Wintertime Low-Frequency Weather Variability in the North Pacific–American Sector 1949–93

Alison C. Renshaw Hadley Centre for Climate Prediction and Research, U.K. Meteorological Office, Bracknell, Berkshire, United Kingdom

Search for other papers by Alison C. Renshaw in
Current site
Google Scholar
PubMed
Close
,
David P. Rowell Hadley Centre for Climate Prediction and Research, U.K. Meteorological Office, Bracknell, Berkshire, United Kingdom

Search for other papers by David P. Rowell in
Current site
Google Scholar
PubMed
Close
, and
Chris K. Folland Hadley Centre for Climate Prediction and Research, U.K. Meteorological Office, Bracknell, Berkshire, United Kingdom

Search for other papers by Chris K. Folland in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A study of the impact of ENSO in the Hadley Centre’s atmospheric climate model HADAM1 is presented, with emphasis on the North Pacific–American (NPA) sector. The study is based both on observational data and an ensemble of six integrations for the period 1949–93, forced with observed global sea-ice and sea surface temperature data. The model is shown to reproduce most of the known features of the worldwide atmospheric response to ENSO in boreal winter (January–March).

Focusing on the NPA sector, the leading modes of low-frequency weather variability in the winter season are identified on their natural timescales for both the modeled and observed atmospheres. These modes are analyzed via rotated EOF analysis of daily 500-hPa height data, filtered to remove synoptic timescale variations. The model gives a reasonably skillful simulation of the main features of the four leading modes in the NPA region:the Pacific–North American (PNA), the west Pacific (WP), the east Pacific (EP), and the North Pacific (NP) modes. The sensitivity of these modes to SSTs is investigated. In particular, sensitivity to SSTs associated with ENSO is analyzed in terms of the shift in frequency of occurrence of the opposing phases of a mode between warm event (El Niño) and cold event (La Niña) years. Three of the observed modes show such a sensitivity: the PNA, WP, and NP modes. Of the corresponding model modes, only the PNA responds significantly to ENSO (but too strongly in warm event years), which is clearly illustrated by changes in both the frequency and duration of PNA episodes between warm and cold event years. The EP mode shows no sensitivity to ENSO, in either model or observed atmospheres. Finally, although the model is able to reproduce the pattern of decadal anomalies seen in the North Pacific in the years 1977–87, which is related to the prevalence of the positive phase of the PNA in this period, it does so with a much reduced amplitude; possible reasons for this discrepancy are discussed.

Corresponding author address: Dr. David P. Rowell, Hadley Centre for Climate Prediction and Research, U.K. Meteorological Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.

Email: dprowell@meto.gov.uk

Abstract

A study of the impact of ENSO in the Hadley Centre’s atmospheric climate model HADAM1 is presented, with emphasis on the North Pacific–American (NPA) sector. The study is based both on observational data and an ensemble of six integrations for the period 1949–93, forced with observed global sea-ice and sea surface temperature data. The model is shown to reproduce most of the known features of the worldwide atmospheric response to ENSO in boreal winter (January–March).

Focusing on the NPA sector, the leading modes of low-frequency weather variability in the winter season are identified on their natural timescales for both the modeled and observed atmospheres. These modes are analyzed via rotated EOF analysis of daily 500-hPa height data, filtered to remove synoptic timescale variations. The model gives a reasonably skillful simulation of the main features of the four leading modes in the NPA region:the Pacific–North American (PNA), the west Pacific (WP), the east Pacific (EP), and the North Pacific (NP) modes. The sensitivity of these modes to SSTs is investigated. In particular, sensitivity to SSTs associated with ENSO is analyzed in terms of the shift in frequency of occurrence of the opposing phases of a mode between warm event (El Niño) and cold event (La Niña) years. Three of the observed modes show such a sensitivity: the PNA, WP, and NP modes. Of the corresponding model modes, only the PNA responds significantly to ENSO (but too strongly in warm event years), which is clearly illustrated by changes in both the frequency and duration of PNA episodes between warm and cold event years. The EP mode shows no sensitivity to ENSO, in either model or observed atmospheres. Finally, although the model is able to reproduce the pattern of decadal anomalies seen in the North Pacific in the years 1977–87, which is related to the prevalence of the positive phase of the PNA in this period, it does so with a much reduced amplitude; possible reasons for this discrepancy are discussed.

Corresponding author address: Dr. David P. Rowell, Hadley Centre for Climate Prediction and Research, U.K. Meteorological Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.

Email: dprowell@meto.gov.uk

1. Introduction

In the 100 years since its discovery (Hildebrandsson 1897), the Southern Oscillation has been the subject of much interest and research. Following the work of Bjerknes (1969), we now understand the Southern Oscillation to be intimately linked with the cycle of warm (El Niño) and cold (La Niña or El Veijo) sea surface temperature (SST) anomalies in the equatorial Pacific, part of a complex ocean–atmosphere interaction usually termed ENSO (El Niño–Southern Oscillation). For ease of reference we shall use the term ENSO to encompass both extremes of the tropical Pacific cycle, which we shall refer to separately as warm and cold events. The interest in ENSO stems from its importance to weather patterns throughout the globe. Such teleconnections were first recognized by Walker (1924) and have been explored in many subsequent papers, the 1980s being a particularly prolific decade for such research (e.g., Horel and Wallace 1981; Pan and Oort 1983; Ropelewski and Halpert 1987, 1989; Kiladis and Diaz 1989). One extratropical area that stands out from these studies as having a strong ENSO response is the North Pacific–American (NPA) sector, where a characteristic pattern of anomalies is found to recur in ENSO years, the so-called Pacific–North American (PNA) pattern. The response of NPA modes of low-frequency weather variability is the primary focus of this paper.

Alongside the observational studies, a growing body of work focused on the response of atmospheric general circulation models (AGCMs) to imposed SST anomalies in the tropical Pacific. These studies ranged from those using idealized anomalies (e.g., Rowntree 1972), or a composite warm event anomaly (e.g., Blackmon et al. 1983; Shukla and Wallace 1983), to multiyear integrations with observed SSTs (e.g., Lau 1985), which allowed the evolution of ENSO events to be followed. In recent years the need to perform such integrations in ensemble mode has been increasingly recognized (e.g., Barnett et al. 1994; Smith 1995), so that SST-forced signals in the extratropics might be better distinguished from the noise of the internal atmospheric variability (Rowell 1998). In general such modeling studies have given reasonable simulations of the global response to warm events, including the PNA response, although some studies looking separately at cold events have been less successful (e.g., Cubasch 1985). Recent multiyear integrations have been able to capture some aspects of the so-called climate shift seen in the North Pacific in the late 1970s–early 1980s (e.g., Graham et al. 1994), which involved large temperature anomalies over North America and which has been related to changes in ENSO. The ability to understand the origin of such regional variations is particularly important in light of current interest in the detection of anthropogenic climate change. The ability to simulate the atmospheric response to tropical Pacific SSTs is also of great importance to seasonal forecasting, especially as the skill in predicting these SSTs has increased considerably in recent years (Barnston 1995).

While AGCMs can provide us with a realistic atmosphere in which to investigate the mechanisms linking warm and cold events to the extratropics, this verisimilitude, and hence complexity, also makes such investigation difficult. Insight into these mechanisms has more usually been gained through the use of more simplified models. For example, the work of Hoskins and Karoly (1981), later extended by Sardeshmukh and Hoskins (1988), provided a framework for understanding the PNA response to warm and cold events. Warm SST anomalies in the tropical Pacific enhance local precipitation and give rise to upper-level divergence; advection of vorticity by the divergent flow gives an effective Rossby wave generator in the western subtropical Pacific. Wave trains emanating from this region carry energy into the extratropics in a great circle path, which, in the Northern Hemisphere, arches over the North Pacific and North America to give characteristic patterns in the midlatitude jet stream. Other studies have indicated that the associated shifts in storm track positions in warm and cold events, and consequent changes in transient vorticity fluxes, also have an important role in maintaining the extratropical teleconnections (Held et al. 1989;Lau 1988; Hoerling and Ting 1994).

The relationship between ENSO events and the PNA pattern is, however, less deterministic than the above discussion suggests. Strong PNA patterns are produced in non-ENSO years both in observed atmospheres and in modeled atmospheres forced with observed SSTs, while a strong warm or cold event does not always give the expected PNA response. Moreover, the PNA pattern is identified as a preferred mode of variability in AGCMs forced with climatological SSTs (Lau 1981; Chervin 1986). These results indicate that the PNA mode, and presumably other patterns, are natural modes internal to the atmosphere and that the effect of a warm or cold event is to bias the probability that the PNA mode will occur in its positive or negative phase. Palmer (1993) gives a very useful discussion of this view in which Lorenz’s (1963) simple model of a nonlinear system provides an analogy for the chaotic extratropical circulation. The two wings of the system’s attractor represent two distinct weather regimes, between which transitions occur at random intervals, but which in the long term are equally likely states for the system. Palmer shows that if a linear oscillator, representing the tropical atmosphere, is coupled to the Lorenz model, the effect is not to alter the phase space location of the two regimes, but rather to alter their relative population density. The system remains chaotic and transitions between regimes still occur. Such behavior corresponds well to the PNA response to ENSO seen in modeling (Molteni et al. 1990) and observed (Livezey and Mo 1987) studies.

It is with this theoretical framework in mind that we approach the analysis of the response to ENSO seen in multidecadal atmospheric simulations and in the observed record. We use a six-member ensemble of integrations, forced with observed SSTs and sea-ice data, to first give a brief analysis of our model’s skill in simulating the worldwide boreal winter response to ENSO. Then, focusing on the NPA sector, the leading modes of low-frequency weather variability are identified. The ensemble provides the opportunity to evaluate the relative importance of SSTs to the variability of each of the NPA modes. These modes are identified on their natural timescales, by analysis of daily data, low-pass filtered to remove synoptic timescale variations. This allows the influence of ENSO on the frequency of occurrence of the modes to be investigated and any shift in probability distributions, as suggested by Palmer (1993), to be identified. This would not be possible with an analysis of monthly or seasonal data.

The paper is arranged as follows. A brief description of the model and experimental design is given in section 2. In section 3, the skill of the model in simulating the seasonal mean response to ENSO is assessed, by comparing model and observed composites for warm and cold events. In section 4, the leading modes of variability in the NPA region are identified using rotated EOF analysis. Model and observed modes are compared, and for each of the modes, the influence of warm and cold events is assessed in terms of shifts in population distribution. The influence on duration of events is also studied for the PNA mode. Finally, in section 5, we present some discussion and conclusions.

