Observed year-to-year sea surface salinity variability in the Bay of Bengal during the 2009–2014 period
The present study describes the observed sea surface salinity (SSS) interannual variability in the Bay of Bengal over the 2009–2014 period. It is based on an original compilation of all available in situ SSS observations in that region, assembled in a 2°-resolution trimonthly gridded field. We find that year-to-year SSS variability is particularly strong in the north-eastern part of the bay. Over recent years, this variability takes the form of two successive and opposite phases: a saltening phase from mid-2009 to late 2010, immediately followed by a freshening phase from late 2010 to late 2011. The typical magnitude of each anomalous spell is about one in the practical salinity scale, making this area one of the most variable of the tropical oceans at interannual timescales. A simple mixed-layer salt budget indicates that year-to-year large-scale SSS variability in the Northern Bay of Bengal is primarily driven by freshwater flux variability with a correlation of 0.68, with rather independent contributions from precipitation and river run-off. The oceanic surface circulation variability contributes less systematically to the large-scale SSS evolution in the Northern Bay of Bengal over the entire record with a correlation of 0.13, despite a strong contribution at times, in particular, during the 2011 positive Indian Ocean Dipole (IOD) freshening.
KeywordsSSS Bay of Bengal Argo Ganges Brahmaputra IOD
The Bay of Bengal (henceforth BoB) forms the north-eastern part of the Indian Ocean and is blocked by landmasses to the north of 23° N. Ocean currents within this semi-enclosed basin are strongly influenced by the seasonally reversing monsoon winds. This basin is also the recipient of intense precipitation and continental river run-off, with most of the freshwater flux entering the northern half of the basin during summer monsoon and post-monsoon season. The yearly freshwater flux received by the BoB largely exceeds the freshwater flux evaporated back to the atmosphere (Shenoi et al. 2002; Sengupta et al. 2006). As a result, sea surface salinity (SSS) is consistently very low in the northern BoB (typically inferior to 321). This low SSS in the BoB is thought to have various implications for regional climate through its stabilizing effect of the water column. These fresh surface waters indeed maintain shallow mixed layers in the bay, which are efficiently warmed by surface heat fluxes (e.g. de Boyer et al. 2007). They are thus held responsible for the consistently high sea surface temperature (SST) in the BoB, compared to the neighbouring tropical basins (Han et al. 2001; Shenoi et al. 2002; Rao et al. 2002a; Girishkumar et al. 2013). As the SST is often above the critical threshold of 28 °C (considered as the limit above which deep atmospheric convection can develop, e.g. Gadgil et al. 1984), the BoB is prone to intense air-sea interactions. A well-known manifestation of these is the development of tropical depressions and cyclones over the bay, in which intensity could be partly controlled by this salinity stratification. These salt-stratified waters indeed inhibit vertical mixing under tropical cyclones and reduce the resulting surface cooling (Sengupta et al. 2008; Neetu et al. 2012), therefore promoting an intense surface evaporation that contributes to the cyclones intensification. This control of upper ocean temperature by BoB salinity stratification not only is limited to the influence of tropical cyclones but also occurs at longer timescales. Two recent studies indeed suggested that intraseasonal SST variability in the Northern BoB could be modulated by salinity stratification (Vinayachandran et al. 2012; Girishkumar et al. 2013). These influences of upper ocean salinity on surface temperature, and potentially on the atmosphere, call for an in-depth description of the SSS spatial and temporal variability within the bay.
Beyond this seasonal picture, much less is known about BoB SSS variations at other timescales. At intraseasonal timescales, Parampil et al. (2010) described a substantial variability of mixed-layer salinity based on a few drifting profilers in the central BoB and suggested that it is mainly driven by small-scale advection of riverine and atmospheric fresh water by ocean turbulent circulation. Rao et al. (2011) also suggested a dominant role of oceanic circulation over air-sea freshwater fluxes on SSS intraseasonal variability in the BoB based on ship-borne hydrographic observations.
