Recent observations indicate that the shallow meridional overturning circulation in the tropical Pacific Ocean has rebounded since 1998, following 25 years of significantly weaker flow. Convergence of cold interior ocean pycnocline water towards the equator increased to 24.1 ± 1.8 × 106 m3 s−1 during 1998–2003 from a low of 13.4 ± 1.6 × 106 m3 s−1 during 1992–98. Intensified circulation led to the development of anomalously cool tropical Pacific sea surface temperatures, which may have affected Pacific marine ecosystems and global climate. The abruptness of the rebound also obscures presumed anthropogenic warming trends in the instrumental temperature record of the tropical Pacific.
 The Pacific basin is characterized by prominent modes of natural decadal variability, one representation of which is the Pacific Decadal Oscillation (PDO) [Mantua et al., 1997]. The time evolution of the PDO has been described alternately in terms of fluctuations across a broad band of periods between 20–70 years or in terms of regime shifts which manifest themselves as abrupt changes in climatic conditions and marine ecosystems over large areas of the Pacific basin. The most studied of these regime shifts occurred in 1976–77, tropical manifestations of which included a slowdown of the shallow meridional overturning circulation and a warming of the sea surface by nearly 1°C in the cold tongue of the eastern and central equatorial Pacific Ocean [McPhaden and Zhang, 2002]. The meridional overturning circulation involves equatorward transport of cold water in the upper pycnocline (roughly the upper 50–300 m of the ocean) of both hemispheres, upwelling near the equator, and divergent poleward flow in a shallow surface Ekman layer [McCreary and Lu, 1994].
 It has recently been argued that a new regime shift occurred in the late 1990s in the North Pacific associated with the PDO [Chavez et al., 2003; Peterson and Schwing, 2003]. However, the tropical Pacific experienced a nearly coincident abrupt shift from an intense El Niño to a strong La Niña in mid-1998. Year-to-year fluctuations associated with El Niño and La Niña (which represent the warm and cold phases of ENSO, respectively) have similar spatial signatures to those of the PDO both in the tropics and at midlatitudes where ENSO teleconnections leave their imprint on oceanic and atmospheric variability. Thus, in order to detect decadal variations in the physical system, it is necessary to average over at least one ENSO cycle, equivalent to about 5 years of data. Enough time has elapsed since 1998 that this requirement can now be met.
 In this study, we compare mean conditions for the 6-year period July 1992–June 1998 (referred to hereafter as 1992–1998) with the more recent 5-year period July 1998–June 2003 (referred to as 1998–2003). Both periods span at least one complete ENSO warm and cold phase cycle [McPhaden, 2004] while the average PDO index [Mantua et al., 1997] based on North Pacific SSTs dropped from +0.84 in 1992–1998 to −0.36 in 1998–2003. Coincidentally, SSTs in the equatorial Pacific cold tongue decreased, the trade winds strengthened, and sea level sloped up towards the west more steeply along the equator (Figure 1). Qualitatively, we would expect that off the equator the intensified east-west sea level slope would lead to increased equatorward geostrophic flow extending into the upper pycnocline. Likewise, stronger trade winds would be expected to increase Ekman divergence from the equator, leading to stronger equatorial upwelling and colder eastern and central Pacific SSTs. The purpose of this paper is to quantitatively evaluate these changes in wind-driven ocean circulation and their relation to SST.
2. Data and Methods
 We utilize hydrographic data and four different surface wind analyses spanning the period 1992–2003. The hydrographic data allow us to calculate geostrophic meridional volume transports in the upper pycnocline following procedures described in McPhaden and Zhang . We use 900 db (1 db = 0.99 m) as a reference level for the geostrophic calculation though levels between 600 db and 1200 db lead to similar results. The total number of hydrocasts to 900 db is 11,585 for July 1992–June 1998 and 6,729 for July 1998–June 2003 (Figure 2). The upper pycnocline is defined as the range of density classes that encompasses equatorward moving water masses immediately below the surface mixed layer, i.e., density classes between 22–26 kg m−3 in the Northern Hemisphere and between 22.5–26.2 kg m−3 in the Southern Hemisphere. The wind products we use are the Florida State University (FSU) pseudostress, the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis winds, the European Centre for Medium Range Weather Forecasts (ECMWF) reanalysis and operational analysis winds, and a combination of satellite scatterometer winds from the European Remote Sensing (ERS) missions and the National Aeronautics and Space Administration (NASA) QuikScat mission. In each case, winds were converted to stress using a wind speed dependent drag coefficient.
