1. Introduction
The role of the Indian Ocean (IO) in interannual climate variability is now well recognized. It has been firmly established that there exists an independent coupled ocean–atmosphere mode in the Indian Ocean, known as the Indian Ocean dipole (IOD; Saji et al. 1999; Webster et al. 1999), characterized in its positive phase by warm SST anomalies in the western Indian Ocean and cold SST anomalies in the eastern Indian Ocean.
Apart from SST, the sea surface salinity (SSS) is another oceanic parameter of interest because of its importance in ocean–atmosphere coupling in the Indian Ocean. Because of the upcoming satellite salinity missions like Soil Moisture and Ocean Salinity Mission (Kerr et al. 2003) and Aquarius (see online at http://aquarius.nasa.gov), the analysis of SSS variability simulated by ocean circulation models assumes increased importance. Han et al. (1999, 2001) developed and used a 4½ layer model to study the influence of salinity on various aspects of Indian Ocean dynamics, thermodynamics, and mixed layer physics. While Han et al. (2001) used climatological forcings, Perigaud et al. (2003) studied the impact of interannual rainfall anomalies on SSS using the same 4½ layer model. In the present work, the interannual variability of oceanic parameters has been studied using the model developed by Han et al. (1999) with special emphasis on the analysis of sea surface salinity. The distinguishing feature of the present study is the use of satellite scatterometer–derived winds and rainfall forcings. Also, unlike the study by Perigaud et al. (2003), river runoff has been included in the model simulations. The use of scatterometer winds follows the recommendation of previous researchers (Grima et al. 1999; Hackert et al. 2001). Normally, ocean circulation models are driven by winds from numerical weather prediction centers. A widely used wind product is from the National Centers for Environmental Prediction (NCEP) reanalysis. However, Goswami and Sengupta (2003) found that the NCEP reanalysis surface winds are underestimated in the equatorial Indian Ocean. Also, Babu et al. (2004) compared the European Remote Sensing Satellite-2 (ERS-2) wind speed with wind speed observed by a moored buoy in the central Arabian Sea (AS). They found a reasonable agreement between the two wind speeds with a correlation of 0.88 and root-mean-square error (RMSE) of 1.68 m s−1. The present authors have found that the correlation between NCEP wind speed and the wind speed measured by the same buoy is 0.74 and the RMSE is 2.29 m s−1. In light of these findings, it seems logical to study the impact of ERS-2 scatterometer winds on the simulations of the model used in the present work. The use of satellite-derived rainfall as freshwater flux forcing follows the recommendation of previous researchers (Doney et al. 2003). The other distinguishing feature of the present study is the detailed validation of the parameters simulated by the model using in situ observations. In fact, Perigaud et al. (2003) appealed for such validation of model salinity with SSS measurements taken in the Indian Ocean.
The paper is organized in the following manner: In section 2, the model and the forcing parameters are described. Section 3 describes the sensitivity experiments carried out with different wind, heat flux, and precipitation products. This section also includes the detailed comparison of the two wind products (NCEP and ERS-2) used to force the model as well as a comparison of model-simulated variables with in situ observations. In section 4, the interannual variability of the simulated sea surface salinity has been analyzed in detail. Section 5 discusses the role of various oceanic processes like advection, upwelling, etc., in determining SSS variations. Section 6 summarizes the main findings of the study.
2. Model and forcing
The model used is a 4½ layer thermodynamic Indian Ocean model (IOM) that extends throughout the IO north of 29°S (Fig. 1). It is an extension of the McCreary et al. (1993) model. The main difference from the original McCreary et al. (1993) model is that it has more layers (4½). Also, there is a salinity equation in each model layer. Other distinctive features of the present model are velocity shear between the mixed layer and the fossil layer for simulation of a better Wyrtki jet (RWJ) and the inclusion of the effect of Red Sea water.
The model details are available in Han et al. (1999, 2001) and Han and McCreary (2001). For the sake of completeness, the model is described in brief. The model ocean consists of four upper-ocean layers with velocities Vi = (ui, υi), layer thicknesses hi, temperatures Ti, and salinities Si (i = 1, 2, 3, 4 is a layer index) overlying a deep inert ocean with temperature Td = 3°C and salinity Sd = 34.8 psu.
The layers correspond to distinct oceanic region types, namely the surface mixed layer (layer 1), the diurnal thermocline (layer 2), the seasonal thermocline (layer 3), and the main thermocline (layer 4). Fluid is allowed to transfer between the layers with velocities w1, w2, and w3, and the system is thermodynamically active in that the temperature and salinity in each layer are allowed to vary horizontally in response to surface heat and buoyancy fluxes, horizontal advection, entrainment, and detrainment. Finally, h1, h2, and h3 are not allowed to become thinner than prescribed minimum values. The horizontal resolution of the model is 0.5° in zonal and meridional directions and the time step is 0.8 h.
