1. Introduction
Studies have suggested that the decadal North Pacific sea surface temperature anomalies (SSTAs) are dominated by their first empirical orthogonal function, often called the Pacific decadal oscillation (PDO) (Mantua et al. 1997; Zhang et al. 1997). Several studies have indicated that the PDO is not a single physical mode of oceanic variability, but rather the sum of several processes with different dynamic origins (Newman et al. 2003; Vimont 2005; Schneider and Cornuelle 2005; Qiu et al. 2007). Among them, variability of the North Pacific oceanic gyre circulation, El Niño–Southern Oscillation (ENSO) teleconnections, and the stochastic atmospheric forcing play equally important roles in the SSTA variability associated with PDO (Liu 2012). Alternatively, the decadal variability has been described in terms of regime shifts and abrupt coherent changes of the equilibrium states of the climate systems, the most pronounced of which in the past 50 years took place in 1976/77, affecting a large area of the Pacific basin (e.g., Trenberth and Hurrell 1994; Minobe 2000). The regime shifts and abrupt changes suggest nonlinearity of the decadal variability.
On decadal time scales, the tropical Pacific Ocean circulation is in a quasi-steady balance with the surface wind forcing, due to the fast propagation of the Rossby and Kelvin waves. The dynamics of the ocean general circulation in this quasi-steady state are controlled by surface Ekman pumping, which gives rise to the meridional movement of the ocean. The zonal currents are determined by the meridional transports through the continuity equation. Thus it is of great importance to understand the variability of oceanic meridional transports in the study of the decadal variability of the tropical North Pacific Ocean. The mean meridional transports of the ocean are traditionally understood in terms of Sverdrup theory (Sverdrup 1947), which assumes a linear dynamic framework and obtains the meridional transport of the wind-driven ocean circulation by integrating the wind stress curl, but provides no information about oceanic baroclinicity. So far, the decadal variability of the oceanic meridional transport has not been estimated based on observations. Their agreement with Sverdrup theory remains to be examined.
Sverdrup transports are found to underestimate the mean gyre circulation of the low-latitude northwestern Pacific Ocean significantly based on comparisons with Argo data collected in the past 10 years or so (Zhang et al. 2013; Yuan et al. 2014; Yang and Yuan 2016). Early studies indeed have shown that Sverdrup transport is inconsistent with observations in the tropical North Pacific Ocean (Meyers 1980; Hautala et al. 1994). Kessler et al. (2003) argued that the tropical Pacific Ocean circulation, which is critically important for the coupled ocean–atmosphere variations on interannual to decadal time scales, is not in Sverdrup balance with the winds. The difference has been attributed to various sources, including oceanic nonlinearity and inaccuracies of the wind stress calculation (Kessler et al. 2003; Yuan et al. 2014).
It is important to simulate the meridional ocean transport and its variations realistically in modern climate system models in the study and prediction of the decadal climate variations. However, most of the modern climate system models [e.g., models from phase 5 of the Coupled Model Intercomparison Project (CMIP5)] are of coarse resolution (ocean model resolution >1° in the zonal direction) and are essentially linear in dynamics so that their simulated steady-state ocean circulation follows Sverdrup dynamics. The decadal variability of the meridional transports in these climate system models remains to be examined using observational data.
The observed meridional transport variability needs to be simulated well in CMIP5 models to reproduce realistic structure and predictability of the decadal climate variations associated with ocean gyre variability. However, climate model simulations cannot be compared with observations directly, because coupled model simulations have internal variability and generally have phase differences with the real ocean and atmosphere variability. An effective way of assessing coupled model simulations is to compare simulated meridional geostrophic transports with Sverdrup theory computed from the model. If coarse-resolution climate system models simulate a meridional transport in close agreement with Sverdrup theory at the decadal time scales (as expected), but the observed transports differ from Sverdrup theory significantly, then the climate models can be considered deficient in simulating and predicting the observed decadal variability. Vigorous decadal variability in the tropical Pacific Ocean and its climatic effects suggests that this kind of study is important.
The data and methods of the study are described in the next section. In section 3, the decadal variations of the meridional ocean transport in the tropical North Pacific are investigated using the P-vector (Chu 1995) absolute geostrophic currents (AGCs) based on historical hydrographic data. Deviations from Sverdrup transport of the decadal variations of the meridional transports in the observations, in two of the climate system models, and in an ocean general circulation model are compared in section 4. In section 5, time-dependent Sverdrup transport is evaluated and compared with the steady-state Sverdrup transport. Sensitivity and errors of our analysis are evaluated. Conclusions are summarized in section 6.
2. Data and methods
a. Observational data
The temperature and salinity anomaly data used here are the yearly anomalies of 5-yr running mean (pentadal) global ocean heat and salt content data (GOHSCD; https://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/anomaly_data.html), which are built from the World Ocean Database 2009 (Boyer et al. 2009) and some additional data that had been processed through the end of 2016 (Levitus et al. 2012). The first frame of the pentadal data is the running mean of 1955–59 and the last frame is the running mean of 2012–16. The time series represent the 5-yr running means from 1957 to 2014. The horizontal resolution is 1° longitude × l° latitude. The vertical extent of the pentadal data is from the surface to the 2000-m depth of the ocean, which is consistent with the Argo profiles up to a nominal depth of 1750–2000 m.
The temperature and salinity (T/S) anomalies in each 1° square for the 5-yr running period were mapped using the objective analysis procedure of Locarnini et al. (2010). The anomalies are referenced to the World Ocean Atlas 2009 (WOA09) climatology. The density of seawater was calculated from these pentadal data by adding back the mean climatological values of the WOA09. In this way, the nonlinearity of the seawater state equation is considered in the calculation of the absolute geostrophic currents. The anomalies of the AGCs are then calculated based on the climatology of yearly pentadal data from 1955–59 to 2012–16. The above is the standard usage of the pentadal data suggested by the user guide. Comparisons between the geostrophic calculations based on the smoothed T/S data and the nonsmoothed T/S anomalies with the climatology added back show negligible small differences in the upper ocean, suggesting that the nonlinearity effect of the density state equation on the geostrophic velocity of the upper ocean is small.
b. Climate model simulations
The model simulations from the CMIP5 are used for this study. The analyses of two of the historical simulations using the GFDL-CM3 and CSIRO Mk3.6.0 models are presented in this study. The other simulations show similar structure of the ocean meridional transports, so their presentations and analyses are omitted here. The GFDL-CM3 historical simulation covers the period of 1955–2005. The CSIRO Mk3.6.0 simulation runs from 1950 through 2005.
In addition, the eddy-permitting simulation of the Ocean General Circulation Model (OGCM) for the Earth Simulator (OFES) (Masumoto et al. 2004) is used to diagnose the decadal variations of the meridional geostrophic transport in the North Pacific Ocean. The experiment is forced by the winds and heat fluxes of the National Centers for Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) reanalysis from 1950 through 2011. The surface salinity of the simulation is relaxed to the World Ocean Atlas 1998 (Levitus et al. 1998) climatology at a time scale of 6 days. The model horizontal resolution is 0.1° longitude × 0.1° latitude, with 54 levels in the vertical. Table 1 lists some parameters of these model configurations.
