1. Introduction
The North Atlantic Oscillation (NAO) and the Pacific–North American (PNA) teleconnection patterns are the most prominent modes that dominate the climate variability over the Northern Hemispheric midlatitudes on monthly to interannual time scales (Wallace and Gutzler 1981; Kushnir and Wallace 1989; Hurrell 1995). Both patterns are the result of internal variability of midlatitude dynamics, as well as being externally driven (Saravanan 1998; Feldstein 2000; Hoerling et al. 2001, 2004; Hurrell et al. 2003). However, the external forcing that is responsible for the variability of the two modes, especially the NAO, is still an open question. So far, the tropical sea surface temperatures (SSTs) have been identified as a source of the low-frequency variability for both teleconnection patterns, but it is unclear which tropical regions are most important. Here we examine which ocean regions strongly influence the two teleconnection patterns, with the ultimate goal to improve the prediction skill of the NAO and PNA.
Regarding the oceanic influence on NAO and PNA variability, the dominant ocean basins may differ depending on time scales. For example, Hoerling et al. (2001), Bader and Latif (2003), and Hoerling et al. (2004) stressed the importance of the Indian Ocean on decadal variability of the NAO using an atmospheric general circulation model (AGCM). Lin et al. (2005) underscored the contribution of the tropical Pacific on the interannual variability of the NAO and PNA patterns using a coupled atmosphere–ocean general circulation model (AOGCM). In this paper, we focus on the sensitivity of winter season NAO and PNA indices to the SST anomalies and specifically on two questions: 1) Where are the ocean forcings most effective for generating the winter season NAO and PNA patterns? and 2) Is it possible to reproduce the NAO and PNA variability using only spatiotemporal variability in tropical SST anomalies?
It is well known that the variability of PNA can arise through atmospheric Rossby waves that are forced by the tropical convection associated with SST anomalies (Trenberth et al. 1998). While this chain of events is straightforward, it does not preclude other dynamical pathways linking the tropics and midlatitudes. The internal variability of the NAO involves eddy-mean flow interactions in the midlatitude jet stream (which may be linearly parameterized in a statistical sense; Whitaker and Sardeshmukh 1998), and the NAO is more remote from the strong forcing regions of the tropical Pacific, implying more opportunity for interactions and indicating a more nonlinear relationship between the NAO and the ocean forcing compared to the PNA.
We use a statistical method based on empirical Green’s functions (described in section 2) to explore the sensitivity of NAO and PNA patterns to patterns of SST anomalies. A similar method has been employed to investigate the sensitivity of the large-scale response to tropical SST anomalies and predictable skills of regional climate change (Barsugli and Sardeshmukh 2002; Barsugli et al. 2006; Li et al. 2012). The linear reconstruction of the NAO and PNA patterns and their variability are presented in section 3. We summarize and discuss the results in section 4 focusing on the relative importance of the tropical ocean basins for driving the variability of the NAO and PNA.
2. Data and methodology
a. Estimating the linear global teleconnection operator
To obtain the linear operator G, we used a random perturbation method (RPM) to generate an ensemble of SST perturbation fields as described in Li et al. (2012). To ensure an SST perturbation is strong enough to result in a significant response yet weak enough to avoid undesired climate noise (Barsugli et al. 2006), we designed each random perturbation field to have a dimension of 16 × 16 in the global domain (i.e., 22.5° × 11.5°) with magnitudes between −2° and 2°C, which follows the uniform distribution. In total, we created a 200-member independent sample set of the SST perturbation fields and added each to the climatological (a 12-month cycle) SST field. We then forced the model with the perturbed climatological SST field and integrated the model for 20 months for each ensemble member. Each perturbed run uses a different initial condition (each initial condition is for 1 January) generated from a long-term control run forced by the climatological SST. From the 20-month simulation, we calculated the seasonal output from the last 12 months (months 9–20) for the analyses.
In this work, we applied this method to test the sensitivity of the two atmospheric teleconnection patterns (i.e., NAO and PNA) to SST anomalies, and therefore, ΔR in Eqs. (1)–(3) refers to the NAO or PNA index.
b. Models and data
We primarily use two AGCMs in this study: the Community Atmospheric Model, version 5.0 (CAM5.0), developed at the National Center for Atmospheric Research (NCAR; Neale et al. 2010a,b) and the Geophysical Fluid Dynamics Laboratory (GFDL) Atmospheric Model, version 2 (AM2; Anderson et al. 2004). To highlight the model structural differences, we used similar horizontal resolutions for the two models (finite volume 1.9° × 2.5° for CAM5 and 2° × 2.5° for GFDL). The climatological SST data used for the control runs are different in the models. CAM5 uses the merged Hadley Centre SST and National Oceanic and Atmospheric Administration (NOAA) optimal interpolation analysis (HadOIB1 SST), while GFDL uses Reynolds Optimum Interpolation (OI) SST as boundary conditions. To ensure that the sensitivity information obtained from different models is comparable with each other, we applied the identical SST data used in CAM5 to GFDL in the perturbed SST simulations.
Additionally, we use the NCAR CAM3.1 at T42 resolution (~2.8° × 2.8°) to validate the RPM method. For this model version, in addition to the perturbed SST runs, we also performed an ensemble experiment forced with climatological SST plus single patch SST forcing over different ocean basins. This type of experiment is called the “patch method,” and detailed information concerning the experimental design was documented in Barsugli and Sardeshmukh (2002), Barsugli et al. (2006), and Li et al. (2012). Because the spatial dimensions of perturbed SST anomalies in the RPM and patch method are similar, the availability of these two types of experiments allow us to verify the linear model described by Eq. (3) in reproducing the teleconnection patterns. The results are presented in section 3.
