Temporal Transferability of Satellite Rainfall Bias Correction Methods in a Data-Limited Tropical Basin

The Philippines experiences intense rainfall but has limited ground-based monitoring infrastructure for flood prediction. Satellite rainfall products provide broad coverage but contain systematic biases that reduce operational usefulness. This study evaluated three correction methods—Quantile Mapping (QM), Random Forest (RF), and Hybrid Ensemble—for improving Satellite Rainfall Monitor (SRM) estimates in the Cagayan de Oro River Basin, Northern Mindanao. When trained on comprehensive 2019-2020 data, Random Forest and Hybrid Ensemble substantially outperformed Quantile Map-ping, achieving excellent calibration accuracy (R² = 0.71 and 0.76 versus R² = 0.25 for QM). However, when tested on an independent year with substantially different rain-fall patterns (2021: 120% higher mean rainfall, 33% increase in rainy-day frequency), performance rankings reversed completely. Quantile Mapping maintained satisfactory operational performance (R² = 0.53, RMSE = 5.23 mm), showing improvement over training conditions, while Random Forest and Hybrid Ensemble both failed dramati-cally, with R² dropping to 0.46 and 0.41 respectively despite their excellent training performance. This highlights that training accuracy alone poorly predicts operational reliability under changing rainfall regimes. Quantile Mapping's percentile-based cor-rection naturally adapts when rainfall patterns shift without requiring recalibration, while machine learning methods learned magnitude-specific patterns that failed when conditions changed. For flood early warning in basins with limited data, equipment failures, and variable rainfall, only Quantile Mapping proved operationally reliable. This has practical implications for disaster risk reduction across the Philippines and similar tropical regions where standard validation approaches may systematically mislead model selection by measuring calibration performance rather than operational transferability.