A Large Ensemble Analysis of the Influence of Tropical SSTs on Seasonal Atmospheric Variability

Peitao Peng Climate Prediction Center, Camp Springs, Maryland

Search for other papers by Peitao Peng in
Current site
Google Scholar
PubMed
Close
and
Arun Kumar Climate Prediction Center, Camp Springs, Maryland

Search for other papers by Arun Kumar in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Based on a 40-member ensemble for the January–March (JFM) seasonal mean for the 1980–2000 period using an atmospheric general circulation model (AGCM), interannual variability in the first and second moments of probability density function (PDF) of atmospheric seasonal means with sea surface temperatures (SSTs) is analyzed. Based on the strength of the SST anomaly in the Niño-3.4 index region, the years between 1980 and 2000 were additionally categorized into five separate bins extending from strong cold to strong warm El Niño events. This procedure further enhances the size of the ensemble for each SST category. All the AGCM simulations were forced with the observed SSTs, and different ensemble members for specified SST boundary forcing were initiated from different atmospheric initial conditions.

The main focus of this analysis is on the changes in the seasonal mean and the internal variability of tropical rainfall and extratropical 200-mb heights with SSTs. For the tropical rainfall, results indicate that in the equatorial tropical Pacific, internal variability of the tropical rainfall anomaly decreases (increases) for the La Niña (El Niño) events. On the other hand, seasonal mean variability of extratropical 200-mb height decreases for the El Niño events. Although there is increase in the seasonal mean variability of 200-mb heights for the La Niña events, results are rather inclusive. Analysis also indicates that for the variables studied, the influence of the interannual variability in SSTs is much stronger on the first moment of seasonal means compared to their influence on the internal variability. As a consequence, seasonal predictability due to changes in SSTs can be attributed primarily to the shift in the PDFs of the seasonal atmospheric means and less to changes in their spread.

Modes of internal variability for 200-mb extratropical seasonal mean heights for different SST categories are also analyzed. The dominant mode of internal variability has little dependence on the tropical SST forcing, while larger influence on the second mode of internal variability is found. For SST forcing changing from a La Niña to El Niño state, the spatial pattern of the second mode shifts eastward. For the cold events, the spatial patterns bear more resemblance to the Pacific–North American (PNA) pattern, while for the warm events, it more resembles the tropical–North Hemispheric (TNH) pattern. Change in the spatial pattern of this mode from strong cold to a strong warm event resembles the change in the spatial pattern of response in the mean state.

Corresponding author address: Dr. Peitao Peng, Climate Prediction Center, W/NP51, 5200 Auth Road, Room 800, Camp Springs, MD 20746. Email: peitao.peng@noaa.gov

Abstract

Based on a 40-member ensemble for the January–March (JFM) seasonal mean for the 1980–2000 period using an atmospheric general circulation model (AGCM), interannual variability in the first and second moments of probability density function (PDF) of atmospheric seasonal means with sea surface temperatures (SSTs) is analyzed. Based on the strength of the SST anomaly in the Niño-3.4 index region, the years between 1980 and 2000 were additionally categorized into five separate bins extending from strong cold to strong warm El Niño events. This procedure further enhances the size of the ensemble for each SST category. All the AGCM simulations were forced with the observed SSTs, and different ensemble members for specified SST boundary forcing were initiated from different atmospheric initial conditions.

The main focus of this analysis is on the changes in the seasonal mean and the internal variability of tropical rainfall and extratropical 200-mb heights with SSTs. For the tropical rainfall, results indicate that in the equatorial tropical Pacific, internal variability of the tropical rainfall anomaly decreases (increases) for the La Niña (El Niño) events. On the other hand, seasonal mean variability of extratropical 200-mb height decreases for the El Niño events. Although there is increase in the seasonal mean variability of 200-mb heights for the La Niña events, results are rather inclusive. Analysis also indicates that for the variables studied, the influence of the interannual variability in SSTs is much stronger on the first moment of seasonal means compared to their influence on the internal variability. As a consequence, seasonal predictability due to changes in SSTs can be attributed primarily to the shift in the PDFs of the seasonal atmospheric means and less to changes in their spread.

Modes of internal variability for 200-mb extratropical seasonal mean heights for different SST categories are also analyzed. The dominant mode of internal variability has little dependence on the tropical SST forcing, while larger influence on the second mode of internal variability is found. For SST forcing changing from a La Niña to El Niño state, the spatial pattern of the second mode shifts eastward. For the cold events, the spatial patterns bear more resemblance to the Pacific–North American (PNA) pattern, while for the warm events, it more resembles the tropical–North Hemispheric (TNH) pattern. Change in the spatial pattern of this mode from strong cold to a strong warm event resembles the change in the spatial pattern of response in the mean state.

Corresponding author address: Dr. Peitao Peng, Climate Prediction Center, W/NP51, 5200 Auth Road, Room 800, Camp Springs, MD 20746. Email: peitao.peng@noaa.gov

1. Introduction

It is well known that for a specified boundary condition, for example, sea surface temperature (SST), the seasonal mean atmospheric state is not unique. This can be documented from an ensemble of AGCM simulations forced with identical SSTs but starting from different atmospheric initial conditions (e.g., Kumar and Hoerling 1995; Barnett 1995; Rowell 1998; Hazrallah and Sadourny 1995). In such simulations, different realizations of seasonal mean atmospheric states demonstrate substantial variability, particularly in the extratropical latitudes. Therefore, for a given SST forcing, a complete specification for the seasonal mean atmospheric states requires knowledge of their probability density functions (PDFs). It is also well known that the interannual variations in the tropical Pacific SSTs related to ENSO influence different moments of PDF of seasonal means (for a review, see Trenberth et al. 1998).

Understanding and documenting changes in the PDF of the observed seasonal mean atmospheric states with SSTs is fundamental to the practice of seasonal prediction efforts. This, however, remains an elusive goal, because, in the first place, reliable estimates of PDFs for seasonal means cannot be constructed as such a procedure requires a large sample of seasonal means with similar SSTs, a requirement that cannot be met by the shortness of the observed data record. Confronted with this shortcoming, neither the PDFs for atmospheric seasonal means nor the influence of anomalous SSTs on these PDFs can be accurately inferred from the observational data alone. This problem is further exacerbated for extreme ENSO events, for example, the warm event of 1997–98, for which similar SST states in the observed data are even harder to find.

For the seasonal prediction problem, the first and second moments of the PDF and their variations with SSTs are of fundamental importance. Whereas the dependence of the first moment on the SST is useful in understanding the average atmospheric response, the dependence of the second moment on SSTs documents the change in spread of the seasonal atmospheric means for a fixed SST forcing. A schematic of the SST impact on the first and second moment of the PDF of the seasonal mean atmospheric states is shown in Fig. 1. Taken together, the information about the first and second moment for different SST states provides estimates for the predictability of seasonal means (for examples of PDFs for atmospheric variables, see Renshaw et al. 1998; Straus and Shukla 2002).

