[Previous Article] [Next Article]
Forecasts of Niño 3 SST Anomalies and SOI Based on Singular
Spectrum Analysis Combined with the Maximum Entropy Method
Amira Saunders, Michael Ghil and J. David Neelin
Department of Atmospheric Sciences and Institute of Geophysics and Planetary Physics
University of California, Los Angeles, California
Singular spectrum analysis (SSA: Vautard and Ghil 1989) and the maximum entropy method
(MEM: 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 47 years of observed data. This forecast
follows up on earlier forecasts using combined SSA-MEM methodology by C. Keppenne and M.
Ghil for the SOI index, starting in the March 1992 issue of this Bulletin, and on those of N. Jiang,
M. Ghil and J.D. Neelin for Niño-3 SST anomalies, starting with the March 1995 issue. 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: Keppenne and Ghil 1993)
combined with MEM is documented in the March 1995 issue of this Bulletin (see also 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.
The skill in this method is based on the low-frequency, oscillatory components of the system, so
forecasts tend to filter out the short-lived from longer-term anomalies.
Figure 1 shows area-averaged Niño 3 SSTAs, forecast and observed, since 1990, using the SSA-
and MSSA-MEM schemes for 6- and 12-month lead. The latest forecasts, for the next 1-4
seasons, using data through November 1997, are shown in Fig. 2. The vertical bars are one
standard deviation in length, based on forecast verification over the 1984-93 time span. The
current forecast indicates a slow decline in the warm anomalies throughout 1998, with significant
warmth persisting through the summer and fall. A caveat on this forecast is that the model has
been consistently predicting a decline over the last few months' forecasts.
Figure 3 shows the SSA-MEM forecast for the SOI from December 1997 through November
1998. The SOI is expected to start rising slowly in the spring of 1998 but to still remain well
below its mean through the next year. The anticorrelation between the present SOI and Niño-3
SSTA indicates higher predictability (Ghil and Jiang 1997), thus enhancing our confidence in the
forecast.
Ghil, M. And N. Jiang, 1997: Recent forecast skill for the El Niño/Southern Oscillation. Geophys.
Res. Lett., 102, in press.
Jiang, N., D. Neelin and M. Ghil, 1995: Quasi-quadrennial and quasi-biennial variability in the
equatorial Pacific. Clim. Dyn., 12, 101-112.
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: An application to subannual variability in atmospheric angular momentum. Intl. J. Bif. &
Chaos, 3, 625-634.
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.
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.
Fig. 1. Forecasts of the area-averaged Niño-3 SST anomalies (SSTA) using the SSA-MEM (star)
and MSSA-MEM (open circle) schemes. The solid line indicates the observed Niño-3 SSTA.
The latest forecast starts from December 1997. Shown for leads of (a) 3-months, (b) 6-months,
(c) 9-months, and (d) 12-months.
Fig. 2. The forecasts of the Niño-3 SSTA for the upcoming 4 seasons using the SSA-MEM
scheme. The solid line indicates the observed Niño-3 SSTA through November 1997.
Fig. 3. SSA-MEM forecast of the SOI for December 1997 through November 1998. The circles
are the monthly SOI values based on a 5-month running mean without the seasonal cycle and the
solid line is the SSA-filtered SOI. The dashed line indicates the forecast for the next 12 months.