Predictability of Java Monsoon Rainfall Anomalies: A Case Study

Abstract

A substantial portion of the interannual variability of rainfall at Jakarta, Java, can be predicted from antecedent pressure anomalies at Darwin, northern Australia; the pressure persistence, the concurrent correlation of pressure and rainfall, and the predictability of rainfall from antecedent pressure are all largest during the ?east? monsoon (June-November). Because of the relatively simple large-scale circulation setting, warranting a single predictor (Darwin pressure), this region is chosen for a series of experiments aimed at exploring the seasonality and secular variations of predictability, optimal length of dependent record, and updating of the regression base period used for predictions on the independent data set. The major features of pressure-rainfall relationships are common through much of the 1911?83 record, namely sign and general magnitude of correlations and the closer relationships during the east, as compared to the west monsoon. Considerable differences are, however, apparent between decades. Them may stem from both sampling deficiencies (noise) and real long-term changes of the pressure-rainfall couplings due to secular alterations in the large-scale circulation setting. The competition between these two factors is relevant concerning the optimal length of the dependent record used for predictions into the independent data set, as well as the updating of the regression base period.