The seasonal upwelling along the west coast of India
(WCI) brings nutrient-rich, oxygen-poor subsurface waters to the continental
shelf, favoring very low oxygen concentrations in the surface waters during
late boreal summer and fall. This yearly-recurring coastal hypoxia is more
severe during some years, leading to coastal anoxia that has strong impacts
on the living resources. In the present study, we analyze a 1/4
The continental shelf off the west coast of India (WCI) is home to the
largest coastal hypoxic system of the world ocean, spreading over an area of
Observations from a series of ship cruises during September–October 1999 (Naqvi et al., 2000) off the WCI and time-series measurements from a fixed site off Goa since 1997 (Naqvi et al., 2009) indicate the occurrence of severe hypoxic conditions over almost the entire shelf and anoxic conditions close to the WCI, with the most intense anoxic event reported in fall 2001 and moderate ones in fall 1998 and 1999. In contrast, these data indicate that fall 1997 was characterized by far less hypoxic conditions. The frequent anoxic conditions which occurred during the 1998–2002 period were accompanied by a 3- to 5-fold decline in demersal fish catch in 1999 and 2001 compared to 1997. The total fish landing also remained low between 1998 and 2002, adversely affecting the economy based on fisheries, and the pelagic fish catches shifted from the dominance of mackerel to oil sardine from 1998 to 1999 (Krishnakumar and Bhat, 2008). There has been a remarkable revival of fisheries since 2003, apparently due to a relaxation of the intensity of oxygen deficiency, with no severe anoxic event reported over the recent years. Subsurface oxygen concentrations have also been reported to be significantly lower for the 1997–2004 period than for the 1971–1975 period (Naqvi et al., 2009). These observations thus reveal large interannual and decadal fluctuations in the oxygen deficiency along the WCI, but the processes responsible for these variations have not yet been understood (Naqvi et al., 2009).
As opposed to the other coastal hypoxic systems that have generally
developed as a result of human activities (largely eutrophication) in the
last few decades (Diaz and Rosenberg, 2008), the seasonal surface oxygen
deficiency along the WCI is naturally driven. Indeed, it has been suggested
that the oxygen-deficient conditions that develop in early fall along the
WCI result from the seasonal upwelling, which brings poorly oxygenated
subsurface waters from the oxygen minimum zone (OMZ) in the interior
Arabian Sea (e.g., Morrison et al., 1999; Naqvi, 1987; Sarma, 2002) towards
the surface along the continental shelf. This connection between the
offshore and the shelf oxygen content has been discussed by several studies
on the basis of in situ oxygen observations along different cross-shelf
transects. For example, Banse (1959) showed a clear association between the
seasonal upwelling and coastal hypoxia on the shelf off Cochin at
The seasonal upwelling along the WCI starts in April, but the shallowest thermocline marking the peak of the upwelling is usually observed during September–October (e.g., Schott and McCreary, 2001). Local alongshore winds are, however, only favorable to upwelling during the southwest monsoon (i.e., they only have an equatorward alongshore component from June to August). This is an indication that upwelling along the WCI is to a large extent forced by remote winds (Shetye et al., 1990). Wind variations in the equatorial band and Bay of Bengal do indeed force coastal Kelvin waves that travel along the rim of the bay and up the WCI to influence the thermocline depth there (e.g., McCreary et al., 1993; Shetye, 1998). Recently, Suresh et al. (2016) demonstrated that wind variations in the vicinity of Sri Lanka are responsible for a large fraction of the seasonal upwelling along the WCI. By contrast with this seasonal variability of the upwelling and upper-ocean oxygen, there are currently little clues as to what causes the interannual variability along the WCI.
