Identifying relevant hydrological and catchment properties in active subspaces: An inference study of a lumped karst aquifer model

The use of the active subspace method was recently proposed to reduce the dimension of complex hydrological models, perform sensitivity analysis of model parameters and quantify the uncertainty affecting model parameters. Although this inversion method is highly promising in terms of computational performance, a clear hydrological interpretation of the meaning of the active subspace that it identifies is missing. In this work, we infer how the active subspace changes in dimension and feature depending on geometrical and hydrological properties of the karst aquifer model LuKARS. We find that both the hydrotope area coverage and model parameters describing the catchment characteristics (here: water storage and discharge properties of the hydrotopes) have major impacts on the active subspace. Our results show that the active subspace method can be used to investigate the relation between the model structure, the area of a hydrotope and the simulated spring discharge.