Abstract

Patterns of outgoing longwave radiation (OLR) have been analyzed over the tropical Indian and Pacific Oceans in order to identify the varying influence of their associated convective anomalies on the circulation at higher latitudes. Particular attention has been given to the changes related to El Niño–Southern Oscillation (ENSO) events.

The two leading EOFs (emperical orthogonal functions) of monthly OLR anomaly patterns for the region between 20°N–20°S and 70°E–120°W, express complementary variations between centers located 1) near 170°W just south of the equator and over the Philippines, and 2) slightly south of the equator near 145°W and slightly north of the equator near 165°E. Cluster analysis over a smaller area between 10°S–10°N and 140°E–140°W has highlighted ENSO-related changes with two of the six clusters associated with “moderate” (EN) and “strong” (EN+) El Niño events, and a third including most La Niña (LN) events. The OLR anomaly patterns associated with the 1986/87 and 1991/92 warm events fell within the moderate category, whereas those for the mature and decaying phases of the 1982/83 and 1997/98 events were associated with the strong pattern.

For the EN cluster, composites of the global 1000-hPa height and 300-hPa streamfunction showed wave trains propagating poleward and eastward in each hemisphere from the main area of enhanced convection. These originated approximately 20° farther east for the EN+ composites. Apart from a deeper Aleutian low in the EN+ composites, the differences over North America were comparatively small. Significant changes were observed over New Zealand, where EN events were associated with weak southwesterly anomalies, but strong west-southwesterly anomalies were observed for the months included in the EN+ class. The overall circulation anomaly patterns associated with La Niña clusters were generally weak, and similar to those of EN but with the opposite sign. The composite patterns show little change between summer and winter months in the Southern Hemisphere, but the influence of the anomalous tropical convection on the Northern Hemisphere during the boreal summer is weak.

The responses to the anomalous convection indicated by the OLR anomalies have also been modeled by applying a linearized version of the barotropic vorticity equation at the 300-hPa level. The results obtained support a number of the key differences observed in the streamfunction composites and highlight the lack of summertime Rossby wave sources in the Northern Hemisphere.

For regions with climate variability sensitive to the location of tropical convection, these results suggest that a single index of tropical circulation, such as the Southern Oscillation index, is not sufficient on its own to specify ENSO-forced climate anomalies.

1. Introduction

The recently completed decade of Tropical Ocean Global Atmosphere (TOGA) research has brought significant advances in understanding of tropical influences on the climate of higher latitudes. Although extratropical climate predictability is smaller, due to the random effects of migratory weather systems, large perturbations in the Tropics associated with El Niño–Southern Oscillation (ENSO) events can produce predictable and climatologically significant variations in many parts of the globe.

The extensive literature relating to global teleconnection patterns arising from tropical sea temperature variations has recently been reviewed by Trenberth et al. (1998). Particular emphasis is placed on the tropical Pacific where the temperature anomalies associated with warm (El Niño) and cold (La Niña) events bring about changes in convection patterns and the divergent circulation. These in turn excite Rossby waves that propagate to higher latitudes and lead to persistent climatic anomalies on interannual timescales.

Although this overall pattern of cause and effect is now well established, many of the details remain to be resolved. Rossby wave propagation, for example, is modified by the mean flow at higher latitudes (e.g., Simmons et al. 1983; Sardeshmukh and Hoskins 1988), by interactions with extratropical weather systems (Kok and Opsteegh 1985; Held et al. 1989; Hoerling and Ting 1994), and by orographic influences (Nigam and DeWeaver 1998). Combinations of these features create waveguides, as explained in Simmons et al. (1983), Hoskins and Ambrizzi (1993), and Ambrizzi and Hoskins (1997), so that global or regional teleconnection patterns may not be particularly sensitive to the precise location of tropical heating. Nevertheless, there are some indications of sensitivity in the Northern Hemisphere response to variations in the tropical SST anomaly patterns. Mo and Higgins (1998), for example, present evidence of systematic changes in the response over North America during the northern winter as the center of enhanced convection is displaced across the Pacific. In the Southern Hemisphere, with its greater zonal symmetry, it is possible that the higher-latitude response is less sensitive to the location of the tropical forcing.

While there have been few major El Niño or warm events in the equatorial Pacific in the last two decades, since more extensive observational data have become available, it is apparent that each differs in some respects from the overall pattern described by Rasmusson and Carpenter (1982). In particular, most El Niño events since 1980 were not preceded by surface warming along the South American coast. Wallace et al. (1998) provide a comprehensive review of observational studies stimulated by the TOGA program, concluding that a single modal structure is incapable of representing the full range of spatial patterns of ocean–atmosphere interaction in the tropical Pacific, and that a further phenomenological description of ENSO is needed.

An overall relationship between outgoing longwave radiation (OLR) and sea surface temperature (SST) anomalies over the ENSO cycle is well established (Matthews and Kiladis 1999)—during El Niño events, a warm SST anomaly in the central tropical Pacific is associated with an eastward displacement of deep convection. However, the details of the OLR–SST connection are less clear as processes responsible for the coupling at the air–sea boundary are not well observed or understood (e.g., Chen et al. 1996). Observational studies have established that a high absolute SST is necessary for active deep convection, that the intensity of convection increases with SST between about 26° and 30°C, and then weakens above 30°C (Graham and Barnett 1987; Zhang 1993). Even when SSTs are high, convection may be suppressed by other factors such as upper-level subsidence generated by remote forcing. Relationships between OLR and SST anomalies may differ in the western Pacific warm pool and the eastern equatorial cold tongue. On the monthly timescale, Steiner and Khalsa (1987) found increases in convection in the eastern Pacific lagging rises in SST by one or more months.

This study was originally motivated by a desire to see whether observed differences in ENSO response in the New Zealand region could be traced back to variations in the tropical convection patterns and whether there may be other patterns of tropical forcing causing climate variations on the interannual timescale. Traditionally the Southern Oscillation index (SOI) has been used as a predictor in local forecasting schemes (Mullan 1995; Francis and Renwick 1998; Renwick et al. 1999) but we believe that this has overly simplified the significant effect that ENSO variability has on the New Zealand climate.

Our analysis has made extensive use of OLR data that has been available with one 9-month break since July 1974, and is a widely used indicator of convective precipitation in the Tropics (Chelliah and Arkin 1992). We considered that it was a more direct indicator of the tropical convection forcing the Rossby wave trains affecting extratropical latitudes than could be obtained from sea temperature patterns.

OLR data have been widely used as a diagnostic tool in the analysis of ENSO and interannual variations, in association with SST and other indicators of the atmospheric and oceanic circulation, and this has often involved emperical orthogonal function (EOF) analysis to determine the principal modes of OLR variation (e.g., Kayano et al. 1995; Nigam and Shen 1993; Chelliah and Arkin 1992; Guo et al. 1998). Nevertheless there does not appear to have been the same detailed examination of year to year differences in the Pacific and eastern Indian Ocean areas that we have undertaken here.

While our 20-yr period of analysis is longer than many previous studies of interannual variability referred to by Kiladis and Mo (1998) and Kidson (1999), it covers only four strong El Niño events (1982/83, 1986;cl87, 1991/92, and 1997/98). This has limited our opportunities to examine seasonal changes in the ENSO patterns that a number of previous studies have shown to be significant (Mitchell and Wallace 1996; Nigam and Shen 1993). In order to support our observational studies, we have also examined the response to tropical forcing in a linear barotropic vorticity equation (BVE) model, both to provide a degree of validation for the overall composite patterns and to see how the ENSO response may vary between summer and winter seasons.

In the sections that follow, we relate the differing patterns of tropical convection obtained through EOF and cluster analysis to the SST anomalies associated with ENSO events (section 3). Composite global anomaly patterns are obtained for the upper- and lower-tropospheric circulation in section 4, and their evolution through the four major warm events since 1982/83 is examined in section 5. Partial validation of the changing midlatitude response is obtained through solutions of the linearized BVE in section 6. This section also briefly describes unsuccessful attempts to simulate the differing responses in the Hadley Centre Hadam3 general circulation model (GCM) by specification of sea temperature anomaly patterns. We finish with an assessment of the results in section 7.

