HTML
--> --> -->In addition to the amplitude asymmetry, the characteristics of the evolution of El Ni?o and La Ni?a events during their decay phases are markedly different (Kessler, 2002; Larkin and Harrison, 2002; McPhaden and Zhang, 2009). Generally, El Ni?o events tend to turn into La Ni?a events in the following June-July after their mature phases; however, the negative SSTAs associated with La Ni?a events can persist for more than one year after peaking, and tend to strengthen again in the next winter (Okumura and Deser, 2010; Okumura et al., 2011), resulting in a longer duration than that of El Ni?o. (DiNezio and Deser, 2014) pointed out that a large fraction (35%-50%) of La Ni?a events can sustain for more than two years. Such remarkable differences between the characteristics of the evolution of La Ni?a and El Ni?o thus challenges traditional ENSO cycle theories (Suarez and Schopf, 1988; Battisti and Hirst, 1989; Jin, 1997) and ENSO forecast (Jin and Kinter III, 2009; Ohba and Watanabe, 2012).
The evolution of ENSO is tightly connected with the zonal wind stress anomalies over the equatorial western Pacific (WP). (Zhang and Huang, 1998) pointed out that the intensity of the zonal wind stress anomalies over the equatorial WP are closely related to the termination of ENSO, because the anomalous easterly over the equatorial WP during the mature phase of El Ni?o can stimulate cold equatorial Kelvin waves by upwelling and cooling, leading to the transition from El Ni?o to La Ni?a (Huang et al., 2001; Yan et al., 2001). (Ohba and Ueda, 2009) suggested that the distinct characteristics of evolution between El Ni?o and La Ni?a decay phases are mainly due to the different distributions of zonal wind stress anomalies over the equatorial WP between ENSO warm and cold phases. During an El Ni?o mature phase, the evident easterly anomalies in the equatorial WP can induce eastward cold Kelvin waves that eliminate the positive SSTAs in the central-eastern equatorial Pacific (CEEP), leading to the phase transition from El Ni?o to La Ni?a. However, during La Ni?a the westerly anomalies are considerably weaker, and thus the resulting downwelling Kelvin waves cannot counteract the negative SSTAs in the CEEP, meaning La Ni?a can persist for longer compared with El Ni?o. The authors argued that the nonlinear response of atmospheric deep convection to SSTAs is the main reason for the distinct anomalous zonal wind over the equatorial WP between El Ni?o and La Ni?a (Hoerling et al., 1997; Kang and Kug, 2002; Okumura et al., 2011; Dommenget et al., 2013). During La Ni?a, the anomalous precipitation center shifts westward by about 10°-15° relative to that of El Ni?o. Therefore, the easterly anomalies associated with negative precipitation anomalies will efficiently reduce the westerly anomalies over the equatorial WP during La Ni?a, resulting in an asymmetric distribution of zonal wind anomalies over the equatorial WP between El Ni?o and La Ni?a mature phases.
Except for the asymmetry of zonal wind anomalies, the southward shift of westerlies during El Ni?o is considered favorable for its termination (Harrison and Vecchi, 1999), which can change the zonal mean equatorial heat content (HC) and establish a meridional asymmetry of thermocline depth in the turnaround phase of ENSO, leading to a durational asymmetry between El Ni?o and La Ni?a (McGregor et al., 2012; Abellán and McGregor, 2016). In addition to atmospheric asymmetric processes, oceanic processes can play a role in prolonging La Ni?a. (Nagura et al., 2008) showed that tropical instability waves slow the transition from La Ni?a to El Ni?o. (Hu et al., 2014) considered that reflected Rossby waves may interrupt the recharge process and prevent the transition from La Ni?a to El Ni?o. (DiNezio and Deser, 2014) proposed that nonlinearity in the delayed thermocline feedback is the sole process prolonging the duration of La Ni?a in a nonlinear delayed-oscillator model.
