Emerging Pacific Quasi‐Decadal Oscillation Over the Past 70 Years

Quasi‐decadal variation in various Pacific regions has been recognized for decades but its spatial‐temporal characteristics and generation mechanisms remain poorly understood. Here we show that the sea surface temperature (SST) in the equatorial central Pacific (ECP) exhibits a prominent 11‐year periodicity since 1951 but not before. An episodic‐like quasi‐decadal warm event is initiated in a northeast‐southwest tilted belt from the United States' west coast to the ECP. The initial development involves positive feedback between atmospheric heating‐induced Rossby waves and underlying SST in the tropical North Pacific. The amplification of ECP warming is primarily due to equatorward heat advection and deepening thermocline, while the zonal advective feedback mainly controls its decay. The quasi‐decadal oscillation (QDO) tends to follow the 11‐years solar cycle with a 1‐to‐2‐year phase delay, meanwhile, it possibly links to the nonlinear behavior of ENSO. The cause of intensification of the QDO over the past 70 years, however, remains elusive.

Observational studies found QDO and ENSO signals share similar spatial structure and relative phasing of SST and SLP anomalies, but QDO evolution is amplified from the northeast tropical Pacific into the central Pacific (e.g., Tourre et al., 2001), and the warmest SST anomalies displaced into the central-western equatorial Pacific (e.g., Allan, 2000). The majority of these studies are based on narrow band-pass filtered data, leading to exaggeratedly periodic QDO pattern. Thereby, the QDO is generally regarded as an extended Pacific ENSO signal, and its characteristic spatial structure and development are diluted. As such, the mechanisms that trigger and control the QDO evolution remain poorly understood.
The origins of the Pacific quasi-decadal variability remain controversial, involving both the internal dynamics and external forcing. The phase transition of the QDO was thought to be governed by an ENSO-like delayed action oscillation mechanism (Battisti & Hirst, 1989;Schopf & Suarez, 1988) in the tropical Pacific (White et al., 2003b;White & Liu, 2008a, 2008b. However, the QDO could trigger further poleward off-equatorial Rossby waves in the central tropical North Pacific (Capotondi & Alexander, 2001). On the other hand, observation and conceptual model results suggested that the Pacific QDO is modulated by the 11-years solar cycle (White et al., 2003a;White & Liu, 2008a, 2008b. Moreover, the decadal solar oscillation effects on the tropical Pacific have been explored widely. During the solar peak years, the intensified precipitation in the intertropical convergence zone and south Pacific convergence zone strengthens the Hadley and Walker circulations, increasing trade winds, enhancing equatorial ocean upwelling, and lowering eastern equatorial Pacific SSTs Van Loon et al., 2004. However, the cooling is significantly different from the La Niña event in the stratosphere structure (Van Loon & Meehl, 2008).
The QDO can significantly influence tropical cyclone activity and continental climate anomalies (Anderson et al., 2016;S. Wang et al., 2014). This study aims to address the following three questions: (a) What are the distinct spatial-temporal structure of Pacific QDO? (b) What physical processes control their development and decay? And (c) what factors set up the decadal time scale of the Pacific QDO?

Data
The monthly mean SST data are obtained from the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) (Rayner et al., 2003), and the National Oceanic and Atmospheric Administration Extended Reconstructed SST (ERSST) version 5 (Huang et al., 2017). To reduce each SST data set's uncertainty, we have made a "merged" monthly mean SST data set by simply taking their arithmetic means over 1871-2018. The ocean reanalysis data sets, including potential temperature, zonal and meridional currents, vertical velocity, and total downward heat flux at the surface, are derived from the Global Ocean Data Assimilation System (GODAS) with 0.33° latitude × 1° longitude grid for 1980-2018 (Behringer & Xue, 2004).

