meyer, submitted manuscript, 2005]. However, until this time, the exact role of the BL in the mixed layer heat budget has not been examined.

[8] This paper is intended to provide a comprehensive description of the mixed layer heat budget and to identify the role of ocean dynamics, especially the ITF, the upwelling, and the BL in determining the seasonal variation of SST off Java and Sumatra. Results from a very high resolution ocean general circulation model (OGCM) will be used for this analysis. The model configuration is described in section 2. In section 3, we validate the model output using available observations and provide a general description of the circulation and thermal structure from the model. The ML and BL are discussed in section 4. In section 5, we identify the role of different processes in the mixed layer heat budget. Effects of horizontal advection and vertical entrainment are assessed in the different regions. Results of this analysis are summarized in section 6.

2. OFES Model Description

[9] Results from an OGCM for the Earth Simulator (OFES) were used for this study [Masumoto et al., 2004]. The computational domain extends from 75°S to 75°N, with horizontal resolution of 0.1°. Vertical resolution varies from 5 m in the upper levels to 330 m near the bottom, with 54 vertical levels. With high vertical resolution from 5 m to 15 m in the upper 200 m, fine structure of the temperature and salinity of the upper layer is well resolved by this model. Up to now, OFES is the highest-resolution global OGCM, well resolving the Indonesian archipelago. Surface momentum, heat and fresh water flux were specified from climatological monthly mean NCEP/NCAR reanalysis data. In addition, sea surface salinity also was restored to the climatological monthly SSS of the World Ocean Atlas 1998 (WOA98). The model spins up for 50 years from 3-D climatological annual mean temperature and salinity of the WOA98 and no motion. The K-profile Parameterization (KPP) scheme was employed for the vertical mixing. We averaged the last 5 years of results to produce a climatological monthly mean data set for this analysis. The study region is limited to between 85°E–130°E and 25°S–5°N.

3. General Characteristics of the Model Output and Comparison With Observations

[10] We expect ITF to play an important role in the surface heat budget and therefore we have validated the model result with observations. The mean seasonal cycle of model ITF above 432 m was evaluated on the IX1 XBT line (Sunda Strait to northwest Australia) (Figure 1a) and compared to the earlier analysis of XBT data [Meyers et al., 1995]. The model has an annual mean transport of 9.0 Sv (1 SV = 106 m3/s) from the Pacific to the Indian Ocean, with a minimum of 6.0 Sv in November and a maximum of 14.4 Sv in July (Figure 3a). The annual mean is similar to the generally accepted value mentioned earlier. The phase of the annual cycle agrees with the estimates from XBTs. The maximum transport estimated from XBT is one month earlier, but, interestingly, this phase difference disappears if model transport above 200 m is used. Most of the volume transport is confined to the upper layers (Figure 3c). Above 200 m, the model ITF has a transport of 8.9 Sv, with a minimum of 5.7 Sv in February and a maximum of 13.4 Sv in August. The annual range of model transport above 432 m is 8.4 Sv and is less than the range estimated from XBTs. Taking into account the contribution of southwestward Ekman transport that is included in the model estimates, and the monthly temporal resolution of the model output, the difference between model and observations is reasonable.

Figure 3.

Transport of ITF across (a) the IX1 XBT line, (b) Timor Sea, Sawu Sea, Lombok Strait, and sum of transport through straits in the upper layer (Sv = 106 m3/s), and (c) their vertical structure of the annual mean. IX1 XBT line is a merchant shipping route from Shark Bay to Sunda Strait, with regularly repeated temperature observation. Positions are shown in Figure 1.

[11] The net flow from the Banda and Flores Seas into the Indonesian Australian Bight (IAB) may also be a factor in the heat budget (and will be relevant to analysis of direct measurements of ITF that have started and will end in 2006). The net transport from top to bottom across the three passages that feed ITF into the Indian Ocean, Timor Sea, Sawu Sea and Lombok Strait, show a mixture of semiannual and annual signals (Figure 3b), as expected in a region with strong monsoonal wind forcing (both local and remote) and reflections of the Wyrtki jets. The phase of maximum transport progresses from east to west with maximum in the Timor Sea in April, in the Sawu Sea in June and in Lombok Strait in August. Added together the three straits give the same phase as the IX1 line. However, Figure 3 shows differences between inflow and outflow from the IAB during some seasons of the year, suggesting that the region may serve as a buffer between the ITF and circulation in the Indian Ocean [Meyers et al., 1995].

[12] Depth of the thermocline in the study region is known to be partly controlled by Rossby and Kelvin wave propagation [Wijffels and Meyers, 2004]. A strong annual cycle Rossby wave originates from the ITF region and travels from about 110°E in January to about 85°E in July [Périgaud and Delecluse, 1992; Masumoto and Meyers, 1998]. The wave appears in both depth of the thermocline and sea surface height (SSH) anomaly [Birol and Morrow, 2001]. The model dynamic height is compared to TOPEX/ERS merged altimetric data in Figure 4. OFES simulates the phase of the Rossby wave and the growth in amplitude [Masumoto and Meyers, 1998] very well, but the strength of the signal is too small by almost a factor of two. The weakness of the signal in the model suggests too much spatial smoothing of the wind stress curl in the NCEP/NCAR reanalysis wind fields. It is worth mentioning that both the model and observations show a high value center near the coast of Java and Sumatra during the northwest monsoon and a low value center during the southeast monsoon (figure not shown).

