Models of the class used in the CMIP6 experiment to make global
climate projections are imperfect representations of reality that
differ widely in regard to the overall magnitude of warming, in their
regional projections, and in their short-range predictions. While
better models of the underlying physical processes are ultimately
needed, immediate improvement may come simply from better methods to
combine existing models. Contributions are solicited on new methods to
fuse models of climate and weather ranging from output averaging techniques to methods that
dynamically combine model components in a synchronizing, interactive
ensemble. The importance (or lack thereof) of nonlinearities in
determining the sufficiency of output averaging is a topic of special
interest.