In these equations, υi = (ui, υi), hi, Ti, and Si are instantaneous values of velocity, layer thickness, temperature, and salinity, respectively, and w0 = υ0 = T0 = S0 = 0. The wind stress is τ, Qi is the heat flux, E is evaporation, P is precipitation, and f is the Coriolis parameter. The wi terms are velocities at the base of layer i that specify how water transfers between layers i and i + 1. No transfer is allowed between layer 4 and the deep ocean so that w4 = 0. Velocity w1 is essentially determined as in the Kraus and Turner (1967) mixed layer model, and w2 primarily relaxes h1 + h2 back to a prescribed thickness (65 m) to parameterize the process of subduction (McCreary et al. 1993). This subduction process is not allowed to happen within the equatorial band of 5°S–5°N, as given by Han and McCreary (2001). Velocity w3 is included to keep h3 from being thinner than h3min = 50 m. Here, wi+ = max (wi, 0) and wi− = min (wi, 0), the positive and negative parts of wi, ensure that momentum and mass are conserved when water moves between layers.
The climatological forcings (winds, thermal, and precipitation) used for spinning up the model from rest are from Legler et al. (1989), Rao et al. (1989), and Legates and Willmott (1990). In the spinup run and subsequent interannual runs, thermal forcing consists of net shortwave and longwave radiation at the ocean surface, near-surface air temperature (Ta), and near-surface specific humidity (qa). The model computes its own latent and sensible heat flux using bulk formulas. SST used in these formulas is the model-computed SST. For the interannual runs, ERS-2 scatterometer–derived wind has been used as dynamic forcing. Thermal forcings have been taken from the NCEP reanalysis and rainfall data are from the Global Precipitation Climatology Project (GPCP; Huffman et al. 2001). GPCP datasets are available globally on a 2.5° × 2.5° latitude–longitude grid and are generated by merging infrared and microwave satellite estimates of precipitation with rain gauge data from more than 6000 stations. All the data used in the interannual runs (winds, thermal, and rainfall data) are for the period 1993–2000 and are monthly averaged. For validating model outputs, the hydrographic data from the World Ocean Circulation Experiment (WOCE) Global Data products (version 3.0) have been used. These datasets contain data from the U.S. Joint Global Ocean Flux Study (JGOFS) program. Arabian Sea biology and upper-ocean physics were intensively studied from 1994 to 1996 under this program. The JGOFS program consisted of six cooperating and complementary activity elements. Also, the data from the central AS buoy (Weller et al. 1998) deployed by Woods Hole Oceanographic Institution (WHOI) during the period 1994–95 have been used for comparison.
3. Model experiments
The model has been spun up from rest for a period of 60 yr to achieve a steady state using climatological forcings. A suite of experiments (Table 1) has been carried out. They differ in the nature of their winds, fluxes, and rainfall forcings.
a. Experiments 1 and 2
In experiment 1, the model is forced by NCEP winds, NCEP radiative fluxes [net shortwave and net longwave fluxes at the ocean surface; Ta and qa), and climatological precipitation. The model-simulated SST for October 1997 has been compared with (Advanced Very High Resolution Radiometer) AVHRR-derived SST. In Fig. 2a, the difference in SST observed by AVHRR and simulated by the model is shown. There is a big discrepancy in the simulation of contrasting SST patterns in the south equatorial Indian Ocean (SEIO) and western Indian Ocean during October 1997, an Indian Ocean dipole year. The difference reaches up to 3°C in the southeastern region of tropical Indian Ocean. In the absolute SST distribution, the SST in the southeastern part of the Indian Ocean observed by AVHRR is of the order of 25°–26°C (2°–3°C below the long-term average), whereas in the model simulation with NCEP winds, it is greater than 28°C.
In experiment 2, the NCEP winds are replaced by ERS-2 winds. A distinct improvement can be seen in the model simulation, when NCEP winds are replaced by ERS winds (Fig. 2b). Except for a small patch north of the equator toward the western side, where the differences are slightly high (greater than 1°C), there is an overall improvement in the model simulation. One more experiment has been performed in which the NCEP fluxes are replaced by climatological fluxes. The ERS-2 winds are used as dynamical forcing. It is found that even in this run, the model is able to pick up the dipole signature in the SST pattern. This possibly brings out the dominance of wind over radiation in the generation of dipole structure. It has to be mentioned here that wind plays a dual role in simulating the SST pattern. The wind enters the evolution equation of SST primarily through the advective terms. However, it enters the equation via the heat flux term also, since it is used in the bulk formulas for computing the latent and sensible heat fluxes. It is thus not entirely surprising that the model is able to generate the dipole structure even when forced by climatological fluxes.
