Mechanism of intraseasonal oceanic signature in the region off southern tip of India during boreal summer


Correspondence to: A. Jayakumar, Indian Institute of Tropical Meteorology, Dr. Homi Bhabha Road, Pune 411 008, Maharashtra, India. E-mail:


Sea surface temperature (SST) in the region off southern tip of India (STI, 75–83°E, 5–8°N) exhibited a prominent variability in the intraseasonal time scale (both 30–90 d and 10–30 d band) during boreal summer. Mechanisms associated with this intraseasonal variability are studied using three-dimensional ocean general circulation model (OGCM) sensitivity experiments, satellite observed outgoing longwave radiation, SST and winds for the period 1998–2007. The background oceanic structure of the STI characterized by a shallow thermocline and moderate mixed layer provided ideal conditions for strong oceanic sub-surface processes. The model mixed layer heat budget reveals that the oceanic processes such as horizontal advection and vertical processes are the dominant mechanisms in the STI region as compared with air-sea flux. Sensitivity experiments with the OGCM reveals that the ocean dynamical processes contribute to most of the intraseasonal SST variability and the wind stress contributes to 85% of the variability whereas surface flux contributes to only 15% for the 30–90 d SST variability. Higher amplitude of surface flux perturbation and its contribution to SST in the 10–30 d as compared with the 30–90 d band are evident in the model experiment and are consistent with the observational analysis. There is year-to-year variability in the relative role of horizontal and vertical processes for different intraseasonal SST events over STI.

1. Introduction

Winds over the tropical Indian Ocean exhibit a seasonal reversal during boreal summer, accompanied by prominent upper ocean variability on different spatial and temporal scales. The seasonal upper ocean circulation in the region off southern tip of India (STI, 75–83°E, 5–8°N) reflected a complex feature, by linking Bay of Bengal with rest of the Indian Ocean through passage of eastward southwest monsoon current and Rossby wave interaction (Schott et al., 1994). Forcing mechanism of this seasonal monsoon current around Sri Lanka and STI (McCreary et al., 1993; Vinayachandran and Yamagata, 1998; Shankar et al., 2002) and its intraseasonal fluctuation (Sengupta et al., 2001) are well documented.

Using Tropical Rainfall Measuring Mission Microwave Imager (TMI), Rao et al. (2006) observed sub-monthly (10–30 d band) variability in sea surface temperature (SST) over STI. They attributed the variability in this upwelling region to changes in the surface divergence and wind mixing. SST variability in this region is influenced by the active-break cycle of the summer monsoon over the Bay of Bengal (Vecchi and Harrison, 2002; Joseph et al., 2005; Joseph and Sabin, 2008; Vialard et al., 2012). Ganer et al. (2009) speculated that the evaporative cooling along with the upwelling associated with the active-break cycle is responsible for this variability. In addition to the local forcing the dynamical remote forcing may have some role in the intraseasonal variability in this upwelling region (Smitha et al., 2008; Webber et al., 2010). But none of these studies looked at the processes associated with STI intraseasonal SST variations in detail. Duncan and Han (2009) carried out model investigations to understand SST variability over areas of maximum intraseasonal variance over Indian Ocean domain, but they have not conducted detailed analysis over STI, as their model performed poorly over this region. Our earlier studies using the ocean general circulation model (OGCM) showed good agreement with the TMI SST over STI in the 30–90 d band (Jayakumar et al., 2011; Jayakumar and Gnanaseelan, 2012; Vialard et al., 2012). Further Vialard et al. (2012) showed that STI is one of the key regions where intraseasonal SST variability is very significant. These help in the model fidelity in understanding the SST variability over STI using Modular Ocean Model Version 4 (MOM4) (Griffies et al., 2004).

The intraseasonal signature in SST over STI is widely influenced by oceanic processes associated with wind forcing, advection through summer monsoon current and the surface flux changes (Rao et al., 2006; Ganer et al., 2009), hence numerical simulations using an OGCM will give insight into intraseasonal SST mechanism by quantifying aforementioned processes better. We studied the mechanisms associated with intraseasonal SST variability in STI by quantifying the relative role of wind-driven oceanic processes and surface flux during the evolution of the event with carefully designed sensitivity experiments (Table 1). Section '2. Data and methodology' presents the data, methodology and description of modelling experiments. In Section '3. Results', we elucidate the observed patterns of SST variability in both the intraseasonal bands (10–30 and 30–90 d) and associated surface flux variations. Further, we quantified the different processes that are responsible for SST variability using OGCM sensitivity experiments and finally in Section '4. Summary', we summarized our results.