For the present availability of the combined oceanic and atmospheric historical data that are of approximately 50-yr duration, the change in the first moment of the PDF can be estimated for a composite ENSO event. For example, upon selecting 10 warm ENSO events out of the 50 yr of data record, a composite analysis provides the expected value of the seasonal atmospheric means. A shift in the seasonal mean state for the composite relative to a similar composite for the normal SST years is the change in the first moment of the PDF. This response, however, only provides a gross estimate for the change in the first moment for an average warm ENSO event, which in the present case, is the average of SST anomaly of the 10 different warm ENSO events. A deeper understanding of how the expected value of seasonal means varies from one ENSO SST to another is beyond the capability of the current archive of the observational data (Horel and Wallace 1981).

The estimation of the second moment of the PDF, and its dependence on SSTs, is an even harder problem. Unlike for the first moment of the PDF for which at least an estimate for an average ENSO state can be made with some confidence, an estimate of the second moment based only on a sample size of approximately 10 ENSO events cannot be done with any statistical confidence whatsoever. This is because an accurate estimation of sample variance requires a much larger sample size than what is needed for the estimation of the sample mean (see discussion in Sardeshmukh et al. 2000).

Use of the atmospheric general circulation models (AGCMs) provides an alternate procedure to analyze the changes in the PDF of the seasonal atmospheric means with SSTs. The advantage of the AGCMs is that a large ensemble of realization of seasonal atmospheric means for indentical SST forcings can be generated. From these simulations, changes in the first and second moment in the PDF of seasonal atmospheric means can then be analyzed for individual SST states. As with all AGCM studies, a caveat is that the conclusions based on AGCM simulations could be influenced by the AGCM biases, and consistency across many different AGCMs is essential to gain confidence in results obtained based on AGCM studies.

Following this approach for different SST forcings, changes in the first moment of the PDF were studied by Kumar and Hoerling (1998), Lau (1997), Rowell (1998), and Shukla et al. (2000), among others. In general, previous studies indicate that over the Pacific–North American (PNA) region, the amplitude of the change in the first moment of the PDF depends quasi-linearly on the amplitude of the tropical Pacific SST anomaly. On the other hand, compared to degree of spatial variability amongst different low-frequency modes of extratropical variability, the spatial pattern of ENSO response has much smaller dependence on the details of SST anomalies from one ENSO to another (Kumar and Hoerling 1997).1 Although most of the studies were based on the ensembles size of approximately 10 realizations, the robustness of results across many different AGCMs lend credibility to these conclusions.

In a series of papers, Kumar et al. (2000), Sardeshmukh et al. (2000), and Schubert et al. (2001) extended the analysis to study the changes in the second moment of the PDFs with SSTs. Unlike for the first moment, the results from these studies have been inconclusive. For example, Kumar et al. (2000) found small changes in the internal variability for the cold events, and Schubert et al. (2001) demonstrated that over the PNA region, the variability of seasonal means increased for the cold ENSO event of 1989. On the other hand, the analysis of Sardeshmukh et al. (2000) indicated a decrease in internal variability for the cold event of 1989. This divergence among the results could be due to the differences in the experimental set up. For example, the analysis of Sardeshmukh et al. (2000) and Schubert et al. (2001) was based on specific case studies for warm and cold ENSO events, while the analysis of Kumar et al. (2000) was based on regression and the compositing approach, which only provides information about the change in the moments of the PDF for some average ENSO SST states. Furthermore, as shown in Table 1, ensemble sizes of the AGCM realizations between these studies also differed.

The experimental set up and the analysis procedures outlined in Table 1 have different advantages and shortcomings. The analysis for the case studies, although done with large ensemble sizes, may be too case specific and may not represent the general impact of SST anomalies on different moments of the PDF. On the other hand, use of regression and composites to analyze the impact of SSTs on the PDF of the seasonal atmospheric means, although it reduces the impact of case specificity of the results, is generally based on smaller-sized ensembles.

Our approach in this paper falls somewhere in between the two extremes outlined above. For the 1980–2000 period, an ensemble size of 40 simulations is used to analyze the impact of SSTs on the first and second moments of the PDF of the seasonal mean atmospheric states. During this period, because several warm and cold events took place, our analysis is not restricted to a few case studies alone. At the same time, a sample size of 40 may be adequate to estimate the first-order impact of SSTs on the different moments of the PDF. Although these estimates may still suffer because of the sampling errors, an expectation that SST influence from one ENSO SST state to another should have continuity provides an additional consistency check. This is because we expect that incremental changes in the tropical Pacific SSTs should also result in incremental changes in the atmospheric response. And therefore, if the atmospheric responses are arranged in the order of increasing amplitude of the strength of anomalous SSTs, a gradual continuity in the atmospheric response is also to be expected.

Apart from the impact of the SSTs on the different moments of PDFs, another problem of interest is how the dominant modes of seasonal mean atmospheric variability change from one SST state to another (Bladé 1997, 1999; Straus and Shukla 2002). We have already pointed out that for a fixed SST forcing, different realizations of seasonal atmospheric means differ, and among different spatial structures for individual seasonal mean anomalies, certain spatial structures, for example, the Pacific–North American or North Atlantic Oscillation, dominate. How these patterns vary from one SST state to another is also a question of fundamental importance in understanding atmospheric low-frequency variability.

In this paper, we focus on the analysis of the impact of interannual variability in the SSTs both on the first and the second moment of the PDF of the January–February–March (JFM) seasonal atmospheric means. We also analyze how the dominant modes of atmospheric variability change for different SST states. The AGCM, data, and the analysis procedure used in this study are described in section 2. Results are presented in section 3, and in section 4, a summary and a discussion are included.

2. Data and analysis procedures

The atmospheric general circulation model used in the present analysis is the former National Centers for Environmental Prediction (NCEP) seasonal forecast model (Kanamitsu et al. 2002a). This dynamical seasonal forecast system consists of a suite of hindcasts and forecasts that are carried out once per month. The hindcasts are for the 1979–99 period and consist of an ensemble of 10 AGCM simulations for each calendar month. For example, for each December between 1979 and 1999 (i.e., 21 yr), an ensemble of 10 AGCM simulations, starting from 12-hr-apart observed atmospheric initial states between the first and the fifth of the month, is made. All the atmospheric simulations are approximately of 7 months duration, and simulations from the December initial states last until the end of June.

From the above set of AGCM simulations, we analyze the variability of the JFM seasonal mean. For all the JFMs between 1980 and 2000, 10 AGCM simulations from December 1979–99 initial conditions for each year can be used. JFM seasonal means from such simulations are approximately 1 month from the initial date. Additional sets of JFM simulations, although with longer lead times, are also available from November, October, and September initial conditions. To illustrate, 10 simulations each for JFM of 1980 are available from December, November, October, and September 1979–99 initial conditions. If all the simulations starting from different initial conditions are pooled, then for each JFM in the period 1980–2000, an ensemble of 40 AGCM simulations is available. That such a procedure is justified depends on the similarity of statistical properties of JFM seasonal means with different lead times and is demonstrated in the next section.

