Abstract

The impact of climate change on the Pacific decadal oscillation (PDO) is studied using a fully coupled climate model. The model results show that the PDO has a similar spatial pattern in altered climates, but its amplitude and time scale of variability change in response to global warming or cooling. In response to global warming the PDO amplitude is significantly reduced, with a maximum decrease over the Kuroshio–Oyashio Extension (KOE) region. This reduction appears to be associated with a weakened meridional temperature gradient in the KOE region. In addition, reduced variability of North Pacific wind stress, partially due to reduced air–sea feedback, also helps to weaken the PDO amplitude by reducing the meridional displacements of the subtropical and subpolar gyre boundaries. In contrast, the PDO amplitude increases in response to global cooling.

In the control simulations the model PDO has an approximately bidecadal peak. In a warmer climate the PDO time scale becomes shorter, changing from ~20 to ~12 yr. In a colder climate the time scale of the PDO increases to ~34 yr. Physically, global warming (cooling) enhances (weakens) ocean stratification. The increased (decreased) ocean stratification acts to increase (reduce) the phase speed of internal Rossby waves, thereby altering the time scale of the simulated PDO.

1. Introduction

The observed 1976/77 climate shift over the North Pacific Ocean featured a decadal-scale transition from one pattern of sea surface temperature (SST) anomalies to a comparable pattern of opposite sign (Mantua et al. 1997). Many studies have examined the potential mechanisms influencing this decadal variability (e.g., Deser and Blackmon 1995; Schneider et al. 1999; Seager et al. 2001; Wu et al. 2005) and its potential climate impacts (e.g., Mantua and Hare 2002; Cayan et al. 2001). Observations show this decadal SST variability is different from El Niño–Southern Oscillation (ENSO) and is focused on the extratropical North Pacific with largest decadal SST variability along the Kuroshio–Oyashio Extension (KOE) region. This decadal variability was called the Pacific decadal oscillation (PDO; Mantua et al. 1997).

The PDO is commonly defined as the first empirical orthogonal function (EOF) of SST anomalies in the North Pacific Ocean north of 20°N. The positive phase of the PDO is characterized by a horseshoe-like pattern, with negative SST anomalies in the western and central North Pacific surrounded by positive SST anomalies along the North American coast. Note that the PDO definition is different from the interdecadal Pacific oscillation (IPO), which focuses on the entire Pacific (e.g., Power et al. 1999), although the two are very closely related with similar pattern and time scale (Han et al. 2014). Observations and ocean–atmosphere fully coupled models consistently show decadal peaks in the PDO time series. However, the mechanisms explaining these decadal fluctuations are quite different in different models.

In early work, Latif and Barnett (1994) suggested that the PDO is attributed to an unstable ocean–atmosphere interaction between the subtropical ocean gyre circulation and the Aleutian low pressure system in the fully coupled ECHO-1. The positive feedback arises from strong midlatitude air–sea feedback, while the delayed negative feedback comes from westward-propagating Rossby waves in the subtropical region. Schneider et al. (2002), however, illustrated that the decadal time scale of the PDO results from the integration along Rossby waves trajectories of stochastic atmosphere forcing in ECHO-2. Kwon and Deser (2007) pointed out that the midlatitude ocean feedback to the atmosphere is weak but very necessary to the PDO decadal peak in the NCAR Community Climate System Model version 2 (CCSM2). The PDO decadal time scale is mainly provided by the westward Rossby wave propagation along the KOE region. Zhang and Delworth (2015) examined the PDO behavior in several versions of climate models from the Geophysical Fluid Dynamics Laboratory (GFDL). They concluded that the PDO in the GFDL models is mainly independent of ENSO, and the existence of a bidecadal spectral peak of the PDO depends on active air–sea coupling over the North Pacific. In addition, the bidecadal time scale is strongly influenced by the propagation speed of oceanic Rossby waves in the subtropical and subpolar gyres, as they provide a delayed feedback to the atmosphere.

Despite the diversity of PDO mechanisms in coupled model analyses, several recent studies agree on some key points (e.g., Kwon and Deser 2007; Zhong and Liu 2009; Zhang and Delworth 2015). First, the KOE SST anomaly is mainly generated by meridional displacement of the subtropical and subpolar gyre boundary, rather than a change in strength of the subtropical gyre. This is consistent with observations and ocean-only model results (e.g., Seager et al. 2001; Nakamura and Kazmin 2003). Second, the KOE positive (negative) SST anomalies can feed back to the overlying atmosphere, inducing an anomalous high (low) pressure that can further amplify initial SST anomalies via anomalous Ekman transport. This positive air–sea feedback over the North Pacific Ocean is also in agreement with observations (Frankignoul and Sennéchael 2007; Zhang et al. 2011). Third, the ocean Rossby wave is the most important contributor to the PDO decadal time scale selection, although the relative importance of propagation is different in different models.

The climatic and ecological influences of the PDO are also widely examined. Mantua et al. (1997) documented that salmon production decreases in the northwest United States and Alaska when the PDO shifts from the negative phase to the positive phase. Positive PDO phases are associated with a deficit of precipitation and positive air temperature anomalies in the northwestern and central United States, and an excess of precipitation and negative air temperature anomalies in the southwestern United States and northern Mexico (e.g., Mantua and Hare 2002; Zhang and Delworth 2015). Opposite conditions occur for the negative phase of the PDO. In this regard, a change in PDO behavior in response to anthropogenic climate change may have very serious climatic and ecological consequences, particularly in the downstream U.S. region. However, few studies have focused on this topic. d’Orgeville and Peltier (2009) demonstrated that the PDO may have a longer period in the twentieth-century run compared to the present day. But, they only used a very short time series (100–200 yr) to study the PDO response to global warming, which is not statistically significant. They also did not explain why the PDO period has such a frequency shift. On the other hand, Fang et al. (2014) argued that the PDO 50–70-yr period disappears in a warm climate and it seems the period moves to a more high-frequency band using a very low-resolution model. However, the frequency shift in their model is not clear.

The purpose of the current study is to examine the PDO response to changed climate. We attempt to address how the PDO responds to global warming in terms of its spatial pattern, amplitude and spectral characteristics. We also investigate a “global cooling scenario” to more broadly assess the response of the PDO to various types of altered radiative forcing.

2. Model description and experimental design

The coupled model used in this study is GFDL Coupled Model version 2.5, using the forecast-oriented low ocean resolution (FLOR) version (CM2.5_FLOR; Vecchi et al. 2014). The atmospheric model has a relatively high horizontal resolution of approximately 50 km × 50 km, with 24 levels in the vertical. The ocean and sea ice components of this FLOR model are based on the low-resolution GFDL Coupled Model version 2.1 (CM2.1; Delworth et al. 2006). The horizontal resolution of the ocean model is 1° in the extratropics, with finer meridional grid spacing in the tropics (~⅓°). The ocean model has 50 levels in the vertical, with 22 evenly spaced levels over the top 220 m. The coupled model runs for 920 yr with atmospheric constituents and external forcing held constant at 1860. The PDO spatial structure and spectrum in this long-term control simulation is quite similar to those found in observational analyses (Zhang and Delworth 2015), which gives us confidence in examining the PDO response to global warming.

We conduct a double CO2 experiment to simulate a global warming scenario, in which atmosphere CO2 concentration is suddenly doubled from a preindustrial value of 286 to 572 ppm. The simulation starts from an arbitrary point in the control simulation and continues for 500 yr. This double CO2 experiment is called 2CO2. The difference between the 2CO2 experiment and the control run is interpreted as the response to global warming. We also perform a half CO2 experiment, in which the CO2 concentration is suddenly halved from the original 286 ppm to 143 ppm and held constant thereafter; this experiment is named 0.5CO2.

