Satellite observations of sea surface height (SSH) and wind stress for the period of 1993–2006 reveal a near-coherent large-scale decadal variability in much of the Indo-Pacific region with a phase change at the turn of the 20th century. Trade wind variations in the tropical Pacific and South Indian Ocean are anti-correlated with each other as are SSH differences across these two basins, implying anti-correlated variation of the subtropical cells in the two oceans. Decadal changes in large-scale SSH in the extra-tropics are mostly associated with well-defined patterns of wind stress curl indicating a near-coherent decadal variation in the strength of subtropical and subpolar gyres. Together, these variations reflect a linkage in the circulation of the Pacific and Indian Oceans via atmospheric and oceanic bridges. The phase change in the tropical Pacific tends to occur earlier than elsewhere, suggesting a potential role of the tropical Pacific in regulating decadal variability of the entire region.
 The meridional and horizontal ocean circulations in the Indo-Pacific region play important roles in regulating climate variability on decadal and longer time scales by redistributing heat geographically. McPhaden and Zhang  (hereinafter referred to as MZ02) found that the shallow meridional overturning circulation connecting the tropical and subtropical Pacific, the subtropical cell or STC [McCreary and Lu, 1994], experienced a multi-decadal slowdown. McPhaden and Zhang  (hereinafter referred to as MZ04) reported a rebound of its strength during 1998–2003. Zhang and McPhaden  showed that decadal variability of the STC has a larger magnitude than the multi-decadal decreasing trend on which it is superimposed. Lee  identified a substantial weakening of the Indian-Ocean STC [Schott, 2005] during the 1990s. Qiu  noted the linear trends in the strength of horizontal gyres in the North Pacific during 1992–2000. Roemmich et al.  discussed the decadal spinup of the subtropical gyre in the South Pacific during the 1990s.
 A systematic study has not been conducted to examine the potential linkages between these decadal variations in ocean circulation. Satellite altimeters and scatterometers have provided continuous near global measurements of sea surface height (SSH) and wind stress for about 15 years, offering an unprecedented opportunity to study this issue. In the present study, we use these measurements to investigate decadal variability of the Pacific and Indian-Ocean STCs and mid- to high-latitude horizontal gyres. We identify a near-coherent decadal variation in these circulations with a phase change during 1999–2000 and highlight associated oceanic and atmospheric processes that establish the linkages between the Pacific and Indian-Ocean STCs.
 The data used for this study include altimeter measurements of SSH and scatterometer measurements of wind stress. The SSH data, processed by JPL, are obtained from the joint US-French missions of TOPEX/POSEIDON (October 1992 to October 2005) and JASON-1 (January 2001 to present). For the overlapping period between TOPEX/POSEIDON and JASON-1, the data from the latter are used (bias correction was applied to JASON-1). Wind stress data are obtained from European Research Satellites (ERS)-1 (August 1991 to June 1996), ERS-2 (March 1995 to January 2001), and QuikSCAT (July 1999 to present) scatterometers. During the overlapping period between ERS-1 and ERS-2, data from the latter are used (the two are very close to each other). The ERS and QuikSCAT data are downloaded from http://cersat.ifremer.fr where documentation is available. Because of the mean bias between ERS-2 and QuikSCAT data, the tendency of ERS data up to 2000 and QuikSCAT data after 2000 are examined separately. NCEP/NCAR reanalysis wind [Kalnay et al., 1996] is also analyzed.
3. Oceanic and Atmospheric Linkage Between Pacific and Indian-Ocean STCs
 The Pacific STCs are associated with the divergence of warmer Ekman flow out of the tropics forced by easterly trade winds (upper branch) and the convergence of colder pycnocline waters into the tropics (lower branch). Much of the geostrophic flow in the pycnocline lies below the Ekman layer, facilitating the separation of the upper and lower branches. The one-sided STC in the South Indian Ocean is associated with southward Ekman transport in response to easterly trade winds centered near 10°S (upper branch) and northward transport of pycnocline water (lower branch). The variability of the lower branches can be inferred from SSH difference across the basin (i.e., SSH at the east coast minus that at the west coast) because it is proportional to the net meridional pycnocline transport by the geostrophic relation [Lee and Fukumori, 2003; MZ04]. The strength of the upper branch can be assessed using zonal wind stress.
 Large-scale SSH in much of the Indo-Pacific region is found to exhibit decadal variations that are characterized by a change of tendency at the turn of the 20th century (Figures 1a and 1c, respectively). The SSH trend in many regions in the Indo-Pacific domain are large and have opposite signs between 1993–2000 and 2000–2006. The largest trends in the Atlantic Ocean on the other hand are weaker than the largest trends in the Pacific and Indian Ocean and do not show a coherent change in sign for these time periods. We therefore focus our discussion on the Indo-Pacific domain.
