1. Introduction
Owing to the profound weather and climate impact of El Niño–Southern Oscillation (ENSO), significant effort has been devoted to predicting ENSO events several seasons in advance. Despite the tremendous progress in ENSO research, the prediction skill of ENSO is not constantly improving but exhibits a significant decadal variation (Hu et al. 2020). The predictability of ENSO was relatively higher in the 1980s to 1990s than in the 1960s to 1970s (Torrence and Webster 1998). Higher uncertainties for long lead-time ENSO predictions may indicate the skill decline of ENSO forecasting in recent decades, such as for the 2010/11 La Niña (Zhang et al. 2013), the 2012/13 El Niño, and the 2014/15 borderline El Niño (McPhaden 2015; Zhu et al. 2016; Levine and McPhaden 2016). The fluctuations of ENSO predictability have been attributed to the decadal variations of the ENSO spring persistence barrier (PB) (e.g., Barnston et al. 2012).
The predictability of ENSO tends to drop dramatically during boreal spring (Barnston et al. 1994; Webster and Yang 1992; Torrence and Webster 1998). This low predictability is closely related to the spring PB, which usually refers to the rapid decline in persistence (lagged autocorrelations) of ENSO-related sea surface temperature (SST) anomalies across boreal springs (Liu et al. 2019; McPhaden 2003; Ren et al. 2016; Tippett and L’Heureux 2020; Torrence and Webster 1998; Webster and Yang 1992). As one of the ENSO properties, the PB intensity also exhibits a decadal variation (Fang et al. 2019; Y. Jin et al. 2020; Torrence and Webster 1998; Yu and Kao 2007; Zhu et al. 2015).
Although various hypotheses have been suggested, the cause of ENSO PB and its decadal variability have not yet been fully understood. Besides the potential impacts of the ENSO biennial component (Clarke and Van Gorder 1999) and thermocline (Yu and Kao 2007; Zhu et al. 2015), it is commonly suggested that the ENSO PB is closely related to the seasonal change of tropical atmosphere–ocean interactions, especially in spring (e.g., Fang and Mu 2018; Nicholls 1979; Torrence and Webster 1998; Webster and Yang 1992; Wright 1979). For example, Webster and Yang (1992) suggested that the weaker Walker circulation in spring may produce the barrier by allowing random perturbations to disrupt the persistence of ENSO anomalies with a lowest signal-to-noise ratio (Torrence and Webster 1998), which is closely related to the Asian monsoon activity (Lau and Yang 1996). Fang et al. (2019) proposed that the spring PB is mainly caused by the ENSO stability in spring, and thermal damping processes in boreal spring regulate the decadal variations in ENSO spring PB.
However, Ren et al. (2016) have pointed out that the timing of PB is not always fixed in spring, and the maximum decline in persistence (lagged autocorrelations) of ENSO-related SST anomalies can occur in summer. They proposed to utilize the maximum rate of persistence decline as the measurement of ENSO PB. It means that the effect of spring alone cannot account for all the variations in ENSO PB. There are several studies emphasizing the important role of the annual cycle of ENSO growth rate in PB (Levine and McPhaden 2015; Liu et al. 2019; Stein et al. 2014, 2010). For example, following the PB definition of Ren et al. (2016), the theoretical solution from Liu et al. (2019) has indicated that the ENSO PB is caused primarily by the declining growth rate instead of the minimum growth rate during spring. The ENSO growth rate can be determined by the tropical Pacific background mean state (An and Bong 2016; Fedorov and Philander 2000). Accordingly, the variations in ENSO PB are usually attributed to the changes in the background mean state (Jin et al. 2019; Fang et al. 2019). However, considering the disturbance from the stochastic atmospheric forcing, it is not clear to what extent the observed long-term changes are systematic due to the changing in the background state (Capotondi and Sardeshmukh 2017).
Changes in the background mean state (e.g., mean SST, thermocline depth, and surface winds) can influence the efficiency and timing of the ENSO dynamics, thereby altering the PB, frequency, amplitude, and other statistics of ENSO events simultaneously (Wang and An 2002; Capotondi and Sardeshmukh 2017; An and Bong 2016; Fedorov and Philander 2000). Apparent changes in ENSO properties, including amplitude, frequency, and PB intensity, were observed in the late 1970s. Recently, using a multicomponent linear inverse modeling technique, Capotondi and Sardeshmukh (2017) proved that changes in crucial El Niño properties observed after the late 1970s did not occur “by chance” and statistically significant systemic changes have indeed occurred in ENSO dynamics since the late 1970s, which is related to the mean state. Based on the results of Capotondi and Sardeshmukh (2017), it is reasonable and meaningful to investigate the shift in the 1970s in ENSO PB associated with the background mean state. Understanding the dynamic processes accounting for the 1970s shift in ENSO PB carries important implications for the relationships among the different ENSO properties and the regime shift in ENSO due to global warming. Therefore, our present study will focus on revealing the main factors that control the interdecadal variability of the ENSO PB intensity before and after the late 1970s.
