A comparison of two representations of subgrid‐scale cloud structure in a global model: radiative effects as a function of cloud characteristics

Two approaches to account for the radiative impacts of subgrid‐scale variability of cloud in a general circulation model (GCM) are compared: (i) deterministic reduction of cloud optical depth imbedded within the radiative transfer scheme and (ii) stochastic subcolumns employed with the Monte Carlo Independent Column Approximation (McICA). This article analyzes impacts, as a function of cloud phase and cloud fraction, due to replacement of deterministic method (which is used currently in the GCM) with the McICA method, as well as the introduction of cloud‐water horizontal inhomogeneity and changes to the description of cloud vertical overlap. The largest radiative effects are produced by changing horizontal inhomogeneity of cloud water, which, when enhanced, generally decreases cloud albedo and emissivity, with the exception of some ice clouds, the albedo of which increases. Reducing the extent to which clouds overlap vertically has smaller and opposite effects relative to increasing horizontal inhomogeneity, with the exception of some ice clouds, where both effects have the same sign. These effects are, however, less pronounced than the increases to both cloud albedo and emissivity that stem from replacement of the deterministic optical depth reduction method by McICA. In essence, deterministic reduction of cloud optical depth has a more aggressive impact on radiative transfer than does McICA's stochastic sampling of horizontal inhomogeneity. When the McICA methodology is applied interactively in the GCM, both cloud fraction and water content for low‐level clouds are reduced relative to the deterministic method.