Principal Component Time Series Research Articles

Bayesian wavelet PCA methodology for turbomachinery damage diagnosis under uncertainty

Centrifugal compressor often suffers various defects such as impeller cracking, resulting in forced outage of the total plant. Damage diagnostics and condition monitoring of such a turbomachinery system has become an increasingly important and powerful tool to prevent potential failure in components and reduce unplanned forced outage and further maintenance costs, while improving reliability, availability and maintainability of a turbomachinery system. This paper presents a probabilistic signal processing methodology for damage diagnostics using multiple time history data collected from different locations of a turbomachine, considering data uncertainty and multivariate correlation. The proposed methodology is based on the integration of three advanced state-of-the-art data mining techniques: discrete wavelet packet transform, Bayesian hypothesis testing, and probabilistic principal component analysis. The multiresolution wavelet analysis approach is employed to decompose a time series signal into different levels of wavelet coefficients. These coefficients represent multiple time-frequency resolutions of a signal. Bayesian hypothesis testing is then applied to each level of wavelet coefficient to remove possible imperfections. The ratio of posterior odds Bayesian approach provides a direct means to assess whether there is imperfection in the decomposed coefficients, thus avoiding over-denoising. Power spectral density estimated by the Welch method is utilized to evaluate the effectiveness of Bayesian wavelet cleansing method. Furthermore, the probabilistic principal component analysis approach is developed to reduce dimensionality of multiple time series and to address multivariate correlation and data uncertainty for damage diagnostics. The proposed methodology and generalized framework is demonstrated with a set of sensor data collected from a real-world centrifugal compressor with impeller cracks, through both time series and contour analyses of vibration signal and principal components.

Forced Atmospheric Teleconnections during 1979–2014

Abstract Forced atmospheric teleconnections during 1979–2014 are examined using a 50-member ensemble of atmospheric general circulation model (AGCM) simulations subjected to observed variations in sea surface temperatures (SSTs), sea ice, and carbon dioxide. Three primary modes of forced variability are identified using empirical orthogonal function (EOF) analysis of the ensemble mean wintertime extratropical Northern Hemisphere 500-hPa heights. The principal component time series of the first and second modes are highly correlated with Niño-3.4 and trans-Niño (TNI) SST indices, respectively, indicating mostly tropical sources. Their impacts are largely confined to the Pacific–North American (PNA) sector. The leading mode describes the canonical atmospheric teleconnection associated with El Niño–Southern Oscillation (ENSO) resembling the tropical/Northern Hemisphere pattern. The second mode describes a wave train resembling the classic PNA pattern resulting from atmospheric sensitivity to ENSO asymmetry and from sensitivity to a tropical precursor SST for ENSO development. The third mode is characterized by a hemisphere-scale increasing trend in heights. Based on a comparison with 50-member coupled ocean–atmosphere model simulations, it is argued that this mode is strongly related to radiatively forced climate change, while the other two forced teleconnections are principally related to internal coupled variability. A trend in the leading forced mode is related to ENSO-like decadal variability and dominates the overall observed 500-hPa height trend since 1979. These model results indicate that the trend in the first mode is due to internal variability rather than external radiative forcing.

The inter‐annual variability of southerly low‐level jets in North America

The inter-annual variability of southerly low-level jets (SLLJs) over North America during the warm (April–September) and cool (October–March) seasons is investigated. SLLJ occurrences over a 31-year period (1979–2009) were identified from the North American Regional Reanalysis (NARR) vertical wind profiles. The first empirical orthogonal function (EOF) modes of the SLLJ frequency during the warm and cool seasons account for about 30 and 20% of the total variance, respectively. Both modes can be interpreted as a strengthening or weakening of the core area of SLLJ anomalies. The principal component (PC) time series display significant positive trends, suggesting an increase in SLLJ activity during both seasons on inter-decadal time scales and are significantly correlated to the summertime Pacific Decadal Oscillation (PDO) and Atlantic Multidecadal Oscillation (AMO) for the warm season and the wintertime PDO, AMO and El Nino Modoki for the cool season. The second modes account for about 20 and 15% of the total variance for the warm and cool seasons, respectively, and are interpreted as primarily a subseasonal latitudinal shift in SLLJ activity between the central Great Plains and the western Gulf of Mexico and southern Texas during the warm season and a longitudinal shift between the western Gulf of Mexico and the Caribbean during the cool season. The second mode appears to be significantly correlated to El Nino Modoki for the warm season and to Nino 3.4 for the cool season.