The zonal wind anomalies over the equatorial WP during ENSO mature phases are tightly linked to an anomalous low-level western North Pacific (WNP) anticyclone (WNPAC) and cyclone (WNPC) (Zhang et al., 1996, Zhang et al., 2017; Wang et al., 1999; Li et al., 2017). The easterly anomalies located in the south wing of the WNPAC during an El Ni?o mature phase can extend southward to the equatorial WP, which is conducive to motivating cold equatorial Kelvin waves; while a westward shifting NWPC during La Ni?a leads to a weaker equatorial thermocline anomaly, which acts as a weaker dynamic forcing to produce a weaker effect on SSTAs, bringing about a longer duration of La Ni?a that persists to the next year (Chen et al., 2016; Tao et al., 2017). It is suggested that Indian Ocean SSTAs may partially contribute to the occurrence of zonal wind anomalies over the equatorial WP during ENSO mature-to-decay phases (Ohba and Ueda, 2009; Okumura and Deser, 2010; Ohba and Watanabe, 2012). The atmospheric Kelvin wave response to warming in the Indian Ocean basin can induce easterly anomalies over the equatorial WP and enhance the low-level anticyclone (Xie et al., 2009; Okumura et al., 2011); nevertheless, it has been noted that the role of Indian Ocean basin warming is more pronounced in the summer of decaying El Ni?o events.
(Zhang et al., 2015) pointed out that the intraseasonal oscillation over the WNP is weak and the interannual variation dominates the wind variability during El Ni?o winters; whereas, during La Ni?a winters the intraseasonal oscillation is dominant and the interannual variation is weak. Such a difference leads to much stronger anomalous anticyclones during El Ni?o than the anomalous cyclones during La Ni?a, causing an asymmetric effect on the precipitation over southern China. The SSTAs over the tropical WP play a crucial role in the different intensities of atmospheric intraseasonal variability between El Ni?o and La Ni?a. The Walker circulation can be affected by the zonal gradient of SSTAs and changes in atmospheric convection are a clue to the Walker circulation slowdown (Tokinaga et al., 2012). Negative SSTAs during El Ni?o can weaken the zonal gradient of SSTAs and lead to a stronger anomalous anti-Walker circulation, resulting in anomalous descending motion and convective cooling over the tropical WP, which is unfavorable for the development of atmospheric intraseasonal oscillation. Meanwhile the reverse is true during La Ni?a (Gao et al., 2018).
The study of (Zhang et al., 2015) mainly focused on the asymmetry of atmospheric circulation during the wintertime of ENSO years, i.e., the mature and decay phases of ENSO. Thus, a question arises: can the distinct intensity of intraseasonal oscillation between El Ni?o and La Ni?a influence the asymmetry of zonal wind anomalies over the equatorial WP during ENSO mature-to-decay phases, and consequently lead to asymmetric decays in El Ni?o and La Ni?a? To address this, the present study begins with an analysis of the anomalous distributions and asymmetric characteristics of SSTAs and atmospheric circulation anomalies during El Ni?o and La Ni?a mature-to-decay phases, based on observation data. Then, we discuss the relationship between the asymmetric characteristics of wind anomalies and intraseasonal variability. Finally, numerical experiments are performed using a global ocean general circulation model (OGCM) to investigate the contribution of zonal wind anomalies over the equatorial WP to the CEEP SSTAs during El Ni?o and La Ni?a decay phases.
2.1. Data and analysis method
The sea level pressure (SLP), 850 hPa wind, and surface wind stress utilized in this study are from the ERA-Interim dataset, with a resolution of 1°× 1° (Dee et al., 2011). The SST data are from HadISST, with a 1°× 1° horizontal resolution (Rayner et al., 2003). The outgoing longwave radiation (OLR) is derived from NOAA/AVHRR data, with a 2.5°× 2.5° horizontal resolution (Liebmann and Smith, 1996). The surface air temperature, SST, and specific humidity used in the model are from COADS (Da Silva et al., 1994). Except for the COADS data that used in the model, the period of other data is from 1979 to 2016, and an anomaly is defined as the departure from the seasonal cycle averaged over this period. In this study, the intraseasonal component is obtained using Lanczos bandpass filtering (10-50 days), and the interannual component is calculated using a three-month running mean based on monthly anomalies.As in (Zhang et al., 2015), the criteria for selecting ENSO events is as follows: if the averaged SSTA over the Ni?o3 region (150°-90°W, 5°S-5°N) in a winter half year (November to April) is greater (less) than 0.5°C, then the winter half year is considered as an El Ni?o (La Ni?a) episode. Eight El Ni?o events (1982/83, 1986/87, 1991/92, 1994/95, 1997/98, 2002/03, 2009/10, 2015/16) and ten La Ni?a events (1984/85, 1985/86, 1988/89, 1995/96, 1998/99, 1999/00, 2005/06, 2007/08, 2010/11, 2011/12), are identified based on these criteria during 1979-2016.
