Decadal Variability in the North Pacific: The Eastern North Pacific Mode*

Transcription

1 3111 Decadal Variability in the North Pacific: The Eastern North Pacific Mode* LIXIN WU ANDZHENGYU LIU Center for Climatic Research, University of Wisconsin Madison, Madison, Wisconsin (Manuscript received 25 November 2002, in final form 7 April 2003) ABSTRACT Decadal variability in the North Pacific is studied in a series of coupled global ocean atmosphere simulations using coupled modeling surgery a set of modeling approaches that can be used to identify the origins and causes of a specific variability mode in the coupled climate system. Both modeling and observational studies suggest two distinctive internal modes in the North Pacific: the North Pacific mode (NPM) and the eastern North Pacific mode (ENPM). The ENPM originates from atmospheric stochastic forcing through spatial resonance. Both local ocean atmosphere coupling and remote tropical teleconnective forcing can enhance the ENPM, but none of them is a necessary precondition. The influence of the tropical forcing in the midlatitudes is dominated by atmospheric teleconnection, while the oceanic teleconnection is negligible. The upper-ocean heat budget reveals that SST anomalies in the central North Pacific and the eastern North Pacific are generated by anomalous Ekman advection and surface heat flux, respectively. In contrast to the ENPM, the NPM critically depends on local ocean atmosphere coupled feedback, although the atmospheric stochastic forcing can generate a NPMlike mode with much reduced amplitudes and no preferred timescale. 1. Introduction Sea surface temperature (SST) in the Pacific basin exhibits significant decadal to multidecadal variations (e.g., Nitta and Yamata 1989; Nakamura et al. 1997; Deser and Blackmon 1995; Zhang et al. 1997; Manuta et al. 1997). Observations seem to suggest two important decadal variability modes in the Pacific: a decadal to bidecadal ENSO-like mode (e.g., Zhang et al. 1997) and a multidecadal North Pacific mode (e.g., Deser and Blackmon 1995; Nakamura et al. 1997; Mestas-Nunez and Enfield 1999). Modeling studies further suggest that the ENSO-like mode seems to originate mainly from the tropical Pacific region, while the North Pacific mode is generated predominantly in the North Pacific (Barnett et al. 1999b; Pierce et al. 2001; Liu et al. 2002; Wu et al. 2003). These studies imply that decadal variability over the Pacific could be caused by multiple modes and associated with multiple mechanisms. Further observational studies suggest that even within the North Pacific region, there appears to exist multiple decadal modes (e.g., Nakamura et al. 1997; Barlow et al. 2001; Luo and Yamagata 2002). For example, in * Center for Climatic Research, University of Wisconsin Madison Contribution Number 816. Corresponding author address: Lixin Wu, Center for Climatic Research-IES, University of Wisconsin Madison, 1225 West Dayton Street, Madison, WI lixinwu@facstaff.wisc.edu addition to the North Pacific mode that is concentrated predominantly in the subarctic frontal zone of the North Pacific (e.g., Deser and Blackmon 1995), Nakamura et al. (1997) suggest another mode is concentrated in the eastern North Pacific and extends into the tropical Pacific. This mode is also detected in other observational studies where longer datasets and different statistical approaches are used (e.g., Mestas-Nunez and Enfield 1999; Barlow et al. 2001). Based on those observational studies alone, it is difficult to tell whether this mode is internally generated in the North Pacific, or is forced by the tropical teleconnective forcing. Indeed, in these previous observations studies, the various Pacific decadal variability modes are extracted from different datasets and by using different statistical methods. As such, the distinction among the various decadal variability modes has remained largely elusive. Recent coupled modeling studies have advanced our understanding of decadal climate variability in the Pacific significantly (Miller and Schneider 2000). Latif and Barnett (1994) proposed that coupled ocean atmosphere interaction can generate an interdecadal mode in the North Pacific without invoking tropical teleconnection. Further studies also highlighted the importance of atmospheric stochastic forcing (e.g., Saravanan and McWilliams 1997; Jin 1997; Frankignoul et al. 1997; Neelin and Weng 1999; Barnett et al. 1999b). Some studies proposed that the Pacific decadal variability is generated in the combined extratropical tropical climate system (e.g., Gu and Philander 1997; Kleeman et al. 1999) with the extratropical tropical teleconnective 2003 American Meteorological Society

2 3112 JOURNAL OF CLIMATE VOLUME 16 FIG. 1. Spatial patterns of the two leading modes of observed SST in the North Pacific, obtained by regressing SST against the PC time series of REOF modes of North Pacific SSTs. (a) NPM (REOF1), (b) ENPM (REOF2), and (c) normalized PC time series. Only cold season (Oct Mar) SSTs are used and data have been low-pass ( 8 yr) filtered.