The limited coverage of available SSS observations over the BoB has prevented the oceanographic community from assessing SSS variations at longer timescales and in particular interannual variations. Over the past few years though, the SSS observing system over the BoB has been improving consistently, and it has now become possible to monitor the year-to-year basin-scale SSS variations. The objective of the present study is threefold. We first present a novel SSS dataset encompassing the whole BoB, with a pluri-annual time span from 2009 to 2014 (Sect. 2). We then describe the year-to-year variations of SSS over the north-eastern BoB during this period for the first time, based on this dataset (Sect. 3). The mechanisms driving interannual SSS variability are then discussed using a simple mixed-layer salinity budget (Sect. 4). Section 5 summarizes our main results and discusses the limitations and perspectives of our study.
2 Methods and datasets
2.1 In situ SSS datasets
The main contributor to our SSS dataset is the Argo database, obtained from the Ifremer Data Centre (ftp.ifremer.fr/ifremer/argo). Argo profilers are autonomous drifters that provide temperature and salinity profiles from 2000-m depth to the surface, every 5 to 10 days. The amount of Argo floats drifting in the BoB gradually increased from 30 to 60 over the 2009–2014 period. They yielded between 1200 (in 2009 and 2010) to 4000 (in 2013) yearly salinity profiles (Fig. 2b). We extracted the uppermost valid measurement within the 5–15-m layer, with “good” or “ probably good” quality flag (see Argo quality control manual, 2012 available on www.argodatamgt.org/content/download/341/2650/file/argo-quality-control-manual-V2.7.pdf). For most profiles, this uppermost record is typically situated at about 8–9-m depth. We will discuss the influence of the sampling depth on the SSS estimate in Sect. 2.2.
We also considered SSS measurements collected from both bucket samples and XCTD temperature and salinity profiles along two repeated merchant ship tracks between Chennai (label “ C” in Fig. 2a) and Port Blair (label “ PB”) and between Kolkata (label “ K”) and Port Blair. Bucket measurements (953) and 249 XCTD profiles were collected along the C-PB and K-PB transects over the 2009–2014 period. Along both routes, the spatio-temporal resolution of the bucket samples and XCTD casts varies but remains close to 1° and 2 months, with about 20 such voyages between July 2009 and December 2013 (see Fig. 2c for an illustration of the spatio-temporal distribution of the XCTD data collected on the Kolkata-Port Blair transect). The bucket samples were analyzed using a Guild Line 8400 Autosal salinometer following standard international procedures. This ensures typical accuracy of salinity of about 10−3. XCTD salinity records over the upper 4 m were discarded as it is usually done to account for the delay in the conductivity sensor response (e.g. Tanguy et al. 2010), and we considered the 5-m record as SSS.
Three RAMA buoys have been episodically measuring salinity at 1-m depth over the 2009–2014 period, at 90° E–8° N, 90° E–12° N and 90° E–15° N. Data return rates vary between 50 % at 8° N and 12° N and 75 % at 15° N. We used the 5-day averaged data obtained from the RAMA website (http://www.pmel.noaa.gov/tao/disdel/disdel-rama.html). Further details on the mooring design and data quality control can be found in McPhaden et al. (2009).
OMNI deep-ocean moorings are similar to RAMA moorings, except that they are inverse catenary moorings. Six moorings were deployed in various regions of the BoB in 2011 (11° N–86.5° E, 8° N–85.5° E, 14° N–83° E, 16.5° N–88° E, 10.5° N–94° E, 18.2° N–89.7° E) with an average data return of 99 %. Data from all sensors are collected during the last 10 min of every hour and stored in the internal storage facility. Data received are transmitted to the shore station using INMARSAT geostationary satellite communication every third hour (Venkatesan et al. 2013). Both high- and low-resolution data are archived at the Indian National Centre for Ocean Information Systems (INCOIS) after standard quality control procedures. They are delivered to end users through the Ocean Data Information Systems (ODIS) (Shesu et al. 2013). The conductivity temperature (CT) sensors are fitted at 5, 10, 15, 20, 30, 50, 75, 100, 200 and 500-m depth along the mooring line. The CT sensors are Seabird SBE37-IM model with accuracy of 0.002 °C for temperature and 0.003 mS/cm for conductivity. Salinity data at 5-m depth from all the six moorings over mid-2011 to mid-2013 were used in this study.
