Observed strengthening of interbasin exchange via the Indonesian seas due to rainfall intensification

Abstract

A proxy of the Indonesian Throughflow (ITF) transport, developed using in situ hydrographic measurements along with assimilations, shows a significant strengthening trend during the past decade. This trend is due to a freshening and subsequent increase in the halosteric component of the ITF transport associated with enhanced rainfall over the Maritime Continent over the same period. The strengthening of the ITF transport leads to a significant change in heat and freshwater exchange between the Pacific and Indian Oceans and contributes to the warming and freshening of the eastern Indian Ocean. The combined effect of the ITF transport of mass and freshwater along with tropical rainfall plays a very important role in the climate system.

Key Points

• Salinity effect dominates an increasing trend of the ITF transport during the past decade

• Freshening and strengthening of the ITF transport are related to enhanced rainfall

• Strengthening of the ITF transport contributes to warming and freshening in the Indian Ocean

1. Introduction

The Indonesian Throughflow (ITF) from the tropical Pacific into the Indian Ocean via the Indonesian seas results in an interocean transport of about 15 Sverdrup (Sv, 1 Sv ≡ 10⁶ m³ s⁻¹) and occurs primarily as a strong core within the thermocline [Gordon et al., 2008, 2012; Sprintall et al., 2009]. The ITF transport is important to the global thermohaline system [Gordon, 1986] and is thought to play a part in determining the spatial structure of heat and salinity distribution in the Pacific and Indian Oceans [Sprintall et al., 2014; Hu et al., 2015; Lee et al., 2015]. Water masses from the Pacific Ocean are strongly mixed within the Indonesian seas [Ffield and Gordon, 1996; Koch‐Larrouy et al., 2010] and so the ITF waters are apparent within the Indian Ocean thermocline as a relatively cool, low‐salinity core [Ffield and Gordon, 1996; Gordon and Fine, 1996; Song et al., 2004; Sprintall et al., 2014]. Fluctuations of the volume, heat, and freshwater transports by the ITF play an essential role in regulating the atmospheric and oceanic circulations and in the redistribution of the heat and water masses in the Indo‐Pacific region on various time scales [Godfrey, 1996; Schneider, 1998; Lee et al., 2002; Talley and Sprintall, 2005; Sprintall et al., 2014; Feng et al., 2015; Lee et al., 2015].

An early hypothesis by Wyrtki [1987] posited that on annual and lower frequencies the ITF is governed by the pressure gradient across the Indonesian archipelago. This enabled the use of sea level or steric height difference between the two ocean basins to produce prediction indices of the ITF transport [Andersson and Stigebrandt, 2005; Sprintall and Révelard, 2014; Susanto and Song, 2015]. These proxy multiyear time series were used to explore how the ITF variability responds to the wind and buoyancy forcing that modulates the Indo‐Pacific pressure gradient. However, the pressure gradient is determined by both the thermal (temperature) and haline (salinity) variations. Hu and Sprintall [2016] found that about (36 ± 7)% of the total interannual variability of ITF transport can be attributed to the salinity effect. The Indonesian region receives heavy rainfall and has experienced a decadal increase in rainfall related to the negative phase of the Interdecadal Pacific Oscillation (IPO) [Dong and Dai, 2015], as well as a longer‐term trend under the warming climate scenarios of the “wet‐gets‐wetter” [Chou et al., 2009] and “warmer‐gets‐wetter” [Tan et al., 2015]. A key question remains is how the thermosteric and halosteric ITF transports will change in response to the warming/cooling and freshening/salinifying within the context of the known significant interdecadal variability and global warming trends [Chou et al., 2009; Du et al., 2015; Feng et al., 2015; Tan et al., 2015]. Some recent numerical studies suggest that the future ITF (2050–2100) is projected to decrease due to wind and deep circulation changes [Hu et al., 2015; Sen Gupta et al., 2016], but the role of buoyancy forcing changes has been largely unexplored.

