Time series measurements of temperature, salinity and surface meteorological parameters recorded at 8°N, 90°E in the southern central Bay of Bengal (BoB) from a Research Moored Array for African-Asian-Australian Monsoon Analysis and predication (RAMA) buoy are used to document temperature inversions and their influence on the mixed layer heat budget during the winters, defined as October to March, of 2006–2007 (W67) and 2007–2008 (W78). There is a marked difference in the frequency and amplitude of temperature inversion between these two winters, with variations much stronger in W78 compared to W67. The formation of temperature inversions is favored by the existence of thick barrier layers, which are also more prominent in W78 compared to W67. Inversions occur when heating in the barrier layer below the mixed layer by penetrative shortwave radiation is greater than heating of the mixed layer by net surface heat flux and horizontal advection. Our analysis further demonstrates that intraseasonal and year-to-year variability in the frequency and magnitude of temperature inversions during winter have substantial influence on mixed layer temperature through the modulation of vertical heat flux at the base of mixed layer.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Sea surface temperature (SST) in the Bay of Bengal (BoB) has significant influence on the climate and monsoon precipitation over the surrounding land masses [Li et al., 2001; Shenoi et al., 2002; Vecchi and Harrison, 2002; Shankar et al., 2007; Jiang and Li, 2011]. Several researchers have previously tried to understand SST variability in the BoB through the mixed layer heat budget analyses using observations [Rao and Sivakumar, 2000; Sengupta et al., 2002; Shenoi et al., 2002; Parampil et al., 2010] and modeling [De Boyer Montégut et al., 2007]. These studies have highlighted the dominant role of net surface heat flux on the modulation of SST on different time scales. Further, several studies have reported the existence of large amplitude intraseasonal and interannual SST variability in the BoB [Sengupta and Ravichandran, 2001; Vecchi and Harrison, 2002; Shenoi et al., 2002; Rao et al., 2006; Duncan and Han, 2009; Vialard et al., 2012].
 In the Bay, surface salinity is relatively low compared to the subsurface layer because of large fresh water fluxes from intense precipitation over evaporation and redistribution of low-saline river discharge water by horizontal circulation (Figure 1). Thus, there is strong haline stratification in the near surface layer [Rao and Sivakumar, 2003; Thadathil et al., 2007; Girishkumar et al., 2011] often associated with an intermediate “barrier layer”, i.e., the layer between the base of the mixed layer and the top of the thermocline [Lukas and Lindstrom, 1991; Rao and Sivakumar2003; Thadathil et al., 2007; Mignot et al., 2009; Agarwal et al., 2012]. The barrier layer, as its name implies, acts as a barrier to turbulent mixing of cooler thermocline waters into mixed layer and thereby plays an important role in the ocean surface layer heat budget [Lukas and Lindstrom, 1991]. Barrier layers can help to explain the warm SSTs (>28°C) in the BoB [Shenoi et al., 2002]. Rao et al.  have also shown that the presence of a highly salt stratified near surface layer can limit the effectiveness of thermocline depth variations associated with Indian ocean dipole (IOD) [Saji et al., 1999; Webster et al., 1999] in producing SST anomalies in the BoB.
 In addition, barrier layers can contain inversions in the vertical temperature profile [Anderson et al., 1996; Mignot et al., 2012]. Several previous studies have reported the existence of temperature inversions in the BoB, particularly during the winter monsoon [Rao et al., 1983; Suryanarayana et al., 1993, Shetye et al., 1996; Vinayachandran et al. 2002; Thadathil et al., 2002; Thompson et al., 2006], which coincides with the peak season for thick barrier layers [Rao and Sivakumar, 2003; De Boyer Montégut et al., 2007]. However, these earlier studies were based on sparse data collected during individual ship surveys. Using individual XBT track data collected along the Andaman-Calcutta section during the international Tropical Ocean and Global Atmosphere (TOGA) programme, Thadathil et al.  examined the interannual variability of temperature inversions in the BoB during winter. They reported that this variability was large and caused by interannual variability in fresh water fluxes and surface cooling in the northern BoB. However, due to lack of systematic measurements of temperature and salinity with high-temporal resolution, most of these studies focused mainly on the seasonal time scale. Hence, much less is known about intraseasonal variability of the barrier layer and how that variability changes from year to year in the Bay.