A merchant ship (M/S Lavender) equipped with a thermosalinograph continuously measuring the near-surface salinity has been crossing the southern bay for ten times from 2009 through September 2013 (typically once every 5 months). The thermosalinograph data went through a delayed-mode quality control including comparison with climatology and correction with external water samples. Although it is hard to assign a specific depth to the thermosalinograph measurements (because of varying ship speed and load), the data may be representative of the 0–10-m upper ocean layer (see Delcroix et al. 2005 for full details on the data acquisition and processing procedures). A few near-surface measurements from ship-borne CTD casts were also made available by the National Institute of Oceanography Data Centre (India). Most of the observations were collected in the coastal western bay (Fig. 2a), at 0–10-m depth, in 2009–2011.
2.2 Data reduction
The spatio-temporal sampling achieved by these various datasets is very heterogeneous. First, these data originate from different depths within the upper 10 m of the water column, which may induce errors in the estimation of SSS evolution. A vast majority of the salinity data are provided by Argo floats, for which salinity has consistently been measured at a depth close to 8 m. The resulting SSS error has been estimated using data from the 15° N, 90° E RAMA mooring. The root-mean-square difference between salinity sampled at 1 and 10-m depth is only 0.17, about four times smaller than the 1-m salinity variations with a standard deviation of 0.63. This difference is also typically much smaller than the ∼0.5 to 1 SSS signals that we analyze in the rest of the study. We checked that the spatial patterns of SSS variability that we obtained were not altered when using RAMA 10 m (instead of 1 m) and XCTD 8 m (instead of 5 m) values (not shown). It is therefore unlikely that the varying depth of data collection induces a significant error in our analyses. Second, although our datasets cover the bulk of the BoB, one has to keep in mind that some sub-regions are left completely unsampled (Fig. 2a; see in particular the Andaman sub-basin between 93° E and the eastern boundary, as well as the far northern bay to the north of 20° N).
In a similar way to de Boyer et al. (2014), we merged the six datasets by computing the median of all available individual measurements on a 2° × 2° × 3 months grid. This grid cell size was chosen after several trials (not shown) and found to give the best compromise between the resolved scales and actual data coverage. We did not attempt to fill the data gaps in the resulting gridded dataset, to avoid introducing any spurious signal in the merged product. Rather, we simply keep the data-void regions out of our study.
2.3 Freshwater fluxes
In order to quantify the atmospheric and continental control of SSS variability, we have to account for evaporation (E), precipitation (P) and river run-off (R). Daily TropFlux 1° × 1° gridded product (Praveen Kumar et al. 2012) provides the evaporation data. This product is available from 1979 to 2013 (see www.locean-ipsl.upmc.fr/tropflux/ for details) and is largely derived from ERA-Interim re-analysis (Dee et al. 2011) with ad hoc corrections derived from observations. We used the three hourly, 0.25° × 0.25° Tropical Rainfall Measuring Mission (TRMM) 3B42 precipitation gridded product (Huffman et al. 2007) over the 2009–2014 period. It is a merged product from multi-satellite and in situ data from rain gauges. This dataset is available from http://disc.sci.gsfc.nasa.gov/precipitation. Finally, run-offs from the three largest river systems in the BoB are included in our study: the Brahmaputra, the Ganges and the Irrawaddy (their locations are shown in Fig. 1). For the Ganges-Brahmaputra river system, we use monthly estimates of continental freshwater flux into the BoB from Papa et al. (2012). This time series is derived from a combination of in situ elevation-discharge relationships (the so-called rating curve, Papa et al. 2010) with river water level retrieved from multiple altimetry satellites (TOPEX/Poseidon, ERS-2, ENVISAT and Jason-2). For the Irrawaddy River, in situ discharge time series are unfortunately not available over the recent period. Furuichi et al. (2009) published monthly estimates from in situ observations over the 1966–1996 period. We used their monthly in situ discharge data, together with TOPEX/Poseidon river level time series over their common period of availability (1993–1996) to establish a rating curve following a similar approach to that of Papa et al. (2012). We then used this rating curve to derive a continuous, monthly estimate of the Irrawaddy River discharge at the river mouths for 1993–2012 using TOPEX/Poseidon (1993–2002), ENVISAT (2002–2008) and Jason-2 (2008 onwards). During the period of the present study (2009 onwards), the estimates of river discharge rely on Jason-2 observations. Note that no run-off data are available after December 2012.
