Abstract
The availability of large samples of catchment data such as the CAMELS
data set for the US and other countries has been a game changer in
hydrological modelling. Model applications to hundreds of catchments and
comparative analyses of model simulations provides new insights into
hydrological catchment modelling. The comparative model application to
many catchments allows to better assess the performance of different
model variants. In general, our results show that bucket-type models can
be surprisingly good in terms of runoff simulations even when there is
no calibration. Furthermore, we found that even limited runoff
observations can be informative for modelling and that regionalisation
remains challenging. "