Cross-correlation analysis is the standard tool for quantifying how the El Niño-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) influence Indonesian rainfall, and it routinely yields an "optimal time lag" offered as operational lead time. For Sumbawa Regency (West Nusa Tenggara) we show that this lag structure is an artefact. Using 44 years of NASA POWER precipitation (1982-2025; 528 months) with the SOI, Niño 3.4 and Dipole Mode Index from NOAA PSL, the conventional cross-correlation function returns significant coefficients at non-causal negative lags (rainfall leading the index), with the naive maximum for Niño 3.4 falling at a negative lag. After Box-Jenkins pre-whitening the structure collapses: only the concurrent SOI signal survives (r = 0.239), while the concurrent oceanic coefficients also fall to zero, so the reported "climate memory" is predictor persistence. Predictability exists at seasonal aggregation. Using June-August index means to forecast September-November (SON) rainfall across 43 years under leave-one-year-out cross-validation, a LASSO model attains an out-of-sample R2 of 0.405 and an RMSE skill score of +0.247 against climatology, and predicts the wet-season onset month with a skill score of +0.219. Skill is carried almost entirely by the JJA-mean SOI; the IOD contributes nothing detectable here, and a random forest does not outperform linear regression. As a categorical warning for below-normal SON rainfall the forecast reaches a critical success index of 0.53 against chance 0.20. Usable lead time therefore comes from seasonal aggregation, not monthly lag structure, and reported optimal lags in the regional literature should be re-examined.
Afiliasi berasal dari OpenAlex; metadata OAI-PMH jurnal tidak memuatnya. Penulis tanpa ID OpenAlex dicocokkan berdasarkan nama, sehingga dua orang bisa tergabung atau satu orang bisa terpisah.