2. Methodology

a. Model formulation

All integrations described in this paper use the first version of the Hadley Centre Atmospheric Model, known as HADAM1. This is a gridpoint model with a horizontal resolution of 2.5° lat × 3.75° long and 19 hybrid levels in the vertical. Boundary layer processes are represented on the lowest five levels. There is also a four-level soil model determining heat transport and a single-layer soil model and vegetation canopy for surface hydrology. The model calculates cloud amounts (convective and large scale) and cloud liquid water and ice contents. The radiation scheme, which resolves six shortwave and four longwave bands, responds to these cloud variables. The atmospheric carbon dioxide concentration is fixed at 321 ppmv, whereas ozone varies with season and latitude. HADAM1 is the model submitted in 1993 by the U.K. Meteorological Office to the Atmospheric Model Intercomparison Project, and a full description of its physics and dynamics is given in Phillips (1994). Rowell (1996) provides a discussion of the skill of its global climatology.

b. Experimental design and data

Model results in this paper are based on an ensemble of six multidecadal integrations forced with observed SSTs; the same set of experiments analyzed in Rowell (1998) and Davies et al. (1997). Each integration was started at 0000 UTC on 1 October 1948, using its own, unique set of initial conditions, selected from recent U.K. Meteorological Office operational analyses for early October. Initial values of soil moisture and snow depth were provided from an adaptation of Willmot et al.’s (1985) climatologies, whereas deep soil temperatures were initialized using equations given in Warrilow et al. (1986). After initialization, heat and moisture contents over the land surface were free to interact with the model atmosphere. As the model atmosphere takes some time to adjust from its initial state to one in quasi equilibrium with the historic SSTs, the first two months of the integration were discarded. The integrations were run on to 0000 UTC on 1 December 1993, giving 45 complete years for analysis.

The runs were forced throughout by observed data from the GISST (Global Sea-Ice and Sea Surface Temperature) dataset, version 1.1 (Parker et al. 1995a). GISST provides globally complete, monthly mean fields of SST and sea-ice data on a 1° × 1° grid (calculated by projecting 5° SST anomalies onto a 1° climatology). Grid points are either set to ice covered or ice free. The SSTs were interpolated to the larger model grid by taking the area-weighted average of temperatures in nonice 1° squares; where ice cover exceeded 50%, model grid boxes were designated as ice points. Finally, the monthly SST and sea-ice data were interpolated to pentad values to provide a smoother boundary forcing to the model atmosphere.

Several observed datasets are used in this paper to assess the model performance. Observed mean sea level pressure (MSLP) data were taken from the latest version (GMSLP2.1d) of the Allan et al. (1996) and Basnett and Parker (1997) global dataset, which is a quality controlled blend of existing gridded datasets. The surface air temperatures used are a blend of land–air temperature data from the Climate Research Unit, University of East Anglia (Jones 1994), and night marine air temperature data, much as described by Bottomley et al. (1990) though since improved as described in Parker et al. (1995b). The 500-hPa height data (seasonal and daily data) were taken from the U.K. Meteorological Office archive (M. Jackson 1986, personal communication).

3. Modeling the mean response to ENSO

In this section we summarize the skill of the HADAM1 model in simulating the seasonal mean atmospheric response to warm and cold events in the tropical Pacific. We consider first the model’s skill in simulating the Southern Oscillation index (SOI) and then describe the global response of MSLP in winter. Last, we begin to focus on one of the extratropical regions most strongly affected by tropical Pacific SSTs, the North Pacific–American sector.

a. Modeling the SOI

The SOI is widely used as a measure of the tropical atmospheric changes associated with warm and cold events (Ropelewski and Jones 1987). The index shown in Fig. 1 is the Tahiti minus Darwin MSLP difference calculated from 5-month running mean values of normalized monthly anomalies (with respect to 1951–80). The SOI values of the six model runs are in very good agreement, indicating a strong SST-forced component. Given an ensemble of runs with a common forcing, the“analysis of variance” technique allows us to assess the fraction of variance, ρ, attributable to that forcing; the remaining variance being the result of random variations internal to the atmosphere. Details of the technique are given in Rowell et al. (1995) and Rowell (1998). For the SOI shown in Fig. 1, the oceanic forcing explains 86% of the variance [with 90% confidence limits of 81% and 90%, calculated as in Rowell 1998].

Even if the model were perfect, the expected correlation between the ensemble mean and the observed SOI would be less than unity, since both time series are subject to random variations. The perfect model correlation, rEM PM, between the ensemble mean and a further model run is easily calculated from ρ (Rowell 1998), and this provides an upper limit against which model-observed correlations may be assessed. For the SOI in Fig. 1, the correlation between the ensemble mean and observed values is 0.85, which is close to the rEM PM of 0.92; thus the SOI is skilfully simulated. However, the model SOI does have a slight positive bias in more recent years, such that the response is too strong in the 1985 cold event and too weak in the 1982–83 and 1987 warm events. This bias has also been seen in other modeling studies (e.g., Konig et al. 1990; Latif et al. 1990; Smith 1995; Livezey et al. 1995) and could be due to errors in the forcing SST data, or in the verifying MSLP data, or could possibly reflect a common error in the models’ response at Tahiti and/or Darwin to the global SST forcing over this period.

b. Global response of MSLP to ENSO in winter

Here we examine the skill of the model’s mean wintertime response to ENSO to demonstrate that our model is adequate for the Pacific SST-based methodological study of section 4 (an analysis for all seasons is beyond the scope of this paper). A composite analysis is used, for which years were selected from those simulated (1949–93) by ranking them in order of their mean SST in the Niño3 region (5°N–5°S, 90°–150°W); the top and bottom quartiles were used, and these years are shown in Table 1. To compute observed composite values, data were required in at least half of these warm or cold years. MSLP is used for this global view as the observed data are nearly complete and fairly trustworthy. For our definition of winter, we choose the months January–March (JFM) as being most appropriate for our focus on the NPA sector, partly because this is the season most strongly influenced by SSTs here (Barnston 1994), partly because the February and March ENSO composite responses are more consistent with each other than the December and January responses (not shown), and partly because the observed and modeled composite differences are in rather better agreement in this season (also not shown). The latter point is likely a consequence (at least to some extent) of the two preceding points.

Local significance is identified with a two-tailed t test, but the nominal sample size must be reduced to account for possible serial correlation between the selected events. Assuming a Markov process, the effective number of points in a time series, ne, can be approximated by
i1520-0442-11-5-1073-eq1
where n0 is the actual number of points and r1 is the first serial correlation coefficient (WMO 1966). Since the years selected for each composite are not generally consecutive, ne provides a low estimate for the effective sample size. Thus significance calculated using ne tends to be conservative. To then compute this significance of warm minus cold composite differences, the t statistic is calculated as follows (e.g., Zwiers and von Storch 1995):
i1520-0442-11-5-1073-eq2
where X1 (X2) = warm (cold) event seasonal mean, ne1 (ne2) = effective sample size in warm (cold) years, and S is the estimate of the standard deviation of the population sampled (here the standard deviation of all seasonal values over 1949–93). The probability of such a value occurring by chance at the 95% confidence level is then found from the t statistic, using ne1 + ne2 − 2 degrees of freedom. Finally, we note that because a larger sample is available from the model, the areas of significant difference between warm and cold composites tend to be more extensive than those in the observed data.

The modeled and observed warm and cold event composites for MSLP are shown in Fig. 2. First, in warm event years there is fairly good agreement between observed and modeled composites. One important exception is that the modeled low anomaly in the North Pacific is displaced to the southeast. Over much of the North Atlantic the ensemble anomalies achieve significance, although the observed anomalies do not, perhaps because of their smaller sample size. The ensemble mean pattern in this region shows increased pressure near Iceland and reduced pressure in the midlatitude North Atlantic, corresponding to increased blocking and the low phase of the North Atlantic Oscillation. This is consistent with the patterns identified by Rogers (1984) and Fraedrich and Muller (1992). In cold event years, anomalies are generally slightly weaker, and over North America, the observed composite suggests a fairly low consistency of response between events.

Responses elsewhere are less important for the purpose of this paper, although it is worth noting the general similarity between observed and modeled composites. In the warm phase, pressure increases over the Antarctic and sub-Antarctic, and the classic responses of increased MSLP over Australia with decreases to the south and east are also seen. The model’s cold event responses also show a general similarity with those observed, and although many anomalies are the opposite of those for the warm phase, the details are less well modeled especially at higher southern latitudes.

Overall the model is acceptably similar to the observations in both phases of ENSO when stratified according to SST anomalies in the tropical East Pacific.

c. ENSO response in the extratropical North Pacific–American sector (20°–80°N, 50°W–120°E)

The remainder of the paper will focus on the North Pacific–American (NPA) sector, which has a moderate winter response to ENSO in both the model and observations. It therefore provides an ideal test bed for developing techniques to investigate the response of the atmosphere’s natural modes to ENSO. As above, winter is defined as the months January–March (JFM).

In section 4 we shall analyze the natural modes of variability in the NPA sector using 500-hPa height data. For ease of comparison, we show here composites in the NPA sector using 500-hPa height data rather than MSLP. However, although tending to have more extensive areas of significance, composite anomalies for MSLP show very similar patterns to those of 500-hPa height, demonstrating the equivalent barotropic nature of anomalies in this region (Blackmon et al. 1983). Figures 3a and 3b show observed and modeled warm minus cold event composite anomalies of 500-hPa height for JFM for (approximately) the NPA sector. Model and observed analyses both show three strong anomaly centers: negative over the North Pacific and southeast United States and positive over Canada. This general pattern of response is well known from many observed and modeling studies (e.g., Horel and Wallace 1981; Chen 1982; Lau and Nath 1994). Although the model captures the general nature of the observed pattern, there are discrepancies in the location of the anomaly centers, most noticeably over the North Pacific, where the model’s negative anomaly is centered too far south and east.

This model verification has also been extended to include a comparison of modeled and observed composite temperature differences (Figs. 3c and 3d). The main error is that the observed positive differences over Canada and the northern United States extend to the east coast but do not do so in the model; this is consistent with the 500-hPa height errors implying too much cold air advection over the eastern United States. Finally, the warm minus cold event precipitation composite difference (not shown) also reveals a strong modeled response to ENSO in the NPA region. Broadly, this is similar to the observed composite differences over North America using data from Hulme (1994) and also to the studies of Ropelewski and Halpert (1986) and Kiladis and Diaz (1989).

