Abrupt Shift in Subsurface Temperatures in the Tropical Pacific Associated with Changes in El Niño

See allHide authors and affiliations

Science  10 Jul 1998:
Vol. 281, Issue 5374, pp. 240-243
DOI: 10.1126/science.281.5374.240


Radiocarbon (14C) content of surface waters inferred from a coral record from the Galápagos Islands increased abruptly during the upwelling season (July through September) after the El Niño event of 1976. Sea-surface temperatures (SSTs) associated with the upwelling season also shifted after 1976. The synchroneity of the shift in both 14C and SST implies that the vertical thermal structure of the eastern tropical Pacific changed in 1976. This change may be responsible for the increase in frequency and intensity of El Niño events since 1976.

Several studies have noted that the pattern of El Niño–Southern Oscillation (ENSO) variability changed in 1976, with warm (El Niño) events becoming more frequent and more intense (1). This “1976 Pacific climate shift” has been characterized as a warming in SSTs through much of the eastern tropical Pacific. A recent study (2) proposed that this shift originated when a subsurface warm water anomaly in the North Pacific penetrated through the subtropics and into the tropics. This model is consistent with an association of the shift in tropical temperatures with changes in North Pacific sea-level pressures (3). However, this interpretation is controversial, and other mechanisms might be responsible. Unfortunately, hydrographic observations have spatial and temporal biases that do not allow for a definitive solution.

To examine changes in the origin of water upwelling in the eastern Pacific during the 1970s, we obtained a record of radiocarbon variability (Δ14C) from a coral from the Galápagos Islands (4). Skeletal carbonate from coral colonies records the Δ14C of the dissolved inorganic carbon in the surrounding water and can be used as a tracer of ocean circulation (5, 6). Surface waters are enriched in 14C relative to deeper water because of the air-sea exchange with14C that was produced by the atmospheric testing of nuclear weapons in the 1950s and 1960s. This enrichment difference makes the distribution of 14C very sensitive to vertical mixing. In the Galápagos Islands, Δ14C values at the surface are affected by the vigor of upwelling and by the origin of upwelling water. Our new Δ14C time series extends from 1957 through 1982 (ending when the colony died during the 1982–83 El Niño) and has an average resolution of eight samples per year (Fig. 1).

Figure 1

Galápagos coral Δ14C [in per mil (‰)] record from 1957 to 1983 (1σ error bars). For reference, WOCE P19 surface measurements for stations near the Galápagos taken from 29 March to 1 April 1993 average 70 per mil (8). Upwelling season maxima are indicated by solid circles; nonupwelling season maxima are indicated by open circles. Linear trends of the upwelling and nonupwelling season extremes are shown (dashed lines) (4) and are as described in the text. Shaded areas highlight El Niño years when the Niño-3 SST anomaly was ≥1°C.

The Δ14C values increase about linearly from pre-1960 values of −70 to −90 per mil, which is consistent with previous measurements of Galápagos corals (5, 7), through maximum values of 60 per mil in 1982. Water samples from nearby locations collected as part of the World Ocean Circulation Experiment (WOCE) program in late March and early April of 1993 give values of ∼70 per mil (8), implying that surface Δ14C values have stabilized and may have peaked in this region since 1983. In addition to the long-term trend, the coral record shows large seasonal and interannual variability; low Δ14C values are measured during July through September, and higher values are measured during January through March. The amplitude of the seasonal cycle increased from 20 to 30 per mil in the early 1960s to 50 to 100 per mil in the mid-1970s. After 1976, the seasonal variability decreased to 40 per mil because the minimum values increased abruptly by ∼60 per mil. The rate of increase of Δ14C during the upwelling season was 4.4 per mil year−1 between 1960 and 1976, and the rate decreased to 3.0 per mil year−1 thereafter. During the non-upwelling season, these rates were 7.0 per mil year−1, decreasing to 2.2 per mil year−1 after 1976.