3 3113 forcing playing the central role. Using a modeling surgery approach, Liu et al. (2002) and Wu et al. (2003) (hereafter LW) demonstrate that Pacific decadal variability possesses two distinct modes: a decadal to bidecadal tropical Pacific mode (TPM), and a multidecadal North Pacific mode (NPM). The TPM and NPM are generated predominantly by local ocean atmosphere interaction in the tropical and North Pacific, respectively, but are also affected by tropical extratropical teleconnections via the atmosphere and the ocean. This paper is a follow-up of LW with the focus now on the North Pacific region. We will investigate the mechanisms of the eastern North Pacific mode (ENPM). To better understand ENPM, we will also show NPM as a comparison but without a comprehensive discussion because NPM has been studied previously in LW. To avoid the confusion of the definition, here NPM refers to the mode of North Pacific decadal variability that is centered in the subarctic frontal zone (e.g., Deser and Blackmon 1995), while ENPM refers to the mode that is predominantly in the central to eastern North Pacific (Nakamura et al. 1997; Mestas-Nunez and Enfield 1999; Barlow et al. 2001; Luo and Yamagata 2002). This latter ENPM, as discussed earlier, has sometimes been referred to as the Pacific decadal oscillation (PDO) in previous observations (e.g., Mestas-Nunez and Enfield 1999; Barlow et al. 2001). In this paper, we will pay special attention to the mechanism of the ENPM. This is because the ENPM is relatively less understood. Based on a series of coupled ocean atmosphere GCM experiments, we propose here that the ENPM, together with the NPM are the two dominant internal decadal variability modes in the North Pacific. The ENPM originates from atmospheric stochastic forcing through spatial resonance (Saravanan and McWilliams 1997), and can be enhanced by both local ocean atmosphere coupling and tropical teleconnective forcing. The paper is organized as follows. Sections 2 and 3 present evidence of the distinction between the NPM and ENPM in observations and a control experiment of our coupled GCM, respectively. In section 4, we will briefly describe the strategies of the modeling surgery used in coupled GCM sensitivity experiments. The origins and mechanisms of the ENPM are studied in section 5. Section 6 concludes the paper with some further discussions. FIG. 2. (a) Autocorrelation of the PC time series of REOF modes corresponding to the NPM and ENPM. (b) Cross correlation between ENP, NP, and TP SST indices (NP: N, 140 E 180 W; ENP: N, W; TP: 10 S 10 N, W). The 95% confidence level for the correlation between ENP and TP SSTs is indicated. 2. Evidence for the eastern North Pacific mode a. The observed ENPM The observational SST dataset used in this study is the reconstruction of historical SST anomalies (SSTAs) generated by Kaplan et al. (1998, hereafter K98) on a 5 5 grid for the period In this study, cold season (from October to March) SSTAs are used. To focus on decadal timescale, most analyses are performed after the data are low-passed filtered to retain variability longer than 8 yr. To extract the modes in the North Pacific, we used rotated EOF (REOF) rather than unrotated EOF, which may yield an unphysical mode due to the orthogonal constraints on the temporal and spatial domains. A detailed discussion about the application of the rotated EOF versus unrotated EOF in extracting the Pacific variability modes can be seen in a recent paper by Barlow et al. (2001). The NPM and ENPM correspond to the two leading rotated EOF modes of the North Pacific (north of 20 N) SSTAs. The two modes explain about 26% and 20% of the field variance, respectively. To show the large-scale pattern associated with these two modes, the full Pacific basin SSTs are regressed on the normalized temporal coefficient of each mode (Fig. 1). The NPM (Fig. 1a) is largely restricted in the midlatitude North Pacific, and has some weak amplitudes of opposite phase in the tropical Pacific, reflecting perhaps the influence of the NPM in the Tropics (e.g., Barnett et al. 1999a; Vimont et al. 2001; LW). This mode is very similar to that extracted from the annual mean SSTAs (Mestas-Nunez and Enfield 1999; Wu et al. 2003) and is also similar to other previous studies that used different datasets (e.g., Deser and Blackmon 1995; Nakamura et al. 1997). In contrast to the NPM, the ENPM (Fig. 1b) has large amplitudes

4 3114 JOURNAL OF CLIMATE VOLUME 16 FIG. 3. Same as Fig. 1, but for the SST residual after removing the component of linear response to the TP SST change. predominantly in the eastern North Pacific, which extends northward along the coast of North America and southward into the central equatorial Pacific and eastern tropical South Pacific. The ENPM is somewhat similar to the ENSO-like decadal mode of Zhang et al. (1997) and the PDO mode of Manuta et al. (1997), but with a signal weaker in the North Pacific and equatorial eastern Pacific. The NPM and ENPM explain about 80% of the variance in the Kuroshio Extension region and eastern North Pacific, respectively (not shown), indicating the

5 3115 FIG. 4. REOF patterns of observed low-passed ( 8 yr) cold season SSTs in the Pacific. (a) TPM (REOF1), (b) NPM (REOF3), (c) ENPM (REOF4), and (d) normalized PC time series of REOF modes. dominant role of these two modes in local climate variability. The ENPM and NPM seem to evolve with different timescales, as shown by the normalized principal components (PCs; Fig. 1c). While the NPM is characterized predominantly by multidecadal timescales, the ENPM seems to evolve at decadal to interdecadal timescales. The difference of timescales between the NPM and ENPM can be seen in the autocorrelation of each PC (Fig. 2a). Relatively, the decorrelation timescale of the ENPM is about half that of the NPM, although the decorrelation timescales of both modes can be affected by the low-pass filtering. The ENPM seems to evolve largely independent of the NPM, as shown by the correlation between the low-passed time series of SSTAs averaged over the region around the maximum loadings of each mode (Fig. 2b) (each SST time series is very similar to the PC time series of the corresponding REOF mode with a correlation coefficient of above 0.9). For no lag is the ENPM correlated significantly to the NPM (Fig. 2b). The ENPM seems to be associated with SSTA in the central equatorial Pacific. This is in contrast to the NPM, which is largely independent of the Tropics (LW). The maximum correlation between the ENPM and the central equatorial SSTs occurs at 0 lag with a value of 0.68, significant at 95% level (Fig. 2b). The influence of the tropical teleconnective forcing can be further estimated by performing REOF on the residual North Pacific SST in which the linear regression to the tropical SST is removed. The REOF modes of residual SST in the North Pacific reproduce the broad features of the NPM and ENPM (Fig. 3 versus Fig. 1). In the meantime, the amplitudes of the ENPM (Fig. 3b) is reduced by about 30% 40% in both the central and eastern Pacific (Fig. 3b versus Fig. 1b), in contrast to the NPM, which remains virtually unchanged (Fig. 3a versus Fig. 1a). This tends to suggest that the origin of the ENPM may not be in the Tropics, although it can enhance the amplitudes of the ENPM. The ENPM and NPM can also be seen in the basinwide REOF modes (Fig. 4). While the first REOF mode (Fig. 4a) corresponds to the TPM, having large amplitudes predominantly in the tropical Pacific, the third and fourth REOF modes correspond to the NPM and ENPM, respectively (Figs. 4b,c). Both the ENPM and NPM explain almost the same fraction (12%) of the basin SST variance. The NPM shown in the basin REOF is virtually identical to the local REOF mode (Fig. 4b versus Fig. 1a), while the ENPM broadly resembles the mode after removing the linear response to the tropical SST change (Fig. 4c versus Fig. 3b), with a much reduced signal in the Tropics and somewhat reduced amplitudes in the North Pacific (Fig. 4c versus Fig. 1b). In summary, analyses of the observational SST provide tentative support for the existence of two distinctive modes in the North Pacific: the ENPM and the NPM. The main focus of our study below is to identify the origin of the ENPM using a coupled climate model. b. The model-simulated ENPM We used the Fast Ocean Atmosphere model (FOAM, version 1.0), which was jointly developed at the Uni-

6 3116 JOURNAL OF CLIMATE VOLUME 16 FIG. 5. Spatial patterns of the two leading REOF modes of the simulated (control run) SST variability in the North Pacific: (a) ENPM (REOF1), (b) NPM (REOF2), and (c) normalized PC time series. Only cold season (Oct Mar) SSTs are used, and data (400 yr) have been low-pass ( 8 yr) filtered.