Figure 4.

Sea surface height anomaly (SSHA) averaged at 11°–14°S from (a) TOPEX/ERS and (b) OFES (cm). TOPEX/ERS SSHA is averaged from 1992 to 2002.

[13] Validation of the subsurface temperature field is also necessary for the heat budget study. Upper layer temperature from the model is compared with the XBT data along IX1 line averaged from 1983 to 2003 [Meyers, 1996]. In general, they show a good agreement (Figure 5). The thickness of the thermocline in OFES matches the observations, unlike many models that have an unnaturally thick thermocline. Both model and observations show a downward slope toward the pole below the middle of the thermocline, indicating the westward currents (ITF and South Equatorial Current (SEC)), and an upward slope in near surface levels, indicating eastward shear (Eastern Gyral Current (EGC)). The surface temperature near Java ranges from 26°C to 28°C in both model and observations. Nevertheless, there are some differences worth noting. The model thermocline is quite a bit tighter than observed in the northern part of IX1 line, suggesting that the model is underestimating the effects of vertical mixing. This could also be due to shortcoming in the model ventilation and subduction much further south in the open ocean. A downward slope of isotherms toward Java in the upper 100 m indicates the eastward flowing shallow South Java Current (SJC), and is stronger in the model than observations. Below 200 m, isotherms slope downward toward the pole to 22°S in the observations, but this slope is weaker or even reversed in the model. With a higher resolution, temperature from the model shows more small-scale features than that from the XBT observations and the model probably captures more of the equatorial and coastal dynamics than the observations. Near Australia toward the southern end of the section, the 24°C isotherm outcrops in winter, whereas the 22°C isotherm outcrops in the model. Near surface thermal stratification does not change much throughout the year near the Sunda Strait, indicative of weak seasonal variation in ML and BL (Qu and Meyers, submitted manuscript, 2005). After comparing to observations, we feel that, despite the weakness noted above, OFES gives one of the best simulations of regional thermal structure yet available and is suitable for an analysis of the upper ocean heat budget.

Figure 5.

Temperature from (right) observation and (left) model along the IX1 XBT line in February and August (°C).

[14] Annual mean model velocity field across IXI shows essentially the same major currents as observations [Meyers et al., 1995], including the SJC, the SEC, the ITF, the EGC, and the Leeuwin Current (LC) (Figure 6). The part of the SEC that carries ITF waters has maximum strength occurring at the surface, as noted by Wijffels et al. [2002], with an axis located at 10°–15°S (Figure 6a). The deep branch of the SEC south of 15°S is a weak feature in the model, compared with the previous observations [Sprintall et al., 2002].

Figure 6.

(a) Annual mean velocity against depth and (b) the seasonal variation at 37.8 m across the IX1 XBT line.

[15] The temporal variation of current at 37.8 m depth (Figure 6b) shows the southward flowing LC persists all year-round. The EGC is a surface-trapped eastward flow, which extends to a depth no more than 150 m in the strongest period from January to May. The axis of the ITF swings about 1.5 degrees of latitude in the meridional direction, from about 10°S in February to about 11.5°S in August. The intensity of ITF at the surface reaches its seasonal maximum in August. Between the main axis of the ITF and Sunda Strait, a semiannual current, SJC, reaches its maxima in May and November, reflecting the strong influence of local rainfall, wind forcing and Kelvin waves from the equatorial region [Quadfasel and Cresswell, 1992]. The ITF and SJC affect the T-S characteristics of water masses and further have influence on the upper thermal structure in the region [Song and Gordon, 2004].

[16] The velocity in the upper layer changes rapidly in response to the annually reversing monsoon winds. Surface currents through the straits of Indonesia show strong variability (Figure 7). In the southern winter, when the South China Sea (SCS) Monsoon blows from the northeast, water surges into the Flores Sea from the SCS across the Sunda Shelf and through shallow Karimata Strait, and eventually enters the Timor Sea. At this time, the surface current from Makassar Strait is weak. When the monsoon reverses, a strong southward current flows through Makassar and Lombok Straits, and enters the Indian Ocean directly, with only a small part flowing into the SCS from the Flores Sea. Water from the Banda and Arafura Seas forms another part of the surface ITF in the southeast monsoon. Most of it flows into the Timor Sea and the rest into the eastern Indian Ocean through the Sawu Sea. The surface ITF originates from the Pacific through different parts of the Indonesian archipelago, which is consistent with the flow inferred from data analyses [e.g., Qu and Meyers, 2005].

Figure 7.

Water temperature (°C) and velocity (cm/s) averaged in the mixed layer in (a) February and (b) August.

[17] Like the thermal structure, the field of currents in OFES is one of the best available in a model. Usually direct observations are not complete and cannot be used for heat budget studies. The model on the other hand is a complete set of heat budget variables and can be used to estimate the dominant mechanisms of the heat budget.

