13 June 2016 - 14 June 2016
The workshop specifically focuses on
1) State-of-the-art in different communities in what concerns using multiple models, as well as constructing and working with multi-model ensembles; examples of existing multi-model ensembles and open questions; comparison of advantages and shortcomings of current practices;
2) Statistical methods and techniques that have been used to “integrate” alternative models, examples of their applications in different fields; comparison of their power and limitations;
3) Concepts of model validation, specifically focusing on multi-model ensembles;
4) Challenges and good practices related to communicating uncertainty and its sources to end-users and decision makers due to multiplicity of models and their stochastic nature.
Composite social-environmental systems are at the focus of systems analysis nowadays. Due to their high complexity, a plethora of models has been put forward to represent the dynamics of these systems. In particular, a few models attempting to represent the same phenomenon may differ because of initial and boundary conditions, parametric and structural uncertainty and, therefore, produce different results.
Climate science is the field, which pioneered recognition of the need and challenge of model inter-comparison and integration. CMIP – Coupled Model Inter-comparison Project under the World Climate Research Programme (WCRP) – was one of the first attempts to standardize and consolidate the study of coupled atmosphere-ocean general circulation models across different groups in the systematic fashion. Later projects, such as CORDEX-MIP, Water-MIP, ISI-MIP, Ag-MIP etc., extend the model inter-comparison to other dimensions relevant to integrated assessment. For example, the ISI-MIP project focuses on providing cross-sectoral global impact assessments, Water-MIP focuses on inter-comparisons of land-surface-hydrology models and global hydrology models, and so on.
Generally, it is believed that using knowledge from several available models representing the same phenomenon should improve our confidence in the overall results, which they produce; this bringing together knowledge from model ensembles is referred to as “integration of models”. IPCC adopted the practice of using multiple models for deriving future projections of climate. Several methods have been proposed to “integrate” (or “average”) alternative models, among which computing a multi-model mean with equal weights is the simplest. Many researchers argue for using weights representing our experience and/or expectations regarding model skill, which relates to a more general issue of model validation. The raise of computational power and accumulation of historical data of appropriate quality have triggered the use of past-experience-based Bayesian techniques. One of the more recent advances is the idea of deriving weights based on the pseudo out-of-sample predictive ability of models. Despite considerable progress in particular applications, development of well-justified tools for integration of the results of alternative models into a consolidated picture remains a strong challenge in systems analysis of complex social-environmental systems.
Last edited: 21 June 2016
International Institute for Applied Systems Analysis (IIASA)
Schlossplatz 1, A-2361 Laxenburg, Austria
Phone: (+43 2236) 807 0 Fax:(+43 2236) 71 313