Since the simulations have been obtained for 8 yr (1993–2000), it will be interesting to see the overall bias of the two (NCEP and ERS-2 forced) SST solutions and their root-mean-square (RMS) variability patterns. Figure 3 shows the model bias in terms of differences between AVHRR-derived mean SST and model-simulated mean SST in the NCEP run (Fig. 3a) and in the ERS-2 run (Fig. 3b). Once again, one can notice a large bias in the NCEP solution in the SEIO sector. The biases in the equatorial region are of opposite signs toward east (warmer SST in the model simulation) and west (cooler SST in the model simulation). Outside the region of the equator, the NCEP SSTs are in agreement with observed SSTs. In contrast, the equatorial region exhibits practically zero bias in the scatterometer simulation. It can be concluded that the maximum impact of scatterometer winds in the Indian Ocean is felt in the equatorial region. In other regions, the two simulations are in close agreement with each other. Figures 4a–c show the distribution of standard deviations of the AVHRR SST and SST simulated using NCEP and ERS-2 winds, respectively. The largest variability is located in the western equatorial Indian Ocean (WEIO) and eastern equatorial Indian Ocean (EEIO) in agreement with the findings of Murtugudde and Busalacchi (1999) and Behera et al. (2000). Once again, the SST variability in the scatterometer solution is closer to observed variability. High variability in the SEIO observed in AVHRR SST is practically absent in the NCEP solution but has been picked up well in the ERS-2 solution.
b. The two wind products (NCEP and ERS-2)
In the last section, it was found that winds play a dominant role in the generation of significant large-scale SST variability in the Indian Ocean. Hence it is logical to compare the two wind products to understand the reason for the difference in the two simulations. It has been mentioned in the introduction that the ERS wind speed is in better agreement with buoy measurements in the central AS than the NCEP wind speed. It is, however, instructive to compare the two products in the entire basin. In Fig. 5, the mean difference (1993–2000) between NCEP and ERS wind speeds is shown. It is seen that the NCEP wind is underestimated in comparison with the ERS-2 wind in the SEIO. It is known that cooling in this region during dipole years is primarily due to upwelling induced by Ekman pumping caused by anomalous winds. Since the NCEP wind in this region is found to be weaker than ERS-2, the cooling is not as pronounced in the NCEP solution as in the ERS-2 solution. The difference is also large in some pockets of the southern Indian Ocean. Figure 6 compares the mean seasonal cycles of zonal and meridional winds from NCEP reanalysis and ERS-2 scatterometer in two regions of the Indian Ocean calculated over the study period (1993–2000). It is seen that during summer, NCEP wind becomes weakly westward in the central equatorial Indian Ocean (lhs of Fig. 6a). This is in conformity with the result reported in Schott and McCreary (2001), and this difference causes differences in model simulations when forced by the two wind products. The meridional winds do not differ much. In the northwestern AS (8°–15°N, 54°–60°E), the difference between the two wind climatologies is practically negligible (Fig. 6b). Thus, the difference seems to be pronounced in the central and eastern equatorial Indian Ocean, in agreement with the result of Goswami and Sengupta (2003). The interannual anomalies for the NCEP and ERS winds have been computed in the same two regions. The results are shown in Fig. 7. The large westward zonal anomalies can be seen in the figure during the second half of 1994 and 1997 (the latter being stronger), which resulted in the generation of the IOD. However, such an anomaly is practically absent in the other region (lhs of Fig. 7b). The meridional anomalies are smaller compared to the zonal anomalies. Overall, ERS-2 winds in the equatorial region are stronger (the mean as well as anomalies) than the NCEP winds that resulted in the difference in model simulations.
c. Experiment 3
In this section, two more experiments differing in the nature of their rainfall forcing, climatological rainfall forcing (RClim) and interannually varying rainfall forcing (RGPCP), are reported. The impact of these two forcings on the simulation of SSS has been studied while keeping the dynamical (ERS-2 winds) and heat flux forcings (NCEP) the same. Salinity effects are known to be quite important in the Indian Ocean (Vinaychandran et al. 1999; Han et al. 2001).
On the lhs of Fig. 8a, SSS anomalies for October 1997 in the RClim solutions are shown. The monthly anomaly of SSS is the departure from the corresponding 8-yr averaged monthly mean SSS. In Fig. 8b, the distribution of the two corresponding rainfall products is shown. The magnitudes of the SSS anomalies are different in the equatorial region in the two experiments. In other regions like the AS and the southern Indian Ocean, the differences are practically negligible. However, the anomalies are quite large in the Bay of Bengal (BOB). In the eastern BOB, the anomalies are slightly less in magnitude in RGPCP than in RClim, as the rainfall in 1997 was more than the climatological rainfall in that region. One noteworthy feature is the low-salinity tongue present in the central equatorial Indian Ocean. It can be seen that this feature is present in both the runs. However, the freshening is more in RGPCP than in RClim. This feature is also present in October 1994 but absent in other years. Another point to be noted is that this region witnessed lesser rainfall in October 1997 than in 1998. Hence, processes affecting the central equatorial freshening are more due to advection caused by anomalous winds than precipitation. This is discussed in detail later. In the SEIO near the Sumatra coast, the SSS anomaly is positive in both the runs. However, the surface water is much more saline—the anomaly being greater than 1.2 psu in RGPCP, whereas in RClim, its value is 0.9 psu. Also, the spread of this high saline water is more in RGPCP than in RClim. This spread is in accordance with the precipitation distribution. In 1997, the precipitation was less compared to climatology in this region. However, it needs to be mentioned that higher salinity in RGPCP is not due to deficient precipitation alone. In fact, it will be shown later that upwelling had a major role to play in modulating the SSS pattern in this region.