2. Data sources and processing

The principal sources of data used in this study are the OLR dataset compiled by the National Oceanic and Atmospheric Administration (NOAA), the Reynolds SST dataset, and monthly mean 1000-hPa heights and 300-hPa streamfunctions derived from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis dataset.

a. OLR data

The OLR data were obtained from twice-daily measurements from NOAA polar-orbiting satellites1 that were checked for gross errors and interpolated in time and space (Liebmann and Smith 1996). Preliminary results indicated significant discontinuities across the gap in this dataset during 1978, and we have confined the EOF and cluster analysis to monthly mean data for the period from January 1979 to December 1999. Since spatial patterns are less sensitive to calibration errors, we have used the full dataset to obtain composite patterns, together with data for January 2000–February 2001 downloaded from the NOAA Cooperative Institute for Research in Environmental Sciences (CIRES) Climate Diagnostics Center, Boulder, Colorado.

Monthly anomalies were obtained by subtracting the long-term monthly mean, and their standard deviation is shown in Fig. 1. The maximum variance in the tropical Pacific between 160°E and 140°W comes largely from the 6-month period centered on January (Trenberth et al. 1998). As we wished to concentrate on ENSO-related variations in the tropical Pacific, further analysis was confined to the area between 20°N–20°S and 70°E–120°W, which contains the strongest month to month variability. Initially an S-mode EOF analysis (e.g., Richman 1986) was carried out on the covariance matrix between each grid point, leading to 2 patterns that the scree test (Craddock and Flood 1969) suggested were clearly significant and two more that appeared to be only marginally significant. These were not rotated because the leading patterns appeared capable of a sensible physical interpretation. The two leading EOFs presented in Fig. 2 show opposing patterns of convection, centered in the western and central Pacific, and together account for 37% of the monthly variance. The third expresses variations in the strength of convection in the Indian Ocean, west of Australia. The fourth is dominated by in-phase variations to the north of Australia and along the equator near the data line.

Fig. 1.

Standard deviation of monthly OLR anomalies between Jan 1979 and Apr 1998. The area chosen for further analysis (20°S–20°N, 70°E–120°W) is indicated by the rectangular box. Units: K

Fig. 1.

Standard deviation of monthly OLR anomalies between Jan 1979 and Apr 1998. The area chosen for further analysis (20°S–20°N, 70°E–120°W) is indicated by the rectangular box. Units: K

Fig. 2.

The four leading EOFs of monthly OLR anomalies between Jan 1979 and Apr 1998. The percentages represent the monthly variance explained by each EOF

Fig. 2.

The four leading EOFs of monthly OLR anomalies between Jan 1979 and Apr 1998. The percentages represent the monthly variance explained by each EOF

In view of the asymmetric distribution of the time coefficients of these EOFs, cluster analysis was also undertaken to identify the most commonly occurring convection patterns, without the requirement for the principal positive and negative anomalies to occur in the same location. The cluster analysis was initially applied over the same area as for the EOF computations, but failed to reveal any differences in convection patterns between warm events as shown in the two leading EOF patterns. Analysis was then carried out for a more restricted area between 10°N–10°S and 140°E–140°W, which contains much of the variance captured by the two leading EOFs. The clustering process started with the 252 sets of anomalies at each grid point and used the rms difference in OLR anomalies as a measure of similarity. A variation of the k-means procedure of MacQueen (1967) was adopted so that the individual monthly patterns could be reassigned to the closest cluster following each merge.

A scree test based on the total within group variance (Kalkstein et al. 1987) suggested that 10 or fewer clusters were likely to be statistically significant. The sequence of merges between clusters and the composite OLR patterns was then examined to decide how many clusters should be kept. In particular, we wished to retain patterns clearly associated with the ENSO phenomenon without the reduction in amplitude that comes from excessive merging. A set of 6 clusters appeared to be most appropriate for the analysis that will be presented in the following section. This set was virtually identical to the 7-cluster set, apart from deletion of 1 cluster containing only 1 member, and there was no apparent reason to retain the additional pattern in the 8-cluster set. The principal ENSO-related patterns in the 6-cluster set were also present in the 5-cluster set, but with some reduction in amplitude.

The failure of the wider-area analysis to capture the difference between warm event patterns probably relates to the scale of the separation between the convective centers relative to the size of the analysis area. On a smaller area, and with other variability removed, the separation of these two patterns contributes more to the reduction in variance by the clustering process.

A complete list of the cluster patterns best correlated with the OLR anomalies for each month is given in Table 1. As noted above, the clusters were derived from the data for 1979–99, but the classifications are also valid for the data from 1974 to 1978 since the spatial patterns are not affected by differences in calibration. The series in Table 1 has been extended to February 2001, using data downloaded from NOAA–CIRES in order to extend coverage of La Niña conditions.

Table 1.

List of OLR cluster assignments to each month from Jun 1974 to Feb 2001. Cluster numbers are defined as 1 = EN+, 2 = LN−, 3 = LN, 4 = X (transitional state), 5 = SP, 6 = EN. Months where the OLR anomalies have a pattern correlation with the mean of the preferred cluster exceeding >0.5 are shown in bold type. Those shown in lighter type have their best match with the indicated cluster, but should be regarded as difficult to classify

List of OLR cluster assignments to each month from Jun 1974 to Feb 2001. Cluster numbers are defined as 1 = EN+, 2 = LN−, 3 = LN, 4 = X (transitional state), 5 = SP, 6 = EN. Months where the OLR anomalies have a pattern correlation with the mean of the preferred cluster exceeding >0.5 are shown in bold type. Those shown in lighter type have their best match with the indicated cluster, but should be regarded as difficult to classify
List of OLR cluster assignments to each month from Jun 1974 to Feb 2001. Cluster numbers are defined as 1 = EN+, 2 = LN−, 3 = LN, 4 = X (transitional state), 5 = SP, 6 = EN. Months where the OLR anomalies have a pattern correlation with the mean of the preferred cluster exceeding >0.5 are shown in bold type. Those shown in lighter type have their best match with the indicated cluster, but should be regarded as difficult to classify

b. SST data

The monthly mean SST analyses were obtained on a 1° global grid as averages of weekly values obtained by optimum interpolation (Reynolds and Smith 1994). The analysis uses both in situ measurements and satellite data adjusted for biases using the method of Reynolds (1988) and Reynolds and Marsico (1993). From 1981 to 1989 the in situ data were obtained from the Comprehensive Ocean–Atmosphere Data Set for the 1980s (Slutz et al. 1985; Woodruff et al. 1993), and subsequently from the Global Telecommunication System. The satellite observations came from operational data produced by the National Environmental Satellite, Data and Information Service. The dataset from November 1981 to December 1998 was downloaded from NCAR, and the 1999 data was direct from NCEP.

c. NCEP–NCAR reanalysis data

We have made use of the NCEP–NCAR dataset comprising 12-h analyses at 2.5° spatial resolution between January 1958 and December 2000 (Kalnay et al. 1996). Daily streamfunctions at the 300-hPa level were computed as described in Kidson (1999) who also provides more details of this dataset, and some problems experienced with the Southern Hemisphere analyses.

3. OLR composites

Composite patterns of OLR for each cluster are shown in Figure 3 with the exception of one cluster, containing only 7 members, that appeared to have no physical significance. We refer to the clusters by the names EN, EN+ (for El Niño, moderate and strong, respectively), LN, LN− (La Niña, moderate and weak, respectively), and SP [South Pacific convergence zone (SPCZ)], for reasons that will become apparent later. As with the other composites presented here, they have been obtained from months where the OLR anomalies have a pattern correlation of ≥0.5 with one of the overall cluster means. On average, around 25% of the monthly anomaly patterns failed to reach this threshold, but the fraction varied between 33% for those best correlated with the LN pattern and 15% of those classified as EN+. The resulting composites have somewhat stronger anomalies but similar patterns to those obtained from all members of the cluster.

Fig. 3.

OLR composite patterns for the five leading clusters (EN, EN+, LN−, LN, and SP) obtained for the area 10°N–10°S, 140°E–140°W. (The number of members each cluster is shown next to its identifier.) Units: K. The dashed line shown on the SP pattern is the mean position of the SPCZ from FRSM. Shading represents areas with a magnitude of plus and minus 15 K or greater

Fig. 3.