Pacific QDO Index and its Characteristic Spatial-Temporal Structure
To detect the geographic locations where the SST variation has a significant quasi-decadal (9-to 13-year) signal, we conducted spectral analysis over each grid in the Pacific. The significant quasi-decadal signals are found primarily in a northeast-southwest tilted belt from the west coast of the United States to the equatorial Central Pacific (ECP) near the dateline ( Figure S1). The ECP has the most prominent and season-independent decadal signal and is the action center of the Pacific QDO. For convenience, we define an ECP index by the mean SST anomalies averaged over (10°S-10°N, 165°E-165°W).
The power spectrum of the yearly-mean ECP index ( Figure 1a) exhibits a conspicuous 11-year periodicity during 1951-2018 ( Figure 1b). This 11-year peak is also seen from the interannual variation of each seasonal mean SST anomalies, i.e., March-April-May (MAM), June-July-August (JJA), September-October-November (SON), and December-January-February (DJF), suggesting that the QDO exists throughout the seasonal cycle ( Figure S2). The index can be used to represent the Pacific QDO, which is also referred to as the QDO index hereafter.
Can the QDO be seen in a long record? We find that the quasi-decadal peak is absent during 1871-1950 ( Figure S3). This absence might be affected by the degraded quality of the SST data, but it likely suggests the nonstationary nature of the QDO. Given the QDO phenomenon has become significant after the 1950s, our analysis will focus on the past 70 years.
To depict coherent QDO structure and to find how a QDO warm event evolves, we first use the annual mean, detrended data to construct the lead-lag correlation maps with reference to the yearly mean QDO index from the −3 to +3 years ( Figure 2 and Figure S4). Note that the lead-lag regression maps show essentially the same evolution processes as shown by the lead-lag regression maps ( Figure S5).
During the transition phases (-3 and +3 years), the significant SST anomalies are almost absent, suggesting that the warming looks like an episodic event (Fig. S4). This view is different from the previous results that used narrowly bandpass-filtered data, which showed a cyclic-like event with significant continuing signals in the transition phases (e.g., Tourre et al., 2001   The blue (red) curve represents the upper bound of the Markov red noise spectrum at the 90% (95%) confidence level. (c) Lead-lag correlation coefficient between the seasonal mean ECP index and seasonal mean TSI during 1951-2018. The blue (red) lines indicate statistical significance at the 90% (95%) confidence level with the effective degrees of freedom ranging from 52 to 57 (Bretherton et al., 1999). A positive value for the season in (c) means TSI leads ECP index. Un-detrended SST data were used. ECP, equatorial central Pacific; TSI, total solar irradiance. . Shown are the lead-lag correlation maps of the yearly mean anomalies of (a) SST, (b) precipitation (color), SLP (contour), and 1,000-hPa winds (vectors) with reference to the annual mean ECP index during 1951-2018 from −2 years to +2 years. The dotted areas denote significance at the 90% confidence level, which is conducted by a two-tailed t-test, in (a) for SST and in (b) for precipitation. The red solid (blue dashed) lines represent the positive (negative) pressure, the contour interval is 0.05, and only the contours and vectors significant at a 90% confidence level are shown in (b). The black boxes show the key locations used in Figure 3. SLP, sea-level pressure.
Two years before the QDO peak warming, a significant cooling is seen in the eastern equatorial Pacific coupled with an SLP seesaw with an anomalous low over the western Pacific and an anomalous high over Central America ( Figure 2). The eastern Pacific cooling generates northward flow and increases precipitation in the subtropical eastern North Pacific (ENP) (10°N-25°N, 150°W-120°W), followed by significant warming over the Gulf of Alaska and a northeast-southwestward tilted warming belt from Baja California to the ECP at −1 year ( Figure 2). The QDO warming then rapidly amplifies at the ECP and reaches its peak at year 0 with a mega-El Niño-like pattern (Wang et al., 2013a).
However, different from El Niño, the warming develops from the tropical-subtropical North Pacific rather than along the equator; besides, the maximum warming occurs near the dateline (160°E-160°W) and the SLP seesaw anomalies are equatorially symmetric rather than biased to the southern hemisphere (SH) as in ENSO. The off-equatorial triangle warming resembles a PDO pattern, but a notable equatorial asymmetry is seen in the eastern Pacific with enhanced warming along the northern hemisphere (NH) ITCZ.
The QDO warming rapidly decays along the equator after the peak. From +1 to +2 years, the SH warming disappears while a remnant NH warming remains visible. Notably, the equatorial eastern Pacific (EEP) cooling is absent in the QDO decay phase, indicating a development-decay asymmetry and an NH-SH asymmetry in the Pacific QDO life cycle. The development and decay processes seem to feature a standing expansion/retraction without apparent propagation of the SST anomaly. Thus, Pacific QDO appears to be a standing oscillation.
Note that Figure 2 is made of yearly mean data, which does not fully filter interannual variability. To further remove the ENSO signals, we used 3-years running mean data to present the QDO evolution ( Figure S6). The evolution shown in Figure S6 resembles, to a large extent, the evolution shown in Figure 2, suggesting that the QDO evolution shown in Figure 2 is robust. But, by removal of the ENSO signals, the warming in the equatorial far eastern Pacific disappears, suggesting that QDO is more concentrated in the central Pacific.