2
2.2. Model and experiments
The model used in this study is the Modular Ocean Model, version 3, developed by the Geophysical Fluid Dynamics Laboratory (Pacanowski and Griffies, 1999). This model adopts a realistic topography, with the model domain ranging from 77°S to 65°N meridionally and reaching down to 5300 m vertically. The horizontal resolution is 1°× 1°, with the meridional resolution varying densely to 1/3° around equator. There are 35 vertical levels, with 20 even levels above 300 m and a 10 m thickness for the uppermost layer. The model adopts a barotropic-baroclinic time splitting algorithm and the explicit free surface scheme is used in this study. Physical parameterizations include the K-Profile Pacanowski-Philandar vertical mixing scheme, the isoneutral mixing scheme, and shortwave solar penetration. To remove the restoration effect of surface air temperature and specific humidity on SST, in this study we use the algorithm of (Rong et al., 2011) to calculate the sensible and latent heat fluxes by bulk formula. The surface air temperature (T a) and specific humidity (q a) are derived by empirical formulas according to the SSTA:\begin{eqnarray} T_{\rm a}&=&T_{\rm ac}+\alpha(x,y)\Delta T_{\rm s} , \ \ (1)\\ q_{\rm a}&=&q_{\rm ac}+\beta(x,y)\Delta T_{\rm s} , \ \ (2)\end{eqnarray} where T s is the observed SST and ? T s=T s-T sc; and T ac, T sc and q ac are the observed climatological surface air temperature, SST and specific humidity, respectively. The coefficients α(x,y) and β(x,y) vary with space and are calculated by regressing the observed monthly SST anomaly onto the monthly specific humidity and surface air temperature anomaly fields at each grid point. Here, the climatological mean wind speed is derived from COADS, while the wind stress and net shortwave and downward longwave flux are derived from ERA-Interim. In this study, the model sea surface salinity (SSS) is simply restored to the Levitus climatological SSS (Levitus, 1982) with a timescale of 30 days.
The model is initiated from a resting ocean. The initial temperature and salinity are derived from the January climatology of Levitus. First, the model is spun up for 50 years, forced by climatological wind speed, wind stress, and net shortwave and downward longwave radiation, with a Newtonian damping term applied to the model SST by forcing it toward the Levitus climatology. The last 20 years' Newtonian damping terms are then averaged and used as the "flux correction" terms in an additional 50-year spin up run. Therefore, the first 100 years' integration is sufficient to make the model upper ocean reach a quasi-equilibrium state. Starting from the above 100-year spin up integration, a further 10-year integration, using the same forcing as the last 50-year spin up integration, is performed and regarded as the control experiment. Then, four sensitivity experiments are conducted starting from the same initial condition as the control experiment. All the forcing fields between the control and sensitivity experiments are the same, except the zonal wind stress. In this respect, an anomalous zonal wind stress is superposed onto the climatological wind stress in the sensitivity experiments. Detailed descriptions of the sensitivity experiments are presented in section 5.
3.1. SSTA evolution
ENSO events are characterized by a significant seasonal phase-locking that peaks in winter and decays after the following spring. A common metric used to represent ENSO is the Ni?o3 index, defined as the averaged SSTA over the Ni?o3 region (5°S-5°N, 150°-90°W). Figure 1 shows the composite Ni?o3 index for El Ni?o and La Ni?a, respectively. Note that the sign of the Ni?o3 index for La Ni?a is reversed to facilitate comparison. The seasonal phase-locking feature of the SSTA during ENSO warm and cold episodes can be clearly observed from Fig. 1. The composite Ni?o3 index generally peaks in winter and declines from the next spring. Both the amplitude and decay speed of El Ni?o are noticeably stronger than those of La Ni?a. The composite Ni?o3 index of El Ni?o crosses the zero line and turns into negative values around the following July after its peak. However, during the La Ni?a decay phase the negative SSTA reinforces again after the following summer and tends to develop as a secondary cold event. By calculating the tendency of Ni?o3 index from January to July of the decay year, it is shown that the averaged decay rate of El Ni?o Ni?o3 index is 0.24°C month-1, while that of La Ni?a is only 0.15°C month-1, indicating an evident asymmetry in decay speed between El Ni?o and La Ni?a events.