OLR composite patterns for the five leading clusters (EN, EN+, LN−, LN, and SP) obtained for the area 10°N–10°S, 140°E–140°W. (The number of members each cluster is shown next to its identifier.) Units: K. The dashed line shown on the SP pattern is the mean position of the SPCZ from FRSM. Shading represents areas with a magnitude of plus and minus 15 K or greater

The OLR clusters are more readily interpreted in conjunction with the matching composite anomalies in SST, shown in Fig. 4. However, we note first that the OLR anomalies for clusters EN and LN have similar spatial patterns to EOF1 shown in Fig. 2, while EN+ and LN− are qualitatively similar to EOF2. Clusters EN and EN+ are both linked to positive SST anomalies along the equatorial Pacific that differ in their westward extent. For these two composites, the area of enhanced convection lies over the western end of the equatorial positive SST anomaly where the absolute temperatures are highest (see Hoerling et al. 1997). Compared to the EN composites, the “strong” class (EN+) has weaker warm anomalies, they are located around 20° farther east and have a greater meridional spread.

Fig. 4.

SST anomaly composites for the OLR clusters EN, EN+, LN, and SP shown in Fig. 3. Units: K. Shading represents values equal to or greater than plus and minus 1.0 K

Fig. 4.

SST anomaly composites for the OLR clusters EN, EN+, LN, and SP shown in Fig. 3. Units: K. Shading represents values equal to or greater than plus and minus 1.0 K

From the cluster time series (Table 1), we see that the very strong 1982/83 and 1997/98 El Niños began as cluster 6 (EN) and then became cluster 1 (EN+) in the mature stage when the area of positive SST anomaly advanced farther eastward toward the South American coast. Both these events ended with the warming confined near South America, rather than beginning that way as described in the Rasmusson and Carpenter (1982) composites. Moreover, cluster 1 persisted for some months beyond the decay of these El Niño events as defined by more conventional ENSO measures such as the SOI or Multivariate ENSO index (MEI; Wolter and Timlin 1998).

The SOI had returned to neutral conditions by May/June 1983, and by July 1998 the 1997/98 El Niño had switched into developing strong La Niña conditions. Cluster 1 also has a very persistent period in 1984 when there are weak negative SST anomalies across the equatorial Pacific, but the OLR still showed the EN+ pattern of Fig. 3. Thus, we use the EN+ terminology because of the association of this cluster with the mature and decaying phases of the two strongest El Niños of the twentieth century (Wolter and Timlin 1998; Barnston et al. 1999), while recognizing that the OLR pattern (and associated circulation anomalies, see section 4) can have some independent existence. The strong negative OLR anomaly center (EN+) is slightly more intense, located farther east, and is centered a little south of the equator. A prominent area of suppressed convection also appears centered near 155°E just north of the equator. The composites for 1985–95 presented in Wallace et al. (1998) show the peak negative anomalies in 1987 near 170°W, where the EN pattern is centered; while in 1992 the OLR minimum is around 150°W, closer to the principal center in EN+. The peak SST anomalies are found eastward of 165°W in both these years, though they are more persistent in 1987.

The clustering procedure was unable to distinguish between cold episodes, which are grouped together in the largest cluster, LN. Convection is suppressed in the vicinity of the date line, and the SST pattern is similar to that for the “moderate” warm episodes though with a smaller amplitude (and a change in sign).

Of the two remaining clusters, LN− appears to be a weak La Niña–type pattern, but its SST composite was found to contain a mixture of both positive and negative SST anomaly patterns and the cluster will be excluded from further analysis. SP shows increased tropical convection aligned along and north of the SPCZ, whose position is shown in Fig. 3, as computed from divergence patterns for November–April (Folland et al. 2001, manuscript submitted to Nature, hereafter FRSM). It is not associated with any significant tropical SST anomaly. As we will see in the next section, it may result from extratropical forcing associated with the positive phase of the high-latitude mode (see Kidson 1999).

4. Circulation anomalies

While the OLR cluster patterns and their corresponding patterns of SST anomalies apply to all months, the response of the atmospheric circulation to this convective forcing is expected to vary with the seasonal changes in the mean wind, as noted earlier. Because of the few ENSO events covered in our analysis period, we have less confidence in the seasonal response patterns but will present some summer/winter differences to indicate possible changes. These differences will be examined further in the barotropic vorticity equation model results discussed in section 6.

a. 300-hPa streamfunction patterns

The 300-hPa streamfunction anomaly patterns are shown in Fig. 5 for the moderate and strong El Niño patterns, and the La Niña and SP composites. These do not include data for 1999, reducing the number of months included in the La Niña composite. The shading is intended to clarify the circulation patterns rather than indicate levels of significance. However, for the EN and LN patterns, the shaded areas at higher latitudes are typically significant at the 1% level or better, as are the smaller circulation centers in the tropical Pacific and Atlantic. For the EN+ composite, with fewer members, departures in the composite mean exceeding 40 × 105 m2 s−1 are significant in the 5%–1% range.

Fig. 5.

Composite 300-hPa streamfunction anomaly patterns corresponding to clusters EN, EN+, LN, and SP. Centers with light shading are anticyclonic in the Southern Hemisphere, and cyclonic in the Northern Hemisphere. Centers with dark shading are cyclonic in the Southern Hemisphere and anticyclonic in the Northern Hemisphere. Units: 105 m2 s−1

Fig. 5.

Composite 300-hPa streamfunction anomaly patterns corresponding to clusters EN, EN+, LN, and SP. Centers with light shading are anticyclonic in the Southern Hemisphere, and cyclonic in the Northern Hemisphere. Centers with dark shading are cyclonic in the Southern Hemisphere and anticyclonic in the Northern Hemisphere. Units: 105 m2 s−1

Both the EN and EN+ patterns have the expected anticyclone pairs straddling the equator (e.g., Gill 1980; Branstator 1985), near to the area of strongest convection indicated by the negative OLR anomaly center. For the moderate EN cluster, the anticyclone centers lie near 165°W, but for the strong cluster they are displaced to around 135°W. Zonal winds in the eastern Pacific near 30°–35°N are strengthened in both cases, with maximum anomalies of around 4 m s−1 increasing to 6 m s−1 for the strong composite with an eastward extension. Along the equator, easterly anomalies are also stronger in the eastern Pacific for the strong case, reaching approximately 5 m s−1. As might be anticipated, the flow over North America is not greatly changed between the two cases, despite the difference in location of the tropical forcing. The Aleutian low becomes more intense in the strong composite and the anticyclonic anomaly center over Canada is strengthened and displaced to the east. In the Southern Hemisphere, the cyclonic anomaly over western Australia is weaker for the strong composites and the westerly flow over New Zealand is appreciably strengthened.

For the La Niña composites (LN) the anomaly patterns are less intense and, with a reversal in sign, would provide a better match to the moderate rather than the strong El Niño pattern.

We also show the composites for the EN and LN classes for the periods November–March and May–September in Fig. 6. These periods were chosen to represent the summer and winter seasons but have been extended to 5 months in length to increase the sample size. The EN pattern is similar in both periods but it is stronger in the winter hemisphere. The LN pattern is again similar to EN, and the main seasonal effect is a strengthening of the Northern Hemisphere departures during November–March. The EN+ pattern has not been included in this figure because it occurs less frequently and sample sizes are small. It too indicates similar patterns in the winter and summer seasons and a much stronger amplitude in the northern winter. The confidence levels were assessed using a Student's t-test and generally support the conclusions drawn above. The t values with a magnitude exceeding 3.0 (99% confidence limit) are found in both hemispheres in the principal centers of action in the northern winter but disappear from the Northern Hemisphere during the northern summer.

Fig. 6.

Composite 300-hPa streamfunction anomalies for the EN and LN clusters over the periods Nov–Mar and May–Sep. Details of contour intervals and shading as for Fig. 5 

Fig. 6.

