Identification of parsimonious rainfall-runoff models for gauged and ungauged catchments
PublisherImperial College London (University of London)
MetadataShow full item record
Rainfall-runoff models are indispensable tools in most hydrological studies. However, the identification of appropriate models remains a difficult problem despite decades of research. This dissertation analyses the identification of conceptual, parsimonious (which may be best translated as parameter efficient), lumped continuous models for gauged and ungauged catchments. The emphasis of this study lies on review and development of new methodologies, not so much on extensive applications. Though different investigative examples of how these new approaches can be applied are shown and a limited case study is provided. A modelling toolkit has been developed after an extensive review of current model structures and modelling procedures. It consists of two components, a Rainfall-Runoff Modelling (RRMT) and a Monte Carlo Analysis Toolbox (MGAT). These allow for the quick implementation and evaluation of rainfall-runoff model structures, although the latter has much more general applicability. The review of local modelling procedures, i.e. applicable to gauged catchments, has lead to the analysis and further development of multi-objective procedures for model performance and identifiability analysis. A novel DYNamic Identifiability Analysis (DYNIA) methodology has been developed which, amongst other things, can be used to analyse inadequacies in a model structure. Both approaches are combined in a framework of corroboration and rejection which can be applied to any dynamic mathematical model structure. This local procedure is extended to the regional case, i.e. applicable to ungauged catchments. Regional information about model parameters is used to derive statistical relationships with catchment characteristics. Various approaches are discussed and analysed. The most promising techniques, combined with new ways of considering uncertainty in the regional modelling process, are integrated in a regional modelling framework. Both frameworks are applied in a case study to 23 catchments located in the Thames river basin in the UK, which have a particularly wide range of geologies.