We also used two additional ensembles to examine the variability of NAO and PNA. In the first ensemble, we generated a 15-member ensemble of CAM5 AMIP simulations. Each realization was forced by time-varying SST fields from January 1949 to December 2005. We selected the result during 1950–2000 and calculated the December–February (DJF) mean for our analysis. In the second ensemble, we used sea level pressure and geopotential height anomaly during January 1950 to December 2000 from CESM historical simulations (Taylor et al. 2012) to calculate the NAO and PNA indices. Only three realizations were available for Community Earth System Model (CESM) historical runs. As described in the Network Common Data Form (netCDF) data for each model, each realization is a fully coupled 1850–2000 transient run with forcing of the solar irradiance, greenhouse gases, volcanic aerosol, sea salt, dust, anthropogenic aerosol, black carbon, mineral dust, organic carbon, ozone, and land use change included.
The data we used to calculate the NAO and PNA indices are the monthly geopotential height and sea level pressure with a resolution of 2.5° × 2.5° for the period of January 1950 to December 2000 from the National Centers for Environmental Prediction (NCEP)–NCAR Reanalysis-1 (Kalnay et al. 1996). The monthly data are averaged to calculate the seasonal data, and we use the DJF mean data to calculate the NAO and PNA indices, as we did for the model output.
The SST data that we used for the linear construction of the time-varying NAO and PNA indices is the HadOIB1 SST data during January 1950 to December 2000.
c. NAO and PNA indices
Traditionally, two ways are used to define the NAO and PNA indices: principal component (PC)-based and station-based definition. PC-based NAO and PNA are usually defined by the first two leading modes of a principal component analysis applied to the normalized 500 height anomaly over 20°–90° of the Northern Hemisphere. This definition is used at the Climate Prediction Center (CPC) of NOAA/National Weather Service (NWS)/NCEP and in many additional studies (Cohen and Barlow 2005; Johansson 2007; Cattiaux et al. 2010; Seager et al. 2010). This PC-based definition of the NAO indices could also be applied to geopotential height of the 300-hPa surface considering the representativeness of transient eddy vorticity fluxes at upper troposphere, but the result was in good agreement with the definition from the CPC (Feldstein 2000, 2003).
PC-based and station-based definitions each have their advantages and limitations. The PC-based definition is able to optimally represent the spatial pattern of the teleconnections yet could be dependent on source data (https://climatedataguide.ucar.edu/climate-data/hurrell-north-atlantic-oscillation-nao-index-pc-based; Osborn 2004). The station-based definition uses fixed spatial grids and is thus less dependent on the source data, yet it is subject to more noise because of the limitation in collecting the data. To investigate the applicability of the two definitions in this work, we first examined the NAO and PNA pattern represented in the perturbed SST runs (Figs. 1, 2). The pattern correlation between the PC [calculated based on Barnston and Livezey (1987)] and station-based NAO patterns for CAM5 and GFDL is 0.84 and 0.92 respectively. The corresponding correlation for the PNA pattern is 0.85 and 0.8; all passed the null hypothesis test at 99% confidence level. As such, the station-based teleconnection is consistent with the PC-based definition in our models. For the NAO pattern, both models produce a strong signal in the Pacific basin that is not present in the observational data and that was also found by other work (Peng et al. 2002). Considering the PC-based indices can be model dependent because of the internal and SST-forced variability, this leads to optimal patterns in one model and less optimal patterns in the other (Osborn 2004). To avoid the potential inconsistency between different models and the inconsistency between model and observations, we choose to use the station-based definition for all models and observations.
Station-based and PC-based NAO patterns. The PC-based NAO pattern is defined as the rotated first empirical orthogonal function (EOF1) of 500-hPa height anomaly of 20°–90°N from the 200 perturbed SST runs. The EOF1 explains 20.1% and 28.6% of the total variance in the CAM5 and GFDL models, respectively.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
As in Fig. 1, but for the PNA pattern. The PC-based PNA pattern is defined as the rotated EOF2 of the H500 anomaly of 20°–90°N from 200 RPM runs for CAM5 and EOF3 for GFDL. EOF2 explains 14.5% of the total variance in CAM5 and 8.8% of the total variance in GFDL.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
To be consistent with the indices calculated from model data, we calculated the NAO and PNA indices from the NCEP reanalysis data instead of using the NAO and PNA indices based on direct observations. To verify our calculation, the calculated indices from the NCEP reanalysis data are compared with the indices from two other datasets [seasonal North Atlantic Oscillation index (station based) at https://climatedataguide.ucar.edu/climate-data/hurrell-north-atlantic-oscillation-nao-index-station-based and http://jisao.washington.edu/data/pna/]. The correlations between the two datasets are 0.95 for NAO and 0.99 for PNA index.
3. Results
a. Sensitivity of the NAO and PNA indices to SST anomaly patterns
The sensitivity maps of the wintertime NAO and PNA indices to SST anomalies from different models (Fig. 3) show the regions where SST anomalies are most effective at forcing the teleconnection patterns, either in a positive or negative sense. Overall, the NAO index is strongly sensitive to the SST anomalies over the tropical Indian Ocean. The NAO index is also sensitive to the SST anomalies over the tropical Atlantic and tropical central Pacific with a weaker response and opposite sign. The PNA index is strongly sensitive to the SST anomalies over the tropical central Pacific, especially the Niño-4 region. The “nodal line” over the warm pool with opposite sign sensitivity is consistent with previous results (Branstator 1985; Kumar and Hoerling 1997; Sardeshmukh and Hoskins 1988; Barsugli and Sardeshmukh 2002). The PNA index is also sensitive to the SST anomaly over the tropical Indian Ocean and tropical and North Atlantic Ocean with opposite sign, which is consistent with Annamalai et al. (2007), who suggested that El Niño–induced SST anomalies over the tropical Indian Ocean have an impact on the PNA pattern.
