1. Introduction
The tropical Pacific Ocean is home to the Earth’s largest source of interannual climate variability—the El Niño–Southern Oscillation (ENSO). ENSO refers to an interannually recurring warming (El Niño) and cooling (La Niña) of the eastern and central tropical Pacific Ocean, and a related large-scale seesaw in atmospheric sea level pressure between the Australia–Indonesian region and the south-central tropical Pacific, known as the Southern Oscillation. El Niño and La Niña events typically last approximately 12–18 months and occur about every 2–7 yr with large variations in their maximum amplitude and other statistical characteristics. The onset of ENSO events generally occurs during the boreal spring [March–May (MAM)] and/or summer [June–August (JJA)], while peak amplitudes in the sea surface temperature anomalies (SSTA) are generally attained during the boreal winter [December–February (DJF)]. ENSO’s termination phase usually occurs during boreal spring (MAM; Larkin and Harrison 2002; Chang et al. 2006). Even though many state-of-the-art coupled general circulation models can reproduce this feature reasonably well, the underlying physical mechanisms for El Niño terminations still remain elusive.
Our theoretical understanding of the dynamics and oscillatory nature of ENSO has significantly increased over the last three decades (Neelin et al. 1998). It is generally accepted that the generation of an ENSO event requires a positive ocean–atmosphere feedback (known as the Bjerknes feedback) to amplify the original anomalous zonal equatorial SST gradient. The termination of ENSO events, on the other hand, requires a delayed negative feedback. The last three decades have seen various mechanisms proposed to explain the negative feedback required for the termination of ENSO events (see e.g., Neelin et al. 1998; Wang and Picaut 2004; Chang et al. 2006). Termination mechanisms include the following: (i) ocean wave reflection at the Pacific Ocean western continental boundary (Schopf and Suarez 1988; Battisti and Hirst 1989), (ii) ocean temperature advection by zonal currents associated with the reflection of oceanic waves at the eastern and western continental boundaries of the Pacific Ocean (Picaut et al. 1997), (iii) the recharge/discharge of warm water volume due to Sverdrup transport (Jin 1997), and (iv) oceanic Kelvin waves forced by wind anomalies in the western equatorial Pacific Ocean (Weisberg and Wang 1997; Wang et al. 1999). These mechanisms consistently point to ocean dynamical adjustments as the restoring force of the unstable air–sea interactions of ENSO. Whereas each of these proposed delayed ocean feedback mechanisms can be identified to a certain extent in observational datasets (Delcroix et al. 2000; Meinen and McPhaden 2000; Bosc and Delcroix 2008; Boulanger et al. 2003; McPhaden and Yu 1999), none of them directly resolve ENSO’s apparent synchronization to the seasonal cycle.
Some recent studies have argued that more than one type of El Niño event exists based on the spatial distribution of SSTA (Kug et al. 2009; Kao and Yu 2009). For instance, Kug et al. (2009) use the spatial structure of SSTAs to separate El Niño events in two main types, warm pool (WP) and cool tongue (CT) type events. The CT-type El Niño events refer to El Niño events displaying the classical canonical eastern equatorial Pacific warming, while WP El Niño events refer to those events displaying warming prominently located in the central equatorial Pacific near the edge of the western Pacific warm pool. These central Pacific WP-type El Niño events have also been called trans-Niño (Trenberth and Stepaniak 2001), date line El Niño (Larkin and Harrison 2005), or El Niño Modoki (Ashok et al. 2007) events. How the recent advances in our theoretical understanding of ENSO events relate to the newly identified WP-type El Niño remain unclear. However, research to date suggests that they operate with different dynamical mechanisms to CT El Niño events (Kug et al. 2009).
The focus of this manuscript is an observed feature of El Niño events that is often not considered by conceptual models of ENSO; namely, the southward shift of El Niño–related equatorial zonal wind anomalies (Harrison 1987; Harrison and Larkin 1998; Larkin and Harrison 2002). Near the end of the calendar year, when El Niño events typically reach their peak amplitude, the associated zonal wind anomalies (which prior to this are quasi-symmetric about the equator) abruptly shift southward, so that the maximum anomalous zonal wind is located around 5°–7°S (Harrison 1987; Harrison and Larkin 1998; Harrison and Vecchi 1999; Vecchi and Harrison 2003; Lengaigne et al. 2006). Prior to this meridional shift, in the case of El Niño (La Niña) events the westerly (easterly) zonal wind anomalies on the equator are responsible for maintaining the deep (shallow) eastern equatorial Pacific thermocline. As such, the southward wind shift and the associated weakening of westerly winds right on the equator ultimately allows the eastern equatorial Pacific thermocline depth to return to near-normal values. This mechanism has been proposed to explain the termination of El Niño events in MAM and, hence, plays a key role in the synchronisation of ENSO to the seasonal cycle (Harrison and Vecchi 1999).
Vecchi and Harrison (2003, 2006) have shown that this southward anomalous zonal wind shift was the dominant factor in the observed shoaling of eastern equatorial Pacific thermocline depth in the 1997/98 and 2002/03 El Niño events. In addition, Lengaigne et al. (2006) and Lengaigne and Vecchi (2009) find that the rapid southward migration of these anomalous zonal winds during DJF is responsible for intense eastern Pacific thermocline shoaling during simulated extreme El Niño events in a coupled climate model. This shoaling preconditions the system for a rapid demise of El Niño events in the following months. The intensity of this seasonal southward wind migration is found to be somewhat weaker for their simulated moderate El Niño events, resulting in a less pronounced thermocline shoaling. However, the system is still preconditioned for rapid event termination in both cases (Lengaigne and Vecchi 2009).
Spencer (2004) demonstrated that this southward shift of the El Niño–related westerly wind anomalies can arise solely in response to anomalous interactions with the seasonal cycle. Both Lengaigne et al. (2006) and Vecchi (2006) go one step further and show that the southward shift of the anomalous zonal winds is driven by the southward displacement of the warmest western-central Pacific SSTs during DJF in response to the seasonal evolution of solar insolation. In these studies it is also noted that anomalous El Niño–related convection also shifts south along with the warmest SST and zonal wind anomalies. It is alluded to, although not explicitly shown, that this southward shift of anomalous convection is the cause for the southward wind shift. To date, however, the detailed dynamical/thermodynamical mechanisms responsible for this southward shift of zonal wind anomalies still remain unknown. Here we investigate the causes of this observed southward zonal wind shift using a simplified atmosphere model that is based on a well-mixed boundary layer and a free troposphere represented by the gravest baroclinic mode. The simple nature of this model allows us to identify the causes for the southward shift under consideration.
To start with the main conclusion upfront, we find that the DJF southward surface wind shift is due to interactions between the momentum damping associated with the climatological surface winds and the anomalous boundary layer winds induced by a lower-tropospheric Gill-type Rossby–Kelvin wave. Furthermore, using a linear shallow-water ocean model we show that the southward shift of wind stress anomalies during mature El Niño events in boreal winter causes the asymmetric discharge of the observed equatorial Pacific Ocean heat content (EOF mode 2 of observed 20°C isotherm depth; Meinen and McPhaden 2000) and plays a significant role in the rapid termination of El Niño events toward the turn of the calendar year. This process is shown to strongly accelerate the termination of large El Niño events relative to those that only utilize the classical Sverdrup-transport-induced discharge.
The rest of this paper is organized as follows. In section 2 we analyze the observed wind stress response to an El Niño event and show that the southward wind shift during DJF is represented by the quadrature of the first two principal component modes of equatorial wind stress. In section 3 we present the simple atmospheric model along with the experiments carried out to establish the cause for the observed southward wind shift. Section 4 details the linear ocean shallow-water model and the experiments used to demonstrate that the southward shift of the ENSO-related zonal wind anomalies is responsible for the asymmetric spatial structure of the equatorial Pacific heat content recharge/discharge and strongly contributions to the termination of large El Niño events. The main results are summarized and briefly discussed in the conclusions in section 5.
2. Observed wind response to ENSO
To identify the observed southward shift in zonal winds during El Niño events we first composite wind stress anomalies (1958–2001 long-term monthly mean removed) of the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) dataset for the period 1958–2001 (Uppala et al. 2005) around WP, CT, and “mixed” El Niño years. Mixed El Niño events are those events that display spatial features consistent with both WP and CT El Niño events. The definition of WP, CT, and mixed El Niño years follows that of Kug et al. (2009); 1977/78, 1990/91, and 1994/95 El Niño events are classified as WP type, while the El Niño events of 1972/73, 1976/77, 1982/83, and 1997/98 are classified as the CT type. The El Niño events occurring in 1986/87, 1987/88, and 1991/92 display features consistent with both WP and CT classes of El Niño, as such they are classified as mixed El Niño events. The zonal mean of the wind stress anomalies and their associated curl is then calculated between 160°W and 120°E for the three different composites (Fig. 1). Looking at the latitude of the maximum zonal wind stress anomalies, identified by the zero wind stress curl contour closest to the equator, there is a clear southward zonal wind shift around DJF during CT and mixed El Niño events consistent with previous studies (Harrison 1987; Harrison and Larkin 1998; Harrison and Vecchi 1999; Vecchi and Harrison 2003; Lengaigne et al. 2006). Interestingly however, wind stress anomalies during WP El Niño events display no distinct southward zonal wind shift during DJF.