Initiation and Development Mechanisms of the QDO Warming
Where does the ECP warming originate? How is the warming triggered? To address these questions, we have to examine the seasonal mean evolutions. Based on Figure 2, we select several air-sea interaction variables at various QDO active locations to examine their time evolutions and lead-lag relationships ( Figure 3).
The earliest precursor that preludes the ECP warming is the significantly positive SLP anomalies over Central-North America at −10 season (Figure 3a), which, by generating equatorial easterly anomalies, favors the EEP cooling that peaks in −8 and −7 seasons ( Figure 3b). Meanwhile, the EEP cooling enhances the subtropical ENP (10°N-25°N, 150°W-120°W) precipitation, which starts at −7 season and reaches its peak around −5 season ( Figure 3c). The chain scenarios from Figure 3a to 3c suggest that the central-North American rising pressure and EP cooling triggers the initial development of the ENP precipitation.
The enhanced ENP precipitation is followed by decreasing pressure in the subtropical central North Pacific (CNP, 5°N-25°N, 160°E-160°W) (Figure 3d). With the deepening of the anomalous CNP low, the climatological northeasterly trade winds weaken, as shown by the negative wind speed anomalies around −5 and −4 season (Figure 3e). The weakened trades lead to rapid warming in the ENP, as evidenced by the maximum SST tendency at −4 season (Figure 3e).
The chain scenarios from Figure 3c to 3e, suggest that the development of QDO warming in the subtropical NP may involve positive atmosphere-ocean feedback. This positive feedback is characterized by the atmospheric heating-induced Rossby wave response and associated dipolar SST anomalies (warming to the southeast and cooling to the northwest of the Rossby low) (Wang et al., 2000;Wang et al., 2013b). As shown in Figure 2 (panel −1 year), to the southeast of the cyclonic (low-pressure) anomalies, the weakened trades-reduced evaporation and entrainment cooling induce sea surface warming. On the other hand, the southwestward tilted warming belt from Baja California to the ECP enhances moisture convergence and precipitation heating that, in turn, generates ascending Rossby-wave low to strengthen the cyclonic anomaly during the westward propagation of the Rossby waves.
Of interest is that while the ENP and ECP warming is nearly simultaneous (Figure 3f), the initial warming at the ENP occurs earlier than ECP, suggesting the QDO warming may extend from ENP southwestward to the ECP. Note also that the ENP warming decays slower than the ECP, meaning that the decay of QDO warming may extend from ECP northeastward to ENP.