Sensitivity of the wintertime (DJF) NAO and PNA indices to the SST anomaly from CAM5 and GFDL AM2 given by G(x′) in Eq. (1). For each model, 200 ensembles are used to obtain the sensitivity map. Shaded regions refer to where the sensitivity is significant using a two-tailed t test at 5% significance level. [Units of the sensitivity map of NAO and PNA indices are (1× 105 km2 K)−1].
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
Because uncertainties exist when using the RPM method to estimate the sensitivity maps, a significant concern is whether the results are sensitive to the ensemble size. To address this question, we compared the sensitivity maps from 200 ensemble members to that from 500 members for both models (not shown). There is no significant change of the result (pattern correlation between 200 and 500 ensemble members is 0.98 for both NAO and PNA indices in CAM5. Pattern correlation between 200 and 500 ensemble members is 0.99 for NAO and 0.98 for PNA in GFDL). As such, we considered 200 ensemble members to be sufficient to represent the sensitivity information for the teleconnection patterns.
b. SST-forced NAO and PNA patterns
The sensitivity patterns show the regions where SST forcing perturbations can be most effective. It is theoretically possible to have a very sensitive region with little or no forcing amplitude on the time scales of interest. We examine whether the actual SST forcing in the sensitive regions is sufficient to reproduce the NAO and PNA patterns.
To further test the sensitive forcing regions for the two teleconnection modes, we reconstruct the positive phase of the two teleconnection patterns by multiplying the sensitivity map with 1°C SST anomaly over the Indian Ocean and Niño-4 region for the NAO and PNA patterns, respectively [i.e., ΔSST(i′, j′, t) = 1 in Eq. (3)]. This is motivated by Hoerling et al. (2004), who tested the equilibrium response to 1°C Indian Ocean SST anomaly (approximately the magnitude of SST trend during 1950–99) by AGCMs and found that the NAO is reasonably reproduced. As we mentioned in section 2, in reconstructing the NAO pattern, we recalculated the sensitivity map of the NAO index using 500-hPa height data. In addition, before using CAM5.0, we tested the linearity and accuracy of the RPM method by comparing the reconstructed patterns from RPM with the result from AGCM run forced by a same localized SST (Fig. 4).
Comparison of the linear reconstruction of (left) 500-hPa geopotential height anomaly with (right) the modeling result. The linear reconstruction (i.e., RPM) is calculated by multiplying the sensitivity map by 1°C SST anomaly over the tropical Indian Ocean (10°S–10°N, 50°–100°E) and Niño-4 (5°S–5°N, 160°E–150°W) regions. The model simulation (i.e., patch method) of the 500-hPa anomaly is obtained by forcing the model with climatological SST plus SST anomaly patch over the Indian Ocean and Niño-4 region. The model used for this comparison is CAM3.1 T42, and the unit is hectometers.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
The pattern correlation between the reconstruction and model simulation is 0.82 for the NAO pattern and 0.68 for the PNA pattern. Both pass the significance test at the 1% level, indicating that the result from the RPM is in good agreement with the model simulation. This suggests that there is a linear component in the atmospheric response to tropical SST forcing and that the RPM works well for that purpose. We also compared our reconstructed result with the Community Climate Model, version 3 (CCM3) result in Hoerling et al. (2004, their Fig. 7) and found our reconstructed 500-hPa height anomaly is consistent with their result from the AGCM simulation.
Overall, the linear reconstruction of the 500-hPa height anomaly by SST anomaly over the Indian Ocean resembles the NAO patterns over the Arctic–Atlantic sector (Fig. 5). The reconstructed PNA pattern is more consistent between the two AGCMs given that the pattern correlation between the two models is 0.88, which is greater than the case of the NAO pattern (pattern correlation is 0.81). Based on Figs. 1 and 2, this difference suggests that the NAO pattern is more sensitive to the model difference than the PNA pattern.
Linear reconstruction of the 500-hPa geopotential height anomaly from CAM5 and GFDL AM2. The linear reconstruction is calculated by multiplying the sensitivity map [i.e., G(x′)] by 1°C SST anomaly over the tropical Indian Ocean and Niño-4 regions. The Indian Ocean is defined as (10°S–10°N, 50°–100°E) following Hoerling et al. (2004); the Niño-4 region is defined as (5°S–5°N, 160°E–150°W).
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
The NAO and PNA patterns are highly correlated in our result. This is caused by the ambiguity of the two patterns or ambiguity of the NAO and Artic Oscillation (AO) represented in the AGCMs (Figs. 1, 2), in which strong response over the North Pacific basin exists in both NAO and PNA patterns. This is a limitation in AGCMs in reproducing the NAO pattern precisely.
c. SST-forced NAO and PNA variability
The oceanic contribution to variability of the NAO is less clear compared to that for the PNA. While the internal noise is identified to be one possible mechanism (Saravanan 1998; Feldstein 2000), SST variability over the tropical Indian Ocean and North Atlantic Ocean is also found to contribute to the low-frequency variability of NAO (Deser and Blackmon 1993; Sutton and Allen 1997; Rodwell et al. 1999; Hoerling et al. 2001; Bader and Latif 2003). While we place emphasis on estimating which ocean basins are driving the response of the NAO and PNA patterns in this paper, we discuss how it is possible to explain a sizable fraction of the NAO and PNA variability using only SST patterns in this section. To examine the SST-forced variability of the NAO and PNA, we reconstructed the time series of the NAO and PNA indices by multiplying the sensitivity map with time-variant SST and compared it with the time series calculated from reanalysis data. To highlight the contribution of the ocean basins to which the NAO and PNA indices are sensitive, we performed the reconstruction using SST anomaly over only the tropical Indian Ocean, central Pacific, and tropical Atlantic instead of the whole tropics. As shown in Fig. 6, the time series of the NAO index is reconstructed using Eq. (3) and combined the SST anomalies over the tropical Indian Ocean (10°S–10°N,50°–100°E) plus the Niño-4 region (5°S–5°N, 160°E–150°W) plus the Caribbean Sea (5°S–5°N, 50°–20°W). The time series of the PNA index is reconstructed using SST anomalies only over the Niño-4 region. To demonstrate the impact of tropical SST on the variability of NAO and PNA indices on different time scales, we applied a 5-yr running mean on the time series to isolate the variability on longer time scales (decadal and longer scales). We calculated the correlation between the reconstructed and observed time series for both raw data (i.e., without applying a 5-yr running mean) and smoothed data (with the 5-yr running mean) for further analysis.
