Neglect by GCMs of subgrid-scale horizontal variations in cloud-droplet effective radius:: A diagnostic radiative analysis

Output from a global climate model (GCM) that employed a low-resolution two-dimensional cloud-system-resolving model (CSRM) in each column is used to assess the radiative impact of neglecting subgrid-scale horizontal variations in cloud-droplet effective radius r(e). For this diagnostic study, only liquid-phase variations in r(e) are addressed: the ice-cloud particle distributions are assumed to be constant. For reference calculations, values of r(e) in the CSRM cells are computed assuming that the droplet-number concentration N(cld) and the effective variance of droplet-size distribution are constant in a GCM cell. The independent-column approximation is used to produce flux profiles for each GCM column. Three alternative methods of setting horizontally-invariant re are examined. each of which resemble how r(e) is set in one-dimensional radiative-transfer models. Relative to the reference calculations. the other methods lead to positive spurious radiative forcings at the surface and at the top of the atmosphere. These stem from overestimation of optical-depth variability and, thus, reduced short-wave albedo of clouds. Globally averaged, these forcings range from 1 W m(-2) to 3 W m(-2), with tonal-mean biases reaching almost 15 W m(-2). The most severe biases arise from use of constant values of re over the land and the ocean. In addition, radiative effects due to unacknowledged uncertainty in Ncld (or r,) are assessed. It is shown that ad hoc, but not outlandish, estimates of unbiased uncertainty in N(cld) impart biases on estimates of the earth's solar-radiation budget (tantamount to a spurious radiative forcing). These arise through the chain of nonlinear relations that link N(cld) to solar radiative transfer.