3. Simulated global warming and cooling scenarios

Figure 1 shows the SST responses in the 2CO2 and 0.5CO2 experiments. The global mean SST increases rapidly within the first three decades (Fig. 1a), with an increase up to 1.4°C at the end of the 30th year. After about 30 yr, the SST continues increasing but at a reduced rate. The SST response in the 0.5CO2 experiment is almost opposite to the 2CO2 run, with a rapid decrease in the first several decades and a subsequent slower decrease thereafter (Fig. 1a).

Fig. 1.

(a) Time series of global mean sea surface temperature (SST) response in the 2CO2 and 0.5CO2 experiments. Also shown is the spatial pattern of the time averaged (year 51–500) SST and sea level pressure (SLP) responses in (b) 2CO2 and (c) 0.5CO2. Units are °C for SST and hPa for SLP. The contour interval is 0.5 hPa for SLP. The solid black (dashed gray) lines denote the positive (negative) SLP anomalies.

Fig. 1.

(a) Time series of global mean sea surface temperature (SST) response in the 2CO2 and 0.5CO2 experiments. Also shown is the spatial pattern of the time averaged (year 51–500) SST and sea level pressure (SLP) responses in (b) 2CO2 and (c) 0.5CO2. Units are °C for SST and hPa for SLP. The contour interval is 0.5 hPa for SLP. The solid black (dashed gray) lines denote the positive (negative) SLP anomalies.

We now turn our attention to the SST spatial structure. The SST response shows broad warming over the global ocean in the 2CO2 experiment except over the deep convection regions (North Atlantic and Southern Ocean) (Fig. 1b). A close examination finds that the North Atlantic is characterized by a dipole SST structure, with a cold anomaly in the subpolar region and a strong warm anomaly along the Gulf Stream path. These SST anomalies are accompanied by a strengthened subpolar gyre and northward shift of the Gulf Stream, which is a fingerprint of a weakened Atlantic meridional overturning circulation (AMOC) (Zhang 2008; Zhang and Wang 2013). The AMOC weakening in the 2CO2 run is attributed to the anomalous radiative heating and Arctic freshwater input as a result of the greenhouse gas effect (e.g., Cheng et al. 2013; Zhang et al. 2014; Zhang and Wu 2012). Similarly, the weak SST cooling over the Southern Ocean is due to the weakened Antarctic Bottom Water (AABW) formation because of the Antarctic sea ice melting under the global warming scenario (e.g., Ma and Wu 2011).

The North Pacific SST, as expected, shows significant and uniform warm responses in the 2CO2 run, with a primary maximum over the polar Bering Strait along with a secondary maximum over the midlatitude KOE region. The former is a classic polar amplification phenomenon, which can be explained by the sea ice decrease and subsequent positive ice-albedo feedback (e.g., Holland and Bitz 2003). The latter suggests that the ocean circulation experiences some changes under global warming scenario. Figure 2a exhibits the zonally averaged (140°–180°E) zonal current response to global warming, which shows a strong positive (weak negative) anomaly north (south) of 32°N. This implies that the Kuroshio slightly expands northward in a warm climate [also seen in Wu et al. (2012)]. We also calculate the Kuroshio volume transport across the 140°E section and it shows a 4-Sv (1 Sv ≡ 106 m3 s−1) increase in a warmer climate as compared to the fully coupled control run. The accelerated Kuroshio here is consistent with the atmosphere circulation response (Fig. 1b). The North Pacific subtropical high strengthens in 2CO2 run, which leads to a spinup of the subtropical gyre and the Kuroshio, an enhanced northward warm advection, and therefore a positive SST anomaly over the KOE region that is larger than in surrounding areas of the North Pacific. The SST spatial pattern response in the 0.5CO2 experiment is almost anticorrelated with that in the 2CO2 run (Fig. 1c).

Fig. 2.

Zonally averaged (140°–180°E) zonal current response in (a) 2CO2 and (b) 0.5CO2. The black contours in (a) and (b) are the long-term mean zonal current in the fully coupled control run, while the shading denotes the zonal current difference between the changed climate (2CO2 or 0.5CO2) and the control run. (c),(d) As in (a),(b), but for the zonally averaged (0°E–0°W) atmosphere temperature. Units are cm s−1 for zonal current and K for atmosphere temperature.

Fig. 2.

Zonally averaged (140°–180°E) zonal current response in (a) 2CO2 and (b) 0.5CO2. The black contours in (a) and (b) are the long-term mean zonal current in the fully coupled control run, while the shading denotes the zonal current difference between the changed climate (2CO2 or 0.5CO2) and the control run. (c),(d) As in (a),(b), but for the zonally averaged (0°E–0°W) atmosphere temperature. Units are cm s−1 for zonal current and K for atmosphere temperature.

The sea level pressure (SLP) response to global warming indicates that the midlatitude westerly wind enhances and shifts poleward in both hemispheres in a warm climate (Fig. 1b). To elucidate how global warming can drive such a westerly wind shift, we exhibit the vertical structure of zonal mean air temperature response (Fig. 2c). The amplified tropical upper tropospheric warming is seen clearly in Fig. 2c, which is largely due to increased latent heat release through enhanced moist convection (e.g., Frierson 2006; Lu et al. 2008). In the middle and high latitudes, warming anomalies are exhibited in the low and middle troposphere, whereas cooling anomalies appear above 350 hPa. These temperature features indicate an increased static stability in the midlatitudes as well as an increased equator-to-pole temperature gradient in the upper troposphere and lower stratosphere, which could push the westerly jet poleward and strengthen westerly winds. The opposite conditions occur in a cold climate (Figs. 1c and 2d).

4. Simulated amplitude response of the PDO in 2CO2 and 0.5CO2

a. PDO spatial pattern and amplitude responses

To understand the PDO amplitude response in warm and cold climates, we first present the SST standard deviation (STD) over the North Pacific (Fig. 3). Here, we use data from years 51 to 500, and remove the linear trend before calculating the STD. The left panels in Fig. 3 are the unfiltered annually averaged SST standard deviation in the control run, 2CO2, 0.5CO2, and their differences, while the right panels denote the STD of 7-yr low pass filtered SST. In the fully coupled control run, the SST amplitude has its primary maximum center in the KOE region along with a secondary maximum over the North American coast (Fig. 3a). It is interesting to see that the SST amplitude decreases (increases) in a warm (cold) climate (Figs. 3c,e). These amplitude characteristics and changes are still retained on decadal time scales, albeit with a smaller magnitude (Figs. 3b,d,f). The STD difference fields further show that the largest decrease (increase) of SST amplitude occurs in the KOE region (Figs. 3g,i), which is significant at 95% level using an F test. Notable changes are also found over the central eastern North Pacific, the Bering Sea, and the Aleutian Trench. A comparison between the total STD difference (Figs. 3g,i) and the decadal STD difference (Figs. 3h,j) reveals that these SST amplitude changes are mainly contributed from the decadal variability, while the interannual variability plays a negligible or even a damping role.

Fig. 3.

Standard deviation (STD) of (left) unfiltered and (right) 7-yr low-pass filtered annually averaged SST in the (a),(b) fully coupled control run, (c),(d) 2CO2, and (e),(f) 0.5CO2 experiments. (g),(h) STD difference between the 2CO2 run and fully coupled control simulation. (i),(j) As in (g),(h), but for the difference between the 0.5CO2 run and control experiment. Unit is °C. The gray points overlapped on shading in (g)–(j) mean the STD difference is significant at 95% confidence level using an F test. Here, we use the data from years 51 to 500 and the linear trend is removed before calculating STD.

Fig. 3.

