GC13V-05 Why Pattern Scaling Works in Earth System Models - And When It Does Not

Earth System Models (ESMs) show an approximately linear scaling between the regional temperature response to climate forcing and global mean temperature anomalies. Remarkably, this feature is found across a wide range of scenarios and ESMs, and forms the basis of “pattern scaling”, a technique used to efficiently compute the regional temperature responses from changes in global mean temperature. We present physical arguments to explain why this technique successfully describes simulated regional temperature responses across many ESMs and commonly used scenarios, but also to formally define what are the assumptions and limits of pattern scaling. Specifically, we use the local energy balance framework to show that regional warming scales exactly linearly with global mean temperature under multiple assumptions: (i) constant regional feedback parameters, (ii) homogeneous and (iii) exponential forcing. As a result, we show that pattern scaling fails when local feedbacks are temperature-dependent, the forcing substantially deviates from an exponential and/or evolves inhomogeneously across regions. Under non-exponential forcing, non-linearity arises from spatial differences in heat capacities and from latitudinal differences in ocean heat uptake. We demonstrate the effects of breaking assumptions (i)-(iii) with simple idealized numerical experiments, and we interpret the results in the context of CMIP6 simulations. For typical end-of-century CMIP6 scenario projections, assumptions (i)-(iii) are decently met for most of the planet and therefore the linear scaling approximation is remarkably accurate, except for high aerosol loading regions (where the homogeneous forcing assumption does not hold) and high-latitude regions (where the local feedback parameter changes with melting sea ice). Peak-and-decline scenarios, as well as experiments where radiative forcing is abruptly changed also display considerable deviations from the linear scaling owing to the different heat capacities mechanism. We discuss the practical implications of our findings for climate impact assessment studies.