Figure1. Composite time series of the Ni?o3 index (units: °C) for El Ni?o (red) and La Ni?a (blue), where the time axis runs from January of the El Ni?o/La Ni?a year (Jan0) to December of the following year (Dec1). The Ni?o3 index for La Ni?a is multiplied by -1.2
3.2. Anomalous atmospheric circulation over the tropical WNP
(Zhang et al., 1996) found that an anomalous anticyclone appears in the lower troposphere over the tropical WNP during El Ni?o mature phases by compositing the 850 hPa wind anomalies of the 1986/87 and 1991/92 El Ni?os, and explained it as the atmospheric Rossby wave response to the anomalous convective cooling over the WNP Maritime Continent. Figures 2a and b show the composite SSTAs and 850 hPa wind anomalies over the WNP during El Ni?o and La Ni?a mature-to-decay phases, respectively. A pronounced anomalous anticyclone over the WNP during the El Ni?o mature phase (D0JF1, where D0 represent the December of the mature phase and JF1 represents the January to February of the following year) can be observed in Fig. 2a, with its center located in the east of the Philippines. Moreover, in the equatorial area of its southern wing, prominent easterly anomalies extend from the east Indian Ocean to 150°E latitudinally, and from 10°S to 5°N meridionally. Corresponding to the anomalous anticyclone, a negative SSTA appears in the east of the Philippines, which is responsible for the atmospheric anomalous convective cooling and arouses the anomalous anticyclone. Both the reduction in evaporation induced by northerly anomalies in the east of the WNPAC (Wang et al., 2000) and the oceanic upwelling Rossby waves induced by wind stress curl anomalies on both sides of equator, which corresponds to the westerly anomalies (Wang et al., 1999), are favorable to the generation and maintenance of the negative SSTA over the WNP. The anomalous cyclone over the WNP during the La Ni?a mature phase is evidently weaker than the anticyclone during El Ni?o, with warm a SSTA occurring in the east of the Philippines at the same time. Furthermore, westerly anomalies over the south of the anomalous cyclone are remarkably weaker than easterly anomalies over the south of the anomalous anticyclone, the extension of which is smaller too (Fig. 2b). Easterly anomalies over the equatorial WP tend to strengthen and extend eastward to 160°E during the El Ni?o decay phase (MAM1) (Fig. 2c), whereas the westerly anomalies during La Ni?a basically remain unchanged (Fig. 2d). Accordingly, both the anomalous anticyclone and its southern easterly anomalies in the mature-to-decay phase (D0JF1 and MAM1) during El Ni?o are noticeably stronger than the anomalous cyclone and its southern westerly anomalies during La Ni?a. Because of the critical influence of zonal wind anomalies over the equatorial WP on the decay of ENSO during D0JF1 and MAM1, next we focus mainly on this period and conduct a composite analysis.
Figure2. Composite SSTAs (shading; units: °C) and 850 hPa wind anomalies (vectors; units: m s-1) during (a, c) El Ni?o and (b, d) La Ni?a (a, b) mature winter (D0JF1) and (c, d) decay spring (MAM1). Dotted areas and plotted vectors are significant above the 99% confidence level.The composite SLP and 850 hPa zonal wind anomalies during El Ni?o and La Ni?a mature-to-decay phases (D0JFMAM1) are shown in Fig. 3. It is demonstrated in the SLP field (Figs. 3a and b) that there are positive anomalies in the WNP during El Ni?o, corresponding to the anomalous anticyclone in the lower troposphere. Its maximum center is located in the eastern ocean of the Philippines, with the maximum exceeding 1.4 hPa. The negative anomalies during La Ni?a are much weaker and located westward compared to the positive anomalies, and its maximum value is only about -1.2 hPa. This indicates that the asymmetry between the intensity of the anomalous anticyclone during El Ni?o and the anomalous cyclone during La Ni?a is clearly reflected in the SLP field.