4. Mixed Layer (ML) and Barrier Layer (BL)

[18] Both the ML and BL are important factors influencing the SST. The ML is an upper turbulent layer of the ocean, usually defined as the quasi-homogeneous surface density layer. We calculate the depth of ML (MLD) by specifying a difference in σθ (potential density) from the surface value, which is equivalent to a net increase in temperature [Lukas and Lindstrom, 1991; Sprintall and Tomczak, 1992]. This criterion takes into account both temperature and salinity stratification. When the effect of salinity stratification is negligible in the density profile, the ML is equivalent to the isothermal layer (IL), which is defined as an upper layer with temperature higher than SST − ΔT. When the effect of salinity stratification is positive and not negligible, the IL is deeper than the ML, and a layer between them is known as the barrier layer (BL) [Lukas and Lindstrom, 1991; Sprintall and Tomczak, 1992]. Different ΔTs have been used in literature, ranging from 0.2°C to 1.2°C. Kara et al. [2000] compared different criteria and suggested that ΔT = 0.8°C is probably the most reliable selection. In this study, we compared estimates from different criteria with ΔT ranging from 0.5°C to 1.0°C and found little difference between them. So, we just simply selected ΔT = 0.8°C as Kara et al. [2000] suggested.

[19] The annual mean MLD and barrier layer thickness (BLT) from the model (Figures 8a–8b) are consistent with observations [e.g., Sprintall and Tomczak, 1992; Qu and Meyers, submitted manuscript, 2005]. The MLD is deep (>40 m) offshore along the equator due to the convergence of surface water forced by westerly wind. Near the coast of Sumatra, the MLD shoals to <40 m, and its contours are more oriented with the coastline. On the basis of existing evaporation and precipitation data, Sprintall and Tomczak [1992] and Qu and Meyers (submitted manuscript, 2005) noticed that there is a large fresh water input in this region, which is probably the cause of shallow ML. The salinity stratification forced by fresh water input changes the density structure in the upper layer, and thus produces a shallower ML and a thicker BL (Figures 8a–8b). The thickest BL (>20 m) is in the equatorial region west of Sumatra. Besides the surface fresh water flux, Qu and Meyers (submitted manuscript, 2005) pointed out that northward advection of fresh water associated with the meridional wind stress may also contribute to this BL maximum. The MLD increases southeastward, and as a consequence, the BL completely disappears near the coast of northwest Australia. At the eastern tip of Java, the BL is apparent but thinner than 10 m. A thick BL is also seen at (97.5°E, 22.5°S) in the model, which was documented in the study from Sprintall and Tomczak [1992]. However, so far, no other studies have reported this thick BL [Annamalai et al., 2003; Masson et al., 2002, 2004]. This region is far away from the continent, and the surface fresh water flux is negative (Qu and Meyers, submitted manuscript, 2005), so that the formation of BL does not depend on rainfall and river runoff. Sprintall and Tomczak [1992] related it to the subduction induced by Ekman pumping in the subtropical region. We suggest that the horizontal advection, especially the Ekman drift in the southern winter, transports the surface low-salinity water from the eastern STIO to this area and forms a thick BL there.

Figure 8.

Mixed layer depth (MLD) and barrier layer thickness (BLT) in meters: (top) annual mean, (middle) averaged in July–September, and (bottom) seasonal cycle averaged in three boxes as indicated in Figure 8d. Open circles represent the region off northwest Australia (105°–125°E, 13°–20°S). Open squares represent the region off Java and farther to the east (105°–125°E, 7°–12°S). Solid square represents the region off Sumatra (95°–105°E, 5°S–2°N).

[20] In the upwelling favorable season, when the southeast monsoon prevails, the shallowest ML and thickest BL draw back close to Sumatra (Figures 8c–8d). The outflow of Lombok Strait reaches its seasonal maximum and strongly affects the upper layer thermal structure in its outlet region. As a result, the coastline oriented MLD becomes shallower near the strait. The surface fresh water flux has a maximum at the coast of Sumatra and turns to negative at the coast of Java (Qu and Meyers, submitted manuscript, 2005). The BLT increases gradually northwestward, with a maximum (20 m) near the coast of Sumatra and a minimum around the eastern tip of Java (5 m). The spatial distribution of BLT has almost the same pattern as the MLD.

[21] In the region off northwest Australia, the model shows spotty BL regions, which seem to originate from the Timor Sea. Along the IX1 XBT line, the southern branch of the ITF (Figure 6b) flows westward from March to September at ∼20°S, which transports fresh water from the Pacific and induces the BL in the region off northwest Australia.

[22] The STIO can be divided into three distinct thermodynamic regimes based on the maps of MLD and BLT in Figure 8. The area off Sumatra (95°–105°E, 5°S–2°N) has a shallow ML and a very thick BL. The area off Java and farther to the east (105°–125°E, 7°–12°S) also has a shallow ML, but a relatively thin BL. The region off northwest Australia (105°–125°E, 13°–20°S) has a deep ML and almost no BL.

