Science:
Earth is a complex system in which many biological and physical
components interact across all space and time scales. To
understand this system, earth scientists have traditionally
built large, multi-component models. However, it is difficult
to know when such a model has become sufficiently detailed for
its task and how confident one can be in its predictions. In a
generic linear system with feedbacks, Roe and Baker have shown
that normally distributed feedbacks give rise to a highly
skewed distribution of responses, similar to those seen for
climate sensitivity in ensembles of global models. Even
relatively narrow ranges of uncertainty in the feedbacks can be
amplified in the response. Thus, besides refining the feedback
uncertainties in traditional earth system models, scientists
and policy-makers must explore complementary approaches to
modeling.