Interannual variability of heat waves in South Korea and their connection with large‐scale atmospheric circulation patterns

ABSTRACTThis study investigates the interannual variation of heat wave frequency (HWF) in South Korea during the past 42 years (1973–2014) and examines its connection with large‐scale atmospheric circulation changes. Korean heat waves tend to develop most frequently in late summer during July and August. The leading Empirical Orthogonal Function accounting for 50% of the total variance shows a mono‐signed pattern over South Korea, suggesting that the dominant mechanisms responsible for the heat wave are linked in a spatial scale much larger than the nation. It also exhibits a regional variation with more occurrences in the southeastern inland area. The regression of the leading principal component (PC) time series of HWF with large‐scale atmospheric circulation identifies a north–south dipole pattern between the South China Sea and Northeast Asia. When this large‐scale circulation mode facilitates deep convection in South China Sea, it tends to weaken moisture transport from the South China Sea to Northeast Asia. Enhanced deep convection in the South China Sea triggers a source of Rossby wave train along southerly wind that generates positive geopotential height anomalies around Korea. The anomalous high pressure pattern is accompanied by large‐scale subsidence in Korea, thereby providing a favourable condition for extreme hot and dry days in Korea. This study highlights that there is a decadal change of the relationship between Korean heat waves and large‐scale atmospheric circulation patterns. The tropical forcing tends to be weakened in the recent decade, with more influences from the Arctic variability from the mid‐1990s.

Spatial and temporal features of summer extreme temperature over China during 1960–2013

Based on daily air temperature data from 772 stations in China, the present study uses absolute index and percentile index to investigate the spatial and temporal features of summer extreme temperature over China during the period 1960–2013. The analysis indicates that Xinjiang and southeastern China are two major domains where extreme heat events frequently occur and that the number of heat day (NHD) and the frequency of heat wave (FHW) both show an increasing trend throughout the country except for Shandong and Henan provinces where a decreasing trend is identified. Although the two leading empirical orthogonal function (EOF) modes of the NHD and the FHW based on the absolute index and percentile index have differences, the time series of the first principal components (PC1) are consistent; PC1 depicts opposite trends from 1960 to late 1990s and during the late 1990s to 2013. According to the climatology and EOF modes, four sub-regions are chosen: Chuanyu, Huanghuai, Southeast, and Xinjiang area. The inter-decadal variation over the four sub-regions differs, but the NHD and the FHW significantly increase after the mid-1990s. Based on Mann-Kendall method, it is found that the NHD and the FHW over Chuanyu exhibited abrupt shifts in 1978 and 2000; sudden shifts occurred in 1973 and 2000 over Huanghuai; an abrupt shift occurred over the Southeast area in 2003.

Predictable signals in seasonal mean soil moisture simulated with observation-based atmospheric forcing over China

The Community Land Model version 3.5 is driven by an observation-based meteorological dataset to simulate soil moisture over China for the period 1951–2008. A method for identifying the patterns of interannual variability that arise from slow (potentially predictable) and intraseasonal (unpredictable) variability is also applied; this allows identification of the sources of the predictability of seasonal soil moisture in China, during March–April–May (MAM), June–July–August (JJA), September–October–November (SON) and December–January–February (DJF). The potential predictability (slow-to-total) of the soil moisture above 1 m is high, with lowest value of 0.76 in JJA and highest value of 0.94 in DJF. The spatial distribution of the potential predictability comprises a northwest–southeast gradient, with a minimum center over East China and a maximum center over the northwest. The most important source of predictability is from the soil moisture persistence, which generally accounts for more than 50 % of the variability in soil moisture. The SSTs in the Indian Ocean, the North Atlantic and the eastern tropical Pacific Oceans are also identified as important sources of variability in the soil moisture, during MAM, JJA and SON/DJF, respectively. In addition, prolonged linear trends in each season are an important source. Using the slow principal component time series as predictands, a statistical scheme for the seasonal forecasting of soil moisture across China is developed. The prediction skills, in terms of the percentage of explained variance for the verification period (1992–2008), are 59, 51, 62 and 77 % during MAM–DJF, respectively. This is considerably higher than a normal grid prediction scheme.