7 3117 FIG. 6. Same as Fig. 2, but for the model control simulation. versity of Wisconsin Madison and Argonne National Laboratory (Jacob 1997). The atmospheric model is a fully parallel version of the National Center for Atmospheric Research (NCAR) Community Climate Model (CCM2) [the Parallel Community Climate Model (PCCM2) at R15], but with the atmospheric physics replaced by those of CCM3 and 17 vertical levels. The ocean model is conceptually similar to the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model (MOM) with a resolution of 1.4 latitude 2.8 longitude 16 vertical levels. A simple thermodynamic sea ice mode is incorporated. The model has been run over 800 yr without apparent climate drift. The last 400 yr of the simulation are used here as the control experiment (CTRL). Consistent with the observation, we only use the cold season (October March) SST anomalies. In addition to the 8-yr high-frequency cutoff, a low-frequency cutoff of 200 yr is also applied in the analysis to remove any possible contribution from the effects of slow change in the deep ocean. The ENPM and NPM appear as the first and second REOF modes of the North Pacific SSTs in the CTRL simulation, explaining 18% and 15% of field variance, respectively. The basin-scale structures of these two modes are derived by regressing the basin SST to the PC time series of each mode (Fig. 5). The model-simulated ENPM and NPM capture the main features of the observed ones. As observed, the simulated ENPM has the largest amplitudes in the eastern North Pacific at about 25 N, which extends northward along the coast of the North America, and southward to the central equatorial Pacific. There are also some modest amplitudes in the central North Pacific with phase opposite to the eastern North Pacific. The difference between the observed and the model-simulated ENPM pattern is shown west of the date line and southeast of the equatorial Pacific. As in the observation, the simulated ENPM and NPM evolve independently, with SSTs in the center area of the ENPM and NPM uncorrelated at all lags (Fig. 6b). Visually, the simulated ENPM also seems to exhibit a shorter timescale than the NPM (Fig. 5c), with the decorrelation timescale shorter than the NPM (Fig. 6a). Similar to the observation, the simulated ENPM tends to be correlated with the tropical SST change, with a maximum correlation 0.28 at lag 0 (significant at 95% level; Fig. 6b). To further assess the influence of the tropical forcing on the North Pacific in the model, we repeat the rotated EOF analyses on the residual North Pacific SSTs after removing the component that is linearly regressed with the tropical SST change (Fig. 7). Both the ENPM and NPM modes remain similar to the original ones (Fig. 7 versus Fig. 5). The ENPM is reduced significantly (30%) in the central North Pacific, as in the observation, but is not reduced significantly in the eastern North Pacific, which is somewhat different from the observation. The ENPM and NPM also appear in the REOFs of the Pacific basin SSTs. While the first REOF shows the TPM (Fig. 8a), the second and third REOF modes correspond to the ENPM and NPM, respectively (Figs. 8b,c). Again, the ENPM shown by the basin REOF resembles the ENPM after removing the tropical influence (Fig. 7a). As in the observation, this linear analysis indicates that the ENPM may originate from local ocean atmosphere interaction in the North Pacific, although it could be enhanced by the tropical teleconnective forcing. The power spectrum of the ENPM and the NPM are calculated using a multitaper spectrum method (Mann and Lees 1996). To exclude the nonphysical power peaks caused by filtering, the power spectrum is based on the unfiltered temporal coefficient that is obtained by projecting the REOF mode onto the original unfiltered SST field in the North Pacific. The power spectrum of the ENPM shows some substantial variance at decadal to interdecadal timescales with a weak peak at around 11 yr (Fig. 9a). In contrast, the NPM shows substantial variance at longer multidecadal timescales (Fig. 9b), consistent with the observation. It is noted that the power spectrum of the TPM also exhibits a weak decadal peak at around 11 yr (Fig. 9c). In addition, both the ENPM and NPM modes show some interannual variability. Associated with these two North Pacific SST modes, the atmosphere exhibits distinct variability patterns. Figure 10 shows the regression of October March 500-hPa geopotential height against the normalized temporal coefficient of the NPM and ENPM, respectively. The 500-

8 3118 JOURNAL OF CLIMATE VOLUME 16 FIG. 7. Same as Fig. 5, but for the residual SST after removing the component of linear response to the TP SST change. hpa pattern associated with the ENPM shows a dipolelike structure over the Pacific North America region, with anomalies occurring in the vicinity of the Aleutian low, and anomalies of the opposite polarity over northwestern Canada (Fig. 10a). This corresponding pattern was seen in some previous observation (e.g., Luo and Yamagata 2002) and bears some similarity to the observed Pacific North American (PNA) pattern (Horel