d. Comparison of the oceanic variables with JGOFS and WOCE observations
In Fig. 9a, comparisons of monthly averaged SST, SSS, and mixed layer depth (MLD) simulated by RClim and RGPCP at a buoy location in the central Arabian Sea are shown. This buoy, mounted by Woods Hole Oceanographic Institution, took measurements from October 1994 to October 1995. The buoy data (temperature and salinity) are also averaged on a monthly basis. MLD simulated by the model is the thickness of the first model layer. MLD from the buoy is computed using fixed density difference criteria following Levitus (1982). The general pattern of time variation of SSS has been simulated reasonably well in both the experiments. Generally the model and the observations agree to within 0.2 psu, while from April 1995 to October 1995, the departure of RGPCP simulations from observations is slightly more (even more than 0.3 psu). The result may seem to be somewhat surprising. The possible reason for this discrepancy could be the coarse resolution (2.5°) of the GPCP product compared to the fine resolution (0.5°) of climatological rainfall product. Therefore, the interpolation error might have contributed to the mismatch between RGPCP simulations and buoy observations. However, the model has nicely picked up the dip in salinity during the monsoon months. This dip in salinity is due to the addition of freshwater at the surface. This freshwater is from upwelling at the coast, which is further augmented by condensation in the areas where the surface is cooler than the condensation point. In the same figure, the model-simulated MLDs in the two runs have been compared with buoy MLD. The two experiments yield almost identical MLD during the entire year at the buoy location. However, it is interesting to note that the MLD simulated by the model does not differ from the in situ MLD by more than 30 m most of the time. The comparison of model-simulated MLD in the central Arabian Sea with mooring observations shows relatively large differences in March and September. In these months, the MLD shoals because of positive net heat gain and suppressed turbulent wind mixing. In fact, it is only in these 2 months that the model shows deviation of approximately 30 m. In other months, the model MLD shows much better agreement with the observed MLD.
It is interesting to note that McCreary et al. (2001) used climatological and daily wind fields to force a somewhat similar model and concluded that the solution driven by daily fields with diurnal forcing reproduced the observed MLD field most faithfully. Incidentally, they used the same buoy observations in the AS for validating the model results. However, the daily wind field used by them was a blend of daily-averaged winds measured by the buoy at the mooring site and other products. Such in situ observed winds in the ocean are generally not available at all the locations. In the present analysis, the model has been run using satellite-observed winds only, albeit monthly averaged. No in situ observed wind has been blended in our forcing. It is interesting to note that even then the model is able to perform a reasonably good simulation of MLD at the buoy location.
One can see that there is negligible impact of interannually varying rainfall on the simulation of SST in this region. Both of the runs simulate the SST reasonably well during most parts of the year except for June and July. This can be attributed to the errors in the fluxes due to uncertainty in the cloud cover during these months.
The large bias (24 m) in simulated MLD might have been due to the following reason: We used WOCE temperature profiles to compute MLD. In this, MLD was taken to be the depth where temperature decreases by 1°C from that at the surface. This might have created a large error in computing observed MLDs. However, at the central Arabian Sea buoy point, we have used a density criterion for computing observed MLD. Hence, at this point, there is no significant bias. However, there is large overestimation of MLD in September by the model. The impact of this overly deep MLD would be to reduce the surface currents because the wind stress is now distributed over a deeper-than-normal mixed layer.
4. Interannual variability of sea surface salinity
First, the statistics of SSS over the tropical Indian Ocean simulated in the RGPCP run during the study period (1993–2000) are discussed. Figure 10a shows the mean SSS field averaged over the entire study period. The most obvious feature is the high salinity (in excess of 35 psu) in the AS and also toward western IO and low salinity in the eastern IO, which encompasses BOB and EEIO regions. The low-salinity areas are consistent with high rainfall and freshwater influx from BOB Rivers, while the high salinity in the AS is due to strong evaporation. While the broad features are in agreement with Perigaud et al. (2003), the magnitudes are quite different, especially in the low-salinity areas. The reason could be the effect of river runoff that has been considered in the present work. The model simulates pockets of very low-salinity regions along the east coast of India and also in the north BOB. In these regions, the salinity gradients are also quite strong. It is expected that SSS will have strong temporal variability. Accordingly, in Fig. 10b, temporal variability is shown in terms of the standard deviation of the simulated SSS. The AS exhibits small variability. Strong SSS variabilities (>0.4 psu) simulated by RGPCP are located in the EEIO, southwest of the Indian subcontinent and in the northern BOB. These regions are also the regions with the low salinity (Fig. 10a) due to excess precipitation over evaporation. In fact, Webster (1994) has shown that the buoyancy flux contributions from the net surface heat flux and freshwater flux are of about equal magnitude in the Indo-Pacific warm pool. This is in contrast to much of the remainder of the Tropics, where the relative heat flux contribution is considerably larger.
The amplitude of the SSS variability simulated by RGPCP in the eastern and central EIO region is much stronger than that simulated by Perigaud et al. (2003). Strong variability in this region could be due to the differences in the forcings (winds and rainfall) used in this study from the forcings used by Perigaud et al. (2003). Moreover, the anomalies in the BOB circulation due to the inclusion of river discharge in the present work can affect the SSS variability in the EIO region via remote effect. It would be instructive to investigate the equatorial current patterns and their interannual variability in order to understand the salinity changes associated with oceanic processes.