Information about models. T63 is a horizontal spectral resolution of approximately 1.875° lon × 1.875° lat, with 18 levels (L18) vertically.
c. Absolute geostrophic currents
Traditionally, the geostrophic currents are derived from the dynamic height calculation based on a level of no motion, the so-called reference level. The P-vector method, used here, determines the geostrophic velocity at the reference level assuming conservation of potential vorticity and density. The intersections of the potential vorticity and potential density surfaces, the so-called P-vectors, at any two different levels and the thermal-wind vector between the two levels form a closed triangle, which can be used to solve for the P-vector magnitudes. The method employs the least squares fitting if multiple levels of hydrographic data are used (Chu 1995, 2000). Dynamically, the P-vector method is equivalent to the β-spiral method under the Boussinesq and geostrophic approximation, but is able to control the errors of the calculation well through the use of the first-order potential density gradients only (Chu 2000; Yuan et al. 2014).
In this study, the AGCs are calculated from the hydrographic data of the upper 2000 m north of 5°N in the tropical North Pacific Ocean using the P-vector method between the 800- and 2000-m depths. Above 800 m, the geostrophic currents are calculated using the dynamic height calculation in reference to the AGCs at 800 m. This treatment is to avoid the P-vector calculation inside the surface mixed layer of the ocean (Zhang et al. 2013; Yuan et al. 2014), where the potential vorticity and potential density are subject to numerous ageostrophic effects. Our experience suggests that the magnitudes and structure of the AGCs are generally not sensitive to the choices of the depth ranges mentioned above, so long as the P-vector calculation is conducted significantly below the surface mixed layer.
d. Wind data and climate indices
We use the NCEP–NCAR reanalysis wind velocity products (1955–2016) and the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) wind velocity (1958–2001) to calculate the Sverdrup transport in this study. According to Yuan et al. (2014), the deviations from Sverdrup theory of the observed ocean circulation in North Pacific Ocean are not sensitive to the scheme of drag coefficients used in calculation of the wind stress (their Fig. 14). Here we use the same drag coefficient of Garratt (1977) as in Yuan et al. (2014). To be consistent with the GOHSCD dataset, the monthly wind data and model outputs are first averaged into annual means and then processed through the 5-yr running filter into the yearly pentadal data.
The PDO index used in this paper was downloaded from https://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/Data/pdo.long.data, which is the first principal component analysis (PCA) mode of the monthly SSTA in the North Pacific north of 20°N (Mantua et al. 1997). We use the same definition to calculate the model simulated PDO indices in this study. The Niño-3.4 index is downloaded from https://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/Nino34/, which is calculated from HadISST1 (Rayner et al. 2003).
e. Sverdrup balance
To compare with the AGCs based on the GOHSCD data, the AGCs of the CMIP5 and OFES models are calculated based on the temperature and salinity output of the models using the P-vector method. In addition, the model-simulated meridional velocity is integrated to examine the balance of Eq. (3), which shows consistent results with that of AGCs (figure is omitted).
The first term on the rhs is the meridional geostrophic heat transport (MGHT). The surface Ekman transport (per unit width), Vek = −τx/(ρ0f), where ρ0 is the average density (ρ0 = 1025 kg m−3) and Vek is independent of the Ekman layer thickness, is multiplied by the vertically averaged temperature in the top 50 m of the ocean to represent the surface Ekman heat transport (EKHT; Hobbs and Willis 2012). The calculated EKHT is not sensitive to the depth chosen because of the weak stratification in the surface Ekman layer of the tropical ocean, where vertical turbulent mixing is strong.
f. The lag correlation analysis and its significance
A lag correlation analysis is used to investigate the relationship between the decadal variation of the MGT and PDO index. The Student’s t test with Monte Carlo techniques (Livezey and Chen 1983) is used to assess the statistical significance of the correlations due to the reduced degrees of freedom under 5-yr running mean filtering.
g. The standard error of the mean
3. Observed MGT variations
The MGT is integrated from the 2000-m depth to the sea surface and from the eastern boundary of the North Pacific Ocean to the grids 1° away from the western boundary. Experimenting (integration to within 0°–2° from the western boundary) suggests that the MGT is not sensitive to the selected distance away from the western boundary. Because of the narrow width of the western boundary currents, the calculation has essentially excluded the transport of the western boundary currents and thus represents the interior geostrophic streamfunction. The mean MGT is calculated from the 58-yr pentadal data, based on which the decadal variability of the MGT is estimated and studied.
a. The mean meridional transport of the North Pacific Ocean
The mean MGT for the period of 1957–2014 is mostly southward south of 36°N in the North Pacific Ocean, which mainly represents the interior subtropical gyre circulation (Fig. 1a), with the maximum transport between the eastern boundary and south of Japan reaching 45 Sv (1 Sv ≡ 106 m3 s−1) to the south. Areas of northward MGT in the high latitudes and in the eastern equatorial Pacific ocean between 5° and 15°N are also identified with northward transport exceeding 5 Sv.
(a) Integrated mean meridional geostrophic transport, (b) the S–E in the North Pacific Ocean, and (c) their difference (Sv). The contour interval is 5 Sv, and the thick white line represents the zero contour line.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Integrated mean meridional geostrophic transport, (b) the S–E in the North Pacific Ocean, and (c) their difference (Sv). The contour interval is 5 Sv, and the thick white line represents the zero contour line.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Integrated mean meridional geostrophic transport, (b) the S–E in the North Pacific Ocean, and (c) their difference (Sv). The contour interval is 5 Sv, and the thick white line represents the zero contour line.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
The mean S−E transport of the same period (Fig. 1b) based on the NCEP–NCAR wind stress is calculated according to the right side of Eq. (3). The patterns of the S−E transport are qualitatively consistent with that of the MGT in the subtropical gyre and near the eastern boundary in the North Pacific Ocean, suggesting the dominance of wind-driven circulation in these areas. The agreement is also evidenced by the small differences between the S−E transport and MGT in three areas along roughly 15°, 21°, and 35°N in the western Pacific Ocean (Fig. 1c). The three latitudes correspond to the boundary of the tropical–subtropical gyres, the internal boundary of the northern and southern subtropical gyres (defined as the latitude of the change of the zonal current direction, corresponding roughly to the subtropical counter current), and the boundary of the subtropical and subpolar gyres, respectively.
However, the maximum southward MGT are not collocated with those of the S−E streamfunction, with the S−E maximum shifted southward compared with those of MGT. The differences between the S−E transport and MGT are the largest in the Kuroshio and Oyashio recirculation and in the tropical northwestern Pacific Ocean. It is known that the Kuroshio and Oyashio recirculation is driven by vigorous eddy activity (Kawai 1972; Hurlburt et al. 1996; Qiu 2001), the mean circulation of which does not obey the linear Sverdrup dynamics. So the discrepancies south of the Kuroshio and north of the Oyashio main streams are expected. The difference between the S−E and MGT transport is called the non-Sverdrup transport in this paper for convenience.
Significant non-Sverdrup transports are also found in the regions 5°–12°N and 17°–19°N west of the date line, with the maximum differences as large as 10–20 Sv. The S-E transport and MGT even have opposite signs in the area between 5° and 12°N west of 160°E. The significant non-Sverdrup circulation in the low-latitude northwestern Pacific Ocean has only been realized recently, the dynamics of which have been attributed to oceanic nonlinearity [see Yuan et al. (2014) for verification].
The non-Sverdrup mean circulation in the low-latitude northwestern Pacific Ocean is statistically significant. The standard errors of the mean MGT for a 95% confidence level in the northwestern Pacific Ocean are generally smaller than 1 Sv (Fig. 2). The maximum S−E transport SEM at the same confidence in the northwestern Pacific Ocean is 4.6 Sv. Uncertainty of MGT estimates due to density errors in the GOHSCD data is less than 0.1 Sv. The sum of the SEMs of the S−E and MGT is much smaller than the values of the non-Sverdrup mean streamfunction in the tropical northwestern Pacific Ocean (Fig. 1c), suggesting the statistical significance of the regional deviation from the linear Sverdrup dynamics of the ocean circulation.
