Assessment of different metrics for physical climate feedbacks

We quantify the feedbacks from the physical climate system on the radiative forcing for idealized climate simulations using four different methods. The results differ between the methods and differences are largest for the cloud feedback. The spatial and temporal variability of each feedback is used to estimate the averaging scale necessary to satisfy the feedback concept of one constant global mean value. We find that the year-to-year variability, combined with the methodological differences, in estimates of the feedback strength from a single model is comparable to the model-to-model spread in feedback strength of the CMIP3 ensemble. The strongest spatial and temporal variability is in the short-wave component of the cloud feedback. In our simulations, where many sources of natural variability are neglected, long-term averages are necessary to get reliable feedback estimates. Considering the large natural variability and relatively small forcing present in the real world, as compared to the forcing imposed by doubling CO2 concentrations in the simulations, implies that using observations to constrain feedbacks is a challenging task and requires reliable long-term measurements.