2.4 Surface current
The role of oceanic transports on SSS variability is assessed using Ocean Surface Current Analyses - Real time (OSCAR) surface current product of Bonjean and Lagerloef (2002). This product is a combination of altimetry-derived geostrophic current and Ekman drift from scatterometer wind. This product is available at various spatial and temporal resolutions. In this study, we considered the version at 1° × 1° × 5 days, filtered, available from www.oscar.noaa.gov.
3 Observed variability of SSS
Durand et al. (2011) found that the total (seasonal plus interannual) SSS variability in their numerical simulation was also maximum in the northern and western parts of the bay and minimal in the central basin. Akhil et al. (2014) also reported the strongest seasonal SSS variability close to the northern and western boundaries in their simulation. Our observational product strongly suggests the same spatial distribution at interannual timescales. Durand et al. (2011) further diagnosed that the sole Ganges-Brahmaputra year-to-year discharge changes were able to drive SSS interannual signals of ∼0.2 restricted to the northernmost part of the basin (north of about 18° N). Our observational product reveals a much stronger interannual variability, with a standard deviation above 0.45 as far south as 14° N. This suggests that other mechanisms than Ganges-Brahmaputra run-off (P, E and/or oceanic circulation) influence SSS interannual variability in the northern half of the bay. The mechanisms driving the observed SSS interannual evolution will be specifically addressed in Sect. 4. In the following texts, we focus our study on the region that displays the most consistent area of maximum SSS variability: the north-eastern bay (henceforth NEB) defined as the 14° N–23° N, 86° E–96° E box (Fig. 3a).
At this stage, and to summarize this first overview of our new SSS dataset, we have shown that the 2009–2014 period is characterized by a suite of alternating saltening and freshening phases in the NEB, which generate intense SSS anomalies (with magnitudes of 1 or more). The deficient run-off in 2009 appears as a possible candidate to explain the salty anomaly build-up seen from mid-2009 to early 2010. The sustained and increased saltening through 2010 is consistent with the deficient rainfall during summer 2010. The return to fresher than normal conditions in late 2011 could be explained by both excess run-off and excess rainfall during 2011. The mid-2013 fresh anomaly is hard to explain, as we do not have any run-off estimate during this year, and precipitation is only weakly in excess at that time. To quantitatively ascertain the respective role of the various components of freshwater flux in the observed variability of upper NEB salinity, as well as the role of ocean transports, we will use a simple mixed-layer salinity budget in Sect. 4.
4 Mechanisms of NEB SSS variability
4.1 Salt budget computation
where S is the box-averaged mixed-layer salinity, h is the mixed-layer thickness (prescribed, assumed to be constant in time and in space), A is the area of the box, unormal is the horizontal current in the normal direction to the box boundaries (i.e. zonal component for the east and west boundaries and meridional component for the north and south boundaries) considered positive when entering the box, Sboundary is the salinity at the boundary of the box, E is the evaporation, P is the precipitation and R is the freshwater flux of all rivers entering the box. The first term of the right-hand side of Eq. (1) is the advection, computed as a two-dimensional integral along the vertical surfaces of the four boundaries of the box. The second term is a two-dimensional integral over the air-sea interface of the box. This equation neglects the space and time dependencies of the mixed-layer depth over the domain and assumes that the vertical exchanges of salt between the mixed layer and the oceanic subsurface are of second order. This is clearly a limitation of our approach, but this practical choice is dictated by the absence of observational estimates of exchanges between the mixed layer and the deeper oceanic layers in the BoB. We will further discuss the consequences of this approximation in Sect. 5.