2. Data and methods
a. Data description
The SST dataset used in the study is from the Hadley Centre Global Sea Ice and SST (HadISST) with a 1° × 1° grid since 1870 (Rayner et al. 2003). The ocean temperature and currents, surface wind stress, 20°C isotherm depth (D20), and heat flux have been derived from the European Centre for Medium-Range Weather Forecasts Ocean Reanalysis System 5 (ORAS5; Zuo et al. 2019) since 1958. The vertical current velocity in the upper ocean is not provided by ORAS5 and is thus calculated through the mass continuity equation (Ren and Wang 2020). In addition, ORAS4 (Balmaseda et al. 2013) is also used to confirm the results given in the online supplemental material. Both D20 and vertical current velocity for ORAS4 are not provided directly. To focus on the ENSO-related variations, high-frequency variability was removed by an 11-month running mean, similar to Lu et al. (2018), using a 9-month filter. Anomalies and climatic fields are calculated using the selected periods.
b. Conceptual model of ENSO
c. Definition of ENSO PB intensity
The PB is characterized by a band of maximum decline in monthly autocorrelation function (ACF), which is described as a function of initial months t and lag months τ (Ren et al. 2016). According to Liu et al. (2019), the strength of ENSO PB can be derived from the ACF in the following steps:
-
For a calendar month t, SB(t) is the maximum decline of the monthly autocorrelation, which is estimated as the lag decline in the time step of 1 month:where τB(t) is the calendar month with maximum ACF decline.
-
Then, the PB intensity is estimated as
d. Bjerknes–Jin index analysis
The BJ index analysis was first derived by Jin et al. (2006) based on the recharge oscillator framework, which has been widely used as process-based ENSO diagnostics in many perspectives. The BJ index analysis is constantly improving (F.-F. Jin et al. 2020). For example, recent updates with many minor modifications take account of nonlinearity, subgrid-scale, and other processes, which contribute to overcoming some important issues (F.-F. Jin et al. 2020; Chen et al. 2021; Iwakiri and Watanabe 2021). The version of BJ index analysis in the present study referred to Liu et al. (2014) has already been applied to the Langevin equation for ENSO PB (Jin et al. 2019). It should be noted that compared to the versions in Jin et al. (2006), the dynamical ocean adjustment rate, which is largely negative, is not considered in Eq. (10). According to the results of F.-F. Jin et al. (2020), the dynamical ocean adjustment rates of different months are equal, which would strongly affect the annual mean of the ENSO growth rate (weakly stable) (Kim and Jin 2011; F.-F. Jin et al. 2020; Chen et al. 2021), rather than the seasonality. Therefore, this simplified version of BJ index analysis can still be valid for diagnosing the seasonal amplitude of ENSO growth rate.
3. Decadal variability of ENSO PB
The strength of the PB exhibits a significant decadal variability, which was stronger during the 1960s to 1970s and weaker in the 1980s to 1990s (Fang et al. 2019; Torrence and Webster 1998; Yu and Kao 2007; Zhu et al. 2015). A significant reduction in PB intensity occurred in 1977/78 (Fig. S1 in the online supplemental material), in which systematic changes in ENSO dynamics from 1958–77 (P1) to 1978–97 (P2) were also detected (Capotondi and Sardeshmukh 2017). Consistently, these two periods (P1 and P2) are selected for investigation in our study. The persistence of ENSO is usually described by the lagged autocorrelation of the Niño index. The autocorrelations for the two periods (Fig. 1) generally decline with increasing lag months. Regardless of the initial month, the rapid decline of persistence is usually phase-locked to the spring-summer season (the calendar months of April are marked by the white dots in Fig. 1), showing a typical PB feature. Besides the significant reduction in the PB intensity from P1 to P2 (Figs. S1 and S2), the timing of PB during P2 has been delayed several months except for the initial months of September and October, compared to that in P1 (black dots in Fig. 1a vs Fig. 1b). The differences in the PB intensity between P1 and P2 are significant exceeding the 95% confidence level by Student’s t test (Figs. S1 and S2).
The Niño index persistence map (autocorrelation function) for periods of (a) 1958–77 (P1) and (b) 1978–97 (P2). The black dots mark the month of maximum autocorrelation decline for different initial months [τB(t)]. The white dots indicate calendar months of April. The values of the PB intensity are noted in each panel (P1: 4.72 and P2: 2.29).
Citation: Journal of Climate 36, 5; 10.1175/JCLI-D-22-0507.1
4. The role of ENSO growth rate
Many previous studies have suggested that PB is closely related to the annual cycle of the ENSO growth rate (Levine and McPhaden 2015; Liu et al. 2019; Stein et al. 2014, 2010). According to the analytical solution [Eq. (5)] of Liu et al. (2019) and Eq. (6), the theoretical PB intensity can be determined by the annual mean (−b0) and the standard deviation (σ) of the ENSO growth rate (Fig. 2a). It shows that a smaller (larger) annual mean growth rate (−b0) intensifies (weakens) the PB (Y. Jin et al. 2020), while a smaller (larger) standard deviation (σ) of growth rate weakens (intensifies) the PB (Fig. 2a). Compared to standard deviation (σ), the changes of PB are significantly less sensitive to the annual mean (−b0) of growth rate (Fig. 2a). Although the analytical solution in Eq. (5) is derived from the Langevin equation, Levine et al. (2017) have made a same conclusion using a recharge oscillator conceptual model of ENSO. With a series of experiments, they concluded that changes in the annual mean of the growth rate do not significantly affect the PB strength. In contrast, the amplitude of the annual cycle of the growth rate can strongly affect both ENSO amplitude and PB strength. The results suggest that the PB intensity is largely determined by the standard deviation of ENSO growth rate, exhibiting almost a linear relationship between them (Fig. 2a).
