A data-driven gap-filling approach to assess long-term streamflow trends across Colombia

Water availability underpins Colombia’s economic and environmental development, given reliance on agriculture and hydropower. Historical streamflow trends remain uncertain, and no study has yet combined gridded precipitation data and machine learning to model rainfall–streamflow processes across the different hydrological regions of Colombia. We assess trends in monthly mean streamflow and relate them to trends in precipitation and maximum and minimum flows covering 1981–2022. Gaps in streamflow records are filled using satellite-derived predictors via random forest and multilayer perceptron models. Performance is complementary and conditioned by basin size, precipitation regime, and variability, with best skill in large, stable basins. Regionally, increasing streamflow trends dominate the Orinoco, Caribbean, and Pacific, whereas the Magdalena–Cauca and Amazon show complex patterns. These insights provide a foundation for future studies on the drivers of streamflow change and can inform adaptive water resource management and policy decisions to enhance Colombia’s resilience to hydroclimatic variability.