 In the BoB, temperature inversions supported by barrier layers affect surface temperature by entrainment of warm subsurface water into the mixed layer [De Boyer Montegut et al., 2007]. Using a numerical model, De Boyer Montégut et al.  studied the seasonal and interannual variability of mixed layer temperature (MLT) in the northern Indian Ocean. They showed that interannual barrier layer variations do have an impact on SST by affecting vertical processes in the Bay, with thicker barrier layers linked to positive SST anomalies. They also suggested the need of further studies to understand SST variability on intraseasonal timescales. However, again due to a lack of systematic temperature, salinity and surface meteorological measurements with high-temporal resolution, previous studies had not examined the role of intraseasonal variability and its year-to-year variation in barrier layer thickness (BLT), and how those variations affected temperature inversions and their influence on MLT in the Bay.
 Using RAMA temperature and salinity data in the southern BoB (8°N, 90°E), Girishkumar et al.  reported on the existence of intraseasonal and year-to-year variability in the BLT. Their analysis further showed that BLT was modulated primarily by vertical stretching of the upper water column associated with westward propagating intraseasonal Rossby waves, which are forced primarily by intraseasonal wind variability (40–100 days) in the Equatorial Indian Ocean (EIO). This large intraseasonal and year-to-year variability of BLT in the south central BoB presumably influences the modulation of temperature inversions and hence the evolution of SST in this region. However, their study did not address the role of this intraseasonal and year-to-year variability in the BLT on temperature inversions and the mixed layer heat budget.
 The main aim of the present study is to investigate these relationships at 8°N, 90°E in the south central BoB for the winters (defined as October to March) of 2006–2007 (W67) and 2007–2008 (W78). We consider these two winters primarily due to the simultaneous availability of high-temporal resolution oceanic and meteorological parameters and presence of greatly contrasting mean climatological conditions during these periods in terms of the IOD [Saji et al., 1999; Webster et al., 1999] and the El Nino/Southern Oscillation (ENSO) [McPhaden et al., 2006]. Earlier studies have reported the occurrence of a strong positive IOD and weak El Niño conditions during the W67 and weak positive IOD and moderate La Niña during W78 [Vinayachandran et al., 2007; McPhaden, 2008; Rao et al., 2008; Cai et al., 2009; Singh et al., 2011]. Cai et al.  further reported that after October 2007, the Indian Ocean was under the influence of a La Niña. The seasonal average (October to March) of wind stress anomalies, sea surface height anomaly (SSHA) and depth of 23°C isotherm (D23; proxy for thermocline) Figure 2 shows the tendency for easterly wind anomalies in the EIO, negative SSHA and shallow thermocline in the EIO and in the BoB (particularly in the southern part of the BoB) during W67. Opposite tendencies occurred during W78, consistent with expectations based on ENSO and IOD variability during these two periods [Saji et al., 1999; Rao et al., 2002; Vinayachandran et al., 2007; Cai et al., 2009; Xie et al., 2002; Girishkumar et al., 2012; Gnanaseelan et al., 2012; Aparna et al., 2012]. In this study, our focus will be on understanding year-to-year differences in temperature inversions at the 8°N, 90°E buoy location within this large scale context and how those inversions influence the mixed layer heat budget in W67 and W78. The site we have chosen is located in a region of significant mean BLT during winter where temperature inversions occur often between November and February [Mignot et al., 2009].
2. Data and Methodology
 The RAMA buoy in the south central Bay at 8°N, 90°E (Figure 1) [McPhaden et al., 2009] provides daily time series of temperature at depths of 1, 10, 13, 20, 40, 60, 80, 100, 120, 140, 180, 300, and 500 m and of salinity at depths of 1, 10, 20, 40, 60, and 100 m. We consider measurements at 1 m nominally as from the surface. The data are linearly interpolated in the vertical to 1 m intervals to facilitate analysis.
