Hydrological modelling in presence of non-stationarity induced by urbanisation: an assessment of the value of information

The proposed protocol of the workshop is followed, which regards the investigation of the effect of non-stationarity due to urbanisation on the performance of a hydrological model. In particular, the rainfall-runoff component of HYDROGEIOS modelling framework (Efstratiadis et al., 2008) is used. This is a parsimonious model of the conceptual type, based on the idea of Hydrological Response Unit (HRU). It is parameterised per HRU with seven parameters in each. Both a lumped and a semi-distributed version are employed. In the latter, two HRUs are assumed, representing the urban and rural areas of the basin. The Evolutionary Annealing Simplex method is used to obtain the best parameter set along with a large number of other retained parameter sets. Levels 1 and 2 of the proposed protocol provide the necessary information for analysis of Level 3, where a stochastic framework is considered inspired by the ideas proposed by Montanari & Koutsoyiannis (2012). This takes into account external information on urbanised fraction of the studied basin. A relationship is established between data on fraction of urbanised area and one of more parameters of the lumped model, while the semi-distributed one takes into account the fraction of urbanised area explicitly. Comparison of prediction intervals with and without exploiting such relationship allows the assessment of the value of information regarding the factor that induces nonstationarity. The methodology as a whole is applied to one of the two drainage basins that show growing urbanisation (Ferson Creek at St. Charles, USA).