Figure3. Composite (a, b) SLP anomalies (units: hPa) and (c, d) 850 hPa zonal wind anomalies (units: m s-1) for (a, c) El Ni?o and (b, d) La Ni?a during D0JFMAM1. Dotted areas are significant above the 99% confidence level.(Wu et al., 2010) pointed out that the anomalous anticyclone (anomalous cyclone) over the WNP during ENSO mature phases is closely related to the easterly anomalies (westerly anomalies) over the equatorial WP, and the correlation coefficient between them can reach 0.79. Figure 3c shows the composite 850 hPa zonal wind anomalies during El Ni?o mature-to-decay phases (D0JFMAM1). The positive anomaly area, which means westerly anomalies, is situated near 20°N over the WP, and easterly anomalies are near the equator, corresponding to the anomalous anticyclone. The pattern during La Ni?a is almost the opposite (Fig. 3d), corresponding to the anomalous cyclone. Compared with the SLP anomalies, the asymmetry of zonal wind anomalies over the equatorial WP is more pronounced. Easterly anomalies during El Ni?o are more widely distributed and have a larger central value, the strongest of which can reach -3.1 m s-1; whereas, the extension during La Ni?a is smaller, with an eastward location and a smaller maximum value (<2.1 m s-1). The situation during ENSO mature phases (D0JF1) is more pronounced, with the maximum values exceeding -3.5 m s-1 and 2.3 m s-1 for El Ni?o and La Ni?a, respectively (figure not shown).
In order to demonstrate the above asymmetries quantitatively, we calculate the regionally averaged values of the three fields according to their high maximum centers, which are shown in Fig. 4. The averaged SSTAs over the WNP key region during El Ni?o and La Ni?a (D0JFMAM1) are -0.23°C and 0.19°C, respectively, exhibiting a stronger negative SSTA. Moreover, it is obvious that the anomalous anticyclone is stronger than the anomalous cyclone according to the SLP anomalies (1.11 hPa and -0.94 hPa), which is consistent with the result of (Zhang et al., 2015). Nevertheless, the averaged zonal wind anomalies over the key region are -1.23 m s-1 for El Ni?o and 0.69 m s-1 for La Ni?a, and the amplitude of El Ni?o is around twice as large as that of La Ni?a, indicating a more pronounced asymmetry in the zonal wind field. Both the SLP and zonal wind anomalies are statistically significant above the 99% confidence level, except the averaged westerly anomalies during La Ni?a. The above results clearly illustrate that the WNPAC and the associated easterly anomalies near the equator during El Ni?o are significantly stronger than the WNPC and the westerly anomalies during La Ni?a.
Figure4. Regional-averaged SST (0°-20°N, 130°-150°E), SLP (0°-20°N, 120°-150°E) and 850 hPa zonal winds (5°S-5°N, 100°-140°E) anomalies for El Ni?o (blue) and La Ni?a (red) during D0JFMAM1.Figure 5 further shows the temporal evolution of the averaged anomalous 850 hPa zonal winds over the equatorial WP key region during El Ni?o and La Ni?a mature-to-decay phases. The Ni?o3 index is also shown to represent the temporal evolution of ENSO. To facilitate comparison, the sign of zonal wind anomalies is reversed by multiplying by -1. The easterly anomaly around the equator generally appears in the October of an El Ni?o developing year, and gradually declines after it reaches the maximum value of 1.59 m s-1 in December. The evolution of westerly anomalies during La Ni?a is similar to El Ni?o but the maximum value is only -0.81 m s-1, which is approximately half that of the easterly anomalies' maximum value. The westerly anomalies during La Ni?a gradually disappear and turn into easterly anomalies in the following June-July, which is favorable for the formation of a secondary cold event; whereas, the easterly anomalies during El Ni?o can last until the following September and turn into westerly anomalies in October, which seems to be the westerly anomalies in the south of the WNPC during the La Ni?a following this El Ni?o.
Figure5. Composite time series of regional-averaged (5°S-5°N, 100°-140°E) 850 hPa zonal wind anomalies (blue; units: m s-1) and Ni?o3 index (black; units: °C) for El Ni?o (solid line) and La Ni?a (dotted line). The zonal wind anomalies are multiplied by -1. Red dots represent the values exceeding the 95% confidence level.The strengths of asymmetry are different among the atmospheric and oceanic variables during ENSO mature-to-decay phases. For instance, the ratios of the averaged WNP SSTA and SLP anomalies between El Ni?o and La Ni?a are about 1.2-1.3, while those for equatorial zonal wind anomalies can reach 1.78 (Fig. 4) and exceed 2 during the mature phase of ENSO (Fig. 5), implying that the pronounced asymmetry in zonal wind anomalies cannot be merely ascribed to the amplitude asymmetry of the WNP SSTA; instead, it may result from other processes, e.g., the intraseasonal oscillation, which is discussed in the next section.
