Emulating Future Climate Projections from Global Climate Models: Methodologies and Challenges

Abstract:

Pattern scaling methods have been used since the 1990s to estimate the results of global climate models (GCMs), in particular for emissions scenarios for which GCM simulations are not available. The basic method uses global mean surface temperature as the scaling variable, and relies on the assumption of a constant spatial pattern of change per unit global warming. This presentation will briefly review the status of pattern scaling science, using results from the published literature, and from a recent workshop held at NCAR. Successes and challenges will be illustrated of the use of pattern scaling to provide information on future changes for use by the impacts and integrated assessment modelling communities. This activity reflects anticipation of an enhanced role for emulation methods in a new process underway to produce integrated scenarios of future climate and societal change, which extends the number of scenarios of interest beyond the small set of RCPs used in GCM simulations for CMIP5.

Relevant challenges include effects of non-linearities caused by: different responses to different levels of greenhouse gas forcing; different timescales of regional response for alternative forcing pathways leading to the same global temperature response; combining the effects of multiple individual forcing agents. Understanding and projecting the responses to forcing agents such as aerosols and land use is likely to be particularly important in emulating changes during the next few decades. Further challenges include how to represent uncertainties and multivariate changes robustly in order to provide a basis for realistic assessments of impacts and risks, and extensions to the basic pattern scaling paradigm. Such extensions include consideration of scaling variables other than global mean temperature, and the recent development of new approaches to emulation using alternative statistical techniques and different physical assumptions.