• Berbery, E. H., and J. N. Paegle, 1993: Intraseasonal interactions between the Tropics and extratropics in the Southern Hemisphere. J. Atmos. Sci.,50, 1950–1965.

  • Blackmon, M. L., 1976: A climatological spectral study of the 500 mb height of the Northern Hemisphere. J. Atmos. Sci.,33, 1607–1623.

  • Branstator, G., 1987: A striking example of the atmosphere’s leading traveling pattern. J. Atmos. Sci.,44, 2310–2323.

  • Dole, R. M., and N. D. Gordon, 1983: Persistent anomalies of the extratropical Northern Hemisphere wintertime circulation: Geographical distribution and regional persistent characteristics. Mon. Wea. Rev.,111, 1538–1553.

  • ——, and R. X. Black, 1990: Life cycles of persistent anomalies. Part II: The development of persistent negative height anomalies over the North Pacific Ocean. Mon. Wea. Rev., 118, 824–846.

  • Ghil, M., and K. C. Mo, 1991: Intraseasonal oscillations in the global atmosphere. Part II: Southern Hemisphere. J. Atmos. Sci.,48, 780–790.

  • Higgins, R. W., and S. D. Schubert, 1994: Simulated life cycles of persistent anticyclonic anomalies over the North Pacific: Role of synoptic scale eddies. J. Atmos. Sci.,51, 3238–3260.

  • ——, and ——, 1996: Simulations of persistent North Pacific circulation anomalies and interhemispheric teleconnections. J. Atmos. Sci.,53, 188–207.

  • ——, and K. C. Mo, 1997: Persistent North Pacific circulation anomalies and the tropical intraseasonal oscillation. J. Climate,10, 223–244.

  • Kalnay E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc.77, 437–471,.

  • Karoly, D. J., 1989: Southern Hemisphere circulation features associated with El Nino–Southern Oscillation. J. Climate,2, 1239–1252.

  • Kidson, J. W., 1988: Interannual variations in the Southern Hemisphere circulation. J. Climate,1, 1177–1198.

  • ——, 1991: Intraseasonal variation in the Southern Hemisphere circulation. J. Climate,4, 939–953.

  • Knutson, T. R., and K. M. Weickmann, 1987: 30–60 day atmospheric oscillations: composite life cycle of convection and circulation anomalies. Mon. Wea. Rev.,115, 1407–1436.

  • Lau, M. K., and P. H. Chan, 1986: Aspects of the 40–50 day oscillation during the northern summer as inferred from outgoing long wave radiation. Mon. Wea. Rev.,114, 1354–1367.

  • ——, P. J. Sheu, and I. S. Kang, 1994: Multiscale low-frequency circulation modes in the global atmosphere. J. Atmos. Sci., 51, 2753–2750.

  • Liebmann B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 1275–1277.

  • Mo, K. C., and G. H. White, 1985: Teleconnections in the Southern Hemisphere. Mon. Wea. Rev.,113, 22–37.

  • Papoulis, A., 1973: Minimum bias windows for high resolution spectral estimates. IEEE Trans. Info. Theory, 19, 9–12.

  • Rogers, J. C., and H. van Loon, 1982: Spatial variability of sea level pressure and height anomalies over the Southern Hemisphere. Mon. Wea. Rev., 110, 1375–1392.

  • Schubert, S. D., R. B. Rood and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc., 74, 2331–2342.

  • Szeredi, I., and D. J. Karoly, 1987: The horizontal structure of monthly fluctuations of the Southern Hemisphere troposphere from station data. Aust. Meteor. Mag., 35, 119–129.

  • Trenberth, K. E., 1986a: The signature of a blocking episode on the central circulation in the Southern Hemisphere. J. Atmos. Sci.,43, 2061–2069.

  • ——, 1986b: An assessment of the impact of transient eddies on the zonal flow during a blocking episode using localized Eliassen–Palm flux diagnostics. J. Atmos. Sci.,43, 2070–2087.

  • Vautard, R., and M. Ghil, 1989: Singular spectrum analysis in nonlinear dynamics with applications to paleoclimatic time series. Physica,35D, 392–424.

  • 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.

  • View in gallery

    (a) EOF 1 and (b) EOF 2 for the 200-hPa eddy streamfunction. EOFs are normalized to 1 and time 100. Contour interval is 0.5 nondimensional unit. (c) Correlations as a function of lead and lag time between PC 1 and PC 2 for the SH winter for total anomalies (open circles) and the (10–90 day) IS-filtered time series (dark circles).

  • View in gallery

    The 200-hPa eddy streamfunction anomaly composite averaged over all (a) positive PSA 1, (b) positive PSA 2, (c) negative PSA 2, and (d) negative PSA 1 days. Contour interval is 3 × 106 m2 s−1. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

  • View in gallery

    OLRA composite difference between positive and negative (a) PSA 1 and (b) PSA 2 events. Contour interval is 5 W m−2. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

  • View in gallery

    (a) EOF 1, (b) EOF 2, and (c) EOF 3 for OLRA for the SH winter. EOFs are normalized to 1 and time 100. Contour interval is 0.8 nondimensional unit.

  • View in gallery

    Map sequences of OLRA represented as the composite difference between positive and negative PSA 1 for the pentad centered at (a) day −13 (b) day −8, (c) day −3, and (d) day 2. Contour interval is 5 W m−2. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

  • View in gallery

    Same as Fig. 5 but for PSA 2 events.

  • View in gallery

    Same as Fig. 5 but for IS-filtered OLRA. Contour interval is 4 W m−2.

  • View in gallery

    Same as Fig. 6 but for IS-filtered OLRA. Contour interval is 4 W m−2.