Zonal mean between 160°E and 120°W of (a) WP (1977/78, 1990/91, and 1994/95), (b) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (c) mixed (1986/87, 1987/88, and 1991/92) El Niño event composites of ERA-40 surface wind stress anomalies (Pa, vectors) and wind stress curl (shaded).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Zonal mean between 160°E and 120°W of (a) WP (1977/78, 1990/91, and 1994/95), (b) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (c) mixed (1986/87, 1987/88, and 1991/92) El Niño event composites of ERA-40 surface wind stress anomalies (Pa, vectors) and wind stress curl (shaded).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Zonal mean between 160°E and 120°W of (a) WP (1977/78, 1990/91, and 1994/95), (b) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (c) mixed (1986/87, 1987/88, and 1991/92) El Niño event composites of ERA-40 surface wind stress anomalies (Pa, vectors) and wind stress curl (shaded).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
To isolate the interannual wind changes, we carry out an EOF analysis of the ERA-40 wind stress anomalies between 10°N–10°S and 100°E–60°W. Corresponding global spatial patterns are then calculated as the regression coefficients between each of the principal component time series of the EOF modes and the anomalous ERA-40 zonal and meridional wind stresses at each spatial location. The first PC (PC1; Fig. 2c), which accounts for 22% of the equatorial region variance, clearly represents ENSO variability producing a correlation coefficient of 0.84 when compared with SSTA averaged over the Niño-3.4 region (5°S–5°N, 170°–120°W) obtained from the Hadley Centre SST (HadSST) dataset (Rayner et al. 2003). This is statistically significant above the 99% level. Note that the statistical significance of all correlation coefficients reported in this study take account of serial (auto) correlation in the series based on the reduced effective number of degrees of freedom outlined by Davis (1976). The corresponding spatial structure displays the classical El Niño–type response, which features westerly wind stress anomalies in the western-central Pacific that have their maximum amplitude south of the equator (Fig. 2a).
The spatial pattern of regression coefficients for (a) PC1 and (b) PC2 of the ERA-40 equatorial Pacific region surface wind stresses (vectors) where the background shading represents the associated wind stress curl. (c) The corresponding principal components, where the PC1 is solid black and PC2 is solid red.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The spatial pattern of regression coefficients for (a) PC1 and (b) PC2 of the ERA-40 equatorial Pacific region surface wind stresses (vectors) where the background shading represents the associated wind stress curl. (c) The corresponding principal components, where the PC1 is solid black and PC2 is solid red.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The spatial pattern of regression coefficients for (a) PC1 and (b) PC2 of the ERA-40 equatorial Pacific region surface wind stresses (vectors) where the background shading represents the associated wind stress curl. (c) The corresponding principal components, where the PC1 is solid black and PC2 is solid red.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The second wind stress PC (PC2; Fig. 2c) accounts for 16% of the equatorial region variance. We note here that PC1 and PC2 are distinct eigenvectors (i.e., they are not a degenerate multiplet) using the North et al. (1982) rule of thumb for sampling errors. PC2 has virtually no direct correlation with the Niño-3.4 region SSTA (correlation coefficient of 0.18). However, a relationship with ENSO is evident upon visual analysis of the ERA-40 PC2 of wind stress anomalies composited around WP, CT, and mixed El Niño events (Figs. 3d–f). For example, during CT El Niño events PC2 has positive values during the onset phase, which then reverse sign abruptly approximately 1–2 months prior to the event peak. In the case of mixed El Niño events, PC2 is roughly zero prior to the event peak and moves toward more negative values around the time of the event peak. The corresponding spatial structure of PC2 regression coefficients (Fig. 2b) exhibits several interesting features. Most notable is that the weighting of zonal wind stresses north of the equator are of the opposite sign and displaced westward to those south of the equator (Fig. 2b).
(a) WP, (b) CT, and (c) mixed El Niño event PC1 composite members (thin black) and mean (thick red) and (d)–(f) the corresponding PC2 composite members (thin black) and mean (thick red), respectively. The PC1 and PC2 reconstructed (g) WP, (h) CT, and (i) mixed El Niño event zonal mean wind stress anomalies (Pa, vectors) and wind stress curl (shaded) between 160°E and 120°W.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) WP, (b) CT, and (c) mixed El Niño event PC1 composite members (thin black) and mean (thick red) and (d)–(f) the corresponding PC2 composite members (thin black) and mean (thick red), respectively. The PC1 and PC2 reconstructed (g) WP, (h) CT, and (i) mixed El Niño event zonal mean wind stress anomalies (Pa, vectors) and wind stress curl (shaded) between 160°E and 120°W.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) WP, (b) CT, and (c) mixed El Niño event PC1 composite members (thin black) and mean (thick red) and (d)–(f) the corresponding PC2 composite members (thin black) and mean (thick red), respectively. The PC1 and PC2 reconstructed (g) WP, (h) CT, and (i) mixed El Niño event zonal mean wind stress anomalies (Pa, vectors) and wind stress curl (shaded) between 160°E and 120°W.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Reconstructing the wind stress anomalies around El Niño events using the composited ERA-40 PC1 and PC2 reveals that PC2 works in quadrature with PC1 to allow the ENSO-related zonal wind anomalies to have a more equatorially symmetric structure during the onset phase (July 0–October 0) of CT El Niño events when ERA-40 PC2 has positive values (Fig. 4c). The transition of ERA-40 PC2 to negative values around the peak of these CT El Niño events allows for a southeastward shift of the related zonal wind anomalies around the end of the calendar year and a strong meridional asymmetry (Fig. 4d). This feature is highlighted when looking at the zonal mean wind stresses or wind stress curl between 160° and 120°W displayed in Figs. 3g–i. It is clear that PC1 and PC2 of the ERA-40 wind stresses form a quadrature pair that represents the southward shift of zonal wind anomalies during CT and mixed El Niño–type events, or lack of it in the case of WP El Niño events.
CT El Niño event composite mean surface wind stress anomalies (Pa, vectors) and the magnitude of the zonal component wind stress anomalies (shaded) in (left) October prior to the event peak and (right) those in March after the event peak. (a),(b) The ERA-40 anomalies. (c),(d) The PC1 and PC2 reconstructed ERA-40 anomalies. (e),(f) The WLF99 control simulation anomalies.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
CT El Niño event composite mean surface wind stress anomalies (Pa, vectors) and the magnitude of the zonal component wind stress anomalies (shaded) in (left) October prior to the event peak and (right) those in March after the event peak. (a),(b) The ERA-40 anomalies. (c),(d) The PC1 and PC2 reconstructed ERA-40 anomalies. (e),(f) The WLF99 control simulation anomalies.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
CT El Niño event composite mean surface wind stress anomalies (Pa, vectors) and the magnitude of the zonal component wind stress anomalies (shaded) in (left) October prior to the event peak and (right) those in March after the event peak. (a),(b) The ERA-40 anomalies. (c),(d) The PC1 and PC2 reconstructed ERA-40 anomalies. (e),(f) The WLF99 control simulation anomalies.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
3. The atmospheric model and experiment results
a. Model validation
In this section we carry out a control simulation with the WLF99 atmospheric model. The model is integrated from 1958–2001 and monthly mean HadSST anomalies are added to the model’s long-term mean seasonal cycle SST forcing in the Pacific Ocean region (see Table 1). The monthly SSTA values are linearly interpolated to the model time step (20 min), while the resultant model surface (BL) winds are converted to wind stresses. Anomalous wind stresses (1958–2001 long-term monthly mean removed) are then composited around WP, CT, and mixed El Niño years. The zonal mean of the wind stress anomalies and their associated curl is then calculated between 160°W and 120°E for the three different composites and plotted in Fig. 5 for comparison with the ERA-40 composites shown in Fig. 1.
Details of the intermediate complexity 2.5-layer atmospheric model experiments presented in this manuscript. The boundary layer (BL) and lower troposphere (LT). The acronym UA refers to the zonal wind anomalies.
Zonal mean between 160°E and 120°W of (a) WP (1977/78, 1990/91, and 1994/95), (b) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (c) mixed (1986/87, 1987/88, and 1991/92) El Niño event composited WLF99 control simulation surface wind stress anomalies (Pa, vectors) and wind stress curl (shaded). Note that the vector scaling differs between the three (see vector scales in the top right).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Zonal mean between 160°E and 120°W of (a) WP (1977/78, 1990/91, and 1994/95), (b) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (c) mixed (1986/87, 1987/88, and 1991/92) El Niño event composited WLF99 control simulation surface wind stress anomalies (Pa, vectors) and wind stress curl (shaded). Note that the vector scaling differs between the three (see vector scales in the top right).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Zonal mean between 160°E and 120°W of (a) WP (1977/78, 1990/91, and 1994/95), (b) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (c) mixed (1986/87, 1987/88, and 1991/92) El Niño event composited WLF99 control simulation surface wind stress anomalies (Pa, vectors) and wind stress curl (shaded). Note that the vector scaling differs between the three (see vector scales in the top right).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Looking at the latitude of the maximum zonal wind stress anomalies (identified by the zero wind stress curl contour located closest to the equator) it is clear that all three composites display a southward zonal wind anomaly shift during DJF. In the case of the WP El Niño event composite, the clear southward zonal wind stress anomaly shift of the WLF99 atmospheric model is in contrast to the indistinct latitudinal wind changes displayed in the ERA-40 WP El Niño composite. The southward shift produced by the WLF99 mixed El Niño composite is in qualitative agreement with the corresponding ERA-40 composite. However, WLF99 produces a more pronounced southward zonal wind stress anomaly shift than that in the ERA-40 composite namely due to the northern location of the maximum latitudinal wind speed in the months leading up to the southward shift (May–October). During CT El Niño events the clear southward shift of the zonal wind stress anomaly around DJF produced by WLF99, and seen in Figs. 5b (Figs. 4e,f), is qualitatively very similar to the ERA-40 CT El Niño composite displayed in Fig. 1b (Figs. 4a,b).