[23] The seasonal variation of MLD and BLT in these areas are shown in Figures 8e–8f. Near the coast of Sumatra, the MLD has a weak semiannual cycle, consistent with the monsoon reversal and solar radiation cycle (Figure 8e). The BLT also has semiannual variation, superimposed with annual variation in response to the seasonal fresh water flux (Figure 8f). The equatorial downwelling Kelvin waves generated in May–June and November propagate eastward and deepen the thermocline in the region off Sumatra (Figure 9c). Despite the seasonal minimum in surface fresh water flux, the BLT reaches a peak in June, next to its seasonal maximum in November (Qu and Meyers, submitted manuscript, 2005). Near the coast of northwest Australia, the annual variation dominates the seasonal cycle of both MLD and BLT. BLT remains small (<10 m) throughout the year with a weak maximum in the southern winter when the downwelling Rossby waves propagate and deepen the thermocline in the region (Figure 9c). Note that the surface layer loses fresh water to the atmosphere during this period of the year (Qu and Meyers, submitted manuscript, 2005). It is not clear at this time how the surface stratification is maintained with such a negative fresh water flux. We leave this question for a future study.

Figure 9.

Seasonal variation of vertical thermal structure, mixed layer, and isothermal layer in the three boxes selected in Figure 8d: (left) OFES and (right) WOA01. The dotted line represents mixed layer depth, and the dashed line represents isothermal layer depth (m).

[24] At the coast of Java and farther to the east, the superimposed semiannual and annual signals again appear in MLD (Figure 8e). Monsoonal transitions are important forcing mechanism for two minima in MLD, but the effect of surface fresh water flux is not negligible (Qu and Meyers, submitted manuscript, 2005). The maximum of BLT in May is associated with the eastward SJC (Figure 6b), which advects the warm and fresh water from the west where the surface fresh water flux reaches its maximum strength at this season. The signature of Kelvin wave in the thermocline is rather weak and has almost no effect on the MLD and IL depth (ILD) (Figure 9a). In November, despite the similar SJC, the negative surface fresh water flux weakens the upper layer stratification and reduces the thickness of the BL (Qu and Meyers, submitted manuscript, 2005).

[25] We expect the thermodynamics of the surface layer to be different in the three areas due to the differing structure of MLD and BLT. Before identifying the dominant mechanisms of SST variation, we need to ensure that the model reasonably simulates the interaction between the subsurface and surface layers. The model simulates the seasonal cycle of subsurface temperature in the World Ocean Atlas 2001 remarkably well in the three areas (Figure 9), except that the thermoclines are a bit tighter than observed, also shown in the transect along IX1 XBT line (Figure 5). Off Java, the upwelling of 22°–24°C water into the mixed layer is apparent, but only a small cooling of SST results. Off Sumatra the heave of the thermocline, a semiannual signal, is strongest in the middle of the thermocline but it has weaker effect on MLD and ILD, which are primarily maintained by the relatively steady surface forcing. Off northwest Australia the large amplitude of MLD and ILD are clearly forced from the surface and the deepening in winter overtakes a downward heave of the thermocline, implying that winter cooling of SST is dampened by the thermocline variation. The strong similarity in the relationships between MLD, BLT, and thermal structure in the observations and OFES encourages us that the model heat budget may be realistic enough to infer processes that are actually happening in the ocean.

5. Mixed Layer Heat Balance

5.1. Mixed Layer Heat Budget Equation

[26] The mixed layer temperature is a good proxy of SST. The equation governing the mixed layer temperature can be expressed as follows [cf. Qu, 2003]:

equation image

where Tm is the mixed layer temperature, hm is the mixed layer depth, and Td is the temperature of water entrained into the mixed layer, taken to be the temperature at 5 m below MLD. Q0 is the net surface heat flux, and qd is the downward radiative flux across MLD. In computing qd, Paulson and Simpson's [1977] empirical formula is used in the OFES:

equation image

where Qr is the surface radiative heat flux, and the coefficients R, ξ1 and ξ2 are selected to be 0.58, 0.35 and 23, respectively. These selections are well suited for clear upper layer water like that in the southeastern Indian Ocean, despite the fact that spatial-temporal functions may be required to better represent the change of water transparency (Figure 2). Surface radiative heat fluxes from NCEP/NCAR are used to calculate qd. It turns out that qd is generally small compared with Qr, because in most cases MLD is larger than 40 m.

[27] An important term in the heat budget equation is the entrainment rate, went. Following Qu [2003], went can be determined by

equation image

where ∂hm/∂t denotes the rate of the ML deepening, wmb the velocity of water parcel at the base of the ML, and U · ∇hm the horizontal advection of water parcels below the ML.

[28] In the following, the term on the left-hand side of the heat budget equation (1) will be referred to as the temperature tendency (Tt), and the three terms on the right-hand side as the surface thermal facing (Qhm), horizontal advection (Adv), and vertical entrainment (Ent), respectively. All these terms will be calculated by using the model's monthly data, and the relative importance of these terms will shed further light on possible mechanisms that control the seasonal variation of SST.