  • View in gallery

    Map sequences of 200-hPa eddy streamfunction anomaly composite difference between positive and negative (a) PSA 1 for the pentad centered at day −8. (b) Same as (a), but for the pentad centered at day −3. (c) Same as (a), but for the pentad centered at day 2. (d) PSA 2 for the pentad centered at day −8. (e) Same as (d), but for the pentad centered at day −3. (f) Same as (d), but for the pentad centered at day 2. Contour interval is 5 × 106 m2 s−1. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

  • View in gallery

    Same as Fig. 9 but for velocity potential. Contour interval is 2 × 10 6 m2 s−1. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

  • View in gallery

    (a) Hovmöller diagram for total OLRA composite difference averaged from 5°S to 5°N between positive and negative PSA 1 events from 15 days before to 10 days after onset. Contour interval is 5 W m−2. Areas where values are significant at the 95% level are shaded. (b) Same as (a) but for IS-filtered OLRA. Contour interval is 3 W m−2. (c) Same as (a) but for PSA 2 mode. (d) Same as (c) but for IS-filtered OLRA.

  • View in gallery

    (a) Hovmöller diagram for the 200-hPa eddy streamfunction composite difference averaged from 50° to 60°S between positive and negative PSA 1 from 10 days before to 10 days after onset. Contour interval is 5 × 106 m2 s−1. Positive values are shaded. (b) Same as (a) but for PSA 2. (c) Same as (a) but for velocity potential. Contour interval 1 × 106 m2 s−1. Negative values are shaded. (d) Same as (c), but for PSA 2.

  • View in gallery

    Fig. A1. Same as Fig. 1 but for the DAO.

  • View in gallery

    Fig. A2. Same as Fig. 9 but for the DAO.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 391 391 63
PDF Downloads 702 702 53

The Pacific–South American Modes and Tropical Convection during the Southern Hemisphere Winter

View More View Less
  • 1 Climate Prediction Center, NWS/NCEP/NOAA, Washington D.C.
© Get Permissions
Full access

Abstract

Atmospheric circulation features and convection patterns associated with two leading low-frequency modes in the Southern Hemisphere (SH) are examined in multiyear global reanalyses produced by NCEP–NCAR and NASA–DAO. The two leading modes, referred to as the Pacific–South American (PSA) modes, are represented by the first two EOF patterns. The two patterns are in quadrature with each other and are dominated by wavenumber 3 in midlatitudes with large amplitudes in the Pacific–South American sector. In the Pacific, anomalies in the subtropics and in the midlatitudes are opposite in phase. Taken together, the two PSA modes represent the intraseasonal oscillation in the SH with periods of roughly 40 days. The evolution of the PSA modes shows a coherent eastward propagation.

A composite analysis was conducted to study the evolution of tropical convection and the corresponding circulation changes associated with the PSA modes. Outgoing longwave radiation (OLR) anomaly composites during the mature phase of the PSA modes resemble the first two leading EOFs of OLR anomalies (OLRA) in the Tropics. Composites of OLRA show an east–west dipole structure roughly 5–10 days prior to the onset of persistent PSA events. The PSA 1 mode is associated with enhanced convection in the Pacific between 140°E and 170°W and suppressed convection over the Indian Ocean. The PSA 2 mode is linked to tropical heating anomalies in the central Pacific extending from 160°E to 150°W just south of the equator and suppressed convection in the western Pacific with a maximum at 20°N. Contributions are from both interannual and intraseasonal bands.

Corresponding author address: Dr. Kingtse Mo, Climate Prediction Center, NOAA/NWS/NCEP, W/NP52, 4700 Silver Hill Rd., Stop 9910, Washington DC 20233-9910.

Abstract

Atmospheric circulation features and convection patterns associated with two leading low-frequency modes in the Southern Hemisphere (SH) are examined in multiyear global reanalyses produced by NCEP–NCAR and NASA–DAO. The two leading modes, referred to as the Pacific–South American (PSA) modes, are represented by the first two EOF patterns. The two patterns are in quadrature with each other and are dominated by wavenumber 3 in midlatitudes with large amplitudes in the Pacific–South American sector. In the Pacific, anomalies in the subtropics and in the midlatitudes are opposite in phase. Taken together, the two PSA modes represent the intraseasonal oscillation in the SH with periods of roughly 40 days. The evolution of the PSA modes shows a coherent eastward propagation.

A composite analysis was conducted to study the evolution of tropical convection and the corresponding circulation changes associated with the PSA modes. Outgoing longwave radiation (OLR) anomaly composites during the mature phase of the PSA modes resemble the first two leading EOFs of OLR anomalies (OLRA) in the Tropics. Composites of OLRA show an east–west dipole structure roughly 5–10 days prior to the onset of persistent PSA events. The PSA 1 mode is associated with enhanced convection in the Pacific between 140°E and 170°W and suppressed convection over the Indian Ocean. The PSA 2 mode is linked to tropical heating anomalies in the central Pacific extending from 160°E to 150°W just south of the equator and suppressed convection in the western Pacific with a maximum at 20°N. Contributions are from both interannual and intraseasonal bands.

Corresponding author address: Dr. Kingtse Mo, Climate Prediction Center, NOAA/NWS/NCEP, W/NP52, 4700 Silver Hill Rd., Stop 9910, Washington DC 20233-9910.

1. Introduction

The atmospheric circulation tends to reside in a few quasi-stationary states (or low frequency modes). In the Southern Hemisphere (SH), there are two leading low-frequency modes in the extratropics, which are characterized by two wavenumber 3 patterns in quadrature with each other and a well-defined wave train extending from the central Pacific to Argentina. Because they resemble the Pacific–North American teleconnection pattern (Wallace and Gutzler 1981), they are referred to as the Pacific–South American (PSA) patterns (Ghil and Mo 1991; Lau et al. 1994).

