The MITgcm is customized to run over the Arabian Sea and Bay of Bengal regions of the northern Indian Ocean for 17 a from 1 January 1998 to 31 December 2014 using the setup described in Section 2. The model uses the initial temperature, salinity, sea level elevation, zonal and meridional velocity from the spin-up run. We show in Fig. 2 (upper panel) the climatological sea surface temperature (SST) from model over the region of study. The corresponding SST values from the TRMM Microwave Imager (TMI) (Wentz et al., 2015) are also shown in the figure. It is clear from the figure that the model is able to reasonably capture the mean state of the SST. The lowest SST values are observed in the western AS. The spatial correlation coefficient of the temporally averaged model SST with the corresponding TMI values is 0.91 (at 99.9% significance level). The SST bias (TMI-model) is also shown in the figure. We see from the figure that the bias is very small in most parts of the domain. The bias seen along the coasts, however, may be ignored since TMI has known problems along the coasts (Wang et al., 2011). The SST root mean square error (RMSE) between the annually averaged model and TMI is 0.35 °C which is much smaller than the standard deviation of 0.92 °C of the annually averaged TMI SST, thus confirming the quality of model output.

The climatological sea surface salinity (SSS) results are also shown in Fig. 2 (lower panel). The comparison of model SSS is made against the NIO Climatological Atlas (NIOA) (Chatterjee et al., 2012). The model realistically simulates high surface salinity values in the AS and low values in the BoB. The spatial correlation coefficient between the temporally averaged model SSS and the corresponding NIOA SSS is 0.95, with a significance level of 99.9%. The model shows small SSS bias with respect to NIOA in nearly all the regions except in the northern BoB. The RMSE between the annually averaged model and NIOA SSS is 0.36 which is much smaller than the standard deviation of 1.28 of the annually averaged NIOA SSS.

To make an assessment of the model’s ability in capturing the spatio-temporal variability of the SST and SSS in the region of study, we compute their monthly and daily standard deviation from the model’s data during 1 January 1998 to 31 December 2014 (Figs 3a and b, respectively). The corresponding monthly SST and SSS standard deviation values from the TMI and NIOA are shown in Fig. 3a, whereas, daily SST and SSS standard deviation values from the TMI and Aquarius (https://podaac.jpl.nasa.gov/aquarius; weekly values interpolated to daily values during 2012–2014) are shown in Fig. 3b. It is clear from the figures that barring a few places, the model is able to distinguish the region of high and low SST and SSS monthly and daily variability over the northern Indian Ocean domain covering the AS and BoB. Further, the daily variability of SST and SSS is always higher as compared to its monthly variability. The SST of the northern part of the domain is more variable as compared to the southern part. We also see from the figure that except western part of the model domain, the SST variability in general monotonically decreases from north to south. The northern BoB and AS show higher SST variablity compared to the southern portion. The SSS variability on the other hand, is higher in the southern AS as compared to northern AS. In the BoB, however, the northern region shows higher SSS variability. Nevertheless, except northern BoB and southern AS, all the other regions of the model domain exhibit very low SSS variability.

The surface and sub-surface variables in the AS and BoB exhibit significantly interesting changes at a seasonal scale (Rao and Sivakumar, 2003; de Boyer Montégut et al., 2007; Schott et al., 2009). Thus, the correct simulation of the seasonal variability of SST and SSS is very important. The cyclogenesis, and near-surface mixing properties in the AS and BoB are largely understood in terms of the seasonal variability of surface temperature and salinity (Neetu et al., 2012). We show in Fig. 4 the seasonal variability of the SST and compare it with the TMI SST and NIOA SST data. The seasonal averages are obtained for December-Januray-February, March-April-May, June-July-August, and September-October-November. We observe that the SST seasonal variability is very well capured by the model. The highest SST is obtained during spring months of March-April-May and lowest during the winter months of December-January-February, as expected. The northern BoB exhibits high SST values during June-July-August and September-October-November. This may happen primarily due to moderately high shortwave radiation compounded with shallow mixed layer depth as a result of increased freshwater flux and precipitation (see Section 3.2 below). Moreover, the creation of a barrier layer in the Bay of Bengal blocks heat transfer from surface to deep ocean. We also find that western and south-western part of the domain shows particularly low SST values during June-July-August months.

