Ideal point error for model assessment

Abstract

When analysing the performance of hydrological models, researchers use a number of diverse statistics. Although some statistics appear to be used more regularly in such analyses than others, there is a distinct lack of consistency in evaluation, making studies undertaken by different authors or performed at different locations difficult to compare in a meaningful manner. Moreover, even within individual reported case studies, substantial contradictions are found to occur between one measure of performance and another. In this paper we examine the Ideal Point Error (IPE) metric - a recently introduced measure of model performance that integrates a number of recognised metrics in a logical way. Having a single, integrated measure of performance is appealing as it should permit more straightforward model inter-comparisons. However, IPE relies on the adoption of a consistent and recognised benchmarking system. This paper examines one potential option for benchmarking IPE: the use of "persistence scenarios".