LONG-RANGE PREDICTION OF HAWAIIAN WINTER RAINFALL USING CANONICAL CORRELATION-ANALYSIS

Hawaiian rainfall is teleconnected to short-term climate variability i n the Pacific Ocean. The summer Southern Oscillation Index (SOI) and s ummer sea-level pressure (SLP) over the North Pacific are used as pred ictors, and the following winter rainfall indices from three islands o f Hawaii are used as predictands. To consolidate the large data array of the SLP field prior to prediction experiments, lagged correlation a nd empirical orthogonal function analyses are used. Canonical correlat ion analysis has been used for predicting Hawaiian winter rainfall. Am ong many schemes tested, predictions. Cross-validation techniques have also been used to estimate the 'overall' forecast skill of various sc hemes, and the results are consistent with those from prediction exper iments. Winter rainfall in Hawaii can be predicted with a good degree of success two seasons in advance using the summer SOI and the first f our eigenmodes of summer SLP over the North Pacific as predictors.