Notice

The evolution of ENSO and global atmospheric temperatures

Kevin E. Trenberth, Julie M. Caron, David P. Stepaniak and Steve Worley


National Center for Atmospheric Research
P. O. Box 3000
Boulder, CO 80307


email: trenbert@ucar.edu
(303) 497 1318; (303) 497 1333 (FAX)

December 20, 2000


J. Geophys. Res. Atmospheres

Abstract

The origins of the mini global warming accompanying El Niño events and the implications for the role of diabatic processes in El Niño-Southern Oscillation (ENSO) are explored. The evolution of global mean surface temperatures, zonal means and fields of sea surface temperatures, land surface temperatures, precipitation, outgoing longwave radiation, vertically integrated diabatic heating and divergence of atmospheric energy transports, and ocean heat content in the Pacific are documented using correlation and regression analysis. For 1950-98, ENSO linearly accounts for 0.06$^\circ$C of global warming. Individual warming events peak 3 months after SSTs in the 34 region, somewhat less than found in previous studies. During and following El Niño events the heat from the ocean is redistributed within the tropical Pacific. Warming at the surface progressively extends to about $\pm 30$$^\circ$ latitude with lags of several months. A build up of ocean heat content in the equatorial west Pacific progresses eastward and peaks some months before the surface event, before spreading polewards along the coast of the Americas and westwards in the off-equatorial tropics. Only in the equatorial region are the subsurface ocean heat content anomalies linked to SST anomalies locally. While the progressive development of ocean heat content anomalies resembles that of the delayed oscillator paradigm, the damping of anomalies through heat fluxes into the atmosphere introduces a substantial diabatic component to the discharge and recharge of the ocean heat content. This contributes to the delayed atmospheric warming in the tropics and subtropics as El Niño decays, but much of the delayed warming outside of the tropical Pacific comes from persistent changes in atmospheric circulation forced from the tropical Pacific. A major part of the ocean heat loss to the atmosphere is through evaporation and the heat is realized in the atmosphere as latent heating in precipitation. This diabatic heating drives large-scale overturning that influences the response throughout the tropics and subtropics and sets up teleconnections in the extratropics. Reduced precipitation and increased solar radiation in Australia, southeast Asia, parts of Africa and northern South America contribute to surface warming that peaks several months after the El Niño event. Teleconnections contribute to the extensive warming over Alaska and western Canada through a deeper Aleutian low and stronger southerly flow into these regions 0 to 12 months later. Because the temperature response is greater over land than ocean opposite changes contribute to the overall mean. The 1976/77 climate shift and the effects of two major volcanic eruptions in the past two decades are reflected in different evolution of ENSO events. At the surface for 1979-98, the warming in the central equatorial Pacific develops from the west and progresses eastward, while for 1950-78, the anomalous warming begins along the coast of South America and spreads westward. The eastern Pacific south of the equator warms 4 to 8 months later for 1979-98 but cools from 1950-78.


1. Introduction

It has been suggested that the time scale of El Niño-Southern Oscillation (ENSO) is determined by the time required for an accumulation of warm water in the tropics to essentially recharge the system, plus the time for the El Niño itself to evolve [ Wyrtki, 1985]. However, it has not been clear how much of the exchange of heat in the equatorial Pacific Ocean is merely with other parts of the ocean, and thus is adiabatic, as is the case for the so-called delayed oscillator paradigm for ENSO [ Suarez and Schopf, 1988; Jin, 1997], versus heat exchanges with the atmosphere involving diabatic processes.

During and following an El Niño, the global surface air temperature typically warms up by perhaps 0.1$^\circ$C with a lag of about 6 months [ Newell and Weare, 1976; Pan and Oort, 1983; Jones, 1989; Wigley and Santer, 2000]. In an exceptional event such as the 1997-98 El Niño the amount exceeds 0.2$^\circ$C. Christy and McNider [1994] and Angell [2000] show that the entire troposphere warms up with an overall lag of 5 to 6 months, but the lag is slightly less in the tropics and greater at higher latitudes. Consequently, the empirical evidence suggests a strong diabatic component to ENSO. Either the atmospheric circulation and cloudiness change with ENSO in such a way to allow this to happen directly within the atmosphere, or following El Niño, the ocean gives up heat to the atmosphere to produce the delayed warming. In fact both seem to occur. The latter appears to be more likely in the tropics and subtropics, and was confirmed for a limited period by Sun and Trenberth [1998] and Sun [2000]. However, in the extratropics of the Northern Hemisphere, the deeper Aleutian low that accompanies El Niño, advects warm moist air along the west coast of North America bringing warmth to western Canada and Alaska [ Trenberth and Hurrell, 1994]. Similarly, in most parts of the tropics, through teleconnections and because the main precipitation shifts over the central Pacific Ocean, drier and sunnier conditions favor higher temperatures with El Niño [ Klein et al., 1999]. These examples suggest that several mechanisms need to be considered to explain the mini global warming following El Niño. Hence a primary purpose of this paper is to clarify how the global atmospheric surface temperatures respond to ENSO and illuminate the global heat budget associated with ENSO.

The sea surface temperatures (SSTs) are a key in the two-way communication between the atmosphere and ocean but have to be supported by a substantial heat content anomaly of the ocean mixed layer if they are to have a continuing influence on the atmosphere. Such an influence typically means that the anomalous heat is being drained from the ocean and thus a negative feedback occurs, as seems to be the case generally in the tropics. This was shown from observations by Trenberth et al. [2001b], who analyzed a new dataset of the divergence of the vertically integrated energy transports in the atmosphere, along with SSTs, precipitation, atmospheric diabatic heating, and other fields to demonstrate how the atmosphere and ocean respond to ENSO, the dominant large-scale coherent coupled mode of variability. ENSO was dominant in the first two modes emerging from a singular value decomposition (SVD) analysis of the temporal covariance of SST with the atmospheric variables, explaining 62% and 12% of the covariance in the Pacific domain and 39.5% and 15.4% globally, respectively. In the tropical Pacific during major El Niño events, the anomalies in divergence of the atmospheric energy transports exceed 50 W m$^{-2}$over broad regions and primarily come from the surface fluxes from the ocean to the atmosphere. High SSTs associated with warm ENSO events are damped through surface heat fluxes into the atmosphere which fuel teleconnections and atmospheric circulation changes that transport the energy into higher latitudes and throughout the tropics, contributing to loss of heat by the ocean, while the cold ENSO events correspond to a recharge phase as heat enters the ocean.

The two coupled atmosphere-ocean modes can be thought of as reflecting different aspects of the mean tropical Pacific SST characteristics. The first measures the mean SSTs in the tropical central and eastern Pacific and is best represented by SST anomalies in the Niño 3.4 region (170-120$^\circ$W, 5$^\circ$N to 5$^\circ$S); we refer to this index as N34 to distinguish the SST index from area averages of other quantities over the same region. The second measures the contrast in SST differences across the Pacific from about the dateline to coastal South America and can be represented by the normalized difference in SSTs in regions Niño 1+2 (0-10$^\circ$S, 90-80$^\circ$W) minus Niño 4 (5$^\circ$N-5$^\circ$S, 160$^\circ$E-150$^\circ$W) that we call the Trans-Niño Index, TNI [ Trenberth and Stepaniak, 2001]. These two indices are roughly orthogonal at zero lag, but have strong relationships that change with time at various leads and lags up to a year. The evolution of ENSO and the implicit lag relationships manifested in the N34 and TNI indices [ Trenberth et al., 2001b; Trenberth and Stepaniak, 2001] suggest that a systematic exploration of lead and lag relationships is warranted, and that the N34 index can be used as the key index.

