Calibration and Regionalisation of Rainfall-Runoff Models

Current rainfall-runoff models require the calibration of at least some key parameters to yield reliable predictions of the continuous response of catchment systems. Recent years have seen considerable advances in the understanding of this calibration, or, better, parameter estimation, process. Examples are the recognition that estimates of parameter (and therefore prediction) uncertainty should be standard in any hydrological study, the fact that more information can be retrieved from rainfall-runoff time-series using multi-objective approaches, and the realization that model structural uncertainty is larger than has generally been assumed so far. While these findings are considered in state-of-the-art parameter estimation in gauged catchments, they have so far only found very limited recognition in the modeling of ungauged catchments. However, these issues are central to the regionalisation of rainfall-runoff models. This paper examines their effect on the regionalisation process, based on a case study using ten catchments located in the southeast of England. The consequences of parameter and model structural uncertainty are demonstrated and new approaches to consider these effects are introduced.