How much can we gain with increasing model complexity with the same model concepts?

Abstract

The main purpose of this study is to test the hypothesis that, with appropriate structures, increasing model complexity with the same model concepts would lead to an increase in the model efficiency in simulating either runoff or internal variables. Five variants of the Hydrologiska Byråns Vattenbalansavdelning (HBV) model, i.e. a lumped model (LWhole), a semi-distributed model (SBand), a grid-model without routing (GRZero), a grid-model with hillslope routing (GROne), and a grid-model with both hillslope and channel routing (GRTwo) are compared in a cold and mountainous catchment in central southern Norway. The five models are compared with respects to (1) runoff simulation at the catchment outlet and the interior points, and (2) simulations of internal variables, i.e. evapotranspiration, snow water equivalent and groundwater depth. The results show that the models with higher complexity can improve the runoff simulation both at the catchment outlet and the interior points. However, there is no superiority of complex grid-models over simple grid-models in reproducing internal variables in this study.