Scaling in Integrated Assessment: Problem or Challenge?

摘要

It is increasingly recognized that scale is a core methodological problem in many scientific fields. This is particularly true for Integrated Assessment, which operates by definition on multiple scales, both in time and space. Thus, the European Forum for Integrated Environmental Assessment (EFIEA) workshop on scaling organized by the International Centre for Integrative Studies (ICIS) in Maastricht, aimed at collecting state-of-the-art knowledge on scales from a variety of angles, was quite timely. Based on this state-of-the-art representation, the building blocks for a potential research agenda for scaling in Integrated Assessment can be defined. Regarding the state-of-the-art, much scaling "handwork" has been done in Integrated Assessment (IA) modelling. But it mainly concerns statistical up-and down-scaling techniques to move from a lower spatial scale level to a higher one and vice versa. Notwithstanding the usefulness of these statistical techniques, we now realize that much more is needed to represent multiple scales in IA-modelling. Furthermore, other tools commonly used in IA, such as scenario building, are in need of innovative multiple scale methods. Finally, a largely unexplored field is the relation between scaling and uncertainty. In general, this paper gives a portfolio of ideas how to deal with scaling in IA-tools & methods. Rather than discussing in-depth the relation between scaling and a particular IA-instrument, we touch upon a number of scaling issues and present some ideas how to incorporate multiple scales in IA-tools & methods. First of all we address the overall methodological problem behind scaling in Integrated Assessment. Then we discuss scaling in IA-modelling, treating three different heuristic scaling-methods that are currently used. Next, we sketch scaling in IA-scenarios, giving two recent examples of multiple-scale scenario assessments. We then discuss scaling in relation to the representation of agents, followed by a brief discussion of mscaling and uncertaing. We finish with a set of recommendations for future IA-research.