Identifying the main drivers of the near-surface oxygen interannual variations in regions of the main coastal hypoxic systems is an important endeavor as it may ultimately help to predict the occurrence of severe anoxic events. The large-scale climate modes have been suggested to influence the local oxygen variability in several coastal regions. For instance, the El Niño–Southern Oscillation (ENSO) strongly influences the oxygen concentrations along the coasts of Peru and Chile (Helly and Levine, 2004; Arntz et al., 2006; Gutierrez et al., 2008), with intensified oxygenation associated with weak El Niño upwelling and intensified hypoxia associated with strong La Niña upwelling. As in the Pacific, the natural climate variability in the Indian Ocean could also be a potential candidate responsible for the near-surface oxygen interannual variations along the WCI, but that has so far not been evaluated. The main indigenous mode of Indian Ocean interannual climate variability is the Indian Ocean Dipole (hereafter IOD; Saji et al., 1999; Webster et al., 1999; Murtugudde et al., 2000). A “positive” IOD is characterized by a cooling and anomalously shallow thermocline in the eastern Indian Ocean and by a warming and anomalously deep thermocline in the central and western Indian Ocean, driven by anomalous easterlies in the central equatorial Indian Ocean. A “negative” IOD is associated with anomalous signals of opposite polarities. The IOD usually starts developing during boreal summer and peaks in fall (e.g., Saji et al., 1999). El Niño and La Niña events tend to induce, respectively, the positive and negative IODs in the Indian Ocean, but IOD can also occur independent of ENSO (e.g., Annamalai et al., 2003). The IODs induce larger amplitude variations in large-scale wind and thermocline depths than those associated with El Niño events over the Indian Ocean (e.g., Currie et al., 2013) and thus have the potential to affect the upwelling variations along the WCI through planetary wave propagation.
While the spatiotemporal density of observations in the eastern Pacific margin is adequate to accurately describe the monthly oxygen evolution along
the west coast of South America over the past 3 decades (Helly and
Levine, 2004; Arntz et al., 2006), there is a dearth of long-term data from
fixed sites in the Indian Ocean as compared to the Pacific and the Atlantic
(Gupta et al., 2016), which in turn prevents building a reliable time series
that could depict the year-to-year variations. On the other hand,
three-dimensional coupled physical–biogeochemical models that include the
oxygen cycle have contributed to improve the description and understanding
of dynamics of hypoxic events in various coastal regions (Peña et al.,
2010), such as the Gulf of Mexico (e.g., Hetland and DiMarco, 2008), Black
Sea (e.g., Gregoire and Friedrich, 2004), and Baltic Sea (e.g., Eilola et al.,
2009). Such models have also been used to investigate the mechanisms driving
the spatial distribution (McCreary et al., 2013) and seasonal evolution of
the OMZ in the interior Arabian Sea (Resplandy et al., 2012). There is,
however, no dedicated modeling study to date, addressing the mechanisms that
drive the interannual oxygen variability along the WCI. The present study
aims at identifying the physical controls of the WCI interannual oxygen
variability, with the help of a 40-year-long simulation from a
1/4
We used the 1
We also used oxygen measurements from the Candolim Time Series (CaTS) station
located on the WCI on the inner shelf off Goa (
This study uses the NEMO (Nucleus for European Modelling of the Ocean;
Madec, 2008) model that includes the PISCES (Pelagic Interaction Scheme for
Carbon and Ecosystem Studies; Aumont et al., 2015) biogeochemical component.