Equatorial Feedback Mechanisms for the Amplification and Decay of ECP Warming
The ECP's fastest-warming and cooling occur around the −4 and +4 seasons (Figure 3f). To reveal coupled dynamic processes involving in the evolution of the QDO over the ECP, we conducted an ocean mixed layer heat budget analysis (Text S1). Due to the limitation of GODAS data, the analysis is only made for 1980-2018. The diagnosed temperature tendency (Sum, 0.075 °C/month) is generally consistent with the observed tendency (0.093 °C/month) in the development stage (−5 to −3 season) ( Table 1). The nonlinear term is negligibly small, and the surface heat flux tends to damp the warming during both the development and decay stages.
The rapid amplification of ECP warming is mainly attributed to the meridional heat advection by mean (0.039 °C/month) and anomalous currents (0.026 °C/month), and the vertical advection of anomalous temperature by mean upwelling (0.037 °C/month). The two zonal advective terms tend to offset each other, contributing little to the ECP warming. In contrast, the ECP warming decay is primarily controlled by the zonal heat advection (−0.130°C/month) and augmented by the vertical advection by mean upwelling due to CHUNHAN ET AL.
10.1029/2020GL090851 6 of 11 To qualitatively illustrate the major terms' contribution in the ECP SSTA development and decay stages, we present a schematic diagram (Figure 4) with the highlight as follows. The climatological mean thermal state of the ECP mixed layer is characterized by (a) negative (westward) zonal temperature gradients (−4.0 × 10 −4°C /Km), (b) positive (upward) vertical temperature gradients (6.2°C/Km), and (c) poleward meridional temperature gradients (positive in the NH 6.7 × 10 −4°C /Km while negative in the SH, −1.5 × 10 −3°C /Km) as illustrated by the color shading. The climatological mean circulation is dominated by the averaged westward currents (−9.2 cm/s), poleward meridional currents (3.2 cm/s in the NH, and −2.5 cm/s in the SH), and upwelling (2.8 × 10 −6 m/s) as illustrated by the green arrows. Because of the equatorially symmetric feature of the meridional advection, only the NH part of the meridional temperature advection is discussed.
Anomalous quantities are qualitatively illustrated in the figure except the upwelling as its heat advection is negligibly small (Table 1). During the amplification stage (Figure 4a), three main processes contribute to the rapidly rising SST. The first is the anomalous southward meridional currents (  v <0) in the NH that transports mean-state warmer seawater equatorward, contributing to ECP warming. The second contributor is the mean northward currents that transport anomalously cooler water (   / T y < 0) away from the equator, reducing the heat loss by the northward mean currents' advection and enhancing the ECP warming. The third is the mean upwelling that transport anomalously warm water (   / T z < 0), reducing the cooling advected by the mean upwelling and increasing ECP SST. The negative anomalous vertical temperature gradient is mainly due to a deepening thermocline associated with the mixed layer currents convergence. During the decay stage of the QDO, the predominant contributor is the anomalous westward currents that transport cold water from the east to cool the ECP ocean mixed layer.

Discussion: What Determines the Time Scale of the QDO?
We present relevant evidence to support two hypotheses.

The 11-Year Solar Irradiance Forcing
The appealing aspect of the solar forcing hypothesis is to explain the time scale of QDO. The TSI shows a significantly increasing trend since 1871, and the intensity of the TSI variation was amplified to about 0.5 W/m 2 around 1950 (Figure 1a). During 1951-2018, the TSI cycle tends to lead the QDO by six (4-10) seasons (Figure 1c), i.e., the Central Pacific reaches a peak warming about 1-2 years after the maximum solar irradiance. However, the lead-lag correlation coefficient between the QDO and TSI is not significant during 1871-1950, conceivably due to the weak 11-year solar forcing (Figure 1a) as speculated by White (2006) and White et al. (1998). The regressed anomalies with reference to the ECP index are used in the heat budget computation during the development stage (mean of −5 season to −3 season) and decay stage (mean of the +3 to +5 season). Sum means the estimated temperature tendency. Nonlinear (NL) term is the sum of the    . The units are °C/ month. The dominant feedbacks in the development and decay of ECP warming are marked bold.

