An alternative index to NINO3.4 for predicting monthly rainfall anomaly in East Java
Abstract ENSO and NINO3.4 index are known to have some relation with Indonesian monthly rainfall anomaly. There is a gap between scientific studies on one hand and forecasting operational problems on the other hand since previous studies are not giving enough attention to the N+1,2,3 concept. The concept is about giving three next month rainfall anomaly prediction rather than connecting ENSO index with three-monthly rainfall anomaly. Here we propose an alternative index for ENSO. The median of categorical gridded rainfall anomaly of East Java is used as a general representation. Plots of correlation between the median and anomaly sea surface temperature from ReynSmithOIv2 are used to determine locus candidate to be compared with NINO3.4. The Near Maritime Continent (NMC) index is selected and proven to have a significant average difference in correlation between based on bootstrap technique. Verification of prediction used in this study is simulation-based and only uses binary hit-miss final result. Prediction is generated by simple linear regression with three lag times (2,3, and 4). Verification based on three categories shows that NMC’s hits are higher than NINO.34 in lag-2 and lag-3. In lag-2, NMC’s verification is 57.5% compared to only 38.7% for NINO3.4. However, NINO3.4 is a still better predictor in lag-4. Radar charts of monthly verifications are also developed.