HESS Opinions "Crash tests for a standardized evaluation of hydrological models"

As all hydrological models are intrinsically limited hypotheses on the behaviour of catchments, models – which attempt to represent real-world behaviours – will always remain perfectible. To progress on the long road towards improved models, we need demanding tests, i.e. true crash tests. Efficient development and assessment of hydrological models requires demanding tests on large and varied data sets to ensure models' generality, to diagnose their failures and limitations, and ultimately, to improve them.