Table 1. List of experiments used in this study
CTLFull forcing
NO_ISO_FLXLow-pass filtered shortwave and non-solar heat fluxes
NO_ISO_STRESSLow passed 120 d filter on wind stress
NO_ISOLow passed shortwave and non-solar heat fluxes and low passed wind stress


2. Data and methodology

2.1. Data

The interpolated outgoing longwave radiation (OLR) from NOAA is used to identify the regions of atmospheric deep convection and microwave SST product from Level 3 TMI is used to get SST measurement in cloudy condition. Ocean sub-surface temperature of model is validated using NCEP Global Ocean Data Assimilation System (GODAS) pentad data. For winds, we use gridded estimates of 10 m winds from the QuikSCAT scatterometer produced at IFREMER. To estimate the air–sea flux associated with the summer monsoon intraseasonal oscillations (ISO), we used latest version 1° ՠ1° global ocean-surface heat flux products developed by the Objectively Analyzed air–sea Heat Fluxes (OAFlux) project (Yu and Weller, 2007). The mixed layer depth and thermocline depth are two important parameters in the upper ocean response to surface forcing. For that purpose, the climatologies of mixed layer depth (MLD) from De Boyer Montgut et al. (2004) and ocean sub-surface temperature from the World Ocean Atlas 2005 (WOA05, Locarnini et al., 2006) are used.

2.2. Model experiments

The OGCM used in the present study is the Geophysical Fluid Dynamics Laboratory MOM4 (Griffies et al., 2004) and is set up for the region between 40°S–25°N and 30°E–120°E with 30 vertical levels. The upper ocean has 15 vertical levels within a depth of 155 m, and so the mixed layer and thermocline zones are well resolved. The zonal resolution is 1° and the meridional resolution varies from 0.33° at equator to 0.7° at 25°N and 1.5° at 40°S. Vertical mixing is based on the K Profile Parameterization (KPP) scheme (Large et al., 1994) which has significant impact on the diurnal coupling. Bottom topography is derived from the 5 min global topography Earth Topography-5 minute (ETOPO5) database. Temperature and salinity at the southern and eastern boundaries are restored to monthly climatologies of Levitus (1998). The model is initialized using temperature and salinity from Levitus (1998), and spun up for a 20 year period using climatological forcing (Large and Yeager, 2004). Subsequently, it is integrated over the 1958–1995 period using the interannual forcing data sets from the Common Ocean-ice Reference Experiments (Large and Yeager, 2004). The control experiment (hereafter CTL) is then run for the period 1996–2007 using the same interannual forcing. Air–sea fluxes are computed interactively via bulk formulae, using model SST and specified 10 m wind, air-temperature, specific-humidity, and shortwave and longwave radiation fields. Detailed model description and validation with the available satellite and buoy measurements at different timescales over the north Indian Ocean has been carried out by Thompson et al. (2006, 2008), Jayakumar et al. (2011) and Vialard et al. (2012).

Table 1 gives a list of sensitivity experiments carried out for this study. The NO_ISO STRESS experiment retains the full spectrum of heat flux forcing, but has low-passed wind stress forcing to eliminate the influence of intraseasonal Ekman pumping, mixing at the bottom of the mixed layer and entrainment. The NO_ISO_FLX experiment has full wind stress forcing, but all the solar and non-solar components of the heat flux forcing are low-passed to eliminate the influence of intraseasonal flux forcing. An experiment with both wind stress and heat fluxes low-passed was also performed and named NO_ISO. To evaluate the contribution of various processes to this intraseasonal variability using the above experiments:

equation image(1)

where the prime indicates the intraseasonal SST variability obtained by filtering in the time domain. The first two terms on the r.h.s. of Equation ((1)) are defined by the differences: SST′τ = SST′CTL− SST′NO_ISO_STRESS and SST′Q = SST′CTL− SST′NO_ISO_FLX; they estimate the contribution of wind-stress and heat-flux forcing to the intraseasonal variability, respectively. Term SST′F defined by SST′F = SST′NO_ISO is the intraseasonal variability remaining in the NO_ISO experiment. It arises from two sources: internally generated oceanic variability and intraseasonal freshwater forcing. It is important to note that the intraseasonal variability in fresh water is not considered in the forcing. The intraseasonal SST perturbation is linearized by internally generated oceanic variability term in addition to the dominant contributing term of flux and wind stress forcing. The SST′R contribution is computed as a residual, and arises mainly from non-linearities in the ocean response, that is called as ‘error’. Contributions from SST′F and SST′R are weak and weakly correlated to total SST perturbation. The modelling approach followed in this paper is very similar to that followed by Jayakumar et al. (2011) for the winter season and Vialard et al. (2012) for summer season. The regression coefficients are computed for the entire experiment for calculating the percentage of contribution by specific processes.