Standard errors of the mean (a) geostrophic and (b) Sverdrup-minus-Ekman (S−E) transport estimated from the standard deviations of the time series during 1957–2014. The confidence level is at 95% of the Student’s t test. The contour interval is 0.5 Sv, and values smaller than 1 Sv are shaded. The thick dashed line represents the 1-Sv contour line.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Standard errors of the mean (a) geostrophic and (b) Sverdrup-minus-Ekman (S−E) transport estimated from the standard deviations of the time series during 1957–2014. The confidence level is at 95% of the Student’s t test. The contour interval is 0.5 Sv, and values smaller than 1 Sv are shaded. The thick dashed line represents the 1-Sv contour line.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Standard errors of the mean (a) geostrophic and (b) Sverdrup-minus-Ekman (S−E) transport estimated from the standard deviations of the time series during 1957–2014. The confidence level is at 95% of the Student’s t test. The contour interval is 0.5 Sv, and values smaller than 1 Sv are shaded. The thick dashed line represents the 1-Sv contour line.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
b. Decadal variations of the observed MGT
The tropical Pacific Ocean has active air–sea interactions and is also important in the initiation of ENSO. The fact that the mean ocean circulation deviates significantly from the Sverdrup circulation in the area suggests important new dynamics of the ocean circulation, which have not been realized before. It is, therefore, important to examine the dynamics governing the decadal variations of MGT across the tropical Pacific Ocean. The modern climate system models need to incorporate the proper nonlinear dynamics to simulate the MGT variations realistically when predicting decadal climate variations.
Figure 3a shows the time series of the MGT anomalies (MGTA) across the North Pacific tropical gyre derived from the pentadal AGC data. Here the MGTA time series, which are the integrated MGT across the entire interior Pacific ocean, are first averaged within 5°–7°N, 128°–138°E with the mean of 1957–2014 subtracted to remove aliasing by mesoscale eddies. The results are not sensitive to the size of the averaging area. A weakening trend associated with the weakening of the zonal currents (Hsin 2016) is also removed from the time series. Because of the use of the pentadal data, the influence of ENSO has largely been removed. Significant MGT variability associated with climate shift events appears. For example, the MGT shifts from a negative anomaly of 7.1 Sv in 1975 to a maximum positive anomaly of 11.9 Sv in 1979, which is coincident with the 1976/77 climate regime shift indicated by the PDO index. The maximum negative anomaly (−11.7 Sv) occurred in 2009, which may be related to the regime shift in the North Pacific from the mid-1990s to the mid-2000s (Minobe 2000; Chavez et al. 2003; Bond et al. 2003).
(a) Comparisons of the detrended MGTA (black) and EKTA (red) averaged in the area 5°–7°N, 128°–138°E. The shaded bars are the PDO index (°C). (b) Lag correlations between MGTA and the PDO (black) and between EKTA and the PDO (red). Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Comparisons of the detrended MGTA (black) and EKTA (red) averaged in the area 5°–7°N, 128°–138°E. The shaded bars are the PDO index (°C). (b) Lag correlations between MGTA and the PDO (black) and between EKTA and the PDO (red). Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Comparisons of the detrended MGTA (black) and EKTA (red) averaged in the area 5°–7°N, 128°–138°E. The shaded bars are the PDO index (°C). (b) Lag correlations between MGTA and the PDO (black) and between EKTA and the PDO (red). Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Both the geostrophic and surface Ekman transports contribute significantly to the total meridional transport of the tropical oceans. The decadal variations of the EKT anomaly (EKTA) in the tropical North Pacific Ocean are found to have comparable amplitudes with that of the MGTA (Fig. 3a), even though the mean Ekman transport is much larger than the mean MGT in the tropics because of the strong trade winds and the small value of the Coriolis parameter (see the appendix and Fig. A1). Moreover, MGTA and EKTA are found to be consistently out of phase with each other. MGTA and PDO have in-phase variations with each other except for a small time lag whereas EKTA and PDO have out-of-phase variations. The lag correlations between MGTA and PDO and between EKTA and PDO confirm these kinds of relations (Fig. 3b). Significant correlations occur when MGTA and EKTA lead the PDO indices by about 1–3 years with the maximum absolute values of the correlation coefficient of 0.77 and −0.63, respectively, occurring at the lead time of 1 year. The correlations suggest that more than 59% and 40% variance of the PDO are associated with the MGTA and EKTA, respectively, in the tropical North Pacific Ocean on decadal or longer time scales.
Because of the out-of-phase variations of MGTA and EKTA, the correlation between the total meridional transport anomalies (MGTA + EKTA) and the PDO index is reduced, but still significant at the lead time of 1–3 years (Figs. 4a,b). This is consistent with stronger variations of MGTA than those of EKTA (Fig. 3a). We have also calculated the total MHT anomalies (MHTA) according to Eq. (4) in the same fashion as the calculation of the MGTA. MHTA varies in phase with MGTA, suggesting the dominance of ocean gyre circulation in the meridional heat transport in this area. High correlations between MHTA and the PDO index are identified, with the maximum value of 0.75 occurring when MHTA leads the PDO by 1 year.
(a) Comparisons of the anomalies of the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red). The shaded bars are the PDO index (°C). (b) Lag correlations between anomalies of the total meridional transport (blue), non-Sverdrup transport (black), and the total meridional heat transport (red) and the PDO, respectively. Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The thin black dashed line marks the 95% confidence level.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Comparisons of the anomalies of the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red). The shaded bars are the PDO index (°C). (b) Lag correlations between anomalies of the total meridional transport (blue), non-Sverdrup transport (black), and the total meridional heat transport (red) and the PDO, respectively. Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The thin black dashed line marks the 95% confidence level.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Comparisons of the anomalies of the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red). The shaded bars are the PDO index (°C). (b) Lag correlations between anomalies of the total meridional transport (blue), non-Sverdrup transport (black), and the total meridional heat transport (red) and the PDO, respectively. Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The thin black dashed line marks the 95% confidence level.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
It is important to understand the dynamics of MGTA and MHTA. From the comparison between the mean states of the geostrophic streamfunction based on the GOHSCD data and of the S−E streamfunction based on the NCEP–NCAR winds, we have identified that the ocean circulation deviates significantly from the Sverdrup dynamics in the tropical North Pacific Ocean. Do these discrepancies also exist on the decadal time scales? To answer this question, we have estimated the decadal variations of the non-Sverdrup transport and examined its relationship with the PDO index (Figs. 4a,b). (The deviation of total meridional transports from time-dependent Sverdrup transports is discussed in section 5.) Significant correlations occur when the non-Sverdrup transport anomalies lead the PDO index by about 1–3 years, with the maximum correlation coefficient of 0.56 occurring at the lead time of 1–2 years. This maximum correlation is even higher than that of the total meridional transport anomaly, suggesting the importance of the non-Sverdrup circulation in the predictability of PDO.