Time series of NAO and PNA indices calculated from linear reconstruction and reanalysis data during 1950–2000. The time series of the reconstructed NAO index is obtained using Eq. (3) by multiplying the sensitivity map with the 2D SST anomaly over tropical Indian Ocean (10°S–10°N, 50°–100°E) plus SST anomaly over the Niño-4 region (5°S–5°N, 160°E–150°W) plus SST anomaly over the Caribbean Sea (5°S–5°N, 50°–20°W). The time series of the PNA index is reconstructed by multiplying the sensitivity map with the 2D SST anomaly over the Niño-4 region. Black and red lines denote the time series of raw data and the five-point running average, respectively. Each time series has been normalized by the standard deviation of the time series. The correlation coefficients between the time series of reconstruction and reanalysis data are indicated.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
The correlation between the time series of the reconstruction and the NCEP reanalysis for the NAO index is 0.41 and 0.34 for CAM5 and GFDL AM2, respectively, for raw data. The correlation between the two time series of the PNA index is 0.42 and 0.43 for the two models, respectively, but decreases about 26% after the smoothing. We note that our NAO index is yearly DJF data. Different from the PNA index, the time series of NAO is highly autocorrelated, especially in the reconstructed data, and thus largely reduces the degrees of freedom. Taking this into account, the correlation of 0.41 for the NAO index passes the significance test at the 10% level based on Livezey and Chen (1983). Also, for the NAO time series, the linear trend (due to the Indian Ocean SST) accounts for 35% and 56% of the correlation for CAM5 and GFDL reconstructions, respectively, indicating a contribution of the Indian Ocean to the total variability of the NAO index.
The correlation between the two time series from the reconstruction using SST anomalies from other basins are also calculated to test the relative importance of the tropical basins on the SST-forced variability of the NAO and PNA indices (Tables 1, 2). Based on this result, to get the best SST-forced variability of the NAO index, we should take into account the variability of the tropical Indian Ocean, Niño-4 region, and Caribbean Sea, or at least the tropical Indian Ocean and Niño-4 region (where the result also passed the significance test at the 5% level). The Niño-4 region is sufficient to reproduce the total variability of the PNA index. Multiple papers (Huang et al. 1998; Mokhov and Smirnov 2006; Seager et al. 2010) highlighted the impact of ENSO on variability of the NAO as well as the impact of the tropical Atlantic (Bretherton and Battisti 2000; Okumura et al. 2001). Given these results, we must include the contribution of all three ocean basins, or at least the Indian Ocean and Niño-4 regions simultaneously, when we consider the one-way impact of the SST on NAO variability. The result concerning the SST-forced variability of the PNA index shows inconsistent results for different models. The consistent part from the two models are that the correlation using the Niño-4 SST anomalies is greater than the combination of ocean basins, which indicates a cancellation of effects over the corresponding ocean basins. This result is consistent with what we observe from the sensitivity map and the previous work by Sardeshmukh and Hoskins (1988) and Barsugli and Sardeshmukh (2002).
Correlation coefficients between observed and reconstructed time series for NAO from different ocean basins. The critical value is for the significance test at the 10% level. Bold numbers denote the correlation that passed the significance test. The effective sample size Neff of the NAO time series is based on Livezey and Chen (1983). IND is Indian Ocean and CAR is Caribbean Sea.
Correlation coefficients between observed and reconstructed time series for PNA from different ocean basins. The critical value is for the significance test at the 10% level. Bold numbers denote the correlation that passed the significance test. The effective sample size Neff of the NAO time series is based on Livezey and Chen (1983). IND is Indian Ocean and CAR is Caribbean Sea.
To further explore the contribution of each individual ocean basin to the variability of NAO and PNA on different time scales, we performed a cross-spectral analysis (Pozo-Vázquez et al. 2000; Paeth et al. 2003) on the 51-yr NAO and PNA indices and regional mean SST data over different ocean basins (Fig. 7). The significance level for the null hypothesis of zero population of coherence assumes the time series should be normally distributed (Jenkins and Watts 1968; Thompson 1979). To strictly satisfy this assumption, we prewhitened the data by removing the first order of the red part from the time series being analyzed [i.e., y(t) = x(t) − Ax(t − 1), where A is an adjustment coefficient]. For the period of 1950–2000, the NAO index is coherent with the SST over the tropical Indian Ocean, with the dominant period of 3 years and a less dominant 10-yr period with Indian Ocean SST leading for about a quarter phase. Because of the short data we used in this work, there could be a bias when we applied the cross-spectral analysis. However, the result is consistent with that from other work (Pozo-Vázquez et al. 2000). For both peaks, the Indian Ocean SST leads the NAO index for about a quarter phase, which implies the impact of the Indian Ocean SST on the NAO index. Some less significant coherent frequencies exist between NAO and Niño-4 SST (e.g., peaks at 7, 3, and 2 years), which are generally consistent with the peaks for NAO–Indian Ocean SST. Although these did not reach to the 10% significance level, the coherence spectrum between the PNA and Niño-4 SSTs show signs of the impact of the Niño-4 impact on the PNA variability.
