A seasonal rainfall forecasting scheme for Fiji is currently operational, t
aking advantage of strong relationships between the Southern Oscillation In
dex (SOI) and rainfall in Fiji. In this scheme, the three-month mean of the
SOI is used to forecast the immediately following three-month total rainfa
ll (e.g. July-September SOI predicting October-December rainfall). In the p
resent study, exploratory data analysis is undertaken using correlation coe
fficients to determine whether this forecast lead-time can be increased. St
rong correlations are demonstrated out to four months in advance for some p
eriods of the year. Correlation coefficients between equatorial sea-surface
temperatures and Fiji rainfall are found to be higher in many periods of t
he year than those between SOI and rainfall.
The hindcast skill of a seasonal prediction scheme was then evaluated for l
ead-times longer than the currently operational scheme. Strong skill was sh
own several months in advance for a few periods of the year, for both predi
ctions based on the SOI and on equatorial sea-surface temperatures. This su
ggests that the current operational scheme could usefully be extended out t
o longer lead-times. For forecasts of the wet season rainfall as a whole, s
ubstantial skill was shown in one region of Fiji for predictions made as mu
ch as four months in advance.