Journal of Geophysical Research: Oceans

Eastern Indian Ocean warming associated with the negative Indian Ocean dipole: A case study of the 2010 event

Authors


Corresponding author: T. Horii, Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, 2-15, Natsushima, Yokosuka, Kanagawa 237-0061 Japan. (horiit@jamstec.go.jp)

Abstract

[1] Warm sea surface temperature (SST) anomalies of more than 1 °C occurred in the southeastern tropical Indian Ocean and peaked during August to October 2010. The anomalous SST warming was associated with the negative phase of the Indian Ocean dipole (IOD) phenomenon. In this study, observational data from a moored buoy were used together with satellite and atmospheric reanalysis data sets to clarify the processes that produced the anomalously warm SST in 2010. We focused on the location 5°S, 95°E where in situ measurements of more than 10 years by a moored buoy were available. The buoy observations captured the oceanic conditions related to the anomalous warming event of 2010. Heat balance analysis demonstrated that air-sea heat fluxes and horizontal heat advections mainly account for the mixed layer temperature variation. Reduced latent heat loss had a major role in producing the warm SST anomalies. Meridional heat advection also contributed to the warm SST anomalies where the southeastward surface current brought warmer water to the southeastern tropical Indian Ocean. The present results from the observations suggest that air-sea heat exchanges play an active role in the SST anomalies in the southeastern tropical Indian Ocean during the negative IOD. In contrast, in the case of cold SST anomalies in the eastern Indian Ocean during the positive IOD, ocean heat advections mainly control the mixed layer temperature variation. These results suggest that the IOD includes different feedback mechanisms in its positive and negative phases.

1 Introduction

[2] Seasonal to interannual variation of the Indo-Pacific warm pool influences local and global climate through changes in atmospheric and hydrological circulations. The Indian Ocean dipole (IOD), also referred to as the Indian Ocean zonal/dipole mode (IOZDM), is a major phenomenon associated with variations of the Indian Ocean warm pool. A positive IOD is characterized by a cold (warm) SST anomaly in the southeastern (central-western) tropical Indian Ocean [Saji et al., 1999; Webster et al., 1999]. In the opposite phase of the IOD, a negative IOD, a warm SST anomaly in the southeastern tropical Indian Ocean (SETIO) co-occurs with a cold SST anomaly in the central-western Indian Ocean. The changes in the ocean and atmosphere during the positive/negative IOD involve an east-west shift of the Indian Ocean warm pool and unusual ascending and descending air motions around the Indian Ocean rim. The anomalous atmospheric circulation may trigger planetary atmospheric waves, by which the IOD also influences global climate [e.g., Saji and Yamagata, 2003; Cai et al., 2011].

[3] Previous studies have suggested that the occurrence of the IOD is linked to anomalous climatic conditions [Reason, 2002; Guan and Yamagata, 2003; Clark et al., 2003; Ashok et al., 2003; Zubair et al., 2003; Behera et al., 2005]. During the positive (negative) IOD, an anomalous westward (eastward) moisture flux occurs across the tropical Indian Ocean, bringing enhanced rainfall over eastern Africa (Indonesia and Australia) [Saji and Yamagata, 2003]. Therefore, the IOD can cause large-scale natural disasters such as floods in eastern Africa [Black et al., 2003], floods/droughts in Australia [Ummenhofer et al., 2009], and Australian bushfires [Cai et al., 2009]. In addition, the IOD can impact the Pacific El Niño/Southern Oscillation (ENSO). A positive (negative) IOD in Northern Hemisphere fall, through a weakening (strengthening) of atmospheric convection over the Maritime Continent, may drive a La Niña (El Niño) event in the following year [Izumo et al., 2010]. For these reasons, it is essential to acquire a better understanding of the IOD.

[4] After Saji et al. [1999] suggested that Bjerknes feedback is a key process in the development of the IOD, various model-based studies have examined the evolution and processes of the IOD. Previous studies have pointed out that an oceanic process could be responsible for the cooling of the SST in SETIO during the positive phase of the IOD. Utilizing ocean general circulation models (OGCMs), Murtugudde et al. [2000] first conducted heat budget analyses of the mixed layer and concluded that the primary factor cooling the SST in SETIO is ocean vertical process, namely, entrainment cooling due to unusually strong upwelling. Several studies have also conducted surface layer/mixed layer heat budget analyses and pointed out the importance of horizontal heat advection and vertical processes [Iizuka et al., 2000; Vinayachandran et al., 2002, 2007; Du et al., 2008; Du et al., 2009; Santoso et al., 2010]. Other studies have emphasized the primary role of the changes in air-sea fluxes in the development and decay of the positive/negative IOD [Baquero-Bernal et al., 2002; Li et al., 2002, 2003; Tokinaga and Tanimoto, 2004]. Recently, using high-resolution model-based assimilated data, Halkides and Lee [2009] showed that the relative importance of surface heat flux, ocean horizontal heat advection, and ocean vertical processes differs significantly on a regional scale (~500 km) near the coasts of Sumatra and Java. Using observations from moored buoy data, Horii et al. [2009] examined mixed layer temperature balance in SETIO during the 2006 IOD and supported the result of Halkides and Lee [2009]. They concluded that it was the ocean horizontal advection and not the net surface heat flux that produced the anomalously cold SST at the mooring site (5°S, 95°E), which is 500 ~ 600 km from the coast of Sumatra. These research advancements reaffirm the need for reliable observations and high-resolution numerical modeling to quantify the processes that produce anomalous SST related to the IOD.

[5] In 2010, warm SST anomalies of more than 1 °C associated with the negative IOD appeared in SETIO (Figure 1a). The amplitude of the warming events was the second largest, after the 1998 event (negative IOD), since the era of satellite SST measurements began in 1982. The warm SST anomalies in August to October 2010 coincided with a La Niña event in the Pacific. In September 2010, the average SST at the eastern pole of the IOD (Figure 1a) reached 29.7 °C, which corresponds to an anomalous warming of +1.3 °C from the climatological value (28.4 °C). Concurrent with the SST anomalies, enhanced atmospheric convection occurred over Indonesia, East Asia, and Australia with westerly wind anomalies and eastward ocean current anomalies in the equatorial Indian Ocean (Figures 1a and 1b).

Figure 1.

(a) SST and ocean surface current anomalies from August to October 2010. The black square represents the eastern region (90°E–110°E, 10°S–equator) for the calculation of the Indian Ocean dipole mode index [Saji et al., 1999]. The small white circle indicates the location of the TRITON buoy at 5°S, 95°E. (b) OLR anomalies (shading) and surface wind anomalies (vector) for the same period. The anomalies are relative to the mean seasonal cycle for SST from 1998 to 2010, for ocean current from 1993 to 2010, and for the surface winds/OLR from 1981 to 2010.

[6] Abnormal weather conditions including heavy rainfall in Indonesia and Australia were reported during the Northern Hemisphere summer to fall of 2010 by the APEC Climate Center (www.apcc21.net/eng/notice/nl), the Australian Bureau of Meteorology (www.bom.gov.au/announcements/media_releases/climate/change/), and others. These unusual climatic conditions were consistent with a statistical teleconnection pattern related to the negative IOD [Hong et al., 2008c; Hamada et al., 2012]. Recently, Hamada et al. [2012], using rainfall data from stations on Java, suggested that the negative IOD has a stronger impact on Indonesian rainfall during the Northern Hemisphere summer-fall than does a single occurrence of a La Niña event.