9 3119 FIG. 8. Spatial patterns of the three leading REOF modes of the Pacific basin SSTs: (a) TPM (REOF1), (b) ENPM (REOF2), (c) NPM, (REOF3), (d) normalized PC time series. Unit is C. and Wallace 1981). By contrast, the 500-hPa pattern associated with the NPM exhibits an elongated zonal structure of the same polarity stretching from east Asia to North America (Fig. 10b). This zonal structure appears similar to the North Pacific pattern (Barnston and Livezey 1987), with the largest amplitudes also in the vicinity of the Aleutian low. Overall, the model captures the major features of the observed ENPM and NPM, and provides tentative evidences for the existence of these two distinctive modes. While both the observation and model control simulation provide useful insight on the ENPM, they reveal little on the causes and mechanisms of the ENPM. This is because both the observed and the model control climate are already the final product after complex feedbacks. To identify the causes and mechanisms of the ENPM, we will analyze a series of coupled sensitivity experiments that are performed after the coupled modeling surgery. 3. Coupled modeling surgery The coupled modeling surgery (CMS) broadly represents a set of modeling approaches that can be used to identify the origins and causes of a specific variability mode in the coupled climate system. Two CMS strategies have been implemented: the partial coupling (hereafter referred to as PC) and the partial blocking (PB; LW). In the PC experiments, the atmospheric model sees a prescribed annual cycle of SST that is obtained from the CTRL in a specified region (called the PC region) and sees the predicted SST from the full ocean atmosphere coupling elsewhere. The ocean model is forced by the full atmosphere ocean flux calculated by the atmospheric model over the entire domain. Over the PC domain, the surface fluxes that drive the ocean model are calculated using the SST predicted by the ocean model at each time step. Variability can still be generated in both the ocean and atmosphere in the PC region because of internal atmospheric variability. Our oneway coupling over the PC domain exerts a stronger negative thermodynamic feedback than in the coupled case, opposite to the other extreme case where an oceanalone experiment is forced by pure flux with no negative SST feedback at all. The former, relative to the fully coupled case, tends to overdamp the SST variability while the latter overestimates SST variability (e.g., Saravanan and McWilliams 1997). The PC approach may underestimate the atmospheric stochastic variability due to the fixed SST boundary condition (Barsugli and Batissti 1998). Nevertheless, the fixed SST forcing of the AGCM is a frequently used approach in AGCM experiments, and therefore provides a useful reference to study the role of coupled ocean atmosphere interaction in generating climate variability (e.g., Saravanan and McWilliams 1997; Chang et al. 2000). In the PB experiments, sponge walls are placed at specified latitudinal bands of the ocean component of the coupled system, such that oceanic teleconnection in different latitudes is cut off. Within the sponge walls, temperature and salinity are restored toward the annual cycle of the CTRL run.

10 3120 JOURNAL OF CLIMATE VOLUME The ENPM and its origins Both the observation and model control simulation are indicative of the influence of the tropical teleconnective forcing. Some theories (e.g., Gu and Philander 1997) also proposed that the tropical telconnective forcing and extratropical tropical oceanic teleconnection are critical to generate this pan-pacific decadal mode. In this section, the origins and causes of the ENPM will be explored in a series of CMS experiments. In particular, we will address the following issues: 1) What is the role of the tropical teleconnective forcing in the generation of the ENPM? In other words, is the ENPM internal to the North Pacific or forced by the tropical teleconnective forcing? 2) What is the relative role of coupled ocean atmosphere interaction versus atmospheric stochastic forcing in the ENPM? 3) What are the mechanisms of the ENPM? a. The role of tropical teleconnective forcing FIG. 9. Power spectrum of the unfiltered time series associated with the ENPM, NPM, and TPM. The time series is obtained by projecting each mode on the original unfiltered SST. The power spectrum is calculated using a multitaper spectrum method (Mann and Lees 1996) with 50% (lower) and 95% (upper) statistical significance levels indicated. Previous studies have shown that ENSO can affect the North Pacific via atmospheric teleconnections (e.g., Horel and Wallace 1981; Graham 1994; Lau and Nath 1996). The linear regression analysis of both the observation and the model control run in the previous section also indicates that the tropical SST change can enhance the amplitudes of the ENPM. Here, we will further assess the role of the tropical teleconnective forcing in the generation of the ENPM using the CMS approach. We use the PC experiment PC-ET from LW, in which the tropical influence is removed by prescribing the climatological annual cycle of SST of the CTRL run equatorward of 20. In the extratropics poleward of 20, full ocean atmosphere coupling remains active as in the CTRL. Without air sea coupling in the Tropics, ENSO variability is nearly removed with the variance reduced by over 90%. In comparison with the CTRL, this approach allows us to assess the effect of the tropical teleconnective forcing more quantitatively. The ENPM and NPM in PC-ET correspond to the two leading REOF modes in the North Pacific, each of which explains about 16% of low-frequency SST variance over the North Pacific (Figs. 11a,b). The ENPM appears broadly similar to that in CTRL with a pattern correlation of 0.90 north of 20 N. As in the CTRL, the ENPM in PC-ET has strong amplitudes along the western coast of the North America and modest amplitudes of opposite polarity in the central North Pacific. The atmospheric variability patterns associated with the ENPM and NPM in PC-ET also remain broadly similar to those in CTRL (Figs. 12a1,b1 versus Figs. 10a,b). The message from the PC-ET experiment is clear: like the NPM in LW, the ENPM is also predominantly internal to the North Pacific climate system, with no need of invoking the tropical teleconnective forcing.

11 3121 FIG. 10. Spatial patterns of atmospheric 500-mb geopotential height anomalies associated with (a) ENPM and (b) NPM, obtained by regressing the atmospheric field on the normalized temporal coefficient of each mode. Contour interval is 1 m. This conclusion does not exclude the role of atmospheric teleconnective forcing on the ENPM. Without the tropical teleconnective forcing, the amplitude of the ENPM is reduced significantly by 30% in the central North Pacific. This change is consistent with the linear regression analysis in the control run. Indeed, the ENPM in PC-ET is almost identical to that found from the residual SST of the control run in both the pattern and the amplitude (Fig. 11a versus Fig. 7a), indicating a dominant linear response of the North Pacific to the Tropics. The associated atmospheric variability pattern also reveals some changes as the tropical teleconnective forcing is removed. The notable change is seen in North America, where the height anomalies shift toward lower latitudes (Fig. 12a1 versus Fig. 10a). The amplitude of the height anomalies shows some slight reduction of about 20% compared to CTRL. The NPM exhibits a similar change (as shown in LW). The amplitudes of the 500-hPa height anomalies associated with the NPM in PC-ET are slightly reduced while the pattern remains virtually unchanged in PC-ET compared with the CTRL (Fig. 12b1 versus Fig. 10b). In addition to the weakening of the amplitude, the temporal evolution of the ENPM undergoes substantial change in PC-ET (Fig. 13). In the absence of the tropical SST change, for the ENPM, the decadal peak around 11 yr that exists in CTRL disappears (Fig. 13a versus Fig. 9a). Instead, substantial variability occurs at interdecadal timescales around yr. For the NPM, substantial variability occurs at around interdecadal (around 20 yr) and multidecadal (around 50 yr) timescales, respectively (Fig. 13b versus Fig. 9b). The frequency shift of the NPM caused by the elimination of the tropical teleconnective forcing is consistent with our previous analyses using annual data (LW). In summary, the PC-ET experiment suggest that the ENPM is an internal North Pacific mode, which predominantly arises from local ocean atmosphere inter-