Thus, while the WLF99 model may be too simplistic to reproduce the atmospheric response to the more complex spatial structures associated with WP El Niño, the model does a reasonable job reproducing the southward shift of zonal wind stress anomalies during CT El Niño events. We note that the magnitudes of the simulated wind stresses and the related curl are significantly underestimated by the intermediate complexity model. However, we do not expect this to affect the results as the southward wind shift appears to be well represented regardless of its relative magnitude.
b. Controlling dynamics
To investigate the underlying dynamics of the observed and simulated southward shift of the wind anomalies during CT El Niño events, we carry out another series of experiments. The first of these experiments, hereafter titled the temporally fixed idealized forcing experiment (FIF-exp), incorporates a time-independent SSTA, which is added to the model’s climatological SST forcing and the model is integrated for 5 yr (see Table 1). The prescribed SSTA pattern has a spatial structure similar to that of a CT El Niño event and is defined by a maximum value of 1.2°C on the equator between 160° and 80°W and decays westward linearly to zero between 160°W and 160°E. The equatorial SSTAs decay away from the equator with an e-folding length scale of 10° latitude, so there is no meridional SSTA asymmetry or off-equatorial region SSTA forcing. Using this SSTA forcing in the unmodified atmospheric model, the results of the FIF-exp confirm that the southward shift of the ENSO-related zonal wind anomalies around the peak of the ENSO event occurs regardless of the simplicity of the spatial structure of the SSTA forcing (Fig. 6a, black solid line). Furthermore, this result agrees with Spencer (2004) who show that a southward shift of the El Niño–related westerly wind anomalies can arise solely in response to anomalous interactions with the seasonal cycle.
The latitude of the simulated maximum zonal wind response (UA) to ENSO-related SSTA forcing zonally averaged between 160°E and 150°W. For comparison the black line in each panel represents the latitude of the maximum zonal wind response from the FIF-exp. (a) The latitude of the maximum zonal wind response from the BLonly-exp (dashed red) and the LTonly-exp (dashed blue). (b) The latitude of the maximum zonal wind response from the BLcoupled-exp (solid red) and LTcoupled-exp (solid blue). (c) Latitude of the maximum zonal wind speed from the experiment utilizing a stationary annual mean wind speed in the BL turbulent momentum flux (dash–dot blue).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The latitude of the simulated maximum zonal wind response (UA) to ENSO-related SSTA forcing zonally averaged between 160°E and 150°W. For comparison the black line in each panel represents the latitude of the maximum zonal wind response from the FIF-exp. (a) The latitude of the maximum zonal wind response from the BLonly-exp (dashed red) and the LTonly-exp (dashed blue). (b) The latitude of the maximum zonal wind response from the BLcoupled-exp (solid red) and LTcoupled-exp (solid blue). (c) Latitude of the maximum zonal wind speed from the experiment utilizing a stationary annual mean wind speed in the BL turbulent momentum flux (dash–dot blue).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The latitude of the simulated maximum zonal wind response (UA) to ENSO-related SSTA forcing zonally averaged between 160°E and 150°W. For comparison the black line in each panel represents the latitude of the maximum zonal wind response from the FIF-exp. (a) The latitude of the maximum zonal wind response from the BLonly-exp (dashed red) and the LTonly-exp (dashed blue). (b) The latitude of the maximum zonal wind response from the BLcoupled-exp (solid red) and LTcoupled-exp (solid blue). (c) Latitude of the maximum zonal wind speed from the experiment utilizing a stationary annual mean wind speed in the BL turbulent momentum flux (dash–dot blue).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
1) Boundary layer response
Here we calculate the BL-only response to the temporally fixed idealized ENSO forcing described above by essentially decoupling anomalous interactions between the model BL and LT. Since the 850-hPa geopotential height in the atmospheric model is the sum of the BL and LT components we can decouple the model layers and isolate the BL response by prescribing the LT component of the 850-hPa geopotential height to the climatological value (calculated by carrying out a 5-yr simulation with no anomalous forcing; Fig. 7). This experiment is hereafter known as the BL-only experiment (BLonly-exp; see Table 1).
A schematic representation of the coupling and interactions between and within the model components. The dash–dot line indicates thermodynamic coupling, while the dashed line represents dynamic coupling. The solid lines indicate both dynamic and thermodynamic interactions, while the acronym GPH stands for geopotential height.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
A schematic representation of the coupling and interactions between and within the model components. The dash–dot line indicates thermodynamic coupling, while the dashed line represents dynamic coupling. The solid lines indicate both dynamic and thermodynamic interactions, while the acronym GPH stands for geopotential height.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
A schematic representation of the coupling and interactions between and within the model components. The dash–dot line indicates thermodynamic coupling, while the dashed line represents dynamic coupling. The solid lines indicate both dynamic and thermodynamic interactions, while the acronym GPH stands for geopotential height.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The anomalous surface winds of this BLonly-exp represent the classical Lindzen–Nigam-type response to SST anomalies. By climatologically prescribing the LT 850 height, we also inhibit the formation of an anomalous secondary circulation through atmospheric Ekman pumping. The simulated wind anomalies in this experiment do not display any significant southward shift around the end of the calendar year. This becomes apparent when looking at the latitude of the maximum zonal mean zonal wind anomalies displayed in Fig. 6a. As expected, the resulting BL winds induce changes in geopotential height that have a virtually fixed spatial structure for the entire experiment [Fig. 8b(1)]. Thus, the BL response alone cannot reproduce the observed southward shift of ENSO-related zonal wind anomalies near the end of the calendar year.
(a) The idealized SSTA pattern. The mean DJF response of the [b(1)] BLonly-exp and [b(2)] LTonly-exp. Anomalies of model 850-hPa geopotential height (gpha in figure above) are shaded, while changes in the respective model layer wind (m s−1) are displayed as vectors. The magenta contours (0.15–0.45 with a spacing of 0.15) overlaying the LT response shown in [b(2)] are anomalies of the nonlinear SST-dependent conditional heating, δ [Eq. (6)]. The mean DJF BL 850-hPa geopotential height (shaded) and BL winds of the [c(1)] BLcoupled-exp and [c(2)] LTcoupled-exp and (d) the sum of both [c(1)] and [c(2)]. (e) Anomalies of model BL 850-hPa geopotential height (shaded) and BL winds from the unmodified model forced with the idealized SSTA.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) The idealized SSTA pattern. The mean DJF response of the [b(1)] BLonly-exp and [b(2)] LTonly-exp. Anomalies of model 850-hPa geopotential height (gpha in figure above) are shaded, while changes in the respective model layer wind (m s−1) are displayed as vectors. The magenta contours (0.15–0.45 with a spacing of 0.15) overlaying the LT response shown in [b(2)] are anomalies of the nonlinear SST-dependent conditional heating, δ [Eq. (6)]. The mean DJF BL 850-hPa geopotential height (shaded) and BL winds of the [c(1)] BLcoupled-exp and [c(2)] LTcoupled-exp and (d) the sum of both [c(1)] and [c(2)]. (e) Anomalies of model BL 850-hPa geopotential height (shaded) and BL winds from the unmodified model forced with the idealized SSTA.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) The idealized SSTA pattern. The mean DJF response of the [b(1)] BLonly-exp and [b(2)] LTonly-exp. Anomalies of model 850-hPa geopotential height (gpha in figure above) are shaded, while changes in the respective model layer wind (m s−1) are displayed as vectors. The magenta contours (0.15–0.45 with a spacing of 0.15) overlaying the LT response shown in [b(2)] are anomalies of the nonlinear SST-dependent conditional heating, δ [Eq. (6)]. The mean DJF BL 850-hPa geopotential height (shaded) and BL winds of the [c(1)] BLcoupled-exp and [c(2)] LTcoupled-exp and (d) the sum of both [c(1)] and [c(2)]. (e) Anomalies of model BL 850-hPa geopotential height (shaded) and BL winds from the unmodified model forced with the idealized SSTA.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
2) Lower-troposphere response
Here we investigate the LT response to the temporally fixed idealized ENSO forcing described above. This experiment is hereafter known as the LT-only experiment (LTonly-exp; see Table 1). Identifying the LT response to a fixed SSTA in the 2.5-layer model is a little more complicated than the BLonly-exp as the diabatic heating of the LT is partially controlled by the large-scale BL circulation and moisture convergence [Eq. (5)]. Thus, to decouple the LT from the BL we need to prescribe the BL circulation and moisture convergence to climatological values. This essentially means that we inhibit anomalous diabatic forcing due to BL Ekman pumping anomalies. Decoupling of LT and BL is achieved by prescribing the LT component of the 850-hPa geopotential height to the climatological value when calculating the BL winds (Fig. 7). We then add the fixed SSTA forcing to the climatological surface temperature Ts in Eq. (6) only. As such, it does not act as a direct BL forcing, which would act to alter the BL climatology and hence the diabatic heat source. The anomalous change in diabatic heating enters the LT via the nonlinear SST-dependent conditional heating term, δ, as Ts in this experiment is made up of the climatological SST plus the temporally fixed idealized SSTA. The anomalies in this conditional heating term effectively increase the area in which diabatic heating (convection) can occur and this diabatic heating is initially set by the climatological BL moisture convergence. This results in the adjustment of the large-scale LT circulation.