 Following Rao and Sivakumar , the mixed layer depth (MLD) is defined as the depth where the density is 0.125 greater than surface value (1 m). The isothermal layer depth (ILD) is defined as the depth where the temperature is 0.5°C lower than SST. This definition of ILD is consistent with the 0.125 density criterion so that when the there is no salinity variation the mixed layer and isothermal layer depths are identical. BLT is defined as the difference of ILD and MLD. Temperature inversions are defined to occur when temperature at depth is greater than SST by 0.1°C. Instrumental accuracy for RAMA temperature measurements is better than or equal to ±0.02°C, which is much lower than our threshold for defining temperature inversions.
 To demonstrate how accurately temperature inversions, BLT, MLD and ILD can be estimated from these buoy data, which have fairly coarse vertical resolution, we have selected 225 profiles, 62 of which show temperature inversions, from nearby Iridium Argo floats that have typically 2 m vertical resolution (i.e., higher vertical resolution than the normal 10 m resolution in the upper 200 m for most Argo floats). These floats moved typically within 5° of the buoy location (Figure 1). Temperature inversions, BLT, MLD, and ILD we recomputed from these data and compared to the same quantities estimated from Argo data sampled at RAMA buoy sensor depths. Differences between these estimates provide an idea of how well these parameters can be derived from RAMA buoy data. The root mean square difference for temperature inversions, BLT, MLD, and ILD are 0.12°C, 4.9 m, 5.8 m and 5.9 m, respectively. These differences are generally small relative to the amplitude of the variations observed in the BoB, so the above analysis indicates that RAMA buoy measurements provide reasonable estimates of these quantities despite their relatively coarse vertical resolution. We also selected nearby Argo float profiles to complement the analysis of subsurface temperature structures from the buoy data. The locations of the specific Argo float profiles near to RAMA buoy location (within 2° latitude and 3° longitude) are shown in Figure 1 during W67 (red circles) and W78 (pink circles). The temporal and vertical resolution of data from this Argo float (WMO ID 4900670) is 10 days and 10 m, respectively.
 The mixed layer heat budget is examined using the expression given by [Rao and Sivakumar, 2000]
 The individual terms of equation (1) represent (a) temperature tendency, (b) net surface heat flux, (c) horizontal advection, vertical processes (sum of entrainment (d) and vertical diffusion (e)) and (f) residual. The average mixed layer temperature is designated as T, ρ is the density of seawater, Cp is specific heat capacity of seawater, t is time, h is MLD (m), Th is the temperature of water entrained into the mixed layer, taken to be temperature at 5 m below MLD [Du et al., 2005], and Wh is the vertical advection below mixed layer inferred from the vertical displacement of isotherms in the thermocline as suggested by [McPhaden, 1982; McPhaden and Hayes, 1991]. The depth of thermocline is often estimated by the depth of a representative isotherm often times chosen to be the 20°C isotherm [Yu, 2003]. However, this isotherm exhibits large spatial variations [Yang and Wang, 2009]. To choose an appropriate representative isotherm for thermocline depth in our particular region, we computed the vertical temperature gradient at the buoy location. The analysis shows that the strongest vertical temperature gradient occurs around the depth of the 23°C isotherm, with a mean value of approximately 0.26°C m−1. Thus, we have chosen the depth of 23°C isotherm as an indicator of thermocline depth in the BoB for this study, with vertical velocity (Wh, in m d−1) below mixed layer inferred from the time rate of change of the 23°C isotherm depth. The entrainment velocity (m d−1) at the base of the mixed layer is estimated from Wh and the rate of change of MLD (∂h/∂t). H is the Heaviside step function [=0 if (Wh + dh/dt) < 0, =1 if (Wh+dh/dt >0]. ∂T/∂z is the average vertical temperature gradient between the base of mixed layer and 5 m below the mixed layer; Kz is the vertical diffusion coefficient and is taken as 1 cm2s−1 in this study. The vertical heat flux calculation will be sensitive to the exact choice of Kz and Wh. However, estimates of vertical heat flux for a range of Kz (0.1 cm2 s−1 to 2 cm2 s−1) and a range of isotherm depths (24°C and 26°C) indicate that the results do not change greatly, with a maximum root mean square difference relative to our default parameter choices of 11.8 W m−2 (for Kz =2 cm2 s−1) and 4.6 W m−2 (for 26°C isotherm) for the diffusion and entrainment terms, respectively.