The evolution of SST patterns with ENSO and its changes over time are highly relevant to the question of how heat is redistributed, and is explored further in this paper. Several of our datasets, especially those dependent on satellite information, begin in 1979. Accordingly, we provide the most complete analysis for the post-1979 period. But we also wish to know the extent to which our results apply more generally in previous periods. Therefore an important factor taken into account in our analysis is the 1976-77 climate shift of ENSO activity toward more warm phases after about 1976 [ Trenberth, 1990], which is very unusual given the record of previous 100 years [ Trenberth and Hoar, 1996, 1997; Urban et al., 2000], and has been linked to decadal changes in climate throughout the Pacific basin [ Trenberth and Hurrell, 1994; Graham, 1994] and changes in evolution of ENSO. Rasmusson and Carpenter [1982] presented composites of the evolution of ENSO events for 6 warm ENSO events from 1951 to 1972 and showed the ``antecedent'', ``onset'', ``peak'', ``transition'' and ``mature'' phases of the composite event. These run from September in the year before to January the year following the event and became known as the ``canonical'' El Niño, see Wallace et al. [1998] for the historical discussion. Before 1976, ENSO events began along the west coast of South America ( TNI positive) and developed westwards. However, after 1977 the warming has developed from the west so that TNI with reversed sign prevailed some 3 to 12 months before the main peak in N34  and was followed by TNI itself some 3 to 12 months after the peak. Therefore, the evolution of ENSO events changed abruptly about 1976/77 [ Wang, 1995; An and Wang, 2000; Trenberth and Stepaniak, 2001]. We take 1979-98 as representative of the post 1976/77 shift period. As well as the climate shift, another reason why the more recent period may be anomalous in ENSO evolution, especially with regard to effects on surface temperatures, is the presence of two strong volcanic eruptions (El Chichon in April 1982 and Pinatubo in June 1991). Wigley and Santer [2000] estimate a global mean cooling from the two volcanoes peaking at $-$0.2$^\circ$C for El Chichon and $-$0.5$^\circ$C for Pinatubo some 30 months after the eruption. Hence, we also explore the relationships for the 1950 to 1978 interval, whenever adequate data are available, as a way to examine the impact of the 1976/77 shift and volcanic influences and determine to the extent possible how reproducible the results are.

In this paper, we are especially interested in what can be determined with regard to the major interannual variations in the tropical Pacific and how they evolve with time, as this will help determine the role of El Niño in the climate system. Often El Niño has been regarded as simply a superposed fluctuation which undergoes its own cycle driven by either a coupled ocean-atmosphere instability within the tropical Pacific or stochastic forcing (and the Madden-Julian intraseasonal oscillation in particular) [ Power and Kleeman, 1994; Flügel and Chang, 1996; Eckert and Latif, 1997; Blanke et al., 1997; Stone et al., 1998; and Moore and Kleeman, 1999]. The combination of the tropical air-sea instability and the delayed negative feedback due to subsurface ocean dynamics can give rise to oscillations [ Suarez and Schopf, 1988; Battisti and Hirst, 1989; Münnich et al., 1991; Tziperman et al., 1994; Neelin and Jin, 1993; Jin, 1997]. As noted above, some theories of El Niño based on the delayed oscillator paradigm move heat out of the equatorial region during El Niño but move heat back as part of the overall ENSO cycle and the process is essentially adiabatic. The same is true for the Cane-Zebiak model [ Cane and Zebiak, 1985]. Yet during the course of these changes, the amount of warm water in the tropical Pacific builds up prior to and is then depleted during ENSO (see also Wyrtki, 1985; Cane and Zebiak, 1985; Jin, 1997; Tourre and White, 1995; Giese and Carton, 1999; Meinen and McPhaden, 2000). Observed changes in subsurface ocean temperatures [ Zhang and Levitus, 1996, 1997] and sea level [ Smith, 2000] and how they evolve with ENSO support this view. We show that part of the ocean heat content buildup and depletion is through exchanges of heat with the atmosphere and involve diabatic processes.

Therefore we exploit new data sets to help deduce and clarify what can be said about the diabatic processes involved in ENSO, how they relate to SST variations and those of the subsurface ocean, and how the mini-global warming following El Niño arises when it appears that it can not be sustained by the atmosphere alone. Tropical precipitation variations are also explored as a key indicator of the latent heating of the atmosphere, and inferences are made about effects of changes in cloudiness and increased solar radiation, and surface wetness on sensible versus latent heating.

Section 2 outlines the data sets used and their processing. The presentation of results in section 3 proceeds from the relationships between ENSO and global mean temperature, to zonal means of various quantities, and subsequently to full geographic spatial structure, so that we can trace where the relationships are coming from. Results for 1950-78 are presented along with those for 1979-98 where available. In discussing the results in section 4, the relative roles of the links with ENSO through the Pacific Ocean and the teleconnections through the atmosphere are addressed. The conclusions are given in section 5.


2. Data and methods

Most of the in-depth analysis here is for the period 1979 to 1998, as this is the time when high quality atmospheric reanalyses are available from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR), and global fields of precipitation and outgoing longwave radiation (OLR) are also available. Prior to 1979, the absence of satellite data adversely affects the quality of the reanalyses. For the period 1979 to 1998 we computed many quantities for each month including the vertically integrated total atmospheric energy transports ${\bf F_A}$ and their divergence $\nabla \cdot {\bf F_A}$ and the vertically integrated diabatic heating [ Trenberth et al., 2001a]. The total energy consists of the potential and internal energy, the latent energy, and the kinetic energy, while the transports also include a pressure-work term and can be broken down into components from the dry static energy and the moist (or latent) component, which together make up the moist static energy, plus the kinetic energy. The energy tendencies are combined with the computed divergence of the vertically integrated atmospheric energy transports to give the net column change, which has to be balanced by the top-of-the-atmosphere (TOA) radiation and/or the surface fluxes, see Trenberth and Solomon [1994] and Trenberth et al. [2001a] for details. Thus, ignoring the tendency terms, which average to be very small over a few months,

\begin{displaymath}\nabla \cdot {\bf F}_A = R_T + F_s
\eqno (1)\end{displaymath}

where $R_T$ is the net downwards top-of-atmosphere radiation and $F_s$ is the upwards surface flux. The diabatic heating is sometimes called $Q_1$, and, ignoring a tiny frictional heating component ($Q_f$), in vertically integrated form is given by

\begin{displaymath}
Q_1 = R_T +F_s +L(P-E) \eqno (2)
\end{displaymath}

where $P$ is the precipitation and $E$ the evaporation in the column, and $L$ iis the latent heat of vaporization. The term $LE$ cancels a part of $F_s$ leaving the surface radiation and sensible heat flux terms. Therefore, $Q_1$ is not a driver of changes in moist static energy or total energy (cf. (1)) because changes in the state of moisture are internal and simply change the dry static energy at the expense of the moisture content. For this reason there is a strong compensation between the dry static energy and the latent component which occurs as the convergence of moisture in the low levels is realized as latent heating.

Trenberth et al. [2001a] show that the variability of TOA fluxes is small and so $\nabla \cdot {\bf F_A}$ is mainly balanced by the surface fluxes. Trenberth et al. [2001b] exploit the large spatial and temporal scales of ENSO to bring out the ENSO signal from the noise. Over the Niño 3.4 region, results imply a random standard error of $\sim$6 W m$^{-2}$and suggests that signals greater than 12 W m$^{-2}$are significant.

We also make use of the OLR data set from the NOAA series of satellites adjusted using results from Waliser and Zhou [1997], as in Trenberth et al. [2001b]. We utilize the precipitation dataset from Xie and Arkin [1996, 1997], called the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP). Over land these fields are mainly based on information from rain-gauge observations, while over the ocean they primarily use satellite estimates made with several different algorithms based on OLR, and scattering and emission of microwave radiation.

We use the SSTs from NCEP from the optimal interpolation (OI) SST analysis of Reynolds and Smith [1994] after 1982 and the empirical orthogonal function (EOF) reconstructed SST analysis of Smith et al.  [1996] for the period before then. The latter does not contain anomalies south of 40$^\circ$S. However, these SSTs are preferred to those in the global surface temperature dataset from the University of East Anglia (UEA) and the United Kingdom Meteorological Office [ Hurrell and Trenberth, 1999], although the latter is employed to examine values over land and to examine the longer record and the global means.

To examine aspects of the subsurface ocean heat content changes, we have used the ocean analyses of the tropical Pacific from the Environmental Modeling Center at NCEP. These model-based analyses have developed over time and currently assimilate observed surface and subsurface ocean temperatures, as well as satellite altimetery sea-level data from TOPEX/POSEIDEN. We use a monthly mean analysis from 1980 to 1998 derived from weekly analyses using the RA6 schemes described by Behringer et al. [1998].

