Interannual Variability of Lower Equatorial Intermediate Current Response to ENSO in the Western Pacific

The ocean currents of the tropical Pacific Ocean vary with El Niño‐Southern Oscillation (ENSO) cycles. A mooring time series obtained during 2014–2018 in the western Pacific reveal that interannual variability extends to Lower Equatorial Intermediate Current (LEIC). The LEIC velocity anomalies are significantly correlated with the Niño‐3.4 index at an 11‐month lag. Monthly velocity data from the global ocean physical reanalysis product and from a linear continuously stratified ocean model during 1993–2018 capture the 2015–2016 signal and are used to identify the underlying mechanism. During El Niño (La Niña) events, the direct wind forcing in the central tropical Pacific and the reflection of Kelvin waves in the eastern Pacific during the first autumn‐winter period contribute comparably to an eastward (westward) current anomaly in the western Pacific during the second autumn‐winter period. This process is achieved mainly through the generation of the second baroclinic mode and off‐equatorial westward‐propagating downwelling (upwelling) Rossby waves.


Introduction
The Equatorial Intermediate Current (EIC) system contributes to global circulation, climate variability, and distributions of heat, watermass, and biogeochemical quantities (e.g., Ascani et al., 2015;Brandt et al., 2011;Czeschel et al., 2011;Hu et al., 2015;Margolskee et al., 2019;Stramma et al., 2010). As a major component of the EIC system in the Pacific Ocean, the westward-flowing Lower Equatorial Intermediate Current (LEIC) is found beneath the Equatorial Undercurrent (EUC) and EIC between 2°N and 2°S, ranging from 600 to 1,200 m with a core at~800 m. The LEIC weakens and shoals toward the east (Cravatte et al., 2017;Marin et al., 2010).
A recent review by Ménesguen et al. (2019) summarizes observations and formation mechanisms of intermediate currents and stresses the important role of deep equatorial intraseasonal variability in their formation. Furthermore, the EIC and LEIC seasonal variations and their underlying mechanisms have been well studied. The mean westward-flowing EIC and LEIC can seasonally reverse to eastward-flowing currents, ©2020. The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.  (Cravatte et al., 2017;Firing, 1987;Firing et al., 1998;Gouriou et al., 2006). Such a seasonal cycle is generated by the vertical propagation of the first meridional mode (l ¼ 1) Rossby wave, which is directly forced by the westward-propagating annual equatorial zonal wind fields (Kessler & McCreary, 1993;Marin et al., 2010).

10.1029/2020GL089311
Due to the sparse and fragmented observations, the interannual variability of LEIC is rarely studied. Under the dominant influence from the El Niño-Southern Oscillation (ENSO), the upper equatorial currents in the Pacific, including the EUC, the South Equatorial Current (SEC), and the North Equatorial Countercurrent, have exhibited significant interannual variability (e.g., Izumo, 2005;Johnson et al., 2000;Shinoda et al., 2011;Zhao et al., 2013). Using 16-month time series current measurements at 159°W spanning the 1982-1983 El Niño episode, Firing (1987) revealed that the zonal currents at EIC and LEIC depths were westward in May 1982 but eastward in May 1983. A similar interannual signal was also detected at 138°E with a 4-year continuous record from moored current meters during 1998-2002 (Kawabe et al., 2008). The underlying dynamics of the interannual variability, however, were not discussed in these studies. Based on Argo gridded temperature-salinity data and drifting velocity data, Johnson and Birnbaum (2016) revealed a strengthening of the westward-flowing EIC in the western Pacific 3 months after the peak of an El Niño. However, these Argo data are still insufficient to robustly determine the mechanisms that drive the interannual variability of the Pacific EICs (Zanowski et al., 2019). Nevertheless, investigating the mechanisms of intermediate currents at the interannual timescale is very important for climate and oceanographic research. This will help us to understand how and why the intermediate currents respond to climate variability and change.
From September 2014 to December 2018, a subsurface mooring was deployed at 0°, 142°E to monitor the variability of LEIC. This mooring is a part of the Scientific Observing Network of the Chinese Academy of Sciences in the western Pacific (Lyu et al., 2018;Ma et al., 2019;Song et al., 2018;. The temporal coverage of the mooring data allows the investigation of the interannual variability of LEIC in response to the strong 2015-2016 El Niño event. Our approach is to combine data analyses on mooring observation, reanalysis product, and numerical experiments with a linear continuously stratified (LCS) ocean model.