The daily anomalies are prepared based on the 1998–2007 climatology as this being the common period of TMI and model simulation. To isolate the signal associated with the sub-monthly (SUB) and Monsoon intraseasonal oscillation (MISO), we have used 10–30 d and 30–90 d band-pass filtering of daily anomalies with respect to the mean seasonal cycle, respectively. This is similar to the approach followed by Duncan and Han (2009).

3. Results

3.1. Observed intraseasonal SST signature over STI and controlling factors

The standard deviation of both TMI and CTL experiment for 30–90 d (a, b) and 10–30 d (c, d) filtered SST during summer (June–September) around Sri Lanka and STI is displayed in Figure 1. As the summer monsoon atmospheric ISO move northward, SST variations appear first at STI (Vialard et al., 2012). The air–sea coupling over this region has a strong impact on the further northward progression of the intraseasonal signals. So it is very important to understand the mechanism responsible for the intraseasonal SST variability in the STI. This is the main focus of the present study. Both the spectral bands 30–90 d and 10–30 d displayed prominent variability over STI (Red box) though the model underestimates the amplitude. Temporal variability in the 10–90 d SST is illustrated in Figure 1(e). The phase agreement between the model and observation further strengthen the fact that model is able to capture the intraseasonal variability in SST. Further the correlation between model and observation is 0.74 (0.66) for MISO (SUB). This shows that the model is successful in reproducing the large-scale SST variations in both bands in STI regions, despite its slightly weaker correlation in high frequency variability. The weaker correlation may be linked to the model coarse resolution and the error in forcing (both heat flux and wind field) as discussed in Jayakumar et al. (2011) and Vialard et al. (2012). Even though the meso-scale variability finer to this resolution may not be reproduced by the coarse model, the large scale variations associated with summer monsoon intraseasonal signals are well captured by the model.

Figure 1.

Standard deviation (SD) of band passed SST for the June–September during 1998–2007 period using (left) TMI observation and (right) CTL experiment selective for (a, c) 30–90 d and (b, d) 10–30 d bands. The red box indicates the STI region used in this study. Box average 10–90 d band passed SST for TMI SST observation (black) and CTL experiment (red) within STI region is shown in (e).

The background oceanic structure and the prevailing wind forcing pattern over STI are the major controlling factors for the intraseasonal SST variability in this region, thereby making this region highly responsive to the atmospheric ISOs and oceanic subsurface processes. Figure 2(a) shows the prevailing north-easterly surface wind circulation and deeper thermal structure over STI during winter. The oceanic temperature gradient from surface to 100 m depth (shaded) gives the idea of thermocline stratification. During June–September period, shallow thermocline depth is observed with maximum gradient near the tip of peninsular India (Figure 2(a)). Very strong seasonality in the thermal structure over STI, during winter (125 m) and summer (95 m) and MLD (contour) is observed to be minima over STI (25 m) during both seasons. High correlation between the thermocline depth with MLD (0.85) indicates that thermocline variability can modulate MLD in this region, thereby giving oceanic background state and thermocline SST feedback as the major requirement for intraseasonal variability. The south-westerly wind during June–September is rather tangential to the STI region (Figure 2(b)). This induces coastal upwelling by Ekman transport brought out by the alongshore component of the wind stress (Smitha et al., 2008). As the wind direction reverses during winter the seasonal upwelling will be suppressed.

Figure 2.

Ocean temperature difference between surface and 100 m ( °C, shaded) using WOA05, Mixed layer depth (MLD, m, contour) from De Boyer Montgut et al. (2004) and wind stress (N m−2, red) for JJAS period are shown. Dark orange colour indicates the shallow thermocline depth region possibly having strong upper temperature gradient over upper 100 m.

Figure 3 shows the vertical extent of the intraseasonal signals from the upper 100 m of the model for one of the strong intraseasonal cooling event during our experimental period. The thermal structure of this events is comparable with GODAS (both are in pentad time scale) (left panels). The model well captured the passage of intraseasonal events to the sub-surface (right panels). The shoaling of the thermocline along with the eastward advection of cooling is shown in Figure 3. The thermal stratification near the bottom of the mixed layer is a critical parameter for entrainment and Ekman pumping. The upward movement of thermocline (26 °C isotherm depth) is found to be seen during the cooling event along with the SST change (mainly cooling), which may infer the role of sub-surface processes in such variability. Temporal evolution of thermal structure at different longitudinal location over the 5–8°N band showed the advection of the sub-surface cooler water. The shoaling (cooling) observed in the 78°E during the end of July can be persisted until mid September at 84°E, which coincide with the eastward advection of sub-surface water mass (Figure 3(c) and (f)). Thus the role of sub-surface dynamic processes and the advection associated with the intraseasonal events are quantified better using a mixed layer model and relative role of air–sea flux and oceanic processes is studied to understand the forcing mechanism.

