Samo04

In this paper we describe the multi-run simulation experiment environment SimEnv and its application in quality assurance matters for computer models. SimEnv has been developed to provide key working techniques for experimenting with complex models. This includes a wide range of simulation and model output evaluation methods in combination with corresponding visualization techniques. The SimEnv framework facilitates the easy execution of multi-run model simulation experiments for standardized, pre-formed experiment types which represent different sampling strategies of the model’s input space. Further experiment types may easily be included, making SimEnv an open experimentation system. The coupling of models to the environment is supported by a simple interface, requiring only minimal model source code modifications. Uncertainty and sensitivity analyses are enabled in SimEnv by combining experiments available from the pool of predefined experiment types with interactive post-processing, applying sequences of related operators to both model output and reference data. Use of SimEnv as an experimental framework for models in global change research demonstrates the applicability of the approach to multi-input / multi-output problems with large amounts of spatio-temporal model output and emphasizes the importance of graphical result presentation and evaluation by appropriate visualization techniques.