Process for benchmarking models: dialogue between water managers and modellers
Kämäri, J.; Boorman, D.; Icke, J.; Perrin, C.; Candela, L.; Elorza, F. J.; Ferrier, R. C.; Bakken, T. H.; Hutchins, M.
The project Benchmark Models for the Water Framework Directive (BMW) has sought to provide advice and criteria to water managers in their selection, use and evaluation of models to assist the implementation of the Water Framework Directive (WFD). This advice is presented in a model evaluation protocol that comprises four stages, i. e.: Management task definition, Model code selection, Model performance assessment, A posteriori review of modeling. Within each evaluation stage there is a dialogue between the water manager and the modeller during which they reach a mutual understanding of the issues relating to that stage before moving forward with the next part of the modelling study. The dialogue is structured using a set of questions, which are accompanied with guidance notes that expand on the questions and the related considerations. The questions and the guidance notes are also contained in a computer-based model evaluation tool, which can also record the outcome of the discussions. The structured dialogue provides the scheme for benchmarking. The word benchmark can be used in different ways. Generally, it refers to a point of reference for the purpose of comparison, but in a few definitions benchmark is taken to mean the best, or superior, performance. Also while in some definitions the benchmark is the standard, in others it represents the process or approach by which performance or suitability is assessed. Benchmarking is here taken as a method of comparison. That implies that an appropriate benchmark has to be defined for each application, against which the comparison can be performed. It is important to note that the evaluation is closely related to an individual application. In terms of model application this obviously means the particular management task and study location. However, those participating in the dialogue bring to the process their unique experience and expertise, and these will also influence the conclusion from each stage of the discussion. Thus the outcome of a previous evaluation is not automatically valid for other, similar, applications, although it is entirely appropriate that previous evaluations feed into the current evaluation. This paper describes the development of the concept of benchmarking models and of the model evaluation protocol, and the reasoning that went into its development. It does not contain the full text of the guidance and consideration behind each step and question of the model evaluation protocol. Those can be found in the BMW River Basin Manager's Toolbox (http://www.rbm-toolbox.net).