4. Leading modes of winter variability in the North Pacific–American sector and their relation to SSTs

The previous section highlighted the connection between ENSO and the atmosphere in the North Pacific–American (NPA) sector (50°W–120°E, 20°–85°N). We now investigate this link in terms of the influence of warm and cold events on the leading modes of low-frequency weather variability in the NPA region.

a. Methodology and description of modes

We have chosen to identify the modes of low-frequency weather variability by performing a rotated empirical orthogonal function (REOF) analysis of 500-hPa height data in the Northern Hemisphere. Patterns derived by REOF analysis are more robust than their unrotated counterparts (Cheng et al. 1995) and are believed to be more physically realistic, since unrotated EOFs can compound physically independent spatial patterns, when those patterns are substantially smaller than the analysis domain. To investigate the modes on their natural time scales, we have used daily data, low-pass filtered to remove synoptic timescale variations. Our results are therefore closer to those of Richman (1994), who also used low-pass filtered daily data, than to those of Barnston and Livezey (1987) where monthly mean fields were analyzed.

As the model’s modes of variability may be distinct from those found in nature, we have identified model and observed modes separately. All observed and modeled REOFs were calculated using data for the entire Northern Hemisphere north of 20°N, and for the years 1970–93; lack of data in the Pacific region made analysis of the observed fields in earlier years unreliable.

A 20-yr daily climatology (1971–90) was calculated and then smoothed by taking the 5-day running mean. Daily anomalies from the smoothed climatology were calculated, and a low-pass 11-point nonrecursive filter (Walraven 1984) applied to remove synoptic-scale variations with periods of about 8 days or less (i.e., anomalies of less than about 4 days duration). The response of the filter is shown in Fig. 4. The filtered daily anomalies for the winter season (JFM) were then selected for further analysis. The filtered daily anomalies were sampled at 5-day intervals and interpolated to an equal area grid, with a meridional resolution of 5°, and zonal resolution equivalent to 10° at the equator. Correlation EOFS were calculated on this equal area grid, and the leading 15 EOFs rotated using the VARIMAX rotation algorithm. To display the REOFs, which is easier on the model grid, the correlation coefficients between the original filtered daily anomalies and the time coefficients of each REOF were calculated. It is the maps of these correlations that are shown in Fig. 5 (discussed below).

In calculating the model REOFs, daily 500-hPa height data from a single run only were used, due to computational restrictions. However, the process was repeated for two further runs, and the modes discussed below were identified again among the leading REOFs (indicated by spatial correlations usually greater than 0.9, and rankings always in the first six as for the observed modes), even though it should be noted that some were a little less distinct from one another than their observed counterparts. As a further test of robustness, the analyses were repeated with the 5-day sampling starting from a different day, thus giving a different set of daily data; little change was found in the major features of the leading modes.

The four patterns with strongest weights in the NPA sector were chosen for further analysis. These are shown in Fig. 5, and adjacent to each observed REOF is the corresponding model mode. As the REOFs were calculated using hemispheric data, the percentage of variance explained by each mode in the NPA sector was calculated from the correlations shown in Fig. 5 (using the area-weighted average of their squared values). These are listed in Table 2. Calculated in this way the four modes shown in Fig. 5 between them explain about half the variance in the NPA sector (45% for observed, 53% for model).

The observed mode explaining the highest amount of variance in the NPA sector is shown in Fig. 5a. This is the well-known PNA pattern described by Wallace and Gutzler (1981) in an analysis of monthly data and by many subsequent authors. The PNA pattern has four anomaly centers, which take alternate signs in the following order: Hawaii, the North Pacific, northwest America, and the southeastern United States. The pattern is seen only in the cold half of the year (Barnston and Livezey 1987), but on a wide variety of timescales from days and months through seasons to decades (Blackmon et al. 1983). In Fig. 5a, the three extratropical centers of action are clear, showing the conventional“positive” phase of the PNA pattern, that is, with a strong negative center over the North Pacific. The corresponding model mode is shown in Fig. 5b. Agreement between modeled and observed PNA modes is generally good (see Table 2) but the model’s North Pacific center is less extensive and centered to the east of the corresponding observed node. Due to its more limited extent, the model PNA explains slightly less of the variance in the NPA sector than does the observed mode.

The second observed mode of the NPA sector (shown in Fig. 5c) is the west Pacific mode (WP). It corresponds to WP modes identified in monthly 700-hPa height data by Horel and Wallace (1981) and Barnston and Livezey (1987). Its main feature is a north–south dipole in the western Pacific; a weaker north–south dipole of opposite polarity lies over the North American continent. The model simulates the western Pacific dipole well (Fig. 5d), although the southern center is located a little too far north. However, the model’s pattern over North America is rather different to the observed REOF, where this weaker node has shifted southeastward; this reduces the spatial correlation between the modeled and observed modes (see Table 2).

The third observed mode for the NPA sector is an eastern Pacific pattern (EP) (shown in Fig. 5e). The EP mode has a north–south dipole in the eastern Pacific, one node centered over Alaska, the other near Hawaii at around 30°S. A third node, in phase with that over Hawaii, lies over the Great Lakes. The corresponding model REOF pattern (Fig. 5f) agrees very well with the observed pattern (see Table 2). The EP mode is very close to the 12th hemispheric REOF at 700 hPa seen in Richman (1994), and its two western centers closely resemble the east Pacific mode of Barnston and Livezey (1987), although the node over eastern Canada corresponds better with a node of their Tropic–Northern Hemisphere (TNH) pattern (see discussion below).

The final observed mode for the NPA sector (shown in Fig. 5g) is the North Pacific pattern (NP). One center is located over southern Japan, and this merges into a second center of the same sign that lies to the west. A broader northern node of opposite phase lies above these two centers. The corresponding model mode (Fig. 5h) is in good agreement with the observed NP pattern (see Table 2). The NP mode is very close to the eighth hemispheric REOF at 700 hPa seen in Richman (1994).

One pattern prominent in other studies (e.g., Mo and Livezey 1986; Barnston and Livezey 1987) that has not been delineated here as a leading mode is the TNH. This could be because the TNH prefers timescales of a season or more (Mo and Livezey 1986), which would lead to its suppression by the relatively higher frequencies included here (R. E. Livezey 1997, personal communication). This idea is supported by Barnston and Livezey’s (1987) study, which shows a notable drop in the variance explained by the TNH when 10-day means are used rather than monthly means. However, this of course raises interesting questions about the physical origins of the TNH, and although these are beyond our current paper, they clearly demand further research.

In summary, the four modeled REOFs shown in Fig. 5 show a reasonable likeness with their observed counterparts computed in the same way, and so we conclude that these winter NPA modes of low-frequency variability are well simulated by the model. We now investigate how these modes are influenced by SSTs and in particular SSTs in the tropical Pacific.

b. Influence of SSTs on the Pacific North American mode

Interannual variations of the PNA mode (and other modes; see subsequent subsections) are investigated using a number of complementary approaches. First, we study the interannual time series by calculating seasonal (JFM) means of the filtered daily REOF amplitudes. These “seasonal time coefficients” are shown in Fig. 6a for modeled and observed data; the latter were calculated only when spatial coverage was considered sufficient (from 1968). Figure 6a shows the individual model runs to be in good agreement; almost half their interannual variance can be explained by the common SST and sea-ice forcing (see Table 3). Agreement with observations is also good; the modeled interannual evolution of the PNA mode is very well captured, given the degree of random variability (cf. rEM and rEM PM in Table 3).

To identify which areas of SST most influence the PNA mode, Fig. 6b (6c) shows correlations of GISST data versus observed (model ensemble mean) seasonal PNA time coefficients. For consistency, both figures are based on data for 1968–93; use of model data over the full period of integration makes little difference. Note that for the calculation of significance, account was taken of serial correlation following Folland et al. (1991). The model and observed patterns of correlation are fairly well matched, with similarly extensive areas of significant correlations. They differ significantly only over the central North Pacific, where the observed correlations are more strongly negative. Elsewhere, model correlations tend to be stronger than the observed, which is expected given higher signal-to-noise ratios of ensemble mean values. The link between the PNA mode and ENSO, highlighted in previous studies using monthly mean data (e.g., Horel and Wallace 1981; Yarnal and Diaz 1986), is clearly revealed in the pattern of correlations in Figs. 6b and 6c. One further point of interest is the strong correlations found in the extratropical North Pacific; positive along the coast of America and negative in the central Pacific. Recent studies indicate that such SST anomalies are driven by the extratropical atmosphere (e.g., Frankignoul 1985; Alexander 1992; Miller et al. 1994), which thus links extratropical and tropical Pacific SSTs. If our model’s PNA mode responds correctly to tropical Pacific SSTs, correlations with the prescribed extratropical Pacific SSTs will inevitably also be obtained (Graham et al. 1994), though they may not be quite as strong. The influence of extratropical SSTs on the atmosphere is thus hard to assess. However, there is evidence that this pattern of North Pacific SST anomalies may influence storm track positions, which in turn influence the anomalous large-scale flow, such that the likelihood of a PNA event is increased (e.g., Zhang et al. 1997). A positive feedback between the extratropical atmosphere and ocean may thus be established (Trenberth and Hurrell 1994; Palmer 1988a, 1993). Since in our experiments such a feedback is not possible, then a relatively small error in the model—for example, in storm track position or strength—could perhaps prevent the influence of extratropical SSTs from being properly simulated (cf. Palmer 1995). It is possible that this may account for the lower simulated correlations in the North Pacific compared with those observed.

Cross-spectral analysis of modeled and observed seasonal PNA time coefficients (see Fig. 7) reveals that they are significantly coherent on timescales of approximately 2.5–8 yr [significance at the 95% level, calculated as in Bloomfield (1976)]. This frequency range matches that associated with ENSO. Filtering the PNA time series to isolate this frequency band increases the percentage variance explained by SSTs to 58%, and the correlation between filtered ensemble mean and observed values to 0.78.

To give more insight into the relation between El Niño–type patterns of SST and the PNA pattern, the relative frequency of occurrence of the positive and negative PNA modes in warm and cold event years has been investigated. Using JFM data for all the years 1968–93, a probability distribution estimate (PDE) was calculated for the daily filtered PNA mode time coefficients. Then, as in our composite calculations, the available years were ranked in order of the Niño3 SSTs, and data from the top and bottom quartiles pooled to give “warm event” and “cold event” subsets, for which PDEs were then calculated separately. Each PDE was calculated using a Gaussian kernel estimator, which gives the PDE of a variable x as
i1520-0442-11-5-1073-eq3
where x are the observed data, σ their standard deviation, and N their number (Molteni et al. 1990). Here, α determines the level of smoothing and was set to 0.3; this corresponds approximately to a histogram with eight bins over ±2 σ; however, since the distributions are only used for display, the value of α is not critical.