Atmospheric initial states for all the simulations are taken from the Reanalysis-2 dataset (Kanamitsu et al. 2002b) and are forced with the observed sea surface temperatures (Reynolds and Smith 1994). The atmospheric general circulation model has T42 spectral resolution (approximately 300-km horizontal resolution) and has 28 vertical levels. Land surface boundary conditions, for example, soil wetness and snow, are initialized as the climatology from the Reanalysis-2 dataset and are predicted during the AGCM integrations.

The analysis in this paper is based on JFM seasonal means of rainfall and 200-mb heights. In the analysis of atmospheric response to interannual variability in SSTs, rainfall plays an important role. In response to SSTs, particularly in the tropical latitudes, the interannual rainfall anomalies are the first link in the chain of the tropical–extratropical interactions. All JFM anomalies for the analysis are computed from the AGCM simulated climatology for the 1980–2000 period.

For the analysis period of 21 yr, El Niño and La Niña events are defined as the years when the Nino-3.4 SST index for the JFM season is above or below the one sigma standard deviation of the Nino-3.4 SST index. Based on this classification for the El Niño and La Niña events within the 21-yr analyis period, subjectively we further classify the events into weak and strong events. The classification of events into weak and strong El Niño and La Niña, together with the neutral years, is shown in Table 2. Finally, our analysis is based on pooling (or averaging) all AGCM simulations in each of the five categories in Table 2. This serves to increase the clarity of presentation and also increases the ensemble size of AGCM simulations for each SST category. Following this approach, results are presented as a continuum of tropical Pacific SST states extending from strong La Niña, weak La Niña, neutral, weak El Niño, to strong El Niño SST states.

3. Results

a. Justification that JFMs from different lead times can be pooled

We begin our analysis by demonstrating that for the atmospheric variables analyzed, their atmospheric statistics for the JFM mean does not depend crucially on the lead time; that is, JFM means from December, November, October, and September initial conditions can indeed be pooled together. This analysis, for JFM 200-mb seasonal mean heights, and their standard deviation, is shown in Figs. 2 and 3.

Shown in Fig. 2a is the difference between the 200-mb height climatology for the AGCM simulations from December and September initial conditions. AGCM-simulated JFM height climatology, for respective initial conditions, is computed as the mean of 210 simulations (i.e., 10 initial conditions each for 21 yr). For reference, 200-mb eddy height climatology, for the December initial conditions, is also shown in Fig. 2b. It is apparent that the differences between the JFM eddy height climatology from December and September initial conditions is much smaller than its absolute value, implying that the AGCM bias quickly saturates.

A similar analysis for the standard deviation of 200-mb for JFM height appears in Fig. 3 where the difference between the 200-mb height standard deviation for the AGCM simulations from December and September initial conditions is shown. Once again, it is apparent that the difference in the variability of 200-mb JFM seasonal means is quite insensitive to the lead time. Similar conclusions hold for mean and standard deviation of rainfall (not shown). This analysis provides a justification for pooling JFMs from different initial conditions with the advantage that for each JFM between the 1980–2000 period, we have access to a 40-member ensemble.

We should point out that the above conclusions may not hold for all the AGCM fields, for example, the soil moisture field, for which adjustment time to AGCM climatology could be much longer than a month. However, adjustment time and variability in land surface fields tend not to effect variability in the free troposphere (Wang and Kumar 1998; Kumar and Yang 2003). This is also true for the wintertime rainfall variability in the extratropical latitudes that is largely controlled by the upper-tropospheric circulation features. In the tropical latitudes, the largest rainfall variability is located over the tropical oceans and has little influence from the longer adjustment times related to land surface variables.

b. Analysis of mean response for different anomalous SSTs

For the five different categories of SST events in the equatorial tropical Pacific, the composite SST anomaly for each category is shown in Fig. 4. Because the anomalies are computed based on the entire 21-yr period, anomalies for the neutral category need not be zero. In fact,the anomaly for the neutral category is a measure of asymmetry between the El Niño and La Niña states.2

In Fig. 4, there are differences in the spatial structure between strong El Niño and La Niña during the 1980–2000 period. Maximum amplitude for the strong El Niño events is located east of the minimum for the La Niña events. Also, above-normal SST anomalies for the strong El Niño events extend all the way to the South American coast. This asymmetry in the spatial structure and the amplitude between the strong El Niño and La Niña is reflected in the SST anomalies for the neutral events. For example, for the neutral events, the SST anomalies in the eastern equatorial Pacific are below normal (i.e., opposite of the warm event composite), while they are above normal in the western tropical Pacific. For the weak El Niño and La Niña events, the location of the strongest anomaly is geographically more similar and is located slightly east of the date line in the equatorial tropical Pacific. We should point out that the features discussed might be specific to the 1980–2000 period.

For the different SST categories, the corresponding tropical rainfall and extratropical 200-mb height anomalies are discussed next. Anomalies for each SST category are the average of ensemble mean anomaly for the years in that SST category. The statistical significance of ensemble mean response is assessed based on the t test for an appropriate choice for the ensemble size (i.e., 120 for strong and weak warm and cold event composites, and 360 for the neutral ENSO state).

The tropical rainfall anomalies for different SST categories are shown in Fig. 5. As expected, for the El Niño events, in the tropical equatorial Pacific the positive rainfall anomalies extend from the date line to the South American coast and are flanked by the negative rainfall anomalies both to the south and the north. A similar, but opposite, pattern tends to happen for the rainfall anomalies in the cold events.

The spatial pattern of the rainfall anomalies for the neutral events also indicates asymmetry between the El Niño and La Niña events. To the east of the date line, where for the neutral events the SST anomalies in the equatorial Pacific are negative, rainfall anomalies are also negative and are larger than 2 mm day−1. Furthermore, similar to the SST anomalies for the neutral years, the spatial structure of the rainfall anomalies is such that to the east of the date line, rainfall is the opposite of the rainfall for the warm event composite, whereas to the west of the date line, it is the opposite of the rainfall for the cold event composite. The asymmetry for the rainfall composites for the warm and cold events is similar to that discussed by Hoerling et al. (1997), and is due to 1) the east–west variation in the SST climatology on which SST anomalies are superimposed and 2) the fact that convective rainfall anomalies favor SSTs larger than 27°C (Graham and Barnett 1987; Gadgil et al. 1984).