 Latent (QLatent) and sensible (QSensible) heat fluxes are estimated from the Coupled Ocean-Atmosphere Response Experiment (COARE) bulk flux algorithm [Fairall et al., 2003] using the mooring SST, air temperature, relative humidity, and wind speed. The buoy does not have a longwave radiation sensor; hence net longwave radiation (Qlongwave) from TropFlux is used [Praveen Kumar et al., 2012]. Net surface shortwave radiation (QShortwave) is obtained from downwelling shortwave flux measured on the mooring and corrected for albedo at the sea surface (downwelling shortwave radiation × 0.945). Downwelling shortwave radiation from the RAMA buoy shows abnormally low values during March 2007, hence heat flux data during that time were replaced with data from TropFlux. Qnet is the net surface heat flux term (Qshortwave−Qpen+Qlongwave+QLatent+QSensible). Following Morel and Antoine  and Sweeny et al. , the penetrating shortwave radiation (Qpen) below the mixed layer is estimated by Qpen =0.47 and Qshortwave [V1e−h1/ζ1 + V2e−h2/ζ2], where ζ1 and ζ2 are the attenuation depths of long visible and short visible and ultraviolet wavelengths, and h is the MLD in meters. The values of V1, V2, ζ1, and ζ2 are estimated from MODIS 8 day composite chlorophyll-a (mg m−3) data using the method from Morel and Antoine . Approximate values of these parameters at the buoy location during the study period are 0.38, 0.61, 1.52, and 18 m, respectively, which are within the prescribed range as suggested by Sweeny et al. .
 We assume that the temperature is approximately uniform from the surface to the base of the mixed layer, so that optimal interpolated (OI) TRMM Microwave Imager (TMI) + Advanced Microwave Scanning Radiometer for EOS (AMSR-E) SST product with quarter degree resolution (∼25km) [Gentemann et al., 2004] is equivalent to MLT when computing the horizontal advection term in equation (1). A statistical comparison between OI SST and MLT shows reasonably good agreement with a correlation of 0.93 and root mean square difference of 0.20 (°C). Following Vialard et al. , OI SST averaged over 50 Km (four-grid point) on either side of buoy location were used to estimate the horizontal gradient of SST. The zonal (U) and meridional (V) component of current measurement at 10 m is obtained from the RAMA buoy. We did not consider horizontal diffusive heat fluxes in equation (1). Previous literature suggests that this term is very small, so we have neglected it [Kurian and Vinayachandran, 2007; Halkides and Lee, 2011]. Its effects will therefore appear in the residual term. The residual also includes measurement errors associated with horizontal currents and surface heat fluxes; errors in parameterizing vertical processes; computational errors associated with finite differencing; and sampling errors [Foltz and McPhaden, 2009, Vialard et al., 2008].
 A satellite gridded SSHA product [AVISO Altimetry, 2009] and the near-surface thermal structure derived from Simple Ocean Data Assimilation (SODA) analysis [Carton et al., 2005] were utilized to characterize the nature of propagating Rossby waves. QuikSCAT wind data [Wentz et al., 2001] are utilized to characterize the wind variability over the EIO.
3.1. Temperature Inversion Variability in the South Central BoB
Thadathil et al.  and Thompson et al.  show that temperature inversions in the BoB exhibit strong seasonality. A relatively large number of temperature inversions occur during the winter season (November to March) when the barrier layer is thickest and a relatively small number of temperature inversions occur during spring (March to May) and during the peak of the summer monsoon (August) when the barrier layer almost disappears. These studies suggested that the reduction in thermal inversions during spring and summer was associated with a weakening of the barrier layer linked to decreasing river runoff and strong warming in the near surface layer due from intense solar heating.