The significance of the change in population between warm and cold event years, and changes from all years, can be assessed by applying the standard Kolmogorov–Smirnov (KS) test (e.g., Hoel 1971; Press et al. 1992) to the daily filtered time coefficients, once the degrees of freedom have been adjusted to take account of the high serial correlation in the data. Using an F test, the significance of differences in standard deviation between the warm and cold event subsets may also be assessed, again with the reduced degrees of freedom. All results of these tests pass or fail at the 95% confidence level, unless stated otherwise.

PDEs for the observed PNA mode are shown in Fig. 6d. The “all years” curve (solid line) is asymmetric, with more high-amplitude cases when the mode is negative and more small-amplitude cases when it is positive. This implies less variability in the positive PNA mode, consistent with the suggestion of Palmer (1988b, 1993) that the atmosphere is more barotropically stable when it is in this phase. In cold event years (dashed line) the distribution shifts toward negative values, and the number of positive cases is markedly reduced. Conversely, in warm event years (dotted line) the distributional change is much smaller, with only a slight increase in the number of positive cases, and a small reduction in the number of negative cases. The KS test shows the warm and cold event populations are significantly different at the 99% confidence level, but only the cold event population is significantly different from the all years population (again at the 99% level). The nonlinear nature of this response, with a stronger preference toward a PNA-type pattern in cold event rather than warm event years, is also reflected in the composite analysis of Livezey et al. (1997). In addition, the standard deviation of the filtered daily coefficients in cold event years is significantly more than that in warm event years, consistent with the greater variability of the negative phase. This larger intraseasonal variability in these years is also likely to enhance the random component of seasonal mean variations and so could contribute to the finding of some studies that the link between cold events and the negative PNA phase can appear rather tenuous (e.g., Cubasch 1985).

The corresponding results for the model are shown in Fig. 6e. The behavior of these time coefficients is broadly similar to that observed, with the exception that they fail to capture the nonlinearity noted above, in that the model’s probability of a PNA mode in warm event years appears too high. The KS test shows all distributions to be significantly different from one another at a confidence level of 99%. The standard deviation of the filtered daily coefficients in warm event years is also significantly less than that in cold event years.

The ENSO-induced changes in the frequency of occurrence of the opposing phases of the PNA mode are also clearly illustrated in Fig. 9. Here, the daily filtered time coefficients of the PNA mode have been used to assess the duration of PNA episodes. The critical value for defining an episode was chosen as a time coefficient of |1.25|. This includes (approximately) the highest 10% and lowest 10% of all time coefficients from modeled and observed analyses; from visual inspection these are associated with patterns that correspond reasonably well to the REOF modes. If the critical value is exceeded on n consecutive days, then an episode of n days duration is said to have occurred. Filtered daily coefficients, rather than the unfiltered daily values, were used to assess duration. In general this makes little difference to the calculated duration of the PNA episodes, but in some cases it does prevent short timescale variations from breaking up what is clearly a single episode. The diagram shows distributions from observed and pooled model data. The observed distributions are more noisy, since they are based on less data, but are in generally good agreement with the modeled distributions. The occurrence of observed positive and negative PNA episodes is not governed solely by the phase of ENSO; one of the longest positive PNA episodes, of 23 days duration, occurred in a weak cold event year (1985). Nevertheless, the influence of ENSO is clearly apparent in Fig. 8 with episodes tending to be longer and more frequent when the phase of ENSO is favorable. The longest observed episode, lasting 41 days, was a positive episode in 1983, the warmest of the warm events in the period studied. The model positive PNA episodes tend to be of shorter duration than the observed, the longest in 1983 was 16 days, perhaps because the model fails to respond correctly to extratropical SST patterns or perhaps because of errors in the stability of the model’s basic flow. As we have few observed episodes, conclusions about the relative durations of positive and negative events are hard to draw; however, the model does show a tendency for negative events to be of longer duration. This is consistent with the observed results of Dole and Gordon (1983) and the ideas expressed by Palmer (1988b) that the negative PNA phase is more barotropically unstable, allowing the growth of perturbations with greater amplitude.

Although at a maximum on ENSO timescales, the coherence of model and observed PNA time series is still high on longer timescales (see Fig. 7). Figure 9a shows the ensemble mean (solid line) and the observed time series (dashed line) after filtering to remove fluctuations on timescales of less than 8 yr. The filtered Niño3 SST time series is also shown (dotted line). All three time series show a shift toward positive values over the period 1977–87, although this is considerably less for the ensemble mean values (amplitude less than half that of the observed PNA time series). The period 1977–87 has often been referred to as one of climate shift in the North Pacific (e.g., Graham 1994). The most striking change in this period was a deepening of the Aleutian low in the winter season (e.g., Nitta and Yamada 1989), corresponding to a shift toward the positive phase of the PNA (though this was not seen in every year). Concurrent changes in surface winds, in sea-ice extent and ocean temperatures, and in temperature and precipitation over the United States were also observed (e.g., UNESCO 1992; Yamagata and Masumoto 1992; Trenberth and Hurrell 1994). Analyzing the model runs, we find that 54% of the variance in the filtered time series can be attributed to oceanic forcing. The corresponding expected correlation (rEM PM) between the ensemble mean and the observed filtered time series is 0.69; the actual correlation is slightly higher at 0.73. With the few degrees of freedom in the filtered data, this correlation is on the borderline of significance at the 95% level. It is therefore difficult to claim that the model reproduces the observed decadal fluctuation. However, the decadal anomalies of modeled and observed surface temperature in the NPA sector shown in Figs. 9b and 9c show that the model does capture the pattern of anomalies well, although again the anomalies are considerably weaker than those observed, as are those found by Kumar et al. (1994) using a different model.

Further analysis reveals why the model did not pick up the strong average PNA state in 1977–87 with adequate amplitude. The model simulates the positive PNA response in warm event years (1977, 1983, 1987) well but fails to capture the observed PNA response in the non-ENSO years (1978, 1981, 1984, 1986) (see Fig. 6a). Taking the average of the JFM SSTs in these non-ENSO years reveals the only region with strong anomalies to be the North Pacific, where the east–west dipole pattern usually seen in conjunction with warm and cold events is apparent. As previously discussed, this pattern of SSTs may be primarily the effect, rather than the cause, of the PNA pattern. So, on the one hand, the strong PNA response in non-ENSO years may be unrelated to direct SST forcing and be solely the result of random variations in the extratropical atmosphere [including patterns centered on the North Pacific and North Atlantic (Hurrell 1996)]. However, the strength of the observed PNA index in these non-ENSO years is unusual, outside the range of the ensemble spread, and this may alternatively be partly due to a possible failure of the model to respond correctly to the prescribed extratropical SSTs (as discussed above), or indeed to non-ENSO-related tropical SSTs.

c. Influence of SSTs on the West Pacific mode

In Fig. 10a the model and observed seasonal time coefficients of the WP mode are shown. The poor agreement between the different model runs indicates that the SST forcing has little influence on the model mode, and the correlation between the ensemble mean and the observed time coefficients is correspondingly low, less than expected (but not significantly so) given the level of variance explained (see Table 3). Cross-spectral analysis fails to reveal any timescale on which the observed and modeled JFM WP modes are significantly coherent.

Figures 10b and 10c show correlations of GISST data with the seasonal WP time coefficients using observed and ensemble mean values, respectively. Had the model’s WP pattern been responding to tropical Pacific SSTs correctly, then its correlations with GISST would have been generally stronger than the observed, as for the PNA mode, due to the higher signal-to-noise ratio of ensemble mean values. Instead we find the model correlations to be weaker and generally insignificant, while there is a distinct El Niño–like pattern of strong, significant correlations in the observed analysis.

Figure 10d shows PDEs of the observed filtered daily WP time coefficients in warm event, cold event, and all years, calculated as for the PNA mode. The main impression is of a stronger response in warm event years than the PNA mode, with a notable shift toward the positive WP phase and also a moderate shift toward the negative WP phase in cold event years. The KS test shows that the observed warm and cold event distributions are significantly different from one another at the 99% confidence level and that each is significantly different from the all years distribution at the 99% and 95% levels, respectively. The corresponding distributions for the model coefficients, Fig. 10e, show PDEs in warm event and cold event years to be very similar;their difference is not significant.

In summary, the model simulates the spatial pattern of the WP mode generally well, although it does differ from the observed mode over the American continent. However, the model mode shows no significant dependence on tropical Pacific SSTs, and little dependence on SSTs in general, while the observed WP mode is strongly influenced by tropical Pacific SSTs. Not surprisingly, therefore, the interannual evolution of the observed and modeled modes are poorly correlated. The link between warm events and the positive phase of the WP mode in nature has been identified in earlier studies such as Livezey and Mo (1987) and Horel and Wallace (1981). As discussed by Hoerling et al. (1992), the extratropical response to ENSO is sensitive to errors in the model climatology. It appears that in our model there is some error in the chain that should couple tropical Pacific SSTs and extratropical anomalies in the western Pacific. Such an error would also explain the discrepancies between model and nature seen in our composites over the northwest Pacific (Fig. 2).

d. Influence of SSTs on the East Pacific mode

In Fig. 11a the seasonally averaged time coefficients of the EP mode are shown. Agreement between different model runs is very poor, and agreement between ensemble mean and observed EP time coefficients is correspondingly low (Table 3), indicating interannual variations of this pattern are probably random. Cross-spectral analysis reveals no timescale on which observed and modeled time series are significantly coherent.

Correlating the EP time coefficients with the GISST dataset reveals few significant values, as seen in Figs. 11b and 11c. The strongest correlations of the observed analysis are with SSTs in the North Atlantic. This banded pattern of correlations is a well-known mode of Atlantic SSTs (Lau and Nath 1990; Kushnir 1994), usually associated with the North Atlantic Oscillation (e.g., Davies et al. 1997). Model correlations form a similar pattern in this North Atlantic region but are rather weaker, significantly so to the south of Greenland. Over the rest of the world’s oceans there are few areas of significant difference between the two correlation maps. Figures 11d and 11e show the PDEs for the observed and modeled modes, respectively. There is good agreement between their “all year” PDEs; both are slightly skewed toward small negative values and have more high-amplitude positive cases. The PDEs for warm event years differ little from those for cold event years for model values, and although there appears to be a larger shift for observed values, KS tests indicate that neither difference is significant at the 95% level.