Table 1 Ocean Mixed Layer Heat Budget Analysis for the Development and Decay Stages of the QDO Over the Equatorial Central
Pacific (5°S-5°N, 165°E-165°W).
The problematic aspect of the solar forcing hypothesis is to establish a cause-effect relationship. Previous studies have shown that a weak La Niña-like response occurs during the 11-year solar cycle peak followed by a Central Pacific El Niño-like response 1−2 years later Misios et al. 2016Misios et al. , 2019. By analyzing the SST and SLP anomalies with reference to the composite TSI peak years (1957, 1968, 1979, 1989, 2000, and 2014), we find that a La Niña condition occurs one year before the solar peak ( Figure S7). At the solar peak, an anomalous low (cyclone) developed in the subtropical NP, reducing the northeasterly trade wind, resulting in ENP warming. One year after the solar peak, ECP warming reaches its maximum. Phenomenologically, the evolution of anomalies associated with the TSI resembles that of QDO except with a one-year lead, suggesting that the QDO might be a response to the 11-year solar cycle. The solar forcing may induce QDO by triggering the eastern Pacific cooling. Besides, the subtropical ocean is an enormous heat reservoir; thereby, the impacts of the enhanced solar forcing may be 'memorized' by the ocean. However, the precise processes that link the TSI and QDO remain elusive.

Summary
It has been difficult to describe the defining features of Pacific QDO because ENSO, PDO, and IPO dominate the Pacific SST variability whereas QDO cannot be clearly identified on the basin scale. This work describes QDO as a distinct regional mode of SST variability in the tropical North Pacific and explains the coupled dynamics responsible for its characteristic evolution.
We find that the QDO is a nonstationary phenomenon with an eye-catching 11-year periodicity during 1951-2018 but not during 1871-1950. The ECP warming appears to be an episodic-like and quasi-standing event, originated in a northeast-southwest tilted belt from the west coast of the United States to the ECP. QDO warming development involves both positive feedback between atmospheric heating-induced Rossby waves and underlying SST in the tropical North Pacific and equatorial meridional advective and thermocline feedback. In contrast, the decay of warming relies on zonal advective feedback. The QDO dynamics distinguish from ENSO dynamics. The QDO warming develops in the subtropical North Pacific by off-equatorial atmosphere-ocean interaction rather than along the equator as in ENSO, and the amplification of ECP warming mainly due to meridional heat advection, not the zonal advection feedback as in ENSO.
The decadal time scale of the QDO might be regulated by the 11-year solar cycle and/or the nonlinear behavior of ENSO. The first is hinted by the fact that QDO lags TSI by 1-to-2 years and the maximum TSI seems to trigger the subtropical North Pacific warming of QDO. The latter is implied by the relationship between the clustering El Niño or persisting La Niña. The two hypotheses are distinctive but not mutually exclusive.
However, the precise physical processes that determine the quasi-decadal time scale remain to be established.
In addition to the origin of the QDO time scale, several other issues call for future investigation. The main question to be addressed is what has caused the Pacific QDO strengthening after 1950. Is the strengthening related to the enhanced 11-year solar cycle or other factors, especially the increasing greenhouse gas forcing? The QDO, the prominent mode of the SST variability in the equatorial central Pacific, involves a sequence of air-sea interaction in the subtropical North Pacific. Therefore, QDO's off-equatorial SST patterns in the developing stage (−3, −1, and 0 years) resemble those of two leading modes of the North Pacific decadal variability, i.e., NPGO (Di Lorenzo et al., 2008) and the Pan-Pacific Decadal variation mode or PDO (Nigam et al., 2020). However, the QDO also involves the equatorial coupled dynamics on the decadal time scale. The QDO primarily involves variability in the equatorial central Pacific and subtropical North Pacific and linkage with South Pacific through the ECP teleconnection. On the other hand, the PDO and NPGO involve more variabilities and processes in the subarctic-midaltitude North Pacific. Previous studies have also shown that the Atlantic variability may trigger the PDO-like SST pattern (Gong et al., 2020;Johnson et al., 2020;Yang et al., 2020;Zhang & Delworth, 2007). Therefore, the linkages between the QDO and the tropical inter-ocean basin interactions require further investigation. Besides, given the critical location of the QDO, its impacts on tropical and global climate variations deserve in-depth exploration.