The seasonal variability of the SSS is shown in Fig. 5. It is clearly evident from the figure that the model very well simulates the SSS seasonal variability. We notice that the AS has higher SSS than BoB in general. The northern BoB has particularly low SSS values during June-July-August (monsoon) and September-October-November (post-monsoon season) as a result of increased freshwater transport from major adjoining rivers (Ganga, Brahmaputra and Irrawaddy) (Sengupta et al., 2006). The southern BoB has higher SSS than northern BoB in all the seasons. We see, however, an exactly opposite nature in the AS. In the southern AS, we find lowest SSS during December-January-February which may be due to coastal runoff, and also due to movement of less saline BoB water towards AS and its mixing with the saltier AS water.

To guage the quality of the model output in simulating the temporal evolution of the SST and SSS, we make an index of these variables in the upper and lower BoB and AS region. The monthly SST and SSS values during 1998–2014 (17 a) are area averaged over the region (7°–11°N, 67°–72°E) in the southern AS (SAS), (16°–20°N, 67°–72°E) in the northern AS (NAS), (87°–92°E, 7°–11°N) in the southern BoB (SBoB), and (16°–20°N, 87°–92°E) in the northern BoB (NBoB). We show in Fig. 6a, the time series of the SST indices for the model as well as TMI. The model and TMI SST time series are highly correlated at values 0.93 and 0.90 in the SAS and NAS, respectively. Similarly, the correlation of 0.94 and 0.96 exists between the model and TMI SST time series in the SBoB and NBoB, respectively. These correlations are significant at 99.9% level. The RMSE between the model and TMI SST time series is 0.33°C (0.55°C) and 0.31°C (0.47°C) in the SAS (NAS) and SBoB (NBoB), respectively. These values are much smaller than the standard deviation of 0.82°C (1.21°C) and 0.87°C (1.51°C) of the TMI SST time series over these regions. The SST time series shows near-periodic (12 months) interannual fluctuations but the trend is nearby absent in all the selected regions over this time period. It is interesting to note that the SST variations in the NAS and NBoB are almost identical to each other, though magnitude is little higher in the BoB. We notice that the range of SST variability is higher in the northern domain (i.e., NAS and NBoB) as compared to southern domain. In the northern domain, the SST peaks during the month of May with a smaller peak occuring in the month of October also. Similarly, the minimum SST is obtained during January–February with another minima occuring during August. In the SAS and SBoB, the SST values are highest during May and lowest during January–February. In the bottom panel of Fig. 6a, we show the SST bias map between the model and TMI values. It may be seen from the figure that the biases are of the same order in all the regions. Morever, barring few exceptions, the biases are generally low. In the years of 2013 and 2014, the March–April SST bias in the NBoB and SBoB, respectively are relatively higher and the model is found to underestimate the SST values. On the other hand, the model overestimates August SST values of 2007 and 2009 in the SAS and NAS regions, respectively.

The monthly time series of the SSS indices in the SAS, NAS, SBoB and NBoB is shown in Fig. 6b for the model and ORAS4 data. The model and ORAS4 monthly SSS time series are significantly correlated with each other in all the regions (Table 3). The RMSE between the two time series is also smaller than the corresponding standard deviations (Table 3). It is seen from the figure that the SSS of the AS is much higher than BoB. The SSS in the SAS remains high with lowest values (~35) observed during March–April. The NAS, on the other hand shows very high SSS values (>36) throughout the year with very small monthly variations. The SBoB SSS values also show small monthly fluctuations around 33. The SSS in the NBoB, however, shows highest monthly variability and ranges between 30 and 34. The lowest SSS values in the NBoB are obtained during the month of October. The model-data bias between the monthly SSS time series of model and ORAS4 is shown in the bottom panel of Fig. 6b. It may be seen from the figure that the SSS bias is very small in all the regions, except in the NBoB. In other words, the model underperforms in the NBoB region, whereas, it performs very well in all other regions in terms of monthly SSS simulation. This may be due to incorrect vertical parametrization and boundary conditions in the NBoB. The underperformance of the model’s SST and SSS during monsoon (and to some extent post-monsoon) season may also be due to use of very coarse river runoff and precipitation data from the NCEP during these seasons.