The PISCES model has 24 compartments, which include two sizes of sinking
particles and four “living” biological pools, representing two
phytoplankton (nano-phytoplankton and diatoms) and two zooplankton
(microzooplankton and meso-zooplankton) size classes. Phytoplankton growth
is limited by five nutrients: NO
When oxygen falls below a threshold value set to 6
The response of oxygen to dynamical processes is computed as
Finally,
The NEMO–PISCES coupled biophysical model has been successfully applied to
various studies in the Indian Ocean (e.g., Koné et al., 2009; Resplandy et
al., 2009; Currie et al., 2013; Keerthi et al., 2016), including the Arabian
Sea OMZ (Resplandy et al., 2011, 2012). A more detailed model description is
provided in the manuals for NEMO and PISCES, available online at
Here, we specifically use a 1/4
The simulation starts from the rest and the WOA13 climatology temperature and salinity (Boyer et al., 2013). PISCES biogeochemical tracers are initialized from the WOA13 database for nutrients and a global simulation climatology for other tracers (Aumont and Bopp, 2006). After 5 years of spin-up with climatological surface and lateral boundary forcing, the model is forced with the Drakkar Forcing Set version no. 4.4 (DFS4.4; Brodeau et al., 2009) from 1958 to 2012. This forcing is a modified version of the CORE dataset (Large and Yeager, 2004), with atmospheric parameters derived from ERA40 reanalysis until 2002 (Uppala et al., 2005) and ECMWF analysis after 2002 for latent and sensible heat flux computation. Radiative fluxes are taken from the corrected International Satellite Cloud Climatology Project-Flux Dataset (ISCCP-FD) surface radiations (Zhang et al., 2004), while precipitation is specified from a blend of satellite products described in Large and Yeager (2004). All atmospheric fields are corrected to avoid temporal discontinuities and to remove known biases (see Brodeau et al., 2009, for details). In the following, the 1960–2012 period is analyzed.
Northern Indian Ocean surface chlorophyll (mg m
The model's ability to capture the climatological surface chlorophyll
concentrations during the summer and winter monsoons is illustrated in Fig. 2.
During the summer monsoon (Fig. 2a), seasonal blooms are observed along the
coasts of the Arabian Peninsula and along the WCI in response to coastal
upwelling that brings nutrients into the euphotic layer (e.g., Wiggert et
al., 2005; Levy et al., 2007; Koné et al., 2009). The chlorophyll signal
along the Somalia and Omani coasts extends offshore towards the central
Arabian Sea (Fig. 2a) through offshore lateral advection of nutrients from
upwelling regions, either by large-scale circulation or by eddy activity
(e.g., Lee et al., 2000; Resplandy et al., 2011). During the winter monsoon (Fig. 2b),
the cool, dry northeasterly winds in the Arabian Sea induce convective
mixing and entrain nutrient-rich waters to the surface, triggering the
chlorophyll bloom observed north of 15
The comparison of the modeled horizontal and vertical climatological oxygen
distribution with that of WOA13 is shown in Fig. 3. In observations, the
core of the OMZ is confined to the northern part of the basin (Fig. 3a) and
expands vertically between 150 and 1000 m depth (Fig. 3b), with lowest
subsurface oxygen concentrations found in the central and eastern part of the
basin. The OMZ is thus shifted to the east of the region of highest
biological production located along the west coast of the Arabian Sea (Fig. 2).
The oxycline lies around
Both temperature and dissolved oxygen decrease with increasing depth below
the mixed layer (Fig. 3b and d). The oxycline or thermocline depths are
defined as the depths of maximum gradient. It is, however, customary to
approximate those depths from a fixed isocontour, especially in tropical
regions. Resplandy et al. (2012) used the depth of 100
To characterize the IOD variability, we used the standard definition of the
Dipole Mode Index (DMI; Saji et al., 1999), which is calculated as the
difference between the interannual sea surface temperature (SST) anomalies in the western
(50–70
Seasonal evolution of oxygen (
Maps of correlation between the mean seasonal cycle of
oxycline and thermocline depths over the northern Indian Ocean from
The seasonality of the oxygen concentration within the core of the OMZ is
very weak, but this is not the case for the OCD (Resplandy et al., 2012).
Figure 4 shows the depth–time section of the oxygen and temperature seasonal
cycle off the WCI (see black frame in Fig. 5 for the WCI box location) from
both the model and the observations. This figure illustrates that the seasonal
evolution of temperature and oxygen are very similar, with both OCD and TCD
starting shoaling in April (
The above close link between the thermocline and oxycline has also been
observed in many regions (e.g., Morales et al., 1999; Prakash et al., 2013).