The 8-yr simulation clearly demonstrates a large interannual variability in surface currents (Fig. 11). In the earlier 2 1/2 layer model (McCreary et al. 1993), the fall Wyrtki jet could not be reproduced. A possible cause for this difference and other differences (e.g., differences between model currents and ship-drift currents in the near-equatorial ocean) could be attributed to the lack of vertical shear between the mixed and fossil layers, which underestimated the importance of wind-drift flows confined to the mixed layer. It has been mentioned earlier by us that in the present model, there is a velocity shear between the mixed layer and the fossil layer. Hence, the model could correctly simulate the semiannual eastward-propagating Wyrtki jet (Wyrtki 1973) during May–June and October–November of each year. In fact, this fall Wyrtki jet, or rather the reversal of it during dipole years, is the main focus of the present study. The model clearly shows a complete reversal of the Wyrtki jet during October 1994 and October 1997. It is much stronger and more organized in 1997 (70 cm s−1) than in 1994 (40 cm s−1). This RWJ due to the anomalous easterlies is reinforced by the south equatorial current that flows west and also by another branch coming from the eastern side of the Bay of Bengal (Fig. 12). One can notice a complete change in the current patterns in the BOB during these extreme events. In the normal years, the reflection of the equatorial Kelvin waves from the eastern boundary of the ocean radiates into and around the perimeter of the Bay of Bengal (McCreary et al. 1993). A test run carried out by McCreary et al. that switched off winds in the equatorial waveguide resulted in practically no reflection from the eastern boundary, suggesting the role of equatorial winds in the Bay of Bengal circulation. Ship-drift observations show a northward current from November to January (Shetye and Shenoi 1988; Rao et al. 1989, 1991). Satellite images reveal that the current develops first in the south and then spreads northward (Shenoi et al. 1992). In the present study, we found that a well-developed equatorward current appears during 1994 and 1997 that is practically absent during other years. This equatorward current coming from the BOB feeds the RWJ, making it stronger. It can also be noticed that equatorward currents in the eastern AS cease during dipole years, whereas they are very prominent during normal years. This strong RWJ results in large-scale changes in the upper-ocean thermohaline properties (described below), which eventually might alter the air–sea interaction mechanism. These currents have a large role to play in the variation of salinity patterns in the central EIO. The strong boundary current in the eastern BOB carries low-salinity water equatorward and then joins the RWJ making central EIO fresher. There is another reinforcing effect to this low-salinity tongue formation. The strong equatorward current in the AS bringing salty water toward the equator in the normal years is absent during the anomalous dipole year. The absence of this current also contributes to the low-salinity tongue formation.
In the normal years from November to February, the East India Coastal Current (EICC) is rather strong (Cutler and Swallow 1984) through which the nutrient rich BOB water intrudes into the AS and feeds the west India coastal current (WICC) flowing northward. It is found that during dipole years, the WICC weakens (figure not shown) because of the weakening of the EICC. This weakening of the WICC is likely to have implications on the air–sea interaction process and also on the productivity of the ocean. Prasanna Kumar et al. (2004), in a recent work, investigated the role of advected nutrients from the BOB in altering the biogeochemistry of the western shelf of India.
5. Role of advection and upwelling in determining salinity variations
In the Indian Ocean, Han et al. (2001), Prasad and Ikeda (2002), Shenoi et al. (2004), and Masson et al. (2004) have carried out salinity-related studies. In this section, the mechanisms that explain the observed SSS variations in the EIO are examined. In the RGPCP run carried out in the present study, it has been found that surface water in the southeastern EIO (7°S, 92°E) in 1997 is saltier by 0.2 psu than in 1994, though rainfall at this location is of similar magnitude in both the years. This finding seems to be in agreement with that of Perigaud et al. (2003). To explore the mechanisms responsible for contrasting salinity behavior in the EIO during the two dipole years, different terms involved in the evolution equation of SSS corresponding to the local rate of change (∂S1/∂t), horizontal advection (υ1 · ∇S1), upwelling [w+1 (S2 − S1)/h1], horizontal mixing (κs∇2S1 − κs4∇4S1), and air–sea net water fluxes (E − P) are investigated on a daily basis over the study period (1993–2000). For clarity, the time series of the rate of change of salinity and combined response of oceanic processes is shown at two locations, one at 3°S, 102°E and another at the equator and 65°E (Fig. 13), only for 2 yr (1996–97). The nature of the rate of change of salinity and processes/parameters governing it are the same for the 1994 dipole year also. In the central EIO (bottom panel of Fig. 13), where a low-salinity tongue was found in October 1997 and 1994, three pulses of SSS decrease are observed during September–October 1997. It can be seen that it is the combined effect of advection and upwelling, which explains the salinity decrease during October 1997. It was described in the earlier section that the RWJ is responsible for bringing low-salinity water from EEIO, making this region freshwater. However, the striking inference is the role of upwelling in determining the salinity change. It can be observed from the figure that upwelling, which brings high-saline water from the deep ocean to the surface, was quite strong in the central EIO during this period. Had there not been upwelling, the salinity in this region would have been much lower than observed. Hence it can be said that horizontal advection combined with upwelling and not the freshwater flux is the dominant factor in regulating the variations in SSS in the equatorial region, especially during the latter half of 1994 and 1997. In fact, the GPCP precipitation was not very different from the climatological precipitation in these years. This salinity tongue is caused by the transport of fresher surface waters from the eastern side of the equatorial region via RWJ. The southward current (which is absent in normal years) from the eastern BOB (bringing fresher water) further strengthens the RWJ.