[29] As noted by Qu et al. [1994], surface heat flux dominates the mixed layer heat budget and is thus the primary mechanism determining the seasonal variation of SST. The correlation coefficient between temperature tendency and surface heat flux is >0.7 throughout the region, with its maximum exceeding 0.9 in much of the area south of 5°S (Figure 10a). A wedge of low correlation extends from the equator to the coasts of Sumatra. If the seasonal variation of MLD is included, the correlation coefficient increases slightly (Figure 10b), but no significant improvement is seen near the coast of Sumatra as expected due to the weak seasonal variation of MLD there (Qu and Meyers, submitted manuscript, 2005). The correlation coefficient significantly increases along the equator and off Sumatra if horizontal advection and vertical entrainment are further included (Figures 10c–10d). This result implies that adding the contribution from ocean dynamics better explains the temperature tendency, particularly in the region off Sumatra. An area of low correlation remains very close to the coast, which may be due to the strong intraseasonal variations in the coastal waveguide (Figure 10d).

Figure 10.

Linear correlation coefficient (a) between temperature tendency (Tt) ∂Tm/∂t and net heat flux (Nhf) Q0qd, (b) between temperature tendency and surface thermal forcing (Qhm) (Q0qd)/ρCphm, (c) between temperature tendency and sum of surface thermal forcing, and horizontal advection (Qhm + Adv), and (d) between temperature tendency and sum of surface thermal forcing, horizontal advection, and vertical entrainment (All). The light shading denotes areas where the correlation coefficient is larger than 0.7.

[30] The mean absolute values of difference (MAVD) are calculated between temperature tendency and other terms of the heat budget equation (1):

equation image

where A = Qhm, Qhm + Adv, or Qhm + Adv + Ent. The results show that the low correlation near the coast of Sumatra is not dynamically significant (Figure 11). The maxima occur near the coasts of Java and northwest Australia. The MAVD is small in the region off Sumatra. Large MAVDs appear in the coastal region off Java and northwest Australia (Figure 11b), where intense intraseasonal variations were observed [Feng and Wijffels, 2002]. Figure 11 also shows that the difference decreases if horizontal advection and vertical entrainment are further included, implying that ocean dynamics contributes to the seasonal variation of SST.

Figure 11.

Absolute difference (a) between temperature tendency (Tt) and surface thermal forcing (Qhm) (Q0qd)/ρCphm, (b) between temperature tendency and sum of surface thermal forcing, horizontal advection, and vertical entrainment (A11), and (c) between temperature tendency and sum of surface thermal forcing and horizontal advection (Qhm + Adv). The light shading denotes areas where the difference is larger than 0.8 (10−7°C/s(=0.26°C/month)).

5.2. Annual Mean Heat Budget in Three Box-Averaged Regions

[31] The results of the annual mean heat budget analysis for the three regions (Java, Sumatra, and northwest Australia) are shown in Table 1. As documented in the previous study [Qu et al., 1994], in the region off Java and farther to the east, horizontal advection by ITF and vertical entrainment by upwelling are by far the most important processes in balancing the annual mean heat budget. The surface thermal forcing is positive but not significant. In the region off northwest Australia, surface thermal forcing is negative, and horizontal advection is the primary heating process. Compared with that near the coast of Java, horizontal current in this region is rather weak (Figure 7b). However, with a larger horizontal SST gradient, horizontal advection is nearly enough to counterbalance the surface cooling and vertical entrainment. It is worthwhile to note that the residual term is relatively large in the region off northwest Australia. The eddy heat fluxes associated with intraseasonal variations may contribute to this large discrepancy (called the residual heat flux hereinafter). Compared with the fixed depth heat budget given by Qu et al. [1994], our analysis has a larger contribution from vertical entrainment. This probably reflects the importance of ML in determining the SST variations. In the region off Sumatra, all heat budget terms are small. Horizontal advection and vertical entrainment are not significant. The positive surface thermal forcing seems to be counterbalanced mainly by the residual heat flux, implying the importance of mesoscale eddies/subgrid processes in this region.

Table 1. Annual Mean Heat Budget in Three Regionsa
 QhmAdvEntRe
  • a

    Abbreviations are as follows: Qhm, surface thermal forcing; Adv, horizontal advection; Ent, vertical entrainment; Re, residual heat flux. Unit: 10−9°C/s.

Java13.091.2−108.03.9
Sumatra9.11.6−3.1−7.6
NW Australia−25.247.0−52.230.4

[32] In addition, the monthly averaging of model output may result in oversmoothing of MLD distribution, and partially accounts for the large residual heat fluxes described above. For example, when the MLD is shallow during the transition season of monsoon in the region off Sumatra, the mixed layer heat budget becomes more sensitive to the synoptic and diurnal variations (Figure 8e). Oversmoothing of MLD distribution at this particular season may reduce the radiative heat flux through the base of the mixed layer, and as a result overestimate the surface thermal forcing, leading to a negative residual heat flux in the region.