Figure 5 quantifies this relationship for the entire northern Indian Ocean by
displaying the correlation between the seasonal OCD and TCD variations for
both observations and the model, and these correlations exceed 0.7 everywhere in
the Arabian Sea, except in a small region off the Horn of Africa. These
correlations drop in the equatorial region, presumably because our oxycline
definition (100
Maps of the seasonal climatology of oxycline (m; shaded)
and thermocline depths (m; contours with 10 m interval) from WOA13 during
The strong dynamical control on the oxygen variability is further
illustrated in Fig. 6, which displays spatial maps of observed and modeled
OCD and TCD seasonal climatologies. During the spring inter monsoon
(March–May), the TCD and OCD are spatially quite uniform and deep
(
In order to further illustrate the remote forcing effects discussed above, and the connection between the offshore oxygen variability and that on the shelf, Fig. 7 shows the seasonal evolution of remote and local winds (Fig. 7a) and a comparison of the climatological near-surface oxygen contents (0–40 m average) obtained from the model and WOA13 offshore of the WCI (i.e., in the WCI box) with those obtained from the CaTS on the shelf (Fig. 7b and c). The modeled and WOA oxygen climatologies in the WCI box match very well for the entire year. In contrast, the CaTs in situ measurements show lower oxygen concentrations. The oxygen concentration on the shelf is expected to be lower than that offshore because of a higher oxygen consumption on the shelf, particularly associated with the sediment respiration process, which in our model is a very simple representation. Despite lower oxygen content on the shelf, there is a very good phase agreement between the oxygen seasonal variations from CaTS in situ measurements and the corresponding offshore variations derived from the WOA and model, suggesting a strong offshore–shelf connection through dynamical upwelling process, as suggested by Banse (1959), Carruthers et al. (1959), and Gupta et al. (2016).
The seasonal wind evolution shown in Fig. 7a allows discussing the respective influences of local (WCI box) and remote forcing (at STI) in driving the seasonal oxycline variations along the WCI. As discussed above (e.g., Fig. 4), the upwelling off WCI starts developing (reduction in upper-ocean oxygen content) at the beginning of the summer monsoon (April–May), reaches a maximum (minimum upper-ocean oxygen content) by September–October, and decays (increased upper-ocean oxygen content) by November. The local winds along the WCI (black continuous curve in Fig. 7a) are favorable to upwelling only during the southwest monsoon (i.e., they have an alongshore southward component only from June to August), which indicates that the upwelling along the WCI is to a large extent driven by remote winds (Shetye et al., 1990). The wind near the STI is upwelling-favorable from about April to October and hence matches the seasonality of the upwelling of cold and low-oxygen waters, in agreement with the results of Suresh et al. (2016).
The gaps in the CaTS observational dataset do not allow constructing a reliable time series of interannual upper-ocean oxygen content, to which our model could be validated. However, the data can still provide some estimate of the amplitude of interannual variability, which can be compared to that of our model (the whiskers in Fig. 7b and c indicate the amplitude of the variability around the mean seasonal cycle). Despite a slight underestimation in our model, the amplitude of the near-surface oxygen interannual variability is largest during SON, both in the shelf observations and the model. This further corroborates the offshore–shelf connection discussed above.
A seasonally shallow OCD combined with a larger interannual variability in fall creates a window of opportunity for the occurrence of coastal anoxic events. Figures 7d and e display the monthly percentages of the occurrence of hypoxic profiles from CaTS (on the shelf) data and the model (offshore). While the general patterns of oxygen (oxycline) variability on the shelf and offshore of the coast remain similar, the actual upper-ocean oxygen content and vertical oxygen profiles are different. Hence, we have used different thresholds to detect hypoxic profiles in the observation and the model. Consistent with previous literature, anoxic events are most likely to occur from August to November in the model and the shelf data, as expected from the very shallow background oxycline at that time of the year. This justifies our focus on the fall period for analyzing the processes that drive the modeled interannual variability of the WCI oxycline in the following.