We used interannual anomalies of the products described in Sect. 2 to estimate terms in Eq. (3): OSCAR currents, TropFlux evaporation, TRMM precipitation and our altimetric-derived discharge dataset for the Ganges, Brahmaputra and Irrawaddy rivers. We computed the time derivative of the salinity anomalies using a forward difference. We used the mixed-layer depth observed estimate from de Boyer et al. (2004). It is defined as the depth where density exceeds the density at 10 m by an amount equivalent to a 0.2 °C drop in temperature. Its climatological and spatially averaged value over the NEB box is 18 m. Note that on account of the limited availability of run-off data, we were only able to perform this budget over the 2009–2012 period.
4.2 Interannual SSS budget of NEB domain
Correlation of the various components of the estimated mixed-layer salt budget to the observed storage term in the NEB box
River run-off forcing
Table 1 provides an overview of the role of the various processes on the salt budget over the period. The freshwater forcing term correlates best with the observed SSS tendency, with a correlation of 0.68, indicating the primary role of freshwater forcing in driving the NEB average SSS. Both precipitation and run-off contribute to this over the entire period, with correlations to the observed tendency of 0.55 and 0.42, respectively. On the other hand, while the advection term is sometimes comparable in amplitude to the freshwater forcing (Fig. 7), it is only weakly correlated (0.13) to the observed storage over the entire period.
Over the entire period, freshwater forcing hence dominates large-scale SSS evolution in the NEB box. Let us now describe the budget throughout the period in more detail. During summer monsoon (June–July–August), the total freshwater forcing term (E-P-R) is generally prominent (in 2010, 2011 and 2012). Consistently with our findings of Sect. 3, the contribution of river run-off dominates the contribution of precipitation in mid-2009, while the opposite holds true in mid-2010 (Fig. 7). In mid-2011, both the river run-off and precipitation have a comparable influence. At some other times, particularly during the inter-monsoon seasons, the advection term dominates (March–April–May 2010 and September–October–November 2011). The relative contribution of zonal advection (through the western boundary of the NEB domain) and meridional advection (through its southern boundary) is also highly variable: For instance, the saltening advective peak of March–April–May 2010 is almost purely driven by the zonal advection, whereas the freshening peak of September–October–November 2011 is mostly driven by meridional advection with a minor contribution of zonal advection. Note that over the whole period, there is a slight tendency of meridional advection to be directly opposed to zonal advection. This can be simply explained by the flux formulation used for advection in our box model (Lee et al. 2004).
Figure 8 also provides the time integral of the various forcing terms of SSS evolution. The 2009–2010 saltening event appears to be largely driven by freshwater forcing and hence reflects the dominant control by this term over the entire period (Table 1). This term shows two periods inducing SSS increase (during the monsoons of 2009 and 2010), separated by a period when it induces hardly any change (from late 2009 to mid-2010). As seen in Sect. 3, the respective contributions of precipitation and river discharge during these two successive monsoons are contrasted. During monsoon 2009, precipitation is close to normal, whereas river discharge is deficient. During monsoon 2010, in contrast, precipitation is deficient and river discharge is slightly higher than normal. Globally, these two terms act together to force a sustained saltening trend during two successive monsoons. The advection tendency terms show that the meridional component (through the southern boundary of NEB box) hardly plays any role, whereas the zonal component (through the western boundary) induces a marked saltening during the first half of 2010. Globally, during the first period of integration, the role of advection is secondary as compared to the role of the freshwater fluxes. From September–October–November 2010 to December 2011–January–February 2012, both freshwater and advective fluxes contribute to the freshening, with advective fluxes contributing about twice more (approximately −0.9 unit) than freshwater fluxes (approximately −0.5 unit). This hence illustrates that, while freshwater forcing tends to dominate over the entire period (Table 1), advection is not negligible at certain times. The freshening driven by the freshwater fluxes is most intense during monsoon 2011. It is explained by both excess rainfall and excess run-off as seen in Fig. 6. Regarding advection, the zonal and meridional terms roughly cancel each other during the first part of the period (until mid 2011); then, both terms contribute to freshen the domain (particularly so during September–October–November 2011), until the end of the integration period (early 2012).