Cross spectrum between the NAO and PNA indices and the SST anomaly over different ocean basins during 1950–2000 (only DJF mean is included in each year). The solid and dashed lines denote the coherence squared and phase spectrum, respectively. Positive (negative) phase angle indicates the NAO leads (lags) the SST. The horizontal solid and long dashed lines in each plot represent the 10% and 5% significance level, respectively, for coherence squared. The unit of the frequency is cycles per year and the bandwidth of all spectral estimates is 0.05 cycles per year.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
For the cross spectrum between the PNA and SST anomaly, the most significant feature is the coherence between PNA and Niño-4 SST at a 4–5-yr period. But, different from the NAO, there is no obvious lead in SST.
To ensure the coherence peaks in the cross-spectral analysis of NAO are significant, we performed a power spectral analysis on the detrended (and the prewhitening) time series of teleconnection indices and SST over different ocean basins (figure not shown). We found that an approximately 3-yr and a less significant 10-yr dominant period in the power spectra of NAO, but the two periods are less significant in the Indian Ocean SSTs. In this case, the coherence between NAO and Indian Ocean SSTs is significant for NAO index but less significant for Indian Ocean SST. The Indian Ocean mostly contributes to the trend of the NAO variability and influences the NAO variability primarily at the 3- and 10-yr periods during subperiods of the study.
The mechanism for the teleconnection impact of Indian Ocean SST on NAO is complex. Czaja and Frankignoul (2002) showed that a “more local” pan-Atlantic SST forcing can lead NAO by up to 6 months through the positive feedback between the heating-forced anomalous flow and transient eddy (Peng et al. 2003). Using a coupled modeling approach, Li et al. (2006) showed that the Atlantic SST tripole is the result of Indian Ocean forcing due to the positive feedback between the NAO and North Atlantic SST. Paeth et al. (2003) suggested the predictability of NAO on decadal time scales from the mean state of the North Atlantic tripole several years (i.e., 2–4 years) in advance. Based on the coherence spectra between the Indian Ocean SST and Caribbean Sea SST (Fig. 7), the impact of Indian Ocean SST on NAO is likely through the influence of Indian Ocean warming on the Caribbean Sea at 2.5–4-yr periods (with Indian Ocean SST leading up to ⅙ of the dominant periods). This result is consistent with the conclusion of Li et al. (2006). A later stage of the causal chain is the impact of the North Atlantic tripole SST pattern on the NAO on short time scales (less than 2-yr period based on Fig. 7) as suggested by Peng et al. (2003).
Another point to consider is the influence of Pacific ENSO on the interannual variability of NAO. Sutton and Hodson (2003) stressed this impact by investigating the oceanic influence on North Atlantic climate for the period of 1871–1999. Based on the significant coherence between PNA and Niño-4 SST at 3.5–5-yr periods and PNA and Caribbean Sea at roughly a 4-yr period, it is also possible that PNA acts as a bridge in the Indo-Pacific impact on NAO, especially during ENSO years. But this has to be proven by a coupled modeling study, which is beyond the ability of our method.
The linear trend of the NAO is also associated with the constant warming of the tropical Indian Ocean (e.g., Hoerling et al. 2001) during the late twentieth century. Because our cross-spectral analysis is performed for the detrended data, the contribution of the trend was not detected, although we have confirmed this from Fig. 5 and other work (Hoerling et al. 2001). The low-frequency variability of the Indian Ocean SST at 3- and 10-yr periods (decadal scale) should also have contribution to the total variability of the NAO index. The Niño-4 shows less significant impact on the variability of the NAO on a similar time scale, while the contribution of the Caribbean Sea on the variability of the NAO seems to be on an interannual scale. Since our linear method cannot completely separate the internal noise from the SST-forced variability, we should also take into account the contribution of the internal noise associated with the SST forcing on the total variability of the NAO and PNA, yet it should be on short time scale (Feldstein 2000).
d. NAO and PNA variability in AMIP and coupled models
We compared the time series of the NAO and PNA indices from AMIP ensembles forced by the same time variant SST (as we used in the linear reconstruction but over global ocean domain) and the time series from the CESM historical runs with the reanalysis data to further test the contribution of each ocean basin to NAO and PNA variability. Because we only have 20 years of data for GFDL AMIP runs, we only use CAM5 AMIP and CESM historical runs to demonstrate the result. Figure 8 shows the time series from the ensemble mean of the 15 CAM5 AMIP runs and three CESM historical runs as compared to that from the observation. For the NAO, although an increase of the correlation for coupled model run is estimated, neither of the correlations are statistically significant. The correlation between the ensemble mean of AMIP increases for the PNA index compared to the linear reconstruction but reduces a lot for the coupled model runs. Given the result of the cross spectrum between the two indices and the SST over the dominant ocean basins (Fig. 9), the year-to-year variability of the wintertime NAO during 1950–2000 is not a result of the global SST forcing, and neither is it a result of an air–sea coupling. The AMIP ensemble of the CAM5 can reproduce the PNA variability well because of the coherence relationship between the simulated PNA index and Niño-4 SST resembles that of the observation.
As in Fig. 6, but for CAM5 AMIP and CESM NAO and PNA indices. The time series for AMIP is the ensemble mean of 15 members. The time series for CESM is the ensemble mean of three CESM historical simulations from CMIP5.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
As in Fig. 7, but for the cross spectrum between (left) NAO and Indian Ocean SST and (right) PNA and Niño-4 SST.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
4. Conclusions and discussions
Knowing that both NAO and PNA variability contain a tropical SST-forced component, we ask, where is the tropical ocean important for driving the winter season variability in the NAO and PNA indices? To answer this, we examined the sensitivity of the NAO and PNA indices to the SST anomalies across the global ocean and then reassess the relative importance of individual tropical ocean basins to explain the variability of the NAO and PNA by a linear statistical method.