It has been argued in recent researches that understanding the intraseaonal variability of the ocean surface variables (more importantly SST) and how it interacts with the atmospheric intraseasonal oscillations is critical for the correct prediction of Indian monsoon rainfall (Schiller and Godfrey, 2003; Li et al., 2015 and references therein; Keerthi et al., 2016). Thus, it will be worthwhile to also study the intraseasonal variability of the SST and SSS over the model domain. For this purpose, we extract the 20–60 d intraseasonal mode from the daily data of model SST area averaged over the NAS, SAS, NBoB and SBoB regions for each season (Fig. 7). The corresponding intraseasonal modes from the TMI observations are also extracted to quantify the model’s performance at intraseaonal scale. Table 4 summarizes the correlation and RMSE between the intreaseasonal modes of observation and model. The mean and standard deviation of intraseasonal TMI values are also given in Table 4. The model’s ability in simulating the SST variability at intraseasonal scale is clearly evident from Fig. 7 and Table 4. It is clear from the Table that correlation between the model and observation is significantly high and RMSE between the model and observation is less than the standard deviation of observation in all the regions and for the intraseaonal modes of all the seasons. This confirms the realistic simulation of model simulated intreaseaonal modes. Table 4 also shows that the intraseasonal SST of the southern part of AS and BoB (i.e., SAS and SBoB) is in general cooler and more variable as compared to the respective northern parts (i.e., NAS and NBoB). The highest intraseasonal SST variability is seen in the spring season of March-April-May in the SAS (0.64°C) and SBoB (0.61°C). On the other hand, the lowest intraseasonal SST variability is obtained in winter months of December-January-February in the NAS (0.30°C) and NBoB (0.28°C) regions. The intraseasonal variability of the SSS over the model domain is not given since the continuous daily SSS observations for the period of simulation are not available.

To demonstrate the quality of model simulation in representing the sub-surface characteristics of temperature and salinity, we show in Fig. 8, the latitude-depth maps of temperature and salinity at 70°E in the AS and 90°E in the BoB. The model caputures the layered structured of the upper ocean with correct magnitude very well. One of the possible regions for the small discrepencies, if any, in the mixed layer depth and thermocline observed between model and NIOA may be their different horizonal and vertical resolutions. The sub-surface temperature and salinity strucutures are very different in the AS and BoB. The latitude-depth map of the temperature shows that the thermocline in the AS has a tilted structure with values increasing from south to north. The BoB, however, shows no such nature and thermocline is nearly flat. The salinity diagram shows that the values are nearly opposite in the AS and BoB at higher latitudes. For example, in the region 15°–22°N, where the highest salinity values are observed in the AS, the BoB region shows the lowest salinity. The low salinity of the BoB in this region is mainly attributed to freshening of water as a result of river runoff from major adjoining rivers of the region. On the other hand, higher evaporation than precipitation in the northern AS is the principal reason of high salinity in that region. The presence of salty outflow from the Red Sea and Persian Gulf is also the reason for high salinity of the region.

The estimation of mixed layer depth (MLD) is very important for oceanic investigations including upper ocean productivity, air-sea exchange processes, and long-term climate change (Thomson and Fine, 2003, Keerthi et al., 2013). We compute the MLD of the model domain using Lorbacher et al. (2006) criterion, which is based on the shallowest extreme curvature of near surface density profiles. We show in Fig. 9 (upper panel) the spatial maps of the seasonal variation of the MLD over our region of study. The corresponding MLD maps from ARGO (Holte et al., 2016) are also shown in Fig. 9 (lower panel) for comparison. Holte et al. (2016) computed MLD using more than 1 250 000 Argo profiles (collected between January 2000 and May 2016) and a hybrid algorithm (Holte and Talley, 2009).

Figure 9 clearly demonstrates the model’s ability in correct estimation of the MLD seasonal variations (both spatially as well as temporally). The model MLD values are close to the ARGO MLD estimates in all the seasons. Moreover, our results also suggest that the MLD is deeper in the AS as compared to the BoB (de Boyer Montégut et al., 2007, Keerthi et al., 2013). The MLD remains low in the upper northern BoB in all the seasons. The lowest MLD values are observed in the months of March-April-May and highest during December-January-February. The mean MLD in the AS region during December-January-February, March-April-May, June-July-August and September-October-November is 50, 22, 44, 27 m, respectively. On the other hand, the mean MLD in the BoB region during the same season is 31, 18, 32, 23 m, respectively. Thus, the mean MLD of the AS remains higher as compared to the BoB during all the seasons. The mean MLD during March-April-May is very shallow in the AS as well as BoB. We, however, observe intermediate MLD values during monsoon season of June-July-August. Interestingly, in the post-monsoon months of September-October-November, the mean MLD of the AS and BoB remain very close to each other. The maximum MLD is obtained during winter (December-January-February) with AS and BoB having MLD values of 150 m and 67 m, respectively. The minimum MLD in the AS and BoB (nearly 11–12 m) is, however, obtained during the spring (March-April-May). It is also interesting to note that the AS and BoB MLD during December-January-February and June-July-August exhibits dipole like structure with opposite signs during these seasons. For example, during the December-January-February, the northern and central AS and BoB show high MLD whereas southern part of domain shows low MLD values. Contrary to this, we observe higher values of MLD in southern domain as compared to northern and central AS and BoB during June-July-August.