 The temporal evolution of temperature inversion with magnitudes greater than 0.1°C at the buoy location (8°N, 90°E) in the BoB during October 2006 to October 2008 is shown in Figure 3. Consistent with earlier studies, we find in our data that temperature inversions show strong seasonality, with greatest frequency of occurrence during October to March and least during May to September [Thadathil et al., 2002; Thompson et al., 2006; De Boyer Montegut et al., 2007]. Temperature inversions appear beginning in October and reach maximum strength (0.5°C–0.75°C) during winter, after which they decrease in frequency and amplitude. It is also interesting to note that frequency and amplitude of temperature inversions show strong year-to-year variability. During W67, temperature inversions were weak (maxima of 0.2°C to 0.35°C) compared to W78 (maxima of 0.50°C–0.75°C). The number of days with temperature inversions during W67 (20% or 13% of the time) was relatively small compared to W78 (52% or 29% of the time). Further, temperature inversions during W67 are shorter lived on average compared to W78. In addition temperature inversions exhibited significant intraseasonal variability, especially during W78.
 To complement our analysis of the RAMA data, we show the temporal evolution of vertical temperature and salinity structures obtained from a nearby Argo float (Figures 4c, 4d, 5c, and 5d) along with RAMA time series (Figures 4a, 4b, 5a, and 5b). These figures show reasonable agreement between these two temperature and salinity data sets, and in particular between the occurrence of temperature inversions and barrier layers. The correspondence is not exact as would be expected because the float and the buoy are typically situated several dozens of kilometers part and because of the differences in temporal and vertical resolution. However, for both data sets W67 exhibits fewer temperature inversions and thinner barrier layers and W78 exhibits a greater number of temperature inversions and thicker barrier layers. The basic consistency between the two data sets builds confidence in our ability to accurately detect temperature inversions and barrier layers in the Bay. The probable mechanism for the formation of these temperature inversions and their year-to-year variability is explained in the following section.
3.2. Mechanisms Contributing to Temperature Inversion Formation
 On seasonal time scales, the formation of shallow salt stratified mixed layers and barrier layers are the necessary condition for the formation of temperature inversions [Thadathil et al., 2002]. In the presence of salt stratification and a thick barrier layer, temperature inversions can occur when the surface layer cools over warmer water in the barrier layer, provided that the overall density profile is statically stable [Thadathil et al., 2002; Kurian and Vinayachandran, 2006; Nisha et al., 2009]. The strong salinity stratification in the BoB during W67 and W78 is evident in Figure 5, which shows relatively low-salinity waters ranging from 32.20 to 34.60 in the upper 30–40 m and higher salinity waters below. This vertical salinity structure maintains shallow salt-stratified mixed layers and allows the formation of barrier layers [Rao and Sivakumar, 2003, Girishkumar et al., 2011].
 Figure 6a shows the BLT and magnitude of temperature inversions at the buoy location in the BoB. The figure clearly shows that strong temporal correspondence between the presence of barrier layers and temperature inversions and as is also evident in Figures 4a and 4b. The large year-to-year and intraseasonal variability in the BLT and temperature inversions is noteworthy, with strong temperature inversions forming only when the barrier layer is thicker than ∼20 m. The study by Girishkumar et al.  showed that intraseasonal Rossby waves, which are forced remotely by intraseasonal wind (40–100 days) variability along the equator, significantly influences thermocline depth and BLT in the BoB [Girishkumar et al., 2011, Figure 9]. The standard deviation at the buoy location of 40–100 day period band pass filtered D23 is 5.5 m during W67 and 11.1 m during W78 respectively, indicating the greater forcing and response on intraseasonal time scales during W78. Furthermore, the connection between thermocline depth variations and temperature inversions is also evident in Figures 4a and 4b.
 A scatter plot (Figure 7) characterizes the statistical relationship between BLT and temperature inversion magnitude and, as seen in Figure 6a, there is a clear tendency for strong temperature inversions during periods of thick barrier layer, though the relationship is not monotonic. It is also important to note that while temperature inversions are large when the barrier layer is thick, not all thick barrier layers exhibit temperature inversions (Figure 6a).
 A thorough analysis of the mechanisms responsible for the formation of temperature inversions would require simultaneous heat budget analyses in both the mixed layer and barrier layer. However, we lack sufficient data at the depth of the barrier layer (e.g., for horizontal currents and temperature gradients) to perform a fully three-dimensional analysis. Even so, we do have sufficient data to examine processes that are likely to be important in the mixed layer and below, including the role of penetrative shortwave radiation in the barrier layer. Penetrative shortwave radiation is a key mechanism to warm the barrier layer most notably when the mixed layer is shallow [Anderson et al., 1996; Kurian and Vinayachandran, 2006; Mignot et al., 2012]. That process combined with mixed layer cooling by net surface heat loss and/or horizontal advection can generate temperature inversions.