Standard deviation (STD) of (left) unfiltered and (right) 7-yr low-pass filtered annually averaged SST in the (a),(b) fully coupled control run, (c),(d) 2CO2, and (e),(f) 0.5CO2 experiments. (g),(h) STD difference between the 2CO2 run and fully coupled control simulation. (i),(j) As in (g),(h), but for the difference between the 0.5CO2 run and control experiment. Unit is °C. The gray points overlapped on shading in (g)–(j) mean the STD difference is significant at 95% confidence level using an F test. Here, we use the data from years 51 to 500 and the linear trend is removed before calculating STD.

The KOE SST amplitude has a strong seasonality in both control run and sensitivity experiments (Fig. 4), with a maximum in the summer and a minimum in the winter. This is because the mixed layer depth shallows (deepens) during the summer (winter), which favors generating a larger (smaller) than normal SST amplitude with the same atmosphere forcing. In a warm (cold) climate, the SST amplitude decreases (increases) throughout the whole year. These amplitude changes are more significant on decadal time scales, which is consistent with the annually averaged analysis (Fig. 3). These results suggest that the PDO, a dominant decadal SST mode over the North Pacific, should have significant changes in amplitude in future/past climate.

Fig. 4.

Seasonal standard deviation (STD; °C) of unfiltered (solid lines) and 7-yr low pass filtered (dashed lines) SST averaged over the Kuroshio–Oyashio Extension (KOE) region (35°–45°N, 140°–180°E) in the fully coupled control run and 2CO2 and 0.5CO2 experiments.

Fig. 4.

Seasonal standard deviation (STD; °C) of unfiltered (solid lines) and 7-yr low pass filtered (dashed lines) SST averaged over the Kuroshio–Oyashio Extension (KOE) region (35°–45°N, 140°–180°E) in the fully coupled control run and 2CO2 and 0.5CO2 experiments.

We show in Fig. 5 the PDO spatial pattern in different scenarios, which is calculated as the first EOF of SST anomalies over the North Pacific north of 20°N. Note that the principal component is normalized by its standard deviation, so the magnitude of the spatial pattern represents the PDO amplitude. The PDO horseshoe-like patterns in both the warm and cold climates are very similar to the control run (Figs. 5a,c,e). In agreement with the STD analysis (Figs. 3 and 4), the PDO amplitude becomes much weaker (stronger) in the 2CO2 (0.5CO2) run compared to the fully coupled control experiment (Fig. 5a vs Figs. 5c,e). The maximum decrease (increase) appears in the KOE region (Figs. 5g,i), and these changes are significant at 95% level based on a t test. Zhang and Delworth (2015) pointed out that the PDO in CM2.5_FLOR model is generated by coupled air–sea interaction over the extratropical North Pacific. This implies that the SST variability associated with the PDO in a changed climate may imprint on the overlying atmosphere. We therefore plot the sea level pressure response associated with the PDO in different runs, as exhibited in the right panels of Fig. 5. As expected, the Aleutian low response is weaker (stronger) in the global warming (cooling) simulation (Figs. 5b,d,f,h,j). This further confirms that the PDO has smaller (larger) amplitude in a warm (cold) climate as compared to the present day.

Fig. 5.

Spatial patterns of the (left) Pacific decadal oscillation (PDO) and (right) corresponding SLP responses in the fully coupled (a),(b) control run, (c),(d) 2CO2, and (e),(f) 0.5CO2. The PDO spatial pattern is defined as the first EOF mode of SST anomalies over the North Pacific. Note that the associated principal component is normalized by its standard deviation and thus the magnitude of PDO spatial pattern can represent the PDO amplitude. The corresponding SLP response is derived by regressing the SLP anomaly upon the normalized principal component of the PDO. (g),(h) Difference of the PDO amplitude and associated SLP responses between 2CO2 and the fully coupled control run. (i),(j) As in (g),(h), but for the difference between 0.5CO2 and control run. Units are °C for SST and hPa for SLP. The gray points overlapped on shading in (g)–(j) mean the SST difference is significant at 95% confidence level based on a t test.

Fig. 5.

Spatial patterns of the (left) Pacific decadal oscillation (PDO) and (right) corresponding SLP responses in the fully coupled (a),(b) control run, (c),(d) 2CO2, and (e),(f) 0.5CO2. The PDO spatial pattern is defined as the first EOF mode of SST anomalies over the North Pacific. Note that the associated principal component is normalized by its standard deviation and thus the magnitude of PDO spatial pattern can represent the PDO amplitude. The corresponding SLP response is derived by regressing the SLP anomaly upon the normalized principal component of the PDO. (g),(h) Difference of the PDO amplitude and associated SLP responses between 2CO2 and the fully coupled control run. (i),(j) As in (g),(h), but for the difference between 0.5CO2 and control run. Units are °C for SST and hPa for SLP. The gray points overlapped on shading in (g)–(j) mean the SST difference is significant at 95% confidence level based on a t test.

b. Possible physical processes controlling the PDO amplitude response

First, we focus on the KOE region, which is the key center of the PDO and has the maximum decadal SST amplitude response in a changed climate. Zhang and Delworth (2015) demonstrated that the decadal SST anomaly over the KOE region in CM2.5_FLOR model is determined by the temperature advection by the anomalous meridional current as a result of the meridional shift of the subtropical and subpolar gyre boundary. As suggested by Thompson and Kwon (2010), the meridional displacement of the KOE in the coupled model can be assessed using , where d is the displacement of the sea surface height (SSH) contours, y is the meridional coordinate, and is SSH. The overbar denotes the time-averaged response, while the prime indicates deviation from the long-term mean. Accordingly, the SST amplitude over the KOE region is roughly estimated by the simple diagnostic relationship listed as follows:

 
formula

where T is SST. This equation means that the SST amplitude over the KOE region is given by the standard deviation of the boundary shift of the KOE multiplied by the mean SST meridional gradient.

The SST amplitude over the KOE region is reproduced very well by the above simple diagnostic relationship in all scenarios. The middle column in Fig. 6 shows the STD of SST over the western North Pacific as estimated by the diagnostic relationship, while the left column shows the STD of SST calculated directly from the model output. The zonal mean of the SST STD over the KOE region calculated using these two methods is very similar, shown in the right panels of Fig. 6. There are also some discrepancies outside the KOE region, since the dynamics controlling the SST anomaly over these regions cannot be represented by Eq. (1). These agreements confirm that there are two factors to determine the SST amplitude over the KOE region in both the present-day and changed climate; one is the mean meridional SST gradient and the other is the amplitude of the meridional displacement of the gyre boundary.

Fig. 6.

(left) Spatial patterns of SST standard deviation over the western North Pacific calculated by the SST output from model and (middle) simple diagnostic relationship shown in the main text in the fully coupled (a),(b) control run, (d),(e) 2CO2, and (g),(h) 0.5CO2. Zonal mean (140°–180°E) (right) SST standard deviation calculated from the SST output (red line) and simple diagnostic relationship (blue line) in (c) the control run, (f) 2CO2, and (i) 0.5CO2. Unit is °C.

Fig. 6.

(left) Spatial patterns of SST standard deviation over the western North Pacific calculated by the SST output from model and (middle) simple diagnostic relationship shown in the main text in the fully coupled (a),(b) control run, (d),(e) 2CO2, and (g),(h) 0.5CO2. Zonal mean (140°–180°E) (right) SST standard deviation calculated from the SST output (red line) and simple diagnostic relationship (blue line) in (c) the control run, (f) 2CO2, and (i) 0.5CO2. Unit is °C.

Figure 7 shows the zonal mean KOE SST amplitude difference between the global warming/cooling scenario and the present day. Again, the decreased (increased) SST amplitude in a warm (cold) climate is remarkably well reproduced by the diagnostic relationship in Eq. (1). We further decompose these responses into the contribution from changes in the mean meridional temperature gradient and the contribution from changes in the amplitude of the meridional displacement of the KOE. Both of these two factors contribute positively to the KOE SST amplitude response and their relative contributions are comparable.

