Selection of conceptual models for regionalisation of the rainfall-runoff relationship

Abstract

This study attempts to identify relationships between suitable conceptual rainfall-runoff model structures and catchment types, and demonstrates an objective procedure for selection of model structures for use in model regionalisation studies. 12 conceptual model structures have been tested on 28 UK catchments. The model structures are combinations of three soil moisture accounting modules (the catchment wetness index model structure, a modified Penman model structure, and a probability distributed soil moisture model), and four routing models (two linear reservoirs in parallel, an adaptation that represents macropore flow, three linear reservoirs in parallel, and a model that includes leakage from the groundwater store). The 28 catchments were selected to cover a range of hydrological types based on known catchment characteristics (designed combinations of catchment size, baseflow index, and annual rainfall are used). The important points in objectively selecting model structures for regionalisation are proposed as: performance in calibration and validation, trade-off between high and low flow performance, and model complexity (here represented as the number of parameters). Although, the results in this study provide no evidence of relationships between catchment type and preferred model structure, the study indicates that out of the 12 trial model structures, the following four may be the most suitable for regionalisation across UK catchments: the modified Penman model with two parallel linear routing reservoirs, and the probability distributed soil moisture model with two parallel linear routing reservoirs, three parallel linear routing reservoirs, or the macropore adaptation.