Both PSA modes are teleconnection patterns in the SH (Mo and White 1985). Rogers and van Loon (1982) performed an EOF analysis using monthly mean 500-hPa height anomalies from the Australian analyses. They identified two wave 3 patterns as the primary eigenmodes in the SH and related them to standing waves in the atmosphere. Similar patterns have also been found by Szeredi and Karoly (1987) using monthly mean station data. Kidson (1988, 1991) observed two wavenumber 3 modes in both interannual and intraseasonal bands using the ECMWF analyses. Ghil and Mo (1991) found that the second pair of modes of the SH intraseasonal oscillation is represented by two eastward propagating wave 3 modes in quadrature with each other. All of the above studies used 500-hPa height or sea level pressure anomalies so features in the Tropics and subtropics are weak. Lau et. al (1994) examined low-frequency modes in 200-hPa streamfunction anomalies using the NMC (National Meteorologial Center, now National Centers for Environmental Prediction) operational analyses and found two PSA modes in the SH (Lau et al. 1994). In addition to a wave 3 pattern in midlatitudes, their eigenfunctions clearly show a phase reversal between anomalies in the subtropics and anomalies in midlatitudes. The fact that the PSA modes were obtained from many different datasets suggests that they are robust and not artifacts of the analyses.

In the Northern Hemisphere (NH), Higgins and Mo (1997) demonstrated that the development of persistent North Pacific anomalies is often linked to the tropical intraseasonal oscillation (IO). In the SH, the subject of tropical–extratropical interactions is less well understood. Several studies have focused on tropical linkages to persistent anomaly patterns in the SH. For example, one of the PSA patterns is the SH response to the El Niño–Southern Oscillation (ENSO) (Karoly 1989), suggesting a tropical origin. Berbery and Paegle (1993) found that the teleconnections between the Tropics and the extratropics are strongly related to the 30–60-day oscillation. On the other hand, Ghil and Mo (1991) did not find a coherent relationship between the IO in the SH and in the Tropics. Lau et al. (1994) in their study of the low-frequency global modes, suggested that these two wave 3 modes were associated with internal dynamics and not forced by the Tropics.

Studies of tropical–extratropical interactions in the SH are limited by inconsistencies of the divergence or heating fields in analyses due to changes in models and assimilation schemes and by the lack of sufficient observations over the southern oceans. Recently, the National Centers for Environmental Prediction–National Center for Atmospheric Research (hereafter NCEP) and the Data Assimilation Office at NASA’s Goddard Space Flight Center (hereafter DAO) have started reanalysis projects. NCEP (DAO) has completed assimilations for the period from 1958 to 1997 (1980 to 1995). Documentation of the reanalyses and data archives can be found in Kalnay et al. (1996) and Schubert et al. (1993), respectively. The reanalyses were produced with fixed assimilation systems and the input database was not restricted by the operational data cutoff time. In data-sparse regions, the reanalyses depend on the model and analysis scheme so the intercomparison becomes important.

This study focuses on relationships between PSA modes and tropical convection. We use 200-hPa streamfunction anomalies to represent circulation anomalies instead of 500-hPa heights to enhance features in the subtropics. The zonal mean at each latitude is removed to get eddy streamfunction. In this study, we concentrate on the SH winter defined as the period from 15 May to 15 September. Datasets used in this study are described in section 2. Results using the NCEP reanalyses are presented next. Composites of circulation features during the mature phase of the PSA modes are given in section 3. The evolution of PSA modes is related to tropical convection in section 4. A summary and discussion are given in section 5. Key results are reproduced for the DAO reanalysis in the appendix.

2. Data

The primary data used in this study are daily global gridded data from the NCEP–NCAR reanalysis for the period from 1973 to 1995 on a 2.5° × 2.5° latitude–longitude grid. Daily averages of the NOAA satellite outgoing longwave radiation (OLR) data are used to represent tropical convection (Liebmann and Smith 1996). OLR data are available for the period from 1 June 1974 to 31 December 1995 with a gap from March to December 1978.

An error was found in the NCEP reanalysis. The Australian surface pressure bogus data (PAOBs) were used as input. Because of a decoder error, the PAOBs were read in 180° out of phase from 1979 to 1992. The PAOBs data were not used over land and roughly 50% of the data were rejected. This error has no impact on the Northern Hemisphere circulation. However, daily fields south of 40°S especially near the surface are affected by this error. This has minimal impact on the monthly mean fields and low-frequency variability because the accepted PAOBs vary each day. To assure that results are not influenced by the PAOBs, all calculations were performed for both the NCEP and the DAO reanalyses since the PAOBs were not used in the DAO. The NCEP data cover a longer period than the DAO, but all major results are reproduced in both sets.

The seasonal cycle at each grid point is defined as the grand mean plus the first and second harmonics with periods of 12 and 6 months, respectively. The difference between the field and the seasonal cycle is defined as the anomaly at that grid point. Low-pass-filtered fields (periods greater than 10 days) were obtained by applying the low-pass filter of Blackmon (1976). We refer to this as the low-frequency band (LF). To yield the intraseasonal signal with periods between 10 and 90 days, the daily anomalies were filtered using the minimum-bias taper developed by Papoulis (1973), which is the same filter used by Ghil and Mo (1991). We refer to this as the intraseasonal frequency band (IS).

