Technical note: Diagnostic efficiency – specific evaluation of model performance

Abstract. A better understanding of the reasons why hydrological model performance is unsatisfying represents a crucial part meaningful evaluation. However, current evaluation efforts are mostly based on aggregated efficiency measures such as Kling–Gupta (KGE) or Nash–Sutcliffe (NSE). These provide relative gradation performance. Especially in case weak it important to identify different errors which may have caused unsatisfactory predictions. originate from parameters, structure, and/or input data. In order more insight, we define three types be related their source: constant error (e.g. by consistent data precipitation), dynamic structural deficient storage routine) and timing routines/parameters). Based these errors, propose novel diagnostic (DE) measure, accounts for types. The disaggregation DE into its metric terms can visualized plain radial space using polar plots. major advantage this visualization technique that contributions clearly differentiated. proof concept, first generated time series artificially with (i.e. simulations surrogated manipulating observations). By computing plots reproduced could then supply evidence concept. Finally, tested applicability our approach modelling example. For particular catchment, compared streamflow realized parameter sets observed streamflow. example, plot suggests explain overall large extent. proposed provides tool developers users facilitates interpretation measure well similar well-established hydrology.