Fig. 7.

(a) Difference of zonal mean SST standard deviation between the 2CO2 and fully coupled control run calculated from the SST output (red line) and simple diagnostic relationship (blue line). We further decompose the STD difference into the contribution from changes in the mean meridional temperature gradient (green line) and the contribution from changes in the amplitude of meridional shift of the subtropical and subpolar gyre boundary (black line). (b) As in (a), but for the STD difference between the 0.5CO2 and control run. Unit is °C.

Fig. 7.

(a) Difference of zonal mean SST standard deviation between the 2CO2 and fully coupled control run calculated from the SST output (red line) and simple diagnostic relationship (blue line). We further decompose the STD difference into the contribution from changes in the mean meridional temperature gradient (green line) and the contribution from changes in the amplitude of meridional shift of the subtropical and subpolar gyre boundary (black line). (b) As in (a), but for the STD difference between the 0.5CO2 and control run. Unit is °C.

The positive role of meridional temperature gradient in the SST amplitude response is more clearly seen in Fig. 8. In the fully coupled control run, the meridional SST gradient has its maximum center over the KOE region (Fig. 8a). In a warm climate, the SST meridional gradient decreases in the KOE region, mainly south of 40°N (Fig. 8b). This corresponds to the strong SST warming response in the midlatitude in a warm climate (Fig. 1b). The weakened SST meridional gradient also occurs in the eastern subtropics, southern Bering Sea, and Aleutian Trench (Fig. 8b). On the contrary, the SST meridional gradient increases in most regions of the North Pacific in a cold climate, including the southern KOE, central North Pacific, and Aleutian Trench. A close inspection finds that the regions with large SST gradient change coincide very well with the regions with large SST amplitude response (Figs. 8b,c vs Figs. 3g–j). Supposing that there is no change in meridional current in a warm (cold) climate, the decrease (increase) of meridional temperature gradient over the North Pacific could lead to weakened (enhanced) amplitude of the PDO.

Fig. 8.

(a),(d) Long-term mean (left) SST meridional gradient and (right) mixed layer depth (MLD) in the fully coupled control run, and responses in (b),(e) 2CO2 and (c),(f) 0.5CO2. Units are 10−5 °C m−1 for SST meridional gradient and m for MLD. Note that the SST meridional gradient is multiplied by −1. The gray points overlapped on shading in (b),(c),(e) and (f) mean the SST gradient or MLD differences are significant at 95% confidence level based on a t test.

Fig. 8.

(a),(d) Long-term mean (left) SST meridional gradient and (right) mixed layer depth (MLD) in the fully coupled control run, and responses in (b),(e) 2CO2 and (c),(f) 0.5CO2. Units are 10−5 °C m−1 for SST meridional gradient and m for MLD. Note that the SST meridional gradient is multiplied by −1. The gray points overlapped on shading in (b),(c),(e) and (f) mean the SST gradient or MLD differences are significant at 95% confidence level based on a t test.

The other positive contributor to changes in PDO-related SST variability is the meridional displacement of the KOE. We assess this as follows: for each longitude, we find the latitudinal position of the maximum zonal current over the western North Pacific. The zonal mean latitudinal position of maximum zonal current within the 140°–180°E band is taken as the position of the KOE. The mean KOE position in the fully coupled control run is at 34.5°N, while it is located at 35.8°N in a warm climate and 33.1°N in a cold climate. This northward (southward) expansion of the subtropical and subpolar gyre boundary in response to global warming (cooling) is consistent with the dipole zonal current structure shown in Figs. 2a and 2b. We also find the standard deviation of the KOE latitudinal position shift is 1.7° in the control run. In a warm (cold) climate, this amplitude decreases (increases) to 1.1° (3.12°). These results are in good agreement with the budget analysis in Fig. 7.

Zhang and Delworth (2015) have pointed out that the meridional shift of the subtropical and subpolar gyre boundary in the GFDL model is mainly determined by the interior wind stress curl over the North Pacific. Moreover, these anomalous wind stress curl variations are associated with extratropical SST feedback. Note that the interior wind stress can also be represented by SLP, which does exhibit decreased (increased) amplitude in a warm (cold) climate (Figs. 5h,j).

We now explore how the strength of the air–sea feedback is altered in a changed climate. Figure 9 displays the characteristics of extratropical North Pacific air–sea coupling in different scenarios. The ENSO signal is removed in this analysis using linear regression. For negative values along the x axis the geopotential height leads the SST, which mainly reflects the atmosphere forcing of the ocean. Positive values along the x axis mean that the SST leads the geopotential height, which mainly reflects the ocean feedback to the atmosphere. In the fully coupled control run (Figs. 9a,b), the positive geopotential height anomaly during the winter season tends to warm the midlatitude SST via reduced latent heat loss and anomalous northward Ekman transport. The SST warming could feed back to the overlying atmosphere, inducing a barotropic positive geopotential height anomaly that can further amplify the initial SST warming. Opposite conditions occur for the negative geopotential height. Note that the magnitude of atmosphere forcing to the ocean is much stronger than the ocean feedback. Thus, there is weak air–sea coupling over the extratropical North Pacific. In a warm (cold) climate, the atmosphere forcing to the ocean becomes stronger (weaker) as compared to the control run (Figs. 9c–f). This is because the ocean mixed layer depth shoals (deepens) in response to global warming (cooling) due to the strengthened (weakened) ocean stratification (Figs. 8d–f). The shallow (deep) mixed layer depth favors an increase (a decrease) of SST response with the same atmosphere forcing, and thus generates a strong (weak) coefficient of atmosphere forcing to the ocean. In contrast, the ocean feedback to the atmosphere (lead 4–6 in each panel) is much weaker (stronger) in the 2CO2 (0.5CO2) experiment than that in the control run (Figs. 9a–f). This weakened (strengthened) atmosphere response could generate a small (large) meridional shift of the subtropical and subpolar gyre boundary and therefore a weak (strong) SST anomaly over the KOE region after the slow adjustment of ocean Rossby wave. Thus, the altered nature of air–sea feedback as the climate warms or cools contributes to the changing amplitude of the PDO.

Fig. 9.

Seasonal lead–lag regression of the North Pacific area averaged (30°–60°N, 150°–150°W) geopotential height anomalies at (left) 850 and (right) 250 mb vs the SST anomalies averaged over the KOE region (35°–45°N, 140°–180°E) in the fully coupled (a),(b) control run, (c),(d) 2CO2, and (e),(f) 0.5CO2. The y axis indicates the calendar month taken for the KOE SST anomaly lead–lag regressed upon the geopotential height anomalies for a particular lag exhibited on the x axis. Unit is m °C−1. The ENSO signal is removed by using the linear regression. Shaded area means the regression coefficient exceeds 95% confidence level based on an F test.

Fig. 9.

Seasonal lead–lag regression of the North Pacific area averaged (30°–60°N, 150°–150°W) geopotential height anomalies at (left) 850 and (right) 250 mb vs the SST anomalies averaged over the KOE region (35°–45°N, 140°–180°E) in the fully coupled (a),(b) control run, (c),(d) 2CO2, and (e),(f) 0.5CO2. The y axis indicates the calendar month taken for the KOE SST anomaly lead–lag regressed upon the geopotential height anomalies for a particular lag exhibited on the x axis. Unit is m °C−1. The ENSO signal is removed by using the linear regression. Shaded area means the regression coefficient exceeds 95% confidence level based on an F test.

We note that the atmosphere circulation response to midlatitude SST anomaly is barotropic (Fig. 9). This is because the midlatitude has strong transient eddies and the associated transient eddy vorticity fluxes can reverse the linear, near-surface response to shallow heating (cooling), replacing a baroclinic downstream low (high) pressure anomaly with an equivalent barotropic high (low) (Kushnir et al. 2002). Therefore, it is reasonable to speculate that the magnitude of the atmospheric response to midlatitude SST anomaly strongly depends on the strength of transient eddy activity.