3. The Pacific–South American modes

Empirical orthogonal functional (EOF) analysis was performed on the LF-filtered mean 200-hPa streamfunction anomalies for the SH winter. The spatial domain is from the South Pole to the equator. The zonal mean at each latitude was removed to obtain eddy streamfunction. Anomalies were not normalized but a latitudinal cosine weighting factor was used in computing the covariance matrix. The first two EOFs explain 7.2% and 6.9% of the variance in the LF band. These are also the first two leading rotated EOFs. The PSA modes are also found in the IS-filtered data. They explain 5.4% and 5.1% of the variance in the LF band. The two PSA (EOF) modes (Figs. 1a and 1b) are in quadrature with each other, consistent with oscillatory behavior. Both patterns have wave 3 in midlatitudes with large amplitudes in the Pacific–South American sector. In the Pacific, anomalies in the subtropics are out of phase with anomalies in midlatitudes. Patterns in the SH resemble rotated EOF patterns observed by Lau et al. (1994) in their study of multiscale low-frequency circulation modes using earlier NMC operational analyses. The PSA modes were also reported by many studies (Szeredi and Karoly 1987; Kidson 1988, 1991; and Ghil and Mo 1991) using different analyses over different time periods. Therefore, both PSA modes are not artifacts of the analyses.

The zonally asymmetric 200-hPa streamfunction anomalies are projected onto two PSA modes to yield the corresponding principal components (PCs). The relationship between the two PSA modes is evident from the lagged correlations between PC 1 and PC 2 during winter (Fig. 1c, open circles). The decorrelation time computed from the time-lagged autocorrelations is about 15 days. Assuming eight degrees of freedom per season, the correlation should be larger than 0.14 to be statistically significant at the 95% level. The two PCs are significantly correlated from lag 3 to lag 9 with the maximum lag correlation at 4–5 days. The correlations are higher in the IS band (dark circles) suggesting that the relationship is not just due to persistent anomalies in the interannual band. The significant correlations between the two modes suggest that these two modes represent the SH intraseasonal oscillation.

To determine the period of the oscillation, we performed singular spectrum analysis (SSA) on the PCs. SSA is more effective than conventional spectral analysis because SSA is not restricted to periodic cycles (Vautard and Ghil 1989). Quasi-periodic signals appear as pairs of degenerate eigenmodes and their corresponding eigenfunctions are in quadrature with each other. For details, readers are referred to Vautard and Ghil (1989) and Ghil and Mo (1991). We carried out SSA for the two PSA PCs for a window length of 60 days to highlight oscillations at intraseasonal timescales. Results are not sensitive to the particular window length used. Results for IS-filtered PSA PCs are given in Table 1, which lists the periods T and the variance associated with each SSA mode with a window size of 60. The error in the percentage of explained variance of the given PSA PC was estimated by a formula adapted by Vautard and Ghil (1989). Since the SSA modes are not pure sines and cosines, the dominant periods listed are estimated using a Blackman–Tukey analysis with a bandwidth of 0.0074. The first and second SSA modes for both PSA 1 and PSA 2 are degenerate with eigenfunctions in quadrature with each other so they represent the IO in the SH with periods of about 40 days. Together they explain 4.5% (3.6%) of the variance in the IS (LF) band. In addition to the 40-day modes, the five and six SSA modes for both PSA PCS represent the IO with periods of 18 days and there is also a 25-day mode found in PSA 2 PC. For unfiltered PCs, the first SSA mode represents a low-frequency component that the current window size cannot resolve. The 40-day modes come as the second and third SSA modes.

If we do not require persistence, we can define positive or negative PSA 1 and PSA 2 events when the corresponding PC is greater than 0.8 standard deviations (or less than −0.8 standard deviations). On average, the flow will persist for 6.2 and 5.8 days for positive and negative PSA 1, respectively. The persistence time for positive and negative PSA 2 is 5.2 and 4.6 days, respectively.

In this paper, we focus on strong persistent events. We select persistent PSA 1 and PSA 2 events according to the procedure outlined by Dole and Gordon (1983). A positive (negative) PSA 1 or PSA 2 event is identified when the corresponding PC is greater than (less than) 1.2 standard deviations for at least 8 days. The events vary when the threshold changes from 0.8 to 1.5 standard deviations, but results reported here are not sensitive to the threshold used to select events. There are 21 positive and 20 negative PSA 1 events with an average duration of 12 days. There are 20 positive and 16 negative PSA 2 events with an average duration of 9.5 days. Composites of 200-hPa eddy streamfunction anomalies and OLR anomalies (OLRAs) for positive and negative PSA 1 and PSA 2 events were produced by averaging over the duration of each event and by averaging all events in the same category. The statistical significance of each composite was assessed by assuming that anomalies obey a normal distribution. Each event is considered as one degree of freedom. Areas where values are statistically significant at the 95% levels are shaded. Figure 2 shows the evolution of the composite mean 200-hPa eddy streamfuction anomalies. The evolution of two PSA modes is summed up as follows:

After the flow leaves one category of events, the path outlined above is also the most likely path for the flow to take. In many cases, the flow pattern evolves from one PSA mode to another and becomes stationary for some time and then propagates eastward again. In some cases, the flow does not belong to any PSA mode after exiting from one. For example, after the flow leaves negative PSA 2, it may go to the next PSA mode immediately, or it may not have large projection on any PSA mode for a while. Statistically, the next PSA pattern mostly likely to occur is a positive PSA 1. The average transition time from one category to the next along the path is 4.9 days.

For positive PSA 1 (Fig. 2a), a dipole straddles the equator in the western Pacific, accompanied by a downstream wave train. As the pattern evolves, a four-cell quadrupole pattern in the central Pacific and a wavenumber 3 pattern in midlatitudes propagate eastward together (Figs. 2b–d). Generally, the anomaly patterns for positive and negative events are similar but with a sign reversal. Long-lived traveling patterns have also been observed in the NH. Branstator (1987) identified a large-amplitude traveling flow anomaly during the NH winter of 1979–80. Unlike the PSA modes, the waves in the NH propagate westward with a period of 23 days.