Mooring Data
The subsurface mooring has one upward-and one downward-looking TRDI (Teledyne RD Instruments) 75 kHz Acoustic Doppler Current Profilers (ADCPs) mounted on the main float at~500 m. The ADCPs returned the velocity profiles over the depth range of 50-950 m with a vertical bin size of 8 m and a frequency of 35 pings per hour. The data were first interpolated vertically to 1-m intervals and then averaged to derive the monthly mean for analysis.

LCS Model
To elucidate the wind-driven equatorial wave dynamics, the LCS ocean model is used. This model is configured to the Pacific Ocean between 20°S to 20°N and 100°E to 70°W with horizontal resolutions of 0.25°× 0.25°. The background potential density is obtained from World Ocean Atlas 2013 temperature and salinity data averaged over 5°S to 5°N and 130°E to 80°W. The resultant profile is then interpolated from 0 m to 4,000 m with a 10 m vertical resolution. Vertical modes are computed based on background density, and the characteristic wave speeds (c n ) for the first four baroclinic modes are 266, 159, 100, and 72 cm s −1 , respectively. These values are consistent with those of a similar computation in Yu and McPhaden (1999). For model equations and parameterizations, readers can refer to Chatterjee et al. (2013); Huang et al. (2018); and Shankar et al. (1996). In this study, the total solution is the sum of the first 10 baroclinic modes. The model is first spun up for 10 years forced by the monthly mean climatology of cross-calibrated multiplatform (CCMP) satellite winds during 1993-2018 (Atlas et al., 2008). The climatology run is then used as the restart state for all the other solutions. The main run, referred to as LCS-MR, intergrates forward in time for the period of 1993-2018 using monthly CCMP winds.

Reanalysis Product
To investigate the interannual variability of LEIC over a longer period than the observation, we use monthly velocity data during 1993-2018 from the global ocean physical reanalysis product GLORYS12V1 (hereinafter GL12) created by Mercator Ocean. The product has a horizontal resolution of 1°/12°in longitude/latitude and 50 vertical levels. For more details on the reanalysis product, readers are referred to Lellouche et al. (2018). For the EIC and LEIC, two striking features are revealed in the observation. First, a pronounced seasonal cycle was present with a strong westward-flowing phase during autumn-winter (boreal, same hereinafter) and a weak westward-or even eastward-flowing phase during spring-summer in 2014, 2015, 2017, and 2018. Second, unlike the above years, the strong westward-flowing phase atypically disappeared during autumn-winter in 2016 (Figure 1a), exhibiting significant interannual variability. This is particularly true for LEIC, which reversed to an eastward-flowing current during this period. The zonal current anomaly (U a , seasonal cycle removed) averaged over 600-820 m (depth of 820 m is selected according to the depth of available ADCP data) has the largest correlation of 0.82 with the Niño-3.4 index at an 11-month lag ( Figure 1j). The correlation coefficients here and subsequently are all significant at the 95% confidence level. This suggests that the atypically disappeared LEIC (i.e., strong eastward U a ) during autumn-winter in 2016 may be related to the 2015-2016 El Niño event (Figures 1e and 1f). Next, we will focus on the mechanisms underlying the LEIC interannual variability with the aid of GL12 and LCS.