[7] To date, however, no direct observational study has assessed which processes are important for the anomalous SST during the negative IOD. The above mentioned previous studies on the IOD focused on the strong cooling tendency in SETIO during the positive IOD because the cold SST anomalies have larger amplitudes than the co-occurring warm anomalies in the western Indian Ocean or the SST anomalies that occur during the negative IOD. Hong et al. [2008a, 2008b] did investigate some aspects of the negative IOD. Using an atmospheric reanalysis data set and the Simple Ocean Data Assimilation (SODA), Hong et al. [2008a] separately examined the mixed layer heat budget in SETIO during positive and negative IOD events. They pointed out that the amplitudes of anomalous horizontal and vertical advection in SETIO were different during the positive and negative IOD. With their numerical model experiments [Hong et al., 2008b], they concluded that the nonlinear ocean heat advection could cause the different tendency of cooling/warming during the positive/negative IOD. They also reported that the asymmetry of solar radiative forcing tends to enhance the skewness of the cold/warm SST anomalies. Recently, using historical, reanalysis, and Argo observation data, Cai et al. [2012] and Cai and Qiu [2012] compared the negative and positive IOD events to assess the skewness of the SST anomalies. They not only reaffirmed the importance of the nonlinear ocean heat advections for the IOD skewness but also pointed out that the anomalous solar radiation does not contribute to the skewness. As above, there have been much debate on the processes of negative IOD and most of which were based on reanalysis data. Further validation is required to clarify the processes of SST evolution and hence ocean-atmosphere interactions associated with the negative IOD.

[8] From the mid-2000s, the Indian Ocean observational network has been improved dramatically by the development of a new moored buoy array referred to as the Research Moored Array for African-Asian-Australian Monsoon Analysis and Prediction (RAMA) [McPhaden et al., 2009]. The Argo array has also provided abundant subsurface observation data for studying seasonal to interannual variations (http://www.argo.ucsd.edu). These new data have enabled us to quantify and better understand ocean-atmosphere interactions in the tropical Indian Ocean. Buoy observations in the eastern Indian Ocean have revealed the evolution of the 2006 IOD in terms of subsurface temperature and current variations [Horii et al., 2008] and heat balance in the mixed layer [Horii et al., 2009]. The signals related to the anomalous SETIO warming event in 2010 were also captured by the time series observations of the moored buoys. The 2010 event had the largest positive SST anomalies in SETIO for the recent 10 years. To understand ocean-atmosphere interactions related to the anomalous SST, adequate representation of the surface mixed-layer heat balance is crucial, because it allows us to assess ocean-atmosphere coupling feedback such as Bjerknes feedback [Bjerknes, 1969] and wind evaporation-SST (WES) feedback [Xie and Philander, 1994] associated with the IOD. We believe that quantitative findings from direct observations can be beneficial for this purpose, even with limited data in time and space. In this study, by applying the buoy measurements together with Argo, satellite, and reanalysis products, we analyzed the mixed layer temperature balance in SETIO with the goal of clarifying the processes that produced the anomalous SST warming in 2010.

[9] The remainder of this paper is organized as follows. Section 'Data and Processing' presents the data sets and data processing procedures. Section 'Variations in SETIO During 2010' briefly describes the major features of oceanic and atmospheric variations in the eastern equatorial Indian Ocean for the year 2010. Section 'Mixed Layer Heat Balance' examines the processes determining the heat balance at a buoy location. The results and implications of this study are discussed in section 'Discussion', followed by a summary and conclusions in section 'Summary and Conclusions'. Because our focus was the anomalous SST warming in the SETIO region, we use the term “SETIO warming event” in this paper instead of “negative IOD.”

2 Data and Processing

2.1 Data Sets

We used data from the Research Moored Array for African-Asian-Australian Monsoon Analysis and Prediction (RAMA) [McPhaden et al., 2009]. We focused on a site (5°S, 95°E) at which the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) deployed a moored buoy of the Triangle Trans-Ocean Buoy Network (TRITON) since late October 2001. Time series of the subsurface temperature and salinity were continuously measured at 10-min intervals, and hourly mean data were transmitted in real time via Argos satellites. The depth levels of the temperature and salinity observations are shown in Table 1. The data are available at www.jamstec.go.jp/jamstec/TRITON/. See Kuroda [2002] and Ueki et al. [2010] for detailed descriptions of the sensor specifications and data sampling intervals. The location of the buoy allowed us to observe and analyze interannual variation at the eastern pole of the IOD (Figure 1). Although there have been some other RAMA buoys in SETIO, this was the only buoy in SETIO that collected ocean data with few missing values in 2010 (http://www.pmel.noaa.gov/tao/rama/). In addition, the buoy at 5°S, 95°E has provided the longest time series in RAMA from October 2001 to the present. To compensate for the spatial limitation of the buoy measurements, we also used data from satellite and reanalysis products.

Table 1. Depths of Temperature and Salinity Sensors in the TRITON Mooring at 5°S, 95°E (in meters)
 October 2001 to December 2007January 2008 to December 2010
Temperature1.5, 25, 50, 75, 100, 125, 150, 200, 250, 300, 500, and 750 (12 levels)1, 10, 20, 40, 60, 80, 100, 120, 140, 200, 300, and 500 (12 levels)
SalinitySame as above1, 10, 20, 40, 100 (5 levels)

In the 9-year period from 2002 to 2010, the temperature and salinity time series from the buoy had two major data gaps: from 16 November 2003 to 8 July 2004 (about 8 months) and from 30 July 2008 to 15 March 2009 (about 7.5 months). To fill the gaps, we also used a monthly temperature and salinity data set based on Argo floats, the Grid Point Value of the Monthly Objective Analysis (MOAA GPV) [Hosoda et al., 2008]. The MOAA GPV data set provides monthly data for the period of 1999 to present on 1° by 1° grids of the world oceans. We used the four grid points of the data set nearest 5°S, 95°E and interpolated the temperature and salinity data using the spline method. We used the interpolated time series only for the calculation of climatology, the time scale of which is longer than monthly. Therefore, the interpolation errors that resulted from the use of the Argo data should be insignificant.

The buoy also observed meteorological variables and surface currents, which are fundamental values for estimating air-sea and lateral heat fluxes. An ocean current meter was installed at 10 m depth. Wind, air temperature, humidity, and shortwave radiation over the sea surface were observed by atmospheric sensors installed on the buoy and were also measured every 10 min. Although these data have non-negligible missing values owing to sensor troubles and damage by human activities (vandalism), the in situ measurements from the buoys provide long time series of observations at a fine temporal resolution. In contrast with subsurface temperature/salinity observations, the meteorological and surface current observations for 2010 contained major data gaps. We therefore used these data only for validation of data from other data sources (see Appendix 'Validation of Flux').