Figure 3 presents the OLRA differences in the Indian–Pacific Ocean sector between positive and negative composites for both PSA modes. The convection pattern associated with PSA 1 (Fig. 3a) shows enhanced convection in the central Pacific and suppressed convection in the Indian Ocean. Positive anomalies to the north and south of negative anomalies signal an enhanced local Hadley circulation. There is also an enhanced ITCZ. This pattern (Fig. 3a) bears resemblance to the OLRA pattern during warm ENSO events which have been linked to the PSA 1 mode by Karoly (1989) at interannual timescales. Figure 3a also resembles the tropical OLRA pattern during the 30–60-day IO (Lau and Chan 1986). The OLRA composite for PSA 2 shows suppressed convection in the western Pacific with a maximum at 20°N and enhanced convection in the central Pacific extending from 160°E to 150°W just south of the equator and over the Indian Ocean.

The above OLRA composites (Fig. 3) are keyed to the two PSA modes. If the relationship between the PSA modes and tropical convection is robust, we should be able to recover these patterns from composites keyed to tropical convection. To examine this further, we performed EOF analysis on the low-pass-filtered OLRA for the SH winter. Since there is a gap in the OLRA data, we used data from 1979 to 1995. To compare with the OLRA composites (Fig. 3), we concentrated on the Indian–Pacific Ocean sector. Anomalies were not normalized but a latitudinal cosine weighting factor was used in computing the covariance matrix. The first three EOFs explain 6.8%, 5.8%, and 5.1% of the variance in the LF band, respectively. The first EOF is also the first EOF in the IS-filtered data, which explains 5.7% of the variance in the LF band. But the second (third) EOF becomes the third (second) EOF in the IS-filtered data and they explain 4.3% and 3.9% of the variance in the LF band, respectively. Our results agree with the findings of Lau and Chan (1986), even though they used a shorter data period (1975–82) and a different filter (15–180 days). The OLRA data from 1979 to 1995 were projected onto EOF modes (Fig. 4) to yield the corresponding principal components (PCs). SSA analysis performed on the three OLRA PCs indicates that the leading oscillatory modes for all PCs have a period of 48 days. The three EOFs represent a typical IO cycle during the SH winter. Enhanced tropical convection moves eastward from the Indian Ocean (EOF 3) and passes the western Pacific (EOF 1) to the central and eastern Pacific (EOF 2).

The first two OLRA EOFs (Figs. 4a and 4b) bear a striking resemblance to the OLRA composites (Fig. 3) based on persistent PSA events. This suggests that both PSA modes are associated with tropical convection. Next, we compute lead and lag correlations between OLRA PCs and PSA PCs during the SH winter. The OLRA PC 1 leads the PSA 1 PC by 5–8 days with a correlation of 0.36. The OLRA PC 2 leads PSA 2 PC by 2–5 days with a correlation of 0.33. They are significant at the 95% level. Results are similar for unfiltered anomalies and IS filtered anomalies. These results suggest that the PSA modes are linked to tropical convection with contributions from both interannual and intraseasonal bands. Both Ghil and Mo (1991) and Lau et al. (1994) did not find significant linkages between PSA modes and the tropical IO. One possibility is that the analyses in previous studies had discontinuities due to changes in the model and assimilation schemes that would lower correlations. Another major difference is that previous studies pooled all seasons together while here we concentrate on the winter season. Seasonal variations of PSA modes are not large, but seasonal changes for the tropical OLRA are quite large and they are not symmetric about the equator (Knutson and Weickmann 1987). The leading EOFs for tropical OLRA vary from one season to another. Thus, combining seasons would weaken the signal.

The PSA modes represent the IO in the SH, which has shorter periods than the tropical IO. This suggests that the PSA modes are not purely responses to the tropical IO. The tropical IO and the PSA modes may have different origins but they do interact episodically. When the tropical IO is strong, it creates a favorable situation for the PSA modes to intensify. In the next section, we provide evidence that the tropical IO serves as a catalyst in the development of large PSA modes.

4. Evolution of PSA modes associated with tropical convection

Lagged composites of several variables were produced from 15 days before onset to 10 days after onset for both positive and negative PSA 1 and PSA 2 events. Since composites for positive and negative anomalies are similar except for a sign reversal, we present the composite differences. The statistical significance was assessed by assuming that anomalies obey a normal distribution and that each event is one degree of freedom. Areas where values are statistically significant at the 95% level are shaded. Figures 5 and 6 show the composite differences (positive minus negative) of pentad mean OLRA for PSA 1 and PSA 2 events, while Figs. 7 and 8 show the same plots but for IS-filtered OLRA composites. The 200-hPa streamfunction anomalies respond about 10 days before onset so the composite pentad mean differences for 200-hPa velocity potential and streamfunction anomalies are plotted from two pentads before onset to one pentad after onset (Figs. 9 and 10).

For PSA 1, three pentads before onset (day −15 to day −11), the OLRA intensify along the equator with one maximum over the Indian Ocean and another in the central Pacific near the date line (Fig. 5a). As time progresses, enhanced convection in the central Pacific intensifies (negative OLRA) while suppressed convection (positive OLRA) moves from the Indian Ocean to the western Pacific (Figs. 5b–d). The IS-filtered OLRA composites show a similar evolution (Fig. 7). Three pentads before onset, enhanced convection is located in the Indian Ocean. As time progresses, OLRA in the central Pacific intensify while positive OLRA propagate eastward. The magnitudes of the anomalies in the IS band are about 25%–35% weaker. Since the OLRA exhibit large variance in the IS band to the west of the date line, the signals associated with the tropical IO are weaker in the eastern Pacific. The weaker responses in the IO band indicate contributions from the interannual band, consistent with the findings of Karoly (1989), and many others that the PSA 1 mode responds to ENSO.