Composite 300-hPa streamfunction anomalies for the EN and LN clusters over the periods Nov–Mar and May–Sep. Details of contour intervals and shading as for Fig. 5 

b. 1000-hPa height patterns

Compared to streamfunction anomalies, the 1000-hPa height anomalies emphasize changes at higher latitudes and give little weight to those in the Tropics. In Fig. 7, we have presented composites for those months in which the OLR anomalies have a pattern correlation with the El Niño or La Niña cluster means exceeding 0.5. This is to give a clearer indication of the response, but in practice the composite patterns differ little from the means for all states assigned to these clusters. The t values (not shown) indicate significant departures at the 1% level or better over much of the Pacific between 30°N and 30°S for the LN and EN clusters. Similar significance levels are also found for the EN+ cluster over the central Pacific and in a belt extending from the south of Alaska through to the Gulf of Mexico.

Fig. 7.

Composite 1000-hPa height anomaly patterns for monthly means whose OLR anomaly has a pattern correlation >0.5 with OLR composites EN, EN+, and LN. Low centers are indicated by light shading and highs by dark shading. Contour interval: 10 gpm (geopotential meters)

Fig. 7.

Composite 1000-hPa height anomaly patterns for monthly means whose OLR anomaly has a pattern correlation >0.5 with OLR composites EN, EN+, and LN. Low centers are indicated by light shading and highs by dark shading. Contour interval: 10 gpm (geopotential meters)

For the moderate El Niño cluster, EN, the anomalous flow over the Northern Hemisphere is quite weak. For the strong cluster EN+, which is most common in the northern winter and spring, heights are lower by around 40 m (or approximately 5 hPa) in the vicinity of the Aleutian low and there is an anomalous southerly flow onto the west coast of the United States and Canada. Easterly anomalies are also found over the east coast of North America. Over New Zealand and southeastern Australia, the west-southwesterly flow is also stronger. Elsewhere the differences are small. For La Niña composites the anomaly patterns are weak, and unlikely to produce noticeable changes at midlatitudes.

The seasonal differences are again shown in Fig. 8, and are similar to those observed for the 300-hPa streamfunction. The EN and LN responses almost disappear during the northern summer, while their strength is comparable to that for the annual mean in the Southern Hemisphere. The t values (not shown) indicate that the 99% confidence level (t > 3.0) is seldom reached outside the tropical belt.

Fig. 8.

Composite 1000-hPa height anomalies for the EN and LN clusters over the periods Nov–Mar and May–Sep. Details of contour intervals and shading as for Fig. 7 

Fig. 8.

Composite 1000-hPa height anomalies for the EN and LN clusters over the periods Nov–Mar and May–Sep. Details of contour intervals and shading as for Fig. 7 

The Southern Hemisphere anomaly patterns for cluster SP, which showed enhanced convection along the mean January position of the SPCZ, are presented in Fig. 9. The streamfunction anomalies at 300 hPa favor an increase in the zonal westerlies near 60°S, particularly over the Indian Ocean and the South Atlantic. Zonal winds tend to be lighter near 40°S. At 1000 hPa, heights are lower than normal over the polar cap and above normal at midlatitudes and the overall pattern is suggestive of the positive phase of the zonally symmetric high-latitude mode (see Kidson 1999 and references therein). As has already been noted, the convection is not linked to any tropical SST anomalies. It may instead be driven by waves propagating from higher latitudes in the Southern Hemisphere, as has been observed in previous studies (e.g., Lau and Phillips 1986; Kiladis and Weickman 1992a,b; Kiladis 1998). The SPCZ lies near the equatorward boundary of westerly flow in the upper troposphere in both January and July and may serve as a sink of equatorward Rossby wave energy. The qualitative agreement in the patterns is not supported by the distribution of either the interannual or intradecadal indices related to the high-latitude mode that are described by Kidson (1999, Fig. 13). Further work along the lines of Kiladis (1998) will be necessary to see if individual bursts of cloudiness in the vicinity of the ITCZ can be linked to disturbances propagating from higher latitudes in the Southern Hemisphere.

Fig. 9.

Anomalous 300-hPa steamfunction (Ψ300) and 1000-hPa height (Z1000) anomalies over the Southern Hemisphere associated with OLR cluster SP. Contour intervals: 106 m2 s−1 and 10 gpm, respectively. Streamfunction value shading is for values equal to or greater than 20; height value shading is for values equal to or greater than ±10

Fig. 9.

Anomalous 300-hPa steamfunction (Ψ300) and 1000-hPa height (Z1000) anomalies over the Southern Hemisphere associated with OLR cluster SP. Contour intervals: 106 m2 s−1 and 10 gpm, respectively. Streamfunction value shading is for values equal to or greater than 20; height value shading is for values equal to or greater than ±10

5. Warm event sequences

a. OLR cluster types for warm events since 1982/83

To characterize the four major warm events since 1982/83, we have made use of the more general MEI devised by Wolter (1987) and Wolter and Timlin (1993). This index is derived from sea level pressure and winds, SST, surface air temperature, and total cloudiness over the tropical Pacific and is published monthly by the NOAA–CIRES Climate Diagnostics Center (CDC; available online at http://www.cdc.noaa.gov). The time series of this index for the warm events since 1982/83 is shown in Fig. 10 along with the occurrences of the moderate and strong OLR cluster types. While differences between them are apparent, these recent warm events reach peak intensity around northern winter and spring in their second year. From Fig. 10 it is evident that the four warm events may be further separated into the following two categories.

  • 1982/83 and 1997/98 showed large positive values of the index within the first of the two calendar years, with comparable or higher values early in the second year. Moderate OLR clusters (EN) were observed in the second half of the first year, changing to strong clusters (EN+) around November or December and persisting throughout the remainder of the event;

  • 1986/87 and 1991/92 were characterized by a slow buildup in the index to a peak early in the second year. Peak values of the index and its cumulative deviation were appreciably weaker than for the 1982/83 and 1997/98 cases. Moderate OLR clusters were observed from late in the first year, around the same time that transitions to the strong clusters were observed for the 1982/83 and 1997/98 years. Again these persisted through the remainder of the warm episode.

We conclude that over the period of study, warm events of moderate strength fall into the EN category during their mature phase, while strong events pass into the EN+ category with the sea temperature and convective anomalies displaced around 20° farther east.

Fig. 10.

Time series of the NOAA–CIRES CDC multivariate ENSO index for the warm events 1982/83, 1986/87, 1991/92, and 1997/98. The occurrence of OLR cluster types EN (moderate) and EN+ (strong) El Niño are indicated by squares and triangles, respectively

Fig. 10.

Time series of the NOAA–CIRES CDC multivariate ENSO index for the warm events 1982/83, 1986/87, 1991/92, and 1997/98. The occurrence of OLR cluster types EN (moderate) and EN+ (strong) El Niño are indicated by squares and triangles, respectively

b. Seasonal variation in cluster type

The monthly frequency of occurrence of the three ENSO-related OLR cluster types is shown in Fig. 11 for the period January 1979–April 1998. It is apparent that the moderate El Niño cluster, EN, occurs with a similar frequency throughout the year. In contrast both the EN+ and LN clusters show a strong annual cycle with the EN+ clusters peaking in southern autumn and winter, and seldom occurring in the spring. The LN cluster frequency peaks in southern spring and is least common in autumn.

Fig. 11.

Monthly frequency (%) of EN, EN+, and LN clusters for each Southern Hemisphere season (summer = Dec/Feb), between 1979 and 1998. (Only months where the OLR anomalies have a pattern correlation >0.5 with the cluster mean are included.)

Fig. 11.

Monthly frequency (%) of EN, EN+, and LN clusters for each Southern Hemisphere season (summer = Dec/Feb), between 1979 and 1998. (Only months where the OLR anomalies have a pattern correlation >0.5 with the cluster mean are included.)

c. Duration of cluster types

The frequency of sequences of the same OLR cluster type is presented in Table 2. One-half of the sequences of cluster types shown in this table do not persist for more than one month, and the longest mean durations are found for the EN, EN+, and LN types. The maximum recorded duration for these three classes is in the range from 7 to 10 months, indicating strong persistence over both warm and cold ENSO events. The 14 individual spells of 5 months or more are shown in Table 3. The 1986/87 and 1991/92 warm events include the principal EN spells, while long EN+ spells occur during the mature phase of the 1982/83 and 1997/98 warm events and for a 5-month period in mid-1984. The LN sequences in 1988/89 and from 1998/99 occur during well-defined cold events, with series of positive SOI values. The other persistent events listed in the table are linked with weak negative departures in the MEI referred to earlier, but departures in the SOI are not consistently above normal.