The patterns of radiocarbon variations in the time series can be explained in terms of variability in the intensity of upwelling and changes in the source water that feeds the upwelling. The upwelling in the eastern tropical Pacific exposes water from the Equatorial Undercurrent, a subsurface water mass that flows from west to east, approximately bounded by potential density surfaces (isopycnals) of 24 and 26 kg m−3 (9). This water is derived from subduction of surface water in the subtropics and water entrained from greater depths in the tropical thermocline. The subtropical water brings higher bomb-produced 14C levels from the sea surface down into the undercurrent. The entrained component mixes colder, low14C water into the undercurrent and augments the Δ14C contrast between the undercurrent and the sea surface (6, 10). The deeper component with less bomb-produced 14C is at least as deep as the isopycnal of 26.5 kg m−3 (6, 9). The time for transport from the subtropics to the tropics is long enough (10 to 20 years) (11) that, when coupled with lower gas exchange in the tropics because of lower wind velocities, the peak of bomb-produced radiocarbon in the tropics is delayed relative to the peak in the subtropics (5, 6). The maximum Δ14C values that are reached in the eastern tropical Pacific are lower than those in the subtropical regions because of dilution by the entrained component. Upwelling in July through September, when the thermocline is shallow, brings water from the undercurrent to the surface, lowering Δ14C values. In January through March, when the thermocline is deeper, surface Δ14C values increase slightly because of air-sea exchange and lateral advection of off-equatorial high Δ14C water. The amplitude of the seasonal cycle around the Galápagos Islands increased from 1960 through 1976 because the surface water surrounding the upwelling region responded more quickly to the invasion of bomb-produced 14C than did water in the undercurrent.

Interannual variability in the coral record is dominated by ENSO. During warm events, the depth of the thermocline increases in the east, so that Ekman pumping no longer brings water from the undercurrent to the surface. This creates high Δ14C anomalies in the coral time series (1965–1966, 1969–1970, and 1972–1973), which are most pronounced from July through September. The increase of over 100 per mil within a few months in 1976 is unusually large in relation to the magnitude of the warm event. The Δ14C values during upwelling seasons remained anomalously high after this rapid rise. A linear trend through the minimum seasonal Δ14C values showed a 20 per mil offset in 1976, with no equivalent offset in maximum values (Fig. 1). There was no abrupt change in wind velocities in 1976 that could cause a sudden reduction in upwelling intensity, although it is difficult to interpret wind data because the observed and reconstructed wind fields show differing trends since 1960 (12).

The abrupt increase in minimum Δ14C values occurred when SSTs shifted in the eastern tropical Pacific and the frequency and intensity of ENSO warm phases increased. The shift in average temperature (Fig. 2) has a strong seasonal bias and is similar to the bias shown by the radiocarbon data. For the upwelling season, minimum SSTs were relatively stable and low from 1945 to 1955, and in 7 of 11 years, temperatures were lower than 22.5°C. The highest minimum SSTs were measured during the 1951 and 1953 El Niño events. A break occurred in 1955–1956, and through 1975, minimum temperatures were close to 23.5°C and were never as low as 22.5°C, although this break may be an instrumental artifact (13). Minimum SSTs shifted again in 1976 to a minimum value of 24.5°C (ENSO warm events excluded); only 1989 had temperatures that were lower than 23.5°C (14). Maximum temperatures (January to March) increased slightly after 1976 as a result of more frequent ENSO warm phases, with no change occurring in non–El Niño years (14).

Figure 2

SSTs in the Niño-3 region (90° to 150°W, ±5°) as observed in COADS (solid line), GOSTA (dashed line), and NMC/IGOSS (dotted line) databases (20). Warm season SSTs have remained relatively invariant, whereas upwelling season SSTs underwent a step-like warming in 1976. The horizontal lines indicate average minimum temperatures excluding strong El Niño years (Niño-3 SST anomaly ≥ 1.25°C) for the periods 1955–1975 and 1976–1992.