In summary, the impact of SSTs on the EP mode appears negligible. In particular, tropical Pacific SSTs do not have a discernable influence. There is some suggestion of a link with Atlantic SSTs, particularly in the observed data, but the question of cause and effect has not been addressed. The model gives a good simulation both of the EP pattern and of its independence of tropical Pacific SSTs.

e. Influence of SSTs on the North Pacific mode

Figure 12a shows the seasonal time coefficients of the NP mode. Again there is little agreement between model runs, and correlation between the ensemble mean and the observed time coefficients is not significantly different from zero (see Table 3). Cross-spectral analysis reveals no timescales on which observed and modeled time series are significantly coherent.

Figure 12b shows correlations between GISST data and the observed seasonal NP time coefficients. There are some relatively high correlations in the North Pacific, approximately coincident with the NP nodes. Such correlations may reflect the influence of the mode on the SSTs in addition to, or instead of, the reverse influence. In the Tropics, the correlations follow an El Niño pattern, although only a few areas are significant. The model correlations (Fig. 12c) show little correspondence to the observed and few areas of significance. The observed PDEs (Fig. 12d) show a preference for the positive NP phase in warm event years, and a slightly smaller shift to the negative phase in cold event years; the KS test indicates that the two sets of time coefficients are significantly different at the 99% level, but only the warm event distribution is significantly different from the all years distribution. Looking at the number of NP episodes of different durations (not shown) confirms the influence of ENSO on the observed NP mode. Neither the model correlation map, shown in Fig. 12c, nor its PDEs, shown in Fig. 12e, reveal any dependence of the model’s NP mode on tropical east Pacific SSTs.

Although the model and observed NP EOF patterns are in very good agreement, the former is little influenced by SSTs, while the latter is clearly subject to influence by tropical Pacific SSTs. This obviously points to a failure in the simulated atmospheric response.

5. Summary and conclusions

We have studied an ensemble of six 45-yr runs of the Hadley Centre’s HADAM1 atmospheric climate model. The runs are for the period 1949–93 and are forced with observed SSTs and sea ice from the GISST1.1 dataset. Our focus has been the atmospheric response to ENSO over the North Pacific–American (NPA) sector.

Seasonal composites of boreal wintertime (JFM) MSLP show that the response to ENSO is reasonably well simulated throughout the globe, with interannual variations of the SOI being particularly well captured. The model composites, with their larger sample sizes, also serve to highlight weak ENSO responses, such as over the North Atlantic and Antarctica. JFM composites for the NPA sector also show that the model and observed atmospheres respond similarly to ENSO, although the modeled anomaly centers in the North Pacific are offset from those observed.

In section 4a, we showed that the patterns of the leading JFM modes of variability in the NPA sector are reasonably well simulated by the HADAM1 model. These modes were identified on their natural timescales by analyzing daily observed and modeled 500-hPa height data, low-pass filtered to remove synoptic timescale variations. Four leading modes were identified for the NPA sector: the Pacific–North American (PNA), west Pacific (WP), east Pacific (EP), and North Pacific (NP). Further analysis is required to investigate the statistical and physical reasons for the absence among these leading modes of the Tropical–Northern Hemisphere pattern, which has been more prominent in other studies. In sections 4b–e, a correlation analysis of SST with the seasonally averaged (JFM) time coefficients of our four modes revealed a strong influence of ENSO on three of those observed: the PNA, WP, and NP. This was confirmed by analysis of the daily filtered REOF time coefficients; significant shifts (from all years) are seen in the distributions of the PNA time coefficients in cold event years, the WP coefficients in warm and cold event years, and the NP coefficients in warm event years. The model, however, fails to reproduce the influence of ENSO on the WP and NP modes, but the influence of ENSO on the PNA mode is captured, except that it is too linear. With analysis of variance techniques we can quantify the strength of the influence of SSTs on the model’s PNA mode, with the finding that almost half the variance of the seasonal means is attributable to oceanic forcing (47%). We also find that, given the limits due to random variations, the temporal evolution of the modeled and observed PNA patterns are in excellent agreement. Furthermore, since ENSO has its strongest impact on the PNA and WP modes (Figs. 6b and 10b), the seasonal ENSO composites are seen to resemble a combination of these patterns, with differences between the modeled and observed composites largely reflecting errors in the model modes and their responses to tropical east Pacific SSTs. However, we also note that such composites (modeled and observed) are sensitive to the SST index used; for example, Livezey et al. (1997) employ an SST area sited farther west and obtain a somewhat different response pattern (at 700 hPa) in warm event years. Future studies should investigate this sensitivity further.

In nature there was a clear preference for the positive phase of the PNA over the period 1977–87, which resulted in strong decadal anomalies in the NPA sector, very similar to warm event composite anomalies. The years 1977–87 include three strong warm events and only one counteracting weak cold event, thus biasing the decadal average toward the positive phase of the PNA; but the non-ENSO years of this period also make a strong contribution to the observed decadal anomalies. The model response over the same period reproduces the pattern of decadal anomalies well, corresponding to the positive phase of the PNA, but with a magnitude that is much to weak. This is because while capturing a PNA response in warm event years, it fails to do so in the non-ENSO years. These differences in model and observed response in non-ENSO years can largely be attributed to the internal variability of the atmosphere, although it is possible that a failure by the model to respond correctly to extratropical SST anomalies could also have some effect. The decadal anomalies in the North Pacific demonstrate that long timescale variations in modes such as the PNA can have large impacts on regional temperatures. The ability to simulate and understand such variations is important for studies of anthropogenic climate change. In the future, such decadal variations will be studied in 125-yr-long integrations forced with observed SSTs, which are currently under way at the Hadley Centre.

The strong relationship between ENSO and weather patterns in the NPA sector make useful seasonal forecasts for this region viable in the winter half-year (e.g., Livezey 1990). Davies et al. (1997) have shown that the model discussed here is also skilful in simulating the leading modes of variability in the North Atlantic–European (NAE) sector. In the future we also intend to analyze these NAE modes of variability on their natural timescales in terms of the distribution of time coefficients, using the techniques described in this paper. This will help in understanding both the prospects for seasonal forecasting in this region and the long timescale variations of modes such as the North Atlantic Oscillation, and its impact on climate.

Acknowledgments

Thanks are due J. Davies and D. Sexton for helping to run the model integrations. Global-observed MSLP data were kindly provided by T. Basnett and D. E. Parker, Tahiti and Darwin MSLP data were supplied by Phil Jones, and D. P. Cullum provided the air temperature data. Discussions with R. E. Livezey were also very much appreciated. This work was carried out under the U.K. Department of Environment Contract PECD/7/12/37, with further partial support under CEC Contract EVSV-CT92-0121 (“Medium Term Climate Variability”).

REFERENCES

  • Alexander, M. A., 1992a: Midlatitude atmosphere–ocean interaction during El Niño. Part I: The North Pacific Ocean. J. Climate,5, 944–958.

  • ——, 1992b: Midlatitude atmosphere–ocean interaction during El Niño. Part II. The Northern Hemisphere atmosphere. J. Climate,5, 959–972.

  • Allan, R., J. Lindesay, and D. E. Parker, 1996: El Niño Southern Oscillation and Climatic Variability. CSIRO, 416 pp.

  • Barnett, T. P., and Coauthors, 1994: Forecasting global ENSO-related climate anomalies. Tellus,46A, 381–397.

  • Barnston, A. G., 1994: Linear statistical short-term climate predictive skill in the Northern Hemisphere. J. Climate,7, 1513–1564.

  • ——, 1995: Our improving capability in ENSO forecasting. Weather,50, 419–430.

  • ——, and R. E. Livezey, 1987: Classification, seasonality and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev.,115, 1083–1126.

  • Basnett, T. A., and D. E. Parker, 1997: Development of the global mean sea level pressure data set GMSLP2. Climate Res. Tech. Note No. 79, 54 pp. [Available from the Hadley Centre, U.K. Met Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev.,97, 163–172.

  • Blackmon, M. L., J. E. Giesler, and E. J. Pitcher, 1983: A general circulation study of January climate anomaly patterns associated with interannual variation of equatorial Pacific SSTs. J. Atmos. Sci.,40, 1410–1425.

  • Bloomfield, P., 1976: Fourier Analysis of Time Series: An Introduction. John Wiley and Sons, 258 pp.

  • Bottomley, M., C. K. Folland, J. Hsiung, R. E. Newell, and D. E. Parker, 1990: Global ocean surface temperature atlas. Joint Met. Office/MIT Project, U.K. Departments of Energy and Environment, HMSO, London, United Kingdom, 20 pp. + 313 plates.

  • Chen, W. Y., 1982: Fluctuations in Northern Hemisphere 700 mb height fields associated with the Southern Oscillation. Mon. Wea. Rev.,110, 808–823.

  • Cheng, X., G. Nitsche, and J. M. Wallace, 1995: Robustness of low-frequency circulation patterns derived from EOF and rotated EOF analyses. J. Climate,8, 1709–1713.

  • Chervin, R. M., 1986: Interannual variability and seasonal climate predictability. J. Atmos. Sci.,43, 233–251.

  • Cubasch, U., 1985: The mean response of the ECMWF global model to the composite El Niño anomaly in the extended range prediction experiments. Coupled Ocean Atmosphere Models, J. Nihoul, Ed., Elsevier, 329–344.

  • Davies, J. R., D. P. Rowell, and C. K. Folland, 1997: North Atlantic European seasonal predictability using an ensemble of multi-decadal AGCM simulations. Int. J. Climatol.,17, 1263–1286.

  • Dole, R. M., and N. D. Gordon, 1983: Persistent anomalies of the extratropical Northern Hemisphere wintertime circulation: Geographical distribution and regional persistence characteristics. Mon. Wea. Rev.,111, 1567–1586.

  • Folland, C. K., J. A. Owen, M. N. Ward, and A. W. Colman, 1991: Prediction of seasonal rainfall in the Sahel region using empirical and dynamical methods. J. Forecasting,10, 21–56.

  • Fraedrich, K., and K. Muller, 1992:Climate anomalies in Europe associated with ENSO extremes. Int. J. Climatol.,12, 25–31.

  • Frankignoul, C. J., 1985: Sea surface temperature anomalies, planetary waves and air–sea feedback in the middle latitudes. Rev. Geophys.,23, 357–390.

  • Graham, N. E., 1994: Decadal-scale climate variability in the tropical and North Pacific during the 1970s and 1980s: Observations and model results. Climate Dyn.,10, 135–162.

  • ——, T. P. Barnett, R. Wilde, M. Ponater, and S. Schubert, 1994: On the roles of tropical and midlatitude SSTs in forcing interannual to interdecadal variability in the winter Northern Hemisphere circulation. J. Climate,7, 1416–1441.

  • Held, I. M., S. W. Lyons, and S. Nigam, 1989: Transients and the extratropical response to ENSO. J. Atmos. Sci.,46, 163–174.