Figure 9 shows that the season-to-season variability of the MLD is quite significant. In general, the air-sea fluxes are considered as drivers of the MLD variability. It is, therefore, important to illustrate and understand the seasonal MLD variability vis-à-vis seasonal changes in the air-sea fluxes. For this purpose, we show in Fig. 10a the seasonal map of air-sea fluxes, namely, wind stress, net heat flux, and net freshwater flux (buoyancy flux). It is clear from the 1st column of Fig. 10a that the negative buoyancy flux brought by the excess of evaporation over precipitation largely explains the relatively deep MLDs in December-January-February season (de Boyer Montégut et al., 2007) seen in Fig. 9 (the 1st column). In addition to this, low shortwave radiation flux and high negative latent heat flux and sensible heat flux (Fig. 10b, the 1st column) (resulting in negative net heat flux) is also responsible for deep MLDs during December-January-February. Whereas, the main reason behind very low MLD values in the AS and BoB during March-April-May are very weak easterlies (weak wind stress, the 2nd column of Fig. 10a) over the region of interest. Very high shortwave radiation flux (Fig. 10b, the 2nd column) (resulting in positive net heat flux) also contributes to MLD shallowing during March-April-May. During the monsoon (June-July-August) season (Fig. 10a, the 3rd column), high wind stress and low shortwave radiation as well as high negative latent heat flux (Fig. 10b, the 3rd column) (resulting in negative net heat flux) are responsible for high MLD values in the southern AS. On the other hand, due to high freshwater flux (precipitation-evaporation) in the northern BoB, the MLD remain shallower. In the post-monsoon months of September-October-November, the MLDs, along coastal and adjoining areas in the BoB have low (to very low) values, whereas other regions of the domain also have low to intermediate values. This happens as a result of low wind stress in the region of interest (Fig. 10a, the 4th column) and increased freshwater flux (runoff from rivers) in the BoB. The net heat flux in most parts of the AS and BoB remains very small due to cancelling effect of its component fluxes (Fig. 10b; the 4th column).

In order to understand the seasonal variability of the mixed layer temperature in general, and SST in particular, the estimation of mixed layer heat budget is important. Even though it is largely believed that the major contribution in the SST variability comes from seasonally varying air-sea fluxes, however, the horizontal and vertical temperature advections also play a major role at some locations. It will be, therefore, useful to explicitly compute all possible contributions to the mixed layer heat budget. This will help us understand and quantify importance of each individual term in the spatio-temporal variability of SST.

Following Stevenson and Niiler (1983), Moisan and Niiler (1998), and Huang et al. (2010), we compute the mixed layer heat budget as

where $ {T_{\rm{t}}}{\rm{}}\left( { = \frac{{\text{∂} T}}{{\text{∂} t}}} \right)$ is the mixed layer temperature tendency and F is the forcing. The forcing is sum of zonal advection (Q_u), meridional advection (Q_v), vertical entrainment (Q_w), net surface heat flux (Q_q), and vertical diffusion (Q_zz). In other words, F is expressed as

The consistency between T_t and F is used to check the closure of the temperature equation. For further computational details on each of these terms please refer to Appendix A of Huang et al. (2010).

The time-latitude map of different terms of mixed layer heat budget at 67°E is shown in Fig. 11. We see from figure that even though T_t and F are very close to each other (thus giving good closure of temperature tendency), however, they are not exactly the same. The difference between them may be due to missing contributions from (weak horizontal) diffusion term. The differences could also be due to numerical errors. It is clear from the figure that out of all forcing terms, Q_q is most dominant term. The contribution from zonal advection Q_u is generally insignificant except few places. The meridional advection Q_v on the other hand helps in increasing the mixed layer heat budget (and SST) of southern part of domain (south of 5°S) during July to September. The Q_v reduces mixed layer heat budget between 5° and 10°N during June to October and north of 21°N during April to November. The vertical entrainment Q_w remains very low in the AS except during April. The diffusion term Q_zz although remains low, but helps in increasing the mixed layer heat budget north of 15°N during April to November. The contributions from Q_q (and therefore forcing F) suggest that in the Arabian Sea at 67°E (north of 5°N), we get low SST in December–February, high SST in March–May, low SST again in June–August, and intermediate SST values in September–November months. This is consistent with the SST variability (at 67°E) shown in Fig. 4. The F and Q_q values of Fig. 11 also suggest that the SST of the southern part of domain remains low during May–September months of the year.