Result
The performance of GL12 and LCS in simulating the LEIC is first examined. During the mooring measurement period, the aforementioned two striking features in LEIC are successfully reproduced in GL12 and LCS-MR (Figures 1b and 1c). The correlation coefficients (r) between the observed and modeled U a at the mooring site are significant (r > 0.52) at the LEIC depths ( Figure S1). Additionally, the GL12 and LCS-MR monthly U a averaged over 600-820 m at the mooring site agree reasonably well with the observation (Figure 1d). High correlation coefficients of 0.82, 0.82, and 0.79 are exhibited between observation and GL12, observation and LCS-MR, and GL12 and LCS-MR, respectively. Note that both GL12 and LCS-MR show a poor performance in the EIC interannual variability (r < 0.3; Figure S1). This is likely due to the overlapped depth ranges of EUC and EIC, which cannot be well reproduced by the models (Figures 1a-1c).
For a longer period during 1993-2018, the correlations of U a in GL12 and LCS-MR with the Niño-3.4 index (Figure 1g) both show a peak value larger than 0.61 at an approximately 11-month lag ( Figure 1j). This adds solid evidence that the LEIC interannual variability is closely associated with ENSO. Following the NOAA definition, an El Niño (La Niña) event is defined as a period when the 3-month running-mean of Niño-3.4 exceeds 0.5°C (falls below −0.5°C) for at least five consecutive months. During each 2-year El Niño event (i.e., 1994-1995, 1997-1998, 2002-2003, 2004-2005, 2006-2007, 2009-2010, and 2015-2016 El Niño events), the strong westward-flowing LEIC disappears during the second autumn-winter period (gray shadings in Figures 1h and 1i), corresponding with the presence of strong eastward U a . Additionally, during the 1997-1998 El Niño event, the modeled disappearance of LEIC from September to December 1998 is consistent with the mooring measurement records at 0°, 138°E (Kawabe et al., 2008, their Figure 3). Given that La Niña events usually have an opposite impact on currents compared to El Niño events, they may cause strong westward U a in the second autumn-winter period. This is basically established in the observation -2017and 2017, GL12, and LCS-MR (1995, 1998-2001, 2005, 2007-2009-2012-2017, and 2017. In the following, we will detail how El Niño events influence the LEIC interannual variability by examining the dynamics behind the presence of a strong eastward U a . Using the GL12 and LCS-MR outputs, we next construct the composite horizontal distribution of U a averaged over 600-820 m depths during the preceding and following October-November-December months (denoted as OND(0) and OND(1)) of the 2-year El Niño events. Both models reproduce the interannual changes of equatorial U a in the western Pacific caused by El Niño events, that is, weak eastward U a in GL12 or even westward U a in LCS-MR during OND(0), but strong eastward U a in both models during OND(1) (Figures 2a-2d).
As the winds are the only external forcing for the LCS-MR, the well-simulated LEIC interannual variability must bear a close relation to the wind changes during El Niño events. On this basis, we plot the longitude-time variations of the composite monthly zonal wind stress anomaly (τ x a_El Niño ) averaged between 5°S and 5°N during all El Niño events from 1993 to 2018 (Figure 3a). The anomaly is obtained by subtracting the monthly climatology to remove the seasonal cycle. The evolution of τ x a_El Niño can be divided into three stages: the developing stage from January(0) to August(0) with strengthening westerly τ x a_El Niño in the western Pacific, the mature stage from September(0) to April(1) with fully developed westerly τ x a_El Niño in the western and central Pacific and weaker easterly τ x a_El Niño in the eastern Pacific, and the decaying stage from May(1) to December(1) with gradually strengthening easterly τ x a_El Niño in the western Pacific. The τ x a_El Niño during the decaying stage indicates the establishment of La Niña-like conditions over the Pacific basin.
To isolate the key stage of the wind evolution for the presence of strong eastward U a among the three stages, we carry out four numerical experiments using composite monthly wind stresses of the El Niño events over the whole stages (LCS-Case0), only during the developing (LCS-Case1), mature (LCS-Case2), and decaying (LCS-Case3) stages, respectively. LCS-Case0 is used as a control run for the other three experiments. During the residual period in the other three experiments, the monthly climatological wind is exerted. Figure 2e shows that the LCS-Case0 reproduces a strong eastward U a along the equator during OND(1) in the western Pacific. Among the other three experiments, only LCS-Case2 shows a similar result (Figures 2f-2h). This suggests that the wind forcing during the mature stage of El Niño events plays a critical role in the presence of strong eastward U a during OND(1).
To further explore the dynamics associated with the winds during the mature stage of El Niño events, we examine the composite results from LCS-MR. Figures 3b-3d show the longitude-time plots of SLA along 3°S, the equator, and 3°N. During the mature stage, the westerly wind anomaly in the central Pacific excites equatorial downwelling Kelvin waves (SLA > 0, Figure 3c), which propagate into the eastern Pacific and are reflected back by the eastern boundary as off-equatorial downwelling Rossby waves in both hemispheres at the end of the preceding year (North et al., 2015; Figures 3b and 3d). Meanwhile, the anticyclonic wind stress curl associated with the eastern Pacific easterly wind anomaly can induce off-equatorial westwardpropagating downwelling Rossby waves. Under the dominance of the second baroclinic mode (as discussed later), the downwelling Rossby waves, including those from the reflection of Kelvin waves and direct wind forcing, can induce negative pressure anomaly fields (P a ) at 600-820 m along 3°S and 3°N; the values of which are more negative than those at the equator (Figures 3f-3h). The meridional structure of the pressure anomaly fields in the equatorial region is consistent with the meridional eigenfunction of the l ¼ 1 Rossby wave (Lukas & Firing, 1985) and can excite an eastward U a according to the equatorial geostrophy , where ρ is background density and β is meridional variation of the Coriolis parameter;  Chen et al., 2015;Figure 3e). The signal of this eastward U a propagates westward accompanied by off-equatorial westward-propagating Rossby waves at a phase speed of approximately 51 cm s −1 and arrives at the mooring site during September(1) to December(1). Thus, the absence of the strong westward-flowing LEIC in the western Pacific is a delayed (~11 months) ocean response to the strong equatorial wind anomaly over the central-to-eastern Pacific during the mature stage of El Niño events. The above conclusions are also established in GL12 ( Figure S2). From the wave dispersion relation (ω ¼ − kc n 2l þ 1 , where ω is frequency, k is zonal wavenumber, c n is characteristic wave speed, and l is meridional mode number; Kessler & McCreary, 1993), c n is estimated to be 153 cm s −1 (c