 Figure 6b shows shortwave radiation absorbed at depths below the mixed layer (Qpen) and heat flux into the mixed layer by the sum of horizontal advection and net surface heat flux (Qnet−Qpen). For clarity these three terms are plotted in Figure 6c. Figures 6a and 6b show significant temporal correspondence between the formation of temperature inversions and net heat loss (due to the sum of horizontal advection and net surface heat flux) from the mixed layer compared to heat gain below the mixed layer in the presence of thick barrier layer (events marked as blue circles in Figure 6b). To illustrate, in mid December 2007, there was approximately 100 W m−2 heat loss from the mixed layer by horizontal advection and net surface heat flux, which would cool the mixed layer. At the same time, approximately 20 W m−2 penetrated below the mixed layer and was absorbed in the thick barrier layer. In the absence of significant vertical mixing, shortwave radiation reaching the depth of barrier layer would result in heating and an increase of temperature. Cooling in the mixed layer and warming below would therefore lead to the formation of a temperature inversion (Figure 6a). This particular process can explain the formation of temperature inversions at other times during the study period as well (blue circles in Figure 6b).
 Figures 6b and 6c further show that the relative contribution of horizontal advection and net surface heat flux to mixed layer cooling varies from event to event. However, in not all cases does shortwave radiation absorbed below the mixed layer and cooling within mixed layer lead to the formation of a temperature inversion (red dots in Figure 6b). These situations are primarily associated with thin barrier layers. Once such an event occurred at the end of January 2007 (Figure 6b; second red dot in the left plot). During this time, net surface heat flux (Qnet−Qpen) and horizontal advection together cooled the mixed layer at a rate of ∼100 W m−2, while 25 W m−2 was absorbed below the mixed layer (Qpen). These processes could have created a strong temperature inversion as explained in the previous example. In this instance the barrier layer was thin and the ILD was shallow (Figure 4b), so that vertical mixing with cold thermocline water could counteract the formation of a temperature inversion.
4. The Influence of Temperature Inversions and BLT on Mixed Layer Heat Budget
 The relative role of barrier layers and temperature inversions on the mixed layer heat budget is examined using equation (1). Consistent with earlier studies, net surface heat flux is a significant contributor to the mixed layer heat budget in the BoB [Rao and Sivakumar, 2000, Sengupta and Ravichandran, 2001; Sengupta et al., 2002; Parampil et al., 2010] though as reported by Parambil et al. , horizontal advection is also important (Figure 8). Both of these processes vary substantially on intraseasonal time scales.
 Vertical processes in the heat budget equation (Figures 8c and 8d) show large intraseasonal and year-to-year variations in association with variability in BLT and the presence of temperature inversions (Figure 6a). Throughout W67, the vertical processes show strong cooling tendencies with only brief periods of warming (Figure 8c). During W67, relatively thin barrier layers and relatively few temperature inversions allow for entrainment of cold thermocline water into the mixed layer to cool the surface.
 During W78, vertical processes show alternating periods of strong cooling and warming tendencies on intraseasonal time scales associated with variations in BLT. During times when thick barrier layers and temperature inversions occur coincidentally, vertical processes show a strong warming tendency due to vertical mixing of warm subsurface water into mixed layer. Periods when a barrier layer exists without a temperature inversion show no significant mixed layer cooling or warming. One such period is the latter part of November 2007, when a relatively thick barrier layer occurs without a temperature inversion and vertical processes do not indicate a significant cooling or warming tendency (Figures 8d and 6a; right plot). During times of thin barrier layers in the absence of temperature inversions, vertical processes generally indicate a cooling tendency. The presence of a shallow thermocline then allows efficient mixing of cold thermocline water into the mixed layer to cool the surface.
