Hydrological models are frequently used in flood forecasting, water resources assessments, and other environmental management projects. They are also useful tools for advancing scientific understanding of hydrological processes. In many applications to date, especially those using conceptual models, data uncertainty and model conceptualization are tacitly assumed to be the main sources of modeling error.
However, in a recent review of practical model robustness, Kavetski and Clark focus on errors arising from the computational technique used to approximate the time-dependent catchment dynamics and on how the lack of numerical error control can affect model behavior. They found that unless careful attention is paid to numerical calculations, troublesome artifacts arise and severely deform the response characteristics of hydrological models.