 Geostrophic transports are integrated from the eastern boundary to 145°E in the Northern Hemisphere and 160°E in the Southern Hemisphere. The westernmost longitudes of integration are several hundred kilometers away from the western boundary where strong boundary currents and their recirculations are not adequately resolved by the data. We choose a latitude of 9°N for this transport calculation as it represents a latitude at which pycnocline flow between the subtropics and the equatorial zone is constrained by the potential vorticity ridge [McCreary and Lu, 1994] (see also Figure 2). For hemispheric symmetry, we perform the same calculation along 9°S since potential vorticity in the Southern Hemisphere is more uniform in structure with no obvious latitudinal choke point.
 Geostrophic transport values for 1992–1998 are very similar to those in McPhaden and Zhang  for the period 1990–1999 (Figure 3a) as one would expect from the significant overlap in data sets used. On the other hand, extension of the analysis to 1998–2003 shows a remarkable rebound in equatorward transports. Interior ocean pycnocline transports approximately doubled in the Northern Hemisphere and intensified to levels not seen since before the 1976–77 regime shift in the Southern Hemisphere. Pycnocline transport convergence increased by approximately 10 Sv (1 Sverdrup = 106 m3 s−1) from 1992–1998 to 1998–2003 reaching its highest value since the mid-1970s (Figure 3b, Table 1). Consistent with an increased supply of cold pycnocline water to feed equatorial upwelling, SST dropped more than 0.6°C in the region 9°N–9°S, 100°–180°W (Figure 3b).
Table 1. Meridional Transport Convergence Across 9°N–9°Sa
July 1992–June 1998
July 1998–June 2003
In units of 106 m3 s−1. Convergence is calculated as transport at 9°S minus transport at 9°N. Divergence is the negative of convergence. Sverdrup and Ekman transport means and standard errors are based on the ensemble of four wind products used in this study.
Geostrophic observed in pycnocline
13.4 ± 1.6
24.1 ± 1.8
Geostrophic from Sverdrup theory
13.5 ± 4.7
29.1 ± 10.2
Ekman in surface layer
−50.3 ± 4.3
−58.3 ± 5.9
Sverdrup  theory provides an independent estimate of transport changes on the assumption that the ocean is in near-equilibrium with surface wind forcing. This assumption, which neglects the details of baroclinic wave processes in adjusting the ocean to changes in wind forcing, is a reasonable approximation at low latitudes on 5–6 year time scales [Philander, 1979]. We thus compute meridional geostrophic transports from Sverdrup theory for four different wind products (Table 1). All four products independently indicate an increase in the geostrophic equatorward transports between 1992–1998 and 1998–2003, with the ensemble average of the four being 13.5 ± 4.7 Sv for 1992–1998 and 29.1 ± 10.2 Sv for 1998–2003. The increase in modeled equatorward geostrophic volume transport convergence is similar to that observed despite the relatively large uncertainties that result from product-to-product differences in surface wind stress and its spatial gradients (namely, wind stress curl). The good agreement between theory and observations could be fortuitous since transports determined from Sverdrup theory are meant to apply to the entire water column, not just the pycnocline. However, the wind-driven circulation in the tropical Pacific is surface intensified, which leads us to conclude that the rebound in the pycnocline transports we infer from the hydrographic data is real.
 The relationship between pycnocline transport convergence and SST in Figure 3b is consistent with the notion that increasing the transport of cold pycnocline water towards the equator feeds increased upwelling to cool the surface. One would expect therefore that Ekman transport would also increase in the surface layer to carry the increased volume flux of upwelled waters poleward. To test this hypothesis, we compute meridional Ekman transports along 9°N and 9°S using the four wind products. Each independently indicates an increase in Ekman transport divergence across 9°N and 9°S between 1992–98 and 1998–2003. The ensemble average of the four is 50.3 ± 4.3 Sv for 1992–1998 and 58.3 ± 5.9 Sv for 1998–2003 (Table 1) consistent with the premise that upwelling of cold pycnocline water has increased in the latter period.