 A major feature of the decadal change is the rising SSH in the western tropical Pacific and southeast Indian Ocean for the 1993–2000 period and the subsiding SSH for the 2000–2006 period. SSH along the coasts in the eastern tropical Pacific and southwest Indian Ocean show relatively small decadal changes, so that variability in SSH differences across the tropical Pacific and South Indian Oceans is determined mainly by SSH in the western tropical Pacific and southeast Indian Ocean, respectively. The increase of SSH in the western tropical Pacific during 1993–2000 indicates an enhancement of equatorward convergence of pycnocline water, i.e., a strengthening of the lower branch of Pacific STC. The increase of SSH in the southeast Indian Ocean during 1993–2000 suggests a reduction of northward pycnocline flow of the Indian-Ocean STC. The opposite changes of SSH for the 2000–2006 period suggest that the lower branch of the Pacific (Indian-Ocean) STC has weakened (strengthened) since the turn of the century. In the Pacific Ocean, these changes are associated with variations in central Pacific SST (Figure 2a), which are indicative of STC related changes in upwelling intensity and possible coupled ocean-atmosphere feedbacks on decadal time scales (MZ02, MZ04).
 In the northern (southern) hemisphere, a positive trend in wind stress curl would cause a negative (positive) trend of SSH anomaly. The rising SSH in the western tropical Pacific off the equator during 1993–2000 are associated with the negative (positive) trend of off-equatorial wind stress curl north (south) of the equator (see the two arrows pointing from Figures 1a to 1b in the western tropical Pacific). These patterns of wind stress curl are centered to the east of the SSH patterns, reflecting the role of westward-propagating baroclinic Rossby waves in adjusting the ocean to wind forcing. A decadal strengthening of the easterly trade winds (Figure 3a) also contributes to the increase of SSH in the west during 1993–2000 (Figure 1a). Weakening of the Pacific trade winds after 2000 (Figure 3b) is likewise consistent with SSH and SST changes in the tropical Pacific during 2000–2006 (Figures 1c and 2a).
 The rising SSH in the southeast Indian Ocean during 1993–2000 (box d in Figure 1a) is clearly not caused by wind stress curl trends in that region because the latter are negative (Figure 1b). SSH in this region shows a relatively coherent interannual-decadal variability with that in the northwest tropical Pacific with the former lagging the latter (Figures 2b and 2d). The lag times determined from monthly and three-year smoothed time series are 6 and 12 months with corresponding cross-correlation of 0.80 and 0.95 (both being significant at the 95% confidence level). These lags suggest that the SSH signal in the southeast Indian Ocean is forced remotely by winds in the tropical Pacific and transmitted from the northwestern Pacific via the Indonesian Archipelago.
 These relationships highlight an oceanic linkage between the lower branches of the Pacific and Indian-Ocean STCs that causes their anti-correlated variability. Feng et al.  discussed the correlation of sea level at Fremantle of western Australia (near 32°S) with Southern Oscillation index on multi-decadal time scales. Cai et al.  found that thermocline depth signal associated with El Nino/Southern Oscillation (ENSO) could propagate from the northwestern Pacific to the coast of southwestern Australia via wave propagation with a 9-month transmission time that is not very different from our estimates. Shi et al.  discussed the difference in the transmission pathways of ENSO signals (North-Pacific versus equatorial Pacific pathways) to the Indian Ocean before and after 1980 in an ocean data assimilation product. However, none of these studies discussed the implications for coherent inter-basin changes in the STCs on decadal time scales using SSH differences between east and west coasts.
 Is there any relation between upper branches of the Pacific and Indian-Ocean STCs? Figure 3a shows strengthening trade winds in the tropical Pacific and weakening trade winds in the South Indian Ocean during 1993–2000. The former corresponds to an increasing divergence of Ekman transport out of the equatorial Pacific. The latter implies a reduction in southward Ekman transport in the South Indian Ocean. The overall reversal in the trends of trade wind intensity during 2000–2006 (Figure 3b) indicates a corresponding reversal in the tendencies for Ekman transports in both oceans. The time series of annually averaged zonal wind stress in the equatorial Pacific (Figure 3c) and in the Indian Ocean at 10°S (Figure 3d) further illustrate the changes of the trade winds in the two oceans. They tend to be anti-correlated on decadal and interannual time scales, probably because of zonal shifts or coherent oscillations in the centers of deep convection associated with the Walker circulation over the tropical Pacific and Indian Oceans. These large-scale wind changes provide an atmospheric bridge that links the upper branches of the STCs in the two oceans.
 In contrast to the southeast Indian Ocean (box d in Figure 1), the trend of SSH in the south-central tropical Indian Ocean (box c) is due to the trend of wind stress curl associated with weakening Indian Ocean trade winds centered to the east (Figure 1). Because the latter co-varies with the Pacific trade winds via atmospheric bridge described earlier, the phase of SSH in this region (Figure 2c) is close to that in the central tropical Pacific (Figure 2a), providing oceanic evidence for the trade-wind linkage.