We previously demonstrated a tight relationship between the seasonal
variability of OCD and TCD in the eastern Arabian Sea. Figure 8a exhibits a
similar relation in large portions of the northern Indian Ocean for fall
interannual OCD and TCD anomalies in the model. A comparison with
observations is unfortunately not feasible due to lack of a basin-scale
dataset for interannual OCD anomalies. The correlation between interannual
OCD and TCD anomalies are, in general, slightly weaker than that on seasonal
timescales (Fig. 5b) but remain high in a large part of the Indian Ocean
north of 5
The influence of remote forcing at WCI is further established in Fig. 9. Due
to the unavailability of the continuous oxygen observations (see Sect. 2.1),
we cannot directly evaluate the modeled oxygen interannual variability in
the WCI region. However, we evaluate the modeled TCD interannual variability,
which is closely tied to the OCD interannual variability (
Regression patterns of fall interannual anomalies of
modeled
The relationships between the modeled interannual variability of the OCD
along the WCI and that of the OCD, TCD, SST, and wind on the basin scale are
demonstrated in Fig. 10a and b, which show regression maps of fall
interannual anomalies of these variables to the time series of the fall WCI
OCD anomalies shown in Fig. 8b (black line) over the 1960–2012 period. These
maps display the typical basin-scale anomalies corresponding to an
anomalously deep OCD in the WCI. Consistent with Fig. 9, the fall WCI OCD
variations are not merely local but are associated with basin-scale
ocean–atmosphere interannual anomalies over the entire equatorial and
northern Indian Ocean. An anomalously deep OCD off the WCI is usually
associated with deeper OCD and TCD (i.e., positive anomalies) in the
southeastern Arabian Sea and in the vicinity of Sri Lanka and the STI (Fig. 10a and b). Positive OCD anomalies off the WCI are also related to
shallower OCD and TCD in the eastern Indian Ocean and along the eastern rim
of the Bay of Bengal (i.e., negative anomalies). The associated large-scale
wind patterns (Fig. 10a) explain these interannual OCD and TCD patterns.
Anomalous easterlies in the equatorial band force upwelling equatorial
Kelvin waves that shoal the OCD and TCD in the EEIO. These signals further
propagate around the rim of the bay as upwelling coastal Kelvin waves,
thereby shoaling the TCD and OCD there (e.g., McCreary et al., 1993, 1996;
Aparna et al., 2012). Similar to what happens on the seasonal
scale (Suresh et al., 2016), easterly zonal wind stress anomalies in the
vicinity of Sri Lanka and the STI force a downwelling coastal Kelvin wave
that propagates poleward along the western Indian coastline, resulting in
a deepening of the TCD and OCD there. The strong negative correlation between
the WCI OCD and STI zonal winds interannual fluctuations (
The SST variations associated with the OCD signals in the WCI region are
characterized by a clear signal in the EEIO (near the Sumatra coast) and
weaker signals of opposite sign in the western Indian Ocean. As pointed out
before, the patterns shown in Figs. 9b, c and 10a, b are reminiscent
of the IOD signature (Saji et al., 1999; Webster et al., 1999; Murtugudde et
al., 2000), an Indian Ocean coupled ocean–atmosphere climate mode that peaks
in fall, as discussed in the introduction. This is further demonstrated in
Fig. 10c and d, which display regression maps of interannual anomalies of
OCD, TCD, SST, and winds onto the boreal fall DMI. The resulting patterns,
representing the typical anomalies associated with a positive IOD phase, are
strikingly similar to those displayed in Fig. 10a and b (pattern
correlation of
Scatterplot of the fall interannual anomalies of modeled WCI oxycline depth against modeled DMI. Red and blue crosses, respectively, indicate positive and negative IOD events (defined as events when DMI exceeds 1 standard deviation). Years of anoxic events reported by Naqvi et al. (2009) are marked as stars, along with the corresponding years.