For our analysis, we used air-sea flux data from a combination of satellite-based and reanalysis products after validating these with the fluxes calculated from bulk variables measured by the buoy (Appendix 'Validation of Flux'). Turbulent heat fluxes were obtained from the objectively analyzed air-sea heat fluxes (OAFlux; Yu and Weller, 2007). This data set was available on a 1° × 1° grid. Longwave radiation was taken from National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis products [Kalnay et al., 1996]. To investigate large-scale atmospheric variations, we also used NCEP/NCAR surface wind, OAFlux wind speed, and outgoing longwave radiation (OLR) data. The OLR data set was provided by National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites [Liebmann and Smith, 1996].

Shortwave radiation data were obtained from the data set of the International Satellite Cloud Climatology Project (ISCCP) [Zhang et al., 2004]. The shortwave radiation product is available for 1983–2009. We used the four grid points nearest 5°S, 95°E to make the time series. To estimate shortwave radiation at 5°S, 95°E for the year 2010, we constructed a time series of shortwave radiation from the OLR data by the following procedure. First, we prepared daily time series of the OLR data at 5°S, 95°E. Then, we compared the OLR data with in situ shortwave observations from the buoy for 5 day averages during 2007–2009 and calculated monthly regression coefficients of the OLR data to the in situ observation. Finally, we applied the monthly regression coefficients to the daily OLR time series and prepared a shortwave radiation time series. The resulting time series was very similar to that by the ISCCP, with a correlation [root mean square (RMS) difference] of 0.93 (11.2 W m−2) for 11 day averages.

To estimate horizontal heat advection, we used two satellite-based products after validating the products with in situ measurements. SST data from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) [Wentz, 1997] were used to calculate the horizontal temperature gradient. Surface currents at 15 m depth were provided by the Ocean Surface Current Analysis-Real Time (OSCAR) project. The OSCAR data, which were derived from Ocean Topography Experiment (TOPEX)/Poseidon sea surface height, Special Sensor Microwave Imager (SSM/I) winds, and NOAA optimum interpolation (OI) SST, gave surface currents consistent with TAO/TRITON observations [Bonjean and Lagerloef, 2002]. We also used 7-day Mapped Sea Level Anomalies (MSLA) generated by AVISO (www.aviso.oceanobs.com/) for the period 1993–2010 to observe interannual variability of sea surface height (SSH). For discussions of anomalous conditions, climatological mean seasonal cycles were calculated based on data for SST from 1998 to 2010 and for SSH/surface currents from 1993 to 2010.

2.2 Heat Balance Formalism

To understand the processes controlling SST variation in the eastern equatorial Indian Ocean, we analyzed the heat balance in the mixed layer at 5°S, 95°E. We mostly followed the formalism proposed and arranged by earlier works using moored buoy data in the equatorial Pacific Ocean [e.g., Cronin and McPhaden, 1997; Wang and McPhaden, 1999, 2000].

The mixed layer heat balance equation can be written as follows:

display math(1)

where T is the mixed layer temperature, ∂T/∂t represents the time rate of its change, ρ is seawater density, Cp is heat capacity, and H is the mixed layer depth (MLD). Here, ρ and Cp are set as the constant values of 1030 kg m−1 and 3930 J (kg °C)-1, respectively. The terms −UTx and −VTy represent horizontal heat advection, such that U (V) and Tx (Ty) denote the zonal (meridional) current and temperature gradient in the mixed layer, respectively. R is the residual, which consists of components that we could not estimate directly from the observational data, such as the horizontal and vertical diffusivity. In this study, vertical heat fluxes from the bottom of the mixed layer were also discussed as a residual because the coarse vertical resolution of the temperature and salinity sensors on the mooring buoy is inappropriate for estimating entrainment cooling at the bottom of the mixed layer. The residual also includes analysis and sampling errors.

The MLD (H) was estimated by a density criterion because the effect of salinity on near-surface stratification was not negligible at the buoy location. After interpolating vertically to every 1 m by the Akima spline method [Akima, 1970], we defined H as the depth at which σθ(z = H) = σθ(z = 0) + (σθ/∂T)∆T, where σθ is the potential density and ∆T was set to 0.5 °C [Sprintall and Tomczak, 1992].

The first term of the right-hand side of Eq. ((1)) represents the contribution of air-sea heat flux to the mixed layer temperature. Q0 is the net heat flux across the air-sea interface, which consists of radiation and turbulent heat fluxes. Qpen represents the shortwave radiation penetrating below the mixed layer and was estimated using the equation for moderately clear water proposed by Paulson and Simpson [1977]. The albedo of the sea surface was set to a constant value of 0.945. The ISCCP shortwave radiation and turbulent heat flux of OAFlux were averaged for the 2° × 2° box centered on 5°S, 95°E. NCEP/NCAR longwave radiation was taken from the nearest point to 5°S, 95°E.

All flux products were adjusted to the fluxes calculated from bulk variables measured by the buoy (called “buoy flux”), and thus their long-term averages for the period of data coexistence should be equal. In the bulk calculations from buoy observations, we estimated longwave radiation and turbulent heat fluxes using the bulk formulae given by Clark et al. [1974] and Fairall et al. [1996], respectively. We found significant differences in shortwave and longwave radiation between the flux products and buoy fluxes. We adjusted the ISCCP product (NCEP longwave radiation) to the buoy flux by subtracting 8.8 W/m2 (adding 6.4 W/m2) uniformly. Although differences were relatively small, we also adjusted OAFlux latent heat flux (sensible heat flux) by subtracting 2.6 W/m2 (1.0 W/m2). Possible overestimation/underestimation in the net air-sea flux and its impact on the heat balance analysis will be discussed in section 'Mixed Layer Heat Balance'.

To calculate the second and third terms of the right-hand side of equation ((1)), OSCAR data for the zonal and meridional currents (U and V) and the zonal and meridional temperature gradients (Tx and Ty) calculated from TMI SST were combined. Here, we assumed that the surface currents from OSCAR, which were adjusted to the current at 15 m depth, represented the vertically averaged velocity in the mixed layer. We also assumed that the horizontal gradient in SST was nearly the same as that in the mixed layer [Wang and McPhaden, 1999]. After re-gridding TMI SST data to the same 1° × 1° grid of OSCAR, we computed the horizontal advection in each grid using the current and the central difference of temperature gradient. To capture the large-scale variation, we averaged the horizontal heat advection for the area 93°E–97°E and 7°S–3°S, which is centered on 5°S, 95°E. The results were insensitive to minor modification of the area. See Appendix A for details about the validation and errors of the horizontal heat advection.

For simplicity, we calculated all heat balance terms as daily time steps. Buoy observations and flux data were averaged to daily values. OSCAR products for every 5–6 days were interpolated to daily steps using the spline method. To focus on time scales larger than 1 week, all data were smoothed with an 11 day running mean filter. Note that this procedure did not filter out the spectral peak of intraseasonal variation (30–50 days) that stands out in the tropical Indian Ocean. For discussions of the climatology and anomaly of the heat balance, climatologies were defined as the 9 year mean from 2002 to 2010, in which the data were smoothed with a 31 day running mean filter.