12 3122 JOURNAL OF CLIMATE VOLUME 16 FIG. 11. Spatial patterns of the (a), (c) ENPM and (b), (d) NPM modes in the experiments: (a), (b) PC-ET (no air sea coupling in the Tropics) and (c), (d) PC-G (no coupling over the entire global oceans). action in the North Pacific. However, the tropical teleconnective forcing can enhance the amplitudes and affect the temporal evolution of the ENPM substantially. Next, we will study whether the ENPM arises from local ocean atmosphere coupling and/or atmospheric stochastic forcing. b. The role of local ocean atmosphere coupling Previous studies have shown diverse results of the relative effects of local ocean atmosphere coupling and stochastic forcing on the North Pacific decadal variability. Some studies suggest that atmospheric forcing with spatial coherence can generate an oceanic variability mode with a preferred spatial and temporal scale (Saravanan and McWilliams 1997; Jin 1997; Frankignoul et al. 1997; Neelin and Weng 1999); others suggest that the dynamical feedback between the ocean and the atmosphere in the North Pacific is crucial for the generation of North Pacific decadal variability (e.g., Latif and Barnett 1994; Venzke et al. 2001). Our recent CMS studies suggested that the atmospheric stochastic processes can drive a weak NPM-like SST mode, but without any preferred timescale. In contrast, the coupled ocean atmosphere feedback play a critical role in the full development of the multidecadal NPM (LW). Here we will further assess the role of local ocean atmosphere coupling in the generation of the ENPM. To address this issue, we analyze the PC-ET experiment in comparison with another PC experiment, PC-G, in which air sea coupling is shut off over the entire globe by prescribing the model climatological annual cycle of global SST over the entire global ocean for the atmosphere. Without air sea coupling in the North Pacific (in PC- G), both the ENPM and NPM remain robust. The ENPM and NPM correspond to the two leading REOFs of the North Pacific and explain an equal amount (15%) of the field variance (Figs. 11c,d). The pattern correlation of the ENPM and the NPM between PC-ET and PC-G

13 3123 FIG. 12. Spatial patterns of atmospheric 500-mb geopotential height anomalies associated with the (a), (c) ENPM and (b), (d) the NPM in the experiments: (a), (b) PC-ET and (c), (d) PC-G. exceeds 0.9. The pattern similarity also appears in the atmosphere. The 500-hPa height anomalies associated with the ENPM and the NPM in PC-G bear a strong similarity to those in PC-ET as well as in CTRL (Figs. 12c,d). In spite of the pattern similarity, the amplitudes of these two SST modes are reduced significantly due to the lack of ocean atmosphere coupling (Figs. 11c,d). For the ENPM, the most significant reduction occurs along the North America coast, where the amplitude is reduced by over 50% compared to PC-ET, while in the central North Pacific, the amplitude remains the same as PC-ET (Fig. 11c versus Fig. 11a). In comparison, the amplitude of the NPM is reduced significantly in the Kuroshio Extension, but somewhat enhanced along the Alaska coast (Fig. 11d versus Fig. 11b). The weakening of the variability in the absence of the local ocean atmosphere coupling is also seen in the atmosphere, where the 500-hPa height anomalies associated with each mode reveal some modest reduction (20% 30%) in the vicinity of the Aleutian low (Figs. 12c,d versus Figs. 12a,b). In spite of reduction of the amplitudes of the ENPM, the temporal behavior of the ENPM in PC-G does not differ significantly from that in PC-ET (Fig. 13c versus Fig. 13a). Without ocean atmosphere coupling, the ENPM still shows substantial variability around yr. In contrast, without ocean atmosphere coupling, the NPM does not reveal any significant multidecadal oscillation, in sharp contrast to that with active coupling (Fig. 13d versus Fig. 13b). This indicates that the ocean

14 3124 JOURNAL OF CLIMATE VOLUME 16 FIG. 13. Power spectrum of unfiltered time series associated with the (a), (c) ENPM and (b), (d) NPM in the experiments: (a), (b) PC-ET and (c), (d) PC-G. The power spectrum is calculated using a multitaper spectrum method with 50% (lower) and 95% (upper) confidence levels indicated. atmosphere coupling is crucial for the full development of the NPM, but is not critical for the ENPM although it can enhance the amplitudes of the ENPM regionally in the eastern subtropical Pacific. In contrast to SST, the power spectrum of atmospheric variability (500-hPa height) associated with the ENPM is essentially white at decadal timescales even in the presence of air sea coupling (Figs. 14a,c). For the NPM, however, the power spectrum of 500-hPa height also shows a peak around yr (which is somewhat shorter than SST) in the presence of air sea coupling, but no peaks at all without air sea coupling. c. Mechanisms of the ENPM The partial coupling experiments suggest that the origin of the ENPM can be attributed to atmospheric internal variability forcing. In this section, we will further study the mechanisms of the generation and evolution of the ENPM. 1) UPPER-OCEAN HEAT BUDGET First, we regress the low-passed surface wind stress and surface turbulent heat flux (latent plus sensible) against the normalized temporal coefficient of the ENPM in CTRL, PC-ET, and PC-G (Fig. 15). The wind stress pattern associated with the ENPM in these three runs is commonly characterized by a cyclonic circulation south of the Gulf of Alaska, anomalous southwesterly trades in the eastern subtropical North Pacific, and northward winds along the coast of the North America. The heat flux tends to enhance SST anomalies in the lower latitude (south of 35 N) in the region of reduced northeasterly trades, but to damp SST anomalies in the central North Pacific. To further investigate the mechanisms of the generation of SST in different regions, we analyze the heat budget of the upper 50-m water column in PC-ET. We will analyze the various terms in the temperature equation: the net surface heat flux from the atmosphere into the ocean, the advection of the mean temperature gradient by anomalous ocean currents and the advection of temperature anomalies by ocean currents (total six terms in zonal, meridional, and vertical directions), and dissipation including diffusion and vertical mixing. These terms are then regressed against the normalized temporal coefficient of the ENPM (Fig. 16). The regression of the net downward surface heat flux (Fig. 16a) is virtually the same as that of the turbulent surface heat flux (Fig. 15b), implying