The influences of positive and negative phases of IOD on the WCI OCD (and
TCD, not shown) are, however, not symmetrical (Fig. 11a): most of the positive
IOD events cause a deepening of OCD along WCI (e.g., 1961, 1967, 1994,
1997), while negative IOD can either be associated with a shoaling (e.g.,
1996, 1998, 2010) or even a deepening (e.g., 1979–1981). To further illustrate
this asymmetry, Fig. 12 provides a scatterplot of the fall DMI versus the
fall interannual anomalies of OCD off the WCI. This scatterplot confirms
that there is an asymmetric impact of positive and negative IODs on the OCD
along the WCI. The WCI OCD response to positive IODs is very robust, with
all positive IOD events (DMI > 1) except one being systematically
associated with a deepening. The WCI response to negative IOD is weaker and
much less systematic. As discussed above, negative IOD events (DMI <
Seasonal evolution of anomalous composites of
The year-to-year variations of coastal hypoxia along the west coast of India
(WCI) have been identified by Naqvi et al. (2000, 2009), along with their
strong impacts on fisheries and the ecosystem. The mechanisms controlling these
variations have, however, not yet been elucidated. The present study offers
new insights on the physical controls of coastal hypoxia along the WCI. To
that end, we used an eddy-permitting (1/4
The shallow oxycline in fall combines with a large interannual variability at this time of year to create a window of opportunity for coastal anoxic events. Our model analysis further indicates that there is a tight coupling between the thermocline and oxycline variability in this region on both seasonal and interannual timescales, indicative of a strong physical control of the oxygen variability through vertical advection. Interannual thermocline fluctuations along the WCI are related to basin-scale wind, thermocline, and oxycline depth perturbations associated with IOD events, an Indian Ocean coupled ocean–atmosphere climate mode that peaks in fall. Positive IOD events are associated with easterly wind anomalies in the central equatorial Indian Ocean and extend meridionally up to the southern tip of India. These easterly wind anomalies trigger downwelling coastal Kelvin waves that propagate along the WCI and deepen the thermocline and oxycline in boreal fall, thereby preventing the occurrence of coastal anoxia off the WCI during positive IOD events. Our model results also suggest an asymmetry between the impact of positive and negative IOD events on the WCI oxycline depth. The westerly wind anomalies at the southern tip of India do indeed have a smaller amplitude during negative IODs than their easterly counterparts during positive IODs, thus resulting in a weaker and less consistent shoaling of oxycline and thermocline along WCI during negative IOD events.
Previous studies have demonstrated the impact of large-scale climate modes on year-to-year variations of the oxygen deficiencies in coastal hypoxic systems elsewhere in the world ocean. In the Pacific, El Niño conditions lead to intensified oxygenation along the coasts of Peru and Chile as a result of weak upwelling (e.g., Arntz et al., 2006; Gutierrez et al., 2008), while in the Atlantic, the Benguela Niño leads to intensified anoxia along the Namibian shelf (Monteiro et al., 2008). The western continental shelf of India is home to the largest naturally formed coastal hypoxic system in the world. In this study, we identify, for the first time, the IOD as the major climatic driver of the year-to-year oxycline and thermocline variations offshore of the WCI. Though the IOD has a weaker thermocline depth signature on the west than on the east coast of India, it has stronger societal consequences as it influences the WCI seasonal upwelling that brings suboxic waters very close to the surface during fall. Although the IOD influence on the west Indian coast has never been reported so far, it has regularly been reported in the in the Bay of Bengal. In line with our results, Aparna et al. (2012) did indeed show that IOD events drive strong sea-level and thermocline fluctuations along the rim of the bay in fall, through coastal Kelvin wave propagation from the equatorial region. Akhil et al. (2016) further demonstrated that this remote forcing also drives counterclockwise anomalous horizontal currents in fall in the bay, which in turn leads to large interannual variations of sea surface salinity in the southern Andaman Sea. On the biogeochemical side, Wiggert et al. (2002, 2009) and Currie et al. (2013) demonstrated that IOD events are responsible for large interannual chlorophyll variations in the southeastern Bay of Bengal and at the STI. Finally, the IOD signature found in the Arabian Sea in the present study has already been described for the Bay of Bengal in terms of sea level (Aparna et al., 2012) and chlorophyll (Currie et al., 2013).