The time-latitude map of mixed layer heat budget at 90°E is shown in Fig. 12. In this case also, we clearly notice that the largest contribution to the mixed layer heat budget is coming from air-sea fluxes Q_q. While the effect of vertical entrainment Q_w is smaller in this case compared to at 67°E, however, the spatio-temporal contributions from zonal advection Q_u are larger here especially during summer monsoon months (June–September) between 5°–10°N. The meridional advection Q_v plays a similar role in this case as seen at 67°E. In the southern part of domain it tends to increase the SST during May–September. On the other hand, negative Q_v values in northern part of domain (between 0–10°N) during Jun–October helps in maintaining intermediate SST values in these regions during these months. The vertical entrainment and diffusion terms give only small positive contribution (north of 15°N) in F at 90°E. The Q_q (and F) map clearly suggests that at 90°E in the Bay of Bengal (north of 5°N), the SST tendency should remain low during December–January and high during March–April. The SST tendency also remains low at 90°E between 5°–12°N and south of 5°S during May–September. These observations are also consistent with the SST variability shown in Fig. 4 thus demonstrating the robustness of mixed layer heat budget calculations.

For correctly capturing the mechanisms responsible for mixed layer temperature (or SST) variability using mixed layer heat budget analysis, it is important to see if a reasonable closure of heat budget exists using the model output. To demonstrate the closure, we show the seasonal variability of forcing F and mixed layer temperature tendency T_t in Fig. 13. The bias map (bottom panel of Fig. 13) suggests that the mixed layer temperature budgets have a good closure, indicating that the results regarding the mechanisms controlling the mixed layer temperature (or SST) evolution are quite robust. The differences between them point towards the role of missing term of horizontal diffusion processes and numerical errors in our calculations. Nevertheless, the variation of seasonal variability of F and T_t is largely coherent with the seasonal evolution of SST seen in Fig. 4.

The monthly mixed layer heat budget area-averaged over the SAS, NAS, SBoB and NBoB region is shown in Fig. 14. In the SAS region, the highest positive (negative) heat budget is obtained during April (June). We notice that F and T_t values are close to each other. Apart from Q_q, the contribution from all other terms is small. In the NAS region, the highest positive (negative) heat budget is obtained during May (July). We notice that F and T_t values are close to each other except during June–September. In addition to Q_q, the contribution from vertical advection Q_v and diffusion Q_zz also becomes important in these months. These terms try to increase the mixed layer heat so that good closure between F and T_t may be obtained. In the SBoB, the highest positive (negative) heat budget is obtained during April (December–January). We see a good closure between F and T_t except during May and November–December. The maximum (minimum) contribution to the mixed layer heat budget comes from air-sea fluxes (horizontal entrainment). In the NBoB, the highest positive (negative) heat budget is obtained during April (November–December). The F and T_t are close to each other in all the months, except during October-December. In the NBoB, we see significant contribution in the mixed layer heat budget coming from Q_v and Q_zz, although with opposite signs (especially during February–June) with maximum contribution still coming from Q_q. While the meridional advection reduces the mixed layer heat budget of the region, the vertical diffusion tends to increase it. It is clear from the Figure, that the net air-sea heat flux plays most important role in controlling the mixed layer temperature variability of the northern Indian Ocean.