3 Variations in SETIO During 2010

[10] Here, we will briefly describe the major features of oceanic and atmospheric variations associated with the anomalous warming in SETIO. Positive SST anomalies appeared in the eastern part of the Indian Ocean and negative SST anomalies appeared in the west during August to October 2010 (Figure 1a). The positive SST anomalies off the coasts of Sumatra and Java had larger amplitudes than those in the west. Corresponding to the anomalous SST distribution, westerly wind anomalies prevailed in the equatorial Indian Ocean (Figure 1b). Atmospheric convection was enhanced over the SETIO to Indonesia and surface wind convergence occurred there.

[11] The buoy at 5°S, 95°E was located in a part of the anomalous warming in SETIO (Figure 1). The mixed layer temperature anomaly and its evolution observed by the buoy were roughly the same as those averaged for the eastern pole of the IOD, although the variation at the buoy location had a somewhat smaller amplitude (Figure 2). During the first half of 2010, the SST anomaly in the equatorial Indian Ocean was positive and short-term variability resulting from intraseasonal variations stood out between January and May. Then, the SST anomalies in SETIO increased and reached +1.0 °C in August. Concurrently, a westerly wind anomaly, manifesting the strengthening of the Walker circulation with the SST warming, occurred in the central equatorial Indian Ocean. The peak phase of the anomalous warming around September corresponded to the period of climatological SST depression (Figure 2b). The SST anomalies exceeded +2.0 °C off the coasts of Sumatra and Java. In concert with the seasonal monsoon reversal, the warm anomalies ceased from October to December.

Figure 2.

(a) Time series of the SST anomaly averaged for the region 90°E–110°E, 10°S–equator (see also Figure 1) (thin red) and the mixed layer temperature anomaly from the TRITON buoy at 5°S, 95°E (bold red). The data were smoothed with an 11 day running mean filter. The dashed line indicates one standard deviation (0.505) of the SST anomaly for 1998–2010. (b) As in Figure 2a but for the climatologies. The climatologies are defined as the mean seasonal cycle for SST from 1998 to 2010 and for buoy temperature from 2002 to 2010.

[12] The ocean profiles at the buoy location associated with the SETIO warming in 2010 are shown in Figure 3. During the first half of 2010, intraseasonal perturbations appeared in the surface to upper thermocline (around the 25–27 °C isotherm) and were also observed in the salinity profile. The MLD was shallow from January to May and became deeper from May to July (Figure 3a), similar to the climatological evolution (not shown). From July to August, positive temperature and high salinity anomalies occupied the surface to 100 m depth region. The positive temperature anomalies peaked in September together with the relatively shallow MLD. The local minimum of the MLD in mid-September to mid-October (~40 m) was about 10 m shallower than the climatological depth. During the period, there were small peaks of relatively low salinity water from the surface to 20 m (shown by the 34.6 psu contour) that contributed to the shallow MLD. Then, the surface temperature anomaly changed from positive to negative while an unusual surface-to-subsurface high salinity anomaly persisted by the end of the year.

Figure 3.

Time-depth section of (a) temperature and (b) salinity during 2010 observed by the TRITON buoy at 5°S, 95°E. Colors (contours) show the anomalies (raw values). The contour interval is 1°C in Figure 3a and 0.2 psu in Figure 3b. The bold gray line in Figure 3a denotes the MLD estimated by the density (σθ) criterion [Sprintall and Tomczak, 1992]. The data were smoothed with an 11 day running mean filter.

[13] Atmospheric variations in SETIO were closely related with the evolution of SETIO warming. In 2010, the wind speed weakened from July to September, concurrent with the anomalous atmospheric and oceanic conditions in SETIO (Figures 1b and 4a). In this situation, the westerly wind anomaly and the climatological easterly wind canceled each other, resulting in the weak wind. As will be described in section 'Mixed Layer Heat Balance', the wind variation strongly influenced the latent heat flux variations. The weak wind was also concurrent with the relatively shallow MLD from mid-September to mid-October (Figure 3b). The period of anomalously weak wind from July to October corresponded to the period of climatological strong wind associated with the northern summer monsoon.

[14] An anomalous atmospheric convection was also observed in the OLR variations (Figure 4b). The OLR fluctuated with a time scale of 30–40 days. At a longer time scale, in concert with the development of warm SST anomalies, the OLR was decreased and rarely exceeded the climatological mean value from July to October. The decreased OLR continued to the end of the year, accompanying the intraseasonal variations.

Figure 4.

Time series of (a) wind speed (m/s) and (b) OLR (W/m2) for the area around the TRITON buoy locations (90°E–100°E, 10°S–equator). Each blue line shows the data for the year 2010. The data were smoothed with an 11 day running mean filter. The black line shows the climatology. The climatologies are defined as the mean seasonal cycle for wind speed from 1985 to 2010 and for OLR from 1981 to 2010.

[15] In the next section, we will clarify how these oceanic and atmospheric variations resulted in the anomalous SST evolution via heat exchange in SETIO.

4 Mixed Layer Heat Balance

[16] In this section, we examine the processes that determined the heat balance at the buoy location (5°S, 95°E). We first present estimations of air-sea heat fluxes and horizontal heat advection. Then, we describe the heat balance analysis of the mixed layer temperature at the buoy location and error estimates.

4.1 Surface Heat Flux

Heat gain due to shortwave radiation and heat loss due to latent heat flux are the primary components of the net heat flux variation (Figure 5). In the climatology (Figure 5a), reflecting the Intertropical Convergence Zone movement across SETIO that occurs twice in each year, shortwave radiation has a peak larger than 200 W/m2 around April and is depressed around June to July (~165 W/m2) and November to December (~170 W/m2). The climatological latent heat also has a large annual variation from about –150 to –105 W/m2 that clearly reflects the seasonal wind-speed variation (Figure 4a). The net heat flux is slightly positive during January to April and negative mainly in June to September. On annual average, net air-sea heat exchange excluding the shortwave radiation penetrating below the mixed layer is almost zero (+0.4 W m2) in this analysis. This value is significantly low and possibly underestimates ocean heat gain by ~20 W/m2 compared with the mean of the flux products [Yu et al., 2007]. However, the various currently available flux products are widely divergent in the tropical Indian Ocean, and our estimate is within the error range among the products.

Figure 5.

(a) Time series of the climatological air-sea heat flux components (W/m2) at 5°S, 95°E. A positive heat flux indicates a heat gain to the ocean. The lines denote shortwave radiation (bold red), shortwave radiation that penetrates below the mixed layer (thin red), net longwave radiation (black), sensible heat flux (green), latent heat flux (blue), and net air-sea heat flux for the mixed layer (gray). The climatological values are defined as the mean seasonal cycle from 2002 to 2010. (b) As in Figure 5a but for the year 2010. (c) As in Figure 5a but for the anomalies of shortwave radiation, latent heat flux, and net flux. The data were smoothed with an 11 day running mean filter. See Appendix 'Validation of Flux' for the validation of the flux variables.