Figure6. Standard deviations (units: W m-2) of the (a, b) interannual and (c, d) intraseasonal component of OLR for (a, c) El Ni?o and (b, d) La Ni?a during D0JFMAM1. Dotted areas are significant above the 99% confidence level.As mentioned above, the anomalous zonal winds over the equatorial WP play a crucial role in the decay of ENSO events, which is tightly associated with the WNPAC/WNPC (Wang and Fiedler, 2006). To illustrate the effect of intraseasonal oscillation on the equatorial zonal wind anomalies, we calculate the regionally averaged intraseasonal and interannual components of 850 hPa zonal wind anomalies over the equatorial WP area, and their standard deviations are shown in Fig. 7. It is clear that the interannual standard deviation of 850 hPa anomalous zonal winds is greater than the intraseasonal standard deviation during El Ni?o, while the opposite occurs during La Ni?a. Comparing the cases between El Ni?o and La Ni?a, the standard deviation of anomalous interannual zonal wind during El Ni?o is stronger than during La Ni?a, with the ratio between two phases during D0JF1 and D0JFMAM1 being about 2. The interannual standard deviation during El Ni?o is statistically significant above the 99% confidence level, while that during La Ni?a is not significant. However, the standard deviation of intraseasonal zonal wind anomalies during La Ni?a is larger than during El Ni?o, and the ratio between them during D0JF1 and D0JFMAM1 is 1.38 and 1.34, respectively. Unlike its interannual part, the intraseasonal standard variation during El Ni?o is not statistically significant above the 99% confidence level, while that during La Ni?a is significant. As the interannual and intraseasonal standard deviations represent their amplitudes, it implies that the interannual amplitude of 850 hPa zonal wind anomalies during El Ni?o is about twice that during La Ni?a, which is consistent with the results of Figs. 4 and 5. The amplitudes of the interannual and intraseasonal 850 hPa zonal wind anomalies during El Ni?o's D0JF1 are 1.66 and 1.19, and those during La Ni?a are 0.74 and 1.64, respectively, indicating that the sums of the interannual and intraseasonal amplitudes are roughly equal between El Ni?o and La Ni?a. In other words, the kinetic energy is approximately conserved during El Ni?o and La Ni?a. The same result can been found for D0JFMAM1. In summary, during El Ni?o, the suppression of convection over the tropical WNP weakens the intraseasonal oscillation, and thus the energy of the wind field is mainly concentrated on the interannual time scale, resulting in stronger interannual zonal wind anomalies; whereas, the enhanced convection during La Ni?a favors stronger intraseasonal oscillation, and thus the energy from the atmospheric wind field is mainly concentrated on the intraseasonal time scale, leading to a weaker zonal wind anomaly. This process therefore brings about the pronounced asymmetry of anomalous zonal wind between El Ni?o and La Ni?a over the equatorial WP, and ultimately leads to asymmetric decay speeds of El Ni?o and La Ni?a.
Figure7. Standard deviations of 850 hPa zonal wind interannual (blue) and intraseasonal (red) components for El Ni?o and La Ni?a during (a) mature phases and (b) mature-to-decay phases. The region in which the 850 hPa zonal winds are averaged is (5°S-5°N, 100°-140°E).5.1. Experimental design
In order to quantify the effect of the asymmetric equatorial WP zonal wind anomalies on the decays of El Ni?o and La Ni?a, we conduct four sensitivity experiments, which are shown in Table 1. In the experiments, the composite zonal wind stress anomalies over the equatorial WP (15°S-15°N, 100°-160°E) during El Ni?o and La Ni?a mature phases (D0JF1) as well as mature-to-decay phases (D0JFMAM1) are superimposed onto the climatological zonal wind stress field of the control experiment. Since the difference between the control and sensitivity experiments is only the zonal wind stress forcing, the SST difference between two experiments can measure the effect of wind stress. By comparing the simulations of El Ni?o and La Ni?a anomalous zonal wind stress forcing, we can identify how the asymmetric zonal wind stress anomalies impact ENSO decay.Figure 8 shows the composite zonal wind stress anomalies of four sensitivity experiments. In general, each of these anomalous patterns is consistent with that in Fig. 3c. The easterly anomalies are distributed around the equator and the westerly anomalies around 20°N during El Ni?o, and vice versa during La Ni?a. Notably, the easterly wind stress anomalies tend to strengthen and extend eastward with time during El Ni?o, while the westerly wind stress anomalies during La Ni?a show little change, in agreement with the results in Fig. 2.