On the Annual Variations in the Amplitude of 25-70-Day Intraseasonal Atmospheric Oscillations in Central Africa

In this paper we analyzed the annual variations in the 25-70-day intraseasonal atmospheric oscillations in central Africa, for the period 1981-2010, using the Outgoing Longwave OLR data. We then extracted the amplitude time series of the dominant modes of intraseasonal variability in 25-70 days filtered OLR anomalies, using Empirical Orthogonal Functions (EOF) analysis. The EOF analysis has shown that three dominant modes characterized the intraseasonal atmospheric oscillation in Central Africa. The amount of variance explained by these three retained EOFs are 19.3%, 13.6% and 11.8% respectively, and they exhibit higher spatial loading over Northern Congo, Southern Ethiopia, and Southwestern Tanzania, respectively. The analysis of Principal Components (PCs) time series showed that the amplitude and of the intraseasonal oscillations (ISO) exhibit large annual variations. In fact the highest values of ISO amplitude are generally observed during October-April season, and much weakened signal the rest of the year. The fraction of yearly Madden Julian Oscillation (MJO) power, occurring within October-April season are 79.3%, 77.92%, 78.73% for EOF1, EOF2, and EOF3, respectively.

On the Differences in the Intraseasonal Rainfall Variability between Western and Eastern Central Africa: Case of 25 - 70-Day Oscillations

The intraseasonal timescale is critical in Central Africa, because the resources of the region are highly rainfall dependent. In this paper, we use 1DD GPCP rainfall product to investigate the differences in the space-time structures of the 25 - 70-day intraseasonal variability of rainfall, over the Western Central Africa (WCA) and the Eastern Central Africa (ECA), with different climate features. The results of Empirical Orthogonal Functions (EOFs) analysis have shown that the amount of variance explained by the leading EOFs is greater in ECA (58.4%) than in WCA (49.8%). For both WCA and ECA, the power spectra of the Principal Components (PCs) peaked around 40 days, indicating a MJO signal. The seasonality of ISO is evident, but this seasonality is much noticeable in ECA where almost 80% of the total yearly ISO power occurs during November-April season, against only around 60% for WCA. Moreover, the lagged cross correlations computed between WCA and ECA PCs time series showed that most of the WCA PCs led ECA PCs time series with a timescale of 8 - 12 days, revealing that the eastward propagation could potentially be the relationship between WCA and ECA modes. The interannual variations in the ISO activity are weak in WCA, when compared with ECA where the signal exhibits larger interannual variations, quite linked with ENSO.

Effect of Northern Hemisphere Teleconnections on the Hydropower Production in Southern Sweden

Approximately 50% of Sweden’s electricity is produced by hydropower. This makes energy production in the country vulnerable to factors affecting water availability. Research has shown a positive correlation between the North Atlantic Oscillation (NAO) and hydropower production in Norway and northern Sweden during winter months. The correlation is, however, weaker in southern Sweden, which indicates that there might be other low-frequency atmospheric drivers (teleconnections) affecting this area. The aim of this paper is to analyze the natural climate forcing that affects hydropower production in southern Sweden on a seasonal basis. This knowledge may be of key importance in improving long-term (seasonal to yearly) water management and planning for electricity production in southern Sweden. The Spearman correlation coefficient was calculated between the principal component time series of electricity production (1999–2010) from 17 hydropower stations in 5 different rivers located in southern Sweden and 5 different teleconnection indices: the NAO, East Atlantic (EA), East Atlantic/Western Russia (EA/WR), Scandinavia (SCA), and Polar/Eurasia (POL). Results show that the impact of teleconnections on hydropower electricity production varies with season. In winter, the NAO shows the strongest positive relation with hydropower electricity production, explaining 24% of the variability in production, while the SCA has an almost as strong (15%) opposite effect. During spring and summer, the EA/WR alone shows a significant negative relation, explaining respectively 9 and 13% of the variance in production. No significant correlation appeared during the fall season.

Stochastic Population Analysis: A Functional Data Approach

Based on the Lee-Carter (LC) model, the benchmark in population forecasting, a variety of extensions and modifications are proposed in this paper. We investigate one of the extensions, the Hyndman-Ullah (HU) method and apply it to Asian demographic data sets: China, Japan and Taiwan. It combines ideas of functional principal component analysis (fPCA), nonparametric smoothing and time series analysis. Based on this stochastic approach, the demographic characteristics and trends in different Asian regions are calculated and compared. We illustrate that China and Japan exhibited a similar demographic trend in the past decade. We also compared the HU method with the LC model. The HU method can explain more variation of the demographic dynamics when we have data of high quality, however, it also encounters problems and performs similarly as the LC model when we deal with limited and scarce data sets, such as Chinese data sets due to the substandard quality of the data and the population policy.