For the surface heat fluxes in 2010 (Figures 5b and 5c), intraseasonal variations, reflecting the strong influence of the Madden-Julian oscillation (MJO) [Madden and Julian, 1972] or intraseasonal oscillation (ISO) [e.g., Sikka and Gadgil, 1980] in the tropical Indian Ocean, were observed in the net surface heat flux. The shortwave radiation and latent heat variations were strongly related to OLR and wind speed variations, respectively (Figure 4). Concurrent enhanced (depressed) shortwave radiation and small (large) latent heat loss, clearly seen from April to June, were probably a result of the dry (wet) phase of the MJO, in which less (more) convection occurs with a weak (strong) wind condition [e.g., Shinoda et al., 1998; Wheeler and Hendon, 2004]. The high-frequency signals were probably not due to the initial SST warming in the SETIO associated with negative IOD because these signals were not confined in SETIO but appeared over the whole tropical Indian Ocean. The signal did not accompany the successive equatorial westerly wind anomalies, unlike the negative IOD signal from July onward (figure not shown).

The above mentioned balance of surface heat fluxes changed from July 2010 when an anomalously weak wind speeds and enhanced convections were observed in SETIO (Figures 4 and 5). Although the shortwave radiation was relatively reduced for the period, the anomalously small latent heat loss (positive anomaly) caused by the weak wind during mid-July to mid-October was more significant than the shortwave change resulting in an anomalous heat gain for the mixed layer (Figures 5b and 5c). The positive net heat flux anomaly was more than +24 W/m2 (+40 W/m2) for the average (for the maximum) during August to September. The weaker-than-normal shortwave radiation continued to the end of the year, while the latent heat loss became anomalously large (negative anomaly) from mid-October to December. The strengthening wind speed and the latent heat flux were related to an acceleration of the climatological westerly wind with a westerly wind anomaly for the period. As a result, the net heat flux anomaly became negative from mid-October to December.

These results indicate that the anomalous variations of the surface heat fluxes had an active role in developing/weakening the warm SST anomaly in SETIO. Although the shortwave time series for 2010 was limited to the point of 5°S, 95°E, the results were essentially the same as those for a modified study area (such as 90°E–100°E, 10°S–0°, not shown). Therefore, the flux balance should be general in SETIO.

4.2 Horizontal Heat Advections

Horizontal heat advection has a considerable impact on the mixed layer temperature at 5°S, 95°E (Figure 6). In the climatology, the amplitude of the meridional heat advection is, on annual average, more than twice as large as that for the zonal heat advection (Figures 6a and 6b). Although the zonal current typically has larger amplitude than does the meridional current, zonal heat advection contributed little to the mixed layer temperature because of the small horizontal temperature gradient. The meridional heat advection has its peak (>0.6 °C/month) in July to August during the northern summer monsoon period and remains positive to the end of year. In this situation, southward current carries warmer water toward the south.

Figure 6.

Time series of the surface currents (m/s; bold gray), horizontal temperature gradient (10−5 °C/m; thin black), and horizontal heat advection (−U Tx and −V Ty; °C/month; colors) at 5°S, 95°E. Red (blue) color indicates positive (negative) heat advection that warms (cools) the mixed layer. The (a) zonal and (b) meridional climatologies were defined as the mean seasonal cycle from 2002 to 2010. (c and d) Same as Figures 6a and 6b, respectively, but for the year 2010. Notice that < U > <Tx > and < U Tx > do not necessarily match because the data are based on temporal (9 years) and/or spatial (93°E–97°E and 7°S–3°S) averages of < U >, < Tx >, and <U Tx > time series.

In 2010, both zonal and meridional heat advections changed remarkably with intraseasonal variations (Figures 6c and 6d). Different from the climatology, zonal current remained eastward during July to October, which was likely a response to the anomalous westerly wind. However, zonal heat advection contributed little to the mixed layer temperature for the same reason as for the climatological condition. In the case of meridional heat advection, stronger southward currents and enhanced meridional temperature gradients produced southward heat advection from April to September, which accompanied the intraseasonal perturbations. From October 2010, when the anomalously warm SST started to decay, both zonal and meridional advections worked to anomalously reduce the mixed layer temperature. The eastward surface current brought cold water from west to east in October to mid-December. The meridional temperature gradients weakened and the southward advections decreased in October to December, which was also an anomalous condition because the climatological meridional temperature gradients and heat advection are still positive for this period.

To understand how the anomalous heat advections occurred during 2010, we decomposed the surface currents and horizontal temperature gradients to the following expressions:

display math

where the angled brackets denote the mean seasonal cycle (climatology) and the prime denotes the anomaly from the climatology. Thus, the zonal and meridional advections can be decomposed as

display math(2)
display math(3)

In the right-hand sides of equations ((2)) and ((3)), the terms < U > < T x> and < V > <Ty > represent the climatological heat advection and the other terms represent the anomalies.

Figure 7 summarizes each component of the heat advection for the development phase (JJA: June-July-August) and the decay phase (OND: October-November-December) of the SETIO warming event. In the development phase of the anomalous SST warming, the zonal heat advection (UTx) was slightly negative (~–0.07 °C/month) and the difference from the climatology term (< U > < Tx >) was also small. The meridional heat advection (V Ty) for the period was larger than the climatology (< V > < Ty >). The anomalous component was produced by the anomalous temperature gradient and the climatological/anomalous southward current (< V > Ty′ + V′Ty′), as can be understood from Figures 6d and 7b. For the 3 month period, the anomalous warming of the mixed layer temperature of +0.3 °C can be explained by the meridional heat advection. In addition, the meridional advection in April to May also partly contributed to the warming of the mixed layer temperature (Figure 6).

Figure 7.

Heat advection terms for the (a and b) development phase (June–August, 2010) and for the (c and d) decay phase (October–December 2010) of the SETIO SST warming. The components of horizontal heat advection are decomposed to the climatologies and anomalies as (1) < U > <Tx >, (2) < U > Tx′, (3) U′ < Tx >, and (4) U′Tx′. Positive (negative) values indicate heat advections that warm (cool) the mixed layer temperature.

In October to December 2010, for the decay phase of the warm SST anomaly, both the zonal and meridional advection anomalously cooled the mixed layer temperature (Figures 7c and 7d). For the zonal heat advection, the anomalous surface current and anomalous temperature gradient pair (U′Tx′) made a large contribution (approximately −0.2 °C/month). The meridional heat advection (VTy) was slightly positive but significantly smaller than its climatology (< V > < Ty >). The largest term in the anomalous meridional advection (approximately –0.38 °C/month) came from the climatological southward surface current and the anomalous temperature gradient (< V > Ty′).

The contributions of the heat advections were essentially the same for the wide region of SETIO in 2010, especially in the southern part of the eastern domain (around 10°S–5°S, 90°S–100°S; not shown). In summary, these results strongly suggest that oceanic dynamics also played active roles in producing and decaying the SST anomalies during the 2010 SETIO warming event.

4.3 Temperature Balance

We begin this section by showing the mean seasonal variation in the heat balance at 5°S, 95°E. The contributions of the air-sea heat fluxes and horizontal heat advections to the mixed layer temperature, represented by the terms in the heat balance equation ((1)), are shown in Figure 8.

Figure 8.