5 Summary and discussion
The BoB upper ocean salinity stratification is believed to play a key role in the Indian Ocean climate (e.g. Shenoi et al. 2002). Several observational studies have described the BoB SSS seasonal cycle and its driving processes over the past two decades (e.g. Rao and Sivakumar 2003; Chatterjee et al. 2012; Chaitanya et al. 2014). This seasonal cycle is closely linked to the monsoonal control of rainfall, river run-offs and upper ocean circulation (Benshila et al. 2014; Akhil et al. 2014). Beyond this prominent seasonal cycle, there is also considerable interannual variability over the BoB. Precipitation (e.g. Gadgil 2003) and riverine freshwater supply to the bay (Papa et al. 2012) in particular vary significantly from year to year. The resulting BoB SSS interannual variability has, however, so far been little studied. The limited in situ measurement availability has indeed prevented the oceanographic community from assessing the interannual variability of BoB salinity over the past decade. During recent years, the BoB SSS observing system has drastically improved, now reaching a comparable status to that of the other tropical basins. As a result, it is now feasible to monitor the year-to-year basin-scale SSS variability in the BoB.
In this study, we assembled an unprecedented set of SSS observations over the BoB, including Argo profilers, a couple of repeated ship-of-opportunity buckets and XCTD transects, as well as RAMA and OMNI mooring data. We merged these different datasets to obtain a 2° × 2° × trimonthly product that satisfactorily covers the BoB at regional scale over 2009–2014. This product reveals a consistent area of marked year-to-year SSS variability, encompassing the north-eastern quarter of the basin. The typical SSS interannual anomaly magnitude ranges from 0.5 to 1 there. This magnitude is commensurate with signals observed in other highly variable regions of the tropical oceans, such as the equatorial Indian Ocean (Durand et al. 2013) or the far-eastern equatorial Pacific Ocean (Alory et al. 2012). The observed variability displays an 18-month-long saltening period from mid-2009 to late 2010, followed by a 15-month-long freshening period from late 2010 to early 2012. The late 2013–early 2014 period is also fresher than normal.
