Conceptual rainfall runoff (CRR) models usually require calibration to identify their parameter values, whereas their model structure is selected prior to modelling. Consid- erable efforts have been directed into calibration of lumped CRR models. Identification of the model structure on the basis of available data still remains unclear. The data-based mechanistic (DBM) approach does minimal assumptions of the model structure, which is identified using powerful statistic techniques. Moreover, there is a similarity between the CRR and DBM model.formulations. Based on this similarity, an integration of CRR and DBM models is proposed and evaluated. Two calibration strategies are investigated in the Upper Illinois river catchment (USA) for lumped modelling. Results show that the identi- fied TF model improves the simulated flow, especially in the time to peak, in comparison to the modelled flow of the conceptual model. It is suggested that this improvement is di- rectly related to the lag time parameter considered in the TF model between the effective rainfall and the flow. Semi-distributed rainfall runoff models provide advantages over lumped models in representing the effect of spatially variable inputs, outputs and catchment properties. However they are affected by parameter identifiability. Four calibration strategies are considered to analyse the ability to meaningful simulate flow at interior locations. Results show that there are no significant improvements at the catchment outlet when internal gauges are included. The behavioural parameter sets defined at the catchment outlet tend to be non-behavioural at the internal gauges. This tendency increases with the distance from the catchment outlet to the internal gauges. Considering only spatial variability of rainfall rather than also of parameter values did not improve the simulations at the out- let or at the internal gauges, compared to lumped modelling results. Calibration only at the catchment outlet using independent sampling of the internal subcatchments achieved similar results. Identification of lumped and semi-distributed CRR models is carried out using the RRTMSD modelling toolbox developed for this work.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:590036 |
| Date | January 2012 |
| Creators | Orellana Bobadilla, Barbara A. |
| Publisher | Imperial College London |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0021 seconds