The significant lead correlations with the PDO index of both the meridional transport anomalies and heat transport anomalies suggest the predictability of the PDO. The 1–3-yr lead correlations between the meridional transport anomalies or the meridional heat transport anomalies and PDO exist in a broad band from 4° to 12°N, with the strongest correlations occurring in the band between 5° and 7°N (see Fig. A2), suggesting that the identified PDO predictability is a robust character in the tropical North Pacific Ocean circulation. In particular, the MGTA shifts from its maximum negative value in 2009 to almost zero in 2014, suggesting a neutral period of PDO in recent years.
One of the mechanisms driving PDO variability has been suggested to be North Pacific subtropical gyre variability influenced by the tropical air–sea interactions (Nakamura et al. 1997). Although the exact mechanisms behind PDO are not well understood, the significant lead correlations between MGTA (and MHTA) in the tropical North Pacific Ocean and the PDO index provide a piece of evidence, based on historical observations, suggesting a tropical influence on PDO.
The PDO is a long-lived El Niño–like pattern of the Pacific climate variability (Tanimoto et al. 1993; Zhang et al. 1997; Latif and Barnett 1996). The significant lead correlations between MGTA and the PDO index can be explained by the discharge and recharge of the warm pool. During a “warm” or “positive” phase, the western Pacific becomes cooler and part of the eastern Pacific Ocean warms up; during a “cool” or “negative” phase, the opposite patterns occur. Positive MGTA indicates anomalous northward transport of warm water and a heat loss in the tropical Pacific Ocean, which favors a discharge condition and should be associated with a positive PDO phase. The opposite holds for a negative PDO phase.
The PDO SSTA at middle latitudes has been suggested to be associated with the tropical SSTA through the atmospheric bridge (Zhang et al. 1997). This is consistent with the fact that the MGTA across the middle latitudes is not correlated significantly with the PDO index. Our analyses have shown that discharges and recharges of the warm pool on the decadal time scales are consistent with the observed PDO variations. The analysis suggests the predictability of PDO at a lead time of 1–3 years using MGTA as a precursor in the tropical Pacific Ocean.
The above analyses have suggested that the non-Sverdrup MGT is a predictor of PDO variability. To our knowledge, this is the first oceanic predictor of the PDO identified in the observations. In comparison with the non-Sverdrup transport anomalies, the Sverdrup transport anomalies (SVA) have smaller amplitudes than those of the total meridional transport anomalies (Fig. 4a) and are not correlated to the PDO index (Fig. 5a). The lagged correlations between SVA and the PDO index are not significant statistically (Fig. 5b), suggesting a lack of predictability of PDO based on the wind-driven circulation.
(a) Time series of the meridional SVA averaged in the area 5°–7°N, 128°–138°E, and the shaded bars are the PDO index (°C). (b) The lag correlations between the SVA and PDO. Negative lags indicate PDO is lagging. Black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Time series of the meridional SVA averaged in the area 5°–7°N, 128°–138°E, and the shaded bars are the PDO index (°C). (b) The lag correlations between the SVA and PDO. Negative lags indicate PDO is lagging. Black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Time series of the meridional SVA averaged in the area 5°–7°N, 128°–138°E, and the shaded bars are the PDO index (°C). (b) The lag correlations between the SVA and PDO. Negative lags indicate PDO is lagging. Black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
4. Decadal variations of the simulated MGT and Sverdrup transports
The significant lag correlations between the non-Sverdrup oceanic transport anomalies and the PDO index in observations suggest that the non-Sverdrup transport variability in the tropical North Pacific Ocean can be used as a precursor for PDO predictions. The identified predictability of PDO raises the issue if modern climate system models contain this PDO predictability, given that most of the ocean models in these climate system models are coarse resolution and are essentially controlled by linear Sverdrup dynamics. In this section, the predictability of the simulated PDO using the simulated MGTA as a precursor is investigated, based on which the dynamics of the PDO predictability are discussed.
The ability of modern climate system models in simulating the realistic variability and predictability of the climate systems relies critically on the realistic simulation of the meridional ocean transport. For this purpose, the decadal variations of the total meridional volume transport, the non-Sverdrup transport, and their relations with the PDO indices in the CMIP5 model simulations are examined.
Analyses have shown that the non-Sverdrup gyre circulation in the tropical North Pacific Ocean is absent in most of the CMIP5 simulations, except CSIRO Mk3.6.0 and a few other models (Li and Yuan 2018, manuscript submitted to Ocean Modell.). Here, we analyze the CSIRO Mk3.6.0 simulation as an example and the GFDL-CM3 model, whose simulation of the mean state of the North Pacific Ocean circulation is consistent with the Sverdrup dynamics, as a comparison. We have also examined the meridional volume transport variations in the OFES model as an example of an eddy-resolving ocean general circulation model simulation. Because the coupled climate system models contain internal variability and generally have phase differences with observed ocean and atmosphere variability, the PDO indices are calculated from the simulated data in the correlation analysis. Since we are interested in the variations on decadal and longer time scales, all the modeled data are smoothed in the same way as that of the observational data in this study.
Figures 6a–c show the comparisons with the modeled PDO indices of the decadal variations of the total meridional transport, the non-Sverdrup transport, and the Sverdrup transport integrated across the North Pacific Ocean and averaged in the band of 5°–7°N, 128°–138°E in the three models, respectively. EOF analyses demonstrate that the first EOF mode in the simulated SSTA is indeed the PDO mode (see Fig. A3). The corresponding lag correlations are displayed in Figs. 6d–f, respectively. All three model simulations underestimate the observed decadal variations of the total meridional transport anomalies (cf. Fig. 4), with the GFDL-CM3 model being the weakest. Total meridional transport anomalies and Sverdrup transport anomalies track each other well in the GFDL-CM3 model, suggesting that the wind-forced Sverdrup balance dynamically dominates the model. This agreement is also indicated by the smaller amplitudes of the non-Sverdrup transport anomalies and by the similar correlations of either the total or the meridional Sverdrup transport anomalies with the PDO indices shown in Fig. 6e. The simulated decadal variations of either the total or the non-Sverdrup transports have insignificant lag correlations with the PDO indices, suggesting that the simulated PDO in these models and the observed PDO are governed by different dynamics. The non-Sverdrup transports or the total volume transports cannot be used as precursors for predicting the simulated PDO in these models (Figs. 6d–f).
Simulated anomalies of (a) the total meridional transport (MGTA plus EKTA; blue), (b) the non-Sverdrup transport (black), and (c) the Sverdrup transport (SVA; red) averaged in the area 5°–7°N, 128°–138°E in comparison with the standardized model PDO indices (shaded bars). The reference period is 1955–2005. Also shown are lag correlations between (d) the total meridional transport and PDO (blue), (e) the non-Sverdrup transport and PDO (black), and (f) the SVA and PDO (red). Negative lags indicate PDO lagging. Thick black dashed lines mark the 95% significance levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Simulated anomalies of (a) the total meridional transport (MGTA plus EKTA; blue), (b) the non-Sverdrup transport (black), and (c) the Sverdrup transport (SVA; red) averaged in the area 5°–7°N, 128°–138°E in comparison with the standardized model PDO indices (shaded bars). The reference period is 1955–2005. Also shown are lag correlations between (d) the total meridional transport and PDO (blue), (e) the non-Sverdrup transport and PDO (black), and (f) the SVA and PDO (red). Negative lags indicate PDO lagging. Thick black dashed lines mark the 95% significance levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Simulated anomalies of (a) the total meridional transport (MGTA plus EKTA; blue), (b) the non-Sverdrup transport (black), and (c) the Sverdrup transport (SVA; red) averaged in the area 5°–7°N, 128°–138°E in comparison with the standardized model PDO indices (shaded bars). The reference period is 1955–2005. Also shown are lag correlations between (d) the total meridional transport and PDO (blue), (e) the non-Sverdrup transport and PDO (black), and (f) the SVA and PDO (red). Negative lags indicate PDO lagging. Thick black dashed lines mark the 95% significance levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
The above comparisons suggest that the climate model simulations of the decadal MGT variations in the North Pacific Ocean are dynamically deficient in terms of the non-Sverdrup gyre simulation. The eddy-resolving OFES ocean model is then examined and found to fail to simulate the observed significant lead correlations of MGTA in the tropical Pacific with the PDO index (Fig. 6d). It is known that climate models can produce PDOs with similar spatial patterns to observations, but their time scales differ widely from each other (Furtado et al. 2011; Newman et al. 2016). The consistent spatial patterns in climate models are an indication that the atmospheric circulation is simulated reasonably once the SST patterns are simulated well. The disparity of the time scales suggests a high sensitivity of the decadal variability to ocean–atmosphere interactions and/or oceanic dynamics. The failed simulation of the significant lead correlations between MGTA and PDO in the CMIP5 models may be due to the poor simulation of the discharge–recharge time scales of the warm pool in the climate models.