The restriction of the shift in Δ14C values and SSTs to the season of the most intense upwelling implies that a persistent shift in temperature and Δ14C of the subsurface waters in the eastern tropical Pacific occurred in 1976. The Δ14C data require that the source of the upwelling contained a greater proportion of water with bomb-produced radiocarbon after 1976, presumably from subduction in the subtropics. The simplest way to satisfy the SST and Δ14C constraints is to alter the vertical density structure along the equator so that the contribution of deeper, colder, lower Δ14C water to the upwelling region is reduced. A deepened thermocline in the eastern tropics after 1976, superimposed on variations in thermocline depth over seasonal and interannual time scales, is supported by subsurface temperature data (2), although the availability of such data before 1976, particularly data from the Southern Hemisphere, is not sufficient to identify more detailed patterns. A deepened thermocline is consistent with overall weaker zonal winds, although the reliability of the wind data, particularly for any long-term trends, is controversial (12).

A fundamental question is whether the change in the thermocline depth in 1976 is a contributing cause to the change in the frequency and intensity of El Niño or is merely an effect of the stronger warm events and reduced trade winds. A related question is whether the change predominantly represents a shift in winds affecting the tilt of the thermocline along the equator or a shift in ocean circulation affecting thermocline depth in the east without compensation in the west. Although numerous modeling studies have described the coupling of thermocline depth, wind stress, and El Niño dynamics (15), the persistence of warmer SSTs during upwelling seasons since 1976, even during cold phase (La Niña) conditions, suggests that a change in the ocean, independent from wind forcing, might be involved. One such change, suggested by Zhang et al. (2), involves subduction in the North Pacific of a warm temperature anomaly from the late 1960s and early 1970s. However, Zhang et al. note that a cold anomaly in the North Pacific after 1980 should have restored cooler conditions to the eastern tropics, rather than sustaining the warm conditions that have persisted. Furthermore, temperature and salinity data suggest that the South Pacific is the dominant source of water in the undercurrent (9) and therefore that the impact of an anomaly in the North Pacific should be minor. The shift at the equator occurred only 8 years after the maximum warm anomaly in the North Pacific, which is a shorter time period than most estimates for the transport time of water between the subtropics and the equator (11). A warm anomaly in the North Pacific cannot explain the Δ14C shift, for the reason that advection of a warm anomaly along isopycnal surfaces would not affect Δ14C values because Δ14C is a water mass tracer unaffected by temperature. Thus, our data do not support the Zhang et al. hypothesis.

An alternative hypothesis is that the source of the equatorial undercurrent changed and southern water became even more dominant after 1976. This would be expressed as a shift in both temperature and salinity, because South Pacific water from the thermocline is warmer and saltier than North Pacific water along a given isopycnal surface (9). Partial support for this hypothesis is given by limited hydrographic data (16) showing that the temperature shift in the eastern equatorial Pacific was accompanied by a slight increase in salinity. However, a shift in sources of the undercurrent cannot explain the rise in Δ14C values, because Southern Hemisphere water, in general, has lower Δ14C on a given isopycnal surface relative to Northern Hemisphere water (6,10).

Future research should explain why the shifts in temperature and Δ14C appear as step functions rather than as gradual changes. Earlier step-like transitions are also seen in the SST record (for example, in 1955–1956), but it is not clear if these values are authentic or are an artifact of changes in measurement technique or of scarcity of observations (13). Although nonlinear systems exhibit such behavior, there is no existing theory that explains why the processes affecting the vertical structure of the tropical pycnocline should respond in a step-like fashion. Before continuous subsurface monitoring of temperature and salinity in the 1980s, hydrographic data were too sparse to explore this question in detail. It may be possible to obtain data from carbonate-secreting organisms that grow at thermocline depths (for example, sclerosponges and ahermatypic corals) to provide additional constraints on what happened in the ocean in 1976. Additional radiocarbon data on Galápagos corals from times in the past 100 years when the frequency of El Niño is known to have waxed and waned could be used to determine whether similar shifts in thermocline structure occurred at these intervals and whether the “1976 climate shift” was truly unusual.


View Abstract

Stay Connected to Science

Navigate This Article