The anomalous diabatic forcing sets up a classical Gill (1980) LT equatorial Kelvin wave straddled between two Rossby waves [Fig. 8b(2)]. Looking at the western Pacific zonal mean of the anomalous zonal LT winds produced in this experiment (Fig. 6a), it becomes apparent that there is no significant southward shift toward the end of the calendar year. Thus, the response of the LTonly-exp is consistent with expectations given that the zonal equatorial wind anomalies are driven by the symmetric Kelvin wave component of the LT response. We note here that although there is no significant asymmetry in the zonal equatorial wind response (essentially the Kelvin wave component of the response), there is an asymmetry in the off-equatorial LT geopotential height (the Rossby wave component of the response) and the associated anomalous meridional winds [Fig. 8b(2)], which can be explained as a result of the nonlinearity of the diabatic forcing in response to a zonally symmetric SST anomaly and an asymmetric background climatological SST. We conclude from this experiment that even with a seasonally varying asymmetric diabatic forcing [Fig. 8b(2)], the free tropospheric equatorial zonal wind stress response is quasi-symmetric about the equator. As such, the LT response alone cannot reproduce the observed southward shift of ENSO-related zonal wind anomalies near the end of the calendar year.
3) Interactions between the boundary layer and the lower troposphere
We have established that neither the BL nor the LT alone can reproduce the magnitude of the observed southward shift of the zonal wind anomalies near the end of the calendar year. In both cases the establishment of an anomalous secondary circulation, which communicates between BL and LT via anomalous Ekman pumping was inhibited. Given the fact that the unmodified version of the model can reproduce this shift, interactions between the model BL and LT must be responsible. For instance, LT pressure anomalies, such as those seen in Fig. 8b(2), are expected to influence the BL circulation via changes in the BL pressure gradient, while the resulting changes in moisture, convergence, and evaporation will eventually feed back to the LT via diabatic heating. To elucidate the nature of these interactions between the two layers and to assess the role of secondary circulations, we carry out two experiments utilizing the unmodified model (i.e., with the BL and LT fully interactive) and the same temporally fixed idealized SSTA described above (Fig. 8a). As shown in the model schematic in Fig. 7, interaction between the two model components is initiated via the thermodynamic coupling between the BL and the LT. This linkage is represented in the LT diabatic heating equation [Eq. (5)]. The addition of a fixed SSTA to the prescribed SST climatology affects the LT diabatic forcing in two ways, either via (i) direct SSTA-forced changes in the BL circulation, evaporation, and moisture convergence; or (ii) changes in nonlinear SST-dependent conditional heating, δ [Eq. (6)]. Our experiments were devised to effectively separate the response with respect to these two processes.
In the first of these experiments, hereafter titled the BL coupled experiment (BLcoupled-exp), the time-independent SSTA perturbation (added to the mean climatological SST) is used to force the fully interactive model via the BL only. That is, the SSTA perturbation is not added to the climatological surface temperatures Ts in Eq. (6). As such, the area in which the BL moisture convergence can influence the LT remains the same. SSTA-induced changes in the BL circulation and moisture convergence and evaporation are allowed to drive LT circulation anomalies via Eq. (5). The LT anomalies then feed back to the BL via pressure anomalies. The BLcoupled-exp produces a BL response, which is dominated by a negative BL 850-hPa geopotential height anomaly in the eastern/central tropical Pacific [Fig. 8c(1)]. The response is qualitatively similar to the BLonly-exp described above [Fig. 8b(1)]. However, the BL-LT interactions clearly act to amplify the BL response to SSTA forcing as the amplitude of the 850-hPa geopotential height anomalies is significantly larger than in the BLonly-exp. This boundary layer geopotential height response remains relatively stationary, so it is not surprising that no significant latitudinal variability occurs in the ENSO-induced zonal wind anomalies (Fig. 6b).
In the second of these experiments, hereafter titled the LT coupled experiment (LTcoupled-exp), the time-independent idealized SSTA perturbation is only added to the nonlinear SST-dependent conditional heating [Eq. (6)]. The direct BL SSTA forcing is prescribed to climatology. Hence, we bypass the direct Lindzen–Nigam-type response that dominated the solution of BLcoupled-exp. The LT conditional heating anomalies drive changes in the LT circulation that are similar to those in the LT only experiment described above [Fig. 8b(2)]. However, in this new experiment the LT is allowed to feed back to the BL (i.e., the BL 850-hPa geopotential height is not prescribed to climatology). This feedback includes the potential establishment of anomalous secondary circulations. Pressure anomalies in the LT are expected to drive BL wind stress anomalies that will drive an anomalous Ekman pumping velocity. This in turn will provide further diabatic forcing to the LT through Eq. (5).
In Fig. 8c(2), we see that the BL wind response to the diabatically induced LT geopotential height anomaly of the LTcoupled-exp is essentially ageostrophic as the surface winds are approximately perpendicular to the LT geopotential height anomaly contours [Fig. 8b(2)]. This anomalous ageostrophic BL circulation is the surface branch of a secondary circulation cell that connects regions of anomalous Ekman pumping and suction. This secondary circulation leads to the most prominent feature in the response of the BL to this LT geopotential height anomaly—the southward shift of the ageostrophic BL wind anomalies during DJF (Fig. 6b). This feature and its accompanying BL 850-hPa height anomalies, which feature two areas of increased BL geopotential height straddling the equator on the 180° meridian [Fig. 8c(2)], are seasonally modulated (Figs. 9a,c,e,g). Note that the SSTA forcing in these experiments is constant in time.
The BL 850-hPa geopotential height (shaded) and BL winds of the LTcoupled-exp for (a) DJF, (c) MAM, (e) JJA, and (g) SON. The vertical pressure velocity (we, shaded) and BL wind winds of the LTcoupled-exp for (b) DJF, (d) MAM, (f) JJA, and (h) SON, while the overlying magenta contours display the anomalies of wind stress curl divided by the Coriolis force. Contours are drawn between ±9 × 10−4 with a spacing of 6 × 10−4, and negative contours are dashed.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The BL 850-hPa geopotential height (shaded) and BL winds of the LTcoupled-exp for (a) DJF, (c) MAM, (e) JJA, and (g) SON. The vertical pressure velocity (we, shaded) and BL wind winds of the LTcoupled-exp for (b) DJF, (d) MAM, (f) JJA, and (h) SON, while the overlying magenta contours display the anomalies of wind stress curl divided by the Coriolis force. Contours are drawn between ±9 × 10−4 with a spacing of 6 × 10−4, and negative contours are dashed.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The BL 850-hPa geopotential height (shaded) and BL winds of the LTcoupled-exp for (a) DJF, (c) MAM, (e) JJA, and (g) SON. The vertical pressure velocity (we, shaded) and BL wind winds of the LTcoupled-exp for (b) DJF, (d) MAM, (f) JJA, and (h) SON, while the overlying magenta contours display the anomalies of wind stress curl divided by the Coriolis force. Contours are drawn between ±9 × 10−4 with a spacing of 6 × 10−4, and negative contours are dashed.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The underlying dynamics become apparent when we compare the BL wind stress curl anomalies with the BL vertical pressure velocity anomalies (we) from the precipitation equation of the model [Eq. (5)]. The close match between vertical pressure velocity and the wind stress curl suggests that atmopsheric Ekman pumping dynamics are prevalent (Figs. 9b,d,f,h). The ageostrophic BL winds connect regions characterized by anomalous divergent atmospheric Ekman pumping (negative wind stress curl anomaly) with those of surface wind convergence east of the date line and between 20° and 10°S (positive wind stress anomaly). Given the similarities between the BL frictional terms and the quadratic wind stress law, the link between the BL vertical velocity and the BL wind stress curl becomes clear when calculating the curl of Eqs. (1) and (2), as it is seen that BL vertical velocity is roughly equal to the Ekman pumping due to the curl of the frictional terms (Fx and Fy). Thus, as friction is modulated through the seasonally varying wind speeds (Fig. 10), so too are the anomalies of frictional divergence/convergence in the boundary layer along with the forced anomalous secondary ageostrophic circulations and the related Ekman pumping velocities. In the coupled BL/LT system, these Ekman pumping anomalies feed back to the LT via diabatic heating [Eq. (5)], thereby providing a forcing for the LT, in addition to the SSTA forcing included in the conditional heating.