 Figures 8c and 8d indicate that vertical processes overall tend to cool more consistently during W67 than, during W78. To explore this difference in greater detail, Figure 9 illustrates the relationship between BLT and temperature inversions on vertical mixing processes during these two periods at the buoy location. Significant warming tendencies are observed during times when the barrier layer is thick (i.e., > 30 m) and strong temperature inversions occur (Figure 9), conditions that occur more frequently during W78 than during W67. Times when the barrier layer is thick and there are no temperature inversions are associated with neither warming nor cooling, consistent with the premise that the barrier layer isolates the mixed layer from the thermocline. Conversely, when the barrier layer is thin (< 20 m), vertical processes lead to strong cooling tendencies (−40 to −160 W m−2). The above analysis highlights how intraseasonal and year-to-year variability in BLT and temperature inversions influence vertical mixing processes in the mixed layer heat budget.
 How these vertical processes compare to other processes affecting the mixed layer heat balance is illustrated in Table 1 for the November to February average values of (i) mixed layer heat content tendency, (ii) the sum of net surface heat flux and horizontal advection, (iii) the vertical processes, and (iv) the residual heat flux term for W67 and W78. Approximately 6 W m−2 more heat is available for the mixed layer during W67 (21.1±15.3 W m−2) compared to W78 (15.4±10.1 W m−2) from the combination of surface heat flux and horizontal advection. However, cooling from vertical processes varies from −22.5±7.0 W m−2 in W67 to −9.7±4.4 W m−2 in W78. The approximately 13 W m−2 greater heat loss by vertical processes from the mixed layer during W67 compared to W78 is consistent with the fact that barrier layers were thinner on average and temperature inversions were less frequent in W67.
Table 1. November to February Average Values for Different Components of the Mixed Layer Heat Budget (in W m−2) at 8°N, 90°E From Equation (1) with 90% Confidence Intervals for W67 and W78a
Rate of change of mixed layer heat content (i)
Net surface heat flux + Horizontal advection (ii)
Vertical heat flux term (iii)
Residual heat flux term (iv)
The average November to February mixed layer depth was 28 m in W67 and 25 m in W78 at this location.
 Thus, during W67 relatively higher heating of the mixed layer from the sum of surface heat flux and horizontal advection (21.1±15.3 W m−2) was counteracted by a similar heat loss by vertical process (−22.5±7.0 W m−2). Even though relatively lower mixed layer heating occurred during W78 from the sum of surface heat flux and horizontal advection (15.4±10.1 W m−2), the reduction in vertical heat flux loss (−9.7±4.4 W m−2) due to the presence of thick barrier layer and temperature inversions resulted in approximately 7 W m−2 more net heat flux into mixed layer than during W67. This difference in heat flux forcing between the 2 years over a MLD of 25 m would lead to an approximately 0.7°C greater temperature increase from November to February W78 compared to the same period during W67. The evolution of SST for 10 day smoothed data (Figure 10a) and monthly SST differences between W67 and W78 (Figure 10b) shows this difference. SST started about 0.25°C warmer in November 2006 compared to November 2007, but ended about 0.35°C colder in February 2007 compared to February 2008 indicating approximately 0.6°C more warming occurred during W78 compared to W67.
 These year-to-year differences are small but they are physically significant due to the presence of persistently warm SSTs in the BoB higher than 28°C [Shenoi et al., 2002], which is generally considered as a threshold for atmospheric deep convection [Gadgil et al., 1984]. Saturation vapor pressure is an exponential function of SST, so the intensity of deep convection is very sensitive to changes in SST over the warm pool region. As shown by Palmer and Mansfield  in the Pacific and Francis and Gadgil  in the BoB, a small change in SST over high-mean SST can result in a significant atmospheric response.
 This analysis demonstrates that differences in vertical mixing processes on interannual time scales can influence interannual variability in SST in the BoB during the winter when barrier layers are most pronounced. Intraseasonal mixed layer warming events associated with thick barrier layers and temperature inversions in aggregate appear to be an important aspect contributing to these year-to-year differences. Hence, it is important to understand the origins of this variability for a more comprehensive understanding of the mixed layer heat balance in the BoB.
