The phenomenon playing the main role is the PDO (refs 1, 9), alternatively known from a slightly different perspective as the Interdecadal Pacific Oscillation, although it may not be a quasi-linear mode of natural variability10. The PDO was in a negative phase before 1976, but became positive from 1976 to 1998, a period coinciding with strong increases in global mean surface temperatures1,11. Then it switched to a negative phase in 1999 coinciding with the pause in upward trend in global mean surface temperatures. However, since 1999, the deeper ocean below 700 m has taken up more heat and there has not been a reduction in the Earth’s energy imbalance3,8.

The PDO pattern (Fig. 2) emerges from an analysis of the departures from the global mean of sea surface temperature (SST) monthly anomalies using a core region from 20° to 70° N, 110° E to 100° W for an empirical orthogonal function analysis1,11. The base period is 1900–2012. The PDO/Interdecadal Pacific Oscillation has a Pacific-wide pattern in both surface and subsurface temperatures with an El Niño-like pattern throughout the tropics and strong extratropical links in both hemispheres (Fig. 2). The subsequent analysis only uses the PDO to provide markers for specifying the last two climate regimes: 1999–2012 versus 1976–1998 (ref. 1). This is more robust than using short-term linear trends9, although results are similar. The recent PDO signals for the annual means1 are complemented here with further diagnostic fields, including especially precipitation and atmospheric diabatic heating.

Figure 2: The PDO based on an empirical orthogonal function analysis of SST anomalies with the global mean removed from 1900 to 2012 in the 20°–70° N, 110° E–100° W region of the North Pacific, which explains 25% of the variance. The principal component time series, given below in normalized units, is regressed on global sea and land surface temperature anomalies to give the map above. The black curve is a 61-month running average. The light red and blue colours depict the positive and negative phases of the PDO. Note the reversal of the colour key in the top panel so that blue colours are positive, and hence the current negative phase has below-normal SSTs in the blue areas. Here, s.d. is standard deviation. Full size image

Since 1999 the annual mean tropical Pacific easterly trade winds have been much stronger than normal1,9 (see also Fig. 3). In the tropics and subtropics this coincides with striking sea-level pressure anomalies that relate to the Southern Oscillation. Much deeper warm waters have piled up in the western Pacific while cooler waters have prevailed through the top 500 m in the eastern Pacific1. The result has been an increase in sea level of order 20 cm near the Philippines, but slight falls in the east since about 1999 (ref. 12), accompanied by changes in ocean heat content1.

Figure 3: Regime differences between 1999–2012 and 1976–1998. a–f, Mean differences between 1999–2012 and 1979–1998 for NDJFM (left) and MJJAS (right) for surface temperature from GISS (a,b), mean sea-level pressure (MSLP) differences from ERA-I (colours) and surface wind vectors (at 10 m, V10, arrows) with the key at the top right (c,d), and precipitation from GPCP truncated to T63 resolution (e,f). GISS: Goddard Institute for Space Studies; ERA-I: European Centre for Medium-range Weather Forecasts atmospheric reanalysis ERA-Interim; GPCP: Global Precipitation Climatology Project v2.2; see Supplementary Information. Full size image

The pronounced strengthening in Pacific trade winds in the 2000s was unprecedented after 1979 and not captured by 48 climate model projections3,9 but was sufficient to account for the cooling of the tropical Pacific and a substantial slowdown in global surface warming through increased subsurface ocean heat uptake. While the strongest anomalies were indeed in the Pacific, strong connections also existed to the North Atlantic and southern oceans where subsurface ocean heat uptake penetrated to well below 700 m depth1, altering surface winds and the distribution of Antarctic sea ice13, and the winter weather throughout Europe and North America. In the Pacific, the extra uptake came about through increased subduction in the Pacific shallow overturning cells, enhancing heat convergence in the equatorial thermocline9,14,15.

Surface temperature (data sets are described in Supplementary Information) differences (Fig. 3a, b) clearly show that the central and eastern Pacific has failed to warm in the past 14 years, in a pattern associated with the PDO. To show the seasonality, two extended seasons November–March (NDJFM), northern winter, and May–September (MJJAS), northern summer, are examined. The Pacific cooling is stronger in northern winter (Fig. 3a). In general, surface warming deviates considerably from the PDO pattern outside the Pacific basin, perhaps signalling a substantial role for external forcing of the climate system and other factors such as internal variability beyond that associated with the PDO.