The extratropical 200-mb height response to different categories of SST anomalies is shown in Fig. 6. A comparison of the atmospheric height response between strong cold and strong warm ENSO cases indicates an approximately 15° eastward shift in the height pattern over the PNA region and is consistent with the asymmetric atmospheric response described in some earlier studies (Hoerling et al. 2001; Hannachi 2001; Straus and Shukla 2002). For the weak warm and cold ENSO cases, such a shift is not pronounced. We should point out that from the observational point of view, validation of an eastward shift in the response pattern remains a controversial issue, and results could be subject to sampling considerations (e.g., see Hoerling et al. 1997; DeWeaver and Nigam 2002). Finally, for this particular AGCM, the amplitude of 200-mb height response for the cold and warm events is comparable and is not consistent with the previously documented results where the atmospheric height response to the cold events tended to be weaker.

c. Analysis of variability of seasonal mean with SSTs

The analysis in section 3b illustrated the influence of ENSO SSTs on the first moment of seasonal atmospheric means. In this section, we document the same for the second moment of the PDF of the seasonal atmospheric means. For each SST category, the spread of seasonal means is computed in the following manner: 1) for each AGCM realization, we compute the departure from the ensemble mean. This departure is indicative of the atmospheric internal variability (see Fig. 1); 2) from the 40 realizations of internal variability, we estimate the variance of seasonal means for each SST state; and 3) for each SST category, we average the internal variance for different SST years in that category and compute the standard deviation from the averaged variance as an estimate for the spread of atmospheric seasonal means. This definition follows the traditional definition of the internal variability (e.g., Kumar and Hoerling 1995; Rowell 1998). The statistical significance of whether the spread for the strong and weak warm and cold events is different relative to the spread for the neutral SST states is assessed based on the F test.

Shown in Fig. 7 is the spatial distribution of change in the spread of the seasonal mean rainfall with SSTs. As for the previous analysis, this change is illustrated for different SST categories. For rainfall, there is good spatial correspondence between the change in variability and the shift in the mean of the PDF (cf. Fig. 7 with Fig. 5). In other words, regions of increased mean rainfall in response to SSTs also have an increase in their seasonal mean variability. For example, for the extreme cold events, the atmospheric seasonal mean response is a reduction in rainfall near the date line; a similar reduction in the spread of seasonal mean rainfall also occurs. Similarly, for the warm events, an increase in rainfall is also accompanied with an increase in the spread of its seasonal mean value. As the negative rainfall anomalies are associated with reduced variability, whereas the positive rainfall anomalies are associated with increased variability, it implies that the negative tropical rainfall anomalies may have larger seasonal predictability.

The correspondence between the seasonal mean rainfall anomaly and its internal variability may be a feature unique to the rainfall. For example, for the regions of reduced rainfall, as total rainfall is constrained to remain positive, rainfall variations to smaller rainfall cannot occur. This constraint for the total rainfall to be positive could lead to a reduction in the internal variability for the regions where rainfall anomaly itself is negative.

Shown in Fig. 8 is the spatial distribution of change in the spread of 200-mb seasonal mean heights in the extratropical latitudes over the PNA region. In general, for the cold events, seasonal mean heights have a small increase in variability, whereas for the warm events, seasonal mean heights have a reduction in variability. These results are consistent with those reported by Chen (2004) using the same dataset. A similar reduction in variability for warm events was also documented by Kumar et al. (2000) and Schubert et al. (2001); however, they are opposite to the conclusions drawn in the study by Sardeshmukh et al. (2000), where analysis for a specific cold event was associated with a decrease in the spread. Implications of reduced variability during the warm events, because of increased signal-to-noise ratio, would imply an increase in seasonal predictability during warm events relative to the scenario if the spread of seasonal means were to remain constant. On the other hand, an increase in variability during the cold events should lead to a decrease in seasonal predictability.

Possible reasons for an increase in the variability of the seasonal means during the cold events have been discussed in the literature. For example, Schubert et al. (2001) point out that the increased variability of the seasonal means in their analysis was a statistical residue of increased variability on the subseasonal time scales, and furthermore, is mainly due to the change in the mean state during the cold events. From the subseasonal time scale analysis, Chen and Van den Dool (1995), Sheng (2002), and Palmer (1988) indicated that the short-range forecasts initiated during the negative phase of PNA (similar to the 200-mb height response in Fig. 4 during the cold events) have larger growth rates than the short-range forecasts initiated during the positive phase of the PNA. A larger growth rate for the atmospheric perturbations during the negative phase of PNA (the spatial pattern favored during the cold events) could lead to an increase in subseasonal variability that in turn can alias into increased variability for the seasonal means.

In terms of the relative amplitude of change in the first and second moments of extratropical 200-mb heights and the tropical rainfall PDFs, it is apparent that influence of ENSO variability is appreciably larger on the mean of the PDF. For example, for the extratropical 200-mb heights, for the extreme cold and warm events, while the change in the mean is of the order of 100 m, change in the spread is approximately 10 m. A similar large difference in change in the mean versus change in the spread for the tropical rainfall also exists (cf. Figs. 5 and 7). As was discussed in Kumar et al. (2000), a stronger SST influence on the mean state implies that the ENSO-related seasonal predictability is mostly a result of the influence of the interannual variations in SSTs on the shift of the PDFs rather than their influence on the spread of the PDFs. If correct, this has implications for the practice of seasonal predictions. For example, while the shift in the mean of PDF can be readily estimated using a relatively small ensemble size of 10, a “fixed” estimate for the climatological seasonal mean spread can be made from a larger pool of model simulations leading to an average value of seasonal prediction skill close to its maximum possible value (Kumar and Hoerling 2000).

d. Analysis of internal modes of atmospheric variability

In the final analysis, the influence of tropical SSTs on the modes of internal variability of a 200-mb height field is analyzed. Analysis is based on the following procedure: 1) For each year, the seasonal mean anomalies for each model simulation with respect to its ensemble mean are computed. These anomalies, therefore, are indicative of atmospheric internal variability. 2) For different SST categories, all model anomalies are put in a single bin and are considered as independent realizations of atmospheric internal variability. For example, for the strong warm events, the number of realizations was 120 (three strong warm events with 40 realizations for each event). 3) An EOF analysis was performed for 200-mb height anomalies for each SST category separately. The spatial domain for the EOF analysis corresponds to what is shown in Figs. 9 and 10, and a latitudinal weighting to data prior to computing EOFs is applied.

Shown in Figs. 9 and 10 is a comparison of the first and second EOFs of internal variability for different SST categories. The shaded regions in these figures are the regions where the correlation between the EOF time series and individual realizations in the EOF analysis exceeds the 99% significance level. EOFs themselves are shown as regressions between individual realizations and the respective normalized EOF time series. The first dominant mode of internal variability has a north–south structure located over the Northern Pacific Ocean (Fig. 9). The spatial structure of this mode, as well as the percent variance explained, varies little with the state of tropical Pacific SSTs. For example, the smallest spatial correlation between any pair of EOFs in Fig. 9 is 0.65 and occurs between the EOF for the strong cold events (Fig. 9, top) and EOF for the strong warm events (Fig. 9, bottom). Lack of significant spatial variability is consistent with previous analysis by many investigators (e.g., Palmer 1993; Lau 1997; Bladé 1999). Although largely invariant, a small eastward shift in the equatorward center of the north–south dipole around the date line is discernable going from the extreme cold to the extreme warm events. Also, at the western edge of the domain shown in Fig. 9, a small but consistent eastward shift, and a decrease in amplitude, takes place.