One more location closer to the Sumatra coast (3°S, 102°E), where the easterlies were anomalous during 1994 and 1997, has been chosen for the time series analysis. As can be seen from Fig. 13 (top panel), the salinity increase during the latter half of 1997 (the same holds true for 1994 also) at this location correlates strongly with the upwelling term of the salt equation. At this location, the precipitation was in deficit in both of the years by a similar magnitude, but the salinity increase was much more in 1997 than in the 1994 because of strong Ekman pumping caused by anomalous winds. The interesting observation is a clear lead in the salinity increase at this location over the formation of the low-salinity tongue in the central EIO. This is because of the continuity of fluid flow. Once the high-saline water near the Sumatra coast is upwelled, the displaced freshwater flows westward, causing freshening in the central EIO. A curiosity regarding the deeper-layer changes in the eastern EIO associated with dipole events led to the investigation of the relation between changes in MLD and SSS near the Sumatra coast. There is an apparent shoaling in the MLD during the dipole years toward the eastern equatorial side. The time series of normalized MLD anomaly and normalized SSS anomaly averaged over 10°S–equator and 100°–110°E have also been analyzed (Fig. 14). It is found that the two anomalies (MLD and SSS) are exactly in opposite phase. The magnitudes of the anomalies in MLD and SSS are larger in 1997 than in 1994. In addition to this, one interesting observation is the time of peaking of these two variables. It can be seen that MLD tends to lead SSS by about 2 months. One possible explanation of the observed fact could be given in terms of upwelling induced by surface Ekman divergence under the influence of abnormal easterlies. As soon as the upwelling process starts, the isotherms begin to tilt up eastward, and consequently the MLD starts shoaling. Hence, the signature starts appearing in the upper-layer thickness. Yet it takes some time before strong salinity anomalies could be seen at the ocean surface. Since there are no in situ salinity measurements during 1994 or 1997 at this location, it is difficult to confirm this inference. However, the observation definitely opens up the question of whether the signature of such extreme events like a dipole can first be seen in the subsurface layers. In fact, Meyers (1996) had identified El Niño–Southern Oscillation (ENSO) signals in the subsurface temperature at 26°S along the western Australian coast. Recently, Feng et al. (2003) found that on an interannual time scale, the surface dynamic height in the deep ocean tends to lead the sea level variation at 32°S, 115.4°E.
6. Summary
In recent years, there has been increasing interest in understanding the interannual variability in the Indian Ocean. This variability has been simulated using ocean general circulation models as well as coupled ocean–atmosphere models. However, most of these studies have used climatological winds or winds produced by atmospheric general circulation models for forcing these ocean circulation models, although it has been shown previously that the satellite scatterometer–derived wind is a better alternative for forcing ocean circulation models. This fact encouraged us to make use of the ERS-2 scatterometer–derived wind product for forcing a multilayer Indian Ocean circulation model. It has been found that the model with ERS wind forcing improves the simulation of the 1997 Indian Ocean dipole.
The sensitivity of the model to interannually varying GPCP rainfall has also been studied. Two different runs have been carried out by forcing the model first by climatological rainfall and next by GPCP rainfall. In both cases, ERS-2 winds are used as dynamical forcing. It has been found that the simulated SSS is quite sensitive to interannually varying rainfall. A detailed comparison of the SSS, SST, and MLD simulated by the model has been made with the same variables, measured by a central Arabian Sea buoy (1994–95). It has been found that the model is able to provide a reasonably realistic simulation of SST, SSS, and MLD.
Acknowledgments
The authors are grateful to Dr. J. P. McCreary, Director of the International Pacific Research Center, for kindly providing the Indian Ocean Circulation model used in this study. The authors express their sincere gratitude to Dr. R. A. Weller of Woods Hole Oceanographic Institution for providing the surface mooring data used in the study. They are thankful to the Manager of the MOG Computer Facility for computing support. Useful comments and suggestions made by two anonymous reviewers helped the authors to improve the manuscript. ERS-2 scatterometer data and hydrographic data used in this study have been obtained from the WOCE Data Products Committee 2002, WOCE Global Data, Version 3.0, WOCE International Project Office, WOCE Report 180/02, Southampton, United Kingdom. NCEP reanalyzed fields and GPCP precipitation data have been downloaded from Web sites. Finally, the authors are thankful for the Department of Ocean Development for partial financial support under the SATCORE Project.