The RA6 wind forcing and surface heat flux have changed over time. For 1980-96 wind forcing was the Hellerman and Rosenstein [1983] climatological wind stress combined with monthly pseudostress fields from Florida State University. NCEP operational near-surface winds, four times daily, with a constant drag coefficient were used to determine wind stress for 1997 through mid 1998. From mid-1998, wind stress is determined directly from NCEP operational analysis. Transients resulting from the wind forcing changes are minimized by gradually applying a new forcing over six months. Surface heat flux for 1980-96 was the mean annual climatological cycle of Oberhuber [1988]. From 1997 onward the heat flux is taken from the NCEP operational analysis. The impact of these changes has not been quantified.

The heat content is computed using data for the upper 387.5 m (using model levels down to and including 345m but excluding levels at 430 m and below), based on the fact that most expendable bathythermographs sample this layer but not deeper and there is inadequate information at greater depths to determine interannual variability. However, it is apparent that the main variability is above 300 m depth and contributions from deeper layers are believed to be very small. We compute the heat content $H$ as

\begin{displaymath}
H = \sum_{i} \rho C_p T_i \delta z_i
\eqno (3)
\end{displaymath}

over the available layers $i$. In computing the heat content, given that the maximum variability in temperature is in the thermocline and we are interested primarily in the tropics and subtropics of the Pacific, we have selected constants of 1025 kg m$^{-3}$ for density $\rho$ and 3990 J kg$^{-1}$ K$^{-1}$ for $C_p$ corresponding approximately to a temperature of 20$^\circ$C, salinity of 35 to 35.5 per mil and 100 to 150 m depth, although because $C_p$ increases with temperature while density decreases, the product $\rho C_p$ is representative of a much broader range of values.

Because large natural variability on synoptic timescales appears as weather noise in monthly means, and also spurious noise related to sampling is present (especially for SSTs, OLR, subsurface ocean temperature, and $\nabla \cdot {\bf F_A}$), we have smoothed the monthly anomaly fields used in the analyses with a binomial ${1\over 4} (1, 2, 1)$ filter which removes two month fluctuations. We primarily employ correlation and regression analysis to bring out relationships at various lags.


3. The tropical Pacific variability

3.1 Time series in the tropical Pacific

As noted in the introduction, we use N34 as a key index. For correlations between variables with persistence related to ENSO [ Trenberth, 1984] for the 240 months (1979 to 1998), there are about 80 degrees of freedom, suggesting statistical significance at the 5% level if the correlations exceed about 0.23. The N34 and TNI indices are given in Fig. 1 normalized using the means and standard deviations for 1950 to 1979. Both the time series and the spatial patterns of N34 and TNI are essentially orthogonal. Cross correlations of N34 and the Southern Oscillation index (SOI) based upon surface pressures at Tahiti and Darwin are a maximum at zero lag ($-$0.83) for 1950-98 and for the two subperiods. Trenberth et al. [2001b] presented the regression of N34  with the global SST field (see Fig. 9, later), and the correlation is also given later as a function of lag in Fig. 8. Patterns of SST associated with TNI are given in Trenberth and Stepaniak [2001] along with the relations of TNI and N34 with lag. Figure 1 also presents the means before and after 1976/77 for N34 and the global mean temperature to illustrate the shift that occurred in these indices toward warmer and more El Niño-like conditions.

3.2 Links with global mean temperature

For the global mean temperatures for 1979-98 (Fig. 2), the cross correlations (Fig. 3) reveal a broad maximum with the global mean temperature lagging N34  by 4 months. This is less than found in several other studies for reasons that are not fully clear. For instance Wigley and Santer [2000] find a lag of 6 months for 1979-99 with an SOI, although their SOI is formulated differently and is not optimal [ Trenberth, 1984]. They attempt to remove the cooling effects of the two volcanic eruptions (El Chichon and Pinatubo) after which their cross correlation sharpens and increases in magnitude to about 0.7 with a 7 month lag. Their results highlight the convolution of the volcanic signal and ENSO event during this period with a blurring of the ENSO-related relationships as a consequence. This result reinforces the need in Fig. 3 and subsequent analysis to also explore the relationships for the 1950 to 1978 period.

Figure 3 shows that the lag correlations are higher and more sharply peaked in the 1950-78 period than more recently, strongly suggesting either a contaminating influence from the two major volcanic eruptions or an impact of the 1976/77 climate shift, or both. The lag is sharply defined as 3 months and with a peak correlation of 0.65, corresponding to a regression of 0.11$^\circ$ C per N34, compared with 0.08$^\circ$C for 1979-98.

The 1976/77 climate shift, whether part of decadal variability or trend, influences results when the two subperiods are combined. Therefore it is worthwhile to include results for the period (1950-98), as the different means in the two subperiods factor into the results. The correlations for the entire period are intermediate between those of the two subperiods except positive correlations last longer ($-$10 to +16 months) and are centered at +3 months lag. Negative correlations before and after, which signal the ENSO quasiperiodicity, are less in evidence on the longer time scale, no doubt reflecting the influence of the 1976/77 climate shift. Alternatively, there is a significant linear trend to the global mean temperatures which accounts for 41% of the total variance for 1950 to 1998.

Finally in this subsection, we examine how much variance of the global mean temperature is accounted for by ENSO for the entire 1950-98 period using a screening regression, and what the residual series looks like (Fig. 2). For the entire 1950-98 period the maximum lag correlation of 0.53 (28% of the variance) between N34 and global temperature is at +3 months. Surprisingly, this is identical whether or not the two series are detrended. The regression coefficient based on the detrended relationship is 0.094$^\circ$C per N34. This is deemed the more appropriate one and the N34 contribution is given in Fig. 2. It shows that for the 1997-98 El Niño, where N34 peaked at about 2.5$^\circ$C, the global mean temperature was elevated as much 0.24$^\circ$C (Fig. 2) although, averaged over the year centered on March 1998, the value drops to about 0.17$^\circ$C. Correlations of the residual with TNI are not significant. The linear trend in N34 over 1950-98 accounts for 4.1% of the N34 variance, and this accounts for 1.8% of the variance due to the trend in global mean temperature. As this is based on the regressions using detrended data, the relationship is determined by the interannual variations. It means that 13.6% ($= 0.018^{0.5}$) of the linear trend in surface temperature, or 0.06$^\circ$C for 1950 to 1998, arises from the changes in ENSO (out of 0.43$^\circ$C).

3.3 Zonal mean results

Figure 4 presents cross correlations with N34  of zonal means at leads and lags as a function of latitude. Positive values always refer to N34  leading. The top panel shows results for 1979-98, the middle panel is for 1950 to 1978, and bottom panel is for the full period 1950-98. For SSTs (first column) the figure shows the maximum correlations exceeding 0.8 lasting from 2 months before to 3.5 months after N34, and reversing in sign a year before and a year later for 1979-98. Correlations outside about $\pm$ 30$^\circ$ are not significant. However, of considerable interest are the ``wings'' or lobes of positive correlations extending from the equatorial region into the subtropics a year later. For 1950-78 the leads and lags in the equatorial region are more symmetric although the wings into the subtropics are still present, and this pattern also exists for the entire period, without diminution of the correlations, and thus there is an enhancement in statistical significance.

The equivalent results for the full surface temperatures (not shown), which thus include land portions, are quite similar overall, although the central highest correlations are delayed by 2 months in the equatorial region and positive values are slightly higher in the subtropics a year later. As with SST, there is no evidence of any relationship extending to higher latitudes, although such has been claimed by Angell [2000]. Therefore the second column of panels (Fig. 4) shows the results just for land. Over land there is less evidence for delays with increasing latitude, seen over the ocean, but the whole pattern is delayed by 3 to 4 months, and the extension is to slightly higher latitudes, especially in the Northern Hemisphere where correlations 45-50$^\circ$N exceed 0.2 at 0 to 7 months lag, and is barely 0.2 at 65$^\circ$N at 13 months lag. This fits better with Angell's result and indicates that his result is probably biased by the distribution of stations used.