Table 2.

Frequency of duration of OLR cluster types EN, EN+, LN−, LN, and SP persisting from 1 to 10 months, and their mean duration in months. (Data for Jan 1979–Feb 2001.)

Frequency of duration of OLR cluster types EN, EN+, LN−, LN, and SP persisting from 1 to 10 months, and their mean duration in months. (Data for Jan 1979–Feb 2001.)
Frequency of duration of OLR cluster types EN, EN+, LN−, LN, and SP persisting from 1 to 10 months, and their mean duration in months. (Data for Jan 1979–Feb 2001.)
Table 3.

Durations and dates of persistent spells of ENSO-related cluster types. (Data for Jan 1979–Feb 2001; * indicates possible incomplete spell.)

Durations and dates of persistent spells of ENSO-related cluster types. (Data for Jan 1979–Feb 2001; * indicates possible incomplete spell.)
Durations and dates of persistent spells of ENSO-related cluster types. (Data for Jan 1979–Feb 2001; * indicates possible incomplete spell.)

Mean SOI departures were calculated for the EN, EN+, and LN clusters (excluding those months with a pattern correlation of less than 0.5 with the cluster mean). The respective values of −1.59, −0.89, and +0.57 are consistent with the definitions of each cluster, except for the reversal in the magnitude of the departures for EN and EN+. The seasonal means are based on relatively few values but indicate that SOI departures for the EN+ cluster are largest during December–February (mean ∼−2.5) and exceed the value of −1.5 obtained for EN. The overall mean for EN+ is reduced in amplitude by mean values of −0.11 during June–August and +0.07 during September–November.

6. Model validation of extratropical influences

While statistically significant departures in the circulation have been found over a large part of the globe for the principal ENSO-related classes, they are based on the relatively few episodes occurring since the OLR record became available. We have sought, in this section, to validate the observed responses to the tropical forcing patterns with a linear barotropic model and a full atmospheric GCM.

The linear barotropic model is driven at the 300-hPa level by vorticity sources matching the composite OLR anomaly patterns, while the Hadam3 GCM was forced by the corresponding sea temperature anomalies superimposed on the long-term mean annual cycle.

a. Linear model using barotropic vorticity equation

In this section we use a linearized BVE model to explore dynamical links, established statistically by the data analysis of the previous sections, between these anomalous tropical convection and extratropical wave propagation events. Given that midlatitude circulation anomalies have a prominent equivalent barotropic structure, this single-level approximation is commonly used (see Trenberth et al. 1998). Ting (1996) found that applying this technique at the 350-hPa level gave the best match to the results from a full baroclinic model, in general agreement with the results of Held et al. (1989). Here we have used the 300-hPa level, the nearest for which daily and monthly streamfunctions were available from 40 yr of NCEP data (Kidson 1999). The isentropes at this level remain in the upper troposphere over most of the globe.

The model follows that of Sardeshmukh and Hoskins (1988), hereafter SH88). The linearized version used for the current study, described in Renwick and Revell (1999), hereafter referred to as RR99, can be written

 
formula

Here ξ is the relative vorticity, f is the Coriolis parameter, ζ is the absolute vorticity (ξ + f), and v is the wind vector with vψ, vχ the rotational and divergent components and vψ is the mean rotational flow. Terms involving vχ have been neglected as part of the linearization as discussed in RR99. Parameter μ determines the timescale over which the shortest length scale is dissipated and λ determines the time scale over which ξ is relaxed back to the mean, ξ.

We interpret the tropical OLR anomaly fields shown in Fig. 3 (scaled by a suitable constant factor to give realistic numbers) directly as the divergence anomaly ∇·vχ. We assume that the summer or winter mean flow, (ξ and vψ) is being maintained by other processes (e.g., baroclinic and orographic), here represented by the linearization and dissipation terms. Equation (1) was solved by the spectral transform technique, with a triangular truncation at wavenumber 42. Here μ was chosen to give a 6-h e-folding time for dissipating the 2-grid-length wave and λ chosen to restore the vorticity to its original value with an e-folding time of 10 days. Running the model for about 10 days (the time taken for the tendency term to become small), gives the linear stationary wave response ξ′ to tropical forcing.

Solutions ψ′ obtained by this procedure, for the moderate El Niño (EN), strong El Niño (EN+), and moderate La Nina (LN) classes, are shown in Fig. 12 for January conditions. Comparing the EN and LN cases with the corresponding observed 300-hPa streamfunction anomalies in Fig. 5 we see good agreement in the central Pacific, North American, and Asian regions.2 Agreement tends to lessen as we move farther in all directions from the central tropical Pacific with the phase remaining approximately correct but the amplitude becoming too small. The EN and LN forcing (indicated by vectors) that correspond to the observed OLR patterns are nearly mirror images of each other. The correspondence between the model response (which is linear) and the observed EN and LN 300-hPa anomalies is thus a measure of the linearity of the atmospheric response to moderate EN or LN forcing.

Fig. 12.

Streamfunction for the linear, barotropic, Rossby wave response after 10 days to divergent moderate El Niño (EN), strong El Niño (EN+), and La Niña (LN) forcing on the Jan mean 300-hPa flow. Contour interval is 106 m2 s−1. Negative contours are shown by dashed lines and indicate anticyclonic flow in the Southern Hemisphere and cyclonic flow in the Northern Hemisphere. The divergent circulation is indicated by the arrows

Fig. 12.

Streamfunction for the linear, barotropic, Rossby wave response after 10 days to divergent moderate El Niño (EN), strong El Niño (EN+), and La Niña (LN) forcing on the Jan mean 300-hPa flow. Contour interval is 106 m2 s−1. Negative contours are shown by dashed lines and indicate anticyclonic flow in the Southern Hemisphere and cyclonic flow in the Northern Hemisphere. The divergent circulation is indicated by the arrows

The first term on the rhs of (1), called the Rossby Wave Source (RWS) term by SH88, is shown in Fig. 13 for each of the situations depicted in Fig. 12. This term has two factors, the first of which, as explained in Hoskins and Karoly (1981), is ineffective at generating Rossby waves in regions of easterly flow indicated in the figure by light shading. The second factor, arising from the product of the υ component of the divergent wind and the poleward gradient of mean absolute vorticity, is collocated with the major higher latitude jets (SH88). The former tends to dominate in the east Pacific (since the midlatitude jets are weaker here and there are tropical westerlies) whereas the latter prevails in the west Pacific (since the jets are stronger here and there are easterlies in the Tropics). Thus in the west Pacific the local gradient of the background flow can tend to fix the position of the effective Rossby wave source regions, even though the longitude of the tropical source may change. This has been suggested in RR99 and several of the studies reviewed in Trenberth et al. (1998).

Fig. 13.

Rossby wave source terms for the three cases in Fig. 12, representing Jan conditions. Contours are in units of Ω2 s−2 where Ω is the earth's rotation rate in rad s−1. Negative contours are shown by dashed lines and indicate regions of cyclonic forcing in the Southern Hemisphere and anticyclonic forcing in the Northern Hemisphere. Areas where the 300-hPa zonal wind is negative (easterly) are indicated by light shading

Fig. 13.

Rossby wave source terms for the three cases in Fig. 12, representing Jan conditions. Contours are in units of Ω2 s−2 where Ω is the earth's rotation rate in rad s−1. Negative contours are shown by dashed lines and indicate regions of cyclonic forcing in the Southern Hemisphere and anticyclonic forcing in the Northern Hemisphere. Areas where the 300-hPa zonal wind is negative (easterly) are indicated by light shading

For the particular OLR anomalies and mean January flow considered here the Rossby wave source regions shown in Fig. 13 appear to be displaced relative to each other for the EN and EN+ cases. The modeled response in low latitudes, shown in Fig. 12 as a matching pair of anticyclonic anomalies, moves with the negative OLR forcing (corresponding to enhanced convection and positive divergence) in each case. This is in agreement with the observed streamfunction anomalies of Fig. 5. However the OLR signal associated with EN+ (shown in Fig. 3) differs from the other cases in that it has a positive center (corresponding to reduced convection and negative divergence) of almost equal intensity lying to the west. The OLR signal for the EN and LN cases tends to be dominated by the single center (with negative and positive value, respectively) in the central Pacific.