(a) The values of the theoretical solutions (
Citation: Journal of Climate 36, 5; 10.1175/JCLI-D-22-0507.1
The analytical solution for PB intensity in Eq. (5) is derived from the Langevin equation [Eq. (1)] under various assumptions (Liu et al. 2019). For example, Liu et al. (2019) assumed a damped, noise-driven ENSO (b0 > 0) with a weak seasonal cycle (AB = 1). However, these assumptions are not always satisfied by the observation. There are two popular ways to estimate the ENSO growth rate. Besides the BJ index analysis (Jin et al. 2006), the ENSO growth rate also can be simply estimated from the time series of the Niño index by linear regression (Jin et al. 2019). Since these two methods are dependent on linear regression, there may be great uncertainties about the specific values of growth rate (Graham et al. 2014). Therefore, previous studies usually quantified the relative magnitude of the growth rate among different periods or seasons (An and Bong 2016) rather than emphasizing its specific value. Particularly, the annual mean growth rate estimated by previous studies could be negative (b0 > 0) or positive (b0 < 0) (Ren and Wang 2020). Kim and Jin (2011) have pointed out that the BJ index analysis may tend to overestimate the growth rate estimated by the Niño index, and the growth rates estimated by the BJ index analysis in their study are positive. However, the positive linear relationship between the PB intensity and the standard deviation (σ) of the growth rate in Fig. 2a is based on the analytical solution derived from a damped, noise-driven ENSO conceptual model with a negative annual growth rate (b0 > 0) (Liu et al. 2019). Therefore, a more general relationship between the PB intensity and σ needs to be further confirmed.
The standard deviation (σ) of growth rate and the PB intensity can be directly derived by the time series of the Niño index. Consistent with the analytical solution (Fig. 2a), it shows that the observed standard deviation of ENSO growth rate is also linearly proportional to the PB intensity (correlation: 0.91, exceeding the 95% confidence level) (Fig. 2b). The above results imply that the variability of PB intensity can be roughly linearly explained by the seasonal amplitude (σ) of ENSO growth rate. Therefore, to investigate the late 1970s shift in ENSO PB, the growth rates for 1958–77 (P1) and 1978–97 (P2) are estimated.
5. Contribution of ocean dynamics to the late 1970s shift in PB
a. Estimation of the ENSO growth rate
As mentioned above, the growth rate can be simply estimated by the time series of the Niño index. Results show that ENSO has the largest seasonal growth rate in boreal autumn and the smallest in boreal spring (Fig. 3a), which is consistent with the seasonality of the tropical Pacific air–sea coupled instability (Fang et al. 2019; Y. Jin et al. 2020; Moore and Kleeman 1996; Karspeck et al. 2006; Larson and Kirtman 2017). Despite these similarities, the annual cycle of the ENSO growth rate shows some differences between P1 and P2. For instance, compared to P1, the standard deviation of P2 is smaller, and the phase/peak of P2 is delayed (Fig. 3a). The decline of PB intensity from P1 (4.72) to P2 (2.29) is about 51% [(P1 − P2)/P1] (Fig. 1), and is consistent with the reduction (52%) in seasonal standard deviation for growth rate (P1: 0.085, P2: 0.041) (Fig. 2a). In addition, the delayed timing of the PB in P2 (Fig. 1a vs Fig. 1b) also can be attributed to the phase delay of the ENSO growth rate (Fig. 3a), which is consistent with the theory for the PB timing of Liu et al. (2019). However, the specific processes affecting the growth rate cannot be resolved from the growth rate estimated by the statistical linear fitting (Fig. 3a).
(a) The annual cycle of ENSO growth rates (month−1) estimated by the statistical linear fitting for P1 (blue) and P2 (red). (b) The seasonal cycle of the BJ index analysis and its five components: (c) TD, (d) MA, (e) ZA, (f) EK, and (g) TH. The total standard deviation decline (%) for (a) the growth rate and (b) the BJ index analysis from P1 to P2 [(P2 − P1)/P1] is noted at the top-right corner, and the declines caused by different components (“+” or “−” indicates a positive or negative contribution to the decline) are marked in (c)–(g). The blue and red dots indicate the maximum and minimum, and the blue and red dashed lines represent their averages, respectively. Following the method of Ren and Wang (2020), the error bar represents the maximum and minimum of each term obtained when the eastern and western boundaries of the averaging area are varied by 10° to either the east or the west with respect to the date line and 90°W, respectively. The gray bars indicate the difference between the two periods (P2 − P1), and the differences exceeding the 95% confidence level are marked by stars.
Citation: Journal of Climate 36, 5; 10.1175/JCLI-D-22-0507.1
b. Contribution estimated by BJ index analysis
Since the PB decadal variability is largely determined by the ENSO growth rate for the two periods (P1 and P2), one could estimate the contribution of different factors to the PB changes by the BJ index analysis. According to the results of Capotondi and Sardeshmukh (2017), the periods of P1 and P2 are speculated to be suitable for the application of the BJ index analysis. The growth rates estimated from the Niño index by statistical linear fitting (Kim and Jin 2011) and another dataset (Fig. S3) are also used as a reference to support the validity of the BJ index analysis in the present study. The BJ index analysis consists of five feedback components [Eq. (11)]. We separately computed each component for P1 and P2 (Figs. 3c–g). The BJ indices during these two periods (Fig. 3b) are approximately consistent with the growth rate estimated by the Niño index (Fig. 3a), including the variations in seasonal amplitudes and phases.