Figure8. Zonal wind stress anomalies superposed in sensitivity experiments for (a) EL_D0JF1, (b) LA_D0JF1, (c) EL_D0JFMAM1, and (d) LA_D0JFMAM1.2
5.2. Effect of anomalous zonal wind stress on SST
Figure 9 displays the differences in Ni?o3 index between the four sensitivity experiments and the control experiment, which represents the influence of anomalous zonal wind stress on ENSO decay. For the convenience of comparison, the results of EL_D0JF1 and EL_D0JFMAM1 are multiplied by -1. As shown in Fig. 9, the SSTAs over CEEP become visible from the following March after the ENSO peak phase. The negative SSTAs that arise over CEEP correspond to the easterly wind stress anomalies over the equatorial WP during El Ni?o, while westerly wind stress anomalies result in positive SSTAs during La Ni?a. Because easterly (westerly) anomalies over the equatorial WP during El Ni?o (La Ni?a) can stimulate cold (warm) Kelvin waves, the warm (cold) SSTA over CEEP will be declined by eastward propagating cold (warm) Kelvin waves (Zhang and Huang, 1998; Wang and Fiedler, 2006). The eastward propagation of Kelvin waves can be seen clearly from the equatorial longitude-time sections of upper-ocean HC (Fig. 10). Negative HC anomalies propagate eastward from the equatorial WP after the December of an El Ni?o mature phase and arrive at the eastern boundary of the ocean in the following February when the SSTAs over CEEP (Ni?o3 index) begin to be noticeable. The speed of HC propagation is approximately equivalent to the speed of equatorial Kelvin waves (Fig. 9). Conversely, positive HC anomalies propagate eastward during La Ni?a. The strongest HC anomalies appear around the equatorial EP in the May-June during a decaying El Ni?o, when SSTAs peak too. The same results can been found during La Ni?a (Fig. 9).
Figure9. (a) Time series of Ni?o3 index differences between the control and sensitivity experiments (units: °C). (b, c) Composite Ni?o3 index derived from the Exp_unfiltered (black lines) and Exp_WPfiltered (red lines) experiments for El Ni?o and La Ni?a, respectively. The differences between Exp_unfiltered and Exp_WPfiltered are denoted by blue lines.
Figure10. Simulated longitude-time cross sections of equatorial (averaged over 5°S-5°N) upper ocean HC differences between the control and sensitivity experiments for (a) EL_D0JF1, (b) LA_D0JF1, (c) EL_D0JFMAM1, and (d) LA_D0JFMAM1. The HC is defined as the vertical integration of temperature by depth from the ocean surface to 400 m (units: °C m-1).The SSTAs simulated by the sensitivity experiments show features consistent with observations insofar as evident asymmetry exists between El Ni?o and La Ni?a. The maxima of Ni?o3 index anomalies of EL_D0JF1 and EL_D0JFMAM1 are -0.18°C and -0.26°C, respectively; while those of LA_D0JF1 and LA_D0JFMAM1 are only 0.06°C and 0.08°C (Fig. 9), respectively. The negative SSTAs in the Ni?o3 region induced by the equatorial WP easterly wind stress anomalies during El Ni?o are threefold greater than the positive SSTAs induced by westerly wind stress anomalies during La Ni?a. Significant asymmetric characteristics can also be observed in oceanic HC anomalies, with the HC anomalies of the upper 400 m during El Ni?o being four to five times as large as those during La Ni?a (Fig. 10). The asymmetry in SSTAs and HC anomalies correspond well to the asymmetry of wind stress. Both the intensity and extension of the easterly wind stress anomalies during El Ni?o are notably greater than those of westerly wind stress anomalies during La Ni?a (Fig. 8), and these asymmetries are stronger than the asymmetry of 850 hPa zonal winds. Accordingly, it is favorable for El Ni?o to decay faster, while the cold SSTAs during La Ni?a tend to maintain for a longer period. Noting that there is some evident HC signal near 140°W, such a signal may be associated with the tropical instability waves.