Our simple mixed-layer salt budget neglects vertical exchanges between the surface and subsurface through turbulent processes and advection. It, however, yields an SSS budget that lies within the bounds of the observational error on the SSS tendencies, hence suggesting that vertical physics do not prominently contribute to interannual SSS anomalies in the NEB over the period that we analyzed. This may seem contradictory with findings reported in the modelling study of Akhil et al. (2014), who concluded that vertical exchanges of salt between the mixed layer and the thermocline play an important part in the seasonal evolution of mixed-layer salinity in the northern BoB. This may be interpreted in two ways. Either processes responsible for the seasonal cycle and interannual anomalies of BoB SSS are indeed different, and vertical processes do not contribute as much to interannual SSS anomalies than to the seasonal cycle, or there is a compensation between some of the neglected terms in our simple salt budget. The fact that our simple budget generally underestimates the magnitude of the most intense observed freshening and saltening spells in the BoB could indeed be attributed to the lack of a proper representation of vertical physics. This limitation also prevents assessing the potential role of long baroclinic waves (coastal Kelvin waves and planetary Rossby waves), known to affect the thermohaline structure of Bay of Bengal at interannual timescales (e.g. Rao et al. 2002a), on SSS interannual evolution. The BoB western boundary is also a region of strong interannual SSS variability (Fig. 3a), in which our simple mixed-layer salt budget does not perform well (not shown). The poor quality of remotely sensed oceanic currents in coastal areas (e.g. Durand et al. 2009) could explain the inability of our simple budget to reproduce the SSS evolution there, since along-shore advection plays a key role in the seasonal freshening of the western BoB (Chaitanya et al. 2014; Akhil et al. 2014). Finally, the limited coverage of in situ salinity data also prevented us from estimating advection of anomalous salinity gradients by climatological currents. All these limitations related to the simplicity of our salt budget and to observational issues call for revisiting our conclusions using a more sophisticated numerical framework. A general ocean circulation model indeed takes the spatio-temporal variability of mixed-layer depth, vertical exchanges of salt/freshwater between the mixed layer and the deeper layers and all the horizontal advective terms explicitly into account. An investigation of the mechanisms at stake in driving the interannual SSS evolution in the Bay of Bengal using such a tool, with close comparisons to observed SSS variability derived in the current study, is therefore a promising perspective to the present study.
Keeping in mind the potential role of the BoB upper ocean salinity stratification on air-sea interactions (Shenoi et al. 2002; Neetu et al. 2012, and references therein), it would be interesting to explore how interannual variability of SSS can influence the interannual variability of the upper BoB heat budget and surface temperature. We believe that the present study opens the way for future research in this direction. The current observational density is indeed sufficient to observe large-scale seasonal anomalies. The observational dataset that we have developed can hence be used as a ground truth for calibration/validation of present and future space-borne salinity products (Subrahmanyam et al. 2013; Tang et al. 2014) and general circulation models of the Bay of Bengal. The current study also pleads not only for a sustained but also for an improved in situ observing system of the BoB salinity. Indeed, some areas (like the Andaman Sea) are not well sampled by existing observations, and current observations are not sufficient to confidently assess all the terms of the SSS balance.
Salinity is expressed in the Practical Salinity Scale throughout this paper, i.e. in grams of salt per kilogram of seawater, hence without units.
