Model prediksi awal musim hujan di Pulau Jawa dengan menggunakan informasi suhu muka laut di kawasan Pasifik dan India
Abstract
The rainfall variability is one of an important role to play onset of wet season. This research aims to analyze that SST (Sea Surface Temperature) exerting an influence on rainfall onset over Java. The significant scale (15°S – 15°N; 80°E- 100°W) of SST patterns are observed to derive a potential predictor rainfall onset over Java. The onset wet season is define as the first 10-day that receive at least 50 mm followed by 2 consecutive 10-days. In Java generally it starts from 1 September. In this study, The regression Multivariate method was offered to develop a rainfall onset model prediction based on SST circulation patern. The principal component and regression technique are improved to achieve a better models and cross validation technique were applied to validate models. Domain predictor of models were derived using spatial correlation analysis result between Rainfall onset data each ZOM over Java and SST anomaly from Indian Ocean to Pacific Ocean during 1978-2007 . There are 28 from 30 ZOM which have potential predictors that be trained to models. The high spatial coherence of the rainfall onset is SST pattern at Juni-Juli-August that represent SST on eastern equatorial Pacific as characteristic of El Nino Southern Oscillation (ENSO), SST on Western Sumatera Island as characteristic of Indian Ocean Dipole Mode (IOD) and SST on around Java region as Characteristic of local condition. Three representative correlation patterns of Java Rainfall onset with SST Indian-Pacific area are investigated that clarify of ENSO, IOD and local condition. Verification result for those models showed promising skill prediction during test in year of 2008 with standart deviation 10 days. In Java generally, skill prediction when rainfall onset advanced exhibited better result than skill prediction when it delayed. The Indonesian ocean area that has local characteristic have influenced better rainfall onset over Java than others. It indicated that models which developed using predictor of Indonesian SST have better skill prediction than SST of Indian or pacific ocean area. However, predictor which has combination domain area of Indonesian SST and Pacific have main influenced rainfall onset over Java. It can be explained that 18 ZOM’s model have been using predictor from those domain.