Climatic Change

, Volume 104, Issue 2, pp 255–285

How well do integrated assessment models simulate climate change?


    • Netherlands Environmental Assessment Agency
  • Jason Lowe
    • MetOffice Hadley Centre
    • Department of MeteorologyThe University of Reading
  • Elke Stehfest
    • Netherlands Environmental Assessment Agency
  • Laila Gohar
    • MetOffice Hadley Centre
    • Department of MeteorologyThe University of Reading
  • Andries F. Hof
    • Netherlands Environmental Assessment Agency
  • Chris Hope
    • Judge Business SchoolCambridge University
  • Rachel Warren
    • Tyndall Centre, School of Environmental SciencesUniversity of East Anglia
  • Malte Meinshausen
    • Potsdam Institute for Climate Impact Analysis
  • Gian-Kasper Plattner
    • Environmental Physics, Institute of Biogeochemistry and Pollutant DynamicsETH Zürich
    • Climate and Environmental Physics, Physics InstituteUniversity of Bern

DOI: 10.1007/s10584-009-9764-2

Cite this article as:
van Vuuren, D.P., Lowe, J., Stehfest, E. et al. Climatic Change (2011) 104: 255. doi:10.1007/s10584-009-9764-2


Integrated assessment models (IAMs) are regularly used to evaluate different policies of future emissions reductions. Since the global costs associated with these policies are immense, it is vital that the uncertainties in IAMs are quantified and understood. We first demonstrate the significant spread in the climate system and carbon cycle components of several contemporary IAMs. We then examine these components in more detail to understand the causes of differences, comparing the results with more complex climate models and earth system models (ESMs), where available. Our results show that in most cases the outcomes of IAMs are within the range of the outcomes of complex models, but differences are large enough to matter for policy advice. There are areas where IAMs would benefit from improvements (e.g. climate sensitivity, inertia in climate response, carbon cycle feedbacks). In some cases, additional climate model experiments are needed to be able to tune some of these improvements. This will require better communication between the IAM and ESM development communities.

Copyright information

© Springer Science+Business Media B.V. 2009