Two pentads before onset the response of the 200-hPa streamfunction anomaly composite shows a dipole structure at 15°S in the western Pacific near the convective area and a weak wave train downstream (Fig. 9a). The 200-hPa velocity potential anomaly field responds with a dipole pattern dominated by wavenumber 1 (Fig. 10a). Throughout the evolution, the 200-hPa velocity potential anomalies move eastward in concert with the OLRA (Figs. 5, 7, and 10). One pentad before onset, in the SH, the correlation between the 200-hPa streamfunction composite (Fig. 9b) and PSA 2 (Fig. 1b) is −0.56 so the wave train is in phase with negative PSA 2. When the OLRA move eastward, the streamfunction anomalies also propagate eastward. One pentad after onset (day 0 to day 4), the OLRA composite resembles OLRA EOF 1 (correlation 0.75) and the 200-hPa streamfunction anomaly pattern correlates well with positive PSA 1 (correlation 0.92).

For PSA 2, three pentads before onset, the OLRA composite (Fig. 6a) shows enhanced convection in the central and eastern Pacific along the equator with positive OLRA located in the subsiding areas of the local Hadley circulation. At day −10, the correlation between the OLRA composite and OLRA EOF 1 is 0.42. A pentad later, the 200-hPa streamfunction anomaly composite starts to show a dipole near the convective area. Enhanced convection in the central Pacific continues to propagate eastward. When convection moves out of the eastern Pacific, it reappears over the Indian Ocean. Meanwhile, positive OLRA in the western Pacific intensify. The wave train extending from the central Pacific to South America appears in the 200-hPa streamfunction composite (Fig. 9e) between day −6 and day −4. One pentad before onset, the streamfunction anomaly pattern resembles positive PSA 1 (correlation between Fig. 9e and PSA 1 is 0.6). The subtropical dipole and wave 3 propagate eastward together. One pentad after onset (day 0 to day 4), the 200-hPa streamfunction pattern is in phase with positive PSA 2 (correlation is 0.95) and the OLRA composite resembles OLRA EOF 2 (correlation 0.59). Throughout the evolution, the 200-hPa velocity potential anomalies move in concert with the OLRA and show an eastward-moving dipole dominated by wavenumber 1 (Fig. 10). There is a quadrature relationship between the velocity potential and streamfunction anomalies associated with PSA 1 and PSA 2 composites. For PSA 2, the contributions from the IS band are slightly larger. Again, Fig. 8 and Fig. 6 are similar, but the magnitudes of the anomalies in the IS band are about 15%–25% weaker in the Indian Ocean and western Pacific. The signals are much weaker in the eastern Pacific where the tropical IO is weak.

To sum up, we plotted a time–longitude cross section of the unfiltered OLRA as well as the IS-filtered anomalies keyed to PSA PCs along the equator (Fig. 11) and the evolution of 200-hPa velocity potential anomalies along the equator and 200-hPa streamfunction anomalies averaged from 50° to 60°S (Fig. 12). About 15 days before the onset of PSA 1 events, negative OLRA strengthen in the central Pacific and reach a minimum roughly 6–8 days before onset. At the same time, positive OLRA propagate from the Indian Ocean to the western Pacific. The wave train starts to develop about day −5 with large amplitudes in the South Pacific. It propagates eastward and reaches a maximum at day +4. For PSA 2, enhanced convection appears in the central and eastern Pacific with suppressed convection in the western Pacific about 10–12 days before onset. When negative anomalies move out of the Pacific, they reappear in the Indian Ocean. The OLRA keyed to PSA 2 is in quadrature with the OLRA keyed to PSA 1. For PSA 2, the wave train starts to appear about day −5 in the 200-hPa streamfunction composite and moves eastward. As anomalies diminish upstream, they strengthen downstream. The velocity potential anomalies keyed to PSA 1 and PSA 2 show a wavenumber 1 pattern in quadrature with each other. The evolution of the IS-filtered OLRA is similar to the total OLRA. Both show eastward propagation of the OLRA dipole, but the magnitudes of anomalies are 20%–30% weaker. The evolution and tropical linkage here are similar to the development of persistent North Pacific anomalies (Higgins and Mo 1997). Both show that enhanced tropical convection occurs prior to onset and an eastward-moving wave train from the South (North) Pacific downstream to South (North) America. The striking similarities between the two cases suggest that similar mechanisms may be involved.

5. Conclusions

We examined the evolution of the leading low-frequency modes in the SH and their linkage to tropical convection during the SH winter. The leading EOFs for both NCEP and the DAO reanalyses (see the appendix) are two PSA modes in quadrature with each other representing the IO in the SH. They show a wavenumber 3 pattern in midlatitudes with large amplitudes in the Pacific–South American sector. Both patterns have been observed in interannual (Kidson 1988) and intraseasonal bands (Kidson 1991; Lau et al. 1994; Ghil and Mo 1991). Two wave 3 patterns also exist in monthly mean station data. These studies indicate that the two patterns are not artifacts of the reanalyses. The correlations between the PCs for NCEP and the DAO during the overlapping period (1985–93) are 0.99 so the results appear to be robust. Low-frequency variability is not affected by the incorrectly assimilated PAOBs data in NCEP reanalysis from 1979 to 1992.

The evolution of the two patterns shows a wavenumber 3 pattern in midlatitudes propagating eastward. In the western hemisphere, anomalies in the subtropics and anomalies in midlatitudes are opposite in phase and propagate eastward together. The OLRA composites during the mature phase of the PSA modes resemble the first two leading EOF patterns for total OLRA and intraseasonal filtered OLRA.