5. The Role of Remote Wind Forcing on Observed Temperature Structures in the BoB
 As discussed in the introduction, large scale conditions associated with the IOD and ENSO favoured a shallower thermocline in the BoB during W67 than during W78 (Figure 2). This interannual variability is reflected in the mean depth of the thermocline (as measured by D23) at the buoy location, which was 85.5 m in W67 compared to 100.5 m during W78. These conditions would favor relatively thick barrier layers and more inversions at the buoy location during W78 compared to W67 since the mean MLD was not much different between the 2 years at 8°N, 90°E (28 m in W67 versus 25 m in W78). We know that much of the intraseasonal variability in thermocline depth and BLT in the BoB is remotely forced by zonal wind variations in the equatorial waveguide [e.g., Girishkumar et al., 2011]. The Bay is also strongly forced on interannual time scales by anomalous wind variability in the EIO [Rao et al., 2002; Vinayachandran et al., 2007; Cai et al., 2009; Rao et al., 2010]. On both time scales, the mechanism involves the propagation of equatorial waves into the eastern boundary and their reflection into westward propagating Rossby waves in the BoB. The combination of these remote forcing effects across this range of intraseasonal and interannual time scales created conditions much more favorable to temperature inversion formation during W78 compared to W67.
6. Summary and Conclusions
 Time series measurements of temperature and salinity recorded at 8°N, 90°E in the south central BoB from a RAMA mooring are used to describe the observed intraseasonal and interannual variability of BLT and temperature inversions and their influence on the mixed layer heat budget during the winters of 2006–2007 (W67) and 2007–2008 (W78). During the W67, the BLT, and temperature inversions were weak compared to W78. In addition, variations in BLT and in the frequency of temperature inversions were larger on intraseasonal time scales during W78.
 The salinity distribution at the buoy location shows a sharp halocline in the near-surface layer resulting in the formation of a shallow salt stratified mixed layer within a deeper isothermal layer. The presence of these shallow salt stratified mixed layers and thick barrier layers provide favorable conditions for the formation of temperature inversions. Horizontal advection of cooler water and penetrative shortwave radiation below the mixed layer are the major process modulating the formation of temperature inversions in the presence of a thick barrier layer. The analysis shows that the intraseasonal and year-to-year variability in BLT and temperature inversions has significant influence on MLT through the modulation of the vertical heat flux at the base of mixed layer. A schematic diagram (Figure 11) illustrates how the processes described above operate to produce temperature inversions during periods of thick and thin barrier layer, and how the presence of these thermal inversions in turn affect SST variability through their effects on the surface layer heat balance. Our analysis suggests a significant role for barrier layers and temperature inversions on mixed layer heat budget and hence the evolution of SST in the BoB.
 Our study further indicates that realistic simulations of salinity, BLT and temperature inversion fields will be necessary to accurately represent MLT in the BoB in ocean and coupled-ocean atmosphere models. Moreover, much of the variability in upper ocean thermal structure and hence the upper ocean heat budget in the BoB is forced remotely by winds in the EIO on intraseasonal to interannual time scales. Thus, understanding the controls on BoB SST is ultimately a problem that requires a basin scale perspective on wind, heat flux and fresh water forcing.
 The encouragement and facilities provided by the Director, INCOIS are gratefully acknowledged. We would also like to acknowledge two anonymous reviewers, whose comments and suggestions greatly improved the final manuscript. RAMA data are provided by the TAO Project Office of NOAA/PMEL. QuikSCAT data are produced by Remote Sensing Systems and sponsored by the NASA Ocean Vector Winds Science Team. Microwave OI SST data are produced by Remote Sensing Systems (www.remss.com) and sponsored by National Oceanographic Partnership Program (NOPP), the NASA Earth Science Physical Oceanography Program, and the NASA MEaSUREs DISCOVER Project. QuikSCAT and TMI+AMSRE data are downloaded from www.ssmi.com. The altimeter products are produced by SSALTO/DUACS and distributed by AVISO. Argo data were collected and made freely available by the International Argo Program and the national programs that contribute to it (http://www.argo.ucsd.edu, http://argo.jcommops.org). The Argo Program is part of the Global Ocean Observing System. INCOIS contribution Number. 140 and PMEL contribution number 3935.