Among the components of the BJ index analysis, MA and TD are the damping factors. MA has a similar seasonal evolution to the BJ index analysis, but a great reduction of MA damping effect in P2 is found in February–March. The seasonal amplitudes of TD are much smaller in contrast to those of MA, but a greater thermal damping effect in P2 is found in boreal spring associated with the eastern tropical Pacific SST warming (Fang et al. 2019). This greater spring thermal damping in P2 has delayed the phase of the seasonal cycle in ENSO growth rate (Fig. 3b) and then postponed the timing of ENSO PB (black dots in Fig. 1a vs Fig. 1b) following the PB timing theory of Liu et al. (2019).
The other three terms, ZA, EK, and TH, are growing factors. The values of ZA and TH are much larger than EK’s with apparent seasonal cycles (Figs. 3e–g). And the phase of the TH term is opposite to that of ZA or BJ, which shows a higher effect in January–April and a lower effect in July–September. Accordingly, in turn, a larger seasonal amplitude of TH for P2 (Fig. 3g) can reduce the amplitude of the BJ index.
Generally, a 38.6% reduction of seasonal standard deviation in BJ index analysis is observed from P1 to P2. To assess each contribution quantitatively, a simple sensitivity test is designed. For example, to test the contribution by MA, the following steps will be taken: 1) The MA of P2 is used to replace the original MA component in BJ index analysis of P1 to get a new growth rate; 2) Compared to the original growth rate by the BJ index analysis in P1 (σBJ,P1), the standard deviation decline for this new growth rate (
c. The roles of MA and TH
As mentioned in the introduction, many previous studies have investigated the effects of changes in the background mean state in the late 1970s on altering the ENSO events (Wang and An 2002; Capotondi and Sardeshmukh 2017; An and Bong 2016; Fedorov and Philander 2000). Concurrent with the decadal modulation of ENSO, a complete evolution of the Pacific decadal oscillation (PDO), which is specified as a long-lived ENSO-like pattern in tropics to northern midlatitude Pacific climate variability, has occurred from P1 to P2 (An and Bong 2016). The phase of PDO turned from negative to positive from P1 to P2 when the mean state shifted toward a warmer eastern tropical Pacific and a colder extratropical central North Pacific, along with the reduction in trade winds and the shoaling of the mean thermocline depth along the equator (e.g., Nitta and Yamada 1989; Wang 1995; McPhaden and Zhang 2009; An and Bong 2016).
According to the calculation of MA in Eq. (10), the damping by the mean advection is determined by the divergence of mean ocean current over the boundary regions (dashed line in Fig. 4) in the eastern equatorial Pacific, which induces strong outward flow from the equator in the oceanic mixed layer. The MA damping rate in P2 has decreased nearly for all seasons, which is mainly caused by the weakened mean meridional current (Fig. 4). Due to the weakened trade wind, the weakened mean meridional current reduces its effect in draining the equatorial warm/cold surface water toward the off-equatorial region, meaning weaker damping. However, the seasonal reduction of MA is uneven (gray bars in Fig. 3d), with a greater (smaller) reduction in February–March (August–September). The decrease in MA damping rate in P2 in February–March and August–September is attributed to both zonal and meridional currents (Fig. S5). Compared to August–September, the reduction in MA damping in February–March is mainly caused by the weakened southward ocean current over the southern boundary (yellow shading in Fig. 4a); in August–September, the zonal component plays a more important role (Fig. S5). According to Fig. 4b, a decrease in westward currents at the western boundary is the most significant change in August–September. The westward currents at the western boundary are smaller than those at the eastern boundary in P2, which induces a weak zonal convergence for the eastern tropical Pacific and reduces the flow outward the equator. The reduction of MA damping rate by the zonal currents at the boundaries in August–September is smaller than the reduction in February–March (Fig. S5). This uneven reduction causes the decline in the standard deviation of MA and thus the ENSO growth rate.
The climatic horizontal ocean current for the mixed layer (0–50 m) for (a) February–March and (b) August–September. The ocean currents for P1 (P2) are marked by blue (red) vectors. The meridional current differences between P1 and P2 (P2 − P1) are shaded in color (the differences exceeding the 95% confidence level by the Student’s t test are marked by gray dots). The solid (5°S–5°N, 180°–90°W) and dashed (6°S–6°N, 177.5°E–90°W and 4°S–4°N, 182°–85°W) line boxes in (a) and (b) represent the eastern Pacific and its boundary, respectively. The values represent the meridional currents averaged over the south and north boundaries (the region between the two dashed boxes), respectively, and the blue (red) numbers are for P1 (P2).