 How do the decadal changes of the STCs described above relate to previous observations? For the Pacific, MZ02 discussed a multi-decadal slowdown of the STCs from the 1970s to the mid-1990s. MZ04 found that the Pacific STC averaged over the 1998–2003 period was stronger than that in previous decades. However, the latter study acknowledged a sensitivity to the influence of 4–5 year period ENSO variations that could affect inferences about the exact timing of decadal phase changes. The decadal phase change of the STC we describe here is better defined than in MZ04 because of the additional data available since 2003 and because of the use of running averages vs. averages over a fixed time window. For the Indian Ocean, Lee  only discussed the weakening STC in the 1990s, but not the strengthening afterwards because of the limited data available then. Moreover, the anti-correlated variability of the Pacific and Indian-Ocean STCs and the oceanic and atmospheric linkages have not been identified before.
4. Decadal Changes in the Extra-Tropics of the Pacific
 As in the tropics, SSH trends in many regions of the extra-tropical Pacific generally have opposite signs between the 1993–2000 and 2000–2006 periods (Figures 1a and 1c). Examples of SSH time series averaged over several key regions in the subtropical and subpolar regions with large trends (Figures 2e–2i) further demonstrate the change of decadal tendency near the end of the century. Many large-scale SSH trends in the extra-tropics are associated with well-defined wind stress curl trends both for the 1993–2000 and 2000–2006 periods (e.g., arrows in Figure 1). Qiu  and Qiu and Chen  for example have discussed the wind forcing of SSH in the North and South Pacific. However, these studies did not identify coherent decadal phase changes in SSH and winds at the end of the century. Interestingly, the decadal changes of wind stress curl (Figure 1b) resemble those in the North Pacific during the late 1970s reported by Miller et al. , who investigated their impact on thermocline depth and gyre circulation.
 The observed SSH changes have direct implications for the strength of the subtropical and subpolar gyres in the Pacific. The SSH data suggest the following changes in gyre circulation during 1993–2000 with a reversed tendency after 2000: a strengthening Alaska gyre (box f), strengthening subtropical gyre in the North central Pacific (box e), and a strengthening subtropical gyre in the South Pacific (boxes g and h). Roemmich et al.  attributed the spinup of the South Pacific subtropical gyre in the 1990s to a decadal intensification of wind stress curl associated with the atmosphere's southern-hemisphere annular mode. The enhanced wind stress curl is due to strengthening circumpolar westerlies and weakening mid-latitude westerlies. This trend has persisted since the 1970s, affecting gyre circulations in mid- to high-latitude southern-hemisphere oceans [Cai, 2006; Roemmich et al., 2007]. The lack of significant reversal of SSH and wind stress curl in the southwestern subtropical Pacific after 2000 (e.g., box g) may be related to the tendency for a decadal decrease forced from the tropics superimposed on a longer-period increasing trend forced from the high-latitude region. However, other explanations are possible that may not involve tropical forcing.
 The three-year smoothed time series in Figure 2 suggest that the decadal phase change in the tropics (Figures 2a, 2b, and 2c) leads that in the extratropics (Figures 2d–2i). Cross-correlation of SSH in the central equatorial Pacific (Figure 2a) with the SSH in other regions shown in Figure 2 using the monthly, annually smoothed, and three-year smoothed time series (black, red, and blue curves in Figure 4) further illustrate the timing of these changes. The maximum values of the cross-correlation are significantly different from zero at 95% level, but generally not significantly different from the values at zero lag unless monthly time series are used to increase the number of degree of freedom. However, all the cross-correlation curves consistently show SSH in the central-equatorial Pacific (an area directly affected by the Pacific trade winds) leads other regions. This suggests that the tropical Pacific plays an active role in regulating decadal variability in the extra-tropics through atmospheric teleconnections [e.g., Trenberth and Hurrell, 1994; Deser et al., 2004].
5. Concluding Remarks
 Satellite measurements of SSH and wind stress for the period of 1993–2006 reveal a near-coherent decadal variation over much of the Indo-Pacific region that is characterized by a change of decadal tendencies at the end of the 20th century. The change in the tropical Pacific tends to lead those in the other areas suggesting that this coherent inter-basin variation probably originates from ocean-atmosphere coupling in the tropical Pacific. These changes reflect linkages between the Pacific and Indian-Ocean STCs with a stronger Pacific STC corresponding to a weaker Indian Ocean STC. In the extra-tropics, the observed SSH variations also suggest a decadal phase change in the strength of subtropical and subpolar gyres in the North and South Pacific. However, the exact cause of these oscillations and the potential feedback of ocean circulation changes to the atmosphere need to be investigated further. Likewise, the relationship of these oscillations to decadal variations in ENSO (which may result in residual decadal signals) and the Pacific Decadal Oscillation, needs to be determined. The unique observational perspective presented here provides a basis for validating coupled ocean-atmosphere models that can be used to further explore the mechanisms responsible for this coherent inter-basin decadal variability.
 This research was carried out at the JPL, California Institute of Technology, under a contract with NASA, and at PMEL with support from NOAA. PMEL publication 3121.