Accordingly we have separately examined the responses to both the positive and negative components of the forcing of EN+ and these are shown in Fig. 14. We find that the enhanced convective anomaly gives a wave train starting near 150°W and extending over North America that is almost identical to the EN+ pattern. The response to the reduced convective anomaly is quite different to the LN response. This is because the anomalous-reduced convection is located farther west and the wintertime jet to the north acts as a waveguide, preventing propagation to higher latitudes. This effect has been verified by inserting a model source at 15° intervals along the equator. The character of the wave train changes when the source lies near to or east of the jet exit region just east of the date line.

Fig. 14.

Streamfunctions for the linear, barotropic, Rossby wave response after 10 days to the enhanced convection center near 150°W (labeled “+150°W”) and the reduced convection near 155°E (labeled “−155°E”), which together give the strong El Niño (EN+) response in Fig. 12. Other details are as indicated for Fig. 12 

Fig. 14.

Streamfunctions for the linear, barotropic, Rossby wave response after 10 days to the enhanced convection center near 150°W (labeled “+150°W”) and the reduced convection near 155°E (labeled “−155°E”), which together give the strong El Niño (EN+) response in Fig. 12. Other details are as indicated for Fig. 12 

The overall solution obtained by adding the two responses, EN+ in Fig. 12, shows that the positive OLR center acts to move the low-latitude parts of the EN wave train farther east in both hemispheres. The magnitude of the changes near North America is greater for EN than EN+ so wind anomalies will be stronger. Figure 14 shows that over New Zealand, the SW gradient from the displaced EN pattern is effectively reinforced by a NW gradient from the negative convective anomaly, so that net EN+ pattern is a stronger westerly rather than a moderate SW anomaly, matching what the observational composites show.

Both EN and EN+ composites, shown in Fig. 5, have elongated anticyclones extending from the east over Australia. With EN the flow is more anticyclonic but northeasterlies over eastern Australia occur in both cases and the differences in weather may not be significant. The EN and EN+ differences over the Indian Ocean are captured to some extent by the linear model. It also gets the phase right but not the amplitude over South America, but results are poor to the east of North America. We conclude that the simple linear barotropic results can explain some of the remote response to observed El Niño variation during austral summer but they clearly do not provide the whole story.

During the austral winter both the Asian and North American jets in the Northern Hemisphere (NH) nearly disappear and easterlies extend throughout the tropical Pacific. This means the only source region for Rossby waves is now in the subtropics of the Southern Hemisphere (SH) between western Australia and the east Pacific. However, to the south of this band there is now a region where only wavenumbers 1 and 2 will propagate. Model results for July shown in Fig. 15 indicate contrasting behavior in the two hemispheres during their respective winters, with a wavenumber 4–5 Pacific–North American (PNA) pattern across the North Pacific during January but not July, and a wave 1 or 2 pattern over the far South Pacific during January and July. The observational results presented in Fig. 6 support this picture, but are not quite so simply interpreted. There is still a small PNA signal in NH summer for the EN case. This may be related to the fact that the observational results span May–September, whereas the model results were just for July.

Fig. 15.

Streamfunction for the linear, barotropic, Rossby wave response after 10 days to divergent moderate El Niño (EN), strong El Niño (EN+), and La Niña (LN) forcing on the Jul mean 300-hPa flow. Details as for Fig. 12 

Fig. 15.

Streamfunction for the linear, barotropic, Rossby wave response after 10 days to divergent moderate El Niño (EN), strong El Niño (EN+), and La Niña (LN) forcing on the Jul mean 300-hPa flow. Details as for Fig. 12 

b. Nonlinear response from Hadam3 GCM

In view of the probable importance of nonlinear interactions (Hoerling et al. 1997), attempts were also made to reproduce the El Niño and La Niña circulation anomalies in a full general circulation model (GCM). The GCM used here is the third version of the Hadley Centre climate model, HadAM3.

The basic features of this model and its performance during the Atmospheric Model Intercomparison Project (AMIP) period were described by Pope et al. (2000). The HadAM3 was set up to run four 50-yr simulations comprising: 1) control run, 2) moderate El Niño, 3) strong El Niño, and 4) La Niña. Forcing at the sea surface for the control run was provided as a 1961–90 climatological annual cycle obtained from the Hadley Centre's Global Sea-Ice and SST dataset, GISST 3.0 (Parker et al. 1995). For the other three ENSO runs, the corresponding composite SST anomalies shown in Fig. 4 were superimposed over the climatological annual cycle of SSTs used for the control run.

The resulting OLR patterns generated by the model varied with the season but generally showed poor correspondence with the observational composites in Fig. 3, and were unable to represent the differences between moderate and strong El Niño events. We have therefore not presented any results from the GCM simulation. The inability of the model to reproduce the observed extratropical response may be attributed to: 1) The stronger upper-level westerlies in the Southern Hemisphere (Pope et al. 2000), which could alter the quasi-stationary circulation anomalies resulting from the Rossby wave response in this part of the globe; and 2) the observational composites were constructed based on the OLR index, whereas the model composites were obtained from the corresponding composites of SST forced experiments.

Palmer and Mansfield (1986a), who worked with nonlinear models, noted a great sensitivity in the response to the basic climatology and it is necessary to get this right for a good El Niño simulation. In Palmer and Mansfield (1986b), they noted that differences in the SST anomaly patterns were reflected in very different teleconnection patterns over the North Pacific and North America.

7. Summary and conclusions

We have analyzed patterns of variability in OLR in the tropical Pacific in order to identify varying responses in the atmospheric circulation to convective forcing in this region. An initial EOF analysis of monthly OLR anomalies over the tropical Pacific and eastern Indian Oceans resulted in two principal modes expressing ENSO-related variability in the tropical Pacific, and variations in convection over the Indian Ocean. Both of these showed an east–west alignment of compensating OLR anomalies near the equator, with those for EOF2 displaced farther east.

In view of the skewed distribution of the temporal coefficients of these EOFs, implying a lack of symmetry between positive and negative departure patterns, we have preferred to base most of the results presented here on cluster analysis. For the restricted area between 10°N–10°S and 140°E–140°W, we obtained 5 patterns characteristic of interannual variations in tropical convection. Two of these (EN and EN+) were associated with warm SST anomalies in the central and eastern tropical Pacific, and a third, LN, with cold SST anomalies. A fourth pattern, LN−, had weak cold anomalies mirroring those of the EN+ pattern, while the fifth, SP, expressed variations along the equatorward side of the SPCZ that were apparently unconnected with tropical sea temperature forcing. The EN and EN+ clusters, associated with warm (El Niño) events differed in the location of the principal region of enhanced convection. When positive sea temperature anomalies extended from the South American coast to the date line, the convection was centered near 170°W, but on occasions when the anomalous warming extended only to around 160°W, the main convective center was also located around 20° farther east and convection was suppressed on a region centered north of the equator around 150°–160°E.

On average the EN, EN+, and LN clusters were most persistent, with maximum durations in the range of 7–10 months. The EN clusters occur throughout the calendar year, while EN+ clusters are found most frequently during the southern autumn and winter and are seldom observed in spring (September–November). The LN clusters occur most often in September–November and are least common in March–May.

An examination of the sequences for the four warm events covered by the period of analysis, showed that the EN+ cluster occurred principally during the mature phase and decay of the “strong” 1982/83 and 1997/98 events, where it followed a sequence of EN cluster patterns. In these cases, the westward boundary of the equatorial SST warm tongue, and the convective anomaly located near it, progressed farther east than for the 1987/87 and 1991/92 events, where the “moderate” EN cluster was observed throughout.

Analysis of the influence of these changes in convective forcing on the circulation needs to take account of the seasonal variation in the strength and position of the jets in each hemisphere, which can both block propagation of Rossby waves to higher latitudes and provide additional Rossby wave sources outside the Tropics through their interaction with the meridional circulation. The relatively few ENSO events covered by the OLR record made it difficult to have much confidence in the seasonal variations, even though the anomaly patterns are statistically significant. The main conclusion to be drawn is that tropical forcing of the Northern Hemisphere summer circulation is very weak.