15 3125 FIG. 14. Power spectrum of unfiltered atmospheric Z500 height associated with the (a), (c) ENPM and (b), (d) NPM in the experiments: (a), (b) PC-ET and (c), (d) PC-G. The power spectrum is calculated using a multitaper spectrum method with 50% (lower) and 95% (upper) confidence levels indicated. a less important role for the radiative forcing. The heat budget analysis reveals different forcing mechanisms in different regions. We will focus on the two action centers of the ENPM: the eastern North Pacific along the coast of North America (centered at 25 N, 130 W) and the central North Pacific (centered at 35 N, 160 W). In the central North Pacific, cooling is associated predominantly with the advection of the mean temperature gradient by anomalous Ekman flow (Fig. 16b). A strong westerly brings colder subpolar water into this region cooling the surface water. This cooling is damped by the surface heat flux and dissipation (Figs. 16a,d). In contrast, in the eastern North Pacific, warming is forced predominantly by the surface heat flux (Fig. 16a), and damped primarily by the vertical advection by anomalous Ekamn upwelling (Fig. 16c) in response to the local anomalous positive wind stress curl. It is noted that the vertical anomalous advection also makes a remarkable contribution to the cooling in the central North Pacific, implying the potential effect of the thermocline variations on the SST (Fig. 16c). Other terms including the zonal advection and meridional and vertical mean advection are negligible and are not shown in Fig. 16. The heat budget of the upper ocean in PC-G is similar to that in PC-ET (not shown). It is noted that SST in the eastern North Pacific (off California) is driven by surface heat flux and is consistent with a recent modeling study by Di Lorenzo et al. (2003, manuscript submitted to J. Phys. Oceanogr.), who show that the longterm warming (since 1976) of the California Current system is driven by an anomalous surface turbulent heat flux and is cooled by increased upwelling favorable winds and nearshore Ekman upwelling. 2) MECHANISM OF THE ENPM: STOCHASTIC RESONANCE Our partial coupling experiments show that the origins of the ENPM can be attributed to atmospheric stochastic forcing. Some previous studies using simple stochastic models suggested that atmospheric stochastic processes alone can give rise to oceanic variability of preferred timescales if certain conditions are satisfied. For example, Saravanan and McWilliams (1997) proposed a stochastically advective resonance mechanism, which occurs if the white atmospheric forcing has a coherent pattern that changes sign along the propagation of either oceanic currents (Saravanan and McWilliams

16 3126 JOURNAL OF CLIMATE VOLUME 16 FIG. 15. Regression of low-passed wind stress, surface heat flux (latent and sensible, downward is positive) on the normalized temporal coefficient of the ENPM in (a) CTRL, (b) PC-ET, and (c) PC-G. The units are W m 2 and N m 2 for heat flux (heavy lines) and wind stress (arrows); respectively. SSTAs are shaded and contoured with light lines. Contour intervals are 0.08 C and 0.6 W m 2 for SST and heat flux, respectively. 1997) or oceanic waves (Jin 1997; Frankignoul et al. 1997; Neelin and Weng 1999). The conditions for stochastic resonance seems to be satisfied in our model for the ENPM. For the ENPM, the power spectrum of Z500 height is essentially white at decadal timescales (Figs. 14a,c). To demonstrate the propagation feature of the ENPM, we show a longitude lag plot of regression coefficients (on the ENPM) of SST, the upper 400-m ocean heat content (OHC), as well as wind stress and its curl averaged within N (Fig. 17). Both SST and OHC anomalies display a westward propagation. This slow propagation (about 30 within 15 yr) tends to be consistent with the speed of the planetary wave at that latitude, although the wave speed is somewhat slower than the real ocean due to a diffusive thermocline in the model. Along the propagation direction, both SST and

17 3127 having the same positive (negative) sign (Fig. 17b). Similarly, a cold (warm) SST anomaly, as it propagates from the eastern North Pacific to the central Pacific, will be intensified by the anomalous westerlies (easterlies), which can bring cold (warm) subpolar (subtropical) water to enhance it. It should be noted that the amplification of SST along the propagation direction is due to the oceanic process (surface Ekman advection), in contrast to previous studies (e.g., Saravanan and McWilliams 1997) where the heat flux controls SST anomalies. A similar propagation is seen for the ENPM in PC-G (not shown). Given the zonal length scale of the atmospheric forcing (L) and the westward propagation speed of the planetary wave (C R ), the period of the ENPM can be estimated as L/C R. In the model, L is about 6000 km, and C R about 180 km yr 1 (at 40 N), thus the period of the ENPM is about 33 yr, consistent with the spectrum analyses (Fig. 13a). In contrast to the ENPM, there is no such advective resonance for the NPM. Our previous studies (Wu et al. 2003) show that SST anomalies in the Kuroshio Extension are predominantly associated with oceanic anomalous meridional advection. Although SST shows a slow eastward-propagation tendency, the oceanic meridional advection, however, has virtually the same sign in the midlatitude. This is not consistent with the stochastic resonance mechanism. FIG. 16. Regression of upper-ocean 50-m heat budget terms (heavy lines) on the normalized temporal coefficient of the ENPM in PC- ET: (a) net surface heat flux (downward is positive), (b) anomalous meridional and (c) vertical advection of temperature gradient, and (d) convection and diffusion. SSTAs are shaded and contoured with light lines. Contour intervals are 0.06 C and 0.5 W m 2 for SSTs and heat budget terms, respectively. oceanic heat content tend to be amplified by local wind via surface anomalous Ekman advection and wind stress curl, respectively. For example, as a negative (positive) OHC anomaly propagates from the eastern North Pacific ( W) to the central Pacific ( W), it experiences the anomalous wind stress curl virtually 5. Summary and discussion Decadal variability in the North Pacific is studied in a series of coupled global ocean atmosphere simulations using a modeling surgery approach. Both modeling and observational studies suggest two distinctive internal modes in the North Pacific: the North Pacific mode (NPM) and the eastern North Pacific mode (ENPM). The simulated ENPM originates from atmospheric stochastic processes, and is regionally enhanced by local ocean atmosphere coupling in the eastern subtropical North Pacific. The upper-ocean heat budget reveals that the warming (cooling) in the central North Pacific and cooling (warming) in the eastern North Pacific are predominantly associated with the anomalous Ekman advection and surface heat flux, respectively. In contrast to the ENPM, the NPM critically depends on local ocean atmosphere coupled feedback, although the atmospheric stochastic forcing can generate a NPM-like mode with much reduced amplitudes and no preferred timescale. Our coupled GCM studies indicate that the mechanism of the ENPM seems to be consistent with the stochastic resonance, as suggested by some previous stochastic models. The atmospheric spatial pattern sets the length scale of large-scale wave motion in the ocean. This wave propagates to the west due to oceanic planetary waves dynamics and affects SST by the changing

