Effects of Equatorial Ocean Current Bias on Simulated El Niño Pattern in CMIP6 Models

This study utilized the Coupled Model Intercomparison Project Phase 6 (CMIP6) models to examine the simulations of equatorial ocean currents and explore their substantial influences on the systematic bias of westward‐extended sea surface temperature anomalies (SSTA) pattern during El Niño. The results show that models simulate an excessive westward ocean current field over the equatorial central Pacific in the mean state. It tends to suppress the equatorial eastward ocean current anomalies with their maximum centering over the equatorial western Pacific in the El Niño developing phase. As a consequence, an overestimated zonal advective feedback toward the maritime continent exists, subsequently inducing the biased westward extension of SSTA pattern. Our results show that the mean‐state performance of equatorial ocean currents plays a key role on simulations of El Niño evolution in CMIP6 models.


• The Coupled Model Intercomparison
Project Phase 6 (CMIP6) models simulate an excessive westward ocean current field over the equatorial central Pacific in the mean state • Overestimated zonal advective feedback in the warm pool region is the dominant factor for the westward extension of El Niño pattern • Mean-state performance of equatorial ocean current field plays a key role on the simulations of El Niño evolution in CMIP6 models

Supporting Information:
Supporting Information may be found in the online version of this article.
Correspondence to: L.-C. Wang, lcwang@atm.ncu Pacific cold tongue is one of the prevalent mean-state biases in CGCMs (G. Li et al., 2015Li et al., , 2016G. Li & Xie, 2012Wei et al., 2021;Wengel et al., 2021;T. Zhang & Sun, 2014). This relatively cold SST is linked with an unrealistic Walker circulation, biasing the ENSO simulation. For example, the amplifying wind-SST feedback and damping heat flux-SST feedback, both dependent on the location of Walker circulation ascending zone, are underestimated in a variety of CGCMs (Bellenger et al., 2014;S. T. Kim et al., 2014;Lloyd et al., 2009). These two feedback biases may cause unreasonable ENSO phase locking and weak ENSO asymmetry (Bayr et al., 2019). It has also been stated in previous studies that the simulation of the seasonal mean state related to the cold tongue bias in the equatorial Pacific may largely affect the ability of models to simulate ENSO phase locking (Liao et al., 2021), and the cold tongue bias can be a possible reason for the weak ENSO asymmetry with few multiyear La Niña events captured in the CMIP 5/6 models (Fang & Yu, 2020). In addition, the corresponding westward migration of convective response is the main factor that contributes to the westward extension of ENSO SSTA pattern (Bayr et al., 2018(Bayr et al., , 2019. On the other hand, limited attention has been made in the relations between the performance of modeled equatorial ocean currents and ENSO SSTA, even though the oceanic factors also play a role in controlling the ENSO diversity. The velocities of equatorial zonal currents (e.g., North Equatorial Countercurrent, South Equatorial Current and Equatorial Undercurrent) reflect largely on the location of maximum ENSO SSTA (Johnson et al., 2002;L. C. Wang & Wu, 2013). The direction of equatorial upper layer currents can be a predictor of the extreme El Niño events as the climatological zonal currents all change to eastward ocean flows in April-May (W. . For dynamical mechanisms, the thermocline feedback (TH) and the zonal advective feedback (ZA) act as the two dominant feedbacks in the recharge/discharge oscillator that conduce to the growth and phase transition of ENSO (Jin & An, 1999;Ren & Jin, 2013;Q. Zhang et al., 2007). In recent studies, the Bjerknes Stability index (Jin et al., 2006) and the associated oceanic feedbacks have been utilized to examine the ENSO behaviors in Coupled Model Intercomparison Project Phase 6 (CMIP6) models (Jiang et al., 2021;Wengel et al., 2021). Wengel et al. (2021) indicated that the essential ocean mesoscale processes related to ENSO are not well simulated in the lower-resolution greenhouse warming projections. Jiang et al. (2021) further suggested that the biased ENSO pattern in the western Pacific is the consequence of biased ZA associated with a larger zonal SST gradient there.
In this study, we use CMIP6 models to investigate the behaviors of simulated equatorial ocean currents and the effects of their oceanic feedback on El Niño SSTA. Given that the robust biases in the zonal currents over the equatorial Pacific basin are responsible for the misestimated ZA, which to some extent reduce the ability for simulating the El Niño structure, this study is aimed to examine the role of ocean current dynamics on El Niño performance via model simulations. We describe the dataset and the methodology for selecting El Niño events in Section 2. In Section 3, the results are presented. Conclusion and discussions follow in Section 4.

