On the use of multiple criteria for a posteriori model rejection: Soft data to characterize model performance

Abstract

[1] Land surface hydrologic models are commonly evaluated based upon the degree of correspondence between measured and modeled discharge. In this paper we illustrate significant shortcomings associated with the simple discharge based evaluation strategy. A standard conceptual hydrologic model is applied within a Monte Carlo framework to two catchments representing significantly different hydrologic regimes. Time source hydrograph separations are derived, in addition to modeled discharge, and used to more completely characterize model functioning, across the entire a priori parameter distribution. The inclusion of hydrograph separation results in improved characterization of parameter uncertainty, and in one of the cases, complete model rejection.

Citizen science in hydrology and water resources: opportunities for knowledge generation, ecosystem service management, and sustainable development

“…Distributed measurements, even of short duration, may also be particularly useful for regionalization of hydrological models (Vandewiele and Elias, 1995;Parajka et al, 2005;Buytaert and Beven, 2009). Lastly, multi-objective calibration methods can address the issue of heterogeneous data availability of river basin modeling (e.g., Vaché et al, 2004;Parajka et al, 2007).…”

Improving Flood Prediction Assimilating Uncertain Crowdsourced Data into Hydrological and Hydraulic Models

“…provided by experimentalists. Vaché et al (2004) demonstrated the effectiveness of using soft-data for multiobjective calibration of hydrological models.…”

“…Although some efforts to validate model results against such observations have been made, these are mainly done in a post-event analysis (Aronica et al, 1998;Seibert and McDonnell 2002;Vaché et al 2004;Sheffield et al 2006;Seibert and Beven 2009). The added value of information coming from citizens, therefore, is not typically integrated into hydrological and/or hydraulic models Nowadays, model updating occurs only in the form of data assimilation using measurements of streamflow, soil moisture, etc.…”

Integrating heterogeneous landscape characteristics into watershed scale modelling

“…One way to achieve this is the integration of data which can not directly used in the model (soft data) or expert knowledge into the model calibration or validation process (Seibert and McDonnell, 2002). Vaché et al (2004Vaché et al ( , 2006 have used soft data on mean transit time to test and reject several model structures in order to find a model which represents the governing hydrological processes best. Another way to consider soft data is the use of expert knowledge, for example through the analysis of spatial data which are not imperative for the model set up, but could be used as a additional qualitative information.…”