Citation: Journal of Climate 36, 5; 10.1175/JCLI-D-22-0507.1
Thermocline feedback (TH) in February–March (August–September) in P2 is stronger (weaker) than that in P1 (Fig. 3g), enhancing the seasonal amplitude of TH, and it also reduces the standard deviation of the growth rate by the BJ index analysis in P2 (13.8%). The TH consists of four components (Fig. 5): anomalous zonal thermocline slope response to wind anomalies (βh), anomalous subsurface temperature response to thermocline change (ah), mean upwelling velocity (
As in Fig. 3, but for the seasonal cycle of (f) the thermocline feedback (TH) and its related coefficients: (a) μa, (b) βh, (c)
Citation: Journal of Climate 36, 5; 10.1175/JCLI-D-22-0507.1
Similar to the result of Zhu et al. (2015), the sensitivity of subsurface temperature (ah) to the thermocline change is weak (strong) during the boreal spring to early summer (fall to early winter) (Fig. 5e). The physical connections between thermocline and subsurface temperature are quite different among these seasons. During spring to early summer, the east–west SST zonal contrast is weak, with a flatter thermocline, and the subsurface temperature anomalies response to thermocline is mainly caused by the local vertical displacement of the thermocline. In contrast, the easterly is strongest during fall to early winter, and the thermocline is more tilted. And the subsurface temperature anomalies response to thermocline is mainly caused by the variation of thermocline inclination, along with the relaxation of the trade wind, which can induce a stronger response of subsurface temperature anomalies to the thermocline depth. Consistently, the greatest changes for ah during March–April and October–November (Fig. 5e) show quite different features (Fig. S6). Accompanied by the effect of mean upwelling velocity (
Difference (P2 − P1) in climatic subsurface temperature along the equator (5°S–5°N) between P1 and P2 for (a) February–March and (b) August–September (the differences exceeding the 95% confidence level by Student’s t test are marked by dots), and the blue and red solid lines represent the depth of the thermocline (20°C isotherm) for P1 and P2, respectively.
Citation: Journal of Climate 36, 5; 10.1175/JCLI-D-22-0507.1
As mentioned above, there is a vertical lift of the thermocline along the equator in February–March without a notable variation of thermocline inclination (Fig. 6a). The lift of the thermocline in the eastern tropical Pacific with surface warming and significant subsurface cooling increases vertical gradient of subsurface temperature. The vertical subsurface temperature gradient (Fig. S7) above 150 m in the eastern tropical Pacific has increased about 11.3% [from 0.065°C m−1 (P1) to 0.072°C m−1 (P2)], which causes a stronger response of subsurface temperature to the depth change of thermocline and hence a stronger spring thermocline feedback in P2. In August–September, the thermocline lift and the subsurface cooling are concentrated in the western Pacific along with the eastern Pacific warming (Fig. 6b). The thermocline becomes flat with the relaxation of the trade wind, which would reduce the response of subsurface temperature to the thermocline, causing a weaker thermocline feedback in P2. These different changes in subsurface temperature structure for different seasons increase the seasonal amplitude of the thermocline feedback. However, compared to February–March, the changes for TH in August–September is not apparent, with little variations for variation of thermocline inclination.
It should be noted that actual changes in the mean thermocline depth and stratification are difficult to diagnose confidently, given large variations due to ENSO and the sparsity of subsurface data (Fedorov et al. 2020). It is reported that there is a large spread of BJ index analysis, depending on reanalysis datasets (Lübbecke and McPhaden 2014). Moreover, consistent with the previous results (An and Bong 2016), the two extreme El Niño events (1982/83 and 1997/98), causing strong nonlinearity and rectification to the climatological mean state (Timmermann et al. 2003), largely contribute to the results of the BJ index analysis in P2, especially for the TH term (Fig. S8). Without the two extreme El Niño events, the seasonal amplitude of the ENSO growth rate becomes smaller (Fig. S8).
6. Conclusions and discussion
The ENSO PB intensity exhibits a significant decadal variability due to the changes in the seasonal cycle of the ENSO growth rate. The present study reveals that the PB intensity is linearly proportional to the seasonal amplitude of the ENSO growth rate. Particularly, a decadal reduction in PB intensity occurred in the late 1970s due to the weakened seasonal amplitude of ENSO growth rate, mainly attributed to the mean advection (MA) and the thermocline feedback (TH). It is also found that in the 1980s and 1990s, the strengthened thermal damping associated with the eastern equatorial Pacific SST warming in spring delayed the timing of PB.
Some ENSO properties, such as amplitude, frequency, and propagation, have been shown to change coherently on a decadal time scale, which jointly results in the decadal shift of ENSO regime (e.g., Capotondi and Sardeshmukh 2017; Hu et al. 2020). In addition, the change in the mean state of the tropical Pacific due to climate change or anthropogenic warming has been suggested to be responsible for the ENSO regime shift (Fedorov and Philander 2000; Jiang and Zhu 2018, 2020). Here, we find that the changes in PB, including its intensity and timing, are linked with the decadal changes in the background state, such as the ocean current, thermocline, and SST. Our results demonstrate a quantitative evaluation of the PB changes in dynamic processes associated with the background state. This contributes to a systematic understanding of the regime shift in ENSO properties. Moreover, our study is focused on the ENSO growth rate and its relation to the PB. Actually, from the perspective of error growth dynamics, some studies have also shown clearly that the annual cycle in the ENSO coupled instability can regulate the SST perturbation growth and result in the prediction uncertainties in the models (Moore and Kleeman 1996; Karspeck et al. 2006; Larson and Kirtman 2017), which also supports the linkage between growth rate and PB. Combined with these related studies, our result also carries implications for understanding the relationship between climate change and ENSO predictability.
Finally, some remaining issues are discussed. In this study, we mainly discussed the differences in the physical processes associated with the background state changes. One may wonder what caused these background changes. Previous studies suggested that the late 1970s shift of the background climate state in the tropical Pacific can be attributed to the changes in the extratropic by changing the surface wind in the tropic (Wang and An 2002; Dima et al. 2015). However, neither the effects of the climate internal variability nor the external forcing can be ruled out (Capotondi and Sardeshmukh 2017).
