Analysis and development of conceptual rainfall-runoff model structures for regionalisation purposes

Abstract

This paper attempts to investigate, in detail, the behaviour of the selected conceptual rainfall runoff model structures (a Pen manbased model and a probability distribution-based model) by using a novel method of dynamic parameter identifiability analysis (DYNIA). Two UK catchments were used as case studies. This paper shows that potential weaknesses of model structures are uncovered by this analysis; a) the optimum parameter shifts over the time domain, and b) insensitive model parameters shift over the wet period and dry period. However, attempting to interpret these results, and use them as a basis for improving the model structure proved difficult. The dynamic relationships between the model parameter and the variable soil moisture state in the model have been considered based on these analyses. Possible model modifications have been suggested: a) Making the proportion of rainfall that bypasses the soil stores responsive to soil wetness, rather than a constant, and b) Directing this bypassed rain fall to the fas trouting reservoirs rather than splitting it between fast and slow reservoirs. However, substantial and well-founded changes in the model structures had marginal effect on the time-series output. The lack of improved performance raises a number of questions and points the way forward for more research. The results may suggest that improvement of the model performance depends more on the quality of data (or catchment information) rather than the hydrological representation of the different catchments.