5.3. Heat Budget in the Southern Winter

[33] In the upwelling season from July to September, SST decreases in the region (Figure 12a), with cooling rate increasing from the north to the south. In the coastal region off Java and Sumatra, a relative maximum rate of cooling is near 10°S, where the influence of ITF is important. The temperature tendency has a similar spatial pattern to the surface thermal forcing; however, careful examination of the patterns indicates that surface thermal forcing is cooling faster in most cases than temperature tendency (Figures 12a–12b). This difference indicates the influence of ocean dynamics, as documented below.

Figure 12.

Mixed layer heat budget in July–September: (a) temperature tendency, (b) surface thermal facing, (c) horizontal advection, and (d) vertical entrainment. The dashed lines in Figure 12d show the geographic locations of transects used for Figures 13 and 14 (10−7°C/s).

[34] A large part of the difference between temperature tendency and surface thermal forcing can be attributed to horizontal advection. As shown in Figure 1, the averaged SST in the region is about 5°C higher than that at the same latitudes of the southeastern Pacific. The warm SST is in part a result of horizontal advection by the ITF in this season (Figure 12c). By transporting a large amount of warm water from the western Pacific warm pool, the ITF has a large impact on the circulation and thermal structure, as documented qualitatively in earlier studies [Gordon, 1986; Wyrtki, 1987]. In the region southwest of Sumatra, the cooling temperature tendency is about 1 × 10−7°C/s (= 0.26°C/month), and is smaller in magnitude than the surface thermal forcing by a factor about 3. The horizontal advection is the primary mechanism balancing this difference (Figure 12c). Vertical entrainment is significant only near the coast of Java and farther to the east along 10°S (Figure 12d), beneath the core of maximum ITF currents. Entrainment falls below 0.5 × 10−7°C/s in magnitude in most of the region. The outflow of warm water through the various straits and passages of the Indonesia archipelago counterbalances much of the vertical entrainment near the coast of Java and farther to the east. An implication of this result is that, although upwelling does occur as indicated by ocean color observation (Figure 2), its cooling effect on the SST is much reduced due to the presence of ITF. This gives further evidence for the interpretation of small SST depressions near the coast of Java and farther to the east [Qu et al., 1994]. To the northwest of Australia, both horizontal advection and vertical entrainment are small and the surface thermal forcing basically controls the temperature tendency.

[35] It is interesting that the vertical entrainment off Sumatra is small despite the upwelling favorable southeast monsoon from June to September. The effect of upwelling on the SST seems to be counterbalanced by other components of vertical entrainment. According to equations (1) and (3), vertical entrainment also depends on the temperature gradient across the base of the ML. As illustrated in Figure 8f, the presence of thick BL prevents the cold thermocline water from entering the ML, and therefore no significant vertical entrainment occurs near the coast of Sumatra. This result suggests that the relatively small SST depression off Sumatra is due in a large part to the presence of BL.

5.4. Seasonal Heat Budget on Transects

[36] The horizontal advection and vertical entrainment are important processes influencing the regional SST in the STIO, and both mechanisms are related to the complex topography in the Indonesia archipelago. The seasonal variation of the heat budget is documented in this section along two transects through areas of maximum variability, one running east-west at 10.25°S just south of the major straits of the archipelago and the second running along the Sumatra coast and extending to the coast of northwest Australia (Figure 12d).

[37] The heat budget components along 10.25°S (Figure 13) show strong horizontal advection and vertical entrainment at the straits, indicative of strong influence of the ITF and monsoonal wind. To the south of Sumatra and Java, the temperature tendency is dominated by the annual variation, with its minimum in July, in response primarily to the annual variation of surface thermal forcing (Figures 13a–13b). At about 115°E, when Lombok Strait transport exceeds 5 Sv from April to October (Figure 3), the effect of horizontal advection is strongest (>12 × 10−7°C/s) (Figure 13c). At the same time, vertical entrainment also reaches its seasonal maximum (Figure 13d). Modulated by the MLD and temperature gradient at the base of the ML, upwelling contributes the most to vertical entrainment (Figures 14e–14f). Although upwelling is stronger to the east of Lombok Strait than that to the west, its effect on the mixed layer temperature is much reduced due to thick MLD to the east of the strait. The eastward SJC transports more fresh water and further enhances the BL to the east of the Lombok Strait in May–June (Figure 8f), leading to a weak vertical temperature gradient at the base of the ML (Figure 13f). As a result, the vertical entrainment reaches its seasonal maximum in August, but not in May–June, when upwelling is strongest (Figure 13e). This result suggests that the depression of SST south of Java is also due to thick MLD and BLT, in addition to monsoonal wind and ITF.

Figure 13.

Hovmoller diagrams of (a) temperature tendency, (b) surface thermal heat forcing, (c) horizontal advection, (d) vertical entrainment in the ML (10−7°C/s), (e) vertical velocity at the base of the ML (10−4 cm/s), and (f) temperature difference between the ML and 5 m blow along 10.25°S (°C).

Figure 14.

Same as Figure 13, but along the transect from west Sumatra to northwest Australia. The geographic location of the transect is illustrated in Figure 12.