The physical explanation for the changes in ENSO growth rate here depends on the BJ index analysis. The results of the BJ index analysis in our study are consistent with others’ results (e.g., An and Bong 2016; Fang et al. 2019; Jin et al. 2019). The relative differences (P1 vs P2) of ENSO growth rates measured by the BJ index analysis nearly agree with those of the numerical linear growth rates estimated from the statistical linear fitting, but the BJ index analysis may tend to overestimate the value of the ENSO growth rate (e.g., Kim and Jin 2011). Graham et al. (2014) indicated that BJ index analysis, as the dynamical growth rate, may not always accurately portray the ocean dynamics. It is suggested that the BJ index analysis should be conducted carefully. Jin et al. (2006) have pointed out that the general formula of the BJ index analysis is useful for assessing the sensitivity of the coupled stability of ENSO to changes in tropical climate conditions. Further, F.-F. Jin et al. (2020) have checked the horizontal structure of climatological state and anomaly (see their Fig. 6.4) and determined the regions for area average. The BJ index analysis should be applied to periods with relatively stable tropical climate states and used for examining the ENSO changes due to the climate background state. The deficiency of the BJ index analysis may largely come from not being used correctly. The ENSO PB intensity seems to become stronger again in the most recent decades, along with the ENSO regime shift in 1999/2000 (Hu et al. 2020), but the PB intensity since then has become unstable (Fig. S1). The ENSO regime shift in 1999/2000 is featured with a westward shift in the atmosphere–ocean coupling in the tropical Pacific, causing the changes in ENSO diversity (Hu et al. 2020). Accordingly, the ENSO dynamics in recent decades are difficult to be diagnosed in the same way for comparison (Ren et al. 2016; Ren and Wang 2020). The mechanisms underlying the recent enhanced PB intensity need further investigation.
Acknowledgments.
This study was jointly supported by the National Natural Science Foundation of China (41830969, 42005011), and the Basic Scientific Research and Operation Foundation of CAMS (2021Z004).
Data availability statement.
ECMWF Ocean Reanalysis System 5 (ORAS5) and ORAS4 can be accessed at https://www.cen.uni-hamburg.de/icdc/data/ocean/easy-init-ocean/ecmwf-oras5.html and https://www.cen.uni-hamburg.de/icdc/data/ocean/easy-init-ocean/ecmwf-ocean-reanalysis-system-4-oras4.html, respectively. The SST dataset from HadISST can be downloaded respectively from https://www.metoffice.gov.uk/hadobs/hadisst/data/download.html.
REFERENCES
An, S.-I., and H. Bong, 2016: Inter-decadal change in El Niño–Southern Oscillation examined with Bjerknes stability index analysis. Climate Dyn., 47, 967–979, https://doi.org/10.1007/s00382-015-2883-8.
An, S.-I., and H. Bong, 2018: Feedback process responsible for the suppression of ENSO activity during the mid-Holocene. Theor. Appl. Climatol., 132, 779–790, https://doi.org/10.1007/s00704-017-2117-6.
Balmaseda, M. A., K. Mogensen, and A. T. Weaver, 2013: Evaluation of the ECMWF ocean reanalysis system ORAS4. Quart. J. Roy. Meteor. Soc., 139, 1132–1161, https://doi.org/10.1002/qj.2063.
Barnston, A. G., and Coauthors, 1994: Long-lead seasonal forecasts—Where do we stand? Bull. Amer. Meteor. Soc., 75, 2097–2114, https://doi.org/10.1175/1520-0477(1994)075<2097:LLSFDW>2.0.CO;2.
Barnston, A. G., M. K. Tippett, M. L. L’Heureux, S. Li, and D. G. DeWitt, 2012: Skill of real-time seasonal ENSO model predictions during 2002–11: Is our capability increasing? Bull. Amer. Meteor. Soc., 93, ES48–ES50, https://doi.org/10.1175/BAMS-D-11-00111.2.
Capotondi, A., and P. D. Sardeshmukh, 2017: Is El Niño really changing? Geophys. Res. Lett., 44, 8548–8556, https://doi.org/10.1002/2017GL074515.
Chen, H.-C., F.-F. Jin, S. Zhao, A. T. Wittenberg, and S. Xie, 2021: ENSO dynamics in the E3SM-1-0, CESM2, and GFDL-CM4 climate models. J. Climate, 34, 9365–9384, https://doi.org/10.1175/JCLI-D-21-0355.1.
Clarke, A. J., and S. Van Gorder, 1999: The connection between the boreal spring southern oscillation persistence barrier and biennial variability. J. Climate, 12, 610–620, https://doi.org/10.1175/1520-0442(1999)012<0610:TCBTBS>2.0.CO;2.
Dima, M., G. Lohmann, and N. Rimbu, 2015: Possible North Atlantic origin for changes in ENSO properties during the 1970s. Climate Dyn., 44, 925–935, https://doi.org/10.1007/s00382-014-2173-x.
Fang, X.-H., and M. Mu, 2018: Both air–sea components are crucial for El Niño forecast from boreal spring. Sci. Rep., 8, 10501, https://doi.org/10.1038/s41598-018-28964-z.
Fang, X.-H., F. Zheng, Z. Y. Liu, and J. Zhu, 2019: Decadal modulation of ENSO spring persistence barrier by thermal damping processes in the observation. Geophys. Res. Lett., 46, 6892–6899, https://doi.org/10.1029/2019GL082921.
