Predicting runoff in ungauged UK catchments

Abstract

An important challenge for hydrologists is to develop a capability for predicting flow time-series in ungauged catchments, where historic flow records are not available to support model calibration. Using a database of 131 well-gauged UK catchments, this paper develops regression relationships between known catchment characteristics and parameters of a conceptual rainfall–runoff model. These relationships allow predictive models to be specified without calibration. In an attempt to improve the relevance of the relationships, the catchments are grouped into more and less permeable catchment types and regression equations are identified for each group separately. Also, equations more applicable to predicting high flows are identified separately to those more applicable to predicting low flows. Physically meaningful relationships between model parameters and catchment characteristics are generally absent. Despite this, results indicate that the best regression model improves the capability of predicting the higher range of flows in the less permeable catchments; however predicting low flows in these catchments and the full range of flows in more permeable catchments remains a challenge.