Nans Addor, Grey Stephen Nearing, Cristina Prieto, Andrew J. Newman, Nataliya Le Vine, Martyn P Clark

Hydrological signatures are now used for a wide range of purposes, including catchment classification, process exploration and hydrological model calibration. The recent boost in the popularity and number of signatures has however not been accompanied by the development of clear guidance on signature selection, meaning that signature selection is often arbitrary. Here we use three complementary approaches to compare and rank 15 commonly-used signatures, which we evaluate in 671 US catchments from the CAMELS data set (Catchment Attributes and MEteorology for Large-sample Studies). Firstly, we employ machine learning (random forests) to explore how attributes characterizing the climatic conditions, topography, land cover, soil and geology influence (or not) the signatures. Secondly, we use a conceptual hydrological model (Sacramento) to critically assess which signatures are well captured by the simulations. Thirdly, we take advantage of the large sample of CAMELS catchments to characterize the spatial smoothness (using Morans I) of the signature field. These three approaches lead to remarkably similar rankings of the signatures. We show that signatures with the noisiest spatial pattern tend to be poorly captured by hydrological simulations, that their relationship to catchments attributes are elusive (in particular they are not correlated to climatic indices like aridity) and that they are particularly sensitive to discharge uncertainties. We question the utility and reliability of those signatures in experimental and modeling hydrological studies, and we underscore the general importance of accounting for uncertainties in hydrological signatures.