Fedorov, A. V., and S. G. Philander, 2000: Is El Niño changing? Science, 288, 1997–2002, https://doi.org/10.1126/science.288.5473.1997.
Fedorov, A. V., S. Hu, A. T. Wittenberg, A. F. Z. Levine, and C. Deser, 2020: ENSO low-frequency modulation and mean state interactions. El Niño Southern Oscillation in a Changing Climate, Amer. Geophys. Union, 173–198.
Graham, F. S., J. N. Brown, C. Langlais, S. J. Marsland, A. T. Wittenberg, and N. J. Holbrook, 2014: Effectiveness of the Bjerknes stability index in representing ocean dynamics. Climate Dyn., 43, 2399–2414, https://doi.org/10.1007/s00382-014-2062-3.
Hu, Z.-Z., A. Kumar, B. Huang, J. Zhu, M. L’Heureux, M. J. McPhaden, and J.-Y. Yu, 2020: The interdecadal shift of ENSO properties in 1999/2000: A review. J. Climate, 33, 4441–4462, https://doi.org/10.1175/JCLI-D-19-0316.1.
Iwakiri, T., and M. Watanabe, 2021: Contribution of Ekman transport to the ENSO periodicity estimated with an extended Wyrtki index. Geophys. Res. Lett., 48, e2021GL095193, https://doi.org/10.1029/2021GL095193.
Jiang, N., and C. Zhu, 2018: Asymmetric changes of ENSO diversity modulated by the cold tongue mode under recent global warming. Geophys. Res. Lett., 45, 12 506–12 513, https://doi.org/10.1029/2018GL079494.
Jiang, N., and C. Zhu, 2020: Tropical Pacific cold tongue mode triggered by enhanced warm pool convection due to global warming. Environ. Res. Lett., 15, 054015, https://doi.org/10.1088/1748-9326/ab7d5e.
Jin, F.-F., S. T. Kim, and L. Bejarano, 2006: A coupled-stability index for ENSO. Geophys. Res. Lett., 33, L23708, https://doi.org/10.1029/2006GL027221.
Jin, F.-F., H.-C. Chen, S. Zhao, M. Hayashi, C. Karamperidou, M. F. Stuecker, R. Xie, and L. Geng, 2020: Simple ENSO models. El Niño Southern Oscillation in a Changing Climate, Amer. Geophys. Union, 119–151.
Jin, Y., Z. Liu, Z. Lu, and C. He, 2019: Seasonal cycle of background in the tropical Pacific as a cause of ENSO spring persistence barrier. Geophys. Res. Lett., 46, 13 371–13 378, https://doi.org/10.1029/2019GL085205.
Jin, Y., Z. Lu, and Z. Liu, 2020: Controls of spring persistence barrier strength in different ENSO regimes and implications for 21st century changes. Geophys. Res. Lett., 47, e2020GL088010, https://doi.org/10.1029/2020GL088010.
Karspeck, A. R., A. Kaplan, and M. A. Cane, 2006: Predictability loss in an intermediate ENSO model due to initial error and atmospheric noise. J. Climate, 19, 3572–3588, https://doi.org/10.1175/JCLI3818.1.
Kim, S. T., and F. F. Jin, 2011: An ENSO stability analysis. Part I: Results from a hybrid coupled model. Climate Dyn., 36, 1593–1607, https://doi.org/10.1007/s00382-010-0796-0.
Larson, S. M., and B. P. Kirtman, 2017: Drivers of coupled model ENSO error dynamics and the spring predictability barrier. Climate Dyn., 48, 3631–3644, https://doi.org/10.1007/s00382-016-3290-5.
Lau, K. M., and S. Yang, 1996: The Asian monsoon and predictability of the tropical ocean–atmosphere system. Quart. J. Roy. Meteor. Soc., 122, 945–957, https://doi.org/10.1002/qj.49712253208.
Levine, A. F. Z., and M. J. McPhaden, 2015: The annual cycle in ENSO growth rate as a cause of the spring predictability barrier. Geophys. Res. Lett., 42, 5034–5041, https://doi.org/10.1002/2015GL064309.
Levine, A. F. Z., and M. J. McPhaden, 2016: How the July 2014 easterly wind burst gave the 2015–2016 El Niño a head start. Geophys. Res. Lett., 43, 6503–6510, https://doi.org/10.1002/2016GL069204.
Levine, A. F. Z., M. J. McPhaden, and D. M. W. Frierson, 2017: The impact of the AMO on multidecadal ENSO variability. Geophys. Res. Lett., 44, 3877–3886, https://doi.org/10.1002/2017GL072524.
Liu, Z., Z. Lu, X. Wen, B. L. Otto-Bliesner, A. Timmermann, and K. M. Cobb, 2014: Evolution and forcing mechanisms of El Niño over the past 21 000 years. Nature, 515, 550–553, https://doi.org/10.1038/nature13963.
Liu, Z., Y. Jin, and X. Rong, 2019: A theory for the seasonal predictability barrier: Threshold, timing, and intensity. J. Climate, 32, 423–443, https://doi.org/10.1175/JCLI-D-18-0383.1.
Lu, B., F.-F. Jin, and H.-L. Ren, 2018: A coupled dynamic index for ENSO periodicity. J. Climate, 31, 2361–2376, https://doi.org/10.1175/JCLI-D-17-0466.1.