The influence of IOD is further shown to be larger for its positive than its negative phase. Our results suggest that part of the weaker WCI oxycline depth response during negative IOD may be explained by the weaker wind stress anomalies at the STI associated with negative IOD events. This weaker wind amplitude could simply be related to the tendency of negative IOD events to be weaker than their positive counterpart (Saji and Yamagata, 2003; Hong et al., 2008; Cai et al., 2013) or to asymmetries in the spatial patterns of winds associated with the nonlinear response of deep atmospheric convection to SST anomalies of each phase of the IOD. A more precise understanding of this asymmetry would require an in-depth investigation of the processes that control the wind variations at the STI and the thermocline along the WCI in response to positive and negative IOD events.
Our findings partly explain the substantial year-to-year changes in both the duration and intensity of the observed seasonal oxygen deficiency over the western Indian shelf (Naqvi et al., 2009). None of the anoxic events reported by Naqvi et al. (2009) (black stars in Fig. 12) lies in the upper right quadrant of the scatterplot shown in Fig. 12, indicating that positive IODs systematically prevent the occurrence of anoxic events. For instance, the relaxation of anoxic conditions in early fall 1997 reported by Naqvi et al. (2009) is in line with the occurrence of very strong positive IOD during that year. Most anoxic events are found in the lower left quadrant, i.e., near neutral or negative IOD conditions and an anomalously shallow offshore oxycline. Neutral or negative IOD years are, however, not necessarily anoxic, indicating that a neutral or negative IOD is a necessary but not a sufficient condition for severe anoxia. A recent study by Gupta et al. (2016) revealed that the oxygen deficiency in 1959 along the WCI was more severe than in 2012, a conclusion consistent with the occurrence of a negative IOD in 1959 and a positive one in 2012. Similarly, in situ measurements also revealed that subsurface oxygen concentrations were significantly lower at the turn of the 20th century than in the 70s (Naqvi et al., 2009): our simulation exhibits a similar behavior (see Fig. 11a), showing many years with shallower than normal OCD in the later period and systematically deeper than normal OCD during 1970s. The causes for those decadal variations need to be investigated in greater detail.
The
It must, however, be kept in mind that other factors are also likely to contribute to the reported interannual fluctuations of hypoxic conditions in this region. Naqvi et al. (2009) for instance suggested that increased productivity due to increased nutrient loading from land associated with anthropogenic activities might have the potential to trigger a shift from natural suboxic to anthropogenic anoxic conditions during recent decades. This hypothesis, however, cannot explain the relaxation of the intensity of oxygen deficiency in recent decades. Another contributing factor could be related to changes in local hydrographic variations. For instance, interannual variations of the land runoff along the Western Ghats, local precipitation during the summer monsoon, or input of Bay of Bengal freshwater during the northeast monsoon (e.g., Jensen, 2001) could modulate the upper-ocean haline stratification, ventilation of the subsurface waters, and hence the subsurface oxygen content along the WCI. Finally, local alongshore wind variations may modulate the intensity of coastal upwelling and hence the amount of oxygen-depleted waters brought to the shelf. The influence of these factors thus requires further investigation.