2
5.3. Impact of intraseasonal wind stress anomalies on ENSO
Previous studies suggest that the atmospheric intraseasonal variation rectifies the interannual oceanic variation via nonlinear ocean processes (Kessler and Kleeman, 2000; Rong et al., 2011; Zhao et al., 2019). To investigate the contribution of this oceanic route by which the atmospheric intraseasonal oscillation impacts ENSO decay, we conduct two additional numerical experiments. In the Exp_unfiltered experiment, the original unfiltered daily wind stress anomalies from 1979 to 2016 are used to force the model, whereas in the Exp_WPfiltered experiment a 90-day running mean is applied to daily wind stress anomalies over the tropical WP (20°S-20°N, 100°E-160°W). Figures 9b and c show the composite Ni?o3 indexes and differences between Exp_unfiltered and Exp_WPfiltered during El Ni?a and La Ni?a, respectively. It can be seen that both the Exp_unfiltered and Exp_WPfiltered experiments reproduce the observed Ni?o3 indexes for El Ni?o and La Ni?a well. However, the atmospheric intraseasonal wind stress anomalies exhibit a very limited effect on the decay of both El Ni?o and La Ni?a, with the Ni?o3 index being almost unchanged between Exp_unfiltered and Exp_WPfiltered for both warm and cold events. This result is consistent with the simulations of (Rong et al., 2011) and (Zhao et al., 2019), who used different OGCMs to investigate the rectifications of the intraseasonal wind stress on the interannual oceanic variability. Their studies also showed limited-amplitude and small-scale ocean SSTAs in response to the intraseasonal wind stress anomalies. Thus, the above experiments indicate that the effect of intraseasonal wind stress anomalies on ENSO decay is negligible.The OLR and equatorial zonal wind anomalies show significantly stronger interannual standard deviations than their intraseasonal standard deviations over the tropical WP during El Ni?o mature-to-decay phases; however, during La Ni?a the intraseasonal standard deviations are larger than the interannual standard deviations. It seems that the suppressed convection during El Ni?o is able to weaken the intraseasonal oscillation and, as a result, the atmospheric wind anomalies are more energetic on the interannual timescale; whereas, during La Ni?a the enhanced convection tends to strengthen the intraseasonal oscillation, and the atmosphere obtains most of its kinetic energy through intraseasonal variation, leading to a weakened interannual fluctuation. Therefore, the difference in intraseasonal oscillation intensity may play an important role in strengthening the asymmetry of zonal wind anomalies over the equatorial WP during El Ni?o and La Ni?a decay phases.
Numerical experiments show that the asymmetric zonal wind stress anomalies during El Ni?o and La Ni?a decay phases can induce asymmetry in SSTA decay speeds over CEEP by exciting different equatorial Kelvin waves. The maximum negative anomalies of Ni?o3 index induced by the zonal wind stress anomalies during El Ni?o mature phases (D0JF1) and mature-to-decay (D0JFMAM1) phases are -0.18°C and -0.26°C, respectively, which are threefold greater than the positive Ni?o3 index during La Ni?a, indicating that the stronger zonal wind anomalies over the equatorial WP favor a faster decay of El Ni?o. The numerical experiments also show that the intraseasonal wind stress anomalies have negligible impact on ENSO decay. Furthermore, we demonstrate that the initial state of the evolving ocean is an important component of interannual variability. However, using the initial state with regard to El Ni?o or La Ni?a may contain additional signals of other processes or forcings——for instance, the reflection of the off-equatorial Rossby waves in the equatorial WP. Moreover, as we use the composite wind stress anomalies to force the sensitivity experiments, it is difficult to select an appropriate initial state with regard to El Ni?o and La Ni?a, since we cannot use a composite initial state, which is generally dynamically unbalanced.
Nevertheless, it should be pointed out that the Ni?o3 index induced by the zonal wind stress anomalies in the four sensitivity experiments are all less than 0.3°C (Fig. 9); however, as shown in Fig. 1, the observational Ni?o3 index during both El Ni?o and La Ni?a exceeds 1°C. That is, despite the zonal wind anomalies over the equatorial WP being favorable to asymmetric ENSO decay, there might be additional processes that contribute to ENSO decay too. The relative contributions of other processes and zonal wind anomalies over the equatorial WP to ENSO asymmetric decay require further exploration.