We thank the three anonymous reviewers for their constructive comments on our paper. XCTD and bucket salinity observations are supported by the Ministry of Earth Sciences, through the Indian National Center for Ocean Information Services, Hyderabad, India. We thank the numerous people who spent a long time at sea collecting the salinity observations. Thermosalinograph data were collected and processed by the French SSS Observation Service (www.legos.obs-mip.fr/observations/sss). We sincerely thank the team efforts of NIOT technical and scientific staff for providing continuous data from the OMNI moorings. We are thankful to Argo and RAMA programmes for making their salinity data available to all. FD, FP, ML and JV are funded by Institut de Recherche pour le Développement (IRD). Part of this work was done when CAVS and FD visited the Indo-French Cell for Water Sciences (Joint International Laboratory IRD-IISc, Bangalore, India). Support from this laboratory is gratefully acknowledged. We acknowledge constructive comments by Debasis Sengupta. This is CSIR-NIO contribution No. 5689
- Alory G, Maes C, Delcroix T, Reul N, Illig S (2012) Seasonal dynamicsof sea surface salinity off Panama: the far Eastern Pacific Fresh Pool. J Geophys Res 117(C04028). doi:10.1029/2011JC007802
- Chaitanya A V S, Lengaigne M, Vialard J, Gopalakrishna VV, Durand F, Kranthikumar C, Amritash S, Suneel V, Papa F, Ravichandran M (2014) Salinity measurements collected by fishermen reveal a “river in the sea” flowing along the east coast of India. Bull Am Meteorol Soc., in pressGoogle Scholar
- Da-Allada C Y, Alory G, du Penhoat Y, Kestenare E, Durand F, Hounkonnou N M (2013) Seasonal mixed-layer salinity balance in the tropical Atlantic Ocean: mean state and seasonal cycle. J Geophys Res 118. doi:10.1029/2012JC008357
- de Boyer MC, Madec G, Fischer AS, Lazar A, Iudicone D (2004) Mixed layer depth over the global ocean: an examination of profile data and a profile-based climatology. J Geophys Res 109(C12003). doi:10.1029/2004JC002378
- Girishkumar M S, Ravichandran M, McPhaden MJ (2013) Temperature inversions and their influence on themixed layer heat budget during the winters of 2006–2007 and 2007–2008 in the Bay of Bengal. J Geophys Res 118. doi:10.1002/jgrc.20192
- Neetu S, Lengaigne M, Vincent EM, Vialard J, Madec G, Samson G, Ramesh Kumar MR, Durand F (2012) Influence of upper-ocean stratification on tropical cyclone-induced surface cooling in the Bay of Bengal. J Geophys Res 117(C12020). doi:10.1029/2012JC008433
- Papa F, Durand F, Rossow WB, Rahman A Bala SK (2010) Satellite altimeter-derived monthly discharge of the Ganga-Brahmaputra river and its seasonal to interannual variations from 1993 to 2008. J Geophys Res, 115(C12013). doi:10.1029/2009JC006075
- Papa F, Bala SK, Pandey RK, Durand F, Gopalakrishna VV, Rahman A, Rossow WB (2012) Ganga–Brahmaputra river discharge from Jason-2 radar altimetry:an update to the long-term satellite-derived estimates of continentalfreshwater forcing flux into the Bay of Bengal. J Geophys Res 117(C11021). doi:10.1029/2012JC008158
- Parampil SR, Gera A, Ravichandran M, Sengupta D (2010) Intraseasonal response of mixed layer temperature and salinity in the Bay ofBengal to heat and freshwater flux. J Geophys Res 115(C0500). doi:10.1029/2009JC005790
- Rao RR, Sivakumar R (2003) Seasonal variability of sea surface salinity and salt budget of the mixed layer of the north Indian Ocean. J Geophys Res 108. doi:10.1029/2001JC000907
- Rao SA, Gopalakrishna VV, Shetye SR, Yamagata T (2002a) Why were cool SST anomalies absent in the Bay of Bengal during the 1997 Indian Ocean Dipole event ? Geophys Res Lett 29(0). doi:10.1029/2001/GL014645
- Rao SA, Behera SK, Masumoto Y, Yamagata T (2002b) Interannual subsurface variability in the tropical Indian Ocean with a special emphasis on the Indian Ocean Dipole. Deep-Sea Res II(49):1549–1572Google Scholar
- Rao SA, et al. (2011) Modulation of SST, SSS over northern Bay of Bengal on ISO time scale. J Geophys Res 116. doi:10.1029/2010JC006804
- Saji NH, Goswami BN, Vinayachandran PN, Yamagata T (1999) A dipole mode in the tropical Indian ocean. Nature 401:360–363Google Scholar
- Sengupta D, Bharath Raj GN, Shenoi SSC (2006) Surface freshwater from Bay of Bengal runoff and Indonesian throughflow in the tropical Indian Ocean. Geophys Res Lett 33(L22609). doi:10.1029/2006GL027573
- Shenoi SSC, Shankar D, Shetye SR (2002) Differences in heat budgets of the near-surface Arabien Sea and Bay of Bengal: implications for the summer monsoon. J Geophys Res 107(C6). doi:10.1029/2000JC000679
- Venkatesan R, Arul Muthiah M, Ramesh K, Ramasundaram S, Sundar R, Atmanand MA (2013) Satellite communication systems for ocean observational platforms: societal importance and challenges. J Ocean Technol 8(3):47–73Google Scholar
- Vinayachandran, PN, Neema CP, Mathew S, RemyaR (2012) Mechanisms of summer intraseasonal sea surface temperature oscillations in the Bay of Bengal. J Geophys Res 117(C01005). doi:10.1029/2011JC007433