Datasets of Models and Observations
Fourteen CMIP6 models (Eyring et al., 2016) that are available with complete variables including SST, sea surface height (SSH), wind stress, potential temperature, and ocean currents were selected for the current analyses. Please see Table 1 for the detailed description about the selected models. The SST from Met Office Hadley Center Sea Ice and Sea Surface Temperature (HadISST) (Rayner et al., 2003), and the SSH, wind stress, potential temperature, and ocean currents from Global Ocean Data Assimilation System (GODAS) based on the National Centers for Environmental Prediction (Behringer & Xue, 2004) were applied in this study. For the purpose of comparing with the outputs from CMIP6 models, the dataset in HadISST and GODAS are referred to as an observation-based data. All the applied variables in models and observations are monthly data over the period 1980-2014, which were uniformly interpolated into 1° × 1° grid for the ease of assessments. The monthly anomalies are calculated by removing annual cycles and linear trends in this study.

Selection of El Niño Events
For identifying El Niño events, we adopt the criterion as follows: The averaged Niño-3 (5°N-5°S, 150°W-90°W) SSTA of El Niño during ND(0)J(+1) exceed 0.75 standard deviation (STD) (Chen et al., 2016). Here NDJ, the average from November to January, indicates the El Niño mature phase, and (0) and (+1) represent the El Niño year and the following year, respectively; moreover, the developing phase is delineated as the averaged period from April(0) to November(0).

Behaviors of Climatological Mean Field in CMIP6
To evaluate the basic ability of models in simulating the equatorial Pacific circulation system, the bias of mean-state zonal flow patterns in CMIP6 multi-model ensemble (MME) was examined ( Figure 1a; the original  mean-state patterns from the observation and CMIP6 MME are shown in Figure S1 in Supporting Information S1). Noting that the result shows an extremely strong westward zonal current over the equatorial Pacific basin. The area with this symbolic error is defined as the extreme bias region (EBR; 3°S-3°N, 180°-120°W) for the further analyses of how the biased ocean current's influence on ENSO SSTA. Since the mean equatorial Pacific circulation system is dominated by zonal geostrophic currents under the balance between the meridional pressure gradient force and Coriolis force (Lukas, 2001), the meridional SSH gradient (MHG) in the tropics is considered to be primarily responsible for the modification of the pattern of equatorial currents. We discovered that the bias of MHG simulation has a high intermodel correlation coefficient (−0.88) with the westward zonal current bias over the EBR region (see the scatter plot in Figure S2 in Supporting Information S1). In other words, the strengthened MHG tends to be a possible source of the westward ocean current bias in the equatorial Pacific in CMIP6 models.
As a comparison, the SST simulation in climatological mean field was also inspected (Figure 1b) to assess the extent of cold tongue bias, which has been mostly viewed as a key factor for causing an unrealistic ENSO SSTA in the coupled models. However, the underestimated SST over the equatorial Pacific basin depicts that an excessive cold tongue still persists in CMIP6 MME. It tends to be accompanied by the poleward divergent wind stress and expanded trade wind field west of 160°E over the EWP. To give a quantitative information for the aforementioned climatological discrepancies, a Taylor diagram (Taylor, 2001) is conducted to estimate the biases in the simulated structure of SST and zonal currents (Figure 1c). Each point denotes the different modeled behavior against the observation (reference point in Figure 1c). It is worth noting that the two clusters are explicitly separated. The pattern correlations of SST (zonal current) structure in the models are all larger (smaller) than 0.75. On the other hand, most of the CMIP6 models produce reasonable amplitudes of both zonal currents and SST patterns in the tropical Pacific. The ratios of standardized deviations of the two simulated fields are almost near 1, while only the GISS-E2.1-G model simulates weaker amplitudes with the ratios lower than 0.75. Besides, the values of the root-mean-square difference between the models and the observation, exhibited as the distance between the model and reference point, are smaller for SST pattern than for zonal currents pattern in the models. In brief, the performances of CMIP6 simulation in the Taylor diagram show that models have a worse capability in reproducing zonal current field. The biased ocean dynamics associated with this crucial issue might also be partly responsible for the biased simulation of El Niño patterns in CMIP6 models.