Predicting Critical Transitions in ENSO models. Part II: Spatially Dependent Models

Abstract The present paper is the second part of a two-part study on empirical modeling and prediction of climate variability. This paper deals with spatially distributed data, as opposed to the univariate data of Part I. The choice of a basis for effective data compression becomes of the essence. In many applications, it is the set of spatial empirical orthogonal functions that provides the uncorrelated time series of principal components (PCs) used in the learning set. In this paper, the basis of the learning set is obtained instead by applying multichannel singular-spectrum analysis to climatic time series and using the leading spatiotemporal PCs to construct a reduced stochastic model. The effectiveness of this approach is illustrated by predicting the behavior of the Jin–Neelin–Ghil (JNG) hybrid seasonally forced coupled ocean–atmosphere model of El Niño–Southern Oscillation. The JNG model produces spatially distributed and weakly nonstationary time series to which the model reduction and prediction methodology is applied. Critical transitions in the hybrid periodically forced coupled model are successfully predicted on time scales that are substantially longer than the duration of the learning sample.

Baroclinic and Barotropic Annular Variability in the Northern Hemisphere

Abstract Large-scale variability in the Northern Hemisphere (NH) circulation can be viewed in the context of three primary types of structures: 1) teleconnection patterns, 2) a barotropic annular mode, and 3) a baroclinic annular mode. The barotropic annular mode corresponds to the northern annular mode (NAM) and has been examined extensively in previous research. Here the authors examine the spatial structure and time-dependent behavior of the NH baroclinic annular mode (NBAM). The NAM and NBAM have very different signatures in large-scale NH climate variability. The NAM emerges as the leading principal component (PC) time series of the zonal-mean kinetic energy. It dominates the variance in the wave fluxes of momentum, projects weakly onto the eddy kinetic energy and wave fluxes of heat, and can be modeled as Gaussian red noise with a time scale of ~10 days. In contrast, the NBAM emerges as the leading PC time series of the eddy kinetic energy. It is most clearly identified when the planetary-scale waves are filtered from the data, dominates the variance in the synoptic-scale eddy kinetic energy and wave fluxes of heat, and has a relatively weak signature in the zonal-mean kinetic energy and the wave fluxes of momentum. The NBAM is marked by weak but significant enhanced spectral power on time scales of ~20–25 days. The NBAM is remarkably similar to its Southern Hemisphere counterpart despite the pronounced interhemispheric differences in orography and land–sea contrasts.

Interdecadal Changes in Summertime Tropical Cyclone Precipitation over Southeast China during 1960–2009

Abstract This study examines the changes in tropical cyclone (TC) precipitation and the associated contributing factors over southeast China during 1960–2009. Climatologically, TC rainfall accounts for approximately 20%–40% of the total rainfall over southeast China during boreal summer, and the contribution can even reach 50% for some of the coastal provinces, such as Guangdong, Fujian, Zhejiang, and Hainan. The dominant mode of TC rainfall reveals a dipole pattern over southern southeast China (SSC) and eastern southeast China (ESC), and the associated principal component time series exhibits remarkable interdecadal variations, with two potential change points being identified in the late 1970s and early 1990s. These interdecadal shifts in TC rainfall are also found to be synchronous with two regime shifts in total rainfall, and they can account for more than 40% of the total rainfall anomalies over the coastal regions of southeast China. To discover the dominant factors responsible for the interdecadal variations, the overall TC rainfall anomalies are broken down into three different components (rainfall frequency, rainfall intensity, and nonlinear terms) based on a new empirical statistical approach. It is found that the interdecadal variation in TC precipitation over SSC is controlled predominantly by changes in TC rainfall intensity as well as TC rainfall frequency, while that over ESC depends mainly on the intensity and the nonlinear terms. Further examination of the TC passage frequency (TPF) suggests that the significant reduction in TPF and TC rainfall frequency over SSC during 1979–92 is associated mainly with suppressed TC genesis (negative genesis effect), while the increase in TPF and TC rainfall frequency during 1993–2009 can be attributed primarily to the enhanced passage probability (positive track effect) over SSC. Meanwhile, variations in TC rainfall intensity seem to be unrelated to the TC’s own intensity change.