Shown in Fig. 5 are three panels for 1979-1998 giving results for OLR, precipitation and $\nabla \cdot {\bf F_A}$ relations with N34  for the ocean only. Because the main precipitation algorithm over the oceans uses OLR -- low OLR from high cloud tops corresponds to high convective rainfall -- very similar results to precipitation come from OLR. We have examined the same results for just the Pacific Ocean and also for the globe (including land). For $\nabla \cdot {\bf F_A}$  the land values are unreliable and not useful [ Trenberth et al., 2001a], while the main relationship over the oceans clearly comes from the Pacific, where the correlations exceed 0.8 just south of the equator near lag 0, but with a pattern quite similar to that in Fig. 4. For precipitation, the patterns for the Pacific versus total ocean versus global zonal means (not shown) are quite similar, but with values dropping in the near equatorial positive correlation region, while the negative correlations in the subtropics hold steady in all three domains. For both precipitation and $\nabla \cdot {\bf F_A}$, the lagged relationship indicated by the wing into the Northern Hemisphere is absent and only the wing in the Southern Hemisphere is present. As seen later, this wing arises from the changes in the South Pacific Convergence Zone (SPCZ).

Fig. 6 shows similar figures for the oceans for several other quantities, the vertically-integrated diabatic atmospheric diabatic heating ($Q_1 - Q_f$) and $L(P-E)$ from the moisture budget [ Trenberth et al., 2001a], and the ocean heat content in the Pacific. The patterns are almost identical for the land plus oceans combined. The $L(P-E)$ relationship is similar to but much weaker than that for precipitation alone. The diabatic heating, computed from the thermodynamic equation as a residual, is quite similar to that for precipitation, as would be expected if latent heating is dominant, although the fact that it is slightly weaker is probably a deficiency in the NCEP reanalyses [ Trenberth and Guillemot, 1998].

The correlations with Pacific upper ocean heat content, on the other hand, are quite different in character. Fig. 6 reveals strong correlations, indicating the build up in equatorial Pacific heat from 12 to 6 months prior to the peak in N34, and with the maximum correlation occurring 4 months prior to N34. Meanwhile there is cooling from 5 to 20$^\circ$N that is a maximum at 8$^\circ$N 2 months before N34. Significant cooling also follows the peak in the equatorial region culminating about 12 months after N34, but the off-equatorial correlations are not as strong, suggesting some loss of signal through diabatic effects. This pattern is consistent with that depicted by Meinen and McPhaden [2000].


3.4 Evolution of spatial patterns

Fig. 7 shows a breakdown of Fig. 3 by ocean sector for 1950-78 and 1979-98. It seems obvious that the lag of surface temperatures in the Pacific should be closely in phase with N34  because of the close proximity, and indeed this is the case, with a 2 month lag in both subperiods. However, peak correlations are almost double in the earlier period. The Atlantic sector (defined as 90$^\circ$W-0$^\circ$) lags by 4-5 months, while the Indian Ocean sector (defined as 0-120$^\circ$E) shows a lag of 7 months for 1979-98 but much less lag, although skewed and centered around +5 months from 1950-78. The corresponding regressions have peak values of 0.10$^\circ$C in the Indian and Atlantic sectors, versus 0.06$^\circ$ C for the Pacific for 1979-98 and 0.12-0.13$^\circ$C for all three oceans for 1950-79. These results highlight the need to examine the redistribution of heat spatially.

To explore where the heat in the oceans and its effects on the atmosphere actually occur, we have computed correlation and regression maps for several fields at various lags, a subset of which are given here. Generally the regression maps are quite similar in overall patterns but the correlations are presented as they allow significance to be roughly assessed. Figure 8 presents the time sequence for correlations with N34  of surface temperatures (based on the NCEP SSTs over the ocean blended with the UEA land data) for 1979-1998 (right column) at $-$8, $-$4, 0, 4, and 8 months, where, again, positive values signify that N34  leads. This reveals the mean evolution of ENSO. The sequence is quite coherent and traceable from about $-$12 to +12 months lag with N34, but becomes quite weak at larger lags, and this $\Delta$4 month sequence is chosen as a compromise through consideration of the interval between panels and their number. Figure 8 (left column) presents the same sequence for surface temperatures from 1950-1978 to allow us to explore the role of the climate shift in 1976/77 and the possible effects of the two volcanoes in the more recent interval.

For 1979-98, the maximum warming in the central equatorial Pacific develops from off-equatorial tropical regions and from the west, progressing eastward. In contrast, for 1950-78, the anomalous warming begins strongly along the coast of South America and appears to spread westward, as was found by Rasmusson and Carpenter [1982]. This difference in development continues, with the eastern Pacific south of the equator warming during the +4 to +8 months for 1979-98 but cooling from 1950-78. In the Indian Ocean, warming begins 4 months prior to the peak in N34  and is strongest from 0 to +4 months for 1950-78 and at +4 months for 1979-98, and is still strongly in evidence at +8 months. The Atlantic signal becomes strongest about +4 months and is quite similar in the two periods, with warming mainly from 0 to 25$^\circ$ N, but also some warming about 20$^\circ$S. These results are consistent with those of Klein et al. [1999] for the tropical Atlantic and Indian oceans.

Figure 9 (left) presents the same evolution but as a regression for the entire period (1950-1998). It therefore highlights the surface temperature signal that is the same between the two subperiods and converts the signal into actual temperature anomalies, thereby sharpening the focus on the tropical Pacific and extratropical North America, where the variance is greater than over the rest of the oceans. Referring to the warm ENSO phase, at lag 0, the warmth in the tropical central and eastern Pacific is accompanied by substantial warmth along the west coast of the Americas, especially in the western two thirds of Canada and Alaska, and is mirrored to some extent in the South Pacific near 50-60$^\circ$S. Modest but widespread warmth is also in evidence in the Indian Ocean, Africa and southern Asia, and in bands from South America across the Atlantic near 20$^\circ$N and 30$^\circ$S. The effects in the Atlantic and Indian Ocean sectors become strongest at about +4 months and are still evident at +8 months. Australia also tends to be warm from 0 to +8 months. The latter relates to the relatively dry, or even drought, conditions during El Niño and so is at least partly a result of changes in cloudiness and atmospheric circulation, and not just heat from the ocean [ Klein et al., 1999]. Similarly, drier conditions over southeast Asia (mainly prior to 1977, Kumar et al., 1999; see also the differences in Fig. 8) and parts of Africa imply warmer temperatures in the summer half year, as shown next.

Figure 9 (right) therefore also presents the regression sequence of N34 with precipitation, but only for 1979-98. The pattern is, not surprisingly, almost identical with OLR (not shown), except the units are different. Peak values for one standard deviation of N34 of order 1 mm/day in precipitation correspond to about 10 W m$^{-2}$in OLR and correlations over 0.5 in magnitude. The strong east-west dipole/boomerang pattern in ENSO is familiar and has very large cancellation when compiled into the zonal means featured in the previous section. Because much of these patterns arise from movements northeastward of the SPCZ and southward of the Intertropical Convergence Zone (ITCZ), the dividing lines are sharp and do not match the SST changes, although there is a general tendency for wetter and warmer conditions to co-exist in the tropical oceans. Over land the relationship is rather different, and warmer conditions more often than not go hand-in-hand with drier conditions, symptomatic of the changes in monsoons that occur with ENSO.

The relationships with the inferred vertically integrated diabatic heating of the atmosphere and $\nabla \cdot {\bf F_A}$ (Fig. 10) are similar to those at zero lag, given in Trenberth et al. [2001b]. The diabatic heating pattern tends to be similar to that of precipitation, signifying the dominance of the latent heating anomalies. It is also closely related to anomalies in mean vertical motion in the mid-troposphere (not shown). The positive correlations between $\nabla \cdot {\bf F_A}$ and SST anomalies in the Pacific are reversed in middle and higher latitudes and in the tropical Atlantic. Although a clear sequence of $\nabla \cdot {\bf F_A}$ anomalies can be discerned in Fig. 10, it is not as strong as with some other fields and the pattern is strongest at 0 and +4 months, suggesting that it is mainly a response to ENSO.