The mean surface wind (vectors, m s−1) and wind speed (shaded, m s−1) from the model for (top left) DJF and (middle left) SON. The ERA-40 mean surface wind (vectors) and wind speed (shaded) for (top right) DJF and (middle right) SON. The zonal mean (between 160°E and 150°W) (bottom left) intermediate model and (bottom right) ERA-40 wind speeds.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The mean surface wind (vectors, m s−1) and wind speed (shaded, m s−1) from the model for (top left) DJF and (middle left) SON. The ERA-40 mean surface wind (vectors) and wind speed (shaded) for (top right) DJF and (middle right) SON. The zonal mean (between 160°E and 150°W) (bottom left) intermediate model and (bottom right) ERA-40 wind speeds.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The mean surface wind (vectors, m s−1) and wind speed (shaded, m s−1) from the model for (top left) DJF and (middle left) SON. The ERA-40 mean surface wind (vectors) and wind speed (shaded) for (top right) DJF and (middle right) SON. The zonal mean (between 160°E and 150°W) (bottom left) intermediate model and (bottom right) ERA-40 wind speeds.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Looking at the anomalous BL geopotential height and wind response of the fixed SSTA-forced unmodified model displayed in Fig. 8e, it becomes evident that simulated atmospheric circulation anomalies due to (i) changes in direct BL SSTA forcing [i.e., the BLcoupled-exp; Fig. 8c(1)], and (ii) changes in the nonlinear SST-dependent conditional heating [i.e., the LTcoupled-exp; Fig. 8c(2)] both play a prominent role in the total response of the model. We note here that the DJF mean anomalous BL geopotential height and wind response of the fixed SSTA-forced unmodified model (Fig. 8e) can almost be completely reconstructed by adding the results of the above two experiments [i.e., Fig. 8c(1) + Fig. 8c(2) = Fig. 8d ≈ Fig. 8e]. This result also holds when looking at the response during individual calendar months (figure not shown).
Summarizing the results of Fig. 8, the direct SST forcing via the classical Lindzen–Nigam (LN) type response alters the BL geopotential height and wind pattern [Fig. 8b(1)]. Comparing Figs. 8b(1) and 8c(1), we see that the LN-related diabatic anomalies drive a LT response that further enhances the BL response. The LT Gill-type Rossby–Kelvin wave [Fig. 8b(2)] response to the SSTA forcing causes a BL response through pressure gradients that in conjunction with seasonally modulated friction establishes a seasonally modulated ageostrophic BL circulation anomaly [Fig. 8c(2)]. When friction is weak as in DJF/MAM in the southwestern equatorial Pacific, the wind response is stronger and more zonal. Wind stress curl anomalies generate Ekman pumping anomalies in the BL and the resulting diabatic forcing modifies the LT response pattern. Thus, the key feature of the southward wind shift during DJF and MAM essentially emerges from the coupling of BL/LT dynamics through conditional heating and seasonally modulated frictional convergence/divergence and its associated Ekman pumping.
To better demonstrate the effect of the climatological variations of wind speed on the southward shift of the wind response we conduct another sensitivity experiment. In this experiment the wind speed (Vs) component of the BL turbulent momentum flux [Eqs. (3) and (4)] is set to a spatially and temporally fixed value. The momentum equation is still fully prognostic. With only climatological SST forcing this change in the seasonal cycle of momentum dissipation leads to increased simulated wind speeds over most of the model domain; however, a southward shift of mean simulated climatological wind speed similar to that displayed in Fig. 10 still occurs (not shown). Adding a fixed anomalous El Niño–like SSTA to the climatological SSTs results in a wind response that lacks the southward shift during the end of the calendar year (Fig. 6c). Thus, the results of this experiment confirm that climatological variations of wind speed and their effect on BL friction and Ekman pumping are responsible for the southward shift of ENSO-related zonal wind anomalies in DJF.
4. The ocean model and experiment results
As discussed in the introduction, numerous theories have been proposed to explain the termination of ENSO events. One of the most prominent is the recharge/discharge oscillator theory of Jin (1997), which invokes Sverdrup dynamics as the delayed oceanic feedback that triggers the demise of ENSO events. This theory views ENSO as an east–west-tilting mode of the equatorial thermocline, which leads to the development of SST anomalies in the eastern equatorial Pacific, and the recharging/discharging mode, associated with longer-term ocean memory that affects the zonal mean equatorial thermocline depth. Meinen and McPhaden (2000) demonstrated using observational data that these two features of the recharge/discharge oscillator theory are represented by the first and second EOF, respectively, of 20°C isotherm depth (see their Fig. 3).
This result is reproduced here (see Fig. 11) by carrying out an EOF analysis of monthly mean equatorial Pacific (between 20°N–20°S and 100°E–60°W) upper-ocean heat content anomalies (vertically averaged temperature in the upper 300 m) of the ECMWF operational ocean analysis/reanalysis system (ORA-S3; Balmaseda et al. 2008). As in the observational analysis of Meinen and McPhaden (2000), the peak correlation (0.47) between the first two ORA-S3 mode PC time series occurs when the first PC lags the second by 9 months. While statistically significant, this correlation coefficient is relatively weak compared to theoretically expected values (Jin 1997). This could indicate that the second EOF mode is an imperfect indicator of the actual recharge/discharge process or that the observed recharge/discharge of equatorial heat content involves more than simple ocean dynamics. Furthermore, an interesting feature of the second EOF (EOF2) mode (i.e., the recharge/discharge mode) both in the Meinen and McPhaden (2000) manuscript and here, which is not explained by the recharge/discharge oscillator theory, is its meridionally asymmetric spatial structure (Fig. 11b). The spatial structure of this mode is characterized by a north–south (N–S) tilting along an axis centered near the ITCZ at approximately 5°N. Understanding what determines the spatial structure of EOF2 could help further elucidate the termination mechanisms of ENSO events. Comparing the ORA-S3 PC time series (Figs. 11b,c, solid black lines) with those of the ERA-40 wind stresses presented in section 2 of this manuscript (Figs. 11b,c, dashed red lines) reveals an interesting similarity. That is, the PC time series of the ORA-S3 first (second) EOF mode is highly correlated with the PC time series of the ERA-40 first (second) EOF mode of combined zonal and meridional wind stress. The respective correlation coefficients are 0.83 and 0.62, which are both statistically significant above the 99% level. In this section we use a simple ocean model to show that the southward shift of ENSO-related surface wind anomalies near the end of the calendar year (which is represented by the second EOF of the ERA-40 wind stresses) plays a key role in setting up the asymmetric structure of this second (recharge) mode and in terminating El Niño events.
(a) EOF1 (37% variance) and (b) EOF2 (15% variance) of ORA-S3 vertically averaged temperature in the upper 300 m. (c),(d) The corresponding principal component time series is indicated by the solid black lines. The blue lines in (c),(d) display the full ERA-40 forcing SWM first and second EOF principal component time series, respectively; while the red lines in both display the time series of the ERA-40 wind stress first and second PCs, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) EOF1 (37% variance) and (b) EOF2 (15% variance) of ORA-S3 vertically averaged temperature in the upper 300 m. (c),(d) The corresponding principal component time series is indicated by the solid black lines. The blue lines in (c),(d) display the full ERA-40 forcing SWM first and second EOF principal component time series, respectively; while the red lines in both display the time series of the ERA-40 wind stress first and second PCs, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) EOF1 (37% variance) and (b) EOF2 (15% variance) of ORA-S3 vertically averaged temperature in the upper 300 m. (c),(d) The corresponding principal component time series is indicated by the solid black lines. The blue lines in (c),(d) display the full ERA-40 forcing SWM first and second EOF principal component time series, respectively; while the red lines in both display the time series of the ERA-40 wind stress first and second PCs, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
The ocean model used here to further explore the role of wind stress forcing on ENSO-related thermocline dynamics is a 1.5-layer reduced gravity shallow-water model (SWM) of the stratified ocean. The upper and lower layers of the model are separated by an interface that approximates the sharp tropical thermocline separating the warm surface waters from the cold waters of the deep ocean. Motion in the upper layer is driven by the applied anomalous wind stresses (per unit density), while the lower layer is assumed motionless and infinitely deep. The associated response of the ocean to wind stress changes is characterized by the vertical displacement of the thermocline and the horizontal components of the upper-layer flow velocity. The upper-layer ocean dynamics are described by the linear reduced-gravity form of the shallow-water equations (McGregor et al. 2007; Holbrook et al. 2011). We prescribe the reduced-gravity parameter g′ = 0.0265 m s−2 and the mean depth of the upper layer as 300 m, so the first baroclinic-mode Kelvin wave speed is 2.8 m s−1. The model has a 1° horizontal resolution and is configured for the low- to midlatitude Indo-Pacific Ocean (51°S and 51°N, 40°E and 60°W). It also includes realistic continental boundaries that were calculated as the locations where the bathymetric dataset of Smith and Sandwell (1997) has a depth of less than the model mean thermocline depth of 300 m.
a. Model validation
To ascertain whether the SWM is an appropriate tool to investigate the causes of the observed asymmetric recharge mode, we carry out a simulation in which the model is forced by monthly ERA-40 wind stress anomalies for the period 1958–2001. These are the same wind stress data used to force the ORA-S3 model. A point-by-point correlation between the ORA-S3 upper-ocean heat content and the SWM thermocline depth reveals correlations of over 0.7 for the majority of the equatorial Pacific Ocean between 10°N–10°S and 120°E–60°W (Fig. 12). As such, the SWM simulation reproduces a reasonable amount of the ORA-S3 equatorial Pacific Ocean interannual thermocline variance.