18 3128 JOURNAL OF CLIMATE VOLUME 16 FIG. 17. Lon lag plot of regression of (a) SST (shaded) and wind stress (arrows), and (b) upper-ocean 400-m heat content (light lines) and wind stress curl (heavy lines) on the normalized temporal coefficient of the ENPM in PC-ET. All variables are averaged over the latitudinal band N. Units are N m 2 for wind stress, m C for heat content, and N m 3 ( 10 8 ) for wind stress curl, respectively. Contour intervals are 0.02, 1.0, and 0.1 for SST, heat content, and wind stress curl, respectively.

19 3129 FIG. 18. SST std dev at (a) (d) interannual and (e) (h) decadal timescales in CTRL, PB-ET, PBC-ET, and PC-ET experiments. Contour unit is C. of vertical advection. Atmospheric forcing virtually acts to amplify the oceanic response along the direction of the propagation and gives rise to a limited resonance. The period of the ENPM is determined by the zonal length scale of atmospheric forcing and oceanic planetary wave dynamics. Our partial coupling experiments also show that the tropical teleconnective forcing can enhance the ENPM

20 3130 JOURNAL OF CLIMATE VOLUME 16 substantially although it is not a necessary precondition. Without tropical atmospheric teleconnective forcing, SST variance is reduced by 30% in the central Pacific. In addition, the temporal evolution of the ENPM also undergoes a substantial shift. Without tropical teleconnective forcing, the decadal peak (around 11 yr) seen in CTRL disappears, and substantial variability occurs around yr although detail processes remain to be analyzed in a future study. In addition to the atmospheric teleconnection, the oceanic waveguide along the eastern boundary can support Kelvin wave like waves that transmit the tropical thermocline variability to the midlatitudes. These waves can radiate baroclinic Rossby waves into the interior (e.g., Jacobs et al. 1994). The decadal tropical thermocline fluctuations may impose a boundary condition on the extratropical thermocline and could affect the extratropical decadal variations via radiation of planetary waves (e.g., Clarke and Lebedev 1999; Liu 2002). To assess the role of the tropical oceanic teleconnection in the midlatitude, we performed two additional experiments using the partial blocking (PB) approach. The experiments, denoted as PB-ET and PBC-ET, are configured the same as CTRL and PC-ET, respectively, except a sponge wall is placed in the latitudinal band (15 25 ) in both hemispheres. Both experiments are integrated for 150 yr. The effect of the tropical oceanic teleconnection on the midlatitudes can be quantified by comparing CTRL, PC-ET, PB-ET, and PBC-ET. Figure 18 plots SST standard deviation at interannual timescales ( 8 yr, Figs. 18a d) and decadal timescales (8 100 yr, Figs. 18e h) in these four experiments. It can be seen that without tropical atmospheric teleconnective forcing (PC-ET versus CTRL and PB-ET versus PBC-ET), SST interannual variability is reduced substantially by about 30% 40% in the central North Pacific (Fig. 18d versus Fig. 18a, and Fig. 18c versus Fig. 18b). In contrast, SST decadal variability is substantially reduced in the eastern North Pacific (the ENPM) while remains the same magnitude in the western North Pacific (the NPM) (Fig. 18h versus Fig. 18e and Fig. 18g versus Fig. 18f). A closure of tropical extratropical oceanic pathway (PB-ET versus CTRL, PBC-ET versus PC-ET) will substantially reduce SST decadal variability in the region of the Kuroshio Extension (the NPM) by about 40% (Fig. 18f versus Fig. 18e and Fig. 18g versus Fig. 18h), but less significantly for interannual variability in the region of the Kuroshio Extension and the eastern North Pacific (Fig. 18b versus Fig. 18a and Fig. 18c versus Fig. 18d). Given the fact that the eastern North Pacific is less affected by tropical oceanic teleconnection shown in the model, we argue that the reduction of SST decadal variability in the region of the Kuroshio Extension is unlikely caused by the elimination of tropical oceanic teleconnection, which is expected to have substantial impact on the variability in the eastern North Pacific by imposing forcing along the eastern boundary. The reduction of SST decadal variability in the region of the Kuroshio Extension may be attributed to the artificial elimination of thermocline adjustment in the subtropics (15 25 N) due to the imposed sponge wall, which may affect the thermocline adjustment in the Kuroshio Extension region, and thus affect the SST by changing the vertical mixing (Schneider et al. 2002). We conclude that the major influence of the Tropics in the midlatitudes is dominated by atmospheric teleconnection, and concentrated in the central and eastern North Pacific. Our modeling studies suggest that the decadal variability in the North Pacific is a complex phenomena that may involve multiple forcing mechanisms. Further studies are clearly needed to understand the mechanisms of the NPM and ENPM, particularly similar sensitivity experiments from other coupled GCMs. Acknowledgments. This work is supported by NASA, NOAA, and DOE. We thank Drs. R. Jacob and R. Gallimore for helping in setting up the GCM experiments, and P. Behling for preparing the figures. Comments from two reviewers helped to improve the paper. The computer time allocations from NCSA are appreciated. REFERENCES Barlow, M., S. Nigam, and E. H. Berbery, 2001: ENSO, Pacific decadal variability, and U.S. summertime precipitation, drought, and streamflow. J. Climate, 14, Barnett, T. P., D. W. Pierce, M. Latif, D. Dommenget, and R. Saravanan, 1999a: Interdecadal interactions between the Tropics and midlatitudes in the Pacific basin. Geophys. Res. Lett., 26, ,, R. Saravanan, N. Schneider, D. Dommenget, and M. Latif, 1999b: Origins of midlatitude Pacific decadal variability. Geophys. Res. Lett., 26, Barnston, A. G., and R. E. Livezey, 1987: Classification, seasonality, and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev., 115, Barsugli, J. J., and D. S. Battisti, 1998: The basic effects of atmosphere ocean thermal coupling on midlatitude variability. J. Atmos. Sci., 55, Chang, P., R. Saravanan, L. Ji, and G. C. Hegerl, 2000: The effect of local sea surface temperatures on atmospheric circulation over the tropical Atlantic sector. J. Climate, 13, Clarke, A. J., and A. Lebedev, 1999: Remotely driven decadal and longer changes in the coastal Pacific waters of the Americas. J. Phys. Oceanogr., 29, Deser, C., and M. L. Blackmon, 1995: On the relationship between tropical and North Pacific sea surface temperature variations. J. Climate, 8, Frankignoul, C., P. Muller, and E. Zorita, 1997: A simple model of the decadal response of the ocean to stochastic wind forcing. J. Phys. Oceanogr., 27, Graham, N. E., 1994: Decade-scale climate variability in the tropical and North Pacific during the 1970s and 1980s: Observations and model results. Climate Dyn., 10, Gu, D., and S. G. H. Philander, 1997: Interdecadal climate fluctuations that depend on exchange between the Tropics and extratropics. Science, 275, Horel, J., and J. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Mon. Wea. Rev., 109, Jacob, R. L., 1997: Low frequency variability in a simulated atmo-