REFERENCES
Babu, K. N., R. Sharma, N. Agarwal, V. K. Agarwal, and R. A. Weller, 2004: Study of the mixed layer depth variations within the north Indian Ocean using a 1-D model. J. Geophys. Res., 109 .C08016, doi:10.1029/003JC002024.
Behera, S. K., P. S. Salvekar, and T. Yamagata, 2000: Simulation of interannual SST variability in the tropical Indian Ocean. J. Climate, 13 , 3487–3499.
Cutler, A. N., and J. C. Swallow, 1984: Surface currents of the Indian Ocean (to 25°S, 100°E). Institute of Oceanographic Science Tech. Rep. 187, 42 pp.
Doney, S. C., S. Yeager, G. Danabasoglu, W. G. Large, and J. C. McWilliams, 2003: Modeling global oceanic interannual variability (1958–1997): Simulation design and model-data evaluation. NCAR Tech. Note NCAR/TN-552+STR, 48 pp.
Feng, M., G. Meyers, A. Pearce, and S. Wiffels, 2003: Annual and interannual variations of the Leeuwin Current at 32°S. J. Geophys. Res., 108 .3355, doi:10.1029/2002JC001763.
Goswami, B. N., and D. Sengupta, 2003: A note on the deficiency of NCEP/NCAR reanalysis surface winds over the equatorial Indian Ocean. J. Geophys. Res., 108 .3124, doi:10.1029/2002JC001497.
Grima, N., A. Bentamy, K. Katsaros, Y. Quilfen, P. Delecluse, and C. Levy, 1999: Sensitivity of an oceanic general circulation model forced by satellite wind stress field. J. Geophys. Res., 104 , 7967–7989.
Hackert, E. C., A. J. Busalacchi, and R. Murtugudde, 2001: A wind-comparison study using an ocean general circulation model for the 1997–1998 El Niño. J. Geophys. Res., 106 , 2345–2363.
Han, W., and J. P. McCreary, 2001: Modeling salinity distributions in the Indian Ocean. J. Geophys. Res., 106 , 859–877.
Han, W., J. P. McCreary, D. L. T. Anderson, and A. J. Mariano, 1999: Dynamics of the eastern surface jets in the equatorial Indian Ocean. J. Phys. Oceanogr., 29 , 2191–2209.
Han, W., J. P. McCreary, and K. E. Kohler, 2001: Influence of precipitation minus evaporation and Bay of Bengal rivers on dynamics, thermodynamics, and mixed layer physics in the upper Indian Ocean. J. Geophys. Res., 106 , 6895–6916.
Huffman, G. J., R. F. Adler, M. M. Morrissey, S. Curtis, R. Joyce, B. McGavock, and J. Sisskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observations. J. Hydrometeor., 2 , 35–50.
Kerr, Y. H., J. Font, P. Waldteufel, J. P. Wigneron, and M. Berger, 2003: The Soil Moisture and Ocean Salinity mission. Geophys. Res. Abstr., 5 , 14583.
Kraus, E. B., and J. S. Turner, 1967: A one-dimensional model of the seasonal thermocline. II: The general theory and its consequences. Tellus, 19 , 98–106.
Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial variability in gauge-corrected global precipitation. Int. J. Climatol., 10 , 111–127.
Legler, D. M., I. M. Navon, and J. J. O’Brien, 1989: Objective analysis of pseudo-stress over the Indian Ocean using a direct-minimization approach. Mon. Wea. Rev., 117 , 709–720.
Levitus, S., 1982: Climatological Atlas of the World Ocean. NOAA Prof. Paper 13, 173 pp. and 17 microfiche.
Masson, S., J-P. Boulanger, C. Menkes, P. Delecluse, and T. Yamagata, 2004: Impact of salinity on the 1997 Indian Ocean dipole event in a numerical experiment. J. Geophys. Res., 109 .C02002, doi:10.1029/2003JC001807.
McCreary, J. P., P. K. Kundu, and R. L. Molinari, 1993: A numerical investigation of dynamics, thermodynamics and mixed-layer processes in the Indian Ocean. Progress in Oceanography, Vol. 33, Pergamon Press, 181–244.
McCreary, J. P., K. E. Kohler, R. R. Hood, S. Smith, J. Kindle, A. S. Fischer, and R. A. Weller, 2001: Influences of diurnal and intraseasonal forcing on mixed-layer and biological variability in the central Arabian Sea. J. Geophys. Res., 106 , 7139–7155.
Meyers, G., 1996: Variation of Indonesian throughflow and the El Niño–Southern Oscillation. J. Geophys. Res., 101 , 12225–12263.
Murtugudde, R., and A. J. Busalacchi, 1999: Interannual variability of the dynamics of the tropical Indian Ocean. J. Climate, 12 , 2300–2326.
Perigaud, C., J. P. McCreary, and K. Q. Zhang, 2003: Impact of interannual rainfall anomalies on Indian Ocean salinity and temperature variability. J. Geophys. Res., 108 .3319, doi:10.1029/2002JC001699.