2. Data and Method

Andersson and Stigebrandt [2005] and Hu and Sprintall [2016] showed that the ITF transport corresponds to the zonal geostrophic transports estimated between the Indian Ocean region where the ITF exits the Indonesian seas (9°S–15°S, 100°E–120°E) and the eastern equatorial Indian Ocean (6°N–6°S, 80°E–100°E) that is outside of the direct influence of the ITF stream and so represents the “mean” background state of the Indian Ocean. The proxy ITF transport relationship can thus be represented by

$$\mathrm{I}\mathrm{T}\mathrm{F}=\frac{g{H}^2\mathrm{\Delta }\rho }{2\left(f_{{\lambda }_N}-f_{{\lambda }_S}\right){\rho }_0}\text{,}$$ (1)

where g is gravity acceleration, $f_{{\lambda }_N}$ and $f_{{\lambda }_S}$ are the Coriolis parameters at the northern boundary latitude λ_N (10°S) and the southern boundary latitude λ_S (16°S), H is 1200 m (the effective sill depth of Timor Strait), ρ ₀ is the reference density at H, and Δ_ρ is the difference in density between the two regions. Since the sea water density gradient Δ_ρ is a function of ocean temperature (T) and salinity (S), the ITF transport (equation (1)) can also be decomposed as a function of T and S: ITF = ITF(T, S) [Hu and Sprintall, 2016]. Thus, the thermal and halosteric ITF transport are respectively defined as follows: ${\mathrm{I}\mathrm{T}\mathrm{F}}_T=\mathrm{I}\mathrm{T}\mathrm{F}\left(T,\overline{S}\right)$ and ${\mathrm{I}\mathrm{T}\mathrm{F}}_S=\mathrm{I}\mathrm{T}\mathrm{F}\left(\overline{T},S\right)$, where $\overline{T}$ and $\overline{S}$ are climatological temperature and salinity. The sign of the transport is such that positive is transport into the Indian Ocean (i.e., the ITF). Further details of the method can be found in Andersson and Stigebrandt [2005] and Hu and Sprintall [2016]. In particular, Hu and Sprintall [2016] compared the proxy ITF transport to the direct velocity measurements made as part of the INSTANT project over the 2004–2006 period [Sprintall et al., 2009] and found excellent agreement in the mean, standard deviation, and time series variability. Comparisons with the recent geostrophic transport (relative to 400 and 700 m) time series derived from eXpendable Bathy Thermograph (XBT) measurements in the outflow region of the ITF [Liu et al., 2015] also suggest reasonable agreement in variability (not shown). These favorable comparisons provide confidence that the proxy ITF transport does a satisfactory job of representing the real ITF as it enters the Indian Ocean.

Temperature and salinity data used to construct the proxy ITF transports are from a number of sources: (a) the Roemmich‐Gilson Argo Climatology monthly gridded data (RG Argo Climatology) [Roemmich and Gilson, 2009] are produced using a weighted least squares fit to Argo profiles over the period 2004–2014 with a monthly resolution of 1° × 1° horizontally; (b) the EN4 data set (version 4.1.1) are monthly 1° × 1° horizontal objectively mapped global fields of temperature and salinity from the Met Office Hadley Centre [Good et al., 2013]; (c) the European Centre for Medium‐Range Weather Forecasts Ocean Analysis/Reanalysis System 3 (ECMWF‐ORA S3) has a horizontal resolution of 1° × 1° spanning 53 years (1959–2011) [Balmaseda et al., 2008]; and (d) the Ensemble Coupled Data Assimilation v3.1 from the Ocean Data Assimilation Experiment at the NOAA Geophysical Fluid Dynamics Laboratory (GFDL‐ECDA) [Chang et al., 2013] are produced from a fully coupled climate model (1° zonal resolution with meridional resolution varying from 1° to 0.3344°) and available from 1961 to 2012. In order to verify the method and these data sets, we also use direct measurements of the ITF transport from the INSTANT mooring array [e.g., Sprintall et al., 2009].