About 10 days before onset of the PSA 1 mode, the OLRA along the equator show enhanced convection in the central Pacific between 140°E and 160°W. Soon afterwards, a dipole appears at 15°S in the 200-hPa streamfunction composite near the convective area and a wave train extends from the South Pacific downstream. The wave train and dipole propagate eastward together with the OLRA. During the mature phase of PSA 1, the OLRA composite shows a dipole with enhanced convection in the central Pacific and suppressed convection in the western Pacific, which is the signature of the tropical IO (Lau and Chan 1986; Knutson and Weickmann 1987). For PSA 2, both OLRA and 200-hPa streamfunction anomalies are in quadrature with anomalies for PSA 1. The similarity between the evolution of total OLRA and the IS-filtered anomalies indicates that the same tropical–extratropical linkages exist in both interannual and intraseasonal bands. The PSA modes and the tropical IO have different periods suggesting that the PSA modes are not pure responses to the tropical IO, and that they interact episodically.

Results here suggest that tropical convection serves as a catalyst in the development of large PSA modes. For persistent anomalies in the NH, Higgins and Schubert (1994, 1996), and Dole and Black (1990) demonstrated the importance of synoptic-scale forcing during the life cycle of persistent anomaly events. The interaction between zonal flow and synoptic-scale eddies may also be important in maintaining the persistent anomalies in SH as illustrated by a case study reported by Trenberth (1986a,b). Because of the PAOBs error, we are not able to examine the role played by the synoptic-scale eddies. Higgins and Mo (1997) showed that the persistent anomalies in the North Pacific evolve coherently with the tropical IO. The linkages between the tropical IO and the PSA modes are very similar to the North Pacific persistent anomalies. When the tropical IO is strong, it maintains the phase locking between the Tropics and the extratropics. The tropical forcing creates a favorable situation for a particular phase of the PSA modes to strengthen.

Acknowledgments

We would like to thank Dr. George Kiladis for the OLR data. This work was partially supported by the Office of Global Programs under the Pan American Climate Studies and the CDAS reanalysis projects and by Interagency Agreement S-41367-F under the authority of NASA Headquarters.

REFERENCES

  • Berbery, E. H., and J. N. Paegle, 1993: Intraseasonal interactions between the Tropics and extratropics in the Southern Hemisphere. J. Atmos. Sci.,50, 1950–1965.

  • Blackmon, M. L., 1976: A climatological spectral study of the 500 mb height of the Northern Hemisphere. J. Atmos. Sci.,33, 1607–1623.

  • Branstator, G., 1987: A striking example of the atmosphere’s leading traveling pattern. J. Atmos. Sci.,44, 2310–2323.

  • Dole, R. M., and N. D. Gordon, 1983: Persistent anomalies of the extratropical Northern Hemisphere wintertime circulation: Geographical distribution and regional persistent characteristics. Mon. Wea. Rev.,111, 1538–1553.

  • ——, and R. X. Black, 1990: Life cycles of persistent anomalies. Part II: The development of persistent negative height anomalies over the North Pacific Ocean. Mon. Wea. Rev., 118, 824–846.

  • Ghil, M., and K. C. Mo, 1991: Intraseasonal oscillations in the global atmosphere. Part II: Southern Hemisphere. J. Atmos. Sci.,48, 780–790.

  • Higgins, R. W., and S. D. Schubert, 1994: Simulated life cycles of persistent anticyclonic anomalies over the North Pacific: Role of synoptic scale eddies. J. Atmos. Sci.,51, 3238–3260.

  • ——, and ——, 1996: Simulations of persistent North Pacific circulation anomalies and interhemispheric teleconnections. J. Atmos. Sci.,53, 188–207.

  • ——, and K. C. Mo, 1997: Persistent North Pacific circulation anomalies and the tropical intraseasonal oscillation. J. Climate,10, 223–244.

  • Kalnay E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc.77, 437–471,.

  • Karoly, D. J., 1989: Southern Hemisphere circulation features associated with El Nino–Southern Oscillation. J. Climate,2, 1239–1252.

  • Kidson, J. W., 1988: Interannual variations in the Southern Hemisphere circulation. J. Climate,1, 1177–1198.

  • ——, 1991: Intraseasonal variation in the Southern Hemisphere circulation. J. Climate,4, 939–953.

  • Knutson, T. R., and K. M. Weickmann, 1987: 30–60 day atmospheric oscillations: composite life cycle of convection and circulation anomalies. Mon. Wea. Rev.,115, 1407–1436.

  • Lau, M. K., and P. H. Chan, 1986: Aspects of the 40–50 day oscillation during the northern summer as inferred from outgoing long wave radiation. Mon. Wea. Rev.,114, 1354–1367.

  • ——, P. J. Sheu, and I. S. Kang, 1994: Multiscale low-frequency circulation modes in the global atmosphere. J. Atmos. Sci., 51, 2753–2750.

  • Liebmann B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 1275–1277.

  • Mo, K. C., and G. H. White, 1985: Teleconnections in the Southern Hemisphere. Mon. Wea. Rev.,113, 22–37.

  • Papoulis, A., 1973: Minimum bias windows for high resolution spectral estimates. IEEE Trans. Info. Theory, 19, 9–12.

  • Rogers, J. C., and H. van Loon, 1982: Spatial variability of sea level pressure and height anomalies over the Southern Hemisphere. Mon. Wea. Rev., 110, 1375–1392.

  • Schubert, S. D., R. B. Rood and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc., 74, 2331–2342.

  • Szeredi, I., and D. J. Karoly, 1987: The horizontal structure of monthly fluctuations of the Southern Hemisphere troposphere from station data. Aust. Meteor. Mag., 35, 119–129.

  • Trenberth, K. E., 1986a: The signature of a blocking episode on the central circulation in the Southern Hemisphere. J. Atmos. Sci.,43, 2061–2069.