Prasad, T. G., and M. Ikeda, 2002: A numerical study of the seasonal variability of Arabian Sea high-salinity water. J. Geophys. Res., 107 .3197, doi:10.1029/2001JC001139.
Prasanna Kumar, S., and Coauthors, 2004: Intrusion of the Bay of Bengal water into the Arabian Sea during winter monsoon and associated chemical and biological response. Geophys. Res. Lett., 31 .L15304, doi:10.1029/2004GL020247.
Rao, R. R., R. L. Molinari, and J. F. Festa, 1989: Evolution of the climatological near-surface thermal structure of the tropical Indian Ocean. I. Description of mean monthly mixed layer depth and sea surface temperature, surface current, and surface meteorological fields. J. Geophys. Res., 94 , 10801–10815.
Rao, R. R., R. L. Molinari, and J. F. Festa, 1991: Surface meteorological and near surface oceanographic atlas of the tropical Indian Ocean. NOAA Tech. Memo. ERLO AOML-69, 59 pp.
Saji, N. H., B. N. Goswami, P. N. Vinaychandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401 , 360–363.
Schott, F. A., and J. P. McCreary, 2001: The monsoon circulation of the Indian Ocean. Progress in Oceanography, Vol. 51, Pergamon Press, 1–123.
Shenoi, S. S. C., A. D. Gouveia, S. R. Shetye, and L. V. G. Rao, 1992: Satellite observations of the northeast monsoon coastal current. Proc. Conf. for Pacific Ocean Environments and Probing, Vol. 2, Okinawa, Japan, PORSEC, 796–801.
Shenoi, S. S. C., D. Shankar, and S. R. Shetye, 2004: Remote forcing annihilates barrier layer in southeastern Arabian Sea. Geophys. Res. Lett., 31 .L05307, doi:10.1029/2003GL019270.
Shetye, S. R., and S. S. C. Shenoi, 1988: The seasonal cycle of surface circulation in the coastal north Indian Ocean. Proc. Indian Acad. Sci., Earth Planet. Sci., 97 , 53–62.
Vinaychandran, P. N., N. H. Saji, and T. Yamagata, 1999: Response of the equatorial Indian Ocean to an unusual wind event during 1994. Geophys. Res. Lett., 26 , 1613–1616.
Webster, P. J., 1994: The role of hydrological processes in ocean–atmosphere interaction. Rev. Geophys., 32 , 427–476.
Webster, P. J., A. M. Moore, J. P. Loschnigg, and R. R. Leben, 1999: Coupled ocean–atmosphere dynamics in the Indian Ocean during 1997–98. Nature, 401 , 356–360.
Weller, R. A., M. F. Baumgartner, S. A. Josey, A. S. Fischer, and J. C. Kindle, 1998: Atmospheric forcing in the Arabian Sea during 1994–1995: Observation and comparisons with climatology and models. Deep-Sea Res., 45 , 1961–1999.
Wyrtki, K., 1973: An equatorial jet in the Indian Ocean. Science, 182 , 254–262.
The model domain with the locations of various in situ measurements shown.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Difference between AVHRR and model-simulated SST (°C) for October 1997. The models are forced by (a) NCEP and (b) ERS-2 winds.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Mean difference between AVHRR and model-simulated SST (°C) from (a) NCEP and (b) ERS-2 forcing.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Std devs of SST (°C) from (a) AVHRR, (b) the model with NCEP forcing, and (c) the model with ERS-2 forcing.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Mean difference between NCEP and ERS-2 wind speed (m s−1).
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Mean seasonal cycle of (left) zonal and (right) meridional wind speed (m s−1). The solid line is for ERS-2 and the dashed line is for NCEP. Winds are averaged in (a) 1°S–1°N, 60°–90°E and (b) 8°–15°N, 54°–60°E boxes.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Same as in Fig. 6, but for interannual anomalies.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
(a) SSS anomaly in October 1997 in (left) RClim and (right) RGPCP. (b) The corresponding rain-rate distributions (mm day−1) in (left) climatology and (right) GPCP.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
(a) Comparison of monthly averaged sea surface parameters simulated by the model forced by two different rainfall products with buoy observations in the central Arabian Sea and (b) scatterplot of simulated and observed (JGOFS) oceanic variables for (left) SST, (middle) SSS, and (right) MLD.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Spatial distribution of (a) mean SSS and (b) std dev of SSS simulated in the RGPCP run.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Longitude–time plot of model-simulated sea surface currents (cm s−1) averaged over 10°S–equator.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Spatial plots of the SSS anomaly overlaid with sea surface currents for the month of October for four different years (1994, 1996, 1997, and 1998).
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Time series of the SSS changes (10−7 psu s−1) at two different locations in the IO: (top) at 3°S, 102°E showing local rate of change (solid line) and change due to upwelling (dashed line) and (bottom) at equator, 65°E showing local rate of change (solid line) and change due to advection (dashed line) and upwelling (dotted line).
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Time series of the normalized anomalies of SSS (solid line) and MLD (dashed line) averaged over 10°S–equator and 100°–110°E.
Citation: Journal of Climate 20, 5; 10.1175/JCLI4032.1
Types of experiments carried out using different forcing parameters.