Since 1975, roughly 1000–2000 temperature profile measurements were obtained each year in the eastern Indian Ocean (80°E–120°E, 16°S–6°N), mainly due to XBT measurements (Figure 1a). By contrast, salinity measurements were largely absent before the early 1970s and remained relatively rare until the Argo era began in 2004 (Figures 1a and 1b). To account for the discrepancies in the temporal data distributions, linear trends in the ITF transport are calculated over three periods: (1) the whole period in common for the three data sets 1970–2010; (2) the period since 1990 when changes in the wind fields and sea surface height patterns were found in the Pacific Ocean [e.g., Merrifield, 2011; Hu and Hu, 2012, 2014]; and (3) the period 2002–2010 during which many more salinity observations are available. The linear trend of transports and various variables is estimated applying a one‐dimensional linear regression model in the least squares sense. The significance of the linear trend is examined by a modified Mann‐Kendall test [Hamed and Rao, 1998], and all reported significances are at 95% or better.

Figure 1.

(a) Annual number of temperature and salinity profiles in the eastern Indian Ocean (80°E–120°E, 16°S–6°N) and (b) distribution of salinity profiles over 1970–2001 (left) and 2002–2010 (right) with red boxes indicating the eastern Indian Ocean. Salinity profiles in 2001 and 2004 are highlighted by the green dots. The time series of ITF′ (c) composited from the EN4, ECMWF ORA, and GFDL‐ECDA data sets (shaded area indicates the standard deviation) and (d) derived from the RG Argo data set. White and yellow square lines in Figures 1c, 1e, and 1g and black square line in Figures 1d, 1f, and 1h show the linear trends over 1970–2010, 1990–2010, and 2004–2014, respectively. Black dot lines indicate the linear trend during 2002–2010. The time period since about 2004 is highlighted by the boxes. The time series of the thermostericcomponent ITF _T′ (Figures 1e and 1f) and halostericcomponent ITF_S′ (Figures 1g and 1h), with details as in panels above.

Monthly precipitation (P) and evaporation (E) are, respectively, from the Global Precipitation Climatology Project (1979–2015) [Adler et al., 2003] and the Objectively Analyzed air‐sea Fluxes (1985–2012) [Yu and Weller, 2007]. The trend and interdecadal variability of the rainfall and surface freshwater input is examined using “evaporation minus precipitation” (E − P). A tripole index for the IPO is defined as the difference of the average sea surface temperature anomaly between the central equatorial Pacific (10°S–10°N, 170°E–90°W) and the northwest (25°N–45°N, 140°E–145°W) and southwest (50°S–15°S, 150°E–160°W) Pacific, and calculated using the NOAA ERSST V3b SST data set [Henley et al., 2015].

3. Results

The total proxy ITF transport anomaly (ITF′), as well as the separated thermosteric component (ITF_T′) and halosteric components (ITF_S′), is calculated using the composited EN4, ECMWF ORA, and GFDL‐ECDA time series (Figures 1c, 1e, and 1g) and the more recent RG Argo climatology (Figures 1d, 1f, and 1h). During the period between 1970 and 2010, the ITF_T′ has a significant decreasing trend of −2.2 Sv decade⁻¹, while the ITF_S′ shows a significant increasing trend of 1.8 Sv decade⁻¹. Together, this leads to a significant slightly decreasing trend of −0.4 Sv decade⁻¹ in the total ITF′. During 1990–2010, although a slight increasing trend of the ITF_T′ is not significant, the trend in the ITF_S′ increases to 3.4 Sv decade⁻¹, so as a result, there is a significant increasing trend of 3.7 Sv decade⁻¹ in the ITF′. During the final period between 2002 and 2010, the increasing trend of the ITF_T′ rises to 0.9 Sv decade⁻¹ although it is still not significant, while the significant trend in the ITF_S′ increases to 4.6 Sv decade⁻¹. Subsequently, during 2002–2010 the ITF′ shows a significant increasing trend of 5.8 Sv decade⁻¹. Using the gridded RG Argo Climatology over the period 2004–2014, we find a comparable significant increasing trend of about 5.9 Sv decade⁻¹ (Figure 1d), composed primarily of a significant increasing trend of 4.1 Sv decade⁻¹ in the ITF_S′ (Figure 1h) since the increasing trend of 1.8 Sv decade⁻¹in the ITF_T′ (Figure 1f) is not significant. The results indicate that the tendency in the ITF′ is primarily induced by the increase of the halosteric component.

