Regionalization of flow-duration curves through catchment classification with streamflow signatures and physiographic-climate indices

This study addresses the estimation of flow-duration curves (FDC) in ungauged sites through the catchment classification. Forty-six catchments in the Upper Po river basin (Italy) were analyzed and classified through two different frameworks: the first scheme consists of the application of two clustering methods in a series considering six streamflow signatures, and the second one treats indexes of climate, physiography, soil, and land-use with the same clustering procedure. Catchments have been classified into three homogeneous groups: the first one is characterized by the lowest runoff and flash-flood events, the second one includes maximum runoff, and the third one shows intermediate behaviour. The estimation of FDCs was done using a lognormal distribution, whereas the regionalization was constructed applying a stepwise multiple linear regression, followed by a leave-one-out cross-validation. The results show great performance improvement when the regionalization model is found by taking account of the three different hydrological classes, with a mean absolute percentage error that decreases from 11% for the single region case to 7% in the three homogeneous regions case