The simulated MGTA has been compared with the Sverdrup balance to assess the coupled model simulations. Non-Sverdrup transport anomalies are calculated by subtracting the S−E streamfunction from the meridional geostrophic streamfunction (Fig. 7). The zonal integration is from the eastern boundary to 130°E. The focus is on the variability in the band between 5° and 7°N in the North Pacific Ocean. The non-Sverdrup transport between the observed MGT and the S−E streamfunction based on the NCEP–NCAR wind data show vigorous decadal variations (Fig. 7a). The simulated non-Sverdrup transports differ significantly from the observed both in amplitude and phase. The non-Sverdrup transport anomalies in OFES model are qualitatively in agreement with those observed before the mid-1970s but are poorly consistent with the observed after that period. Not only are the lead correlations between the simulated non-Sverdrup transport anomalies and the PDO indices insignificant, but also the non-Sverdrup transport variations differ from the observations. The discrepancies and insignificant lead correlations between MGTA and PDO suggest the failure of the coupled climate models in simulating and predicting realistic PDOs (Kuball 2007).
Time–longitude plots of the (a) observed and (b)–(d) simulated non-Sverdrup meridional transport anomalies (Sv) averaged between 5° and 7°N. The reference period is 1955–2005. The contour interval is 2 Sv.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Time–longitude plots of the (a) observed and (b)–(d) simulated non-Sverdrup meridional transport anomalies (Sv) averaged between 5° and 7°N. The reference period is 1955–2005. The contour interval is 2 Sv.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Time–longitude plots of the (a) observed and (b)–(d) simulated non-Sverdrup meridional transport anomalies (Sv) averaged between 5° and 7°N. The reference period is 1955–2005. The contour interval is 2 Sv.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
The non-Sverdrup gyre circulation discussed here basically represents the circulation driven by oceanic nonlinearity. One may calculate the difference between the observed MGTA and those simulated by a linear model or a coarse-resolution ocean general circulation model. The identified predictability of the PDO essentially lies in the difference of the observed MGTA from the linear simulations, suggesting that the PDO is likely associated with nonlinear processes of the tropical ocean circulation.
5. Discussion
a. Latitudinal dependence
The above analysis of the observational data suggests predictability of the decadal variability of the Pacific Ocean using the interior meridional transport of the tropical Pacific Ocean along 5°–7°N as a precursor. We have also investigated the same kind of predictability based on the meridional transports in northern latitudes. Figure 8 shows the comparisons of the total meridional volume transport anomalies, the total meridional heat transport anomalies, and the meridional non-Sverdrup transport anomalies integrated from the eastern boundary and averaged between 13° and 15°N and between 17° and 19°N, respectively, in the longitudinal range of 128°–138°E. Their correlations with the PDO index based on the pentadal observations are also shown. No significant lead correlations between the volume transport anomalies and the PDO indices at these latitudes are identified (Figs. 8b,d). Only the meridional geostrophic heat transport anomalies (MGHTA) show a marginally significant simultaneous correlation of 0.48 with the PDO index in the band of 13°–15°N, which disappears in the band of 17°–19°N farther north. These analyses suggest that the predictability of PDO comes mainly from the tropical gyre heat content of the Pacific Ocean.
(a),(c) Comparisons of the anomalies of the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red) averaged in the areas 13°–15°N, 128°–138°E and 17°–19°N, 128°–138°E, respectively. The shaded bars are the PDO index. (b),(d) Lag correlations between the total meridional transport and PDO (blue), between the non-Sverdrup transport and PDO (black), and between the total meridional heat transport and PDO (red). Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The thin black dashed lines in (b) and (d) mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a),(c) Comparisons of the anomalies of the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red) averaged in the areas 13°–15°N, 128°–138°E and 17°–19°N, 128°–138°E, respectively. The shaded bars are the PDO index. (b),(d) Lag correlations between the total meridional transport and PDO (blue), between the non-Sverdrup transport and PDO (black), and between the total meridional heat transport and PDO (red). Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The thin black dashed lines in (b) and (d) mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a),(c) Comparisons of the anomalies of the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red) averaged in the areas 13°–15°N, 128°–138°E and 17°–19°N, 128°–138°E, respectively. The shaded bars are the PDO index. (b),(d) Lag correlations between the total meridional transport and PDO (blue), between the non-Sverdrup transport and PDO (black), and between the total meridional heat transport and PDO (red). Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The thin black dashed lines in (b) and (d) mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
The area-averaged MGTA and MGHTA in 5°–7°N, 128°–138°E are also found to be significantly correlated with interdecadal Pacific oscillation index, which is a Pacific-wide manifestation of the PDO (Mantua et al. 1997) (figure omitted). These correlations suggest that the identified predictability of the decadal variability is robust.
The errors of the Sverdrup transports associated with the wind stress errors have been evaluated based on a comparison of the NCEP–NCAR winds with the ERA-40 surface winds during the period of 1958–2001. The SVA values calculated from the NCEP–NCAR and the ERA-40 wind data, using the drag coefficient of Garratt (1977), show consistent variations with each other except before 1963 (Fig. 9a). No significant lead–lag correlations between ERA-40 SVA and PDO are found, which is consistent with the analysis using the NCEP–NCAR winds (cf. Fig. 5). Thus, the identified predictability of the decadal variability in this study is suggested to be independent of the wind products used in the analyses.
(a) Time series of the meridional SVA averaged in the area 5°–7°N, 128°–138°E calculated from NCEP–NCAR (black) and ERA-40 (red) wind data respectively. The reference period is from 1961 to 2000, and the shaded bars are the PDO index. (b) Lag correlations between the SVA calculated from ERA-40 wind data and the PDO index. Negative lags indicate PDO lagging. Black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Time series of the meridional SVA averaged in the area 5°–7°N, 128°–138°E calculated from NCEP–NCAR (black) and ERA-40 (red) wind data respectively. The reference period is from 1961 to 2000, and the shaded bars are the PDO index. (b) Lag correlations between the SVA calculated from ERA-40 wind data and the PDO index. Negative lags indicate PDO lagging. Black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Time series of the meridional SVA averaged in the area 5°–7°N, 128°–138°E calculated from NCEP–NCAR (black) and ERA-40 (red) wind data respectively. The reference period is from 1961 to 2000, and the shaded bars are the PDO index. (b) Lag correlations between the SVA calculated from ERA-40 wind data and the PDO index. Negative lags indicate PDO lagging. Black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
b. Time dependence
The time-dependent and the steady-state Sverdrup streamfunction anomalies averaged in the band between 5° and 7°N compare quite well with each other, both in amplitude and phase (Fig. 10), suggesting that the steady state assumption is a valid approximation of the 5-yr running mean circulation at the low latitudes. The simultaneous correlation coefficient between the time-dependent and the steady-state Sverdrup streamfunction anomalies averaged in the area of 5°–7°N, 128°–138°E reaches 0.87, significantly above the 95% level.
