抄録
Accurate rainfall predictions, especially for tropical monsoonal rainfall, are among the most difficult tasks in hydrology. In this article, we discuss ANN-based long-term rainfall predictions for Oekabiti, West Timor, Indonesia. Due to the remoteness of the area, the only information available is historical rainfall data. However, by basing on previous studies into the time lag relationship between the El Nino Southern Oscillation (ENSO) and Indonesian rainfall, we utilized Sea Surface Temperature Anomaly (SSTA) Zone Nino 3.4 and the Southern Oscillation Index (SOI) to improve accuracy levels of rainfall prediction models efficiently. During our model development, it was revealed that rainfall fluctuation is more influenced by the lag 0 to lag -1 of the SOI than by the SSTA. We also found that the resulting models could dynamically predict long-term rainfall, but tended to underestimate some extreme values, which limited their utility for irrigation management planning.