The spatial structure of the second mode, together with the percent variance explained, is shown in Fig. 10. The percent variance explained by this mode is larger for the extreme cold events and decreases for the warm events. The spatial structure of this mode also undergoes an eastward shift from the cold to the warm events. This is reflected by a small spatial correlation of 0.07 between the EOF for the strong cold events (Fig. 10, top) and the EOF for the strong warm events (Fig. 10, bottom). For the cold events the spatial structure of this mode is similar to the PNA pattern. For the warm event, the spatial pattern is reminiscent of the tropical–North Hemispheric (TNH) pattern. It is also interesting to note that for the strong cold and warm SST categories, the spatial pattern of the second mode resembles the extratropical component of the SST-forced response in the seasonal means in Fig. 6. For example, the anomaly correlation between the spatial structure of atmospheric response in Fig. 6 and the spatial structure of the second EOF in Fig. 10, for strong cold and strong warm events, is 0.55 and 0.81, respectively.

The mechanistic interpretation of the influence of tropical SSTs on the extratropical low-frequency variability has been a contentious issue in the research community. On one hand, it is hypothesized that interannual variations in SSTs selectively influence the frequency of occurrence of internal modes, leading to seasonal mean anomalies (Palmer 1993; Lau 1997; Bladé 1999). One can also envision extratropical variability as a deterministic linear response to tropical heating anomalies associated with ENSO (Hoskins and Karoly 1981) and as an additive noise (i.e., internal variability) that is insensitive to changes in the tropical forcing. Following these arguments, the spatial structure of the seasonal mean response can be identified as one of the internal modes for both the neutral as well as the anomalous SST state. On the other hand, it is argued that the internal modes for different tropical SSTs could have a distinct spatial pattern. For this case, the spatial pattern of the seasonal mean response for anomalous SSTs cannot be identified as one of the modes of internal variability for the neutral SST state (Straus and Shukla 2002).

The modes shown in Figs. 9 and 10 are by definition internal modes of variability for different SST states. It is also apparent that there is a subtle but distinct difference in their spatial structure with SSTs. This is particularly true for mode 2. The analysis also indicates that the seasonal mean response in the extratropical latitudes for extreme events has a spatial structure similar to the spatial structure of mode 2. A mechanistic interpretation of modes of internal variability and influence of tropical Pacific SSTs on the extratropical climate variability can be summarized as follows: For the neutral SST state, the PNA pattern is a mode of internal variability. Cold tropical Pacific SST anomalies selectively favor one particular phase of PNA pattern leading to a seasonal mean response that is also similar to the PNA pattern. Neutral and cold SST states, therefore, have the same mode of internal variability, and the seasonal mean response can be identified with the internal modes for both SST states. For the warm SST state, one of the modes of internal variability is the TNH pattern and is also the seasonal mean response for the warm SSTs. However, unlike the neutral and cold SST states, neutral and warm SST states do not share the same dominant modes of internal variability, and the response pattern for the seasonal mean anomalies can no longer be identified as an internal mode for both the neutral and warm SST state. This conclusion is consistent with the observational analysis of Robertson and Ghil (1999) and Livezey and Mo (1987), who documented that the TNH pattern is primarily observed only during the warm events and also represents the seasonal mean response.

The above arguments imply that mechanistic interpretation of the influence of SSTs on the modes of internal variability and the seasonal mean response may lie somewhere in between the two hypothesis outlined above. For the cold events, the influence of SSTs is to selectively favor a particular phase of PNA, a mode that is also a mode of internal variability for the neutral SST state. In this case, the seasonal mean response can be identified as an internal mode for both the cold and neutral SST states. For the warm SST states, a distinct internal mode, that is, the TNH pattern, also exists. With respect to the neutral SST state, TNH is also the seasonal mean response but cannot be identified as an internal mode for the neutral SST state itself. The dynamical reason as to why it should be so may be related to preferred modes of variability for the base states for the cold, neutral, and warm SSTs and needs to be investigated further.

4. Summary and discussion

From the analysis of a large ensemble of AGCM simulations for the 1980–2000 period, different facets of atmospheric response to interannual variability in SSTs were analyzed. A unique aspect of this study was the use of large ensembles for many different SST states. The analysis was based on 40-member ensembles for each year, and further, depending on the strength of the SST anomaly in the Niño-3.4 region, the simulation years were further grouped into five different SST categories ranging from strong warm events to strong cold events. This leads to further enhancement in the size of ensemble for each SST category. To further ensure that the sampling does not influence the conclusions, the continuity of atmospheric response from one SST category to another provided an additional constraint.

Analysis of internal variability of 200-mb heights in the extratropical latitudes confirmed some of the previously reported analysis that during the cold event, the internal variability of seasonal mean height tends to increase, while it decreases during the warm events. For the equatorial tropical rainfall, on the other hand, internal variability of seasonal mean rainfall decreases during the cold events, while it increases during the warm events.

Origins of extratropical internal variability could be entirely due to the extratropical atmospheric dynamics or could also have some relationship with the internal variability of the rainfall in the tropical latitudes. The latter would be in the same spirit as the interannual variability in SSTs where accompanying rainfall variability is the forcing for the atmospheric response in the extratropical latitudes. In other words, if spatially organized tropical rainfall anomalies for fixed SSTs exist, they could also lead to atmospheric responses in the extratropical latitudes. In the present analysis, however, even if spatially organized tropical rainfall anomalies for fixed SSTs were to exist, as changes in the internal variability in the tropical rainfall and the extratropical heights are of opposite signs—that is, for the cold event, while the internal variability of the tropical rainfall decreases, internal variability of the extratropical heights increases—this may contradict the notion that the extratropical internal variability has its origin in the tropical internal variability of the rainfall anomalies. A similar conclusion could be drawn from the analysis of Schubert et al. (2001), who demonstrated that during the cold event, an increase in the internal variability of heights in the extratropical latitudes was primarily due to the extratropical dynamics related to the changes in the base state.

Although the differences in the base states between the warm and cold events may explain the differences in the amplitude of the internal variability, the analysis also indicates that the changes in the extratropical mean state associated with the tropical SST variability are not large enough to lead to significant changes in the internal modes of atmospheric variability in the extratropical latitudes. The dominant mode of the extratropical internal variability for different SST states also remains almost invariant in its spatial pattern. It is only the spatial pattern of the second dominant mode of internal variability that shows an eastward shift between strong cold and strong warm ENSO events. In fact, in the extratropical latitudes, the spatial structure of the second mode for different SST categories (Fig. 10) resembles the spatial structure of the atmospheric response for strong SST events (Fig. 6); that is, between strong cold and strong warm SST events, both have an eastward shift.

These results also imply that the dynamical characteristics of the extratropical atmosphere may be similar between the cold and neutral SST states, and for both of them, the PNA pattern can be identified as a mode of internal variability. The same pattern also emerges as the response pattern for the seasonal means for the cold events, indicating that a particular phase of PNA is favored during the cold events. On the other hand, the dynamical characteristics of the extratropical atmosphere differ between the strong warm and neutral SST states. For in this case, the TNH pattern is one of the distinct modes on internal variability. The same pattern also emerges as the seasonal mean response for the warm SST events. The dynamical reasons as to why this may be so remain to be clarified.