Spatial and temporal variability of drought in the arid region of China and its relationships to teleconnection indices

Summary We studied the drought patterns in the arid region of northwestern China between 1960 and 2010 using the Palmer Drought Severity Index (PDSI). The general evolution of drought was obtained by empirical orthogonal function (EOF), rotated empirical orthogonal function (REOF), the Mann–Kendall test, and the continuous wavelet transform method. Additionally, relationships between rotated principal component time series (RPCs) and seven selected climate indices were analyzed. The results showed that: (1) Four moisture-related spatial patterns (North Xinjiang, western South Xinjiang, Central Xinjiang, and the Hexi Corridor) were objectively defined by REOF analysis. These patterns are related to distinct geographical areas and are associated with distinct temporal variations. (2) The PDSI increased significantly in most regions of Xinjiang, while decreased in the eastern Hexi Corridor. The significant 4–8 year band is the major period band for the annual and seasonal PDSI derived. (3) The seasonal REOFs (RPCs) and EOFs (PCs) have consistent spatial distribution patterns with the annual REOF. The seasonal trends of PDSI are also the same as the annual PDSI trends, indicating space–time consistency between annual PDSI and seasonal PDSI. (4) The drought evolution in this region is affected by the area of northern hemisphere polar vortex, the Arctic Oscillation, and the North Atlantic Oscillation. In addition, the changes of drought in South Xinjiang and the Hexi Corridor may also be associated with the Tibetan Plateau High. Changes in drought pattern are expected to have a strong impact on the economic livelihood of the region, especially for agricultural production.

Interannual to multidecadal Euro‐Atlantic blocking variability during winter and its relationship with extreme low temperatures in Europe

AbstractThe dominant modes of blocking frequency variability in the Atlantic‐European region are evaluated for the 1871–2010 period. An Empirical Orthogonal Function (EOF) analysis of a two‐dimensional blocking indicator field reveals three dominant EOFs, describing about 35% of interannual to multidecadal blocking variability. The first EOF captures an out‐of‐phase blocking frequency anomaly over Greenland and Western Europe regions. The corresponding principal component time series is strongly correlated with the North Atlantic Oscillation index but shows also significant correlations with indices of the East Atlantic, Scandinavian, and East Atlantic‐Western Russia patterns. The second EOF shows a dominant center over the North Sea region as well as a less pronounced center with anomalies of the same sign over southeastern Greenland. The multidecadal variations of this mode of blocking variability are related with a basin wide North Atlantic sea surface temperature anomaly which projects partly on the Atlantic Multidecadal Oscillation (AMO). The third mode is an east‐west dipole of blocking frequency anomalies from Scandinavian and southern Greenland regions and shows enhanced variability at ~20 year time scales. The coherent variations of the time coefficients of this pattern with open solar flux suggest a possible solar influence on blocking variability at these time scales. Furthermore, the dominant patterns of blocking variability are related with distinct anomaly patterns in the occurrence of extreme low temperature events over Europe at interannual to multidecadal time scales. AMO as well as the solar signals was detected also in the corresponding extreme low temperature blocking patterns. We argue that multivariate analysis of blocking indicators gives additional information about blocking and related extreme climate phenomena variability and predictability comparative with classical sectorial approach.

Temporal variability of observed and simulated hyperspectral reflectance

AbstractMultivariate analysis techniques were used to quantify and compare the spectral and temporal variability of observed and simulated shortwave hyperspectral Earth reflectance. The observed reflectances were measured by the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) instrument between 2002 and 2010. The simulated reflectances were calculated using climate Observing System Simulation Experiments (OSSEs), which used two Intergovernmental Panel on Climate Change AR4 scenarios (constant CO2 and A2 emission) to drive Moderate Resolution Atmospheric Transmission simulations. Principal component (PC) spectral shapes and time series exhibited evidence of physical variables including cloud reflectance, vegetation and desert albedo, and water vapor absorption. Comparing the temporal variability of the OSSE‐simulated and SCIAMACHY‐measured hyperspectral reflectance showed that their Intertropical Convergence Zone‐like Southern Hemisphere (SH) tropical PC1 ocean time series had a 90° phase difference. The observed and simulated PC intersection quantified their similarity and directly compared their temporal variability. The intersection showed that despite the similar spectral variability, the temporal variability of the dominant PCs differed as in, for example, the 90° phase difference between the SH tropical intersection PC1s. Principal component analysis of OSSE reflectance demonstrated that the spectral and centennial variability of the two cases differed. The A2 PC time series, unlike the constant CO2 time series, exhibited centennial secular trends. Singular spectrum analysis isolated the A2 secular trends. The A2 OSSE PC1 and PC4 secular trends matched those in aerosol optical depth and total column precipitable water, respectively. This illustrates that time series of hyperspectral reflectance may be used to identify and attribute secular climate trends with a sufficiently long measurement record and high instrument accuracy.