 The large 20–30 Sv difference between the Ekman transport divergences and geostrophic convergences along 9°N and 9°S (Table 1) for the most part reflects transports in the swift and narrow western boundary currents which are unaccounted for in our calculation. Various model simulations indicate, moreover, that on seasonal to decadal time scales, western boundary current transports tend to partially compensate changes in interior ocean meridional geostrophic volume transports [Lee and Fukumori, 2003; Hazeleger et al., 2004]. Such compensation may be inferred from Sverdrup theory where the modeled increase in meridional geostrophic volume transport convergence (about 15 Sv) is larger than the increase in Ekman divergence in the surface layer (about 8 Sv) between 1992–1998 and 1998–2003 (Table 1). This western boundary current transport compensation can be seen in the difference of the Sverdrup transport streamfunction for the two periods shown in Figure 4b. The inferred decadal changes in western boundary current transports of O(several Sv) are larger than estimated decadal changes in Indonesian Throughflow of ≲1 Sv [Schneider, 1998; McPhaden and Zhang, 2002; Lee and Fukumori, 2003], so the latter can be neglected when considering the decadal mass balance.
 A lack of uniformly distributed hydrographic data extending to 900 db prevents us from estimating pycnocline transports at latitudes higher than 9°. However, the broad latitudinal scale of the wind stress and sea level changes (Figure 1) and calculations based on Sverdrup theory (Figure 4) suggests that circulation changes similar to those we have described along 9°N and 9°S may extend to 20°N and 20°S. Indeed, the anomalous anticyclonic circulation cells within 20° of the equator in the western Pacific (Figure 4b) are qualitatively consistent with the observed pattern of elevated sea level in the region (Figure 1b), since sea level slope is proportional to geostrophic flow in the upper ocean. These results imply possible interactions between the tropical and subtropical gyres in generating the observed variability [Kleeman et al., 1999; Schott et al., 2004].
4. Summary and Conclusions
 The shallow meridional overturning circulation in the tropical Pacific Ocean has recently rebounded to levels almost as high as in the 1970s in association with an intensification of the trade winds. These observations are consistent with theories of decadal variability that invoke tropical-subtropical interactions [Kleeman et al., 1999]. However, we cannot rule out the possibility that the changes we have observed are simply the residual of decadal modulations in the statistics of ENSO variations [Kirtman and Schopf, 1998]. Also, precisely why the trades have intensified is a question beyond the scope of this study. Resolving these issues will require coupled model studies that can realistically simulate fluctuations like those described here.
 Tropical Pacific SSTs significantly cooled in conjunction with the observed circulation changes, presumably in response to increased upwelling along the equator. Cooling of underlying SSTs in the eastern and central Pacific was accompanied by shifts in the patterns of tropical rainfall and deep convection (Figure 1c) which we would expect to affect the global atmospheric circulation through teleconnections to higher latitudes [Trenberth and Hurrell, 1994]. It may be therefore that the prolonged changes in climatic conditions and associated marine ecosystem changes observed in the North Pacific since the late 1990s [Chavez et al., 2003; Peterson and Schwing, 2003] are in part a response of these teleconnections. Also, atmospheric general circulation models forced by observed tropical SSTs indicate that the Pacific Ocean had a significant influence on the development of globe girdling drought that gripped much of the subtropical Northern Hemisphere during 1998–2002 [Hoerling and Kumar, 2003].
 Finally, some coupled ocean-atmosphere model studies have suggested that warming trends in the tropical Pacific Ocean in the second half of the 20th century were due to greenhouse gas forcing [Meehl and Washington, 1995; Knutsen and Manabe, 1998]. Our results indicate that precise magnitude of anthropogenic influences will be difficult to extract with confidence from the instrumental record given the rapidity with which observed warming trends can be reversed by natural variations (Figure 3b). Thus, coupled model assessments of warming trends in the tropical Pacific will be most reliable for those models that can accurately replicate observed energetic decadal variations like those described in this study.
 The authors would like to thank Gary Mitchum of the University of South Florida for providing the blended TOPEX/Poseidon altimeter analysis. This work was supported by the NOAA Office of Global Programs and Office of Oceanic and Atmospheric Research, and the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) at the University of Washington. JISAO contribution 1084 and PMEL contribution 2686.