IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis

3.6 Patterns of Atmospheric Circulation Variability

3.6.1 Teleconnections

The global atmospheric circulation has a number of preferred patterns of variability, all of which have expressions in surface climate. Box 3.4 discusses the main patterns and associated indices. Regional climates in different locations may vary out of phase, owing to the action of such ‘teleconnections’, which modulate the location and strength of the storm tracks (Section 3.5.3) and poleward fluxes of heat, moisture and momentum. A comprehensive review by Hurrell et al. (2003) has been updated by new analyses, notably from Quadrelli and Wallace (2004) and Trenberth et al. (2005b). Understanding the nature of teleconnections and changes in their behaviour is central to understanding regional climate variability and change. Such seasonal and longer time-scale anomalies have direct impacts on humans, as they are often associated with droughts, floods, heat waves and cold waves and other changes that can severely disrupt agriculture, water supply and fisheries, and can modulate air quality, fire risk, energy demand and supply and human health.

The analysis of teleconnections has typically employed a linear perspective, which assumes a basic spatial pattern with varying amplitude and mirror image positive and negative polarities (Hurrell et al., 2003; Quadrelli and Wallace, 2004). In contrast, nonlinear interpretations would identify preferred climate anomalies as recurrent states of a specific polarity (e.g., Corti et al., 1999; Cassou and Terray, 2001; Monahan et al., 2001). Climate change may result through changes from one quasi-stationary state to another, as a preference for one polarity of a pattern (Palmer, 1999), or through a change in the nature or number of states (Straus and Molteni, 2004).

In the NH, one-point correlation maps illustrate the Pacific-North American (PNA) pattern and the NAO (Figure 3.26), but in the SH, wave structures do not emerge as readily owing to the dominance of the SAM. Although teleconnections are best defined over a grid, simple indices based on a few key station locations remain attractive as the series can often be carried back in time long before complete gridded fields were available (see Section 3.6.4, Figure 3.31); the disadvantage is increased noise from the reduced spatial sampling. For instance, Hurrell et al. (2003) found that the residence time of the NAO in its positive phase in the early 20th century was not as great as indicated by the positive NAO index for that period.


Figure 3.26. The PNA (left) and NAO (right) teleconnection patterns, shown as one-point correlation maps of 500 hPa geopotential heights for boreal winter (DJF) over 1958 to 2005. In the left panel, the reference point is 45°N, 165°W, corresponding to the primary centre of action of the PNA pattern, given by the + sign. In the right panel, the NAO pattern is illustrated based on a reference point of 65°N, 30°W. Negative correlation coefficients are dashed, and the contour increment is 0.2. Adapted from Hurrell et al. (2003).

Many teleconnections have been identified, but combinations of only a small number of patterns can account for much of the interannual variability in the circulation and surface climate. Quadrelli and Wallace (2004) found that many patterns of NH interannual variability can be reconstructed as linear combinations of the first two Empirical Orthogonal Functions (EOFs) of sea level pressure (approximately the NAM and the PNA). Trenberth et al. (2005b) analysed global atmospheric mass and found four key rotated EOF patterns: the two annular modes (SAM and NAM), a global ENSO-related pattern and a fourth closely related to the North Pacific Index and the PDO, which in turn is closely related to ENSO and the PNA pattern.

Teleconnection patterns tend to be most prominent in the winter (especially in the NH), when the mean circulation is strongest. The strength of teleconnections and the way they influence surface climate also vary over long time scales. Both the NAO and ENSO exhibited marked changes in their surface climate expressions on multi-decadal time scales during the 20th century (e.g., Power et al., 1999b; Jones et al., 2003). Multi-decadal changes in influence are often real and not due just to poorer data quality in earlier decades.

Box 3.4: Defining the Circulation Indices

A teleconnection is made up of a fixed spatial pattern with an associated index time series showing the evolution of its amplitude and phase. Teleconnections are best defined by values over a grid but it is often convenient to devise simplified indices based on key station values. A classic example is the Southern Oscillation (SO), encompassing the entire tropical Pacific, yet encapsulated by a simple SO Index (SOI), based on differences between Tahiti (eastern Pacific) and Darwin (western Pacific) MSLP anomalies.

A number of teleconnections have historically been defined from either station data (SOI, NAO) or from gridded fields (NAM, SAM, PDO/NPI and PNA):

  • Southern Oscillation Index (SOI). The MSLP anomaly difference of Tahiti minus Darwin, normalised by the long-term mean and standard deviation of the MSLP difference (Troup, 1965; Können et al., 1998). Available from the 1860s. Darwin can be used alone, as its data are more consistent than Tahiti prior to 1935.
  • North Atlantic Oscillation (NAO) Index. The difference of normalised MSLP anomalies between Lisbon, Portugal and Stykkisholmur, Iceland has become the most widely used NAO index and extends back in time to 1864 (Hurrell, 1995), and to 1821 if Reykjavik is used instead of Stykkisholmur and Gibraltar instead of Lisbon (Jones et al., 1997).
  • Northern Annular Mode (NAM) Index. The amplitude of the pattern defined by the leading empirical orthogonal function of winter monthly mean NH MSLP anomalies poleward of 20°N (Thompson and Wallace, 1998, 2000). The NAM has also been known as the Arctic Oscillation (AO), and is closely related to the NAO.
  • Southern Annular Mode (SAM) Index. The difference in average MSLP between SH middle and high latitudes (usually 45°S and 65°S), from gridded or station data (Gong and Wang, 1999; Marshall, 2003), or the amplitude of the leading empirical orthogonal function of monthly mean SH 850 hPa height poleward of 20°S (Thompson and Wallace, 2000). Formerly known as the Antarctic Oscillation (AAO) or High Latitude Mode (HLM).
  • Pacific-North American pattern (PNA) Index. The mean of normalised 500 hPa height anomalies at 20°N, 160°W and 55°N, 115°W minus those at 45°N, 165°W and 30°N, 85°W (Wallace and Gutzler, 1981).
  • Pacific Decadal Oscillation (PDO) Index and North Pacific Index (NPI). The NPI is the average MSLP anomaly in the Aleutian Low over the Gulf of Alaska (30°N–65°N, 160°E–140°W; Trenberth and Hurrell, 1994) and is an index of the PDO, which is also defined as the pattern and time series of the first empirical orthogonal function of SST over the North Pacific north of 20°N (Mantua et al., 1997; Deser et al., 2004). The PDO broadened to cover the whole Pacific Basin is known as the Inter-decadal Pacific Oscillation (IPO: Power et al., 1999b). The PDO and IPO exhibit virtually identical temporal evolution (Folland et al., 2002).