Managing uncertainty in integrated environmental modelling frameworks: The UncertWeb framework

The "Model Web" presents a vision of a future where models are exposed as Web Services in a flexible distributed architecture and can be discovered, combined into complex workflows and executed over a distributed architecture. Such a system provides tremendous opportunities for integrated environmental modelling but raises several significant challenges. This paper focuses on the specific issue of uncertainty management in an era of increasing access to both data and models, where model outputs are increasingly used for policy and decision making. For any modelling framework to be of practical use, tools are needed to support the practical management of uncertainty by addressing gaps in uncertainty information, propagating uncertainty through model chains, managing changes in spatial and temporal scale and communicating the uncertain outputs of modelling workflows. In this paper we describe developments in the UncertWeb project which will support an "uncertainty enabled Model Web". We provide an overview of uncertainty in modelling, review uncertainty management in existing modelling frameworks and consider the semantic and interoperability issues raised by integrated modelling. We describe the scope and architecture required to support uncertainty management including tools which support elicitation, aggregation/disaggregation, visualisation and uncertainty/sensitivity analysis, and highlight areas that require further research.JRC.H.5-Land Resources Managemen