This study confirms that the NAO pattern is most sensitive to the SST anomalies over the tropical Indian Ocean. The PNA pattern is most sensitive to the central tropical Pacific Ocean, especially the Niño-4 region. Setting a constant SST anomaly over the Indian Ocean and Niño-4 region the NAO and PNA patterns, respectively, can be well reconstructed. This suggests an efficient SST forcing over Indian Ocean for the NAO pattern and over Niño-4 region for PNA. Based on previous works (Hoerling et al. 2001) and our result, the contribution of the Indian Ocean on NAO is the linear trend and low-frequency variability at a significant 3-yr and a less significant 10-yr time scale. It is likely that the Indian Ocean SST influences the NAO variability at these two time scales through the influence of Indian Ocean warming on the Caribbean Sea. PNA could also act as a bridge in the influence of the NAO influenced by Indian Ocean SST at longer time scales. The contribution of Niño-4 on the decadal variability should also be considered. The linearly reconstructed variability is closer to the observational data than the AMIP and coupled model run for NAO. This signals the importance of the Indian Ocean, Niño-4 region, and the Caribbean Sea on the low-frequency variability of the NAO and implies a possibility to increase the prediction skill of the NAO variability by climate models. From the correlation between the individual AMIP members and the observations (Fig. 10), we estimate that the NAO index is much more sensitive to the initial condition than the PNA index.
Correlation of the (top) NAO and (bottom) PNA indices between each AMIP simulation and the observed data. None of the correlations for NAO index passed the significance tests. For the PNA index, the critical correlation value at the 5% significance level is 0.23.
Citation: Journal of Climate 27, 24; 10.1175/JCLI-D-14-00231.1
The pure linear model determines the variability of a given response from a one-way impact of the tropical SST forcing. The contribution of the internal noise is considered in our method but is uncertain. Since the sensitivity of the response to the external SST forcing is significantly detectable in our work, we assumed that the internal noise is associated with the SST forcing instead of a separate source of the detectable signal.
Previous work suggests that the mechanism for the low-frequency variability of NAO can be considered an atmosphere–ocean coupled mode (Latif and Barnett 1996; Liu and Alexander 2007). Based on the results from CESM we used in this paper, the coupled ocean–atmosphere component of NAO is probably not the dominant mechanism in this model. Given the limitation of this method, we only examine the impact of SST variability. The impact of other external factors should be considered, including long-term trends in the stratospheric ozone (Zhou et al. 2001), although it is outside the scope of this work.
Because the NAO and PNA patterns and their variability are closely related to the weather and climate of North America, Europe, and Africa (van Loon and Rogers 1978; Leathers et al. 1991; Cohen and Entekhabi 1999; Notaro et al. 2006; Seager et al. 2010; Hurrell et al. 2003) and the prediction skill of NAO and PNA on monthly to seasonal scales is still low (Johansson 2007), the result from this work highlights the need for accurate modeling of the SST. Starting from here, we can apply this method to other research objectives concerning the decadal impact of SST.
Acknowledgments
This work is supported by DOE Grants DE-SC0005399 and DE-SC0005171. The authors are very grateful to Joseph Barsugli for his constructive and valuable comments and many helpful discussions to improve the paper. We appreciate Shuguang Wang and Junhong Wei for their helpful discussions on the spectral analysis. We also thank Steven Feldstein, Byron Steinman, and Haiwen Liu for useful discussions and three anonymous reviewers who helped improve the paper.
REFERENCES
Anderson, J. L., and Coauthors, 2004: The new GFDL global atmosphere and land model AM2–LM2: Evaluation with prescribed SST simulations. J. Climate, 17, 4641–4673, doi:10.1175/JCLI-3223.1.
Annamalai, H., K. Hamilton, and K. R. Sperber, 2007: The South Asian summer monsoon and its relationship with ENSO in the IPCC AR4 simulations. J. Climate, 20, 1071–1092, doi:10.1175/JCLI4035.1.
Bader, J., and M. Latif, 2003: The impact of decadal-scale Indian Ocean sea surface temperature anomalies on Sahelian rainfall and the North Atlantic Oscillation. Geophys. Res. Lett., 30, 2169, doi:10.1029/2003GL018426.
Barnston, A., and R. Livezey, 1987: Classification, seasonality and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev., 115, 1083–1126, doi:10.1175/1520-0493(1987)115<1083:CSAPOL>2.0.CO;2.
Barsugli, J. J., and P. D. Sardeshmukh, 2002: Global atmospheric sensitivity to tropical SST anomalies throughout the Indo-Pacific basin. J. Climate, 15, 3427–3442, doi:10.1175/1520-0442(2002)015<3427:GASTTS>2.0.CO;2.
Barsugli, J. J., S. I. Shin, and P. D. Sardeshmukh, 2006: Sensitivity of global warming to the pattern of tropical ocean warming. Climate Dyn., 27, 483–492, doi:10.1007/s00382-006-0143-7.
Bojariu, R., and L. Gimeno, 2003: The role of snow cover fluctuations in multiannual NAO persistence. Geophys. Res. Lett., 30, 1156, doi:10.1029/2002GL015651.
Branstator, G., 1985: Analysis of general circulation model sea-surface temperature anomaly simulations using a linear model. Part I: Forced solutions. J. Atmos. Sci., 42, 2225–2241, doi:10.1175/1520-0469(1985)042<2225:AOGCMS>2.0.CO;2.
Bretherton, C. S., and D. S. Battisti, 2000: An interpretation of the results from atmospheric general circulation models forced by the time history of the observed sea surface temperature distribution. Geophys. Res. Lett., 27, 767–770, doi:10.1029/1999GL010910.
