Observations from the tropical Pacific Ocean identify an abrupt climate shift in 1976 with surface temperatures changing from cooler than normal to warmer than normal in the span of about 1 year. Model results show that this climate shift originates with subsurface temperature anomalies in the south tropical Pacific Ocean which propagate first to the western boundary, then northward to the equator, and finally eastward along the equator to 140° W where they rise to the surface. The results suggest that changes in the North Pacific respond to changes in the tropical Pacific.
 These Northern Hemisphere mechanisms dominate in the literature in spite of evidence from observational [Lindstrom et al., 1987] and modeling [McCreary and Lu, 1994; Blanke and Raynaud, 1997] studies that show that the Southern Hemisphere contributes as much as 70% to the water masses in the equatorial undercurrent. Historical observations and atmospheric reanalysis products indicate that on both interannual and decadal time scales there are well defined atmospheric anomalies in the Southern Hemisphere with comparable amplitude to those in the Northern Hemisphere [Karoly, 1989; Garreaud and Battisti, 1999]. There is also new evidence that changes in the current structure in the western Pacific near the coast of Australia and New Guinea are closely linked to changes in the equatorial undercurrent during the mid-1970's [Chang et al., 2001].
 In this paper we explore the possibility that the large scale climate change that took place across the Pacific Ocean in 1976 was forced by subsurface changes in the ocean south of the equator and is part of an oscillation intrinsic to the equatorial and southern tropical Pacific.
 An ocean general circulation model with a data assimilation package was used to explore ocean changes associated with the 1976 climate transition. The global version of the model used is an intermediate resolution version of the Modular Ocean Model 3 (MOM3) code developed at the National Oceanic and Atmospheric Administration's Geophysical Fluid Dynamics Laboratory. The model's spatial domain is global in longitude and extends from 60°S to 60°N in latitude. The model incorporates an optimal interpolation scheme which assimilates surface and subsurface temperature data. The assimilation scheme and its incorporation into the model has been described in detail elsewhere [Carton et al., 2000].
 The model is forced with weekly reanalysis winds from the National Center for Environmental Prediction (NCEP) [Kalnay et al., 1996]. The reanalysis winds have a considerable weak bias in the tropics [Putman et al., 2000], thus the reanalysis winds are bias corrected before being used in the model. The bias correction adjusts the NCEP reanalysis mean and variance to the mean and variance of the more accurate NCEP production run winds. This correction varies spatially, with largest differences in the tropical Pacific and Atlantic Oceans, but does not vary in time. SST is damped to NCEP weekly values and the sea surface salinity is damped to monthly mean climatology from the Comprehensive Ocean Atmosphere Data Set (COADS). The damping time scale for both temperature and salinity is 21 days. The model was spun up for 20 years using climatological winds and then run from 1948 to 1999 using data assimilation of hydrographic temperature data.
 To describe the ocean changes that occurred in the mid-1970s we first show the spatial patterns of temperature and wind stress anomalies at three year intervals from 1968 through 1983. The model results have been averaged over a 5 year period to reduce the considerable interannual variability that exists in the tropical Pacific. The first column in Figure 1 shows the evolution of temperature anomalies on the 1025 kg m−3 isopycnal surface. The 1025 kg m−3 isopycnal surface was chosen because this density surface roughly coincides with the core of the equatorial undercurrent, and so water mass anomalies on this surface can be readily advected into the undercurrent. The first panel in column 1 shows positive temperature anomalies in the central and eastern Pacific in both the Northern and Southern Hemispheres. The Southern Hemisphere anomalies extend from 110°W to 160°W centered at a latitude of about 10°S. The Northern Hemisphere anomalies are weaker and cover a smaller region, extending from the coast of Mexico to about 140°W. By 1971 the Southern Hemisphere anomalies have grown considerably, extending across the entire southern Pacific, from the coast of Australia to the region where the 1025 kg m−3 isopycnal surface outcrops near 100°W. Subsequent panels show the continuation of these warm anomalies, with a propagation of the center of the Southern Hemisphere anomalies to the west over the 10 year period from 1968 to 1977. In contrast with the temperature anomalies in the Southern Hemisphere, the temperature anomalies in the Northern Hemisphere weaken and dissipate. Although we can identify weak temperature anomalies that propagate westward to the western Pacific, these anomalies are far weaker than those of the Southern Hemisphere.
 By 1974 Southern Hemisphere temperature anomalies reach the western boundary, where they propagate northward to the equator and subsequently propagate to the east. Since the 1025 kg m−3 isopycnal surface gets shallower towards the east, this represents movement of the temperature anomalies toward the surface, where they eventually reach the surface, outcropping near 140°W.