Figure 3.

The upper 100 m temperature profile for the averaged [5–8°N] region using GODAS (left panels) and Model (right panels) along (a, d) 75°E, (b, e) 78°E and (c, f) 84°E during May–October 2002.

3.2. Mechanism associated with intraseasonal events

OAFlux and TMI SST are used to isolate the typical atmospheric and surface heat flux perturbations associated with the observed intraseasonal SST variability in both bands (30–90 and 10–30 d) over STI (Figure 4). SST in both bands showed similar amplitudes, though the time integral of forcing is longer in the MISO than SUB. Thus it strongly suggests the dependency on the strength of the forcing parameter. This can be inferred from the amplitude of the heat flux and its components (mainly shortwave flux) from both bands (Figure 4(d)). The net heat flux perturbation is 16 W m−2 in 10–30 d whereas it is 8 W m−2 in 30–90 d band. This difference could be due to the relatively higher perturbations in both convection and wind speed in the sub-monthly time scale. Latent heat flux (green dash curve) associated with changes in the wind speed appears to have higher amplitude in SUB than MISO scale (Figure 4(d)). Further the phases of different flux components also play key role to determine the intensity of the intraseasonal signals. In the sub-monthly case they are added positively to the net heat flux perturbation. Thus role of surface heat flux is observed to be seen higher in the intraseasonal events mainly in the SUB.

Figure 4.

STI average 30–90 d (black) and 10–30 d (red) filtered (a) OLR (b) SST and (c) Wind speed regressed to the normalized 30–90 d and 10–30 d SST, respectively. (d) Heat flux (black) and its components: shortwave (red), latent (green), sensible (blue) and longwave (purple) for 30–90 d (thick) and 10–30 d (dash) filtered similarly regressed to the aforementioned normalized SST filtered bands.

In order to understand the processes that are responsible for the intraseasonal SST events in the model in addition to the role of surface heat flux changes, the heat balance of the mixed layer has been computed as in De Boyer Montgut et al. (2007), and Jayakumar and Gnanaseelan (2012). The mixed layer temperature (Tml) tendency can be written as the net effect of surface heat flux, horizontal advection, and vertical processes (Equation ((2))). The vertical processes include the effects of entrainment, vertical advection and diffusion.

equation image(2)

Here h is the time-varying depth of the model mixed layer, u, v, w are the components of ocean currents. Th, and wh are the temperature and the velocity at the base of the mixed layer; Kz is the vertical mixing coefficient for tracers; ρ is the seawater reference density; and Cp is the seawater heat capacity. Qs is the total surface flux available in the mixed layer after considering the fraction of shortwave radiation that penetrates through the base of the mixed layer.

In this calculation, near-closure of the heat budget is achieved. We carried out mixed layer heat balance analysis for the full experimental period and in particular we have selected few anomalous events (1999, 2001 and 2002) which showed maximum intraseasonal SST variance (as in the study of Rao et al., 2006). Figure 5 shows that there are year to year variation in the role of horizontal and sub-surface processes, but these processes are dominating the intraseasonal SST variability whereas the air–sea flux plays a secondary role. These events are found peaking in different months during different years; e.g. during 2002, it peaked in July, whereas August in 2001 and early June in 1999. This could have strong interaction with the active-break cycle of monsoon. The maximum cooling tendency was around 4 °C month−1 for the year 2002 and 2001, where vertical processes dominate over horizontal advection in former and vice versa in later events (Figure 5(a) and (b)). It is difficult to explain this drastic cooling by heat flux alone. Further there is significant event-to-event variability in the details of the spatial and temporal patterns of different intraseasonal events associated with wind variations (see e.g. Duvel and Vialard, 2007).

Figure 5.

Mixed layer temperature tendency ( °C month−1) (black) and its contributions from air–sea flux (red), horizontal advection (blue) and sub-surface processes (blue) for intraseasonal events in (a) 2002, (b) 2001 and (c) 1999. Nine-day running mean is applied to avoid the short period oscillation lesser than 9 d.