Cattiaux, J., R. Vautard, C. Cassou, P. Yiou, V. Masson-Delmotte, and F. Codron, 2010: Winter 2010 in Europe: A cold extreme in a warming climate. Geophys. Res. Lett., 37, L20704, doi:10.1029/2010GL044613.
Cohen, J., and D. Entekhabi, 1999: Eurasian snow cover variability and Northern Hemisphere climate predictability. Geophys. Res. Lett., 26, 345–348, doi:10.1029/1998GL900321.
Cohen, J., and M. Barlow, 2005: The NAO, the AO, and global warming: How closely related? J. Climate, 18, 4498–4513, doi:10.1175/JCLI3530.1.
Czaja, A., and C. Frankignoul, 2002: Observed impact of Atlantic SST anomalies on the North Atlantic oscillation. J. Climate, 15, 606–623, doi:10.1175/1520-0442(2002)015<0606:OIOASA>2.0.CO;2.
Deser, C., and M. L. Blackmon, 1993: Surface climate variations over the North Atlantic Ocean during winter: 1900–89. J. Climate, 6, 1743–1753, doi:10.1175/1520-0442(1993)006<1743:SCVOTN>2.0.CO;2.
Feldstein, S. B., 2000: The timescale, power spectra, and climate noise properties of teleconnection patterns. J. Climate, 13, 4430–4440, doi:10.1175/1520-0442(2000)013<4430:TTPSAC>2.0.CO;2.
Feldstein, S. B., 2003: The dynamics of NAO teleconnection pattern growth and decay. Quart. J. Roy. Meteor. Soc., 129, 901–924, doi:10.1256/qj.02.76.
Grötzner, A., M. Latif, and T. P. Barnett, 1998: A decadal climate cycle in the North Atlantic Ocean as simulated by the ECHO coupled GCM. J. Climate,11, 831–847, doi:10.1175/1520-0442(1998)011<0831:ADCCIT>2.0.CO;2.
Hoerling, M. P., J. W. Hurrell, and T. Xu, 2001: Tropical origins for North Atlantic climate change. Science, 292, 90–92, doi:10.1126/science.1058582.
Hoerling, M. P., J. W. Hurrell, T. Xu, G. T. Bates, and A. S. Phillips, 2004: Twentieth century North Atlantic climate change. Part II: Understanding the effect of Indian Ocean warming. Climate Dyn., 23, 391–405, doi:10.1007/s00382-004-0433-x.
Huang, J., K. Higuchi, and A. Shabbar, 1998: The relationship between the North Atlantic Oscillation and El Niño-Southern Oscillation. Geophys. Res. Lett., 25, 2707–2710, doi:10.1029/98GL01936.
Hurrell, J. W., 1995: Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation. Science, 269, 676–679, doi:10.1126/science.269.5224.676.
Hurrell, J. W., Y. Kushnir, G. Ottersen, and M. Visbeck, Eds., 2003: The North Atlantic Oscillation: Climatic Significance and Environmental Impact. Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 279 pp., doi:10.1029/GM134.
Jenkins, G. M., and D. G. Watts, 1968: Spectrum Analysis and Its Applications. Holden-Day, 541 pp.
Johansson, Å., 2007: Prediction skill of the NAO and PNA from daily to seasonal time scales. J. Climate, 20, 1957–1975, doi:10.1175/JCLI4072.1.
Kalnay, and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, doi:10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Kumar, A., and M. P. Hoerling, 1997: Interpretation and implications of the observed inter–El Niño variability. J. Climate, 10, 83–91, doi:10.1175/1520-0442(1997)010<0083:IAIOTO>2.0.CO;2.
Kushnir, Y., and J. M. Wallace, 1989: Low-frequency variability in the Northern Hemisphere winter: Geographical distribution, structure and time-scale dependence. J. Atmos. Sci., 46, 3122–3142, doi:10.1175/1520-0469(1989)046<3122:LFVITN>2.0.CO;2.
Latif, M., and T. Barnett, 1996: Decadal climate variability over the North Pacific and North America: Dynamics and predictability. J. Climate, 9, 2407–2423, doi:10.1175/1520-0442(1996)009<2407:DCVOTN>2.0.CO;2.
Leathers, D. J., B. Yarnal, and M. A. Palecki, 1991: The Pacific/North American Teleconnection Pattern and United States climate. Part I: Regional temperature and precipitation associations. J. Climate, 4, 517–528, doi:10.1175/1520-0442(1991)004<0517:TPATPA>2.0.CO;2.
Li, S., M. P. Hoerling, and S. Peng, 2006: Coupled ocean-atmosphere response to Indian Ocean warmth. Geophys. Res. Lett., 33, L07713, doi:10.1029/2005GL025558.
Li, W., C. E. Forest, and J. Barsugli, 2012: Comparing two methods to estimate the sensitivity of regional climate simulations to tropical SST anomalies. J. Geophys. Res., 117, D20103, doi:10.1029/2011JD017186.
Lin, H., J. Derome, and G. Brunet, 2005: Tropical Pacific link to the two dominant patterns of atmospheric variability. Geophys. Res. Lett., 32, L03801, doi:10.1029/2004GL021495.
Liu, Z., and M. Alexander, 2007: Atmospheric bridge, oceanic tunnel and global climatic teleconnections. Rev. Geophys., 45, RG2005, doi:10.1029/2005RG000172.
Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111, 46–59, doi:10.1175/1520-0493(1983)111<0046:SFSAID>2.0.CO;2.
Mokhov, I. I., and D. A. Smirnov, 2006: El Niño–Southern Oscillation drives North Atlantic Oscillation as revealed with nonlinear techniques from climatic indices. Geophys. Res. Lett., 33, L03708, doi:10.1029/2005GL024557.