The increase of volume transport results in a sharp strengthening of the heat and salinity transports by the ITF (Figure 2) defined as follows: Q _T = ITF ⋅ ΔT and Q _S = ITF ⋅ ΔS, where ΔT is temperature difference, and ΔS is salinity difference. To investigate the influence of the strengthened ITF transport, here we calculate the temperature and salinity transports in two schemes (Figure 2a). Scheme‐A uses temperature and salinity differences between the eastern Indian Ocean (EIO, 80.5°E–115.5°E, 25.5°S–0.5°S) and western Pacific Ocean (WPO, 125.5°E–150.5°E, 2.5°S–15.5°N), and Scheme‐B uses temperature and salinity differences between the EIO region and the outflow region (115.5°E–125.5°E, 17.5°S–4.5°S). Thus, Scheme‐A aims to examine the heat and freshwater redistribution between the eastern Indian and western Pacific Oceans, while Scheme‐B is to estimate the influence on the heat and freshwater change within the Indian Ocean. In Scheme‐A, the temperature transport by the ITF derived from the RG Argo Climatology during 2004–2014 increased by 47% relative to the decadal mean of 45.7°C Sv (Figure 2b), while the salinity transport by the ITF increased by 75% relative to the decadal mean of −3.6 practical salinity scale (PSS‐78) Sv (Figure 2c). Given that the temperature difference in Scheme‐A shows no significant trend (Figure 2d), the sharply intensified heat being transferred from the Pacific Ocean to the Indian Ocean must be mainly caused by the increase in the ITF volume transport. In Scheme‐B the temperature transport has a small but significant increasing trend of 5.5°C Sv decade⁻¹ (Figure 2b), and the temperature difference increases at a rate of about 0.1°C decade⁻¹ (Figure 2d). This suggests a significant heat input into the Indian Ocean by the ITF. The salinity difference and freshwater transport in Scheme‐A have both decreased during the past decade (Figures 2c and 2e), suggesting that the ITF has transferred less freshwater from the Pacific Ocean to the Indian Ocean. However, the small trend in observed in salinity differences in Scheme‐B suggests that this large salinity difference between the eastern Indian Ocean and western Pacific Ocean was sharply diminished within the Indonesian seas (Figure 2e). While the salinity difference in Scheme‐B has decreased by an insignificant rate of about −0.01 PSS‐78 decade⁻¹ (Figure 2e), the salinity transport from the Indonesian seas shows a significant trend about −2.8 PSS‐78 Sv decade⁻¹ (Figure 2d) due to the enhancement of the ITF. As a result, the ITF trend gives rise to a strong freshening and warming effect on the eastern Indian Ocean.

Figure 2.

Transports and differences (ΔT and ΔS) of temperature and salinity between the EIO region and WPO region (red) and the outflow region (blue) during the Argo era with linear trends (dashed lines) under the two schemes. (a) Design of the two schemes and their (b) temperature transports, (c) salinity transports, (d) temperature differences, and (e) salinity differences. The volume transport of the ITF is derived from RG Argo temperature and salinity data. The differences are that of volume‐weighted mean temperature/salinity in the upper 200 layer averaged over the WPO region or outflow region minus that in the EIO region. The temperature and salinity transports are defined as the volume transport of the ITF multiplied by the corresponding temperature and salinity differences. Positive temperature transport denotes heat transfered from the WPO or Indonesian seas to the Indian Ocean, and negative salinity transport indicates freshwater transfered from the WPO or Indonesian seas into the Indian Ocean. All the time series are low passed by a 13 month running mean filter.