  • Hildebrandsson, H. H., 1897: Quelques recherches sur les centres d’action de l’atmosphere. Svenska Vet. Akad. Handlingar, 36 pp. + 7 plates.

  • Hoel, P. G., 1971: Introduction to Mathematical Statistics. John Wiley, 409 pp.

  • Hoerling, M. P., and M. Ting, 1994: Organization of extratropical transients during El Niño. J.Climate,7, 745–766.

  • ——, M. L. Blackmon, and M. Ting, 1992: Simulating the atmospheric response to the 1985–87 El Niño cycle. J. Climate,5, 669–682.

  • Horel, J. D., and J. M. Wallace, 1981: Planetary scale phenomenon associated with the Southern Oscillation. Mon. Wea. Rev.,109, 813–829.

  • Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci.,38, 1179–1196.

  • Hulme, M., 1994: Validation of large scale precipitation fields in general circulation models. Global Precipitation and Climate Change, M. Debois and F. Desalmaud, Eds., NATO AS1 Series, Vol. I 26, Springer-Verlag, 387–405.

  • Hurrell, J. W., 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett.23, 665–668.

  • Jones, P. D., 1994: Hemispheric surface air temperature variations: A reanalysis and an update to 1995. J. Climate,7, 1794–1802.

  • Kiladis, N. K., and H. F. Diaz, 1989: Global climatic anomalies associated with extremes in the Southern Oscillation. J. Climate,2, 1069–1090.

  • Konig, W., E. Kirk, and R. Sausen, 1990: Sensitivity of an atmospheric general circulation model to interannually varying SST. Ann. Geophys.,8, 829–844.

  • Kumar, A., A. Leetmaa, and M. Ji, 1994: Simulations of atmospheric variability induced by sea surface temperatures and implications for global warming. Science,266, 632–634.

  • Kushnir, Y., 1994: Interdecadal variations in the North Atlantic sea surface temperature and associated atmospheric conditions. J. Climate,7, 141–157.

  • Latif, M., J. Biercamp, H. von Storch, M. J. McPhaden, and E. Kirk, 1990: Simulation of ENSO related surface wind anomalies with an atmospheric GCM forced by observed SST. J. Climate,3, 509–521.

  • Lau, N. C., 1981: A diagnostic study of recurrent meteorological anomalies appearing in a 15 year simulation with a GFDL general circulation model. Mon. Wea. Rev.,109, 2287–2311.

  • ——, 1985: Modeling the seasonal dependence of the atmospheric response to observed El Niños in 1962–76. Mon. Wea. Rev.,113, 1970–1996.

  • ——, 1988: Variability of the baroclinic and barotropic transient eddy forcing associated with monthly changes in midlatitiude storm tracks. J. Atmos. Sci.,45, 2718–2743.

  • ——, and M. J. Nath, 1990: A general circulation model study of the atmospheric response to extratropical SST anomalies observed in 1950–79. J. Climate,3, 990–998.

  • ——, and ——, 1994: A modeling study of the relative roles of tropical and extratropical SST anomalies in the variability of the global atmosphere–ocean system. J. Climate,7, 1184–1207.

  • Livezey, R. E., 1990: Variability of skill of long-range forecasts and implications for their use and value. Bull. Amer. Meteor. Soc.,71, 300–309.

  • ——, and K. C. Mo, 1987: Tropical–extratropical teleconnections during the Northern Hemisphere winter. Part II: Relations between monthly mean Northern Hemisphere circulation patterns and proxies for tropical convection. Mon. Wea. Rev.,115, 3115–3132.

  • ——, C. K. Folland, J. Davies, D. Karoly, J. Walsh, M. Sugi, P. Gleckler, and D. E. Parker, 1995: AGCM intercomparison diagnostics: Key time series. Climate Research Tech. Note 56, Workshop Report. [Available from the Hadley Centre, U.K. Met Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • ——, M. Masutani, A. Leetmaa, H. Rui, M. Ji, and A. Kumar, 1997:Teleconnective response of the Pacific–North American region atmosphere to large central equatorial Pacific SST anomalies. J. Climate,10, 1787–1820.

  • Lorenz, E. N., 1963: Deterministic nonperiodic flow. J. Atmos. Sci.,20, 130–141.

  • Miller, A. J., D. R. Cayan, T. P. Barnett, N. E. Graham, and J. M. Oberhuber, 1994: Interdecadal variability of the Pacific Ocean: Model response to observed heat flux and wind stress anomalies. Climate Dyn.,9, 287–302.

  • Mo, K. C., and R. E. Livezey, 1986: Tropical–extratropical geopotential height teleconnections during the Northern Hemisphere winter. Mon. Wea. Rev.,114, 2488–2515.

  • Molteni, F., S. Tibaldi, and T. N. Palmer, 1990: Regimes in the wintertime circulation over the northern extratropics. Observational evidence. Quart. J. Roy. Meteor. Soc.,116, 31–67.

  • 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, 375–383.

  • Palmer, T. N., 1988a: Large scale tropical–extratropical interactions on timescales of a few days to a season. Aust. Meteor. Mag.,36, 107–125.

  • ——, 1988b: Medium and extended range predictability and stability of the Pacific North American mode. Quart. J. Roy. Meteor. Soc.,114, 691–714.

  • ——, 1993: Extended range atmospheric prediction and the Lorenz model. Bull. Amer. Meteor. Soc.,74, 49–65.

  • ——, 1995: The influence of north-west Atlantic sea surface temperature: An unplanned experiment. Weather,50, 413–419.

  • Pan, Y. H., and A. H. Oort, 1983: Global climate variations connected with sea surface temperature anomalies in the eastern equatorial Pacific Ocean for the 1958–73 period. Mon. Wea. Rev.,111, 1244–1258.

  • Parker, D. E., C. K. Folland, A. C. Bevan, M. N. Ward, M. Jackson, and K. Maskell 1995a: Marine surface data for analysis of climatic fluctuations on interannual to century timescales. Natural Climate Variability on Decade-to-Century Timescales, D. G. Martinson et al., Eds., National Acad. Press, 241–250 + plates 222–228.

  • ——, ——, and M. Jackson, 1995b: Marine surface temperature: Observed variations and data requirements. Climate Change,31, 559–600.

  • Phillips, T. J., 1994: A summary documentation of the AMIP models. PCMDI Rep. 18, University of California, 343 pp. [Available from Lawrence Livermore National Laboratory, University of California, Livermore, CA 94550.].

  • Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, 1992: Numerical Recipes in FORTRAN. Cambridge University Press, 963 pp.

  • Richman, M. B., 1994: On the use of multivariate statistics to validate global climate models. Ph.D. thesis, University of Illinois, 321 pp. [Available from Illinois State Water Survey, Champaign, IL 61820-7407.].

  • Rogers, J. C., 1984: The association between the North Atlantic Oscillation and the Southern Oscillation in the Northern Hemisphere. Mon. Wea. Rev.,112, 1999–2015.

  • Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev.,114, 2352–2362.

  • ——, and ——, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev.,115, 1606–1625.

  • ——, and P. D. Jones, 1987: An extension of the Tahiti–Darwin Southern Oscillation index. Mon. Wea. Rev.,115, 2161–2165.

  • ——, and M. S. Halpert, 1989: Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Climate,2, 268–284.

  • Rowell, D. P, 1996: Using an ensemble of multi-decadal GCM simulations to assess potential seasonal predictability. Climate Res. Tech. Note 69, 39 pp. [Available from the Hadley Centre, U.K. Met Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • ——, 1998: Assessing potential seasonal predictability with an ensemble of multi-decadal GCM simulations. J. Climate,11, 109–120.

  • ——, C. K. Folland, K. Maskell, and M. N. Ward: 1995: Variability of summer rainfall over tropical North Africa (1906–92): Observations and modelling. Quart. J. Roy. Meteor. Soc.,121, 669–704.

  • Rowntree, P. R., 1972: The influence of tropical east Pacific Ocean temperatures on the atmosphere. Quart. J. Roy. Meteor. Soc.,98, 607–625.

  • 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.

  • Shukla, J., and J. M. Wallace, 1983: Numerical simulation of the atmospheric response to equatorial Pacific sea surface temperature anomalies. J. Atmos. Sci.,40, 1613–1630.

  • Smith, I. N., 1995: A GCM simulation of global climate interannual variability 1950–1988. J. Climate,8, 709–718.

  • Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn.,9, 303–319.

  • UNESCO, 1992: Oceanic interdecadal variability. IOC Tech. Ser. 40, 40 pp.

  • Walker, G. T., 1924: Correlation in seasonal variations of weather. Part IX: A further study of world weather. Mem. Indian Meteor. Dept.,24, 275–332.

  • 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.

  • Walraven, R., 1984: Digital filters. Proc. Digital Equipment Users Society, Anaheim, CA, Dept. of Applied Science, University of California, Davis, 23–30.

  • Warrilow, D. A., A. B. Sangster, and A. Slingo, 1986: Modelling of land surface processes and their influence on European climate. Dyn. Climatol. Tech. Note 38, 92 pp. [Available from the Hadley Centre, U.K. Met. Office, Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • Willmot, C. J., C. M. Rowe, and Y. Mintz, 1985: A global archive of land cover and sils data for use in general circulation climate models. J. Climatol.,5, 119–143.

  • WMO, 1966: Climate change. WMO Tech. Note 79, 79 pp. [Available from World Meteorological Organization, 41 Giuseppe-Motta, Case Postale 2300, CH-1211 Geneva 2, Switzerland.].

  • Yamagata, T., and Y. Masumoto, 1992: Interdecadal natural climate variability in the western Pacific and its implication in global warming. J. Meteor. Soc. Japan,70, 167–175.

  • Yarnal, B., and H. F. Diaz, 1986: Relationships between extremes of the Southern Oscillation and the winter climate of the Anglo-American Pacific coast. J. Climatol.,6, 197–219.

  • Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability. J. Climate,10, 1004–1020.

  • Zwiers, F. W., and H. von Storch, 1995: Taking serial correlation into account in tests of the mean. J. Climate,8, 336–351.

Fig. 1.
Fig. 1.

Modeled and observed SOI, 1949–93. Calculated as the Tahiti minus Darwin difference in 5-month running mean values of normalized MSLP anomalies. Each solid line shows the SOI of an individual model integration. Gray shading shows the observed SOI.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 2.
Fig. 2.

Composite MSLP anomalies, wrt 1949–93 mean values, for JFM. (a) Warm event (El Niño) observed composite; (b) warm event (El Niño) model composite; (c) cold event (La Niña) observed composite; (d) cold event (La Niña) model composite. Positive anomalies have solid contours, negative anomalies have dashed contours. Areas of significance at the 95% level are shaded, and the proportion of the globe with such significance is shown in brackets.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 3.
Fig. 3.

