Noise due to the Monte Carlo independent‐column approximation: short‐term and long‐term impacts in ECHAM5

The Monte Carlo independent‐column approximation (McICA), as a method for computing domain‐average radiative fluxes, allows a flexible treatment of unresolved cloud structure, and is unbiased with respect to the full independent‐column approximation, but its flux estimates contain conditional random noise. In our previous study with the ECHAM5 atmospheric general‐circulation model with prescribed sea‐surface temperatures (SSTs), McICA noise caused a slight reduction in low‐cloud fraction. Here, we first demonstrate that this feature originates from an immediate, nonlinear response of precipitation formation to McICA's random errors in radiative‐heating rates, and is subsequently amplified by a radiative feedback. We then study the long‐term impacts of McICA noise on climate in ECHAM5 simulations employing a mixed‐layer ocean model. The use of interactive rather than prescribed SST somewhat amplifies the reduction in low‐cloud fraction, which contributes to an overall warming of simulated climate with increasing McICA noise. For a typical implementation of McICA, the global‐mean 2 m air temperature is 0.33 K higher than for a low‐noise reference simulation. A 0.1 µm systematic increase in cloud‐droplet effective radius r e causes a similar global‐mean warming (0.31 K), with generally similar spatial patterns. However, in the eastern tropical Pacific, McICA noise has locally larger effects than the uniform perturbation in r e . It is concluded that the climatic impacts of McICA noise mainly represent a straightforward response to the systematic perturbation in the surface and top‐of‐atmosphere energy budget related to the initial reduction in low cloudiness.