It is evident from Figure 5 that the ocean dynamic response (both advection and vertical processes) is dominating intraseasonal SST variability, but the air–sea flux plays a secondary role. It is difficult to quantify the relative contribution of net heat fluxes and ocean processes to SST intraseasonal variability from observations and CTL solution alone, as it is widely correlated to each other (Jayakumar et al., 2011; Vialard et al., 2012). The series of sensitivity experiments described in Section '2. Data and methodology' help us to evaluate the respective contributions of heat fluxes and wind stress, in both 30–90 d and 10–30 d SST variability in the model. Table 2 provides the contributions to the average intraseasonal SST variability in these two bands over STI. For both spectral bands, wind stress associated processes (include both advection and vertical processes) dominate the intraseasonal SST variability with contribution of about 95% for 30–90 d and 75% for 10–30 d band. Remaining portion of the SST variance is accounted by heat flux perturbations, internal variability and non-linearity. The relatively higher role of surface flux in 10–30 d spectral band in the model experiment is consistent with the observational analysis as well (Figure 4).

Table 2. Regression coefficient of the 10–30 d and 30–90 d average SST variability associated with heat flux forcing and wind stress forcing (see text for details)
Process10–30 d SST ( °C)30–90 d SST ( °C)
All processes (CTL experiment)1.001.00
Heat flux0.240.04
Wind stress0.750.95

Process study of the intraseasonal SST in the 30–90 d band showed the higher role (85%) of wind stress changes and relatively weak role in flux changes. Figure 6(a) shows the map of wind stress contribution to the intraseasonal SST, with maximum values over STI (CTL minus NO_ISO_STRESS). This shows the significance of ocean dynamic response to the wind stress over this region. This ocean dynamic response is associated with both local processes and remote forcing. The remote effect is calculated as the difference between the model CTL vertical velocity at the Ekman layer depth and the local wind-induced Ekman pumping speed, which is calculated from the model wind stress. Figure 6(b) shows the variability in Ekman divergence associated with the changes in the wind stress curl and the along shore wind stress as discussed by Rao et al. (2006) and Smitha et al. (2008). Variability in remote forcing calculated in the model (Fig 6(c)) shows significant signals over STI and it may be due to both coastally trapped Kelvin waves, and the offshore propagating Rossby waves. The northward propagating coastal Kelvin wave circumnavigates the Bay of Bengal and pass into Arabian Sea through tip of India (Webber et al., 2010). The westward offshore propagating Rossby waves can interact with coastal Kelvin wave, and modulate the SST over STI region.

Figure 6.

(a) SD of intraseasonal SST for CTL minus NO_ISO_STRESS for the JJAS period. In this region the maximum deviation value represents stronger impact of intraseasonal wind stress associated processes. (b) SD of Ekman pumping (EkP) anomaly represents the local upwelling/downwelling effect. (c) SD of remote forcing (RF) anomaly shows the influence of coastal waves and the high frequency mixed Rossby wave using model vertical velocity at Ekman layer depth within STI region.

4. Summary

We have studied the intraseasonal events in the region off STI during the boreal summer using satellite observation and an ocean model. There is prominent display of intraseasonal SST signature in both 10–30 d and 30–90 d bands. The background oceanic condition for the average June–August period shows a shallow thermocline and mixed layer thickness which support the sub-surface processes. The temperature tendency within the mixed layer showed the dominant role of oceanic processes over air–sea flux. There is interannual variation in the supremacy of horizontal and vertical processes over the intraseasonal cooling events. Our modelling approach using sensitivity experiments linearize the intraseasonal SST perturbation by internal generated oceanic variability term in addition to the dominant contributory term of flux and wind stress forcing. Result shows the dominant role of dynamic ocean response to the intraseasonal wind stress (85%) and weaker response to the surface flux impact for both 30–90 d and 10–30 d bands. The wind stress impact in this region may be influenced by both local upwelling and remote waves induced by the intraseasonal wind forcing. Our experiment using forced ocean model showed the importance of ocean dynamic over STI. The modulation of intraseasonal SST within STI region by remote forcing and local wind stress associated processes need to be isolated with suitably designed sensitivity experiments and which will be beyond the scope of this paper. The roles of ocean–atmosphere interaction and dynamical response in the evolution of the intraseasonal SST in this region have implications on the circulation and biology in equatorial Indian Ocean.


A. J. thanks the Council of Scientific and Industrial Research (CSIR), and Nansen Environmental Research Center India for providing Research Fellowships. The authors thank Prof. B. N. Goswami, Director, IITM, for his encouragement and interest in the work. Financial support of Space Application Centre (SAC), Ahmedabad, India and Indian National Centre for Ocean Information Services (INCOIS), Hyderabad, India is acknowledged. We also acknowledge the anonymous reviewer for the constructive suggestions. TMI gridded SST data were produced by Remote Sensing Systems and are available at and QuikSCAT wind from IFREMER site.