Acknowledgments

This research was partially supported by NOAA’s Climate Dynamics and Experimental Program.

REFERENCES

  • Barnett, T. P., 1995: Monte Carlo climate forecasting. J. Climate, 8 , 10051022.

  • Bladé, I., 1997: The influence of midlatitude ocean–atmosphere coupling on the low-frequency variability of a GCM. Part I: No tropical SST forcing. J. Climate, 10 , 20872106.

    • Search Google Scholar
    • Export Citation
  • Bladé, I., 1999: The influence of midlatitude ocean–atmosphere coupling on the low-frequency variability of a GCM. Part II: Interannual variability induced by tropical SST forcing. J. Climate, 12 , 2145.

    • Search Google Scholar
    • Export Citation
  • Chen, W. Y., 2004: Significant change of extratropical natural variability associated with tropical ENSO anomaly. J. Climate, 17 , 20192030.

    • Search Google Scholar
    • Export Citation
  • Chen, W. Y., and H. Van den Dool, 1995: Low-frequency variabilities for widely different basic flows. Tellus, 47A , 526540.

  • DeWeaver, E., and S. Nigam, 2002: Linearity in ENSO’s atmospheric response. J. Climate, 15 , 24462461.

  • Gadgil, S., P. V. Joseph, and N. V. Joshi, 1984: Ocean–atmosphere coupling over monsoon regions. Nature, 312 , 141143.

  • Graham, N. E., and T. P. Barnett, 1987: Sea surface temperature, surface wind divergence and convection over tropical ocean. Science, 238 , 657659.

    • Search Google Scholar
    • Export Citation
  • Hannachi, A., 2001: Toward a nonlinear identification of the atmospheric response to ENSO. J. Climate, 14 , 21382149.

  • Hazrallah, A., and R. Sadourny, 1995: Internal versus SST-forced variability as simulated by an atmospheric general circulation model. J. Climate, 8 , 474495.

    • Search Google Scholar
    • Export Citation
  • Hoerling, M. P., A. Kumar, and M. Zhong, 1997: El Niño, La Niña, and the nonlinearity of their teleconnections. J. Climate, 10 , 17691786.

    • Search Google Scholar
    • Export Citation
  • Hoerling, M. P., A. Kumar, and T-Y. Xu, 2001: Robustness of the nonlinear climate response to ENSO’s extreme cases. J. Climate, 14 , 12771293.

    • Search Google Scholar
    • Export Citation
  • Horel, J. D., and J. M. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Mon. Wea. Rev., 109 , 813829.

    • Search Google Scholar
    • Export Citation
  • 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 , 11791196.

    • Search Google Scholar
    • Export Citation
  • Kanamitsu, M., and Coauthors, 2002a: NCEP dynamical seasonal forecast system 2000. Bull. Amer. Meteor. Soc, 83 , 10191037.

  • Kanamitsu, M., W. Ebisuzaki, J. Woollen, S-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002b: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 16311643.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 1995: Prospects and limitations of atmospheric GCM climate predictions. Bull. Amer. Meteor. Soc., 76 , 335345.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 1997: Interpretation and implications of the observed inter–El Niño variability. J. Climate, 10 , 8391.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 1998: Annual cycle of Pacific–North American seasonal predictability associated with different phases of ENSO. J. Climate, 11 , 32953308.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 2000: Analysis of a conceptual model of seasonal climate variability and implications for seasonal predictions. Bull. Amer. Meteor. Soc., 81 , 255264.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and F. Yang, 2003: Comparative influence of snow and SST variability on extratropical climate in northern winter. J. Climate, 16 , 22482261.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., A. G. Barnston, P. Peng, M. P. Hoerling, and L. Goddard, 2000: Changes in the spread of the variability of the seasonal mean atmospheric states associated with ENSO. J. Climate, 13 , 31393151.

    • Search Google Scholar
    • Export Citation
  • Lau, N-C., 1997: Interactions between global SST anomalies and the midlatitude atmospheric circulation. Bull. Amer. Meteor. Soc., 78 , 2133.

    • Search Google Scholar
    • Export Citation
  • Livezey, R. E., and K. C. Mo, 1987: Tropical–extratropical teleconnections during the Northern Hemisphere winter. Part II: Relationships between monthly mean Northern Hemisphere circulation patterns and proxies for tropical convection. Mon. Wea. Rev., 115 , 31153132.

    • Search Google Scholar
    • Export Citation
  • Palmer, T. N., 1988: Medium and extended range predictability and stability of the Pacific/ North American mode. Quart. J. Roy. Meteor. Soc., 114 , 619713.

    • Search Google Scholar
    • Export Citation
  • Palmer, T. N., 1993: Extended-range atmospheric prediction and the Lorenz model. Bull. Amer. Meteor. Soc., 74 , 4966.

  • Renshaw, A. C., D. P. Rowell, and C. K. Folland, 1998: Wintertime low-frequency weather variability in the North Pacific–American sector 1949–1993. J. Climate, 11 , 10731093.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7 , 929948.

    • Search Google Scholar
    • Export Citation
  • Robertson, A. W., and M. Ghil, 1999: Large-scale weather regimes and local climate over the western United States. J. Climate, 12 , 17961813.

    • Search Google Scholar
    • Export Citation
  • Rowell, D. P., 1998: Assessing potential seasonal predictability with an ensemble of multidecadal GCM simulations. J. Climate, 11 , 109120.

    • Search Google Scholar
    • Export Citation
  • Sardeshmukh, P. D., G. P. Compo, and C. Penland, 2000: Changes of probability associated with El Niño. J. Climate, 13 , 42684286.

  • Schubert, S. D., M. J. Suarez, Y. Chang, and G. Branstator, 2001: The impact of ENSO on extratropical low-frequency noise in seasonal forecasts. J. Climate, 14 , 23512365.

    • Search Google Scholar
    • Export Citation
  • Sheng, J., 2002: GCM experiments on changes in atmospheric predictability associated with the PNA pattern and tropical SST anomalies. Tellus, 54 , 317329.

    • Search Google Scholar
    • Export Citation
  • Shukla, J., and Coauthors, 2000: Dynamical seasonal prediction. Bull. Amer. Meteor. Soc., 81 , 25932606.

  • Straus, D. M., and J. Shukla, 2002: Does ENSO force the PNA? J. Climate, 15 , 23402358.

  • Trenberth, K. E., G. Branstator, G. W. Karoly, A. Kumar, N-C. Lau, and C. Ropelewski, 1998: Progress during TOGA in understanding and modeling global teleconnections with tropical sea surface temperatures. J. Geophys. Res., 103 , 1429114324.