[38] The role of ocean dynamics in the mixed layer heat budget is different along the transect from the coast of Sumatra to the coast of northwest Australia (Figure 14). Along this transect, heat budget components show a large meridional difference. In the equatorial region, heat budget components are generally small (<1 × 10−7°C/s) (Figures 14a–14d). Though upwelling is wide spread during the southeast monsoon season (June–September) (Figure 14e), vertical entrainment is weak all year-round due to the presence of the BL (Figure 14d). In the region northwest of Australia, temperature tendency is dominated by the annual cycle, with its minimum (<−5 × 10−7°C/s) in June, in response primarily to the annual variation of surface thermal forcing (Figures 14a–14b). At about 21°S, the temperature difference across the base of the ML is as large as 1.6°C, and thus vertical entrainment has a significant effect on the SST in the southern summer, though the upwelling is relatively weak at that season. The southern branch of the ITF reaches its seasonal maximum (>10 cm/s) in April (Figure 6b), which induces a horizontal heat advection exceeding 2 × 10−7°C/s. In the region away from the coast between 12°S and 18°S, the seasonal variation of horizontal advection and vertical entrainment are small, and surface thermal forcing seems to the only important process responsible for the seasonal variation of SST. This result is essentially consistent with the earlier study of Qu et al. [1994].

5.5. Area-Averaged Heat Budget

[39] On the basis of the characteristics of ML (section 4) and the heat budget in the southern winter (section 5.3), the mixed layer heat budget components are further averaged over three areas defined in Figure 8d. In the region off Java and farther to the east, all the three heat budget terms (e.g., surface thermal forcing, vertical entrainment, and horizontal advection) are significant (Figures 15a–15b). However, since the horizontal advection and vertical entrainment are nearly counterbalanced (Figure 15b), the seasonal variation of the surface thermal forcing is in good correspondence with that of temperature tendency. During the southeast monsoon season, water near the surface is forced to move offshore, and as a consequence, upwelling occurs along the coast of Java and Sumatra. At the same time, the ITF starts to strengthen toward its seasonal maximum, and this produces an enhanced warming advection in the region, especially nearby the outflow strait (Figure 7b). The cooling entrainment and warming advection are nearly counterbalanced, with a sum only about −1.5 × 10−7°C/s in July–September (Figure 15b). In a previous study, Qu et al. [1994] noted that the depression of SST mainly occurs in May, when the sum of horizontal advection and vertical entrainment reaches its negative maximum. In this regard, our results show a better agreement with observation [Susanto et al., 2001]. The residual heat flux between temperature tendency and the sum of surface thermal forcing, horizontal advection, and vertical entrainment is large in this season. We will come to this point in section 5.6.

Figure 15.

Time series of temperature tendency (bold line), surface thermal forcing (solid line), sum of surface thermal forcing, horizontal advection and vertical entrainment (bold dashed line), and residual flux (solid line with open circles) in left panel, and horizontal advection (bold line), vertical entrainment (bold dotted line), and sum of horizontal advection and vertical entrainment (solid line) in right panel in the three boxes, (top) Java, (middle) Sumatra, and (bottom) northwest Australia, selected in Figure 8 (10−7°C/s).

[40] In the region off Sumatra, all heat budget terms are relatively small (Figures 15c–15d). As discussed in section 5.3, the vertical entrainment becomes negligible due to the presence of BL. Horizontal advection is weak, but still plays a role in balancing the surface thermal forcing. In the southern winter, when surface thermal forcing is negative, horizontal advection reaches its seasonal maximum, leading to a net increase in SST. The temperature tendency is dominated by a semiannual variation (Figure 15c), which apparently cannot be explained by the annual variation in the surface thermal forcing as indicated in Figure 10b. As discussed in Section 5.3, the presence of the BL is an important mechanism responsible for the small SST depression in the region off Sumatra.

[41] In the region off northwest Australia (Figure 15c), all heat budget terms are dominated by the annual variation, as previously pointed by Qu et al. [1994]. Horizontal advection warms SST in March–August, consistent with the maximum transport of the southern surface branch of ITF. As described in section 5.4, horizontal advection and vertical entrainment have different phases. Vertical entrainment occurs in the southern summer, with its minimum <−1.5 × 10−7°C/s in February. From December to March, horizontal advection and vertical entrainment cancel about 50% of the surface thermal forcing. In the southern winter from June to July, they cancel only about 15% of the surface thermal forcing. These results differ from those based on a fixed depth (50 m) heat budget analysis. In such an earlier study, Qu et al. [1994] found that horizontal advection and vertical entrainment counterbalance most of the surface heat flux throughout the year. The difference would be due to the seasonal variation of MLD. Both horizontal advection and vertical entrainment tend to have a larger influence on the SST in the southern summer when MLD is shallow (∼35 m) than in the southern winter when the MLD is deeper (∼90 m).