Geophysical Research Letters
which is close to that of the second baroclinic mode (c 2 ¼ 159 cm s −1 ). Furthermore, the longitude-time variation of U a is largely explained by the second baroclinic mode, which can be clearly inferred from Figures 3e and 3i-3l. These results suggest that the first meridional mode Rossby waves associated with the second baroclinic mode are mainly responsible for the LEIC interannual variability.
The off-equatorial downwelling Rossby waves can be produced from direct easterly wind forcing and reflection of equatorial Kelvin waves in the eastern Pacific. The relative importance of the above two originations deserves to be assessed. To do so, we conduct an experiment using a damper within the region of 7.5°S to 7.5°N and 85°W to 80°W in the eastern Pacific Ocean (LCS-Damp). The experiment is forced by the same winds of LCS-MR. The damper can efficiently absorb the energy of incoming equatorial Kelvin waves; therefore, no Rossby waves are reflected back into the ocean interior from the eastern boundary. We construct the composite zonal current anomaly at 0°, 142°E during El Niño events from LCS-MR and LCS-Damp, denoted as LCS-MR-Composite and LCS-Damp-Composite, respectively. The difference between them isolates the reflected Rossby wave effects, denoted as LCS-Reflect-Composite. During OND(1) when the strong westward-flowing LEIC disappears, U a induced by the direct wind forcing and reflected Rossby waves make comparable contributions to the total U a , inferred from the time-depth plots in Figures 4a-4c.
To further quantify this argument, we obtain the time series of U a averaged between 600 and 820 m from LCS-MR-Composite, LCS-Damp-Composite, and LCS-Reflect-Composite (Figure 4d). The correlations of total U a with that obtained from LCS-Damp-Composite and LCS-Reflect-Composite are 0.95 and 0.96, respectively, suggesting that their temporal variations are generally in phase. The root-mean-square deviations (RMSDs) of total U a with that obtained from LCS-Damp-Composite and LCS-Reflect-Composite are 3.36 and 2.70 cm s −1 , respectively. The comparable RMSDs further indicate that direct wind forcing and reflected Rossby waves are almost equally crucial for the LEIC interannual variability. This finding roughly applies to the LEIC over longitudes of 135°E to 165°E, and the reflected Rossby waves make more significant contributions to the interannual variability over the longitudes east of 165°E ( Figure S3).
For La Niña events, the mature stage wind also plays a decisive role in the interannual variability of LEIC compared to the developing and decaying stages (figure not shown). The zonal wind stress anomaly (τ x a_La Niña ) during the mature stage of La Niña generally has an opposite condition compared to El Niño events, that is, fully developed easterly τ x a_La Niña in the western and central Pacific and weaker westerly τ x a_La Niña in the eastern Pacific. The off-equatorial westward-propagating upwelling Rossby waves are excited by the cyclonic wind stress curl associated with the eastern Pacific westerly wind anomaly and by the eastern boundary reflection of equatorial upwelling Kelvin waves initially triggered by the easterly wind anomaly in the central Pacific. Under the dominance of the second baroclinic mode, the direct wind forcing and reflected upwelling Rossby waves can equally induce a signal of westward U a along the equator at the LEIC depths.