Correlation coefficients between anomalies of ORA-S3 upper-ocean temperature (vertically averaged between the surface and 300 m) and the modeled SWM thermocline depth. The dashed black and solid black lines represent the 0.5 and 0.75 correlation coefficient contours, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Correlation coefficients between anomalies of ORA-S3 upper-ocean temperature (vertically averaged between the surface and 300 m) and the modeled SWM thermocline depth. The dashed black and solid black lines represent the 0.5 and 0.75 correlation coefficient contours, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Correlation coefficients between anomalies of ORA-S3 upper-ocean temperature (vertically averaged between the surface and 300 m) and the modeled SWM thermocline depth. The dashed black and solid black lines represent the 0.5 and 0.75 correlation coefficient contours, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Conducting an EOF analysis of the resultant SWM thermocline depth in the equatorial Pacific (between 20°N–20°S and 100°E–60°W), we observe that the first EOF of simulated thermocline anomalies (Fig. 13a) displays an east–west-tilting mode with a similar spatial structure, although meridionally broader in the eastern equatorial Pacific, to that obtained from the ORA-S3 data (see Fig. 11a). The corresponding principal component time series has a correlation coefficient of 0.91 (statistically significant above the 99% level) when compared with the first principal component time series of the ORA-S3 data. Comparing the SWM EOF2 with that of the ORA-S3 data it is observed both have similar spatial structure throughout most of the Pacific basin. The principal component time series of the second SWM EOF has a correlation coefficient of 0.92 (statistically significant above the 99%) when compared with the corresponding time series of the ORA-S3 data. Thus, the dynamics of the asymmetric spatial structure of the recharge mode appear to be reasonably well reproduced by a linear ocean model forced with observed wind stress anomalies. As such, the SWM is an appropriate tool to further investigate what processes are responsible for the asymmetric structure of the recharge mode.
(left column) The first and (right column) second EOF modes of SWM thermocline depth from the full forcing experiment are displayed in (a) and (b) respectively, while those of PCexp1 [(PCexp1+2)] are displayed in (b) and (c) [(e) and (f)] respectively. (g) The associated principal component time series of the full ERA-40 forcing experiment first (37% variance) and second (17% variance) EOFs indicated by the black line and dashed gray line, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(left column) The first and (right column) second EOF modes of SWM thermocline depth from the full forcing experiment are displayed in (a) and (b) respectively, while those of PCexp1 [(PCexp1+2)] are displayed in (b) and (c) [(e) and (f)] respectively. (g) The associated principal component time series of the full ERA-40 forcing experiment first (37% variance) and second (17% variance) EOFs indicated by the black line and dashed gray line, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(left column) The first and (right column) second EOF modes of SWM thermocline depth from the full forcing experiment are displayed in (a) and (b) respectively, while those of PCexp1 [(PCexp1+2)] are displayed in (b) and (c) [(e) and (f)] respectively. (g) The associated principal component time series of the full ERA-40 forcing experiment first (37% variance) and second (17% variance) EOFs indicated by the black line and dashed gray line, respectively.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
b. Cause of asymmetric recharge mode
Here we utilize the ERA-40 wind stress that were earlier decomposed into PC modes in section 2 (Fig. 2) to force the SWM in a series of experiments. The first of these experiments utilizes wind stress forcing reconstructed from only the first PC mode of the ERA-40 wind stresses (hereafter PCexp1). Each subsequent experiment carried out utilizes an extra PC mode in the reconstruction of the forcing wind stresses (i.e., PCexp1 + 2 = PC1 + PC2, PCexp1 + 2 + 3 = PC1 + PC2 + PC3, …). This experimental setup allows us to identify how many PC wind stress modes are needed to reproduce the asymmetric recharge mode, thereby providing us with insight into the underlying dynamics.
Results from this set of experiments reveal that only the first and second ERA-40 wind stress PC modes (PCexp1 + 2) are needed to reproduce the simulated asymmetric recharge mode of the Pacific Ocean (Figs. 13e,f). The first and second modes of ocean variability from this experiment have nearly identical spatial structures to the respective modes of the full ERA-40 SWM forcing experiment (cf. Figs. 13a,b). The corresponding principal components (not shown) produce correlation coefficients of 0.95 and 0.94, respectively, when compared to the principal components of the full ERA-40 SWM forcing experiment (both correlation coefficients are statistically significant above the 99% level). The first mode on its own (PCexp1) produces the east–west tilting of the thermocline as the dominant EOF mode [Fig. 13b]. However, the second (recharge) mode obtained in PCexp1 is quite symmetric around the equator [Fig. 13b] and resembles the idealized model solutions of a symmetric Rossby wave forced by a stationary wind pattern (Neelin et al. 1998; Fedorov 2010). The corresponding principal components produce correlation coefficients of 0.85 (statistically significant above the 99% level) and 0.70, respectively, when compared to the principal components of the full ERA-40 SWM forcing experiment. Thus, while the second mode of PCexp1 still represents a recharge-type mode (a maximum correlation coefficient of 0.91 when EOF2 is leading EOF1 by 9 months), it does not have the asymmetric spatial features seen in the observations, nor does it correlate as well with the ORA-S3 and the full ERA-40-forced SWM-simulated estimates of equatorial heat content. Adding the second ERA-40 wind stress PC mode in the forcing of PCexp1 + 2 allows the SWM to better reproduce the temporal variability and the N–S spatial asymmetry seen the second EOF (recharge) mode of the observational estimates (ORA-S3) and the full ERA-40 forcing experiment presented above.
To further demonstrate the effect of the second ERA-40 wind stress PC mode on the zonal mean changes in equatorial thermocline depth and the discharging process, we conduct another SWM experiment, which is forced by only the second PC mode of the ERA-40 wind stresses (referred to as PCexp2). We then calculate the zonal mean changes of equatorial region thermocline depth from the full ERA-40-forced SWM simulation along with that from PCexp1 and PCexp2. Confirming the SWM EOF analysis results presented above, the zonal mean thermocline depth changes (a measure for recharge/discharge dynamics) of the full ERA-40-forced simulation can almost be fully reproduced by adding the PCexp1 and PCexp2 simulation zonal mean thermocline depth changes (Fig. 14a, correlation coefficient of 0.85 which is statistically significant above the 99% level). Visual analysis of the role of the individual components, displayed in Fig. 14b, reveals that both wind stress PC modes (the PCexp1 and PCexp2 experiment results) play an essential role in the zonal mean thermocline depth changes of the full ERA-40 forced simulations. This is confirmed when looking at the variance of each time series or their correlation with the full ERA-40-forced SWM simulation. The zonal mean thermocline depths of PCexp1 (PCexp2) have a variance of 8.33 m2 (6.77 m2), while the zonal mean thermocline depth anomalies of PCexp1 (PCexp2) have a correlation of 0.70 (0.64) when compared with those of the full ERA-40-forced simulation (both correlation coefficients are statistically significant above the 99% level). Interestingly though, composites of zonal mean equatorial thermocline depth displayed in Figs. 14c–e reveal that the effective discharge of equatorial heat content in the 4–5 months surrounding the peak of CT El Niño events is clearly dominated by the second PC of the ERA-40 wind stress anomalies and not the Rossby wave contribution that characterizes the classical discharging. As discussed above, wind stress anomalies associated with the first two ERA-40 wind stress PCs act in quadrature to allow the ENSO-related wind stress anomalies to propagate southward toward the end of the calendar year. Thus, the spatial asymmetry in the Pacific Ocean recharge mode along with a significant portion of the zonal mean changes in thermocline depth and the acceleration and synchronization of the discharge process toward DJF/MAM are a consequence of the southward shift of the ENSO-related wind stress anomalies.
(a) The zonal mean equatorial thermocline depth from the full ERA-40 forced SWM simulation (solid black) and the sum of the PCexp1 and PCexp2 zonal mean thermocline depth anomalies (solid gray). (b) The PCexp1 and PCexp2 zonal mean thermocline depth anomalies are shown separately with solid blue and solid red lines, respectively. The composite mean zonal mean equatorial thermocline depth composited around (c) WP (1977/78, 1990/91, and 1994/95), (d) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (e) mixed (1986/87, 1987/88, and 1991/92) type El Niño events, where month 0 refers to January during the El Niño event peak.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) The zonal mean equatorial thermocline depth from the full ERA-40 forced SWM simulation (solid black) and the sum of the PCexp1 and PCexp2 zonal mean thermocline depth anomalies (solid gray). (b) The PCexp1 and PCexp2 zonal mean thermocline depth anomalies are shown separately with solid blue and solid red lines, respectively. The composite mean zonal mean equatorial thermocline depth composited around (c) WP (1977/78, 1990/91, and 1994/95), (d) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (e) mixed (1986/87, 1987/88, and 1991/92) type El Niño events, where month 0 refers to January during the El Niño event peak.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
(a) The zonal mean equatorial thermocline depth from the full ERA-40 forced SWM simulation (solid black) and the sum of the PCexp1 and PCexp2 zonal mean thermocline depth anomalies (solid gray). (b) The PCexp1 and PCexp2 zonal mean thermocline depth anomalies are shown separately with solid blue and solid red lines, respectively. The composite mean zonal mean equatorial thermocline depth composited around (c) WP (1977/78, 1990/91, and 1994/95), (d) CT (1972/73, 1976/77, 1982/83, and 1997/98), and (e) mixed (1986/87, 1987/88, and 1991/92) type El Niño events, where month 0 refers to January during the El Niño event peak.