(a) Observed and (b) modeled warm event (El Niño) minus cold event (La Niña) composite differences of 500-hPa height, for the winter season JFM in the North Pacific–American sector. Positive anomalies have solid contours, negative anomalies are dashed. Areas of significant difference (at the 95% level) are shaded. Cross-hatching indicates missing data. (c) and (d) As (a) and (b) but for surface air temperature.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 4.
Fig. 4.

Response function of the low-pass filter applied to daily 500-hPa height data prior to REOF analysis.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 5.
Fig. 5.

Leading rotated EOFS in the NPA sector, calculated from filtered 500-hPa height data for JFM (1970–93). (a) Observed PNA mode; (b) model PNA mode; (c) observed WP mode; (d) model WP mode; (e) observed EP mode; (f) model EP mode; (g) observed NP mode; (h) model NP mode.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 6.
Fig. 6.

Seasonally averaged time coefficients of the PNA mode for JFM. Each dotted line represents an individual model integration (1949–93), the dashed line is the ensemble mean, and the solid line the observed values (1968–93). (b) Correlation between GISST data and the seasonally averaged time coefficients of the observed PNA mode. Solid contours indicate positive correlations, dashed contours negative. Areas of significant correlation (at the 95% level) are shaded. (c) As (b) but for the ensemble mean model PNA mode. (d) Probability distribution of daily time coefficients of the observed PNA mode. Solid lines are calculated using data from all years; dotted lines use data from warm event (El Niño) years; and dashed lines use data from cold event (La Niña) years. (e) As (d) but for the model PNA mode.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 7.
Fig. 7.

Coherence between modeled and observed seasonally averaged (JFM) PNA time coefficients. Dashed line indicates 95% significance level.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 8.
Fig. 8.

Histogram of the duration of PNA episodes in JFM 1968–93, measured as the number of consecutive days for which the filtered daily time coefficient exceeds |1.25|: (a) positive PNA episodes in warm event years, (b) positive PNA episodes in cold event years, (c) negative PNA episodes in warm event years and (d) negative PNA episodes in cold event years. Heavy thin bars show model values, thicker lighter bars show observed values.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 9.
Fig. 9.

(a) Filtered time series (low pass, >8 yr) showing decadal variations of seasonally averaged (JFM) PNA time coefficients, 1968–93, observed (dashed line) and modeled (solid line). Dotted line shows seasonally averaged (JFM) filtered Niño3 SSTs. (b) Observed and (c) modeled JFM decadal anomaly (1977–87) of surface air temperature in the North Pacific–American region.

Citation: Journal of Climate 11, 5; 10.1175/1520-0442(1998)011<1073:WLFWVI>2.0.CO;2

Fig. 10.
Fig. 11.
Fig. 12.

Table 1.

Years used to create warm and cold event composites.

Table 1.
Table 2.

Characteristics of the patterns of observed and modeled NPA modes.

Table 2.
Table 3.

Influence of SSTs on, and skill of, seasonally averaged time coefficients of each of the NPA modes. Here ρ is the percentage of variance explained by oceanic forcing; rEM is the correlation between ensemble mean and observed values; and rEM PM is the expected correlation for a perfect model.

Table 3.
Save
  • Alexander, M. A., 1992a: Midlatitude atmosphere–ocean interaction during El Niño. Part I: The North Pacific Ocean. J. Climate,5, 944–958.

  • ——, 1992b: Midlatitude atmosphere–ocean interaction during El Niño. Part II. The Northern Hemisphere atmosphere. J. Climate,5, 959–972.

  • Allan, R., J. Lindesay, and D. E. Parker, 1996: El Niño Southern Oscillation and Climatic Variability. CSIRO, 416 pp.

  • Barnett, T. P., and Coauthors, 1994: Forecasting global ENSO-related climate anomalies. Tellus,46A, 381–397.

  • Barnston, A. G., 1994: Linear statistical short-term climate predictive skill in the Northern Hemisphere. J. Climate,7, 1513–1564.

  • ——, 1995: Our improving capability in ENSO forecasting. Weather,50, 419–430.

  • ——, and R. E. Livezey, 1987: Classification, seasonality and persistence of low-frequency atmospheric circulation patterns. Mon. Wea. Rev.,115, 1083–1126.

  • Basnett, T. A., and D. E. Parker, 1997: Development of the global mean sea level pressure data set GMSLP2. Climate Res. Tech. Note No. 79, 54 pp. [Available from the Hadley Centre, U.K. Met Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev.,97, 163–172.

  • Blackmon, M. L., J. E. Giesler, and E. J. Pitcher, 1983: A general circulation study of January climate anomaly patterns associated with interannual variation of equatorial Pacific SSTs. J. Atmos. Sci.,40, 1410–1425.

  • Bloomfield, P., 1976: Fourier Analysis of Time Series: An Introduction. John Wiley and Sons, 258 pp.

  • Bottomley, M., C. K. Folland, J. Hsiung, R. E. Newell, and D. E. Parker, 1990: Global ocean surface temperature atlas. Joint Met. Office/MIT Project, U.K. Departments of Energy and Environment, HMSO, London, United Kingdom, 20 pp. + 313 plates.

  • Chen, W. Y., 1982: Fluctuations in Northern Hemisphere 700 mb height fields associated with the Southern Oscillation. Mon. Wea. Rev.,110, 808–823.

  • Cheng, X., G. Nitsche, and J. M. Wallace, 1995: Robustness of low-frequency circulation patterns derived from EOF and rotated EOF analyses. J. Climate,8, 1709–1713.

  • Chervin, R. M., 1986: Interannual variability and seasonal climate predictability. J. Atmos. Sci.,43, 233–251.

  • Cubasch, U., 1985: The mean response of the ECMWF global model to the composite El Niño anomaly in the extended range prediction experiments. Coupled Ocean Atmosphere Models, J. Nihoul, Ed., Elsevier, 329–344.

  • Davies, J. R., D. P. Rowell, and C. K. Folland, 1997: North Atlantic European seasonal predictability using an ensemble of multi-decadal AGCM simulations. Int. J. Climatol.,17, 1263–1286.

  • Dole, R. M., and N. D. Gordon, 1983: Persistent anomalies of the extratropical Northern Hemisphere wintertime circulation: Geographical distribution and regional persistence characteristics. Mon. Wea. Rev.,111, 1567–1586.

  • Folland, C. K., J. A. Owen, M. N. Ward, and A. W. Colman, 1991: Prediction of seasonal rainfall in the Sahel region using empirical and dynamical methods. J. Forecasting,10, 21–56.

  • Fraedrich, K., and K. Muller, 1992:Climate anomalies in Europe associated with ENSO extremes. Int. J. Climatol.,12, 25–31.

  • Frankignoul, C. J., 1985: Sea surface temperature anomalies, planetary waves and air–sea feedback in the middle latitudes. Rev. Geophys.,23, 357–390.

  • Graham, N. E., 1994: Decadal-scale climate variability in the tropical and North Pacific during the 1970s and 1980s: Observations and model results. Climate Dyn.,10, 135–162.

  • ——, T. P. Barnett, R. Wilde, M. Ponater, and S. Schubert, 1994: On the roles of tropical and midlatitude SSTs in forcing interannual to interdecadal variability in the winter Northern Hemisphere circulation. J. Climate,7, 1416–1441.

  • Held, I. M., S. W. Lyons, and S. Nigam, 1989: Transients and the extratropical response to ENSO. J. Atmos. Sci.,46, 163–174.

  • Hildebrandsson, H. H., 1897: Quelques recherches sur les centres d’action de l’atmosphere. Svenska Vet. Akad. Handlingar, 36 pp. + 7 plates.

  • Hoel, P. G., 1971: Introduction to Mathematical Statistics. John Wiley, 409 pp.

  • Hoerling, M. P., and M. Ting, 1994: Organization of extratropical transients during El Niño. J.Climate,7, 745–766.

  • ——, M. L. Blackmon, and M. Ting, 1992: Simulating the atmospheric response to the 1985–87 El Niño cycle. J. Climate,5, 669–682.

  • Horel, J. D., and J. M. Wallace, 1981: Planetary scale phenomenon associated with the Southern Oscillation. Mon. Wea. Rev.,109, 813–829.

  • Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci.,38, 1179–1196.

  • Hulme, M., 1994: Validation of large scale precipitation fields in general circulation models. Global Precipitation and Climate Change, M. Debois and F. Desalmaud, Eds., NATO AS1 Series, Vol. I 26, Springer-Verlag, 387–405.

  • Hurrell, J. W., 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett.23, 665–668.

  • Jones, P. D., 1994: Hemispheric surface air temperature variations: A reanalysis and an update to 1995. J. Climate,7, 1794–1802.

  • Kiladis, N. K., and H. F. Diaz, 1989: Global climatic anomalies associated with extremes in the Southern Oscillation. J. Climate,2, 1069–1090.

  • Konig, W., E. Kirk, and R. Sausen, 1990: Sensitivity of an atmospheric general circulation model to interannually varying SST. Ann. Geophys.,8, 829–844.

  • Kumar, A., A. Leetmaa, and M. Ji, 1994: Simulations of atmospheric variability induced by sea surface temperatures and implications for global warming. Science,266, 632–634.

  • Kushnir, Y., 1994: Interdecadal variations in the North Atlantic sea surface temperature and associated atmospheric conditions. J. Climate,7, 141–157.

  • Latif, M., J. Biercamp, H. von Storch, M. J. McPhaden, and E. Kirk, 1990: Simulation of ENSO related surface wind anomalies with an atmospheric GCM forced by observed SST. J. Climate,3, 509–521.

  • Lau, N. C., 1981: A diagnostic study of recurrent meteorological anomalies appearing in a 15 year simulation with a GFDL general circulation model. Mon. Wea. Rev.,109, 2287–2311.

  • ——, 1985: Modeling the seasonal dependence of the atmospheric response to observed El Niños in 1962–76. Mon. Wea. Rev.,113, 1970–1996.

  • ——, 1988: Variability of the baroclinic and barotropic transient eddy forcing associated with monthly changes in midlatitiude storm tracks. J. Atmos. Sci.,45, 2718–2743.

  • ——, and M. J. Nath, 1990: A general circulation model study of the atmospheric response to extratropical SST anomalies observed in 1950–79. J. Climate,3, 990–998.

  • ——, and ——, 1994: A modeling study of the relative roles of tropical and extratropical SST anomalies in the variability of the global atmosphere–ocean system. J. Climate,7, 1184–1207.