Time–longitude plot of the (a) time-dependent and (b) steady-state Sverdrup transport anomalies (Sv) averaged between 5° and 7°N across the Pacific Ocean. The reference period is 1955–2016. The contour interval is 2 Sv. (c) Comparison of time series of time-dependent (solid) and steady-state (dashed) Sverdrup transport anomalies averaged in the area 5°–7°N, 128°–138°E. Their simultaneous correlation coefficient reaches r = 0.87.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Time–longitude plot of the (a) time-dependent and (b) steady-state Sverdrup transport anomalies (Sv) averaged between 5° and 7°N across the Pacific Ocean. The reference period is 1955–2016. The contour interval is 2 Sv. (c) Comparison of time series of time-dependent (solid) and steady-state (dashed) Sverdrup transport anomalies averaged in the area 5°–7°N, 128°–138°E. Their simultaneous correlation coefficient reaches r = 0.87.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Time–longitude plot of the (a) time-dependent and (b) steady-state Sverdrup transport anomalies (Sv) averaged between 5° and 7°N across the Pacific Ocean. The reference period is 1955–2016. The contour interval is 2 Sv. (c) Comparison of time series of time-dependent (solid) and steady-state (dashed) Sverdrup transport anomalies averaged in the area 5°–7°N, 128°–138°E. Their simultaneous correlation coefficient reaches r = 0.87.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
c. Equatorial wave dynamics behind the PDO predictability
The 5-yr running mean data used in this study have essentially excluded the ENSO signals oscillating at 2–7-yr periods. Neither the total meridional volume transport anomalies nor the non-Sverdrup transport anomalies have shown significant lead–lag correlations with the pentadal Niño-3.4 index (Fig. 11), suggesting no contribution from the Niño-3.4 predictability to the PDO predictability identified in this study. Although significant 1-yr lead correlations between MGTA or EKTA and the Niño-3.4 index do exist, with the maximum correlation coefficients reaching 0.62 and −0.62, respectively, they basically cancel each other in the total meridional transport anomalies.
(a) Comparisons of the anomalies of the meridional geostrophic transport (MGTA; blue) and the meridional Ekman transport (EKTA; red) averaged in the area 5°–7°N, 128°–138°E. The shaded bars are the Niño-3.4 index. (b) Lagged correlations between MGTA (blue) and EKTA (red) and the Niño-3.4 index, respectively. Trends are removed before the correlation calculations. Negative lags indicate Niño-3.4 lagging. (c),(d) As in (a),(b), but for the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red). The thin black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Comparisons of the anomalies of the meridional geostrophic transport (MGTA; blue) and the meridional Ekman transport (EKTA; red) averaged in the area 5°–7°N, 128°–138°E. The shaded bars are the Niño-3.4 index. (b) Lagged correlations between MGTA (blue) and EKTA (red) and the Niño-3.4 index, respectively. Trends are removed before the correlation calculations. Negative lags indicate Niño-3.4 lagging. (c),(d) As in (a),(b), but for the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red). The thin black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Comparisons of the anomalies of the meridional geostrophic transport (MGTA; blue) and the meridional Ekman transport (EKTA; red) averaged in the area 5°–7°N, 128°–138°E. The shaded bars are the Niño-3.4 index. (b) Lagged correlations between MGTA (blue) and EKTA (red) and the Niño-3.4 index, respectively. Trends are removed before the correlation calculations. Negative lags indicate Niño-3.4 lagging. (c),(d) As in (a),(b), but for the total meridional transport (MGTA plus EKTA; blue), the non-Sverdrup transport (total minus Sverdrup; black), and the total meridional heat transport (MGHTA plus EKHTA; red). The thin black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
However, the tropical Pacific pentadal SSTA, which are related to the thermocline depth, and hence the heat content, of the equatorial Pacific Ocean (McGregor et al. 2004), are significantly correlated with PDO (Fig. 12a). The MGTA dominates the EKTA in determining the thermocline variability of the equatorial Pacific Ocean on decadal time scales (Fig. 3b). The significant 1-yr lead correlation between MGTA and the equatorial Pacific SSTA averaged between 12°S and 12°N is consistent with the dynamical connection between MGTA and the PDO index on decadal time scales (Fig. 12b), which have been identified by our study to be a precursor of the PDO variability.
(a) Time series of SSTA (red) in the equatorial Pacific Ocean (12°N–12°S), MGTA (black; averaged over 5°–7°N, 128°–138°E), and the PDO index (shaded bar). All time series are 5-yr running means with trends removed. (b) Lag correlations between SSTA and PDO (red), and between SSTA and MGTA (black). Negative correlations represent SSTA lagging.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Time series of SSTA (red) in the equatorial Pacific Ocean (12°N–12°S), MGTA (black; averaged over 5°–7°N, 128°–138°E), and the PDO index (shaded bar). All time series are 5-yr running means with trends removed. (b) Lag correlations between SSTA and PDO (red), and between SSTA and MGTA (black). Negative correlations represent SSTA lagging.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a) Time series of SSTA (red) in the equatorial Pacific Ocean (12°N–12°S), MGTA (black; averaged over 5°–7°N, 128°–138°E), and the PDO index (shaded bar). All time series are 5-yr running means with trends removed. (b) Lag correlations between SSTA and PDO (red), and between SSTA and MGTA (black). Negative correlations represent SSTA lagging.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
The facts that the MGTA are significantly correlated with PDO index and insignificantly correlated with the pentadal Niño-3.4 index suggest different dynamics of the equatorial and off-equatorial Pacific Ocean to the discharge–recharge forcing on interannual and decadal time scales, respectively. It is known that the PDO SSTA have a broader meridional distribution than the ENSO anomalies in the eastern tropical Pacific Ocean, the dynamics of which have not been studied in the published literature so far. We speculate that this difference in the SSTA distributions can be explained using the equatorial wave dynamics of the Pacific Ocean.
At the interannual time scale, westerly wind bursts (WWBs) during the developing stage of El Niño force downwelling Kelvin waves to propagate eastward and depress the ocean thermocline in the eastern equatorial Pacific. The short periods of the forcing and the subsequent negative feedback suggest that the SSTA signals are concentrated on the equator, hence the interannual Niño-3.4 index. In contrast, on the decadal time scales, WWBs force downwelling Kevin waves persistently, which are reflected into equatorial Rossby waves at the eastern boundary to propagate westward and redistribute the heat content into the off-equatorial area. In the meantime, the upwelling Rossby waves forced by the same persistent WWBs are reflected at the western boundary and reduce the amplitudes of the wind-forced equatorial Kelvin waves. Therefore, the pentadal Niño-3.4 index has weak PDO signals. The detailed dynamics of the thermocline–SSTA–PDO connections are still an active area of research, the study of which is beyond the scope of this investigation.