  • ——, 1986b: An assessment of the impact of transient eddies on the zonal flow during a blocking episode using localized Eliassen–Palm flux diagnostics. J. Atmos. Sci.,43, 2070–2087.

  • Vautard, R., and M. Ghil, 1989: Singular spectrum analysis in nonlinear dynamics with applications to paleoclimatic time series. Physica,35D, 392–424.

  • 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.

APPENDIX

Results Reproduced for the DAO Reanalysis

In this appendix, we repeated calculations using the DAO reanalysis. The DAO reanalysis covers 10 years from 1 March 1983 to 30 September 1993. The first two EOFs (Fig. A1) from the 200-hPa streamfunction anomalies are essentially the same as the first two leading EOFs for NCEP (Fig. 1). They explain 8.8% and 8.0% of the variance in the LF band, which is slightly larger than for NCEP. All major features in NCEP EOFs are reproduced. Differences are found in the Indian Ocean where features are weak.

The eddy 200-hPa streamfunction anomalies were projected onto these two EOFs to get the corresponding principal components. The simultaneous correlations between PCs for NCEP and the DAO during the overlapping period are 0.99 for both PSA 1 and PSA 2 modes. This suggests that the PAOBs problem has little effect on the PSA modes. Our results are robust. Since the PCs from the two analyses are the same, PSA 1 and PSA 2 persistent events are also the same during the 10 overlapping winters.

To compare the evolution of the two PSA modes, we produced lagged composites of 200-hPa streamfunction anomalies from 10 days before to 10 days after onset for both positive and negative PSA 1 and PSA 2 events for the DAO. Results are shown in Fig. A2, which should be compared to composites from NCEP (Fig. 9). There are differences, but the key features such as the dipole near the area of convection and the wavetrain from the South Pacific to South America are similar. In each case, there is a wavenumber 3 pattern in midlatitudes and a quadrupole in the subtropics propagating eastward together. Again, major differences are found in the Indian Ocean and in Asia where anomalies are weak. These results suggest that our results are robust. The PAOBs problem does not influence our conclusions.

Fig. 1.
Fig. 1.

(a) EOF 1 and (b) EOF 2 for the 200-hPa eddy streamfunction. EOFs are normalized to 1 and time 100. Contour interval is 0.5 nondimensional unit. (c) Correlations as a function of lead and lag time between PC 1 and PC 2 for the SH winter for total anomalies (open circles) and the (10–90 day) IS-filtered time series (dark circles).

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 2.
Fig. 2.

The 200-hPa eddy streamfunction anomaly composite averaged over all (a) positive PSA 1, (b) positive PSA 2, (c) negative PSA 2, and (d) negative PSA 1 days. Contour interval is 3 × 106 m2 s−1. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 3.
Fig. 3.

OLRA composite difference between positive and negative (a) PSA 1 and (b) PSA 2 events. Contour interval is 5 W m−2. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 4.
Fig. 4.

(a) EOF 1, (b) EOF 2, and (c) EOF 3 for OLRA for the SH winter. EOFs are normalized to 1 and time 100. Contour interval is 0.8 nondimensional unit.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 5.
Fig. 5.

Map sequences of OLRA represented as the composite difference between positive and negative PSA 1 for the pentad centered at (a) day −13 (b) day −8, (c) day −3, and (d) day 2. Contour interval is 5 W m−2. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 6.
Fig. 6.

Same as Fig. 5 but for PSA 2 events.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 7.
Fig. 7.

Same as Fig. 5 but for IS-filtered OLRA. Contour interval is 4 W m−2.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 8.
Fig. 8.

Same as Fig. 6 but for IS-filtered OLRA. Contour interval is 4 W m−2.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 9.
Fig. 9.

Map sequences of 200-hPa eddy streamfunction anomaly composite difference between positive and negative (a) PSA 1 for the pentad centered at day −8. (b) Same as (a), but for the pentad centered at day −3. (c) Same as (a), but for the pentad centered at day 2. (d) PSA 2 for the pentad centered at day −8. (e) Same as (d), but for the pentad centered at day −3. (f) Same as (d), but for the pentad centered at day 2. Contour interval is 5 × 106 m2 s−1. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 10.
Fig. 10.

Same as Fig. 9 but for velocity potential. Contour interval is 2 × 10 6 m2 s−1. Zero contours are omitted. Areas where values are statistically significant at the 95% level are shaded.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 11.
Fig. 11.

(a) Hovmöller diagram for total OLRA composite difference averaged from 5°S to 5°N between positive and negative PSA 1 events from 15 days before to 10 days after onset. Contour interval is 5 W m−2. Areas where values are significant at the 95% level are shaded. (b) Same as (a) but for IS-filtered OLRA. Contour interval is 3 W m−2. (c) Same as (a) but for PSA 2 mode. (d) Same as (c) but for IS-filtered OLRA.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Fig. 12.
Fig. 12.

(a) Hovmöller diagram for the 200-hPa eddy streamfunction composite difference averaged from 50° to 60°S between positive and negative PSA 1 from 10 days before to 10 days after onset. Contour interval is 5 × 106 m2 s−1. Positive values are shaded. (b) Same as (a) but for PSA 2. (c) Same as (a) but for velocity potential. Contour interval 1 × 106 m2 s−1. Negative values are shaded. (d) Same as (c), but for PSA 2.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

i1520-0493-126-6-1581-fa1

Fig. A1. Same as Fig. 1 but for the DAO.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

i1520-0493-126-6-1581-fa2

Fig. A2. Same as Fig. 9 but for the DAO.

Citation: Monthly Weather Review 126, 6; 10.1175/1520-0493(1998)126<1581:TPSAMA>2.0.CO;2

Table 1.

SSA modes for PSA PCs in the IS band (10–90 days).

Table 1.
Save