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
We finally note that the zonal mean thermocline depth changes of the ERA-40 forced control SWM simulation along with the PCexp1 + 2 and PCexp2 forced simulations are highly correlated with the principal components of the second PC of the ERA-40 wind stresses with zero lead/lag. Correlation coefficients of 0.64, 0.73, and 0.86 are produced, respectively, of which all are statistically significant above the 99% level.
c. Role of western boundary reflection
The high correlation between the SWM- or ORA-S3-simulated recharge mode (EOF2) and PC2 of the ERA-40 wind stresses implies that ocean dynamics, specifically due to the western boundary reflection of the gravest equatorial Rossby waves, are not required for the recharge/discharge of equatorial heat content during El Niño events. To examine this hypothesis we carry out two wind stress–forced SWM simulations, one of which has the same wind stress forcing as PCexp1, while the other has the same wind stress forcing as PCexp2. The difference between the current experiments and those presented above (section 4c) is that in the current experiment we extend the width of the Pacific basin by moving the Australasian continents 60° to the west (these experiments are referred to as PCexp1-AUS and PCexp2-AUS). If western boundary reflections were important for the zonal mean changes in equatorial thermocline depth we would expect the zonal mean equatorial thermocline depth variations of the experiment to lag those with the original continental mask by 3–4 months. These experiments are somewhat similar to those presented in Harrison and Vecchi (1999). However, as we have shown that the first two PCs of ERA-40 equatorial wind stresses both play prominent roles in the changes in equatorial heat content, we can assess the role of western boundary reflection on these individual components separately.
The results of the wind stress PC1-forced simulation reveal a maximum correlation coefficient of 0.74 is attained when the zonal mean equatorial thermocline depth of the experiment PCexp1-AUS is shifted forward by 4 months relative to the original PCexp1 experiment (figure not shown). This indicates that the zonal mean changes in equatorial thermocline depth forced by PC1 of the ERA-40 wind stresses do rely on the classical ocean dynamics described by the recharge/oscillator paradigm of Jin (1997). However, the results of PCexp2-AUS exhibit a maximum correlation of 0.98 at zero lag between simulated zonal mean thermocline depth changes and those obtained in PCexp2, thus revealing that the reflection of Rossby waves plays virtually no role in the zonal mean equatorial thermocline depth changes induced by the southward shift of ENSO-related wind stresses. Instead the asymmetric component of the recharge mode and its zonal mean change in thermocline depth are a direct response to oceanic Ekman pumping and Kelvin waves induced by the southward-displaced wind stress anomalies. Note also that the wind stress curl anomalies that are responsible for the oceanic Ekman pumping and the asymmetric thermocline anomalies in the southwestern tropical Pacific are the same anomalies that drive the anomalous atmospheric Ekman pumping and the secondary circulations discussed in section 3 and shown in Fig. 9. Thus, an essential element in the termination of strong (CT) El Niño events is the coupling between the thermocline and atmospheric circulation, which changes seasonally via the climatological wind speed–modulated wind stress curl anomalies.
5. Conclusions
This study demonstrates that the seasonally paced southward shift of surface wind anomalies during El Niño events, described by Harrison (1987) and Harrison and Larkin (1998), is well represented by the quadrature of the first two PCs of ERA-40 surface wind stresses in the equatorial Pacific basin. This southward wind shift is shown to be a key component in the termination of observed and simulated El Niño events (Harrison and Vecchi 1999; Vecchi and Harrison 2003; Lengaigne et al. 2006; Vecchi and Harrison 2006; Lengaigne and Vecchi 2009) and has been suggested as one of the mechanisms responsible for the apparent synchronization of ENSO events to the seasonal cycle (Harrison and Vecchi 1999).
Using the atmospheric model of Fu and Wang (1999), which couples an LN-type (Lindzen and Nigam 1987) model of the atmospheric boundary layer (BL) with a Gill-type (Gill 1980) model of the lower troposphere (LT), we have identified the mechanisms that tie this well-documented southward wind shift to the seasonal cycle during CT El Niño events. While generating significantly weaker anomalous wind stresses and their associated curl than those of the ERA-40 reanalysis, the similarities between the modeled and observed (ERA-40) western Pacific anomalous winds and climatological wind speeds (see Fig. 10) suggest that the mechanism identified here may also be operating in reality. When the model is separated into its individual components we find that neither the boundary layer nor the tropospheric model alone can reproduce the southward shift of ENSO-related zonal wind anomalies. Instead interactions between the BL and LT are essential for the generation of this southward shift.
Figure 15 summarizes the main findings from the intermediate complexity atmospheric model. Imposing an idealized equatorially symmetric El Niño SST anomaly reduces the sea level pressure in the eastern equatorial Pacific, while at the same time triggering a classical Gill-type Rossby–Kelvin wave in the LT. The LT geopotential height perturbation induces a secondary circulation characterized by an ageostrophic BL flow that creates two centers of high sea level pressure that straddle the equator near the 180° meridian. These two features combine to give the total surface wind and pressure response. The overall anomalous surface wind response is amplified in the South Pacific during DJF/MAM, compared to the North Pacific, due to the reduced climatological wind speeds and the related BL Ekman pumping in the region. The direct link between atmospheric Ekman pumping and rainfall in regions with SST above the threshold for convection along with the coevolving nature of these phenomena, mean that the results of this study are consistent with recent studies, which allude to the DJF/MAM southward shift of anomalous convection as the cause of the southward zonal wind anomaly shift (Spencer 2004; Lengaigne et al. 2006; Vecchi 2006). We note that the South Pacific region of low climatological wind speed and reduced momentum dissipation is also the region of largest climatological wind convergence—the SPCZ. Hence, we conclude that the development of a strong climatological SPCZ in DJF/MAM and the associated weakening of BL wind speeds is one of the key factors in the seasonal termination of strong (CT) El Niño events.
Schematic diagram showing the mechanism responsible for the southward shift of El Niño–related zonal wind anomalies. Lower troposphere (LT) , boundary layer (BL), sea level pressure (SLP), and eastern equatorial Pacific (EEP).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Schematic diagram showing the mechanism responsible for the southward shift of El Niño–related zonal wind anomalies. Lower troposphere (LT) , boundary layer (BL), sea level pressure (SLP), and eastern equatorial Pacific (EEP).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
Schematic diagram showing the mechanism responsible for the southward shift of El Niño–related zonal wind anomalies. Lower troposphere (LT) , boundary layer (BL), sea level pressure (SLP), and eastern equatorial Pacific (EEP).
Citation: Journal of Climate 25, 16; 10.1175/JCLI-D-11-00332.1
One of the key mechanisms identified here for the southward shift of the wind anomalies is a direct dynamical link between an ENSO-induced Gill-type Rossby–Kelvin wave in the LT and the sea level pressure in the off-equatorial western North and South Pacific. The simulated off-equatorial sea level pressure anomalies in the western Pacific Ocean are consistent with observations of SLPA around El Niño events (Harrison and Larkin 1996). Wang (1995) found that the anomalous sea level pressure in these off-equatorial regions of the western Pacific (including the Philippine anticyclone) are highly correlated with indices of ENSO. They have also been proposed by Weisberg and Wang (1997) in the western Pacific oscillator theory, and Wang et al. (1999) as a possible mechanism for the termination of ENSO events, as they are linked with near-equatorial easterly winds. The mechanism proposed here to explain the dynamical linkage between the LT and BL provides a different perspective on these off-equatorial atmospheric anomalies that play a key role in the termination of El Niño events.
Experiments with a linear reduced gravity shallow-water ocean model were then conducted to illustrate the linkage between the seasonally paced southward wind shift during an El Niño event and the asymmetric discharge of equatorial heat content associated with the termination of the event. These experiments show that EOF2 of the SWM thermocline depth, which is qualitatively similar to the recharge/discharge mode presented in Meinen and McPhaden (2000), is essentially caused by two unrelated processes:
The first of these processes is predominantly a direct oceanic (Ekman pumping/equatorial Kelvin wave type) response to the southward shift of El Niño–related wind stresses during DJF. This component is directly responsible for the asymmetric structure of the observed recharge mode (Meinen and McPhaden 2000). Furthermore, consistent with this quasi-direct relationship suggested by our experimentation, we find a significant correlation between the PC2 of the ERA-40 surface wind stresses and the recharge mode (EOF2) of ORA-S3 ocean reanalysis vertically averaged temperature (correlation coefficient of 0.62, statistically significant above the 99% level).
The second component is the classic ocean dynamically induced recharge/discharge of equatorial heat content consistent with the recharge/discharge oscillator paradigm of Jin (1997). As expected from its links with the Delayed Action Oscillator paradigm for ENSO (Schopf and Suarez 1988; Battisti and Hirst 1989), our experimentation reveals that this component is somewhat reliant on western boundary reflection of equatorial Rossby waves.
The question of which of these two processes is mainly responsible for the discharge of equatorial heat content during the termination of an El Niño event is dependent on the type of El Niño. For instance, as expected due to the lack of a prominent southward wind shift during WP El Niño events, PC1 wind stress forcing (classic ocean dynamic discharge) is primarily responsible for the changes of equatorial heat in the 6-months after the event peak during these events. During mixed-type El Niño events on the other hand, the southward wind shift works together with wave dynamics to enhance the discharge process. The prominent southward shift of ENSO-related wind stresses (wind stress PC2 forced) during CT El Niño events is predominately responsible for the discharge of heat content in the 4–5 months surrounding the event peak. After this time the classical ocean dynamical discharge (wind stress PC1 forced) dominates the heat content discharge. This suggests that the southward shift of ENSO-related wind stress anomalies related to the seasonal development of the SPCZ during DJF/MAM has a considerable contribution to the termination of mixed-type El Niño events and is almost solely responsible for the rather abrupt seasonally synchronized termination of strong (CT type) El Niño events. It also indicates that without this southward shift, strong (CT) El Niño events would last significantly longer and not necessarily have a termination phase tied to the seasonal cycle.