Summary and Discussion
Based on a 4-year mooring record of zonal currents at 0°, 142°E, the mean westward-flowing LEIC is found to have significant interannual variability. The LCS model and GL12 product faithfully reproduce the observed LEIC interannual variability and thus are used to explore its underlying dynamics. The correlations of the observed LEIC U a during 2014-2018 and GL12 and LCS modeled U a during 1993-2018 with the Niño-3.4 index all show significant peak values larger than 0.61 at an approximately 11-month lag. This suggests that the LEIC interannual variability is strongly influenced by the ENSO cycle.
By conducting a series of LCS experiments, we demonstrate that the interannual variability is closely associated with wind changes during the ENSO mature stage (i.e., September(0) to April(1)). For El Niño (La Niña) events, the off-equatorial westward-propagating downwelling (upwelling) l ¼ 1 Rossby waves are excited by the anticyclonic (cyclonic) wind stress curl associated with the equatorial easterly (westerly) wind anomaly in the eastern Pacific and by the eastern boundary reflection of equatorial downwelling (upwelling) Kelvin waves initially triggered by the westerly (easterly) wind anomaly in the central Pacific. Under the dominance of the second baroclinic mode, these Rossby waves make the second order meridional derivative of the intermediate-depth pressure anomaly from 3°S to 3°N negative (positive) and thus a signal of eastward (westward) U a at the equator according to the equatorial geostrophy. This signal of eastward (westward) U a in the eastern Pacific propagates westward and arrives at the mooring site in~11 months during September(1) to December(1). Rossby waves from the direct wind forcing and reflection of Kelvin waves make comparable contributions to the interannual variability of LEIC.
Considering that the observation period in this study only covers the 2015-2016 El Niño event, we cannot yet make full conclusions about the interannual viability of EICs. For example, given that the wind field evolution during each El Niño event is different, the discrepancies among the responses of the LEIC to different El Niño events are overlooked in our composite analysis and would be interesting topics for future studies.
The presence of the western boundary and the Papua New Guinea coast makes the western Pacific a narrowed accumulation region for intermediate currents. Figures 3f-3h show that anomalous intermediate 10.1029/2020GL089311 P a intensifies near the western boundary. This can generate an enhanced magnitude of equatorial U a (Figure 3e), suggesting the potential effect of the western boundary on LEIC variability in the western Pacific. Furthermore, the LEIC and the intermediate New Guinea Coastal Undercurrent (NGCUC) have an interconnected relationship. Previous studies (Kawabe et al., 2008;Zenk et al., 2005) and GL12 outputs (figure not shown) both reveal that the LEIC and NGCUC reverse seasonally with a nearly opposite phase. The seasonal-to-interannual variability of the LEIC and NGCUC can work together to influence the watermass distribution along the equator and in the entrance region of the Indonesian Throughflow (ITF), which is the only pathway for the interocean exchange of tropical waters between the Pacific and Indian Oceans (Gordon et al., 2008(Gordon et al., , 2019Gordon & Fine, 1996;Sprintall et al., 2014). When the LEIC reverses to an eastward-flowing current, the Antarctic Intermediate Water (AAIW) carried by the NGCUC mainly turns clockwise to spread eastward along the equator. In contrast, when the LEIC flows westward, the AAIW can intrude northward along the western boundary into the entrance region of the ITF (figure not shown). The question is still open, and more observations and numerical simulations are necessary to examine the intermediate circulation system in the western Pacific.