Neale, R. B., and Coauthors, 2010a: Description of the NCAR Community Atmosphere Model (CAM 4.0). NCAR Tech. Note NCAR/TN-485+STR, 212 pp. [Available online at www.cesm.ucar.edu/models/ccsm4.0/cam/docs/description/cam4_desc.pdf.]
Neale, R. B., and Coauthors, 2010b: Description of the NCAR Community Atmosphere Model (CAM5.0). NCAR Tech. Note NCAR/TN-486+STR, 268 pp. [Available online at www.cesm.ucar.edu/models/cesm1.1/cam/docs/description/cam5_desc.pdf.]
Notaro, M., W.-C. Wang, and W. Gong, 2006: Model and observational analysis of the Northeast U.S. regional climate and its relationship to the PNA and NAO patterns during early winter. Mon. Wea. Rev., 134, 3479–3505, doi:10.1175/MWR3234.1.
Okumura, Y., S. P. Xie, A. Numaguti, and Y. Tanimoto, 2001: Tropical Atlantic air–sea interaction and its influence on the NAO. Geophys. Res. Lett., 28, 1507–1510, doi:10.1029/2000GL012565.
Osborn, T. J., 2004: Simulating the winter North Atlantic Oscillation: The roles of internal variability and greenhouse gas forcing. Climate Dyn., 22, 605–623, doi:10.1007/s00382-004-0405-1.
Paeth, H., M. Latif, and A. Hense, 2003: Global SST influence on twentieth century NAO variability. Climate Dyn., 21, 63–75, doi:10.1007/s00382-003-0318-4.
Peng, S., W. A. Robinson, and S. Li, 2002: North Atlantic SST forcing of the NAO and relationships with intrinsic hemispheric variability. Geophys. Res. Lett., 29, 1276, doi:10.1029/2001GL014043.
Peng, S., W. A. Robinson, and S. Li, 2003: Mechanisms for the NAO responses to the North Atlantic SST tripole. J. Climate, 16, 1987–2004, doi:10.1175/1520-0442(2003)016<1987:MFTNRT>2.0.CO;2.
Portis, D. H., J. E. Walsh, M. El Hamly, and P. J. Lamb, 2001: Seasonality of the North Atlantic Oscillation. J. Climate, 14, 2069–2078, doi:10.1175/1520-0442(2001)014<2069:SOTNAO>2.0.CO;2.
Pozo-Vázquez, D., M. J. Esteban-Parra, F. S. Rodrigo, Y. Castro-Díez, 2000: An analysis of the variability of the North Atlantic oscillation in the time and the frequency domains. Int. J. Climatol., 20, 1675–1692, doi:10.1002/1097-0088(20001130)20:14<1675::AID-JOC564>3.0.CO;2-C.
Rodwell, M. J., D. P. Rowell, and C. K. Folland, 1999: Oceanic forcing of the wintertime North Atlantic oscillation and European climate. Nature, 398, 320–323, doi:10.1038/18648.
Saravanan, R., 1998: Atmospheric low-frequency variability and its relationship to midlatitude SST variability: Studies using the NCAR climate system model. J. Climate, 11, 1386–1404, doi:10.1175/1520-0442(1998)011<1386:ALFVAI>2.0.CO;2.
Sardeshmukh, P. D., and B. J. Hoskins, 1988: The generation of global rotational flow by steady idealized tropical divergence. J. Atmos. Sci., 45, 1228–1251, doi:10.1175/1520-0469(1988)045<1228:TGOGRF>2.0.CO;2.
Seager, R., Y. Kushnir, J. Nakamura, M. Ting, and N. Naik, 2010: Northern Hemisphere winter snow anomalies: ENSO, NAO and the winter of 2009/10. Geophys. Res. Lett., 37, L14703, doi:10.1029/2010GL043830.
Simmons, A. J., J. M. Wallace, and G. W. Branstator, 1983: Barotropic wave propagation and instability, and atmospheric teleconnection patterns. J. Atmos. Sci., 40, 1363–1392, doi:10.1175/1520-0469(1983)040<1363:BWPAIA>2.0.CO;2.
Sutton, R. T., and M. R. Allen, 1997: Decadal predictability of North Atlantic sea surface temperature and climate. Nature, 388, 563–567.
Sutton, R. T., and D. Hodson, 2003: The influence of the ocean on North Atlantic climate variability 1871–1999. J. Climate, 16, 3296–3313, doi:10.1175/1520-0442(2003)016<3296:IOTOON>2.0.CO;2.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, doi:10.1175/BAMS-D-11-00094.1.
Thompson, R. O. R. Y., 1979: Coherence significance levels. J. Atmos. Sci., 36, 2020–2021, doi:10.1175/1520-0469(1979)036<2020:CSL>2.0.CO;2.
Trenberth, K. E., G. W. Branstator, D. Karoly, A. Kumar, N. C. Lau, and C. Ropelewski, 1998: Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures. J. Geophys. Res., 103, 14 291–14 324, doi:10.1029/97JC01444.
van Loon, H., and J. C. Rogers, 1978: The seesaw in winter temperatures between Greenland and northern Europe. Part I: General description. Mon. Wea. Rev., 106, 296–310, doi:10.1175/1520-0493(1978)106<0296:TSIWTB>2.0.CO;2.
Wallace, J. M., and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109, 784–812, doi:10.1175/1520-0493(1981)109<0784:TITGHF>2.0.CO;2.
Whitaker, J. S., and P. D. Sardeshmukh, 1998: A linear theory of extratropical synoptic eddy statistics. J. Atmos. Sci., 55, 237–258, doi:10.1175/1520-0469(1998)055<0237:ALTOES>2.0.CO;2.
Zhou, S., A. J. Miller, J. Wang, and J. K. Angell, 2001: Trends of NAO and AO and their associations with stratospheric processes. Geophys. Res. Lett., 28, 4107–4110, doi:10.1029/2001GL013660.