The above results are consistent with the horizontal and vertical patterns of the linear trend in temperature and salinity in the broader tropical Indo‐Pacific Oceans (Figures S1 and S2 in the supporting information). During 2004–2014, the eastern Indian and the western Pacific Ocean shows a significant warming in the upper 200 dbar ocean. Salinity trends are opposite in each basin on either side of the Indonesian seas with a freshening in the eastern Indian Ocean and a salinifying in the western Pacific Ocean (Figure S1). This implies that the Indo‐Pacific gradient induced by the thermal difference between the two oceans is small compared to that induced by the halosteric difference. The zonal sections of trends of salinity, temperature, and density along 12.5°S show that significant warming and freshening dominate the upper layer in the eastern Indian Ocean, leading to a decrease of the density in the ITF and increase of the halosteric component of the ITF transport (Figure S2). Other studies have suggested that the ITF is an important contributor to the recent increase of the observed Leeuwin Current transport [Feng et al., 2015] as well as the recent halosteric sea level trend in the southeast Indian Ocean [Llovel and Lee, 2015] likely brought about by salinity changes associated with the advection of ITF surface freshwater fluxes from the Indonesian Seas [Phillips et al., 2005].

The decadal trend of the ITF is also in agreement with the variations of the pertinent dynamical forcing fields within the tropical Indo‐Pacific Oceans (Figure 3). Over the period 2002–2010, a significant increasing trend is clear in the sea level of the western Pacific Ocean and Indonesian seas, but the sea level anomaly trend in the tropical Indian Ocean is quite small (Figure 3a). This suggests that the gradient between the two basins has been enhanced, consistent with the increasing trend of the ITF transport.

Figure 3.

Linear trend of (a) sea level (cm decade⁻¹), (b) wind stress (arrow) and Ekman pumping velocity (color; 10⁻⁵ m s⁻¹ decade⁻¹), (c) E‐P (mm d⁻¹ decade⁻¹), and (d) precipitation (mm d⁻¹ decade⁻¹) over 2002–2010. Areas of significant trends at confidence levels higher than 95% are stippled.

However, the tropical Indian Ocean and the region where the ITF enters the eastern Indian Ocean (Figure 3b), which are used to derive the ITF transport, show similar wind forcing and Ekman pumping trends with the same magnitude over 2002–2010 and hence demonstrate only a small change in the thermosteric gradient. The spatial structure of wind forcing and Ekman pumping velocity trend in the eastern Indian Ocean explains why the linear trend of ITF_T′ is close to zero. As shown in Figures S1b and S2a, the freshening signal that accounts for the increasing halosteric height in the buoyancy pool and increase of ITF_S′ arises directly from within the Indonesian seas. Finally, the linear trends in E‐P and precipitation clearly show the strong enhancement of precipitation and freshwater input over the Indonesian seas during the period 2002–2010 (Figures 3c and 3d). By contrast, there is a significant decrease of precipitation and freshwater input over the eastern Indian and central western Pacific Oceans. This result largely explains the decreasing trend in the ITF salinity and the increasing trend of the halosteric component of the ITF transport. Some of the freshening of the ITF may also be due to advection of freshwater. For example, the Karimata Strait transport is known to regulate the seasonal variability of the Makassar Strait transport via freshwater advection from the South China Sea to the Java Sea [Gordon et al., 2003; Fang et al., 2010; Susanto et al., 2013]. However, the lack of observations within the Indonesian archipelago makes this contribution difficult to assess on a decadal time scale.

The role of freshwater input is further investigated by comparing the observed salinity variation with the salinity variation inferred using a mean temperature‐salinity (T‐S) relationship, which might be expected to better represent the thermocline response to the wind‐driven variability over the past decade. The mean T‐S relationship during 2004–2014 from the RG‐Argo data is used to derive the salinity anomaly induced by thermocline fluctuations, and the linear trends of the derived salinity anomaly and observed Argo salinity anomaly are calculated (Figure S3). The comparison clearly indicates that the derived salinity trend due to thermocline fluctuation is significantly less than the observed salinity trend.