 SST and surface wind stress, shown in the center and right columns in Figure 1, show that these anomalies develop together, confirming that temperature and surface winds are strongly coupled to each other. SST shows strong negative anomalies in the eastern equatorial Pacific in the late 1960s and early 1970s which weaken with time. After 1976 there are considerable positive SST anomalies that extend from the coast of South America to the central Pacific. Wind stress anomalies show a similar evolution, with stronger than normal easterly tradewinds in the central Pacific which weaken and are eventually replaced with anomalous westerlies. Interestingly, the westerly anomalies are shifted to south of the equator, and although less noticeable, the stronger than normal easterly anomalies prior to 1976 also are shifted to south of the equator. In the late 1970s temperature anomalies on the 1025 kg m−3 isopycnal surface become cooler than normal south of the equator, and these cool anomalies also propagate to the west.
 We can illustrate the propagation of these temperature anomalies by showing a time longitude map of temperature anomalies. Figure 2 shows temperature anomalies on the 1025 kg m−3 isopycnal surface, from 120°W to 150°E averaged between 7° and 10°S latitude. The sense of the x-axis has been reversed so that east is on the left hand side of the figure for this section. The central panel shows temperature anomaly on the 1025 kg m−3 isopycnal surface on the equator from 150°E to 140°W (where the isopycnal surface typically outcrops) and the right hand panel shows SST from 140°W to 90°W. Focusing first on the evolution of SST anomalies, shown in the right hand panel, the dramatic shift to warmer SST is clearly identified in 1976. Prior to 1976 SST is anomalously cool by about 1.25°C and after 1976 SST is anomalously warm by the same amount. The subsurface temperature anomalies clearly show that these surface temperature changes were preceded by temperature anomalies in the subsurface southern Pacific Ocean. In fact, we can identify the propagation of warm temperature anomalies on the 1025 kg m−3 isopycnal surface as early as 1968 in the eastern and central Pacific. These temperature anomalies show a clear sense of propagation to the west, where they arrive at the coast of Australia in 1973. The anomalies then migrate equatorward reaching the equator in the far western Pacific in late 1974. The anomalies then propagate relatively quickly across the Pacific along the equator, reaching the surface near 140°W in 1975 and 1976. Once the temperature anomalies reach the surface we can no longer track them on an isopycnal surface, but we can track the evolution of SST anomalies. When the isopycnal temperature anomalies reach the surface on the equator there is rapid growth of SST anomalies, presumably through positive air-sea interaction.
 Most of the evidence regarding a climate shift in 1976 comes from observations made in the North Pacific. The model results shown here do not show any direct linkage to the North Pacific Ocean, so it remains to be explained how the North Pacific might undergo climate changes simultaneously with the equatorial Pacific. Mid-latitude Pacific Ocean variability has been described as a response to changes in the tropics through large scale changes in the atmospheric circulation [Lau and Nath, 1996; Giese and Carton, 1999]. Figure 3 shows the anomalies of the height of the 200 mb surface from the NCEP reanalysis product regressed onto central Pacific (NINO3) SST anomalies. The height anomalies show a large scale change in the height of the 200 mb surface which varies with tropical SST. Some of the largest Northen Hemisphere changes are found in the western portion of the domain, near the Kuroshio extension region, extending away from the coast of Japan and extending into the eastern North Pacific. In fact, this region near 40 N is the region where the largest SST anomalies on decadal time scales are found.
 One interpretation of Figure 3 is that there is a Pacific-wide shift in the atmospheric circulation caused by the tropical heating anomalies, a mechanism for decadal change proposed by [Giese and Carton, 1999]. An equatorward shift in the circulation after 1976 results in stronger than normal westerlies across the Pacific at 30°N to 50°N, in turn resulting in cooler SST in the North Pacific. Since the atmospheric response to changes in SST is rapid, there may be little lag between tropical and North Pacific SST anomalies.
4. Summary and Conclusions
 These results describe an oscillation intrinsic to the equatorial and south tropical Pacific Ocean, with coupling between SST and surface wind stress on the equator and subsurface temperature anomalies off the equator in the Southern Hemisphere. The time delay between surface wind forcing and the resulting change in SST arises because of the time that it takes for the off-equatorial subsurface temperature anomalies to propagate across the Pacific to the western boundary, along the western boundary to the equator and finally across the equator to the eastern Pacific. We estimate the anomalies propagate at a speed of about 4 to 12 cm s−1, slower than the speed of a Rossby wave at this latitude. This scenario is similar to the delayed oscillator mechanism that has been used to describe El Niño events [Suarez and Schopf, 1988; Battisti and Hirst, 1989]. However in this case the feedback between anomalous wind stress and subsurface thermal changes, presumably through subduction of surface anomalies, resides primarily in the Southern Hemisphere.
 The results also suggest that the tropical Pacific changes during 1976 do not result from changes in the North Pacific, but instead force North Pacific anomalies. Since the atmospheric response to changes in SST is rapid, there may be little lag between tropical and North Pacific SST anomalies. This framework suggests that the North Pacific plays a role primarily of responding to equatorial SST anomalies via an atmospheric pathway, instead of forcing climate variations across the Pacific. If so, then an increased effort to understand the dynamics and thermodynamics of the Southern Hemisphere is needed to better understand climate fluctuation on decadal time scales.
 This manuscript benefited greatly from discussions with T. Crowley, N. Slowey, J. Carton, and G. Giese. This work was supported by a grant from NOAA (NA86GP0207).