To test this hypothesis, we shown in Fig. 10 the 500-mb storm track responses in warm and cold climate. Here, the storm track is defined as the variance of 2–8-day bandpass filtered meridional wind. In a warm climate, the storm track shifts poleward, with an obvious weakening south of 45°N [Figs. 10a,c,e; this can also be seen in, e.g., Yin (2005)]. This storm track change could lead to a decreased transient eddy vorticity fluxes from the upper troposphere to the low troposphere south of 45°N and therefore a decreased SLP response to midlatitude SST. Figure 9 displays that the strongest midlatitude ocean feedback occurs when the late summer/early fall SST meets the winter atmosphere. Coincidently, the storm track has a maximum decrease south of 45°N during the winter season (Fig. 10c,e), which further reinforces the weakened atmosphere response. Opposite to the 2CO2 run, the transient eddies become more active in cold climates (0.5CO2), with a maximum increase in the winter season (Figs. 10b,d,f). This eventually leads to a stronger than normal atmosphere response to the SST anomaly over the KOE region (Figs. 9d,e).

Fig. 10.

(a),(b) Annually averaged and (c),(d) winter and (e),(f)summer averaged storm track responses in (left) 2CO2 and (right) 0.5CO2. The storm track here is defined as the variance of 2–8-day bandpass filtered meridional velocity at 500 mb and the unit is m2 s−2. The contours in each subplot denote the long-term mean storm track activity in each season in the fully coupled control simulation, while the shading denotes the difference between the changed climate (2CO2 or 0.5CO2) and the control run.

Fig. 10.

(a),(b) Annually averaged and (c),(d) winter and (e),(f)summer averaged storm track responses in (left) 2CO2 and (right) 0.5CO2. The storm track here is defined as the variance of 2–8-day bandpass filtered meridional velocity at 500 mb and the unit is m2 s−2. The contours in each subplot denote the long-term mean storm track activity in each season in the fully coupled control simulation, while the shading denotes the difference between the changed climate (2CO2 or 0.5CO2) and the control run.

5. Simulated period shift of the PDO in 2CO2 and 0.5CO2

Figure 11 shows the power spectrum of the PDO index in the control run, 2CO2, and 0.5CO2, respectively. The PDO index here is defined as the principal component of the first EOF mode of SST anomalies over the North Pacific north of 20°N. The PDO in the fully coupled control run has an approximately bidecadal (20 yr) peak (Fig. 11a), which has been analyzed in Zhang and Delworth (2015). In the 2CO2 simulation the spectral peaks are at shorter time scales (Fig. 11b). In the cold climate, however, the PDO spectral peaks occur at substantially longer time scales than in the control run, with significant peaks between 30 and 50 yr (Fig. 11c). As shown below there are multiple factors responsible for the change in time scale. A prominent factor, however, is the altered ocean stratification in the warming and cooling experiments. Stratification increases (decreases) in the warming experiment. This altered stratification leads to faster (slower) internal wave speeds in the warmer (colder) climate, thereby shortening (lengthening) the time scale of the PDO as influenced by the travel time of internal waves.

Fig. 11.

Power spectrum of the PDO time series in the fully coupled (a) control run, (b) 2CO2, and (c) 0.5CO2. The PDO time series is derived from the principal component of the first EOF mode of SST anomalies over the North Pacific. Here, the PDO spatial pattern is normalized by the standard deviation and therefore the associated principal component can represent the PDO amplitude.

Fig. 11.

Power spectrum of the PDO time series in the fully coupled (a) control run, (b) 2CO2, and (c) 0.5CO2. The PDO time series is derived from the principal component of the first EOF mode of SST anomalies over the North Pacific. Here, the PDO spatial pattern is normalized by the standard deviation and therefore the associated principal component can represent the PDO amplitude.

a. Subpolar pathway and its influence on PDO time scale

To better understand the change in the time scale of the PDO in response to global warming or cooling, we show the PDO evolution as characterized by regression coefficients at various lags in different scenarios (Figs. 12 and 13). In the fully coupled control run, the reversal of the PDO phase is associated with three pathways as proposed by Zhang and Delworth (2015). The first pathway primarily comes from the subpolar region. At a lag of 0, the PDO is in its mature positive phase with weak warming anomaly in the western subpolar region (Fig. 12a). This warm SST anomaly then grows and gradually extends southward into the midlatitude, leading to a decay of negative SST anomalies in the KOE region (Figs. 12a–d). Note that the warm SST anomaly in the western subpolar region is upwelled from the subsurface (Figs. 13a,b) and is generated by the westward downwelling Rossby wave adjustment induced by the anomalous negative wind stress curl 10 years prior during the mature negative phase of PDO (Fig. 12g) (Zhang and Delworth 2015). The southward advection of subpolar warm temperature anomaly by the Oyashio is mainly trapped in the upper 100 m and gradually strengthens during the propagation (Figs. 13a–e) due to the positive temperature–salinity convective feedback (Zhong and Liu 2009). In association with the southward propagation of warm subpolar SST, the positive wind stress curl during the mature PDO positive phase also triggers a westward propagated upwelling Rossby wave and eventually leads to a cold temperature anomaly in the western subpolar subsurface after about 10 years when the PDO transitions to its mature negative phase (Figs. 13f,g). This cold temperature during the PDO negative phase favors flipping the PDO back to the positive phase by upwelling to the surface and subsequent southward propagation to the KOE region. Thus, the westward propagation of Rossby wave in the subpolar region, mainly in the western basin from the date line to west coast, provides the decadal time scale selection to the PDO (Zhang and Delworth 2015).

Fig. 12.

Lagged regression of SST and surface wind stress anomalies vs the normalized PDO time series in the fully coupled (a)–(g) control run, (h)–(n) 2CO2, and (o)–(u) 0.5CO2. Units are °C for SST and N m−2 for surface wind stress. All data are 10–50-yr bandpass filtered to retain decadal variability. Note that for a given row the lags are different in different columns.

Fig. 12.

Lagged regression of SST and surface wind stress anomalies vs the normalized PDO time series in the fully coupled (a)–(g) control run, (h)–(n) 2CO2, and (o)–(u) 0.5CO2. Units are °C for SST and N m−2 for surface wind stress. All data are 10–50-yr bandpass filtered to retain decadal variability. Note that for a given row the lags are different in different columns.

Fig. 13.

As in Fig. 13, but for the zonal mean (140°–180°E) ocean temperature over the western Pacific.

Fig. 13.

As in Fig. 13, but for the zonal mean (140°–180°E) ocean temperature over the western Pacific.

