Forecasts of Equatorial Pacific SST Anomalies Based on Singular Spectrum Analysis Combined with the Maximum Entropy Method

Singular spectrum analysis (SSA: Vautard and Ghil 1989; Ghil and Vautard 1991; Plaut et al. 1995) and the maximum entropy method (MEM: Burg 1968; Penland et al. 1991) are used here for long-lead forecasts of the sea-surface temperature (SST) anomalies averaged over the Niño 3 area and the Southern Oscillation Index (SOI). The forecast is for up to one year ahead, based on the last 45 years of observed data. More detailed information on the forecast method based on single-channel SSA combined with MEM is given by Keppenne and Ghil (1992), while multi-channel SSA (M-SSA: Kimoto et al. 1991; Keppenne and Ghil 1993; Plaut and Vautard 1994) combined with MEM is documented in the March 1995 issue of this Bulletin (Jiang et al. 1995). Briefly, the time series is filtered first by SSA (if univariate) or M-SSA (if multivariate), so that the statistically significant components are retained, specifically the quasi- quadrennial (QQ) and the quasi-biennial (QB) components of ENSO variability (Rasmusson et al. 1990; Keppenne and Ghil 1992; Jiang et al. 1995). Then MEM is applied to advance these components in time.

Figure 1 shows area-averaged Niño 3 SSTAs, forecast and observed, since 1990, using the SSA and MSSA-MEM schemes for a 3-, 6-, 9- and 12-month lead. The last forecast, for the next 1-4 seasons, using data through January 1996, is shown in Fig. 2. The vertical bars are one standard deviation in length, based on forecast verification over the 1984-93 time span. The forecasts indicate that the presently slightly cooler than normal or near-normal conditions in Niño 3 may continue through 1996. Figure 3 shows the SSA-MEM forecast for the SOI from February 1996 through January 1997. The SOI is expected to remain close to its mean over this year. The present SOI forecast agrees with the Niño 3 SSTA forecast, although the anomalously weak (anti)correlation between these two ENSO signals, which we pointed out in the September 1995 issue of this Bulletin, continues.

Burg, J.P., 1968: Maximum entropy spectral analysis. Modern Spectrum Analysis, 34-48. IEEE Press.

Ghil, M. and R. Vautard, 1991: Interdecadal oscillations and the warming trend in global temperature time series. Nature, 350, 324-327.

Jiang, N., D. Neelin and M. Ghil, 1995: Quasi-quadrennial and quasi-biennial variability in the equatorial Pacific. Clim. Dyn., 12, 101-112.

Jianag, N., M. Ghil and D. Neelin, 1995: Forecasts of Equatorial Pacific SST anomalies using an autoregressive process using singular spectrum analysis. Experimental Long-Lead Forecast Bulletin, 4, 1, 24-27.

Keppenne, C.L. and M. Ghil, 1992: Adaptive filtering and prediction of the Southern Oscillation Index. J. Geophys. Res., 97: 20449-20454.

Keppenne, C.L. and M. Ghil, 1993: Adaptive filtering and prediction of noisy multivariate signals Adaptive filtering and prediction of noisy multivariate signals: An application to subannual variability in atmospheric angular momentum. Intl. J. Bif. & Chaos, 3, 625-634.

Kimoto, M., M. Ghil and K.C. Mo, 1991: Spatial structure of the 40-day oscillation in the Northern Hemisphere extratropics. Proc. 8th Conf. Atmos. & Oceanic Waves & Stability. Amer. Met. Soc., Boston, 115-116.

Penland, C., M. Ghil and K.M. Weickmann, 1991: Adaptive filtering and maximum entropy spectra, with application to changes in atmospheric angular momentum. J. Geophys. Res., 96, 22,659-22,671.

Plaut, G.R., M. Ghil and R. Vautard, 1995: Interannual and interdecadal variability in 335 years of central England temperature. Science, 268, 710-713.

Rasmusson E.M., X. Wang and C. F. Ropelewski, 1990: The biennial component of ENSO variability. J. Mar. Sys., 1, 71-96.

Vautard R., and M. Ghil, 1989: Singular spectrum analysis in nonlinear dynamics with applications to paleoclimatic time series. Physica D, 35, 395-424.

Figures

Figure 1. Forecasts of the area-averaged Ni¤o 3 SST anomalies (SSTAs) by the SSA-MEM (star) and MSSA-MEM (circle) schemes. The solid line indicates the observed Ni¤o 3 SSTAs. The latest forecast starts from January 1996, shown for lead times of (a) 3 months, (b) 6 months, (c) 9 months and (d) 12 months.

Figure 2. SSA-MEM forecasts of Nino-3 SSTAs for the upcoming 4 seasons, with a January 1996 start time. The solid line indicates the observed Nino-3 SSTAs.

Figure 3. SSA-MEM forecast for the SOI for February 1996 through January 1997. The circles represent the raw monthly SOI values without the seasonal cycle, and the solid line the SSA-filtered SOI. The dashed line indicates the forecast for the coming 12 months.