Potential for using Smartphone Accelerometers in Non-laboratory Environments

In recent years, there has been rising interest in using accelerometers as an alternative instrument to measure the center of pressure. Accelerometers are inexpensive, small, sensitive, and can be readily used in non-laboratory environments. In addition to this, smartphones with built-in accelerometers add to their capability with real-time processing of data from sensors. Clinicians and researchers are currently in disagreement from whether these measurements provide the same physiological information about the participant’s balance. In this study, twelve participants were asked to wear smartphone on their right ASIS using a belt clip and stand still on the forceplate. The data was synchronized using a tap by the right foot and collected over a fifty second period for analysis. Various linear and non-linear measures were extracted from the time series of resultant principal component (PC) scores. The results show high correlation in the COP time series from the two instruments (R=0.86).

On the Differences in the Intraseasonal Rainfall Variability between Western and Eastern Central Africa: Case of 10–25-Day Oscillations

In this paper, we analyze the space-time structures of the 10–25 day intraseasonal variability of rainfall over Central Africa (CA) using 1DD GPCP rainfall product for the period 1996–2009, with an emphasis on the comparison between the western Central Africa (WCA) and the eastern Central Africa (ECA) with different climate features. The results of Empirical Orthogonal Functions (EOFs) analysis have shown that the amount of variance explained by the leading EOFs is greater in ECA than WCA (40.6% and 48.1%, for WCA and ECA, resp.). For the two subregions, the power spectra of the principal components (PCs) peak around 15 days, indicating a biweekly signal. The lagged cross-correlations computed between WCA and ECA PCs time series showed that most of the WCA PCs lead ECA PCs time series with a time scale of 5–8 days. The variations of Intraseasonal Oscillations (ISO) activity are weak in WCA, when compared with ECA where the signal exhibits large annual and interannual variations. Globally, the correlation coefficients computed between ECA and WCA annual mean ISO power time series are weak, revealing that the processes driving the interannual modulation of ISO signal should be different in nature or magnitude in the two subregions.

1979~2011년, 북극진동지수 측면에서의 겨울철 남한지역 신적설과 최저 온도 특성

A characteristic of snowfall and minimum temperature variability in South Korea with respect to the variability of Arctic Oscillation (AO) was investigated. The climatic snowfall regions of South Korea based on daily new fresh snowfall data of 59 Korea Meteorological Administration (KMA) stations data corresponding to the sign of AO index during December to February 1979~2011 were classified. Especially, the differences between snowfalls of eastern regions and that of western regions in South Korea were seen by each mean 1000hPa geopotential height fields, which is one of physical structure, for the selected cases over the East Asia including the Korean Peninsula. Daily minimum temperature variability of 59 KMA station data and daily AO index during the same period were investigated using Cyclo-stationary empirical orthogonal function (CSEOF) analysis. The first CSEOF of wintertime daily AO index and that of minimum temperature of 59 KMA stations explain 33% and 66% of total variability, respectively. Correlation between principal component time series corresponding to the first CSEOF of AO index and that of temperature at the period of 1990s is over about -0.7 when that of AO index leads about 40 days.

Barotropic and Baroclinic Annular Variability in the Southern Hemisphere

Abstract The leading patterns of large-scale climate variability in the Southern Hemisphere are examined in the context of extratropical kinetic energy. It is argued that variability in the Southern Hemisphere extratropical flow can be viewed in the context of two distinct and largely independent structures, both of which exhibit a high degree of annularity: 1) a barotropic structure that dominates the variance in the zonal-mean kinetic energy and 2) a baroclinic structure that dominates the variance in the eddy kinetic energy. The former structure corresponds to the southern annular mode (SAM) and has been extensively examined in the literature. The latter structure emerges as the leading principal component time series of eddy kinetic energy and has received seemingly little attention in previous work. The two structures play very different roles in cycling energy through the extratropical troposphere. The SAM is associated primarily with variability in the meridional propagation of wave activity, has a surprisingly weak signature in the eddy fluxes of heat, and can be modeled as Gaussian red noise with an e-folding time scale of approximately 10 days. The baroclinic annular structure is associated primarily with variations in the amplitude of vertically propagating waves, has a very weak signature in the wave fluxes of momentum, and exhibits marked quasi periodicity on time scales of approximately 25–30 days. Implications for large-scale climate variability are discussed.