In a tropical montane catchment of the Colombian Andes (Abanico del Quindío), we analysed a seven-year (2018–2025) monthly record of precipitation isotopes (δ 2H and deuterium excess, d-excess) using a novel combination of Seasonal–Trend decomposition (STL) and generalized additive models (GAMs). This approach separates intra-annual and interannual variability and quantifies their climatic drivers. Results reveal a pronounced seasonal δ 2H cycle (~80–100‰ amplitude) driven by a strong amount effect: rainy months yield depleted δ 2H, whereas drier periods see enriched values. δ 2H and d-excess co-vary seasonally, but their interannual responses diverge under ENSO influences: δ 2H is enriched during El Niño conditions (positive MEI.v2) and more depleted in La Niña, whereas d-excess shows the opposite pattern, rising during La Niña (consistent with enhanced continental moisture recycling) and dropping in El Niño. After 2019, coincident with a prolonged “triple-dip” La Niña, the seasonal δ 2H peak shifted ~1 month earlier and its amplitude dampened (~20% reduction), reflecting modified rainfall seasonality. GAM results further indicate that precipitation amount and Atlantic SST anomalies (TNA index) significantly predict δ 2H variability (p < 0.001), while ENSO (MEI.v2) adds no independent explanatory power after controlling for these factors. For d-excess, the GAM yields significant positive effects of TNA and precipitation and a negative effect of MEI.v2. The best-fit model explains 86% of δ 2H variance and 62% of d-excess variance, underscoring the strength of this integrated STL–GAM approach for unravelling multi-scale climate controls on precipitation isotopes in tropical mountains.