Accurate future flow projections necessitate hydrological models suitable for such a purpose. However, there are no formal guidelines on model selection. We here evaluate suitability of 29 commonly used bucket-type models for climate change impact studies across 50 high-latitude catchments. Model performance is represented by various indicators and the percentage of catchments with well-reproduced hydrological signatures. Models are grouped according to their structural attributes, and performance among the groups is compared to identify key features. Models generally reproduce well mean flows, 30-day maxima and runoff seasonality, but perform less reliably under low-flow conditions. Best-performing models feature balanced complexity with non-linear soil- and routing parameterisations and a limited number of parameters. Complexity of the snow routine does not noticeably influence overall model performance, as opposed to the efficiency in low flows. These findings contribute to good modelling practice by supporting systematic model selection for climate change impact studies in high latitudes. • Conceptual models can reproduce distributions of mean flows and runoff seasonality • oWell-performing models have non-linear soil- and routing routine parameterisations • oBest-performing models have balanced complexity and yield consistent performance • oMore complex snow routines can improve performance in low flows • oComplex interactions among the model routines hinder assessment of their effects
Todorović et al. (Wed,) studied this question.