We find that the subpolar pathway is still retained in a warm climate (Figs. 12h–n vs Figs. 12a–g; Figs. 13h–n vs Figs. 13a–g). The southward advection of western subpolar warm anomaly by the Oyashio in the upper 100 m is clearly seen in the 2CO2 experiment, but with a much faster speed than that in the control run. In the 2CO2 run, the subpolar warming anomaly reaches 40°N at a lag of 4 yr (Fig. 13l), leading to a diminishing of SST cooling anomaly over the KOE region, while the same happens at a lag of 6yr in the control run (Fig. 13e). Meanwhile, the significant appearance of western subpolar subsurface cold temperature anomaly in the 2CO2 run, which is induced by the positive wind stress curl during the positive phase of PDO, is much earlier than that in the fully coupled control run. The former is clearly seen at a lag of ~6 yr (Fig. 13n), while the latter obviously emerges at a lag of ~10 yr (Fig. 13g). These results imply that the subpolar ocean adjustment time associated with the westward propagated Rossby wave differs in these two experiments. To test this hypothesis, we calculate the first baroclinic Rossby wave speed over the North Pacific in both control and warm climate simulations using the eigenvalue method suggested by Chelton et al. (1998) (Figs. 14a,b). In the control run, the estimated Rossby wave speed is 0.70 cm s−1 around 50°N, equivalent to a cross-basin time scale of ~25 yr. This Rossby wave speed increases to 0.9 cm s−1 as the CO2 concentration is doubled, which equals ~18 yr cross-basin time (Figs. 14a,b). We note that the maximum wind stress curl anomaly in the subpolar region in GFDL model is located around 160°W. Therefore, the Rossby wave propagation in the subpolar region takes ~10 yr in the control run and ~6 yr in the 2CO2 experiment. Here, the speeding up of ocean Rossby waves in a warm climate is largely associated with the strengthening of ocean stratification (Fig. 14c). The subpolar pathway is also preserved very well in the 0.5CO2 experiment, but with a much slower propagation speed. For example, the subpolar warm SST approaches 40°N until at a lag of 12 yr (Figs. 12s and 13s), in sharp contrast to a 6-yr lag in the fully coupled control run (Figs. 12e and 13e). Opposite to the 2CO2 run, this slow evolution of the PDO is associated with the weakening of ocean stratification and associated slowdown of Rossby wave speed (Fig. 14d).

Fig. 14.

(a) Zonally averaged first baroclinic Rossby wave speed (cm s−1), (b) basin-crossing time (yr), and (c),(d) seawater temperature (°C) responses over the North Pacific Ocean in the fully coupled control run and 2CO2 and 0.5CO2 experiments, respectively. The black contours in (c) and (d) represents the long-term mean seawater temperature averaged over the North Pacific basin, while the shading denotes the temperature difference between the changed climate (2CO2 or 0.5CO2) and control run.

Fig. 14.

(a) Zonally averaged first baroclinic Rossby wave speed (cm s−1), (b) basin-crossing time (yr), and (c),(d) seawater temperature (°C) responses over the North Pacific Ocean in the fully coupled control run and 2CO2 and 0.5CO2 experiments, respectively. The black contours in (c) and (d) represents the long-term mean seawater temperature averaged over the North Pacific basin, while the shading denotes the temperature difference between the changed climate (2CO2 or 0.5CO2) and control run.

b. Subtropical pathway and its influence on PDO time scale

Zhang and Delworth (2015) suggested that the second pathway during the PDO cycle arises from the subtropical region. In the control run, a warm temperature anomaly emerges in the western subtropical subsurface at a lag of 2 yr (Fig. 13c) and then enhances (Figs. 13c–e), which is largely associated with the weak negative wind stress curl over the subtropical region during the mature positive phase of PDO (Fig. 12a). This warm subsurface temperature anomaly is further entrained to the surface and advected northward to the KOE region by the Kuroshio (Figs. 12c–g and 13c–g). The formation of subsurface warming anomaly associated with the adjustment of westward propagated downwelling Rossby wave takes about 5 yr (Figs. 13a–e), and the entrainment and advection of this warm anomaly to the KOE region take another 5 yr (Fig. 13e–g). The sum of these two processes accounts for 10 yr, which also provides a decadal time scale selection to the PDO (Zhang and Delworth 2015). In a warm (cold) climate, the formation of subtropical subsurface warm anomalies takes a shorter (longer) time as compared to the control run, with a significant warming at a lag of about 3 yr (9 yr) (Figs. 13h–k,o–r). This is in contrast to 4 yr in the fully coupled control run (Figs. 13a–e). The advection of warm western subtropical SST anomaly to the KOE region by the Kuroshio, again, takes a much shorter (longer) time in the 2CO2 (0.5CO2) simulation than in the control run (Figs. 12k–n,r–u and 13k–n,r–u vs Figs. 12d–g and 13d–g). Similar to the subpolar pathway, the fast (slow) evolution of the PDO cycle in the 2CO2 (0.5CO2) run is determined by the fast (slow) adjustment of subtropical ocean circulation due to the speed up (slowdown) of the ocean Rossby wave (Fig. 14). In the subtropical region, the first baroclinic Rossby wave speed is 2.5 cm s−1 around 30°N in the control run, equivalent to a cross-basin time scale of ~10 yr. This Rossby wave speed increases (decreases) to 3.3 (2.0) cm s−1 as the CO2 concentration is doubled (halved), which equals ~6 (16) yr cross-basin time. We note that the wind stress curl anomaly in the subtropical region in the GFDL model is mainly confined west of the date line. Thus, the Rossby wave propagation in the subtropical region takes about 3 (9) yr in the 2CO2 (0.5CO2) experiment. Moreover, the acceleration (deceleration) of the Kuroshio due to the anomalous interior wind stress curl takes a much shorter (longer) time in a warm (cold) climate compared to the control run due to the accelerated (decelerated) Rossby wave speed, and therefore leads to a slow (fast) advection of subtropical SST anomalies to the KOE region.

c. Midlatitude pathway and its influence on PDO time scale

In the fully coupled control run, the warm SST anomaly advected from the western subtropical and subpolar regions leads to a decay of SST cooling anomaly over the KOE region and a formation of a tripole-like SST anomaly over the midlatitudes (Figs. 12b–d). This tripole-like SST anomaly generated by the subtropical and subpolar pathways consists of a narrow cooling in the KOE region surrounded by warm anomalies to the north and south. This pattern favors an anticyclonic atmospheric circulation over the North Pacific (Zhang and Delworth 2015). The anticyclonic circulation first induces SST warming over the central northeastern Pacific by the fast northward Ekman transport and then generates a KOE SST warming after the adjustment of the first baroclinic Rossby wave (Figs. 12d–g). We refer to this midlatitude SST propagation as the midlatitude pathway. Here the subpolar, subtropical, and midlatitude pathways act like a relay game to finish the PDO full cycle (Zhang and Delworth 2015). We note that the tripole-like SST formation takes about 5 years and the warm SST propagation from the central North Pacific to the KOE region associated with the westward propagating Rossby wave takes about another 5 yr. The sum of these two processes accounts for 10 yr, a half cycle of the PDO. In warm climates, the tripole-like SST formation takes a shorter time (~3 yr) due to the fast advection of warm SST anomaly by the subpolar and subtropical pathways as mentioned above (Figs. 12i–k). The tripole-like SST anomaly in the midlatitude in warm climates also corresponds to an anticyclonic circulation over the North Pacific that first induces a SST warming in the central North Pacific at a lag of 3–4 yr and then a SST warming over the KOE region at a lag of 6 yr (Figs. 12k–n). Again, the delayed KOE SST response is due to the slow adjustment of the ocean Rossby wave, which is much faster than that in the fully coupled control run (Figs. 12k–n vs Figs. 12d–g) as a result of strengthened ocean stratification (Fig. 14c). In cold climates, the midlatitude pathway is also preserved but with a much slower speed (Figs. 12r–u). This is again attributed to the slow Rossby wave propagating speed due to the weakened ocean stratification in the midlatitudes (Fig. 14d).

6. Discussion and summary

In the present paper, we explore how the PDO responds to global warming or cooling using a fully coupled climate model, GFDL CM2.5_FLOR. To produce the warm (cold) climate we instantaneously double (halve) the amount of CO2 in the model atmosphere, and then let the model run for 500 yr to adjust to the altered radiative forcing. The model results show that the PDO spatial pattern in the altered climates is very similar to that in the control climate, but the amplitude and time scale are significantly changed.

First examining the case of a warmed climate, we find that the PDO amplitude is significantly weakened, with maximum decreases in the standard deviation of SST over the Kuroshio–Oyashio Extension (KOE) region. The PDO time scale also becomes shorter, with the approximately 20-yr characteristic time scale in the control simulation reducing to 10–15 yr in a warmed climate.