Effects of Biased Zonal Current Simulations on El Niño SSTA
Owing to the finding of obvious systematic bias in the modeled climatological zonal currents field, its interannual variability could be influenced to a certain extent. Figures 2a and 2b respectively illustrate the zonal current anomalies during the developing phase of El Niño events in observation and CMIP6 MME. Notable differences can be found in the equatorial flow pattern between these two figures. As shown in the observed result, the maximum of eastward current anomalies attains 1.5 ms −1 in the equatorial Pacific basin and even reaches 2 ms −1 in the EEP. The model simulations, on the contrary, display weaker current anomalies with the maximum centering near the maritime continent. The bias map demonstrated an east-west asymmetric pattern that the significant positive (negative) anomalies are revealed in the EWP (EEP) Pacific within 3°S-3°N during the developing phase of El Niño (Figure 2c).
To examine how the climatological bias impacts the interannual variability among the CMIP6 models, we calculated the intermodel regression map of the zonal current anomalies in the El Niño developing phase onto the mean zonal current averaged in the EBR (Figure 2d). The current velocity in the climatological mean field was inverted (multiplied by −1) so that greater bias could be shown via larger magnitude. The positive (negative) regression coefficients indicate that a stronger mean zonal current in the EBR tends to result in a stronger (weaker) zonal current anomaly during the development of El Niño. Focusing on the equatorial region, Figure 2d shows negative regression coefficients nearly throughout the Pacific basin, and the positive coefficients lie between 180° and 160°W at around 5°S. The contrast between Figures 2c and 2d suggests the weakened intensities of eastward zonal current anomalies over the equatorial Pacific are suppressed by the strengthened mean-state zonal current in the EBR. Therefore, the maximum of eastward zonal current anomalies may consequently shift toward the maritime continent in CMIP6 simulations compared to the observation.
A mixed-layer heat budget analysis is conducted to diagnose how the oceanic dynamical processes behave during ENSO events under the biased equatorial Pacific circulation system. In general, the TH and ZA are two dominant dynamics related to the recharge/discharge mechanism (Jin & An, 1999) that positively contribute to the mixed layer temperature anomalies during the El Niño development (Chen et al., 2016;Kug et al., 2009Kug et al., , 2010Xu et al., 2020). Following the heat budget equation from Ren and Jin (2013), the TH and ZA terms can be respectively expressed as where the overbar and prime denote the climatology and anomaly. T indicates ocean temperature; u and w indicate zonal and vertical current velocity. H represents constant mixed layer depth of 50 m. Subscript "sub" means the averaged subsurface layer from 50 to 100 m.
To contrast the dynamics between EEP and EWP, the time series of TH and ZA averaged over Niño-3 and Niño-4 during El Niño is shown in Figure 3. During the El Niño developing phase, models generally capture a realistic TH evolution over Niño-3 that the values in CMIP6 MME and the observation are quite similar (Figure 3b). The slightly underestimated TH is however observed over Niño-4 in most of the models (Figure 3a), which could be explained via the particularly strong velocity of mean-state westward zonal current in EBR, leading to a suppressed mean upwelling (described as in Equation 1) in the EWP. Different from the other models' simulations, the magnitude of TH in MIROC6 model (gray line in Figures 3a and 3b) has significantly enhanced from May(−1) before the El Niño develops and results in an overestimation till its decaying phase. On the contrary, most of the CMIP6 models present larger (smaller) magnitude of ZA against the observation over Niño-4 (Niño-3) during the developing phase of El Niño (Figures 3c and 3d). Such a stronger (weaker) modeled feedback tends to correspond with the significant overestimated (underestimated) zonal current anomalies pattern west (east) of 150°W in the equatorial Pacific (Figure 2c). These results suggest that the ZA simulation is sensitively affected by the behavior of zonal current anomalies in CMIP6 models.
To explore how the biased dynamical feedback affect the behavior of El Niño structure, the ND(0)J(+1) SSTA derived from the CMIP6 MME with its bias against the observation were investigated (Figure 4a). Despite using the state-of-the-art CGCMs, excessively warm SSTA with the discrepancy exceeding +0.5°C still exists in the warm pool region. We further statistically evaluated the link between the magnitudes of El Niño-related SSTA (mature phase) and ZA (developing phase) among the 14 CMIP6 models (Figures 4b-4e). In the Niño-4 region, it is shown that the models with a comparatively strong ZA tend to simulate a relatively warm SSTA with the intermodel correlation coefficient up to 0.81 (Figure 4d), whereas the underestimated TH (Figures 3a and 4b) plays a less important role in the contribution to the warm SSTA that the intermodel correlation coefficient is relatively lower (0.55) in most of the models. In the Niño-3 region, the SSTA is dominated by TH with the intermodel correlation coefficient of 0.77 (Figure 4c), and the CMIP6 MME does not show significant biases in the SSTA structure in the EEP (Figure 4a). Therefore, the biased ZA in the EWP is essential to induce a westward-shifted warm mixed layer temperature that results in a westward expansion of SSTA in the El Niño mature phase in CMIP6 simulations.

Conclusions and Discussions
The outputs of the CMIP6 historical run were adopted to determine the influential factors of equatorial ocean currents for the westward extension of El Niño SSTA. Two systematic mean-sate errors are found over the tropical Pacific: the excessive cold tongue and the bias in the zonal current field over EBR. Via a quantitative examination based on the Taylor diagram, the climatological bias in zonal currents is shown to act as a key factor in reducing the ability of models with realistic simulations on El Niño-related ocean dynamics. Through the analyses of intermodel regression, it is demonstrated that this climatological ocean current bias is the origin of the weakened equatorial zonal current anomalies in the equatorial Pacific. The maximum of such anomalies field consequently locates too far toward the west during the developing phase of El Niño. Since the simulated ZA