As suggested by previous studies [Gordon, 1986; Sprintall et al., 2014], the ITF plays a pivotal role in the global thermohaline circulation. Under the forcing of surface wind that has enhanced during the past decade, oceanic heat and freshwater have converged in the Indo‐Pacific Ocean (Figures 3 and S1a). The intensified heat and freshwater transport by the ITF acts an important regulator that may accelerate the global thermohaline circulation and redistribute the global pattern of heat and freshwater (Figure 4).

Figure 4.

Schematic representation of the mechanism responsible for the enhancement of the ITF during the past decade. The linear trends of (a) precipitation in the tropical Indo‐Pacific Ocean. Dashed black arrow lines denote the tendency of ocean surface zonal wind stress over the three ocean basins. The linear trends in (b) temperature and (c) salinity in the western Pacific (along 4.5°N) and eastern Indian (along 12.5°S) Oceans. Contour lines show the mean temperature (Figure 4b) and salinity (Figure 4c) over 2004–2014 from RG Argo data set. The direction of the ITF and the enhancement of the ITF are shown by the dashed blue arrow.

4. Discussion and Conclusions

In summary, the enhancement of the rainfall over the Indonesian seas freshens the Indonesian seas and so strengthens the volume transport of the ITF. The intensified ITF transfers more heat from the Pacific Ocean and the freshening of the ITF associated with increased rainfall within the Indonesian seas leads to a sharp input of freshwater into the Indian Ocean. Thus, the relationship between rainfall and the ITF transport might act as a key process in determining Indo‐Pacific contribution to the global pattern of the heat and freshwater distribution.

Global climate and the ocean experience strong natural internal interdecadal variability as well as long‐term trends induced by external forcing, including increasing concentrations of greenhouse gases and other anthropogenic activities [Dong and Dai, 2015; Chou et al., 2009; Tan et al., 2015]. The IPO was in a positive phase during 2004–2007 but switched to a negative phase after 2008 resulting in a significant decreasing trend (Figure S4). The IPO index leads the ITF_S′ by about 2 years with a significant correlation coefficient of about −0.8. Tropical precipitation and the wind field have a known prominent association with the IPO [e.g., Dong and Dai, 2015]. Composites formed by differencing the positive from the negative phases of the IPO during 1979–2010 show a remarkable increase of E‐P and a decrease in precipitation over the Indonesian seas and the far western Pacific Ocean (Figure S5). Westward wind stress anomaly and upward Ekman pumping anomaly occur in the western Pacific Ocean, while an eastward wind stress anomaly and downward Ekman pumping anomaly control the tropical eastern Indian Ocean relative to the negative IPO phase (Figure S5). The above processes during the negative IPO phase act to freshen the eastern Indian Ocean but salinify the central western Pacific Ocean. This cancels out the zonal pressure gradient over the tropical Indo‐Pacific Ocean and leads to a decrease in the ITF transport. Thus, a switch in the IPO phase from positive to negative is expected to lead to an increase of the ITF transport.

The long‐term trend is probably also important. For example, during the period from 1970 to 2010, the ITF_S′ exhibits a clear trend that is greater than the interdecadal variability (Figure 1g), but the IPO index shows no significant trend during the same period. In contrast, the precipitation during the past four decades shows a significant increasing trend in the Indonesian seas and far western Pacific Ocean but a decreasing trend in the central western Pacific Ocean, which is likely related to the global warming due to external forcing (Figure S6). The long‐term trend in the precipitation field contributes to the increase of the ITF transport. Therefore, we suggest that both the natural interdecadal variability and external forcing are likely responsible for the intensification of the ITF transport during the past decade.

Supporting information

Supporting Information S1

Figure S1

Figure S2

Figure S3

Figure S4

Figure S5

Figure S6

Hu, S. , and Sprintall J. (2017), Observed strengthening of interbasin exchange via the Indonesian seas due to rainfall intensification, Geophys. Res. Lett., 44, 1448–1456, doi:10.1002/2016GL072494.