Forced by a double CO2 concentration, the surface air temperature warms quickly due to the reduced outgoing radiative heat flux that is the so-called greenhouse gas effect. This leads to enhanced moist convection in the tropics as a result of increased latent heat release and therefore generates an amplified tropical upper tropospheric warming. The amplified upper tropospheric warming favors an increased equator-to-pole temperature gradient in the upper troposphere and lower stratosphere in both hemispheres, which eventually produces strengthened and poleward shifted westerly winds in the midlatitudes. The westerly wind anomaly over the North Pacific Ocean is consistent with an enhanced subtropical high, which in turn strengthens the Kuroshio and associated ocean transport and therefore leads to a warm SST anomaly over the midlatitudes. The midlatitude SST warming decreases the mean meridional SST gradient in the KOE region. In the absence of changes in the meridional currents, the weakened mean meridional SST gradient contributes to decreased amplitude of SST variations over the KOE region and decreased amplitude of the PDO as well.

The poleward shifted storm track in a warm climate corresponds to decreased transient eddy vorticity fluxes from the upper troposphere to the low troposphere south of 45°N and therefore generates a decreased coefficient of midlatitude ocean feedback to the atmosphere. The weakened atmospheric circulation response over the North Pacific, as shown in the SLP, first drives a weak SST response in the central North Pacific by the anomalous Ekman transport and then leads to a weakened SST response over the KOE region by reducing the meridional shift of the gyre boundary after the adjustment of ocean Rossby waves. The mean temperature gradient weakens significantly in the KOE region in a warm climate, which further reinforces the decreased KOE SST response to a weakened atmosphere circulation.

In a cold climate, the PDO amplitude shows a significant strengthening. The physical mechanisms associated are similar to that under the global warming scenario but with an opposite sign. It is worth noting that the PDO amplitude response to CO2 forcing is not purely linear. The magnitude of the PDO amplitude increase in 0.5CO2 is larger than the magnitude of amplitude decrease in 2CO2 (Figs. 35). This nonlinear amplitude response of PDO is closely related to the meridional temperature gradient change. As shown in Figs. 8a–c, the magnitude of the SST meridional temperature gradient increase south of 40°N in the 0.5CO2 run is much larger than the magnitude of temperature gradient decrease in the 2CO2 experiment.

The PDO in the fully coupled control run has an approximately bidecadal peak. This quasi-20-yr period disappears in a warm climate and is replaced by higher-frequency peaks (e.g., a quasi-12-yr period). In a cold climate, however, the PDO period shifts to a lower-frequency band compared to the control run, with spectral peaks between 30 and 50 years. Physically, global warming (cooling) enhances (weakens) ocean stratification, increases (decreases) the Rossby wave speed, and in turn shortens (lengthens) the PDO period.

Some studies viewed the PDO as an extension of ENSO. The ENSO can impact the PDO via the atmosphere bridge, which is an important player in the PDO time scale selection (e.g., Newman et al. 2003; Strong and Magnusdottir 2009; Mills and Walsh 2013). The impact of ENSO on the PDO amplitude response in a changed climate is also assessed here. Figures 15a and 15b show that the ENSO amplitude is significantly weakened under a global warming scenario. This suppression in ENSO amplitude may be due to the change of ENSO meridional width that is caused by the change in mean meridional overturning circulation in the equatorial Pacific Ocean (Chen et al. 2015). No matter what mechanisms cause the ENSO amplitude response in a warm climate (this is not the focus of the current study), the ENSO change seems to contribute, at least partly, to the weakened PDO amplitude response in the 2CO2 run. In sharp contrast, the ENSO amplitude shows a much weaker response in a cold climate, with a small increase (decrease) west (east) of 160°E (Fig. 15c vs Fig. 15e). The ENSO amplitude change becomes even weaker on decadal time scales (Figs. 15e,f), whereas the PDO amplitude significantly increases in the North Pacific (Fig. 15d). This indicates that the increase of PDO amplitude in a cold climate is dominated by physical processes inside the North Pacific, rather than teleconnections from the tropical ENSO. A comparison between the decadal SST amplitudes in both warm and cold climates (Fig. 15d vs Fig. 15f) implies that the PDO amplitude response in the 2CO2 simulation is not dominated by the ENSO teleconnection. If the ENSO teleconnection played a dominant role, the PDO amplitude would be much stronger in the warm climate than in the cold climate because of such a strong decrease of the ENSO amplitude in the 2CO2 run. These results are consistent with the arguments by Zhang and Delworth (2015), who pointed out that the PDO in the GFDL model is independent of ENSO, but ENSO can slightly enhance the PDO amplitude, particularly over the central North Pacific. Nevertheless, it is still possible that the impacts of ENSO on the PDO period are model dependent. Wang et al. (2012) found the NCEP CFS model captures the fundamental PDO characteristics with increased amplitude in the ENSO-included run, consistent with our results (Zhang and Delworth 2015). However, the active ENSO events can shift the PDO to lower frequency, which is not seen in the GFDL model. Therefore, comparison of the PDO response to global warming in GFDL model with that simulated by other climate models will be very helpful in future.

Fig. 15.

(a),(b) Standard deviation (STD) of (left) unfiltered and (right) 7-yr low-pass filtered annually averaged SST in the fully coupled control run. (c),(d) STD difference between the 2CO2 run and fully coupled control simulation. (e),(f) As in (c),(d), but for difference between the 0.5CO2 run and control experiment. Unit is °C. The gray points overlapped on shading in (c)–(f) mean the STD difference is significant at 95% confidence level using an F test.

Fig. 15.

(a),(b) Standard deviation (STD) of (left) unfiltered and (right) 7-yr low-pass filtered annually averaged SST in the fully coupled control run. (c),(d) STD difference between the 2CO2 run and fully coupled control simulation. (e),(f) As in (c),(d), but for difference between the 0.5CO2 run and control experiment. Unit is °C. The gray points overlapped on shading in (c)–(f) mean the STD difference is significant at 95% confidence level using an F test.

Acknowledgments

The authors thank Hiroyuki Murakami, Xiaosong Yang, and Wei Zhang for their valuable suggestions and constructive comments on our preliminary manuscript.