The evolution of the Pacific Ocean heat content anomalies (Fig. 11) shows patterns that have been partly documented elsewhere, with the buildup of heat in the western Pacific before $-$8 months, progressing eastwards to reach the west coast of the Americas before lag 0, subsequently spreading polewards into both hemispheres (+4) and then westwards from 10 to 20$^\circ$N at +8 months while building up near 10$^\circ$S but with little spread westward beyond about 150$^\circ$W, near the location of the SPCZ. Figure 12 extends the time sequence by focusing on what appear to be fairly coherent regional averages. This shows that the buildup of heat in the western Pacific, west of 150$^\circ$W, first occurs off the equator ($-$20 to $-$18 months), most strongly south of 8$^\circ$S. It progresses into the equatorial region (8$^\circ$N to 8$^\circ$S), peaking at about $-$15 months. The equatorial and southern region west of 150$^\circ$W becomes negative at $-$3 and $-$5 months, and the maximum negative correlation is at +4 or +5 months. Meanwhile the initial heat deficit from 8$^\circ$N to 30$^\circ$N, which peaks at $-$5 months, becomes positive at +4 months and peaks at +14 months. The maximum heat content from 150$^\circ$W to the Americas in the $\pm 8$$^\circ$ strip coincides with N34  and drops to zero at +9 months. From 8-30$^\circ$S, east of 150$^\circ$W, the warming occurs from $-$7 to +18 months, peaking at +5 months.

While the changes in Pacific heat content lead and appear to determine the SST evolution in the equatorial strip, this does not seem to be the case at higher latitudes. This is shown by the local correlation between the monthly anomalies of ocean heat content and the SST (Fig. 14), which has direct implications for the relevance of a simple two layer ocean model in being able to depict SST variations with variations in the local thermocline. Only in the equatorial region, broadening from 20$^\circ$N to 10$^\circ$S west of 170$^\circ$E and 10$^\circ$N to 15$^\circ$S along the Americas, is the correlation significantly positive. Values are actually negative along most of the Pacific at 10-15$^\circ$ N, north of the ITCZ, and also just northeast of the SPCZ region from 10$^\circ$S 160$^\circ$W to 35$^\circ$S 100$^\circ$W. Giese and Carton [1999] show a somewhat similar figure based on model assimilated data. This result directly reflects the ability of the subsurface ocean heat to sustainably influence the atmosphere.


4. Discussion

4.1 Role of the tropical Pacific Ocean

In the tropical Pacific, and the tropical Indian Ocean to some extent [ Tourre and White, 1995], heat loss from the ocean drives the atmospheric circulation changes, and the redistribution of heat by ocean currents as ENSO evolves, contributes directly to subsequent heating and increased temperatures in the tropics and subtropics as El Niño fades. Magnitudes of observed changes in surface fluxes with moderate El Niños are about 40 W m$^{-2}$ [ Trenberth et al., 2001a] which would correspond to a 1$^\circ$C temperature change over a column of water of 130 m deep over 5 months. Zonal means across the Pacific of the ocean heat content (not shown) reveal typical anomalies of order 0.5$\times 10^9$ J m$^{-2}$ over 5$^\circ$ latitude which corresponds to a 1$^\circ$C change over a layer of 122 m. Maximum anomalies in 1981-83 and 1997-98 are about 3 times this value but these are the main times when there is also a meridional dipole of anomalies of opposite sign present, signifying a significant meridional rearrangement of heat within the ocean. Consequently, the estimated surface heat fluxes are a dominant factor in the change in area-average subsurface ocean heat content and the heat shows up in the atmosphere.

The evolution of ENSO in the tropical Pacific illustrated here supports much of that previously described by Barnett et al. [1991], Zhang and Levitus [1996], Tourre and White [1995], Giese and Carton [1999], Smith [2000], and Meinen and McPhaden [2000] in the way that anomalies of subsurface ocean heat content in the western Pacific develop as they progress eastward across the equatorial Pacific, often with a dipole pattern across the Pacific, and then with anomalies progressing off the equator to higher latitudes. Zhang and Levitus found links only to the North Pacific, perhaps reflecting the available data, while links are strong in our results into both hemispheres. The SST evolution lags somewhat behind that of the subsurface ocean and is damped by surface fluxes and transports out of the region by the atmosphere, emphasizing the dominant role of the surface wind stress and ocean dynamics and advection in producing the local ocean heat content and SST anomalies. This damping of the ocean signal, however, forces the atmospheric anomalies. Moreover, this aspect also emphasizes that in cold La Niña conditions, the surface fluxes of heat are going into the ocean relative to the mean and are warming the ocean, although not locally, as where the heat builds up is determined by the currents.

The evolution of the ocean heat content shows the eastwards progression in the equatorial region, then spreading north and southwards (presumably as Kelvin edge waves) along the west coast of the Americas to higher latitudes, but also with a component spreading westwards at about 10 to 20$^\circ$ latitude in both hemispheres. From 1979-98 the SSTs were strongly positive in the Southern Hemisphere portion but remained negative in the Northern Hemisphere 12 months after the N34  index peaks. The SST development from 1950-78 is rather different and results in strong positive anomalies in the Northern Hemisphere near 10 to 20$^\circ$N 12 months after the peak, and with some residual heat also in the Southern Hemisphere in a region where it can influence the SPCZ. The quite different evolution along the coast of South America in the classical El Niño region has previously been emphasized, with the warming at the surface often occurring first there before 1979, as in the Rasmusson and Carpenter [1982] composites, but after the central Pacific warming in more recent years [ Wang, 1995].

4.2 Role of the atmosphere

In dealing with a time mean heat budget over several months, tendencies are necessarily small and all the terms must be in balance. As a result, the cause of a temperature anomaly may not be easily deduced and the utility of quantities like $\nabla \cdot {\bf F_A}$are limited.

There is little to support the idea that the local ocean heat storage plays much role at higher latitudes. Instead, there is a known strong link through atmospheric teleconnections between ENSO events and changes elsewhere [ Trenberth et al., 1998]. Over the North Pacific, ENSO teleconnections are well captured at the surface by the index of area-average sea level pressure over the North Pacific, called the NP index [ Trenberth and Hurrell, 1994]. Cross correlations between N34 and NP (not shown) are negative for $-$5 to +10 months and peak at $-$0.20 at +2 months for 1979 to 1998. That relationship is strongest in the winter half year. Composites of seasonal anomalies of mean temperature, changes in storm tracks and tendencies in temperature from divergence of transient eddy heat transport [ Trenberth and Hurrell, 1994] show that the latter two are negatively spatially correlated and act to destroy the temperature structures in the lower troposphere. Instead the temperature perturbations are partly set up by advection by the anomalous mean flow, and this is demonstrated by regressing the 700 mb temperature and vector wind on N34 as a function of lag, with N34 leading by 0, 4 and 8 months (Fig. 14).

Thus, the substantial warming with El Niño over Alaska and western Canada is clearly related to the change in atmospheric circulation that is known to be forced through teleconnections from the tropical Pacific. The change that most obviously brings about the warming is warm advection, and this comes about through an anomalous southerly component to the flow advecting the mean temperature gradient. We split the temperature $T$ into the time mean, given by the overbar, and the anomaly perturbation, given by the prime $T = {\overline T}
+T'$, and similarly for other quantities. Thus it is the term $v'{\partial {\overline
T}\over {\partial y}}$ that gives rise to part of the $T'$ (see Fig. 14) while it is offset by changes in storm tracks and the transient eddy temperature flux that acts to destroy the temperature anomaly on about a two week time scale [ Trenberth and Hurrell, 1994]. These counterbalancing effects are both included within $\nabla \cdot {\bf F_A}$ and largely cancel. However, associated changes in moisture advection and cloudiness occur, introducing some radiative anomalies as well, and there is more to $T'$ than advection.

The North American warming at 700 mb extends eastwards from the maximum southerly flow to the upstream anticyclonic flow (see also Wallace et al., 1998). Over the North Pacific, Fig. 14 shows the link between the cooling and the more northerly flow but also to the cyclonic circulation. An approximate mirror image appears over the South Pacific, although the wave train has a slightly shorter wavelength. Cooler conditions east of New Zealand are associated with southwesterly flow anomalies around a cyclonic circulation centered near 40$^\circ$S 150$^\circ$W and a blocking high exists farther downstream near 60$^\circ$S 110$^\circ$W.