Climatological components of the mixed layer temperature balance (1) (°C/month): (a) net heat flux (Qnet/ρCpH, color), (b) horizontal advection (−UTxVTy, color), and (c) the sum of net heat flux and horizontal advection (Qnet/ρCpH − UTx − VTy, gray line). The black lines in Figures 8a–8c (common) show temporal change of the climatological mixed layer temperature. The vertical bars and light shadings indicate the errors in the analysis. Details of the error estimation are presented in the Appendix 'Errors in Heat Balance'.

In the climatology, the mixed layer temperature warms from January to April, cools from May to September, and warms again from October to mid-December. The annual variation in the net surface heat flux is roughly in phase with the mixed layer temperature change, except for in October–December (Figure 8a). The variations in the net surface heat flux are explained by the combination of shortwave radiation and latent heat loss, as was shown in section 'Surface Heat Flux'.

The climatological heat advection clearly has a different evolution from the mixed layer temperature variation (Figure 8b). Except for January to March, a southward current carries warm water from north to south, and the heat advection warms the mixed layer temperature at the study site (Figure 6b). The warming signal from the advection counteracts the cooling tendency from the net surface heat flux during the northern summer monsoon period from May to September. From October to December, the meridional heat advection explains the warming tendency of the mixed layer temperature.

The mean seasonal variation in the mixed layer temperature is well explained by the sum of the net surface heat flux and horizontal heat advection (Figure 8c), suggesting that the heat balance estimate is reasonable. The difference between the two time series indicates the residual term of the heat balance equation ((1)). In general, the sum of the net heat flux and the horizontal heat advection has a positive bias for mixed layer temperature variation, especially in January, April–September, and November. The difference would be caused by the effect of entrainment cooling at the bottom of the mixed layer, although we cannot form a conclusion considering the large errors. If we underestimated ocean heat gain by the net air-sea heat flux, as mentioned in sections 'Heat Balance Formalism' and 'Surface Heat Flux', the contribution of entrainment cooling could perhaps be larger.

We now consider the mixed layer heat balance in 2010 (Figure 9). The anomalies were calculated by subtracting the climatology (Figure 8) from the original data. Throughout the year, intraseasonal variations were apparent in the heat balance. Large intraseasonal temperature variations in January to April were mainly explained by the net surface heat flux anomaly, although values sometimes differed in terms of phase and variability. The net surface heat flux variations were explained by the combination of large (small) solar radiation and small (large) latent heat loss. As was mentioned in sections 'Variations in SETIO During 2010' and 'Surface Heat Flux', this is consistent with flux variation produced by the MJO [Shinoda et al., 1998]. Concurrent with the intraseasonal signals, the large residuals appear in February to April and June. A possible cause of the residuals might be the MLD estimated in this study: a shallower (deeper) MLD would lead to an overestimate (underestimate) of the contribution of net surface heat flux. Although we do not have enough observation data to detect and identify the process, intraseasonal variations of the entrainment cooling would also contribute to the residuals.

Figure 9.

As in Figure 8 but for anomalous components of the mixed layer temperature balance for the year 2010. The data were smoothed with an 11 day running mean filter.

From May to August, the anomalous net surface heat flux changed from negative to positive. Anomalous horizontal heat advection had a positive contribution to the mixed layer temperature variations, the average of which was +0.18 °C/month, corresponding to the +0.7 °C anomalous warming for the period. In September, the net surface heat flux anomaly contributed to the warming of the mixed layer temperature anomaly. The positive net surface heat flux anomaly was caused by a weakening wind from mid-August to September. A small peak in shoaling MLD around September (Figure 3) also could have caused a warming tendency because the shallow MLD would have caused the mixed layer to be more sensitive to net surface heat flux, but the relative importance of this effect was not as large as the anomalous net heat flux itself.

From October to December 2010, both the net surface heat flux and horizontal heat advection anomalies generally contributed to a cooling of the mixed layer temperature anomaly. This period corresponded to the decay phase of the warm SST anomalies in SETIO. The contribution of the horizontal heat advection was about three times as large as that of the net surface heat flux. As can be seen from Figure 7, anomalous zonal and meridional heat advection had a cooling effect on the mixed layer temperature anomaly. In addition, enhanced latent heat loss due to strengthening wind speed anomalously cooled the mixed layer temperature, as did suppressed shortwave radiation, as mentioned in section 'Surface Heat Flux'.

These results indicate that the combination of net surface heat flux and horizontal heat advection produced and decayed the warm SST anomalies in the SETIO during the northern fall of 2010. Interestingly, the process of anomalous SST evolution was quite different in the case of anomalous cooling in SETIO during a positive IOD event [Horii et al., 2009]. This point shall be discussed in the next section.

5 Discussion

[17] The present results indicate that the relative importance of heat fluxes and horizontal advections changed remarkably in the development, peak, and decay phases of the anomalous SETIO warming event in 2010. In this section, we discuss these results and implications with reference to previous findings on the evolution of the SST anomalies in SETIO. We pay special attention to differences between the SETIO warming (negative IOD) and cooling (positive IOD) processes.

[18] The most intriguing point from these observational results is the different process of SST evolution between the warming event (negative IOD) and the cooling event (positive IOD). Horii et al. [2009] studied the SETIO cooling event during the 2006 IOD in almost same way. Comparing the two opposite events, the 2006 cooling and the 2010 warming, the principal difference is whether the surface heat flux had an active role in the positive feedback during the development stage. In mid-July to September 2010, the effect of anomalously weak latent heat loss overcame the relatively weak shortwave radiation and acted to continue warming the SST anomalies in SETIO. The circumstance was consistent with WES feedback [Xie and Philander, 1994], in which the anomalous westerly wind weakened or reversed the background easterly wind and reduced latent heat loss, producing the anomalous warming tendency of SST. In the development-to-peak phase of the cooling event during the same season of 2006, on the other hand, the contribution of the heat flux was small because strong shortwave radiation and large latent heat loss almost canceled each other and the mixed layer was deep [Horii et al., 2009].

[19] The results of our case studies on the warming/cooling event in the 2000s should be compared with previous studies on the negative/positive IOD based on the reanalysis data and OGCMs [Hong et al., 2008a, 2008b; Hong and Li, 2010; Cai et al., 2012]. Regarding the contribution of air-sea flux, the present observational results for the 2006 and 2010 events demonstrate that air-sea heat exchanges played more active role in warming the SST anomalies during the negative IOD, than in cooling during the positive IOD. Although sample cases are small, buoy observations agree well with Cai et al. [2012] but differs from Hong et al., [2008a]. After peaks of warming/cooling events, it was common for the surface heat flux to act as a negative feedback for the SST anomalies. In the decay phase of the warming (cooling) event in 2010 (2006), large (small) latent heat loss related to strong (weak) winds and depressed (enhanced) shortwave radiation caused a cooling (warming) tendency because of strengthening (weakening) wind speed. Whether and how these surface flux contributions are different among the IOD events should be clarified in future report.

