Appropriate Accuracy of Models for Decision-Support Systems: Case Example for the Elbe River Basin

Given the growing complexity of water-resources management there will be an increasing need\ud for integrated tools to support policy analysis, communication, and research. A key aspect of the design is the\ud combination of process models from different scientific disciplines in an integrated system. In general these\ud models differ in sensitivity and accuracy, while non-linear and qualitative models can be present. The current\ud practice is that the preferences of the designers of a decision-support system, and practical considerations\ud such as data availability guide the selection of models and data. Due to a lack of clear scientific guidelines the\ud design becomes an ad-hoc process, depending on the case study at hand, while selected models can be overly\ud complex or too coarse for their purpose. Ideally, the design should allow for the ranking of selected\ud management measures according to the objectives set by end users, without being more complex than\ud necessary. De Kok and Wind [2003] refer to this approach as appropriate modeling. A good case example is\ud the ongoing pilot project aiming at the design of a decision-support system for the Elbe river basin. Four\ud functions are accounted for: navigability, floodplain ecology, flooding safety, and water quality. This paper\ud concerns the response of floodplain biotope types to river engineering works and changes in the flooding\ud frequency of the floodplains. The HBV-D conceptual rainfall-runoff model is used to simulate the impact of\ud climate and land use change on the discharge statistics. The question was raised how well this rainfall-runoff\ud model should be calibrated as compared to the observed discharge data. Sensitivity analyses indicate that a\ud value of R2 = 0.87 should be sufficient