Numerical aspects of applying the fluctuation dissipation theorem to study climate system sensitivity to external forcings

The fluctuation dissipation theorem (FDT), a classical result coming from statistical mechanics, suggests that, under certain conditions, the system response to external forcing can be obtained using the statistics of natural fluctuation of the system. The application of the FDT to the most sophisticated climate models and the real climate system represents a difficult problem due to the huge dimensionality of these systems and the lack of the data available for proper sampling of the system natural variability. As a con sequence, one has to use some regularization procedures constraining the form of permitted perturbations. Naturally, the skill of the FDT depends on the type and parameters of the regularization procedure. In the present paper we apply FDT to predict the response of a recent version of the NCAR climate system model (CCSM4) to salinity and temperature forcing anomalies in the North Atlantic. We study the sensitivity of our results to the amount of available data and to key parameters used in our numerical algorithm.