21 3131 sphere ocean system. Ph.D. thesis, University of Wisconsin Madison, 155 pp. Jacobs, G. A., H. E. Hurlburt, J. C. Kindle, E. J. Metzger, J. L. Mitchell, W. J. Teague, and A. J. Wallcraft, 1994: Decade-scale trans-pacific propagation and warming effects of an El Niño anomaly. Nature, 370, Jin, F. F., 1997: A theory of interdecadal climate variability of the North Pacific ocean-atmosphere system. J. Climate, 10, Kaplan, A., M. A. Cane, Y. Kushnir, A. C. Clement, M. B. Blumenthal, and B. Rajagopalan, 1998: Analyses of global sea surface temperature J. Geophys. Res., 103, Kleeman, R., J. P. McCreary, and B. A. Klinger, 1999: A mechanism for generating ENSO decadal variability. Geophys. Res. Lett., 26, Latif, M., J. P. McCreary, and T. P. Barnett, 1994: Causes of decadal climate variability over the North Pacific and North America. Science, 266, Lau, N.-C., and M. J. Nath, 1996: The role of the atmospheric bridge in linking tropical Pacific ENSO events to extratropical SST anomalies. J. Climate, 9, Liu, Z., 2002: How long is the memory of the tropical ocean dynamics? J. Climate, 15, , L. Wu, R. Gallimore, and G. Jacob, 2002: Search for the origins of Pacific decadal climate variability. Geophys. Res. Lett., 29, 1404, doi: /2001gl Luo, J., and T. Yamagata, 2002: Four decadal ocean-atmosphere modes in the North Pacific revealed by various analysis methods. J. Oceanogr., 58, Mann, M., and J. Lees, 1996: Robust estimation of background noise and signal detection in climate time series. Climate Change, 33, Mantua, N. J., S. R. Hare, Y. Zhang, J. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, Mestas-Nunez, A. M., and D. B. Enfield, 1999: Rotated global modes of non-enso sea surface temperature variability. J. Climate, 12, Miller, A., and N. Schneider, 2000: Interdecadal climate regime dynamics in the North Pacific Ocean: Theories, observation and ecosystem impacts. Progress in Oceanography, Vol. 27, Pergamon, Nakamura, H., G. Lin, and T. Yamagata, 1997: Decadal climate variability in the North Pacific during recent decades. Bull. Amer. Meteor. Soc., 78, Neelin, J. D., and W. Weng, 1999: Analytical prototypes for ocean atmosphere interaction at midlatitudes. Part I: Coupled feedbacks as a sea surface temperature dependent stochastic process. J. Climate, 12, Nitta, T., and S. Yamada, 1989: Recent warming of tropical sea surface temperature and its relationship to the Northern Hemisphere circulation. J. Meteor. Soc. Japan, 67, Pierce, D., T. Barnett, N. Schneider, R. Saravanan, D. Dommenget, and M. Latif, 2001: The role of ocean dynamics in producing decadal climate variability in the North Pacific. Climate Dyn., 18, Saravanan, R., and J. C. McWilliams, 1997: Stochastically and spatial resonance in interdecadal climate fluctuations. J. Climate, 10, Schneider, N., A. J. Miller, and D. W. Pierce, 2002: Anatomy of North Pacific decadal variability. J. Climate, 15, Venzke, S., M. Munnich, and M. Latif, 2001: On the predictability of decadal changes in the North Pacific. Climate Dyn., 16, Vimont, D., D. Battisti, and A. Hirst, 2001: Footprinting: A seasonal connection between the midlatitudes and tropics. Geophys. Res. Lett., 28, Wu, L., Z. Liu, R. Gallimore, R. Jacob, D. Lee, and Y. Zhong, 2003: Pacific decadal variability: The tropical Pacific mode and the North Pacific mode. J. Climate, 16, Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: J. Climate, 10,