The results of our SWM experimentation are consistent with earlier studies that show that the seasonally paced southward wind shift plays a prominent role in the termination of El Niño events (Harrison and Vecchi 1999; Vecchi and Harrison 2003, 2006) and hence, the seasonal cycle phase synchronization of El Niño event termination. Our study has further revealed the atmospheric processes involved in the latitudinal wind shift and has provided a means to quantify the relative contributions of this process to the termination of El Niño events in comparison with the classical Rossby wave–induced discharge theory. Our results also suggest that coupled GCMs that do not represent this southward shift of ENSO-related winds are likely to be biased toward longer El Niño duration and as a result this may affect the model’s predictive skill. Furthermore, our results also effectively link the results of these previous studies with our analytical understanding of ENSO, and suggest that changes in the analytical theories of ENSO are needed to properly explain the observed changes in equatorial heat content and its synchronization to the seasonal cycle.
Acknowledgments
We thank Dr. Fu and Dr. Wang from the IPRC for making the intermediate atmosphere model code available to us and M. J. McPhaden for originally proposing the ENSO asymmetry question at an ENSO summer school in 2008. We are also grateful for initial discussions with S. B. Power and numerous stimulating discussions with F.-F. Jin. A. Timmermann and S. McGregor are supported by NOAA through Grants NA08OAR4320910 and DE-FG02-07ER64469 of the Office of Science (BER), U.S. Department of Energy. N. Schneider is supported by Grants De-FG02-07ER64469 of the Office of Science (BER), U.S. Department of Energy and OCE05-50233 of the National Science foundation. M. F. Stuecker is supported by the U.S. NSF through Grant ATM1034798 and U.S. Department of Energy through Grant DESC005110. M. England is supported by the Australian Research Council.
REFERENCES
Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata, 2007: El Niño Modoki and its possible teleconnection. J. Geophys. Res., 112, C11007, doi:10.1029/2006JC003798.
Balmaseda, M. A., A. Vidard, and D. L. T. Anderson, 2008: The ECMWF Ocean Analysis System: ORA-S3. Mon. Wea. Rev., 136, 3018–3034.
Battisti, D. S., and A. C. Hirst, 1989: Interannual variability in the tropical atmosphere–ocean system: Influences of the basic state, ocean geometry, and nonlinearity. J. Atmos. Sci., 46, 1687–1712.
Bosc, C., and T. Delcroix, 2008: Observed equatorial Rossby waves and ENSO-related warm water volume changes in the equatorial Pacific Ocean. J. Geophys. Res., 113, C06003, doi:10.1029/2007JC004613.
Boulanger, J. P., S. Cravatte, and C. Menkes, 2003: Reflected and locally wind-forced interannual equatorial Kelvin waves in the western Pacific Ocean. J. Geophys. Res., 108, 3311, doi:10.1029/2002JC001760.
Chang, P., and Coauthors, 2006: Climate fluctuations of tropical coupled systems—The role of ocean dynamics. J. Climate, 19, 5122–5174.
Davis, R., 1976: Predictability of sea surface temperature and sea level pressure anomalies over the North Pacific Ocean. J. Phys. Oceanogr., 6, 249–266.
Delcroix, T., B. Dewitte, Y. duPenhoat, F. Masia, and J. Picaut, 2000: Equatorial waves and warm pool displacements during the 1992–1998 El Niño Southern Oscillation events: Observation and modeling. J. Geophys. Res., 105 (C11), 26 045–26 062.
Fedorov, A. V., 2010: Ocean response to wind variations, warm water volume, and simple models of ENSO in the low-frequency approximation. J. Climate, 23, 3855–3873.
Fu, X., and B. Wang, 1999: The role of longwave radiation and boundary layer thermodynamics in forcing tropical surface winds. J. Climate, 12, 1049–1069.
Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106, 447–462.
Harrison, D. E., 1987: Monthly mean island surface winds in the central tropical Pacific and El Niño events. Mon. Wea. Rev., 115, 3133–3145.
Harrison, D. E., and N. K. Larkin, 1996: The COADS sea level pressure signal: A near-global El Niño composite and time series view, 1946–1993. J. Climate, 9, 3025–3055.
Harrison, D. E., and N. K. Larkin, 1998: EL Niño–Southern Oscillation sea surface temperature and wind anomalies, 1946–1993. Rev. Geophys., 36, 353–399.
Harrison, D. E., and G. A. Vecchi, 1999: On the termination of El Niño. Geophys. Res. Lett., 26, 1593–1596.
Holbrook, N. J., I. D. Goodwin, S. McGregor, E. Molina, and S. Power, 2011: ENSO to multi-decadal time scale changes in East Australian Current transports and Fort Denison sea level: Oceanic Rossby waves as the connecting mechanism. Deep-Sea Res. II, 58, 547–558, doi:10.1016/j.dsr2.2010.06.007.
Jin, F., 1997: An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual model. J. Atmos. Sci., 54, 811–829.
Kao, H., and J. Yu, 2009: Contrasting eastern Pacific and central Pacific types of ENSO. J. Climate, 22, 615–632.
Kug, J., F. Jin, and S. An, 2009: Two types of El Niño events: Cold tongue El Niño and warm pool El Niño. J. Climate, 22, 1499–1515.
Larkin, N. K., and D. E. Harrison, 2002: ENSO warm (El Niño) and cold (La Niña) event life cycles: Ocean surface anomaly patterns, their symmetries, asymmetries, and implications. J. Climate, 15, 1118–1140.
Larkin, N. K., and D. E. Harrison, 2005: On the definition of El Niño and associated seasonal average U.S. weather anomalies. Geophys. Res. Lett., 32, L16705, doi:10.1029/2005GL022738.
Lengaigne, M., and G. Vecchi, 2009: Contrasting the termination of moderate and extreme El Niño events in coupled general circulation models. Climate Dyn., 35, 299–313, doi:10.1007/s00382-009-0562-3.
Lengaigne, M., J. Boulanger, C. Meinkes, and H. Spencer, 2006: Influence of the seasonal cycle on the termination of El Niño events in a coupled general circulation model. J. Climate, 19, 1850–1868.
Lindzen, R., and S. Nigam, 1987: On the role of sea surface temperature gradients in forcing low-level winds and convergence in the tropics. J. Atmos. Sci., 44, 2418–2436.
McGregor, S., N. J. Holbrook, and S. B. Power, 2007: Interdecadal sea surface temperature variability in the equatorial Pacific Ocean. Part I: The role of off-equatorial wind stresses and oceanic Rossby waves. J. Climate, 20, 2643–2658.
McPhaden, M. J., and X. Yu, 1999: Equatorial waves and the 1997–98 El Niño. Geophys. Res. Lett., 26, 2961–2964.
Meinen, C. S., and M. J. McPhaden, 2000: Observations of warm water volume changes in the equatorial Pacific and their relationship to El Niño and La Niña. J. Climate, 13, 3551–3559.
Neelin, J. D., D. S. Battisti, A. C. Hirst, F. Jin, Y. Wakata, T. Yamagata, and S. E. Zebiak, 1998: ENSO theory. J. Geophys. Res., 103, 14 261–14 290.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 659–706.
Picaut, J., F. Masia, and Y. du Penhoat, 1997: An advective-reflective conceptual model for the oscillatory nature of the ENSO. Science, 277, 663–666.
Rayner, N., D. Parker, E. Horton, C. Folland, L. Alexander, D. Rowell, E. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.
Schopf, P. S., and M. J. Suarez, 1988: Vacillations in a coupled ocean–atmosphere model. J. Atmos. Sci., 45, 549–566.
Smith, W., and D. Sandwell, 1997: Global seafloor topography from satellite altimetry and ship depth soundings. Science, 277, 1956–1962.
Spencer, H., 2004: Role of the atmosphere in seasonal phase locking of El Niño. Geophys. Res. Lett., 31, L24104, doi:10.1029/2004GL021619.
Trenberth, K., and D. P. Stepaniak, 2001: Indices of El Niño evolution. J. Climate, 14, 1697–1701.
Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131, 2961–3012.
Vecchi, G., 2006: The termination of the 1997/98 El Niño. Part II: Mechanisms of atmospheric change. J. Climate, 19, 2647–2664.
Vecchi, G., and D. E. Harrison, 2003: On the termination of the 2002–03 El Niño event. Geophys. Res. Lett., 30, 1964, doi:10.1029/2003GL017564.
Vecchi, G., and D. E. Harrison, 2006: The termination of the 1997/98 El Niño. Part I: Mechanisms of oceanic change. J. Climate, 19, 2633–2646.
Waliser, D. E., N. E. Graham, and C. Gautier, 1993: Comparison of highly reflective cloud and outgoing longwave radiation datasets for use in estimating tropical deep convection. J. Climate, 6, 331–353.
Wang, B., 1995: Transition from a cold to a warm state of the El Niño–Southern Oscillation. Meteor. Atmos. Phys., 56, 17–32.
Wang, B., and T. Li, 1993: A simple tropical atmosphere model of relevance to short-term climate variations. J. Atmos. Sci., 50, 260–284.
Wang, B., R. Wu, and R. Lukas, 1999: Roles of the western North Pacific wind variation in thermocline adjustment and ENSO phase transition. J. Meteor. Soc. Japan, 77, 1–16.
Wang, C., and J. Picaut, 2004: Understanding ENSO physics: A review. Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 21–48.
Weisberg, R. H., and C. Wang, 1997: A western Pacific oscillator paradigm for the El Niño-Southern Oscillation. Geophys. Res. Lett., 24, 779–782.