In the Southern Hemisphere both the pattern and strength of the EN and LN composite circulation anomalies are similar in the May–September and November–March seasons and we can give more weight to the composites of all events. Composites of the EN OLR anomalies for these seasons (not shown) also indicated no systematic displacement of the forcing patterns.

The composite 300-hPa streamfunction and 1000-hPa height field anomalies for all seasons show similar wave trains originating near 150°–160°W for both EN and LN composites. In the Northern Hemisphere this extends over North America and is similar to the well-known Pacific North American or PNA pattern (e.g., Wallace and Gutzler 1981). Some differences are apparent in the EN+ pattern, which is forced both by enhanced convection in the eastern Pacific and a strong center of reduced convection near 155°E. The low-latitude part of the wave train is displaced to the east, along with the anticyclone over North America. There is a stronger westerly component in the flow on to the western United States near 30°N, as also seen in the composites of Barnston et al. (1999; Fig. 4) for the January–March quarters of the 1982/83 and 1997/98 El Niños.

In the Southern Hemisphere, the main changes occur over southeastern Australia and the South Pacific. The EN composites are associated with weaker wind speeds and a more southwesterly flow anomaly over New Zealand. Strong westsouthwesterlies are seen over southeastern Australia and New Zealand for the EN+ composite. The anomalies associated with the LN composite are very weak.

Application of the barotropic vorticity equation to obtain linear estimates of Rossby wave forcing at higher latitudes has borne out some of the features observed in this short dataset. The model results indicate the key locations for forcing an extratropical Rossby wave response are the divergent regions in the tropical belt and the jet streams of the subtropical latitudes. The extratropical response is critically dependent on the location of the tropical forcing relative to the major jet streams as these also tend to act as waveguides and barriers to the propagation of patterns with higher wavenumber. Variations of these forcing regions with season, for example, the disappearance of the Asian and North American jets and the spreading of easterlies throughout the tropical belt in NH summer, can explain some features of the seasonal variation of the composite observations. Attempts to examine some of the potential nonlinear effects, through forcing of the Hadam3 AGCM by the observed composite SST anomalies, proved unsuccessful.

The different response of the circulation to forcing from the EN and EN+ patterns suggests a more cautious approach to the derivation of statistical relationships between climatic elements and the SOI or MEI. In some areas the response does not simply increase with the departure in the index, but the pattern also changes. Care must be taken in these cases to ensure that the predictions for moderate El Niños are not based on relationships dominated by the strong events, and consideration of the forcing pattern as well as its amplitude becomes important.

We have identified some of these effects on New Zealand climatic elements and regional circulation that will be discussed in a separate paper (Kidson and Renwick 2002).

Acknowledgments

The barotropic vorticity equation model code was kindly supplied by the atmospheric modeling group at Reading University, United Kingdom. We thank Lois Steenman-Clark who helped adapt it to run locally. The NCEP–NCAR reanalysis dataset was provided by the Data Services Section at NCAR, and interpolated OLR and SST data were obtained from the NOAA–CIRES CDC, Boulder, Colorado, through their Web site at http://www.cdc.noaa.gov/. This study was supported by the New Zealand Foundation for Research, Science and Technology through Contracts CO1628 and C01X0030.