    • Search Google Scholar
    • Export Citation
  • Wang, W., and A. Kumar, 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103 , 2863728646.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

A schematic for the change in PDF for a seasonal mean quantity with SSTs. The thin curve is the PDF for neutral SSTs, and a finite spread is indicative of the fact that for a fixed SST forcing, the seasonal mean atmospheric state is not unique. The thick curve is the PDF for anomalous SSTs. The influence of SSTs could be on the mean of the PDF (implied by the shift in the PDF), as well as on the spread of the PDF (implied by the change in the standard deviation of the PDF)

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 2.
Fig. 2.

(a) Difference in the JFM 200-mb eddy height climatology for AGCM simulations from Dec initial conditions (i.e., 1-month lead time) and AGCM simulations from Sep initial conditions (i.e., 4-month lead time). A small difference indicates that AGCM eddy heights drift quickly to the AGCM climatology. The contour interval is 5 m. (b) JFM 200-mb eddy height climatology for AGCM simulations for Dec initial conditions. The contour interval is 30 m. Negative values are dashed

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 3.
Fig. 3.

Same as in Fig. 2, but for the std dev of JFM 200-mb seasonal mean heights. The contour intervals are (a) 3 and (b) 20 m. Negative values are dashed

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 4.
Fig. 4.

JFM SST anomalies for five different SST categories from (top) the strongest cold events to (bottom) the strongest warm events. The composite for the neutral years is shown in the middle panel. Years in the composites for different SST categories are in Table 2. The contour interval is 0.5°C, and negative values are dashed

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 5.
Fig. 5.

JFM composite rainfall anomalies corresponding to the five different SST categories shown in Fig. 4 and arranged from (top) strong cold to (bottom) strong warm ENSO events. Contours are plotted at 0, ±2, ±4, and ±8, and units are in mm day−1. Negative contours are dashed. Shaded regions indicate where the ensemble mean anomaly is significant at the 99% level based on the t test

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 6.
Fig. 6.

Same as in Fig. 5, but for 200-mb height anomalies. The contour interval is 30 and units are in m

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 7.
Fig. 7.

JFM composite for std dev anomalies related to the rainfall internal variability corresponding to the five different SST categories shown in Fig. 4 and arranged from (top) strong cold to (bottom) strong warm ENSO events. The contour interval is 0.5, and units are in mm day−1. Shaded regions indicate where the std dev for strong and weak warm and cold events is significantly different from the standard deviation for the neutral event at the 99% level based on the F test

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 8.
Fig. 8.

Same as in Fig. 7, but for anomalies related to 200-mb rainfall internal variability. The contour interval is 5, and units are in m

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 9.
Fig. 9.

Spatial pattern of the first EOF of 200-mb JFM seasonal mean height internal variability corresponding to the five different SST categories shown in Fig. 4 and arranged from (top) strong cold to (bottom) strong warm ENSO events. The spatial patterns are computed as regression with the normalized EOF time series, and units are in m. The fraction of variance explained by the mode is indicated to the right of each panel. The straight line connects the positive center near the date line across different panels. The shaded regions are the regions where the correlation between the EOF time series and individual realizations in the EOF analysis exceeds the 99% significance level

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Fig. 10.
Fig. 10.

Same as in Fig. 9, but for the spatial pattern of the second mode of 200-mb JFM seasonal mean height internal variability. The straight line connects the negative center close to 40°N across the different panels. Units are in m. The shaded regions are the regions where the correlation between the EOF time series and individual realizations in the EOF analysis exceeds the 99% significance level

Citation: Journal of Climate 18, 7; 10.1175/JCLI-3314.1

Table 1.

A comparison of ensemble size and analysis procedure used by different authors in documenting the influence of tropical SST variability on the internal spread of seasonal mean atmospheric variability

Table 1.
Table 2.

List of JFMs between 1980 and 2000 in different SST categories extending from the strong cold to strong warm ENSO events. Categorization is based on the strength of the Niño-3.4 SST index

Table 2.

1

 This spatial pattern is merely the spatial map of change in the first moment of the PDFs at different geographical locations.

2

 For the anomalies computed based on the entire 21-yr period, the anomaly for the neutral years has to be the weighted mean of anomaly for the warm and cold events in the record.

Save
  • Barnett, T. P., 1995: Monte Carlo climate forecasting. J. Climate, 8 , 10051022.

  • Bladé, I., 1997: The influence of midlatitude ocean–atmosphere coupling on the low-frequency variability of a GCM. Part I: No tropical SST forcing. J. Climate, 10 , 20872106.

    • Search Google Scholar
    • Export Citation
  • Bladé, I., 1999: The influence of midlatitude ocean–atmosphere coupling on the low-frequency variability of a GCM. Part II: Interannual variability induced by tropical SST forcing. J. Climate, 12 , 2145.

    • Search Google Scholar
    • Export Citation
  • Chen, W. Y., 2004: Significant change of extratropical natural variability associated with tropical ENSO anomaly. J. Climate, 17 , 20192030.

    • Search Google Scholar
    • Export Citation
  • Chen, W. Y., and H. Van den Dool, 1995: Low-frequency variabilities for widely different basic flows. Tellus, 47A , 526540.

  • DeWeaver, E., and S. Nigam, 2002: Linearity in ENSO’s atmospheric response. J. Climate, 15 , 24462461.

  • Gadgil, S., P. V. Joseph, and N. V. Joshi, 1984: Ocean–atmosphere coupling over monsoon regions. Nature, 312 , 141143.

  • Graham, N. E., and T. P. Barnett, 1987: Sea surface temperature, surface wind divergence and convection over tropical ocean. Science, 238 , 657659.

    • Search Google Scholar
    • Export Citation
  • Hannachi, A., 2001: Toward a nonlinear identification of the atmospheric response to ENSO. J. Climate, 14 , 21382149.

  • Hazrallah, A., and R. Sadourny, 1995: Internal versus SST-forced variability as simulated by an atmospheric general circulation model. J. Climate, 8 , 474495.

    • Search Google Scholar
    • Export Citation
  • Hoerling, M. P., A. Kumar, and M. Zhong, 1997: El Niño, La Niña, and the nonlinearity of their teleconnections. J. Climate, 10 , 17691786.

    • Search Google Scholar
    • Export Citation
  • Hoerling, M. P., A. Kumar, and T-Y. Xu, 2001: Robustness of the nonlinear climate response to ENSO’s extreme cases. J. Climate, 14 , 12771293.

    • Search Google Scholar
    • Export Citation
  • Horel, J. D., and J. M. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Mon. Wea. Rev., 109 , 813829.

    • Search Google Scholar
    • Export Citation
  • 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 , 11791196.

    • Search Google Scholar
    • Export Citation
  • Kanamitsu, M., and Coauthors, 2002a: NCEP dynamical seasonal forecast system 2000. Bull. Amer. Meteor. Soc, 83 , 10191037.

