This paper explores the reliability of a hydrological modeling framework in a mesoscale (1515 km2) catchment of the dry Andes (30° S) where irrigation water use and snow sublimation represent a significant part of the annual water balance. To this end, a 20-year simulation period encompassing a wide range of climate and water-use conditions was selected to evaluate three types of integrated models referred to as A, B and C. These models share the same runoff generation and routing module but differ in their approach to snowmelt modeling and irrigation water use. Model A relies on a simple degree-day approach to estimate snowmelt rates and assumes that irrigation impacts can be neglected at the catchment scale. Model B ignores irrigation impacts just as Model A but uses an enhanced degree-day approach to account for the effects of net radiation and sublimation on melt rates. Model C relies on the same snowmelt routine as Model B but incorporates irrigation impacts on natural streamflow using a conceptual irrigation module. Overall, the reliability of probabilistic streamflow predictions was greatly improved with Model C, resulting in narrow uncertainty bands and reduced structural errors, notably during dry years. This model-based analysis also stressed the importance of considering sublimation in empirical snowmelt models used in the subtropics, and provided evidence that water abstractions from the unregulated river are impacting on the hydrological response of the system. This work also highlighted areas requiring additional research, including the need for a better conceptualization of runoff generation processes in the dry Andes. @en