The response of the atmosphere to surface temperature changes greatly depends on the spatial pattern of those changes. Recent work has shown that the total response can be expressed as the sum of the responses to the temperature change at each location, taken in isolation. As a result, it is useful to take a Green's function approach, in which atmosphere-only patch experiments performed systematically over the whole globe are used to find the spatial dependence of a variable of interest (such as the net top-of-atmosphere radiation) on the (sea) surface temperature.
The Green's Function Model Intercomparison Project (GFMIP) is a project to standardize these experiments, and collect and analyze their output. We have developed an experimental protocol by conducting a series of sensitivity tests using the HadAM3, ICON, and CanESM5 models, which have helped us determine: a patch layout that allows a proper reconstruction of historical net TOA radiative flux response while minimizing the number of patches used; the utility of using both cooling and warming patch experiments to construct a Jacobian; the best base climatology to employ; and the number of simulation years needed for both the control and patch experiments. In addition to presenting this protocol, we discuss results from our sensitivity tests that give us new insight into the nature of the atmospheric Green's function method, including the tradeoff between patch size, temperature perturbation, and simulation length for reducing the uncertainty of our Jacobian; the nonlinear nature of cooling vs. warming patches; and the dependence of this nonlinearity on the size of patches.