[20] In the present results, anomalous horizontal heat advections contributed to both the warming in the development phase and cooling in the decay phase of the 2010 SETIO warming event. Although the stated importance of ocean advections during the negative IOD is consistent with previous studies [Hong et al., 2008a, 2008b; Zheng et al., 2010; Cai and Qiu, 2012], our results at a particular site (5°S, 95°E) provide a different interpretation from past results. Previous studies emphasized the importance of anomalous Bjerknes feedback (thermocline feedback) during the event. These studies stated that weakening of the seasonal upwelling of cool water in SETIO produced anomalous warming. The importance of Bjerknes feedback should be significant especially in areas near the coasts of Sumatra and Java, where upwelling and entrainment cooling are dominant [Du et al., 2005; Du and Xie, 2008; Hong et al., 2008a; Halkides and Lee, 2009]. In this study, in addition to this, it was suggested that meridional advection related to the anomalous temperature gradient and climatological/anomalous southward current produced the warm mixed layer temperature anomaly in SETIO. A picture of the anomalous horizontal advection can be seen in the SSH anomaly field in the development-to-peak phase (Figure 10). It demonstrates that positive (negative) SSH anomalies spread along the coast of Sumatra and Java (off the equator from 5°S, 80°E to 12°S, 100°E) and anomalous southeastward currents flowed between the crest and trough of the SSH anomalies. The positive and negative SSH anomalies were a response to the equatorial westerly wind anomalies and off-equatorial negative wind curl anomalies that often appear in the eastern Indian Ocean as an interannual wind variation related to ENSO and/or the IOD [Yu et al., 2005] (see also Figure 1). This ocean geostrophic condition is favorable to bringing warmer water to SETIO.

Figure 10.

The SSH and ocean surface current anomalies averaged from June to August 2010. The white square indicates the location of the TRITON buoy at 5°S, 95°E. The anomalies are relative to the mean seasonal cycle from 1993 to 2010.

[21] The different interpretations of the ocean heat advection could reflect differences in study areas: our analysis focused on observations at 5°S, 95°E in the eastern Indian Ocean warm pool, whereas the previous studies examined the heat balance over large areas including the coasts of Sumatra and Java. Although the present study did not consider the interannual variation in entrainment cooling, we identified that horizontal advection warmed and cooled the mixed layer temperature in a wide region of SETIO. Thus, we conclude that the contributions of the horizontal advections were also robust for the SETIO warming event of 2010. In addition, since the warmest SST anomalies along the coasts of Sumatra and Java could not be explained by the anomalous meridional advections, an observational assessment of interannual variation of the entrainment cooling is also desirable for the coastal regions in future work.

[22] The anomalous horizontal heat advections shown in Figures 6 and 7 also raise another question: What produced the anomalous warm SST anomalies that then advected to SETIO? A possible candidate could be the basin-wide warming in the Indian Ocean [Chambers et al., 1999] associated with the 2009 El Niño in the Pacific. From April to June 2010, warm SST anomalies, probably forced by El Niño-induced heat flux anomalies, distributed the central to northern Indian Ocean. In this situation, enhanced shortwave radiation increased the warm SST anomalies in the equatorial to northern Indian Ocean [Du et al., 2008] and strengthened the meridional temperature gradient. Afterward, climatological and anomalous southward currents in SETIO brought the warmer water from north to south, resulting in the anomalous warm heat advection. If this was the case, one might expect that ENSO drove the warming event or the negative IOD in 2010. Observational results indicate that the anomalous warming of SST, enhanced OLR, westerly wind anomalies, and westward surface current anomalies developed in phase during the summer to fall 2010. It is therefore likely that the anomalous components all interacted. It is reasonable to think that ENSO preconditioned the warm SST anomalies but that an ocean-atmosphere coupling interaction developed the warm SST anomalies during the 2010 warming event.

[23] It should be noted that the present observational heat balance does not explicitly address ocean vertical processes, specifically, entrainment cooling and vertical advection at the bottom of the mixed layer. The signal of local entrainment cooling might have appeared in the residual term as a cooling tendency that could not be explained by net heat flux and horizontal advection. To support this, relatively large residuals occurred from April to July and in mid-August to September and correspond to the period when the MLD tended to deepen on average. The vertical temperature gradients around the mixed layer depth were anomalously less (Figure 3), associated with the negative IOD from August to December 2010. The anomalous thermal structure and a possible weakening of local entrainment cooling might have contributed to some of the anomalous warming in 2010, as pointed out by Hong et al. [2008a, 2008b] and Cai and Qiu [2012]. However, the role was probably minor because of the anomalously deeper MLD at the buoy location at this period.

[24] Another remarkable hydrographic feature in SETIO in 2010 was the high salinity water from the surface to 100 m depth that appeared in the latter half of 2010 (Figure 3). A salinity variation can influence a change in the heat balance because ocean mixing could be interrupted by the existence of the halocline in the isothermal layer and the shallow mixed layer could trap heat exchange near the surface layer [Lukas and Lindstrom, 1991; Sprintall and Tomczak, 1992]. Therefore, the effects of salinity on the heat balance should be considered. Qiu et al. [2012] examined interannual variations in the barrier layer in SETIO using gridded temperature and salinity profiles from Argo floats. They showed that a stronger fall eastward current caused by westerly wind anomalies brought high salinity water to the east, resulting in both a thick isothermal layer and a barrier layer in September to November 2010. They also suggested that the advection of saltier water exceeded the increase in local rainfall in the upper layer and that a thick mixed layer did spread in SETIO during the period. Whereas the eastward ocean current anomalies (Figure 1) and the observed high salinity water with a lower vertical gradient (Figure 3) were consistent with that reported by Qiu et al. [2012], our temporally high-resolution salinity time series provides an alternative perspective. From the buoy observations, we found MLD shoaling and barrier layer evolution possibly due to local rainfall. The relatively low salinity signal around mid-September to early October (Figure 3b) was associated with sub-daily- to daily-scale surface water freshening (figure not shown). The data strongly suggest that local rainfall contributed to forming the upper layer halocline. Actually, barrier layers are patchily distributed in the synoptic field in the tropical ocean [e.g., Sato et al., 2006]. Although the 10 m shoaling of the average MLD made a minor contribution to the mixed layer heat balance, as we mentioned in section 'Temperature Balance', these daily- to weekly-scale salinity variations and their possible contribution to a short time scale heat balance should be clarified in future research.

[25] Finally, we briefly describe a small anomalous warming event that occurred in the northern summer-fall of 2005. The Indian Ocean SST anomaly in the eastern domain (Figure 1) reached +1.0 °C in September 2005. Our heat balance analysis for the 2005 warming event indicates that a similar process acted to warm the mixed layer temperature in SETIO, while the contribution of anomalously weak latent heat flux loss was half that of anomalous meridional advection. Because the amplitude of the 2005 warming event was about half that of the 2010 event, a large noise ratio may preclude qualitative assessment of the importance of the air-sea heat flux and horizontal heat advection. Nevertheless, the similar results for the 2005 event support the conclusion that both heat flux and horizontal heat advection contributed to the SETIO warming event.