Lübbecke, J. F., and M. J. McPhaden, 2014: Assessing the twenty-first-century shift in ENSO variability in terms of the Bjerknes stability index. J. Climate, 27, 2577–2587, https://doi.org/10.1175/JCLI-D-13-00438.1.
McPhaden, M. J., 2003: Tropical Pacific Ocean heat content variations and ENSO persistence barriers. Geophys. Res. Lett., 30, 1480, https://doi.org/10.1029/2003GL016872.
McPhaden, M. J., 2015: Playing hide and seek with El Niño. Nat. Climate Change, 5, 791–795, https://doi.org/10.1038/nclimate2775.
McPhaden, M. J., and X. Zhang, 2009: Asymmetry in zonal phase propagation of ENSO sea surface temperature anomalies. Geophys. Res. Lett., 36, L13703, https://doi.org/10.1029/2009GL038774.
Moore, A. M., and R. Kleeman, 1996: The dynamics of error growth and predictability in a coupled model of ENSO. Quart. J. Roy. Meteor. Soc., 122, 1405–1446, https://doi.org/10.1002/qj.49712253409.
Nicholls, N., 1979: A simple air–sea interaction model. Quart. J. Roy. Meteor. Soc., 105, 93–105, https://doi.org/10.1002/qj.49710544307.
Nitta, T., and S. Yamada, 1989: Recent warming of tropical sea surface temperature and its relationship to the Northern Hemisphere circulation. J. Meteor. Soc. Japan, 67, 375–383, https://doi.org/10.2151/jmsj1965.67.3_375.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. 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, https://doi.org/10.1029/2002JD002670.
Ren, H.-L., and R. Wang, 2020: Distinct growth rates of the two ENSO types. Geophys. Res. Lett., 47, e2020GL088179, https://doi.org/10.1029/2020GL088179.
Ren, H.-L., F.-F. Jin, B. Tian, and A. A. Scaife, 2016: Distinct persistence barriers in two types of ENSO. Geophys. Res. Lett., 43, 10 973–10 979, https://doi.org/10.1002/2016GL071015.
Stein, K., N. Schneider, A. Timmermann, and F.-F. Jin, 2010: Seasonal synchronization of ENSO events in a linear stochastic model. J. Climate, 23, 5629–5643, https://doi.org/10.1175/2010JCLI3292.1.
Stein, K., A. Timmermann, N. Schneider, F.-F. Jin, and M. F. Stuecker, 2014: ENSO seasonal synchronization theory. J. Climate, 27, 5285–5310, https://doi.org/10.1175/JCLI-D-13-00525.1.
Timmermann, A., F. F. Jin, and J. Abshagen, 2003: A nonlinear theory for El Niño bursting. J. Atmos. Sci., 60, 152–165, https://doi.org/10.1175/1520-0469(2003)060<0152:ANTFEN>2.0.CO;2.
Tippett, M. K., and M. L. L’Heureux, 2020: Low-dimensional representations of Niño 3.4 evolution and the spring persistence barrier. npj Climate Atmos. Sci., 3, 24, https://doi.org/10.1038/s41612-020-0128-y.
Torrence, C., and P. J. Webster, 1998: The annual cycle of persistence in the El Nino/Southern Oscillation. Quart. J. Roy. Meteor. Soc., 124, 1985–2004, https://doi.org/10.1256/smsqj.55009.
Wang, B., 1995: Interdecadal changes in El Niño onset in the last four decades. J. Climate, 8, 267–285, https://doi.org/10.1175/1520-0442(1995)008<0267:ICIENO>2.0.CO;2.
Wang, B., and S. An, 2002: A mechanism for decadal changes of ENSO behavior: Roles of background wind changes. Climate Dyn., 18, 475–486, https://doi.org/10.1007/s00382-001-0189-5.
Webster, P. J., and S. Yang, 1992: Monsoon and ENSO: Selectively interactive systems. Quart. J. Roy. Meteor. Soc., 118, 877–926, https://doi.org/10.1002/qj.49711850705.
Wright, P. B., 1979: Persistence of rainfall anomalies in the central Pacific. Nature, 277, 371–374, https://doi.org/10.1038/277371a0.
Yu, J. Y., and H. Y. Kao, 2007: Decadal changes of ENSO persistence barrier in SST and ocean heat content indices: 1958–2001. J. Geophys. Res., 112, D13106, https://doi.org/10.1029/2006JD007654.
Zhang, R.-H., F. Zheng, J. Zhu, and Z. Wang, 2013: A successful real-time forecast of the 2010–11 La Niña event. Sci. Rep., 3, 1108, https://doi.org/10.1038/srep01108.
Zhu, J., A. Kumar, and B. Huang, 2015: The relationship between thermocline depth and SST anomalies in the eastern equatorial Pacific: Seasonality and decadal variations. Geophys. Res. Lett., 42, 4507–4515, https://doi.org/10.1002/2015GL064220.
Zhu, J., A. Kumar, B. Huang, M. A. Balmaseda, Z. Z. Hu, L. Marx, and J. L. Kinter, 2016: The role of off-equatorial surface temperature anomalies in the 2014 El Niño prediction. Sci. Rep., 6, 19677, https://doi.org/10.1038/srep19677.
Zuo, H., M. A. Balmaseda, S. Tietsche, K. Mogensen, and M. Mayer, 2019: The ECMWF operational ensemble reanalysis–analysis system for ocean and sea ice: A description of the system and assessment. Ocean Sci., 15, 779–808, https://doi.org/10.5194/os-15-779-2019.
