The Monte Carlo Independent Column Approximation: An Assessment using Several Global Atmospheric Models
H. W. Barker, J. N. S. Cole, J.-J. Morcrette, R. Pincus, P. Räisänen, K. von Salzen, and P. A. Vaillancourt
Quarterly Journal of the Royal Meteorological Society, Vol. 134, pages 1463 - 1478, July 2008. doi:10.1002/qj.303.

Abstract

The Monte Carlo Independent Column Approximation (McICA) computes domain-average, broadband radiative flux profiles within conventional global climate models (GCMs). While McICA is unbiased with respect to the full ICA, it generates, as a by-product, random noise. If this by-product affects statistically significant impacts on GCM simulations, it could spell limit the usefulness of McICA. This paper assesses the impact of McICA's random noise on six GCMs. To this end, the GCMs performed ensembles of 14-day long simulations for various renditions of McICA; each with differing amounts of random noise. As seen in the past, low cloud fraction and surface temperature were impacted most by noise. However, all simulations using operationally viable versions of McICA showed no statistically significant impacts; even forprecipitation, a highly intermittent variable that one might expect to be sensitive to random fluctuations. Two GCMs showed statistically significant responses to an academic version of McICA that generates overly large smapling noise. Time series analyses of high resolution (i.e., typically 2 hourly) data revealed that fluctuations associated with most variables and GCMs are immune McICA noise. Moreover, the nature of these fluctuations can vary substantially among GCMs and most often they overwhelm any noise impacts. Overall, the results presented here corroborate a range of previous studies done on one GCM at a time: random noise produced by recommended versions of McICA has statistically insignicant impacts on GCM simulations.

Download a reprint (3.2 Mb, PDF format) or read the article on-line.
Further reproduction or electronic distribution is not permitted.