6 Summary and Conclusions

[26] In August–October 2010, warm SST anomalies associated with the negative phase of the IOD occurred in the southeastern part of the tropical Indian Ocean and persisted more than 3 months from August to October. To understand the ocean-atmosphere interactions related to the anomalous SST, we analyzed the mixed layer temperature balance in SETIO. Temporally high-resolution data from a moored buoy revealed ocean temperature evolution and intraseasonal fluctuations during a particular period of the negative IOD. The results demonstrate that both air-sea heat fluxes and horizontal heat advection accounted for the mixed layer temperature variation. Consistent with previous studies, an anomalously weak latent heat flux had a major role in maintaining the SST anomalies in September. The anomalous latent heat flux was associated with a weakening wind in the fall of 2010 as the result of a weak southeasterly monsoon wind from August to October. We also found that the meridional current transported anomalously warmer water from north to south and contributed to the warming of the mixed layer temperature from June to August. The decay of the SETIO warming events also occurred by net surface heat flux and horizontal heat advection. From October to December, enhanced latent heat loss due to strengthening wind speed, accompanied by suppressed shortwave radiation, anomalously cooled the mixed layer temperature. Both zonal and meridional heat advection also had a cooling effect on the mixed layer temperature for the period.

[27] The present observational study provides a fundamental example for quantifying the possible feedback that could warm the SST in SETIO. In particular, the results are useful for assessing the asymmetry and skewness of SST anomalies between warming and cooling events in SETIO, which previous studies have analyzed from reanalysis data sets and/or by numerical modeling. Although the results were based on spatially limited data, satellite and reanalysis products suggest that the heat balance should be robust in the wider region of SETIO, including the buoy locations. Because it is still quantitatively unclear if the present thermodynamic balance is general in the coastal area, our next goal will be to diagnose the heat balance at the regional scale near the coasts of Sumatra and Java. In addition, continuous observations in the 21st century will be needed not only to understand seasonal to interannual variation but also to elucidate whether a longer timescale warming trend is occurring in the Indian Ocean [e.g., Du and Xie, 2008].

Appendix A

This appendix describes the data used in this study and the estimated errors in our analysis. The satellite-based and reanalysis data sets were validated by comparison to direct observations from the buoy or to buoy-based fluxes calculated from bulk variables and the COARE 3.0 algorithm. The periods of comparison differed for some components because of periods of missing data. We then estimated errors for each term in the mixed layer temperature balance in SETIO.

A1 Validation of Flux

To assess the satellite-based and reanalysis data, we used the buoy observations and the buoy-based flux calculations containing measurement errors. These errors depended on the accuracy of the sensors and the temporal and spatial sampling resolution. The errors for the meteorological observations made by the buoy were 0.3 m/s for the wind speed, 5% (or 8.6 W/m2) in RMS error for the shortwave radiation, 2% for the relative humidity, 0.1 °C for the air temperature, and 0.002 °C for the ocean temperature. These sensor accuracies produced RMS errors in the bulk fluxes of 2.1 W/m2 for longwave radiation, 1.6 W/m2 for sensible heat flux, and 10.5 W/m2 for latent heat flux.

The ISCCP shortwave radiation data averaged for the 1° × 1° box centered on 5°S, 95°E were assessed with direct buoy observations based on an 11 day average. The period was from December 2006 to December 2009, when a relatively long time series was available from buoy observations. The RMS difference (correlation) was 19.2 W/m2 (0.83). Because we used a reconstructed shortwave time series for 2010, a larger error should be taken for the period. We compared the OLR-based shortwave radiation time series with buoy observations for the period from December 2006 to June 2010. The RMS difference (correlation) was 22.1 W/m2 (0.80).

Turbulent fluxes from OAFlux and longwave radiation from NCEP/NCAR were assessed in the same way. The periods examined were from July 2004 to 31 December 2009 for the latent heat, from July 2004 to July 2010 for the sensible heat, and from December 2006 to December 2009 for the longwave radiation. The RMS differences (correlations) were 16.3 W/m2 (0.83) for latent heat, 2.7 W/m2 (0.60) for sensible heat, and 6.9 W/m2 (0.13) for longwave radiation.

Given the above estimates, errors in the variables were assessed as the RMS differences from these comparison experiments. Although these values are not the actual observational errors, they are two to four times as large as the measurement errors. As a conservative estimate, we applied these errors to the heat balance analysis.

A2 Validation of Horizontal Heat Advection

The OSCAR current data and the horizontal temperature gradient calculated from the TMI SST were assessed in the same way as by Horii et al. [2009, see their appendix]. Because we averaged the horizontal heat advection for the area 93°E–97°E and 7°S–3°S, which is centered on 5°S, 95°E, we checked the spatial correlation scale and selected “2” as the degree of freedom. The average errors of the zonal and meridional advections were 0.19 °C/month and 0.33 °C/month, respectively.

A3 Errors in Mixed Layer Temperature/Depth

Errors in the mixed layer temperature were negligible because the error for the ocean temperature observations was relatively small, only 0.002 °C. The MLD should have a larger error because the estimate was based on buoy profiles of coarse vertical resolution mooring data. The error in the MLD had the largest influence on the first term on the right-hand side of equation ((1)). We estimated these errors by comparison with eight CTD observations from locations near the buoy. The RMS difference of MLD was 7.7 m and we used this value as the error of the MLD.

A4 Errors in Heat Balance

We here discuss the total errors in our heat balance analysis. In addition to observation errors, we also considered errors caused by analysis assumptions. Because we treated surface (10 m) current observations as the mean velocity in the mixed layer in our heat balance analysis, we added an additional 5.0 cm/s to the maximum error to account for this assumption, following Wang and McPhaden [2000].

All of these errors were combined using the sampling theory for propagation of error by Emery and Thomson [1998]. See the appendix of Wang and McPhaden [2000] for the detailed procedure. The errors in our heat balance analysis are shown by the vertical bar and light shading in Figures 8 and 9. The average errors in the net heat flux and horizontal heat advection were 0.57 °C/month and 0.40 °C/month, respectively. Although these errors are not negligible, the amplitudes of these errors do not negate our conclusions.

Acknowledgments

We thank all of the members of the R/V Mirai and R/V Kaiyo cruises and also thank the data processing team for the TRITON buoy cruise operations for their data management efforts. Yasushi Ishihara, Takeo Matsumoto, Masayuki Yamaguchi, and Nobuhiro Fujii greatly contributed to the development, deployment, and data management of the m-TRITON buoy. We also thank Lisan Yu for providing the OAFlux data set, William B. Rossow for providing the ISCCP-FD shortwave radiation data set, and the NOAA-Cooperative Institute for Research in Environmental Sciences (CIRES) Earth System Research Laboratory (ESRL)/Physical Sciences Division (PSD) for providing reanalysis flux and wind data. TMI SST data were produced by Remote Sensing Systems (available online at http://www.remss.com). OSCAR velocity data were provided by Earth and Space Research. The altimeter products were produced by Ssalto/Duacs and distributed by Aviso with support from Cnes. The Argo float data used in this study were collected and made freely available by the International Argo Project and the national programs that contribute to it (http://www.argo.ucsd.edu, http://wo.jcommops.org/cgi-bin/WebObjects/Argo). We thank two anonymous reviewers for helpful comments and suggestions. We also thank I. Iskandar for helpful discussions.