6. Conclusions
Based on the GOHSCD pentadal hydrographic data from 1957 to 2014, we estimate the meridional geostrophic volume transport in the interior tropical gyre circulation of the North Pacific Ocean. The decadal to interdecadal variations of the observed meridional geostrophic transports in the tropical gyre of the Pacific Ocean (centered along 5°–7°N) are found to precede PDO at the lead time of 1–3 years above the 95% significance level. Enhanced (reduced) meridional transport in the tropical gyre precedes a positive (negative) PDO phase, suggesting the equatorial discharge–recharge mechanism behind the identified predictability.
The result in this paper suggests that the decadal variability of the Pacific is predictable if the meridional transport of the tropical gyre is used as a precursor. The predictability originates primarily from the non-Sverdrup meridional transports, which are likely driven by ocean nonlinearity. The Sverdrup transport is found to have no lead correlations with PDO, suggesting that the PDO is not driven by the wind-forced circulation in the tropical Pacific gyres. To our knowledge, the non-Sverdrup meridional transport is the first oceanic predictor identified for PDO prediction in the published literature, suggesting that the PDO might be driven by nonlinear processes in the tropical North Pacific Ocean.
It is well known that modern climate system models contain essentially no skill in predicting the PDO. The meridional transports of the CMIP5 coupled simulations and of the OFES ocean simulation are diagnosed in this study, showing poor simulation of the decadal variations of the non-Sverdrup transport anomalies. The deficiencies of these oceanic and climate system models in predicting the decadal variability of the Pacific Ocean lie in the fact that they are dominated by the linear Sverdrup dynamics, which bear no significant predictability of the PDO.
The significant lead correlations between the meridional geostrophic transport and the PDO indices on decadal time scales identified in this study suggest an important dynamical connection between the tropical Pacific Ocean and PDO through the discharge and recharge of the tropical ocean heat content. The analyses also suggest that the model deficiencies in simulating and predicting the Pacific decadal variability are primarily due to the inability to simulate the nonlinear processes faithfully in the tropical ocean. More studies on the detailed physical processes controlling PDO predictability will follow.
Acknowledgments
We acknowledge the CMIP5 project and the U.S. NODC for sharing the global ocean heat and salt content dataset. The work is supported by the NSFC (41376032, 41421005, and 41720104008), CAS (XDA11010205), QMSNL (2016ASKJ12 and 2016ASKJ04), SOA (GASI-03-01-01-05), and the Shandong Provincial projects (2014GJJS0101 and U1606402). WKD is also supported under NSF Grants 1434780 and 1537304.
APPENDIX
Mean Meridional Ekman Transport in North Pacific Ocean
The integrated mean meridional Ekman transport based on NCEP–NCAR wind data during 1955-2016 shows very strong northward transport in the low-latitude tropical North Pacific Ocean with the maximum amplitude exceeding 25 Sv (Fig. A1), which is due to the strong trade winds and the small value of the Coriolis parameter there.
Integrated mean meridional Ekman transport (Sv) based on NCEP–NCAR wind data in reference to the climatology during 1955–2016. Color contour interval is 5 Sv, and the thick white lines represent the zero contour.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Integrated mean meridional Ekman transport (Sv) based on NCEP–NCAR wind data in reference to the climatology during 1955–2016. Color contour interval is 5 Sv, and the thick white lines represent the zero contour.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Integrated mean meridional Ekman transport (Sv) based on NCEP–NCAR wind data in reference to the climatology during 1955–2016. Color contour interval is 5 Sv, and the thick white lines represent the zero contour.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
There exists a broad area from 6° to 12°N of 1–3-yr lead correlations between the meridional transport anomalies or the meridional heat transport anomalies and PDO index in the low latitude tropical North Pacific Ocean (Fig. A2). The stronger correlations in the lower latitudes suggest that the identified PDO predictability is a robust character in the low-latitude tropical North Pacific Ocean circulation.
(a)–(e) Comparisons of the anomalies of the MGTA (blue), the MHTA (red) averaged in the area 128°–138°E from 6° to 12°N. The shaded bars are the PDO index. (f)–(j) Lag correlations between MGTA (blue), MHTA (red), and the PDO index respectively. Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a)–(e) Comparisons of the anomalies of the MGTA (blue), the MHTA (red) averaged in the area 128°–138°E from 6° to 12°N. The shaded bars are the PDO index. (f)–(j) Lag correlations between MGTA (blue), MHTA (red), and the PDO index respectively. Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
(a)–(e) Comparisons of the anomalies of the MGTA (blue), the MHTA (red) averaged in the area 128°–138°E from 6° to 12°N. The shaded bars are the PDO index. (f)–(j) Lag correlations between MGTA (blue), MHTA (red), and the PDO index respectively. Trends are removed before the correlation calculations. Negative lags indicate PDO lagging. The black dashed lines mark the 95% confidence levels.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
The first EOF modes show similar spatial patterns in the simulated SSTA by OFES (Fig. A3a), GFDL (Fig. A3b), and CSIRO (Fig. A3c), with negative SSTAs in the extratropical central North Pacific Ocean, and anomalies of opposite signs in the northeastern Pacific, which are reminiscent of PDO SSTA spatial pattern.
Spatial patterns of the first EOF of SSTA in the (a) OFES, (b) GFDL-CM3, and (c) CSIRO Mk3.6.0.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Spatial patterns of the first EOF of SSTA in the (a) OFES, (b) GFDL-CM3, and (c) CSIRO Mk3.6.0.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
Spatial patterns of the first EOF of SSTA in the (a) OFES, (b) GFDL-CM3, and (c) CSIRO Mk3.6.0.
Citation: Journal of Climate 31, 15; 10.1175/JCLI-D-17-0639.1
REFERENCES
Bond, N. A., J. E. Overland, M. Spillane, and P. Stabeno, 2003: Recent shifts in the state of the North Pacific. Geophys. Res. Lett., 30, 2183, https://doi.org/10.1029/2003GL018597.
Boyer, T. P., and Coauthors, 2009: World Ocean Database 2009. S. Levitus, Ed., NOAA Atlas NESDIS 66, 216 pp.
Bretherton, C. S., M. Widmann, V. P. Dymnikov, J. M. Wallace, and I. Bladé, 1999: The effective number of spatial degrees of freedom of a time-varying field. J. Climate, 12, 1990–2009, https://doi.org/10.1175/1520-0442(1999)012<1990:TENOSD>2.0.CO;2.
Chavez, F. P., J. Ryan, S. E. Lluch-Cota, and M. Ñiquen, 2003: From anchovies to sardines and back: Multidecadal change in the Pacific Ocean. Science, 299, 217–221, https://doi.org/10.1126/science.1075880.
Chu, P. C., 1995: P-vector method for determining absolute velocity from hydrographic data. Mar. Technol. Soc. J., 29, 3–14.
Chu, P. C., 2000: P-vector spiral and determination of absolute velocities. J. Oceanogr., 56, 591–599, https://doi.org/10.1023/A:1011161313505.
Furtado, J. C., E. Di Lorenzo, N. Schneider, and N. A. Bond, 2011: North Pacific decadal variability and climate change in the IPCC AR4 models. J. Climate, 24, 3049–3067, https://doi.org/10.1175/2010JCLI3584.1.
Garratt, J. R., 1977: Review of drag coefficients over ocean and continents. Mon. Wea. Rev., 105, 915–929, https://doi.org/10.1175/1520-0493(1977)105<0915:RODCOO>2.0.CO;2.