The more complete picture therefore, is that there is a substantial equivalent barotropic component to the atmospheric anomalies. In other words there is a substantial temperature anomaly in phase with the geopotential height field, and surface geopotential height or pressure anomalies increase in magnitude with height, although with some westward slope caused by the baroclinic component seen in the advection. Within the atmosphere, this comes about through the potential vorticity dynamics of teleconnections [ Trenberth et al., 1998] in which upper tropospheric transports of vorticity and momentum by transient eddies are important along with the heat transports. To keep the system in geostrophic and hydrostatic balance, subsidence occurs in anticyclonic regions, producing warming, while rising motion in cyclonic regions results in cooling. At the surface, cyclonic flow is accompanied by Ekman transports that produce upwelling and a shoaling thermocline, and increased storminess that encourages increased ocean mixing and fluxes of heat from the ocean into the atmosphere (e.g., see Deser et al., 1996). Thus the SST signature is similar to the tropospheric signature.

The remarkable persistence of the surface and 700 mb temperature anomalies extending from at least 0 through +8 months, and thus transitioning from winter to summer seasons, is worth noting. Fig. 9 shows that the tropical precipitation anomalies and thus anomalous diabatic heating and forcing of the atmosphere in the tropics continues throughout this period. There is some change in the anomalies with season, however; at +8 months the main contrast is from the Aleutian Islands region to North America, north of 50$^\circ$ N (Fig. 14). It is also of interest to compare the tropical precipitation (Fig. 9) and diabatic heating (Fig. 10) and the atmospheric response in Figs. 8 and 9 for $-$4 with +8 months, which are for the same season but a year apart. The differences in tropical regions are fairly subtle but evidently important. At $-$4 months the tropical Pacific anomalies are oriented more east-west while at +8 months they involve the SPCZ and the subtropical high near Hawaii is more involved in the tropical overturning. The response over North America is much greater in the latter case.

Hence the $T'$ at 700 mb is quite similar to that in Fig. 8 at the surface in the extratropics, except the cooling with El Niño over southern California at 700 mb is not present at the surface. The latter relates to the changes in storm tracks, cloudiness, and precipitation, and perhaps some direct influence of the local warming of the subsurface ocean (Fig. 11).

Some nonlinear effects that come into play depend on whether the temperature anomalies occur over the ocean or land. Over the ocean, the magnitude of that surface temperature change is muted by the heat capacity of the underlying ocean, the mixing of heat through the mixed layer, and possible entrainment and deepening of the thermocline. Consequently, the net surface warming over land is typically much larger than an equivalent response over the ocean (Fig. 9) and this can influence the global mean temperature, although it is not equivalent to a net heat content anomaly. The changes in the North Pacific and North America are mirrored to some extent over the southeast South Pacific as anomalous blocking takes place in that region as part of the Pacific-South American teleconnection but the land-ocean rectification effect is not present.

In the tropics and subtropics over land, the predominant change is one of warming over continental land (Africa, northern South America, Australia, southern Asia), which is associated with less precipitation, less high cloud (OLR), and more solar radiation [ Klein et al., 1999]. The dominant precipitation center moves over the central tropical Pacific, with a southwards shift in the ITCZ and a northeastwards shift in the SPCZ. The result is overturning teleconnections in the atmosphere that create the drier or even drought conditions in these locations. Other studies, such as Dai et al. [1999], which used detailed measurements from the First International Satellite Land Surface Climatology Project Field Experiment (FIFE), have documented the processes involved. They show how increased precipitation and surface moisture results in more evaporation and cooler surface temperatures, whereas decreased cloud increases temperatures mainly through increased solar radiation during the day, and there is an increase in temperature with poleward compared with equatorward winds. Therefore, the associations shown in this paper are consistent with the physical processes known to be operating.

The delayed warming in the tropical Indian and Atlantic Oceans is also a result of changes in atmospheric circulation [ Wallace et al., 1998; Klein et al., 1999]. In the Atlantic, the relationship of $\nabla \cdot {\bf F_A}$ (Fig. 11) with local SST anomalies is the opposite to that in the Pacific, showing warming of the Atlantic from the changes in the atmosphere heat budget with El Niño. In the tropical Indian Ocean, however, especially in the western part, the results are more like those in the Pacific, suggesting that the changes in surface wind stress are responsible for much of the surface temperature changes, but then there is a feedback and forcing of the atmosphere through local convection changes. This was especially noticeable in 1997-98, although claims have been made of independence with ENSO [ Webster et al., 1999; Saji et al., 1999].

These results highlight the fact that over most of the globe away from the tropical Pacific, it is the persistent change in atmospheric circulation, driven from the Pacific that results in the changes in temperature.

4.3 Individual events

To reveal how the correlations presented in the previous figures have eventuated from individual ENSO events, we have examined the latitude-time Hovmöller diagrams for several quantities (not shown, but see Wallace et al., 1998). While these show to some extent how the individual events contribute to the statistical relationships explored earlier, they also show considerable variability from event to event. Moreover, detailed comparisons of the evolution of each event suggest that the time scale is not necessarily fixed and the evolution in some events is stretched out relative to others, making relationships at fixed lags or leads not necessarily the best way to examine the evolution. Whether external influences such as the effects of volcanic eruptions is a factor in this or whether it relates more to the amplitude of each event and the way in which transients perhaps trigger and contribute to the development remain outstanding questions.


5. Conclusions

ENSO events contribute to coherent interannual and even decadal fluctuations in the global mean temperature and, as we have shown here, the nature of the ENSO contribution is quite complex. Part of it involves the recharge and discharge of heat from the tropical Pacific Ocean. During and following El Niño events the heat from the ocean is redistributed within the tropical Pacific and much of it is released to the atmosphere, creating local warming. However, a major part of the ocean heat loss is through evaporation, and the heat is realized in the atmosphere as latent heating in precipitation. This diabatic heating drives large-scale overturning that influences the response throughout the tropics and subtropics as well as other teleconnections within the atmosphere extratropics.

The main tool used in this study is correlation and regression analysis which, through least squares fitting, tends to emphasize the larger events. This seems appropriate as it is those events where the signal is clearly larger than the noise. Moreover, the method properly weights each event (unlike many composite analyses). Although it is possible to use regression to eliminate the linear portion of the global mean temperature signal associated with ENSO, the processes that contribute regionally to the global mean differ considerably and the linear approach likely leaves an ENSO residual. We have shown here that 0.06$^\circ$C of the warming from 1950 to 1998 can be accounted for by the increased El Niño phase of ENSO. The lag of the global temperatures behind N34 is 3 months, somewhat less than found in previous studies. In part this probably relates most to the key ENSO index used, as the evolution of ENSO means that greater or lesser lags arise for alternative indices and these vary across the 1976/77 climate shift.

We have shown how positive correlations of surface temperatures with N34 extend from the equatorial region into the subtropics a year later. Results are fairly similar for both 1979-98 and 1950-78 and this pattern also exists for the entire period, without diminution of the correlations, and thus with an enhancement in statistical significance. However, for 1979-98, the warming in the central equatorial Pacific develops from the west and progresses eastward. In contrast, for 1950-78, the anomalous warming begins along the coast of South America and spreads westward, as shown by Rasmusson and Carpenter [1982]. This difference in evolution continues as the eastern Pacific south of the equator warms during the +4 to +8 months for 1979-98 but cools from 1950-78.

