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

8.4.7 El Niño-Southern Oscillation

During the last decade, there has been steady progress in simulating and predicting ENSO (see Chapters 3 and 9) and the related global variability using AOGCMs (Latif et al., 2001; Davey et al., 2002; AchutaRao and Sperber, 2002). Over the last several years the parametrized physics have become more comprehensive (Gregory et al., 2000; Collins et al., 2001; Kiehl and Gent, 2004), the horizontal and vertical resolutions, particularly in the atmospheric component models, have markedly increased (Guilyardi et al., 2004) and the application of observations in initialising forecasts has become more sophisticated (Alves et al., 2004). These improvements in model formulation have led to a better representation of the spatial pattern of the SST anomalies in the eastern Pacific (AchutaRao and Sperber, 2006). In fact, as an indication of recent model improvements, some IPCC class models are being used for ENSO prediction (Wittenberg et al., 2006). Despite this progress, serious systematic errors in both the simulated mean climate and the natural variability persist. For example, the so-called ‘double ITCZ’ problem noted by Mechoso et al. (1995; see Section 8.3.1) remains a major source of error in simulating the annual cycle in the tropics in most AOGCMs, which ultimately affects the fidelity of the simulated ENSO. Along the equator in the Pacific the models fail to adequately capture the zonal SST gradient, the equatorial cold tongue structure is equatorially confined and extends too far too to the west (Cai et al., 2003), and the simulations typically have thermoclines that are far too diffuse (Davey et al., 2002). Most AOGCMs fail to capture the meridional extent of the anomalies in the eastern Pacific and tend to produce anomalies that extend too far into the western tropical Pacific. Most, but not all, AOGCMs produce ENSO variability that occurs on time scales considerably faster than observed (AchutaRao and Sperber, 2002), although there has been some notable progress in this regard over the last decade (AchutaRao and Sperber, 2006) in that more models are consistent with the observed time scale for ENSO (see Figure 8.13). The models also have difficulty capturing the correct phase locking between the annual cycle and ENSO. Further, some AOGCMs fail to represent the spatial and temporal structure of the El Niño-La Niña asymmetry (Monahan and Dai, 2004). Other weaknesses in the simulated amplitude and structure of ENSO variability are discussed in Davey et al. (2002) and van Oldenborgh et al. (2005).

Figure 8.13

Figure 8.13. Maximum entropy power spectra of surface air temperature averaged over the NINO3 region (i.e., 5°N to 5°S, 150°W to 90° W) for (a) the MMD at the PCMDI and (b) the CMIP2 models. Note the differing scales on the vertical axes and that ECMWF reanalysis in (b) refers to the European Centre for Medium Range Weather Forecasts (ECMWF) 15-year reanalysis (ERA15) as in (a). The vertical lines correspond to periods of two and seven years. The power spectra from the reanalyses and for SST from the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) version 1.1 data set are given by the series of solid, dashed and dotted black curves. Adapted from AchutaRao and Sperber (2006).

Current research points to some promise in addressing some of the above problems. For example, increasing the atmospheric resolution in both the horizontal (Guilyardi et al., 2004) and vertical (NCEP Coupled Forecast System) may improve the simulated spectral characteristics of the variability, ocean parametrized physics have also been shown to significantly influence the coupled variability (Meehl et al., 2001) and continued methodical numerical experimentation into the sources of model error (e.g., Schneider, 2001) will ultimately suggest model improvement strategies.

In terms of ENSO prediction, the two biggest recent advances are: (i) the recognition that forecasts must include quantitative information regarding uncertainty (i.e., probabilistic prediction) and that verification must include skill measures for probability forecasts (Kirtman, 2003); and (ii) that a multi-model ensemble strategy may be the best current approach for adequately dealing with forecast uncertainty, for example, Palmer et al. (2004), in which Figure 2 demonstrates that a multi-model ensemble forecast has better skill than a comparable ensemble based on a single model. Improvements in the use of data, particularly in the ocean, for initialising forecasts continues to yield enhancements in forecast skill (Alves et al., 2004); moreover, other research indicates that forecast initialisation strategies that are implemented within the framework of the coupled system as opposed to the individual component models may also lead to substantial improvements in skill (Chen et al., 1995). However, basic questions regarding the predictability of SST in the tropical Pacific remain open challenges in the forecast community. For instance, it is unclear how westerly wind bursts, intra-seasonal variability or atmospheric weather noise in general limit the predictability of ENSO (e.g., Thompson and Battisti, 2001; Kleeman et al., 2003; Flugel et al., 2004; Kirtman et al., 2005). There are also apparent decadal variations in ENSO forecast skill (Balmaseda et al., 1995; Ji et al., 1996; Kirtman and Schopf, 1998), and the sources of these variations are the subject of some debate. Finally, it remains unclear how changes in the mean climate will ultimately affect ENSO predictability (Collins et al., 2002).