  • Kanamitsu, M., W. Ebisuzaki, J. Woollen, S-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002b: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 16311643.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 1995: Prospects and limitations of atmospheric GCM climate predictions. Bull. Amer. Meteor. Soc., 76 , 335345.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 1997: Interpretation and implications of the observed inter–El Niño variability. J. Climate, 10 , 8391.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 1998: Annual cycle of Pacific–North American seasonal predictability associated with different phases of ENSO. J. Climate, 11 , 32953308.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 2000: Analysis of a conceptual model of seasonal climate variability and implications for seasonal predictions. Bull. Amer. Meteor. Soc., 81 , 255264.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and F. Yang, 2003: Comparative influence of snow and SST variability on extratropical climate in northern winter. J. Climate, 16 , 22482261.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., A. G. Barnston, P. Peng, M. P. Hoerling, and L. Goddard, 2000: Changes in the spread of the variability of the seasonal mean atmospheric states associated with ENSO. J. Climate, 13 , 31393151.

    • Search Google Scholar
    • Export Citation
  • Lau, N-C., 1997: Interactions between global SST anomalies and the midlatitude atmospheric circulation. Bull. Amer. Meteor. Soc., 78 , 2133.

    • Search Google Scholar
    • Export Citation
  • Livezey, R. E., and K. C. Mo, 1987: Tropical–extratropical teleconnections during the Northern Hemisphere winter. Part II: Relationships between monthly mean Northern Hemisphere circulation patterns and proxies for tropical convection. Mon. Wea. Rev., 115 , 31153132.

    • Search Google Scholar
    • Export Citation
  • Palmer, T. N., 1988: Medium and extended range predictability and stability of the Pacific/ North American mode. Quart. J. Roy. Meteor. Soc., 114 , 619713.

    • Search Google Scholar
    • Export Citation
  • Palmer, T. N., 1993: Extended-range atmospheric prediction and the Lorenz model. Bull. Amer. Meteor. Soc., 74 , 4966.

  • Renshaw, A. C., D. P. Rowell, and C. K. Folland, 1998: Wintertime low-frequency weather variability in the North Pacific–American sector 1949–1993. J. Climate, 11 , 10731093.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7 , 929948.

    • Search Google Scholar
    • Export Citation
  • Robertson, A. W., and M. Ghil, 1999: Large-scale weather regimes and local climate over the western United States. J. Climate, 12 , 17961813.

    • Search Google Scholar
    • Export Citation
  • Rowell, D. P., 1998: Assessing potential seasonal predictability with an ensemble of multidecadal GCM simulations. J. Climate, 11 , 109120.

    • Search Google Scholar
    • Export Citation
  • Sardeshmukh, P. D., G. P. Compo, and C. Penland, 2000: Changes of probability associated with El Niño. J. Climate, 13 , 42684286.

  • Schubert, S. D., M. J. Suarez, Y. Chang, and G. Branstator, 2001: The impact of ENSO on extratropical low-frequency noise in seasonal forecasts. J. Climate, 14 , 23512365.

    • Search Google Scholar
    • Export Citation
  • Sheng, J., 2002: GCM experiments on changes in atmospheric predictability associated with the PNA pattern and tropical SST anomalies. Tellus, 54 , 317329.

    • Search Google Scholar
    • Export Citation
  • Shukla, J., and Coauthors, 2000: Dynamical seasonal prediction. Bull. Amer. Meteor. Soc., 81 , 25932606.

  • Straus, D. M., and J. Shukla, 2002: Does ENSO force the PNA? J. Climate, 15 , 23402358.

  • Trenberth, K. E., G. Branstator, G. W. Karoly, A. Kumar, N-C. Lau, and C. Ropelewski, 1998: Progress during TOGA in understanding and modeling global teleconnections with tropical sea surface temperatures. J. Geophys. Res., 103 , 1429114324.

    • Search Google Scholar
    • Export Citation
  • Wang, W., and A. Kumar, 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103 , 2863728646.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    A schematic for the change in PDF for a seasonal mean quantity with SSTs. The thin curve is the PDF for neutral SSTs, and a finite spread is indicative of the fact that for a fixed SST forcing, the seasonal mean atmospheric state is not unique. The thick curve is the PDF for anomalous SSTs. The influence of SSTs could be on the mean of the PDF (implied by the shift in the PDF), as well as on the spread of the PDF (implied by the change in the standard deviation of the PDF)

  • Fig. 2.

    (a) Difference in the JFM 200-mb eddy height climatology for AGCM simulations from Dec initial conditions (i.e., 1-month lead time) and AGCM simulations from Sep initial conditions (i.e., 4-month lead time). A small difference indicates that AGCM eddy heights drift quickly to the AGCM climatology. The contour interval is 5 m. (b) JFM 200-mb eddy height climatology for AGCM simulations for Dec initial conditions. The contour interval is 30 m. Negative values are dashed

  • Fig. 3.

    Same as in Fig. 2, but for the std dev of JFM 200-mb seasonal mean heights. The contour intervals are (a) 3 and (b) 20 m. Negative values are dashed

  • Fig. 4.

    JFM SST anomalies for five different SST categories from (top) the strongest cold events to (bottom) the strongest warm events. The composite for the neutral years is shown in the middle panel. Years in the composites for different SST categories are in Table 2. The contour interval is 0.5°C, and negative values are dashed

  • Fig. 5.

    JFM composite rainfall anomalies corresponding to the five different SST categories shown in Fig. 4 and arranged from (top) strong cold to (bottom) strong warm ENSO events. Contours are plotted at 0, ±2, ±4, and ±8, and units are in mm day−1. Negative contours are dashed. Shaded regions indicate where the ensemble mean anomaly is significant at the 99% level based on the t test

  • Fig. 6.

    Same as in Fig. 5, but for 200-mb height anomalies. The contour interval is 30 and units are in m

  • Fig. 7.

    JFM composite for std dev anomalies related to the rainfall internal variability corresponding to the five different SST categories shown in Fig. 4 and arranged from (top) strong cold to (bottom) strong warm ENSO events. The contour interval is 0.5, and units are in mm day−1. Shaded regions indicate where the std dev for strong and weak warm and cold events is significantly different from the standard deviation for the neutral event at the 99% level based on the F test

  • Fig. 8.

    Same as in Fig. 7, but for anomalies related to 200-mb rainfall internal variability. The contour interval is 5, and units are in m

  • Fig. 9.

    Spatial pattern of the first EOF of 200-mb JFM seasonal mean height internal variability corresponding to the five different SST categories shown in Fig. 4 and arranged from (top) strong cold to (bottom) strong warm ENSO events. The spatial patterns are computed as regression with the normalized EOF time series, and units are in m. The fraction of variance explained by the mode is indicated to the right of each panel. The straight line connects the positive center near the date line across different panels. The shaded regions are the regions where the correlation between the EOF time series and individual realizations in the EOF analysis exceeds the 99% significance level

  • Fig. 10.

    Same as in Fig. 9, but for the spatial pattern of the second mode of 200-mb JFM seasonal mean height internal variability. The straight line connects the negative center close to 40°N across the different panels. Units are in m. The shaded regions are the regions where the correlation between the EOF time series and individual realizations in the EOF analysis exceeds the 99% significance level

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 281 149 8
PDF Downloads 116 28 2