An obvious limitation of the current study is the spatial resolution of our
model (
While our model does not reproduce the details of exchanges between the shelf and open ocean, we have just used it as a proxy of the behavior of open ocean, off the WCI. Several studies have already pointed towards the influence of offshore oxygen variations on the variability of hypoxic conditions along other coastal regions (e.g., Grantham et al., 2004; Helly and Levine, 2004; Arntz et al., 2006; Gutierrez et al., 2008). As was shown in Figs. 6 and 7b, the model and WOA climatology vertical oxygen distributions agree quite well, both in terms of the oxycline depth and near-surface value. The CaTS data, on the other hand, is representative of what happens much closer to the coast and displays much lower oxygen levels than seen further offshore in WOA and the model. This may of course partially be due to shortcomings in representing physical exchanges between the shelf and open ocean at the current resolution of our model and existing oxygen dataset in the region. But biological processes are also known to be a prominent oxygen consumption factor on the shelf, in particular in the benthic zone where the enhanced concentration of particulate matter above sediments is associated with a high oxygen demand (e.g., Cowie, 2005). The crude parameterization of sediments in the model probably does not consume enough oxygen very close to the coast. On the other hand, the good phasing between the oxygen seasonal variability offshore (in the model and WOA) and shelf (CaTS) data (Fig. 7b and c) suggests that the offshore variability is probably an important driver of the oxygen content on the shelf. However, a proper representation of benthic biological processes would probably be needed to represent the low oxygen levels very close to the coast (Fig. 7c). Dedicated studies at a higher spatial resolution with sensitivity tests on the representation of nearshore biological processes will probably be needed in order to better understand how the representation of nearshore biological processes constrains the coastal oxygen representation.
On the observational front, the current spatiotemporal sampling does not allow building reliable long-term time series of the month-to-month oxygen variations along the shelf and offshore. Despite the establishment of frequent measurements of the oxygen profile off Goa since September 1997, the numerous unsampled months (July and August are almost unsampled because of rough weather conditions) and the strong sub-monthly variability prevent a continuous monitoring of oxygen variations along the WCI. A reasonable number of moorings or Argo drifters with oxygen and temperature sensors along the shelf and further offshore would allow a finer description of the oxygen variability and of its relationship with temperature and connection with the offshore variations. In order to establish unequivocal evidence for the shelf–open-ocean interactions, future studies should also consider improved observations such as repeated glider transects or triad of moorings (shelf, shelf break, and open ocean) monitoring both physical and biogeochemical quantities in this region.
Percentage of the occurrence of near-surface hypoxic
conditions in the model during fall. Near-surface hypoxic conditions are
defined as profiles with an oxygen concentration below 80
Though the present study is focused on the WCI, our Indian Ocean
configuration model allows assessing other regions where near-surface
hypoxia can occur. Figure 14 shows the percentage of profiles where oxygen
concentrations below 80
The (CaTS) data archived at the CSIR-NIO data center and Chemical
Oceanography Division have been used in Figs. 1 and 7. These datasets are not
publicly available. Hence we do not provide the data or any access link.
All other datasets are publicly available: WOA13 –
The authors declare that they have no conflict of interest.
This work is a part of a CSIR-funded INDIAS IDEA project. I. Suresh acknowledges financial support from the Council of Scientific and Industrial Research, New Delhi, and INCOIS/MoES (HOOFS program). V. Parvathi is funded by CSIR under a Senior Research Fellowship. We thank the CSIR-NIO data center and Chemical Oceanography Division for making the archived cruise and CaTS data available for the present study. We thank the NEMO–PISCES modeling team. The simulations were performed on HPC Pravah at CSIR-NIO. We thank M. Afroosa for assistance in data processing. M. Lengaigne, C. Ethé, J. Vialard, and M. Levy benefited from Institut de Recherche pour le Développement (IRD) funding for their visits to the CSIR-NIO. This is NIO contribution number 6008. Edited by: L. Bopp Reviewed by: two anonymous referees