Exploring parameter (dis)agreement due to calibration metric selection in conceptual rainfall-runoff models

We examine the extent to which the parameters of different types of catchments are sensitive to calibration criteria selection (i.e. parameter agreement), and explore possible connections with overall model performance and model complexity. To this end, we calibrate the lumped GR4J, GR5J and GR6J hydrological models - coupled with the CemaNeige snow module - in 95 catchments spanning a myriad of hydroclimatic and physiographic characteristics across Chile, using 12 streamflow-oriented objective functions. The results show that (i) the choice of objective function has smaller effects on parameter values in catchments with low aridity index and high mean annual runoff ratio, in contrast to drier climates; and (ii) catchments with better parameter agreement also provide better performance across model structures and simulation periods. More generally, this work provides insights on the type of catchments where it is more challenging to find sub-domains in the parameter space that satisfy multiple streamflow criteria.