  • Livezey, R. E., 1990: Variability of skill of long-range forecasts and implications for their use and value. Bull. Amer. Meteor. Soc.,71, 300–309.

  • ——, and K. C. Mo, 1987: Tropical–extratropical teleconnections during the Northern Hemisphere winter. Part II: Relations between monthly mean Northern Hemisphere circulation patterns and proxies for tropical convection. Mon. Wea. Rev.,115, 3115–3132.

  • ——, C. K. Folland, J. Davies, D. Karoly, J. Walsh, M. Sugi, P. Gleckler, and D. E. Parker, 1995: AGCM intercomparison diagnostics: Key time series. Climate Research Tech. Note 56, Workshop Report. [Available from the Hadley Centre, U.K. Met Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • ——, M. Masutani, A. Leetmaa, H. Rui, M. Ji, and A. Kumar, 1997:Teleconnective response of the Pacific–North American region atmosphere to large central equatorial Pacific SST anomalies. J. Climate,10, 1787–1820.

  • Lorenz, E. N., 1963: Deterministic nonperiodic flow. J. Atmos. Sci.,20, 130–141.

  • Miller, A. J., D. R. Cayan, T. P. Barnett, N. E. Graham, and J. M. Oberhuber, 1994: Interdecadal variability of the Pacific Ocean: Model response to observed heat flux and wind stress anomalies. Climate Dyn.,9, 287–302.

  • Mo, K. C., and R. E. Livezey, 1986: Tropical–extratropical geopotential height teleconnections during the Northern Hemisphere winter. Mon. Wea. Rev.,114, 2488–2515.

  • Molteni, F., S. Tibaldi, and T. N. Palmer, 1990: Regimes in the wintertime circulation over the northern extratropics. Observational evidence. Quart. J. Roy. Meteor. Soc.,116, 31–67.

  • 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, 375–383.

  • Palmer, T. N., 1988a: Large scale tropical–extratropical interactions on timescales of a few days to a season. Aust. Meteor. Mag.,36, 107–125.

  • ——, 1988b: Medium and extended range predictability and stability of the Pacific North American mode. Quart. J. Roy. Meteor. Soc.,114, 691–714.

  • ——, 1993: Extended range atmospheric prediction and the Lorenz model. Bull. Amer. Meteor. Soc.,74, 49–65.

  • ——, 1995: The influence of north-west Atlantic sea surface temperature: An unplanned experiment. Weather,50, 413–419.

  • Pan, Y. H., and A. H. Oort, 1983: Global climate variations connected with sea surface temperature anomalies in the eastern equatorial Pacific Ocean for the 1958–73 period. Mon. Wea. Rev.,111, 1244–1258.

  • Parker, D. E., C. K. Folland, A. C. Bevan, M. N. Ward, M. Jackson, and K. Maskell 1995a: Marine surface data for analysis of climatic fluctuations on interannual to century timescales. Natural Climate Variability on Decade-to-Century Timescales, D. G. Martinson et al., Eds., National Acad. Press, 241–250 + plates 222–228.

  • ——, ——, and M. Jackson, 1995b: Marine surface temperature: Observed variations and data requirements. Climate Change,31, 559–600.

  • Phillips, T. J., 1994: A summary documentation of the AMIP models. PCMDI Rep. 18, University of California, 343 pp. [Available from Lawrence Livermore National Laboratory, University of California, Livermore, CA 94550.].

  • Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, 1992: Numerical Recipes in FORTRAN. Cambridge University Press, 963 pp.

  • Richman, M. B., 1994: On the use of multivariate statistics to validate global climate models. Ph.D. thesis, University of Illinois, 321 pp. [Available from Illinois State Water Survey, Champaign, IL 61820-7407.].

  • Rogers, J. C., 1984: The association between the North Atlantic Oscillation and the Southern Oscillation in the Northern Hemisphere. Mon. Wea. Rev.,112, 1999–2015.

  • Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev.,114, 2352–2362.

  • ——, and ——, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev.,115, 1606–1625.

  • ——, and P. D. Jones, 1987: An extension of the Tahiti–Darwin Southern Oscillation index. Mon. Wea. Rev.,115, 2161–2165.

  • ——, and M. S. Halpert, 1989: Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Climate,2, 268–284.

  • Rowell, D. P, 1996: Using an ensemble of multi-decadal GCM simulations to assess potential seasonal predictability. Climate Res. Tech. Note 69, 39 pp. [Available from the Hadley Centre, U.K. Met Office, London Rd., Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • ——, 1998: Assessing potential seasonal predictability with an ensemble of multi-decadal GCM simulations. J. Climate,11, 109–120.

  • ——, C. K. Folland, K. Maskell, and M. N. Ward: 1995: Variability of summer rainfall over tropical North Africa (1906–92): Observations and modelling. Quart. J. Roy. Meteor. Soc.,121, 669–704.

  • Rowntree, P. R., 1972: The influence of tropical east Pacific Ocean temperatures on the atmosphere. Quart. J. Roy. Meteor. Soc.,98, 607–625.

  • 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.

  • Shukla, J., and J. M. Wallace, 1983: Numerical simulation of the atmospheric response to equatorial Pacific sea surface temperature anomalies. J. Atmos. Sci.,40, 1613–1630.

  • Smith, I. N., 1995: A GCM simulation of global climate interannual variability 1950–1988. J. Climate,8, 709–718.

  • Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn.,9, 303–319.

  • UNESCO, 1992: Oceanic interdecadal variability. IOC Tech. Ser. 40, 40 pp.

  • Walker, G. T., 1924: Correlation in seasonal variations of weather. Part IX: A further study of world weather. Mem. Indian Meteor. Dept.,24, 275–332.

  • 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.

  • Walraven, R., 1984: Digital filters. Proc. Digital Equipment Users Society, Anaheim, CA, Dept. of Applied Science, University of California, Davis, 23–30.

  • Warrilow, D. A., A. B. Sangster, and A. Slingo, 1986: Modelling of land surface processes and their influence on European climate. Dyn. Climatol. Tech. Note 38, 92 pp. [Available from the Hadley Centre, U.K. Met. Office, Bracknell, Berkshire RG12 2SY, United Kingdom.].

  • Willmot, C. J., C. M. Rowe, and Y. Mintz, 1985: A global archive of land cover and sils data for use in general circulation climate models. J. Climatol.,5, 119–143.

  • WMO, 1966: Climate change. WMO Tech. Note 79, 79 pp. [Available from World Meteorological Organization, 41 Giuseppe-Motta, Case Postale 2300, CH-1211 Geneva 2, Switzerland.].

  • Yamagata, T., and Y. Masumoto, 1992: Interdecadal natural climate variability in the western Pacific and its implication in global warming. J. Meteor. Soc. Japan,70, 167–175.

  • Yarnal, B., and H. F. Diaz, 1986: Relationships between extremes of the Southern Oscillation and the winter climate of the Anglo-American Pacific coast. J. Climatol.,6, 197–219.

  • Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability. J. Climate,10, 1004–1020.

  • Zwiers, F. W., and H. von Storch, 1995: Taking serial correlation into account in tests of the mean. J. Climate,8, 336–351.

  • Fig. 1.

    Modeled and observed SOI, 1949–93. Calculated as the Tahiti minus Darwin difference in 5-month running mean values of normalized MSLP anomalies. Each solid line shows the SOI of an individual model integration. Gray shading shows the observed SOI.

  • Fig. 2.

    Composite MSLP anomalies, wrt 1949–93 mean values, for JFM. (a) Warm event (El Niño) observed composite; (b) warm event (El Niño) model composite; (c) cold event (La Niña) observed composite; (d) cold event (La Niña) model composite. Positive anomalies have solid contours, negative anomalies have dashed contours. Areas of significance at the 95% level are shaded, and the proportion of the globe with such significance is shown in brackets.

  • Fig. 3.

    (a) Observed and (b) modeled warm event (El Niño) minus cold event (La Niña) composite differences of 500-hPa height, for the winter season JFM in the North Pacific–American sector. Positive anomalies have solid contours, negative anomalies are dashed. Areas of significant difference (at the 95% level) are shaded. Cross-hatching indicates missing data. (c) and (d) As (a) and (b) but for surface air temperature.

  • Fig. 4.

    Response function of the low-pass filter applied to daily 500-hPa height data prior to REOF analysis.

  • Fig. 5.

    Leading rotated EOFS in the NPA sector, calculated from filtered 500-hPa height data for JFM (1970–93). (a) Observed PNA mode; (b) model PNA mode; (c) observed WP mode; (d) model WP mode; (e) observed EP mode; (f) model EP mode; (g) observed NP mode; (h) model NP mode.

  • Fig. 6.

    Seasonally averaged time coefficients of the PNA mode for JFM. Each dotted line represents an individual model integration (1949–93), the dashed line is the ensemble mean, and the solid line the observed values (1968–93). (b) Correlation between GISST data and the seasonally averaged time coefficients of the observed PNA mode. Solid contours indicate positive correlations, dashed contours negative. Areas of significant correlation (at the 95% level) are shaded. (c) As (b) but for the ensemble mean model PNA mode. (d) Probability distribution of daily time coefficients of the observed PNA mode. Solid lines are calculated using data from all years; dotted lines use data from warm event (El Niño) years; and dashed lines use data from cold event (La Niña) years. (e) As (d) but for the model PNA mode.

  • Fig. 7.

    Coherence between modeled and observed seasonally averaged (JFM) PNA time coefficients. Dashed line indicates 95% significance level.

  • Fig. 8.

    Histogram of the duration of PNA episodes in JFM 1968–93, measured as the number of consecutive days for which the filtered daily time coefficient exceeds |1.25|: (a) positive PNA episodes in warm event years, (b) positive PNA episodes in cold event years, (c) negative PNA episodes in warm event years and (d) negative PNA episodes in cold event years. Heavy thin bars show model values, thicker lighter bars show observed values.

  • Fig. 9.

    (a) Filtered time series (low pass, >8 yr) showing decadal variations of seasonally averaged (JFM) PNA time coefficients, 1968–93, observed (dashed line) and modeled (solid line). Dotted line shows seasonally averaged (JFM) filtered Niño3 SSTs. (b) Observed and (c) modeled JFM decadal anomaly (1977–87) of surface air temperature in the North Pacific–American region.

  • Fig. 10.

    As Fig. 6 but for WP mode.

  • Fig. 11.

    As Fig. 6 but for EP mode.

  • Fig. 12.

    As Fig. 6 but for NP mode.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 213 31 1
PDF Downloads 47 12 2