REFERENCES

REFERENCES
Ambrizzi
,
T.
, and
B. J.
Hoskins
,
1997
:
Stationary Rossby wave propagation in a baroclinic atmosphere.
Quart. J. Roy. Meteor. Soc
,
123
,
919
928
.
Barnston
,
A. G.
,
A.
Leetma
,
V. E.
Kousky
,
R. E.
Livezey
,
E. A.
O'Lenic
,
H.
van den Dool
,
A. J.
Wagner
, and
D. A.
Unger
,
1999:
:
NCEP forecasts of the El Niño of 1997–98 and its U.S. impacts.
Bull. Amer. Meteor. Soc
,
80
,
1829
1852
.
Branstator
,
G.
,
1985
:
Analysis of general circulation model sea-surface temperature anomaly simulations using a linear model. Part I: Forced solutions.
J. Atmos. Sci
,
42
,
2225
2241
.
Chelliah
,
M.
, and
P. A.
Arkin
,
1992
:
Large-scale interannual variability of outgoing longwave radiation anomalies over the global tropics.
J. Climate
,
5
,
371
389
.
Chen
,
S. S.
,
R. A.
Houze Jr.
, and
B. E.
Mapes
,
1996
:
Multiscale variability of deep convection in relation to large-scale circulation in TOGA COARE.
J. Atmos. Sci
,
53
,
1380
1409
.
Craddock
,
J. M.
, and
C. R.
Flood
,
1969
:
Eigenvectors for representing the 500mb geopotential surface over the Northern Hemisphere.
Quart. J. Roy. Meteor. Soc
,
95
,
576
593
.
Folland
,
C. K.
,
J. A.
Renwick
,
M. J.
Salinger
, and
A. B.
Mullan
,
2001
:
The influence of interdecadal and interannual climate variability on the South Pacific Convergence Zone.
Geophys. Res. Lett., submitted
.
Francis
,
R. I. C. C.
, and
J. A.
Renwick
,
1998
:
A regression-based assessment of the predictability of New Zealand climate anomalies.
Theor. Appl. Climatol
,
60
,
21
36
.
Gill
,
A. E.
,
1980
:
Some simple solutions for heat induced tropical circulation.
Quart. J. Roy. Meteor. Soc
,
106
,
447
462
.
Graham
,
N. E.
, and
T. P.
Barnett
,
1987
:
Sea surface temperature, surface wind divergence and convection over tropical oceans.
Science
,
238
,
657
659
.
Guo
,
Y.
,
P.
Zhai
, and
Y.
Ni
,
1998
:
A new index for ENSO monitoring.
Quart. J. Appl. Meteor., Beijing, China
,
9
,
169
177
.
Held
,
I. M.
,
S. W.
Lyons
, and
S.
Nigam
,
1989
:
Transients and the extratropical response to El Niño.
J. Atmos. Sci
,
46
,
163
174
.
Hoerling
,
M. P.
, and
M.
Ting
,
1994
:
Organization of extratropical transients during El Niño.
J. Climate
,
7
,
745
766
.
Hoerling
,
M. P.
,
A.
Kumar
, and
M.
Zhong
,
1997
:
El Niño, La Niña, and the nonlinearity of their teleconnections.
J. Climate
,
10
,
1769
1786
.
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
.
Hoskins
,
B. J.
, and
T.
Ambrizzi
,
1993
:
Rossby wave propagation on a realistic longitudinally varying flow.
J. Atmos. Sci
,
50
,
1661
1671
.
Kalkstein
,
L. S.
,
G.
Tan
, and
J. A.
Skindlov
,
1987
:
An evaluation of three clustering procedures for use in synoptic climatological classification.
J. Climate Appl. Meteor
,
26
,
717
730
.
Kalnay
,
E.
,
and Coauthors,
.
.
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc
,
77
,
437
471
.
Kayano
,
M. T.
,
V. E.
Kousky
, and
J. E.
Janowiak
,
1995
:
Outgoing longwave radiation biases and their impacts on empirical orthogonal function modes of interannual variability in the tropics.
J. Geophys. Res
,
100
, (
D2
).
3173
3180
.
Kidson
,
J. W.
,
1999
:
Principal modes of Southern Hemisphere low-frequency variability obtained from NCEP–NCAR reanalyses.
J. Climate
,
12
,
2808
2830
.
Kidson
,
J. W.
, and
J. A.
Renwick
,
2002
:
Patterns of convection in the tropical Pacific and their influence on New Zealand weather.
Int. J. Climatol., in press
.
Kiladis
,
G. N.
,
1998
:
Observations of Rossby waves linked to convection over the eastern tropical Pacific.
J. Atmos. Sci
,
55
,
321
339
.
Kiladis
,
G. N.
, and
K. M.
Weickmann
,
1992a
:
Circulation anomalies associated with tropical convection during northern winter.
Mon. Wea. Rev
,
120
,
1900
1923
.
Kiladis
,
G. N.
, and
K. M.
Weickmann
,
1992b
:
Extratropical forcing of tropical Pacific convection during northern winter.
Mon. Wea. Rev
,
120
,
1924
1938
.
Kiladis
,
G. N.
, and
K. C.
Mo
,
1998
:
Interannual and intraseasonal variability in the Southern Hemisphere.
Meteorology of the Southern Hemisphere, Meteor. Monogr.,
.
Amer. Meteor. Soc.
,
49
,
307
336
.
Kok
,
C. J.
, and
J. D.
Opsteegh
,
1985
:
Possible causes of anomalies in seasonal mean circulation patterns during the 1982–83 El Niño event.
J. Atmos. Sci
,
42
,
677
694
.
Lau
,
K-M.
, and
T. J.
Phillips
,
1986
:
Coherent fluctuations of extratropical geopotential height and tropical convection in intraseasonal time scales.
J. Atmos. Sci
,
43
,
1164
1181
.
Liebman
,
B.
, and
C. A.
Smith
,
1996
:
Description of a complete (interpolated) outgoing longwave radiation dataset.
Bull. Amer. Meteor. Soc
,
77
,
1275
1277
.
MacQueen
,
J.
,
1967
:
Some methods for classification and analysis of multivariate observations.
Proc. Fifth Berk. Symp. Math. Stat. Prob
,
1
,
281
297
.
Matthews
,
A. J.
, and
G. N.
Kiladis
,
1999
:
Interactions between ENSO, transient circulation, and tropical convection over the Pacific.
J. Climate
,
12
,
3062
3086
.
Mitchell
,
T. P.
, and
J. M.
Wallace
,
1996
:
ENSO seasonality: 1950–78 versus 1979–92.
J. Climate
,
9
,
3149
3161
.
Mo
,
K. C.
, and
R. W.
Higgins
,
1998
:
Tropical convection and precipitation regimes in the western United States.
J. Climate
,
11
,
2404
2423
.
Mullan
,
A. B.
,
1995
:
On the linearity and stability of Southern Oscillation–climate relationships for New Zealand.
Int. J. Climatol
,
15
,
1365
1386
.
Nigam
,
S.
, and
H-S.
Shen
,
1993
:
Structure of oceanic and atmospheric low-frequency variability over the tropical Pacific and Indian Oceans. Part I: COADS observations.
J. Climate
,
6
,
657
676
.
Nigam
,
S.
, and
E.
DeWeaver
,
1998
:
Influence of orography on the extratropical response to El Niño events.
J. Climate
,
11
,
716
734
.
Palmer
,
T. N.
, and
D. A.
Mansfield
,
1986a
:
Study of wintertime circulation anomalies during past El Niño events using a high-resolution general circulation model. Pt. 1: Influence of model climatology.
Quart. J. Roy. Meteor. Soc
,
112
,
613
638
.
Palmer
,
T. N.
, and
D. A.
Mansfield
,
1986b
:
Study of wintertime circulation anomalies during past El Niño events using a high-resolution general circulation model. Pt. 2: Variability of the seasonal mean response.
Quart. J. Roy. Meteor. Soc
,
112
,
639
660
.
Parker
,
D. E.
,
C. K.
Folland
, and
M.
Jackson
,
1995
:
Marine surface temperature: Observed variations and data requirements.
Climatic Change
,
31
,
559
600
.
Pope
,
V. D.
,
M. L.
Gallani
,
P. R.
Rowntree
, and
R. A.
Stratton
,
2000:
:
The impact of new physical parameterizations in the Hadley Centre climate model: HadAM3.
Climate Dyn
,
16
,
123
146
.
Rasmusson
,
E. M.
, and
T. H.
Carpenter
,
1982
:
Variations in the tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño.
Mon. Wea. Rev
,
110
,
354
384
.
Renwick
,
J. A.
, and
M. J.
Revell
,
1999
:
Blocking over the South Pacific and Rossby wave propagation.
Mon. Wea. Rev
,
127
,
1233
1247
.
Renwick
,
J. A.
,
R. E.
Basher
,
A. B.
Mullan
,
M. J.
Salinger
,
C. S.
Thompson
, and
R. I. C. C.
Francis
,
1999
:
Procedures for the production of monthly and seasonal climate outlooks.
NIWA Internal Rep. 56, 38 pp
.
Reynolds
,
R. W.
,
1988
:
A real-time global sea surface temperature analysis.
J. Climate
,
1
,
75
86
.
Reynolds
,
R. W.
, and
D. C.
Marsico
,
1993
:
An improved real-time global sea surface temperature analysis.
J. Climate
,
6
,
114
119
.
Reynolds
,
R. W.
, and
T. M.
Smith
,
1994
:
Improved global sea surface temperature analyses.
J. Climate
,
7
,
929
948
.
Richman
,
M. B.
,
1986
:
Rotation of principal components.
J. Climatol
,
6
,
293
335
.
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
.
Simmons
,
A. J.
,
J. M.
Wallace
, and
G. W.
Branstator
,
1983
:
Barotropic wave propagation and instability, and atmospheric teleconnection patterns.
J. Atmos. Sci
,
40
,
1363
1392
.
Slutz
,
R. J.
,
S. J.
Lubker
,
J. D.
Hiscox
,
S. D.
Woodruff
,
R. L.
Jenne
,
D. H.
Joseph
,
P. M.
Steuer
, and
J. D.
Elms
,
1985
:
Comprehensive Ocean–Atmosphere Data Set: Release 1.
NOAA Environmental Research Laboratory, Boulder, CO, 268 pp
.
Steiner
,
E. J.
, and
S. J. S.
Khalsa
,
1987
:
Sea surface temperature, low-level moisture, and convection in the tropical Pacific, 1982–1985.
J. Geophys. Res
,
92
,
(C13),
.
14217
14224
.
Ting
,
M.
,
1996
:
Steady linear response to tropical heating in barotropic and baroclinic models.
J. Atmos. Sci
,
53
,
1698
1709
.
Trenberth
,
K. E.
,
G. W.
Branstator
,
D.
Karoly
,
A.
Kumar
,
N-C.
Lau
, and
C.
Ropelewski
,
1998
:
Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures.
J. Geophys. Res
,
103
,
(C7),
.
14291
14324
.
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
.
Wallace
,
J. M.
,
E. M.
Rasmusson
,
T. P.
Mitchell
,
V. E.
Kousky
,
E. S.
Sarachik
, and
H.
von Storch
,
1998
:
On the structure and evolution of ENSO-related climate variability in the tropical Pacific: Lessons from TOGA.
J. Geophys. Res
,
103
,
(C7),
.
14241
14259
.
Wolter
,
K.
,
1987
:
The Southern Oscillation in surface circulation and climate over the tropical Atlantic, eastern Pacific, and Indian Oceans as captured by cluster analysis.
J. Climate Appl. Meteor
,
26
,
540
558
.
Wolter
,
K.
, and
M. S.
Timlin
,
1993
:
Monitoring ENSO in COADS with a seasonally adjusted principal component index.
Proc. 17th Climate Diagnostics Workshop, Norman, OK, NOAA/NMC/CAC, NSSL, Oklahoma Climatic Survey, CIMMS and the School of Meteorology, University of Oklahoma, 52–57
.
Wolter
,
K.
, and
M. S.
Timlin
,
1998
:
Measuring the strength of ENSO events: How does 1997/98 rank?
Weather
,
53
,
315
324
.
Woodruff
,
S. D.
,
S. J.
Lubker
,
K.
Wolter
,
S. J.
Worley
, and
J. D.
Elms
,
1993
:
Comprehensive Ocean–Atmosphere Data Set (COADS) Release 1a: 1980–1992.
Earth Syst. Monitor
,
4
(
1
)
6
.
Zhang
,
C.
,
1993
:
Large-scale variability of atmospheric deep convection in relation to sea surface temperature in the tropics.
J. Climate
,
6
,
1898
1913
.

Footnotes

Corresponding author address: Dr. John W. Kidson, NIWA, P.O. Box 14-901, Wellington, New Zealand. Email: j.kidson@niwa.cri.nz

1

Kindly provided by Dr. George Kiladis, NOAA/ERL.

2

Note that the convective forcing is scaled by an arbitrary constant so that spatial patterns may be compared, but not their amplitudes.