Precipitation differences between these two regimes (Fig. 3e, f) feature patterns similar to those associated with La Niña: much drier from 160° E to South America along the Equator, a northward-shifted but weaker northern Intertropical Convergence Zone, a southwestward-shifted South Pacific Convergence Zone, and extensive heavy rains over the Indonesian maritime continent. Wet conditions over northern Australia, Southeast Asia, northern Brazil, Colombia and Venezuela extend over the tropical Indian and Atlantic oceans (Supplementary Fig. 2) and are probably associated with changes in the Walker circulation16. The subtropical Pacific dry zone extends across the southeastern United States and throughout much of the North Atlantic in winter. Considering that the differences in Fig. 3 are for a 14-year period, the extent of the values exceeding 1 mm d−1 is exceedingly large (p < 0.05 based on the interannual standard deviation in the tropical Pacific from 1979 to 199517 and two-sided Student’s t-test), and suggests that much greater values occur in some years.

Complementary perspectives provided by the changes in sea-level pressure and surface winds (Fig. 3c, d) have some similarity in the tropical Pacific in both seasons, with a positive Southern Oscillation Index, but exhibit stronger anomalies in NDJFM. In the tropics, the intensification of easterly winds just south of the Equator near the Date Line in NDJFM exceeds 2 m s−1 versus about 1.1 m s−1 in MJJAS. These compare with mean vector wind speeds of up to 8 m s−1 in December–February and less than 7 m s−1 in June–August for 1950–197918. Seasonal mean standard deviations are mostly less than 1 m s−1, and hence an anomaly of 1 m s−1 over 14 years is highly statistically significant (p < 0.01). Sea-level pressure anomalies are large in both hemispheres in the subtropics, but stronger wave-like teleconnections in each hemisphere occur in winter.

The regime differences for the diabatic heating from radiation and latent and sensible heating of the atmosphere (Fig. 4) have been computed as a residual of the energy equations using ERA-I data19,20. As 1 mm d−1 is equivalent to latent heating of about 29 W m−2, its dominant role in the anomalous diabatic heating is apparent from comparing Fig. 3e, f with Fig. 4, and it is where precipitation anomalies exceed at least 0.5 mm d−1 over broad areas that a significant decadal response in terms of atmospheric teleconnections can be expected21,22.

Figure 4: Vertically integrated diabatic heating and divergent wind component for the difference between 1999–2012 and 1979–1998. Diabatic heating (W m−2) is computed as a residual from the energy equation and lightly smoothed with a 4-point smoother. The vectors show the divergent wind component at 100 hPa for the annual means, based on ERA-I reanalyses. Full size image

The degree to which the extratropical signal is wave-like in sea-level pressure is better seen in a polar projection (Supplementary Fig. 3). The alternating anomalies are clearest in MJJAS in the Southern Hemisphere. However, wave-like responses are best seen in the upper troposphere in the streamfunction field. Accordingly, we focus on the 300 hPa streamfunction differences (Fig. 5), which depict the rotational part of the flow (the flow is along the contours). By far the strongest feature is the dipole structure in the central tropical Pacific straddling both the Equator and the region of the strongest precipitation deficit. There is a single predominant centre in the upper tropospheric tropics of each hemisphere with a cyclonic sign, as expected in conjunction with strong meridional convergence into the equatorial region and subsidence21 which can be inferred from the divergent wind components at 100 hPa (Fig. 4). That convergence is consistent with local atmospheric overturning driven by precipitation latent heat anomalies. The negative heating anomalies associated with the exceptionally dry tongue along the Equator evidently dominate the atmospheric forcing (Fig. 4). The pattern outside the tropics is strongly suggestive of a wave train of Rossby waves emanating from these regions that are strongest in winter21, as indicated schematically by the green arrows. This pattern constructively interferes with the observed long-term trend in tropospheric circulation23 in some regions, and destructively in others.

Figure 5: The 300 hPa streamfunction for the differences between 1999–2012 and 1979–1998. a, NDJFM, and b, MJJAS, based on ERA-I reanalyses, where the rotational wind is parallel to the contours. The green arrowheads show schematically the main wave trains emanating from the tropical Pacific. Full size image

From theory and modelling21,22, tropical SST anomalies in the central and western Pacific are very effective in driving robust extratropical atmospheric responses. High SST anomalies lead to low surface pressure and low-level moisture convergence that feeds high precipitation, and vice versa for low SST anomalies. The latent heat (Fig. 3e, f) release in turn drives overturning circulations brought about in part by anomalous Walker circulations (Figs 4 and 5) throughout the tropics, and forces extratropical Rossby waves through the anomalous upper-level divergence flow (Fig. 4) perhaps involving the lower stratosphere24 in the extratropics. The term in the atmospheric vorticity equation, where β is the meridional gradient of the Coriolis parameter and is the anomalous divergent northerly wind component (Supplementary Information), makes a large contribution to driving the changes in the rotational flow19,25 owing to the large values of (Fig. 4).

The stationary Rossby waves tend to follow a great circle route, and in turn alter the mid-latitude storm tracks, transient eddy heat and momentum transports, and interact with orography and the climatological mean flow. Negative SST anomalies and associated local precipitation deficits tend to have effects of opposite sign.