5.6. Residual Heat Flux

[42] The residual heat flux in the mixed layer heat budget is not always negligible. In addition to the uncertainties caused by monthly averaging as discussed above (section 5.2), two processes are presumably responsible for the residual heat flux. One is the diffusion associated with subgrid-scale processes, and the other is eddy heat flux, especially fluxes associated with mesoscale eddies and/or intraseasonal variations. Both processes cannot be estimated using the present model output. In the region south of Java, for example, Feng and Wijffels [2002] found intense intraseasonal variations during the southeast monsoon, when the ITF and SEC are strongest. They explained these intense intraseasonal variations as a result of enhanced instability waves. We assume that a large part of the residual heat flux in the region south of Java and Sumatra from July to September is due to intraseasonal variations (Figure 15a). Excluding the effect of intraseasonal variations might have underestimated the heat advection by ITF, thus leading to a positive residual heat flux (Figure 15a). The residual heat flux is positive at almost all seasons in the region off northwest Australia (Figure15c). The mechanisms responsible for this positive residual heat flux are not known. The diffusion and mesoscale eddies may explain at least part of it.

[43] The residual heat flux is negative in the southern summer in the regions off both Sumatra and Java (Figure 15e). To illustrate the vertical distribution of the residual heat flux, we also carried out a fixed depth heat budget analysis with the depth varying from the surface to the bottom of the ocean [Feng et al., 1998]. The results of this analysis show that the maximum residual heat flux occurs in the surface layer, implying an important influence of mesoscale eddies activity on the ML and the mixed layer heat budget as well.

6. Summary and Discussion

[44] In this study, we have examined the role of ocean dynamics in determining the seasonal variation of SST off Java and Sumatra, using results from OFES. The mixed layer heat budget analysis indicates that different processes determine the seasonal variation of SST in the regions north and south of 13°S. During the southeast monsoon, upwelling is prominent at subsurface levels off Java and Sumatra, but not near the northwest coast of Australia. The region north of 13°S can be further divided into two parts: one is located to the west of Sumatra and the other lies south of Java and farther to the east. In the latter, warming horizontal advection by the ITF and the cooling entrainment associated with upwelling are nearly counterbalanced and thus their effect on the SST is quite small. Off Sumatra, the presence of a BL is a key process that impedes cold thermocline water from entering the ML. These mechanisms explain why the depression of SST near the coast of Java and Sumatra is small compared with other eastern boundary upwelling regions like that west of Peru in the eastern Pacific.

[45] The very high resolution of OFES allows an analysis of several interesting phenomena in the Indonesian archipelago. First, the outflow through Lombok Strait has a notable impact on the circulation and thermal structure in the nearby area. It impedes the SJC from transporting fresh water further to the east, resulting in a weak BL near the coast of Java. Second, the large residual heat flux implies that the mesoscale eddies and intraseasonal variations are important in the region off Java and farther to the east in the southern winter. Interestingly, these active mesoscale eddy activities mostly occur in the outflow region of the straits, a unique feature of the STIO.

[46] Finally, we note that regional ocean dynamics is quite different from that in the southeastern Pacific. First, at the coast of Java and Sumatra, the surface Ekman current reverses twice a year in response to the seasonal variation of the monsoon. Upwelling occurs only in the southern winter, and as a consequence, vertical entrainment is a strongly seasonal process. This forms a contrast with the situation off South America, where upwelling occurs all year-round. Second, the net fresh water flux is positive throughout the year and this results in a BL that further reduces the effect of upwelling and entrainment as a cooling mechanism for SST. Third, the ITF transports a large amount of heat into the region. According to our analysis of the mixed layer heat budget, vertical entrainment near the coast of Java and Sumatra is counterbalanced largely by horizontal advection. Both horizontal advection and vertical entrainment tend to cool the SST in the southeastern Pacific. In summary, ocean dynamics is primarily responsible for the small depression of SST near the coast of Java and Sumatra and the variation of horizontal advection and upwelling/entrainment are key factors in determining the seasonal cycle of SST.

[47] This study of the mean seasonal cycle of heat budget sets the background for a future study of interannual variations. The variation of Indonesian throughflow is primarily controlled by winds over the Pacific Ocean, while depth of the thermocline off Java and Sumatra is controlled primarily by winds over the equatorial Indian Ocean. The Pacific and Indian Ocean wind fields are not tightly coupled and they display substantially different timescales of variability [Wijffels and Meyers, 2004]. We expect that the delicate balances of advection, upwelling/entrainment and barrier layer formation will not always be maintained at interannual timescales and that this will lead to substantially different realizations of the seasonal heat budget.

Acknowledgments

[48] This work was supported by NASA through grant NAG5-12756, Japan Marine Science and Technology Center through its sponsorship of the International Pacific Research Center (IPRC), and CSIRO Marine Research. The OFES simulation is conducted on the Earth Simulator, and Asia-Pacific Data-Research Center (APDRC) serves its output at IPRC. The AVHRR Oceans Pathfinder SST data and SSM/I wind data were obtained from the Physical Oceanography Distributed Active Archive Center (PO.DAAC) at the NASA Jet Propulsion Laboratory, Pasadena, California, from Web site http://podaac.jpl.nasa.gov. NCEP short-wave flux was provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado, USA, from Web site http://www.cdc.noaa.gov. The SeaWiFs data and TOPEX/ERS merged SSHA data were obtained from APDRC in IPRC-SOEST, University of Hawaii, from Web site http://apdrc.soest.hawaii.edu. This is IPRC contribution number 311 and SOEST contribution number 6546.

Ancillary