Over the oceans, there is an evolution from the maximum anomaly in the equatorial Pacific spreading polewards and to the other oceans over several months. Many tropical land areas tend to be warm from 0 to +8 months. This warmth relates to the relatively sunny and dry, or even drought, conditions during El Niño and so is mainly a result of subsidence and changes in cloudiness and atmospheric circulation. Heat from the ocean is mainly important for the tropical heat sources that drive the teleconnections and tropical large-scale overturning. More generally, over land there is less evidence for delays with increasing latitude, seen over the ocean, but instead the whole pattern is delayed by 3 to 4 months. Also the extension is to slightly higher latitudes at 4 to 8 months, especially in the Northern Hemisphere over North America from 45$^\circ$N to 65$^\circ$N. This fits with Angell [2000] but indicates that his result is probably biased by the distribution of stations used and that it is not representative of the land and ocean combined. Because the peak in N34 tends to occur about November to December and is phase locked to the annual cycle, these results mean that the maximum warmth occurs in the following northern spring and extends well into the summer. The processes involved are discussed in section 4 and are well established in winter and spring [ Trenberth et al., 1998].

OLR is often used as a proxy for precipitation owing to the dominant changes that occur in cloudiness and high cloud tops. However, the algorithms for translating OLR to precipitation do not account for the real effects on OLR of surface temperature changes or small changes associated with heating of the atmosphere. High SSTs are associated with the atmospheric convergence, deep convection and thus low OLR, but also with a flux of latent energy into the atmosphere, condensation and heating of the atmosphere, and transport of heat to higher latitudes where it can be radiated to space. This increase in OLR, seen especially in the subtropics, may be interpreted erroneously by precipitation algorithms as less precipitation, and hence whereas the precipitation amounts should increase when integrated over the domain, they do not appear to and this is probably incorrect [ Soden, 2000]. We infer that the increase in tropospheric temperature in the tropics, shown in Fig. 14, mainly originates from the increase in precipitation and latent heating and the redistribution from overturning and subsidence in the atmosphere, but it is not present in the Xie-Arkin dataset.

There is no doubt that the subsurface ocean heat content in the Pacific leads the SST and atmospheric changes. However, the much lower lag correlations between N34 and temperatures in the Pacific for 1979-98 versus 1950-79 suggest differences in evolution, part of which may have been due to the two volcanic eruptions that occurred, with the Pinatubo event in 1991 estimated to cause a global cooling of $\sim$0.5$^\circ$C, peaking 30 months after the eruption [ Wigley and Santer, 2000]; see especially the residual temperature signal in Fig. 2. But the climate shift in 1976/77 seems to have been a major factor in fundamentally altering the evolution sequence of ENSO events. It suggests that the subsurface ocean evolution since 1980, which has become the paradigm for ENSO, may not be robust across all events, and/or the links between the subsurface and the surface may have changed. Mechanisms for warming in the tropical Pacific depend on different balance of terms within the ocean [ Neelin et al., 1998] and this balance may have shifted. It is known, for instance, that vertical temperature gradients and upwelling in the eastern tropical Pacific play a key role in westward development, while eastward development relies more on east-west temperature gradients and advection in the central tropical Pacific. We noted in Fig. 13 how poorly the SSTs were related to the total ocean heat content anomalies outside of the equatorial Pacific strip and that this will limit the ability of two layer oceans, such as are often employed in intermediate models, to simulate ENSO.

The negative feedback between SST and surface fluxes can be interpreted as showing the importance of the discharge of heat during El Niño events and the recharge of heat during La Niña events. Relatively clear skies in the central and eastern tropical Pacific allow solar radiation to enter the ocean, apparently offsetting the below normal SSTs, but the heat is carried away by ocean currents and adjustments through ocean Rossby and Kelvin waves, and the heat is stored in the western Pacific tropics. This is not simply a rearrangement of the ocean heat, but also a restoration of heat in the ocean. Similarly, during El Niño, the loss of heat, especially through evaporation, into the atmosphere is a discharge of the heat content, and both contribute to the life cycle of ENSO. These observations support the picture put forward by Barnett et al. [1991] based mainly on model results in which the SST anomalies are created by ocean dynamics and response to wind forcing, and not local surface fluxes. However, the role of the surface fluxes and the diabatic component of the ENSO cycle should not be underestimated as it has implications for the role of ENSO in climate.

An alternative view of the change in evolution of ENSO is one based upon time filtering of the data into interannual and interdecadal time scales (e.g., Zhang et al., 1997; Mantua et al., 1997; Giese and Carton, 1999) but because the patterns of SST on each time scale are not orthogonal and the processes are nonlinear, strong interactions are implied and results are difficult to interpret. The reasons why the change in evolution with the 1976/77 climate shift occurred are quite uncertain but appear to relate to changes in the thermocline [ Guilderson and Schrag, 1998; Giese and Carton, 1999] which could be linked to climate change and global warming [ Trenberth and Hoar, 1996; Meehl and Washington, 1996; Timmermann et al., 1999].


Acknowledgments.

This research was sponsored by NOAA Office of Global Programs grant NA56GP0247 and the joint NOAA/NASA grant NA87GP0105. We thank Clara Deser for comments.


References

An, S-I., and B. Wang, Interdecadal change in the structure of the ENSO mode and its impact on the ENSO frequency, J. Clim., 13, 2044-2055, 2000.

Angell, J. K., Tropospheric temperature variations adjusted for El Niño, 1958-1998, J. Geophys. Res.,105, 11841-11849, 2000.

Barnett, T. P., M. Latif, E. Kirk, and E. Roeckner, On ENSO physics, J. Clim., 5, 487-515, 1991.

Battisti, D. S., and A. C. Hirst, Interannual variability in the tropical atmosphere/ocean system: Influence of the basic state, ocean geometry and non-linearity, J. Atmos. Sci., 46, 1687-1712, 1989.

Behringer, D. W., M. Ji, and A Leetmaa, An improved coupled model for ENSO prediction and implications for ocean initialization. Pt I: The ocean data assimilation system, Mon. Weather Rev., 126, 1013-1021, 1998.

Blanke B., J. D. Neelin, and D. Gutzler, Estimating the effect of stochastic wind stress forcing on ENSO irregularity, J. Clim., 10, 1473-1486, 1997.

Cane M. A., and S. E. Zebiak, A theory for El Niño and the Southern Oscillation, Science, 228, 1085-1087, 1985.

Christy, J. R., and R. T. McNider, Satellite greenhouse signal, Nature, 367, 325, 1994.

Dai, A., K. E. Trenberth, and T. R. Karl, Effects of clouds, soil moisture, precipitation and water vapor on diurnal temperature range, J. Clim., 12, 2451-2473, 1999.

Deser, C., M. A. Alexander, and M. S. Timlin, Upper-ocean thermal variations in the North Pacific during 1970-1991, J. Clim., 9, 1840-1855, 1996.

Eckert C., and M. Latif, Predictability of a stochastically forced hybrid coupled model of El Niño, J. Clim., 10, 1488-1504, 1997.

Flügel, M., and P. Chang, Impact of dynamical and stochastic processes on the predictability of ENSO, Geophys. Res. Lett., 23, 2089-2092, 1996.

Giese, B. S., and J. A. Carton, Interannual and decadal variability in the tropical and midlatitude Pacific Ocean, J. Clim., 12, N3402-3418, 1999.

Graham, N. E., Decadal-scale climate variability in the tropical and North Pacific during the 1970s and 1980s: observations and model results, Clim. Dyn., 10, 135-162, 1994.

Guilderson, T. P., and D. P. Schrag, Abrupt shift in subsurface temperatures in the tropical Pacific associated with changes in El Niño, Science, 281, 240-243, 1998.

Hellerman, S., and M. Rosenstein, Normal monthly wind stress over the world ocean with error estimates, J. Phys. Oceanogr.,13, 1093-1104, 1983.

Hurrell, J. W., and K. E. Trenberth, Global sea surface temperature analyses: multiple problems and their implications for climate analysis, modeling and reanalysis, Bull. Am. Meteor. Soc., 80, 2661-2678, 1999.

Jin, F., An equatorial ocean recharge paradigm for ENSO, Part I: Conceptual model, J. Atmos. Sci., 54, 811-829, 1997.

Jones, P. D., The influence of ENSO on global temperatures, Clim. Monitor, 17, 80-89, 1989.

Klein, S. A., B. J. Soden, and N-C. Lau, Remote sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge, J. Clim., 12, 917-932, 1999.

Kumar, K. K., B. Rajagopalan, and M. A. Cane, On the weakening relationship between the Indian monsoon and ENSO, Science, 284, 2156-2159, 1999.

Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, A Pacific interdecadal climate oscillation with impacts on salmon production, Bull. Am. Meteor. Soc., 78, 1069-1079, 1997.

Meehl, G. A., and W. M. Washington, El Niño-like climate change in a model with increased atmospheric 2 concentrations, Nature, 382, 56-60, 1996.

Meinen, C. S., and M. J. McPhaden, Observations of warm water volume changes in the equatorial Pacific and their relationship to El Niño and La Niña, J. Clim., 13, 3551-3559, 2000.

Moore, A. M., and R. Kleeman, Stochastic forcing of ENSO by the intraseasonal oscillation, J. Clim., 12, 1199-1220, 1999.

Münnich, M., M. A. Cane, and S. E. Zebiak, A study of self-excited oscillations of the tropical ocean-atmosphere system, Part II: Nonlinear cases, J. Atmos. Sci.,48, 1238-1248, 1991.

Neelin, J. D., and F.-F Jin, Modes of interannual tropical ocean-atmosphere interaction - a unified view, part III: Analytical results in fully coupled cases, J. Atmos. Sci, 50, 3523-3540, 1993.

Neelin, J. D., D. S. Battisti, A. C. Hirst, F-F. Jin, Y. Wakata, T. Yamagata, and S. E. Zebiak, ENSO theory, J. Geophys. Res.,103, 14261-14290, 1998.

Newell, R. E., and B. C. Weare, Factors governing tropospheric mean temperatures, Science, 194, 1413-1414, 1976.

Oberhuber, J. M., An atlas based on the ``COADS'' data set: The budgets of heat, buoyancy and turbulent kinetic energy at the surface of the global ocean. Rep. 15, Max Planck Inst. fur Meteorologie, 20 pp, 1998.

Pan, Y. H., and A. H. Oort, Global climate variations connected with sea surface temperature anomalies in the eastern equatorial Pacific Ocean for the 1958-1973 period, Mon. Weather Rev., 111, 1244-1258, 1983.

Power, S. B., and R. Kleeman, Surface heat flux parametrization and the response of ocean general circulation models to high-latitude freshening, Tellus, 46, 86-95, 1994.

Rasmusson, E. M., and T. H. Carpenter, Variations in tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño, Mon. Weather Rev., 110, 354-384, 1982.

Reynolds, R. W., and T. M. Smith, Improved global sea surface temperature analyses using optimum interpolation, J. Clim., 7, 929-948, 1994.

Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, A dipole mode in the tropical Indian Ocean, Nature, 401, 360-363, 1999.

Smith, T. M., Tropical Pacific sea level variations (1948-1998), J. Clim., 13, 2757-2769, 2000.

Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. C. Stokes, Reconstruction of historical sea surface temperatures using empirical orthogonal functions, J. Clim., 9, 1403-1420, 1996.

Soden, B. J., The sensitivity of the tropical hydrological cycle to ENSO, J. Clim., 13, 538-549, 2000.

Stone L., P. I. Saparin, A. Huppert, and C. Price, El Niño Chaos: The potential role of noise and stochastic resonance on the ENSO cycle, Geophys. Res. Lttrs., 25, 175-178, 1998.

Sun, D-Z, The heat sources and sinks of the 1986-87 El Niño, J. Clim., 13, 3533-3550, 2000.

Sun, D-Z, and K. E. Trenberth, Coordinated heat removal from the tropical Pacific during the 1986-87 El Niño, Geophys. Res. Lttrs., 25, 2659-2662, 1998.

Suarez M. J., and P. S. Schopf, A delayed action oscillator for ENSO, J. Atmos. Sci., 45, 3283-3287, 1988.

Timmermann A., J. M. Oberhuber, A. Bacher, M. Esch, M. Latif, and E. Roeckner, Increased El Niño frequency in a climate model forced by future greenhouse warming, Nature, 398, 694-696, 1999.

Tourre, Y. M., and W. B. White, ENSO signals in global upper-ocean temperature, J. Phys. Oceanogr.,25, 1317-1332, 1995.

Trenberth, K. E., Signal versus noise in the Southern Oscillation, Mon. Weather Rev., 112, 326-332, 1984.

Trenberth, K. E., Recent observed interdecadal climate changes in the Northern Hemisphere, Bull. Am. Meteor. Soc., 71, 988-993, 1990.

Trenberth, K. E., and C. J. Guillemot, Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses, Clim. Dyn., 14, 213-231, 1998.

Trenberth, K. E., and T. J. Hoar, The 1990-1995 El Niño-Southern Oscillation Event: Longest on record, Geophys. Res. Lttrs., 23, 57-60, 1996.

Trenberth, K. E., and T. J. Hoar, El Niño and climate change, Geophys. Res. Lttrs., 24, 3057-3060, 1997.

Trenberth, K. E., and J. W. Hurrell, Decadal atmosphere-ocean variations in the Pacific, Clim. Dyn., 9, 303-319, 1994.

Trenberth, K. E., and A. Solomon, The Global Heat Balance: Heat transports in the atmosphere and ocean, Clim. Dyn., 10, 107-134, 1994.

Trenberth, K. E., and D. P. Stepaniak, Indices of El Niño evolution, J. Clim., in press, 2001.

Trenberth, K. E., J. M. Caron, and D. P. Stepaniak, The atmospheric energy budget and implications for surface fluxes and ocean heat transports, Clim. Dyn., in press, 2001a.

Trenberth, K. E., D. P. Stepaniak, and J. M. Caron, Interannual variations in the atmospheric heat budget, J. Geophys. Res., submitted, 2001b.

Trenberth, K. E., G. W. Branstator, D. Karoly, A. Kumar, N-C. Lau, and C. Ropelewski, Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures, J. Geophys. Res., 103, 14291-14324, 1998.

Tziperman, E., M. A. Cane, S. E. Zebiak, Y. Xue, and B. Blumenthal, On the locking of El Niño's peak time to the end of the calendar year in the delayed oscillator picture of ENSO, J. Clim., 11, 2191-2199, 1998.

Urban, F. E., J. E. Cole, and J. T. Overpeck, Modification of tropical Pacific variability by its mean state inferred from a 155-year coral record, Nature, 407, 989-993, 2000.

Waliser, D. E., and W. Zhou, Removing satellite equatorial crossing time biases from the OLR and HRC data sets, J. Clim., 10, 2125-2146, 1997.

Wallace, J. M., E. M. Rasmusson, T. P. Mitchell, V. E. Kousky, E. S. Sarachik, and H. von Storch, On the structure and evolution of ENSO-related climate variability in the tropical Pacific: Lessons from TOGA, J. Geophys. Res.,103, 14241-14260, 1998.

Wang, B., Interdecadal changes in El Niño onset in the last four decades, J. Clim., 8, 267-285, 1995.

Webster, P. J., A. M. Moore, J. P. Loschnigg, and R. R. Leben, Coupled ocean-atmosphere dynamics in the Indian Ocean during 1997-98, Nature, 401, 356-360, 1999.

Wigley T. M. L., and B. D. Santer, Differential ENSO and volcanic effects on surface and tropospheric temperatures, J. Clim., submitted, 2000.

Wyrtki, K., Water displacements in the Pacific and the genesis of El Niño cycles, J. Geophys. Res.,90, 7129-7132, 1985.

Xie, P., and P. A. Arkin, Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions, J. Clim., 9, 840-858, 1996.

Xie, P., and P. A. Arkin, Global precipitation: A 17 year monthly analysis based on gauge observations, satellite estimates and numerical model outputs, Bull. Am. Meteor. Soc., 78, 2539-2558, 1997.

Zhang, R-H, and S. Levitus, Structure and evolution of interannual variability of the tropical Pacific upper ocean temperature, J. Geophys. Res.,101, 20501-20524, 1996.

Zhang, R-H, and S. Levitus, Interannual variability of the coupled tropical Pacific Ocean-atmosphere system associated with the El Niño Southern Oscillation, J. Clim., 10, 1312-1330, 1997.

Zhang, Y., J. M. Wallace, and D. S. Battisti, ENSO-like interdecadal variability: 1900-93, J. Clim., 10, 1004-1020, 1997.




Julie Caron
2001-01-16