In an effort to demonstrate the quality of model solutions in simulating the surface circulation, we show in Fig. 15 the seasonally varying surface currents of the region during the period of simulation. The model results are compared with the OSCAR surface current (Bonjean and Lagerloef, 2002) data, which is a combination of in situ and satellite altimeter derived current products. We find that the model surface currents match very well with OSCAR. We notice from Fig. 15 that during December-January-February (winter monsoon season), the southeasterly currents flow northwards in the southern AS and BoB. In the AS, the flow is dominantly meridional (south-north), whereas upper BoB shows a weak anti-cyclonic behavior. During March-April-May strong anticyclonic motion is observed in the BoB covering the northern and parts of central bay. The poleward East Indian coastal current (EICC) (Schott and McCreary, 2001) is also seen in the upper bay. The currents, on the other hand, are found to be weak in the AS. In the summer monsoon season of June-July-August, the BoB currents are dominantly (zonal) eastwards, except in the northernmost bay, where we notice a feeble cyclonic circulation. The model perfectly reproduces the direction of the observed currents although with somewhat greater magnitudes (Fig. 15; the 3rd column). In the AS, we notice northwesterly currents (including West Indian coastal current, WICC) flowing southwards which are almost opposite in direction to December-January-February currents. During September-October-November, we observe cyclonic circulation in the northern bay. Both the EICC and WICC flow equatorwards. The zonal summer monsoon current (SMC) flowing from southern AS towards southern BoB during June-July-August and winter monsoon current (WMC) flowing in exactly opposite direction during December-January-February are also well reproduced by the model. The discrepancies in modeled and observed (OSCAR) surface current data seen in the 3rd column of Fig. 15 may be mainly due to their very different horizontal and temporal resolutions. For example, the OSCAR data is available at 1/3-degree horizontal resolution and 5 d of temporal resolution. On the other hand, our model output uses daily data at a horizontal resolution of 0.09 degree.

We also compute zonal and meridional volume transport across different sections (shown in Fig. 1 by vertical and horizontal lines, respectively) in the AS and BoB using Eq. (3) as below:

where h is the ocean depth at position (x, y), u is the zonal velocity, v is the meridional velocity, Ly and Lx are the widths of the meridional and zonal transects, respectively. The meridional volume transport for the upper 100 m is computed along the section (12°N, 71°–75°E) in the Arabian sea and (12°N, 80°–84°E) in the Bay of Bengal. The zonal transport for the upper 100 m is computed along the section (4°–7°N, 76°E) at the interface of AS and BoB. These sections are so chosen that by studying the transport of water mass across them (which includes surface as well as intermediate and deep ocean), we can infer the variability of the monsoonal currents, namely, WICC in AS, EICC in the BoB, and SMC (WMC) at the interface of AS and BoB on seasonal and interannual time scales. Moreover, this will also help us quantify seasonal transport of water mass from the BoB to AS and vice versa during summer and winter monsoon seasons. In Fig. 16, we show the monthly climatology of the volume transport across these sections and compare the results with the corresponding values from the ECCO state estimates. Figure 16 (upper panel) shows that in the AS section, the peak north to south meridional transport (–2.4×10⁶ m³/s) is found during July (monsoon season), whereas peak south to north transport (+3.7×10⁶ m³/s) is obtained during November-December (post-monsoon season). However, the annually averaged volume transport across this section is very small ((0.1± 1.06)×10⁶ m³/s). Similarly, in the BoB (Fig. 16, middle panel) the mean meridional transport across the chosen section is nearly zero (–0.02±2.2)×10⁶ m³/s, i.e., almost equal amount of water flows through these section in the north to south and south to north directions. The peak north-to-south meridional volume transport (–2.6×10⁶ m³/s) across the BoB section is seen during February-March, whereas highest south to north transport is obtained during monsoon season of October-November (+2.7×10⁶ m³/s). Thus, the signs of meridional volume transport in the AS and BoB are almost opposite to each other (r = –0.75) during different months of the year. However, the difference between the AS and the BoB meridional volume transport should not be simply looked as opposite in phase, because the transport of EICC has a semiannual component, whereas the WICC has a single peak and trough over a year. The correlation between volume transport time series obtained from the model and ECCO state estimate is 0.81 and 0.85 in the AS and BoB, respectively. Even though these correlations are highly significant, however, phase differences between the model and ECCO values are seen in the BoB during some months. The zonal volume transport across a section at the interface of AS and BoB is shown in Fig. 16 (lower panel). It is clear from the Figure that the peak east-west transport (–4.1×10⁶ m³/s) is obtained during January (winter season), whereas the peak zonal transport (+4.5×10⁶ m³/s) in opposite direction i.e., west-east direction is observed during August–September (summer monsoon) months. The correlation between model and ECCO derived zonal transport time series is 0.95. The mean zonal flow across this section is from east to west (i.e., from BoB to AS) with transport in the range of (–0.6± 2.2)×10⁶ m³/s.