[Previous Article] [Next Article]
SOI-Based Forecast of Australian Region Tropical Cyclone Activity
contributed by Neville Nicholls
Bureau of Meteorology Research Centre, Melbourne, Australia
The tropical cyclone season around northern Australia (105-165oE) extends from November to
May. The average number of cyclones per season is 9.4 (mean from 1949/50 season to 1995/96
season), with a standard deviation of 3.3. Cyclone activity in this region is related to the El
Niño/Southern Oscillation, with fewer than normal cyclones during El Niño episodes, and is
predictable from prior observations of simple indices of the El Niño/Southern Oscillation (Nicholls
1979, 1984, 1985, 1992; Solow and Nicholls 1990). Figure 1 shows time series of October
Southern Oscillation Index (SOI) and the number of tropical cyclones (TCs). The SOI values
were provided by the National Climate Centre, Bureau of Meteorology, Melbourne, and are the
standardized (mean = 0, standard deviation = 10) difference between Tahiti and Darwin pressures.
The correlation is +.47, significant at better than 1%.
Solow and Nicholls (1990) and Nicholls (1992) noted that the relationship between the SOI and
TCs is weakened by apparently artificial long term variations in TC activity (e.g., the gradual
increase in numbers up to the mid-1970s, as observation systems improved). Nicholls (1992)
suggested that correlating first differences (i.e., year0 minus year-1 differences) of the SOI and TC
numbers would reduce the influence of any artificial trends. The correlation is +.71, also
significant at better than 1%. (See N. Nicholls' contribution the December 1994 issue of this
Bulletin for plots of the time series and a scatterplot of the first differences). The linear regression,
with zero intercept, between the first differences is:
(TCs) = 0.22 (SOIOctober),
where (TCs) is the predicted difference in TC numbers from last season to the coming season, and
(SOIOctober) is the observed difference in SOI from October last year to the current year. Since the
number of TCs occurring in the previous season is known, this equation can be used to predict the
coming year's TC numbers. Nicholls (1992) demonstrated that such an approach can lead to
skillful forecasts, by calculating the regression between the first differences using 1959/60 to
1978/79 data and then applying it to 1979/80 to 1990/91 data. The RMS error of the hindcasts on
the independent data was 2.8. The RMS error of a persistence forecast for the same period was
4.8.
The change in October SOI from 1995 to 1996 was +4.6. The above equation therefore led to a
prediction that TC numbers in 1996/97 would be about 1.0 greater than in 1995/96, i.e., about 12,
somewhat above average. The actual number was 11, slightly above average.
The October 1997 SOI was -17.8, and the change from the October 1996 value was -22.0. The
above equation therefore leads to a prediction that TC numbers in 1997/98 should be about 5
fewer than in 1996/97, i.e., about 6, well below average.
Nicholls, N., 1979: A possible method for predicting seasonal tropical cyclone activity in the
Australian region. Mon. Wea. Rev., 107, 1221-1224.
Nicholls, N., 1984: The Southern Oscillation, sea-surface- temperature, and interannual
fluctuations in Australian tropical cyclone activity. J. Climatol., 4, 661-670.
Nicholls, N., 1985: Predictability of interannual variations of Australian seasonal tropical cyclone
activity. Mon. Wea. Rev., 113, 1144-1149.
Nicholls, N., 1992: Recent performance of a method for forecasting Australian seasonal tropical
cyclone activity. Aust. Meteorol. Mag., 40, 105-110.
Solow, A., and N. Nicholls, 1990: The relationship between the Southern Oscillation and tropical
cyclone frequency in the Australian region. J. Climate, 3, 1097-1101.
Fig. 1. Time series of October SOI (left scale) and the number of tropical cyclones around
northern Australia (right scale) in the ensuing cyclone season (November-May). The correlation
between the two series is 0.47.