REFERENCES

REFERENCES
Cayan
,
D. R.
,
M. D.
Dettinger
,
S. A.
Kammerdiener
,
J. M.
Caprio
, and
D. H.
Peterson
,
2001
:
Changes in the onset of spring in the western United States
.
Bull. Amer. Meteor. Soc.
,
82
,
399
415
, doi:.
Chelton
,
D. B.
,
R. A.
deSzoeke
,
M. G.
Schlax
,
K. E.
Naggar
, and
N.
Siwertz
,
1998
:
Geographical variability of the first baroclinic Rossby radius of deformation
.
J. Phys. Oceanogr.
,
28
,
433
460
, doi:.
Chen
,
L.
,
T.
Li
, and
Y.
Yu
,
2015
:
Causes of strengthening and weakening of ENSO amplitude under global warming in four CMIP5 models
.
J. Climate
,
28
,
3250
3274
, doi:.
Cheng
,
W.
,
J. C. H.
Chiang
, and
D.
Zhang
,
2013
:
Atlantic meridional overturning circulation (AMOC) in CMIP5 models: RCP and historical simulations
.
J. Climate
,
26
,
7187
7197
, doi:.
Delworth
,
T. L.
, and Coauthors
,
2006
:
GFDL’s CM2 global coupled climate models. Part I: Formulation and simulation characteristics
.
J. Climate
,
19
,
643
674
, doi:.
Deser
,
C.
, and
M. L.
Blackmon
,
1995
:
On the relationship between tropical and North Pacific sea surface temperature variations
.
J. Climate
,
8
,
1677
1680
, doi:.
d’Orgeville
,
M. D.
, and
W. R.
Peltier
,
2009
:
Implications of both statistical equilibrium and global warming simulations with CCSM3. Part I: On the decadal variability in the North Pacific basin
.
J. Climate
,
22
,
5277
5297
, doi:.
Fang
,
C.
,
L.
Wu
, and
X.
Zhang
,
2014
:
The impact of global warming on the Pacific decadal oscillation and the possible mechanism
.
Adv. Atmos. Sci.
,
31
,
118
130
, doi:.
Frankignoul
,
C.
, and
N.
Sennéchael
,
2007
:
Observed influence of North Pacific SST anomalies on the atmospheric circulation
.
J. Climate
,
20
,
592
606
, doi:.
Frierson
,
D. M. W.
,
2006
:
Robust increases in midlatitude static stability in simulations of global warming
.
Geophys. Res. Lett.
,
33
,
L24816
, doi:.
Han
,
W.
, and Coauthors
,
2014
:
Intensification of decadal and multi-decadal sea level variability in the western tropical Pacific during recent decades
.
Climate Dyn.
,
43
,
1357
1379
, doi:.
Holland
,
M. M.
, and
C. M.
Bitz
,
2003
:
Polar amplification of climate change in coupled models
.
Climate Dyn.
,
21
,
221
232
, doi:.
Kushnir
,
Y.
,
W. A.
Robinson
,
I.
Bladé
,
N. M. J.
Hall
,
S.
Peng
, and
R.
Sutton
,
2002
:
Atmospheric GCM response to extratropical SST anomalies: Synthesis and evaluation
.
J. Climate
,
15
,
2233
2256
, doi:.
Kwon
,
Y.-O.
, and
C.
Deser
,
2007
:
North Pacific decadal variability in the Community Climate System Model version 2
.
J. Climate
,
20
,
2416
2433
, doi:.
Latif
,
M.
, and
T. P.
Barnett
,
1994
:
Causes of decadal climate variability over the North Pacific and North America
.
Science
,
266
,
634
637
, doi:.
Lu
,
J.
,
G.
Chen
, and
D. M. W.
Frierson
,
2008
:
Response of the zonal mean atmospheric circulation to El Niño versus global warming
.
J. Climate
,
21
,
5835
5851
, doi:.
Ma
,
H.
, and
L.
Wu
,
2011
:
Global teleconnections in response to freshening over the Antarctic Ocean
.
J. Climate
,
24
,
1071
1088
, doi:.
Mantua
,
N. J.
, and
S. R.
Hare
,
2002
:
The Pacific decadal oscillation
.
J. Oceanogr.
,
58
,
35
44
, doi:.
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
, doi:.
Mills
,
C. M.
, and
J. E.
Walsh
,
2013
:
Seasonal variation and spatial patterns of the atmospheric component of the Pacific decadal oscillation
.
J. Climate
,
26
,
1575
1594
, doi:.
Nakamura
,
H.
, and
A. S.
Kazmin
,
2003
:
Decadal changes in the North Pacific oceanic frontal zones as revealed in ship and satellite observations
.
J. Geophys. Res.
,
108
,
3078
, doi:.
Newman
,
N.
,
G. P.
Compo
, and
M. A.
Alexander
,
2003
:
ENSO-forced variability of the Pacific decadal oscillation
.
J. Climate
,
16
,
3853
3857
, doi:.
Power
,
S.
, and Coauthors
,
1999
:
Inter-decadal modulation of the impact of ENSO on Australia
.
Climate Dyn.
,
15
,
319
324
, doi:.
Schneider
,
N.
,
A. J.
Miller
,
M. A.
Alexander
, and
C.
Deser
,
1999
:
Subduction of decadal North Pacific temperature anomalies: Observations and dynamics
.
J. Phys. Oceanogr.
,
29
,
1056
1070
, doi:.
Schneider
,
N.
,
A. J.
Miller
, and
D. W.
Pierce
,
2002
:
Anatomy of North Pacific decadal variability
.
J. Climate
,
15
,
586
605
, doi:.
Seager
,
R.
,
Y.
Kushnir
,
N. H.
Naik
,
M. A.
Cane
, and
J.
Miller
,
2001
:
Wind-driven shifts in the latitude of the Kuroshio–Oyashio Extension and generation of SST anomalies on decadal timescales
.
J. Climate
,
14
,
4249
4265
, doi:.
Strong
,
C.
, and
G.
Magnusdottir
,
2009
:
The role of tropospheric Rossby wave breaking in the Pacific decadal oscillation
.
J. Climate
,
22
,
1819
1833
, doi:.
Thompson
,
L.
, and
Y.
Kwon
,
2010
:
An enhancement of low-frequency variability in the Kuroshio–Oyashio Extension in CCSM3 owing to ocean model biases
.
J. Climate
,
23
,
6221
6233
, doi:.
Vecchi
,
G. A.
, and Coauthors
,
2014
:
On the seasonal forecasting to regional tropical cyclone activity
.
J. Climate
,
27
,
7994
8016
, doi:.
Wang
,
H.
,
A.
Kumar
,
W.
Wang
, and
Y.
Xue
,
2012
:
Influence of ENSO on Pacific decadal variability: An analysis based on the NCEP Climate Forecast System
.
J. Climate
,
25
,
6136
6151
, doi:.
Wu
,
L.
,
E. L.
Dong
, and
Z.
Liu
,
2005
:
The 1976/77 North Pacific climate regime shift: The role of subtropical ocean adjustment and coupled ocean–atmosphere feedbacks
.
J. Climate
,
18
,
5125
5140
, doi:.
Wu
,
L.
, and Coauthors
,
2012
:
Enhanced warming over the global subtropical western boundary currents
.
Nat. Climate Change
,
2
,
161
166
, doi:.
Yin
,
J. H.
,
2005
:
A consistent poleward shift of the storm tracks in simulations of 21st century climate
.
Geophys. Res. Lett.
,
32
,
L18701
, doi:.
Zhang
,
L.
, and
L.
Wu
,
2012
:
Can oceanic freshwater flux amplify global warming?
J. Climate
,
25
,
3417
3430
, doi:.
Zhang
,
L.
, and
C.
Wang
,
2013
:
Multidecadal North Atlantic sea surface temperature and Atlantic meridional overturning circulation variability in CMIP5 historical simulation
.
J. Geophys. Res. Oceans
,
118
,
5772
5791
, doi:.
Zhang
,
L.
, and
T. L.
Delworth
,
2015
:
Analysis of the characteristics and mechanisms of the Pacific decadal oscillation in a suite of coupled models from the Geophysical Fluid Dynamics Laboratory
.
J. Climate
,
28
,
7678
7701
, doi:.
Zhang
,
L.
,
L.
Wu
, and
J.
Zhang
,
2011
:
Coupled ocean–atmosphere responses to recent freshwater changes over the Kuroshio Oyashio Extension region
.
J. Climate
,
24
,
1507
1524
, doi:.
Zhang
,
L.
,
C.
Wang
, and
S.-K.
Lee
,
2014
:
Potential role of Atlantic warm pool–induced freshwater forcing in the Atlantic meridional overturning circulation: Ocean–sea ice model simulations
.
Climate Dyn.
,
43
,
553
574
, doi:.
Zhang
,
R.
,
2008
:
Coherent surface-subsurface fingerprint of the Atlantic meridional overturning circulation
.
Geophys. Res. Lett.
,
35
,
L20705
, doi:.
Zhong
,
Y.
, and
Z.
Liu
,
2009
:
On the mechanism of Pacific multidecadal climate variability in CCSM3: The role of the subpolar North Pacific Ocean
.
J. Climate
,
39
,
2052
2076
, doi:.