Hautala, S. L., D. H. Roemmich, and W. J. Schmitz Jr., 1994: Is the North Pacific in Sverdrup balance along 24°N? J. Geophys. Res., 99, 16 041–16 052, https://doi.org/10.1029/94JC01084.
Hobbs, W. R., and J. K. Willis, 2012: Midlatitude North Atlantic heat transport: A time series based on satellite and drifter data. J. Geophys. Res., 117, C01008, https://doi.org/10.1029/2011JC007039.
Hurlburt, H. E., A. J. Wallcraft, W. J. Schmitz Jr., P. J. Hogan, and E. J. Metzger, 1996: Dynamics of the Kuroshio/Oyashio current system using eddy-resolving models of the North Pacific Ocean. J. Geophys. Res., 101, 941–976, https://doi.org/10.1029/95JC01674.
Hsin, Y.-C., 2016: Trends of the pathways and intensities of surface equatorial current system in the North Pacific Ocean. J. Climate, 29, 6693–6710, https://doi.org/10.1175/JCLI-D-15-0850.1.
Kawai, H., 1972: Hydrography of the Kuroshio Extension. Kuroshio: Physical Aspects of the Japan Current, H. Stommel and K. Yoshida, Eds., University of Washington Press, 235–352.
Kessler, W. S., G. C. Johnson, and D. W. Moore, 2003: Sverdrup and nonlinear dynamics of the Pacific equatorial currents. J. Phys. Oceanogr., 33, 994–1008, https://doi.org/10.1175/1520-0485(2003)033<0994:SANDOT>2.0.CO;2.
Kuball, C. D., 2007: Diagnosis of the Pacific decadal oscillation in global climate models. B.S. thesis, Dept. of Geological and Atmospheric Sciences, Iowa State University, 16 pp.
Latif, M., and T. P. Barnett, 1996: Decadal climate variability over the North Pacific and North America: Dynamics and predictability. J. Climate, 9, 2407–2423, https://doi.org/10.1175/1520-0442(1996)009<2407:DCVOTN>2.0.CO;2.
Levitus, S., T. Boyer, M. Conkright, D. Johnson, T. O’Brien, J. Antonov, C. Stephens, and R. Gelfeld, 1998: Introduction. Vol. 1, World Ocean Database 1998, NOAA Atlas NESDIS 18, 346 pp.
Levitus, S., and Coauthors, 2012: World ocean heat content and thermosteric sea level change (0–2000 m), 1955–2010. Geophys. Res. Lett., 39, L10603, https://doi.org/10.1029/2012GL051106.
Liu, Z., 2012: Dynamics of interdecadal climate variability: A historical perspective. J. Climate, 25, 1963–1995, https://doi.org/10.1175/2011JCLI3980.1.
Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111, 46–59, https://doi.org/10.1175/1520-0493(1983)111<0046:SFSAID>2.0.CO;2.
Locarnini, R. A., A. V. Mishonov, J. I. Antonov, T. P. Boyer, H. E. Garcia, O. K. Baranova, M. M. Zweng, and D. R. Johnson, 2010: Temperature. Vol. 1, World Ocean Atlas 2009, NOAA Atlas NESDIS 68, 184 pp.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, 1069–1079, https://doi.org/10.1175/1520-0477(1997)078<1069:APICOW>2.0.CO;2.
Masumoto, Y., and Coauthors, 2004: A fifty-year eddy-resolving simulation of the World Ocean preliminary outcomes of OFES (OGCM for the Earth Simulator). J. Earth Simul., 1, 35–56.
McGregor, S., N. J. Holbrook, and S. B. Power, 2004: On the dynamics of interdecadal thermocline depth and sea surface temperature variability in the low to mid-latitude Pacific Ocean. Geophys. Res. Lett., 31, L24201, https://doi.org/10.1029/2004GL021241.
Meyers, G., 1980: Do Sverdrup transports account for the Pacific North Equatorial Countercurrent? J. Geophys. Res., 85, 1073–1075, https://doi.org/10.1029/JC085iC02p01073.
Minobe, S., 2000: Spatio-temporal structure of pentadecadal variability over the North Pacific. Prog. Oceanogr., 47, 381–408, https://doi.org/10.1016/S0079-6611(00)00042-2.
Nakamura, H., G. Lin, and T. Yamagata, 1997: Decadal climate variability in the North Pacific during the recent decades. Bull. Amer. Meteor. Soc., 78, 2215–2225, https://doi.org/10.1175/1520-0477(1997)078<2215:DCVITN>2.0.CO;2.
Newman, M., G. P. Compo, and M. A. Alexander, 2003: ENSO-forced variability of the Pacific decadal oscillation. J. Climate, 16, 3853–3857, https://doi.org/10.1175/1520-0442(2003)016<3853:EVOTPD>2.0.CO;2.
Newman, M., and Coauthors, 2016: The Pacific decadal oscillation, revisited. J. Climate, 29, 4399–4427, https://doi.org/10.1175/JCLI-D-15-0508.1.
Qiu, B., 2001: Kuroshio and Oyashio Currents. Encyclopedia of Ocean Sciences, Academic Press, 1413–1425.
Qiu, B., and R. Lukas, 1996: Seasonal and interannual variability of the North Equatorial Current, the Mindanao Current and the Kuroshio along the Pacific western boundary. J. Geophys. Res., 101, 12 315–12 330, https://doi.org/10.1029/95JC03204.
Qiu, B., N. Schneider, and S. Chen, 2007: Coupled decadal variability in the North Pacific: An observationally constrained idealized model. J. Climate, 20, 3602–3620, https://doi.org/10.1175/JCLI4190.1.
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.
Schneider, N., and B. D. Cornuelle, 2005: The forcing of the Pacific decadal oscillation. J. Climate, 18, 4355–4373, https://doi.org/10.1175/JCLI3527.1.
Sverdrup, H. U., 1947: Wind-driven currents in a baroclinic ocean; with application to the equatorial current of the eastern Pacific. Proc. Natl. Acad. Sci. USA, 33, 318–326, https://doi.org/10.1073/pnas.33.11.318.
Tanimoto, Y., K. Hanawa, Y. Toba, and N. Iwasaka, 1993: Characteristic variation of sea surface temperature with multiple time scales in the North Pacific. J. Climate, 6, 1153–1160, https://doi.org/10.1175/1520-0442(1993)006<1153:CVOSST>2.0.CO;2.
Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn., 9, 303–319, https://doi.org/10.1007/BF00204745.
Vimont, D. J., 2005: The contribution of the interannual ENSO cycle to the spatial pattern of decadal ENSO-like variability. J. Climate, 18, 2080–2092, https://doi.org/10.1175/JCLI3365.1.
Yang, L., and D. Yuan, 2016: Absolute geostrophic currents in global tropical oceans. Chin. J. Oceanol. Limnol., 34, 1383–1393, https://doi.org/10.1007/s00343-016-5092-3.
Yuan, D., Z. Zhang, P. C. Chu, and W. K. Dewar, 2014: Geostrophic circulation in the tropical North Pacific Ocean based on Argo profiles. J. Phys. Oceanogr., 44, 558–575, https://doi.org/10.1175/JPO-D-12-0230.1.
Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10, 1004–1020, https://doi.org/10.1175/1520-0442(1997)010<1004:ELIV>2.0.CO;2.
Zhang, Z., D. Yuan, and P. C. Chu, 2013: Geostrophic meridional transport in tropical northwest Pacific based on Argo profiles. Chin. J. Oceanol. Limnol., 31, 656–664, https://doi.org/10.1007/s00343-013-2169-0.
