Refining Precipitation Estimates: CHIRPS v3.0 Improvements in Caribbean Colombian

Accurate precipitation information is essential for hydrological studies, hazard monitoring, and water resource management. This study assesses the performance of CHIRPS v2.0 and CHIRPS v3.0 against observations from 505 quality-controlled rain gauges for the period 1981–2024, aiming to identify systematic biases, quantify improvements across temporal scales, and evaluate the applicability of CHIRPS v3.0. Daily, 3-day, 5-day, and monthly accumulated precipitation were analyzed using a comprehensive set of statistical performance metrics. Differences between CHIRPS versions were assessed using paired t-tests and Wilcoxon signed-rank tests for statistical significance, Cohen’s d for effect size, and paired bootstrap confidence intervals for uncertainty and robustness. For 3-day and 5-day aggregations, both versions saw reduced data dispersion, but outliers persisted in mountain regions. CHIRPS v3.0 detected rainfall events more frequently across all subregions, though CHIRPS v2.0 had lower average error indices at shorter time scales. At the monthly scale, both versions showed the highest agreement with rain gauge observations. A total of 59% of all rain gauges had KGE scores above 0.7 for CHIRPS v3.0, compared to 38% for CHIRPS v2.0. Paired tests indicated significant differences in several metrics, while Cohen’s d showed meaningful improvements for correlation ( |d| > 0.60 ) and Probability of Detection ( |d| > 0.85 ) across subregions and time scales. Overall, CHIRPS v3.0 improves rainfall occurrence detection and monthly precipitation representation, supporting its use for regional hydroclimatic monitoring, although local validation remains necessary, especially for short-scale error behavior in complex coastal and mountainous environments.