Both climate forcing and climate sensitivity persist as stubborn uncertainties limiting the extent to which climate models can provide actionable scientific scenarios for climate change. A key, explicit control on cloud-aerosol interactions, the largest uncertainty in climate forcing, is the vertical velocity of cloud-scale updrafts. Model-based studies of climate sensitivity indicate that convective entrainment, which is closely related to updraft speeds, is an important control on climate sensitivity. Updraft vertical velocities also drive many physical processes essential to numerical weather prediction. Vertical velocities and their role in atmospheric physical processes have been given very limited attention in models for climate and numerical weather prediction. The relevant physical scales range down to tens of meters and are thus frequently sub-grid and require parameterization. Many state-of-science convection parameterizations provide mass fluxes without specifying vertical velocities, and parameterizations that do provide vertical velocities have been subject to limited evaluation against what have until recently been scant observations. Atmospheric observations imply that the distribution of vertical velocities depends on the areas over which the vertical velocities are averaged. Distributions of vertical velocities in climate models may capture this behavior, but it has not been accounted for when parameterizing cloud and precipitation processes in current models. New observations of convective vertical velocities offer a potentially promising path toward developing process-level cloud models and parameterizations for climate and numerical weather prediction. Taking account of the scale dependence of resolved vertical velocities offers a path to matching cloud-scale physical processes and their driving dynamics more realistically, with a prospect of reduced uncertainty in both climate forcing and sensitivity.

Abstract. Both climate forcing and climate sensitivity persist as stubborn uncertainties limiting the extent to which climate models can provide actionable scientific scenarios for climate change. A key, explicit control on cloud-aerosol interactions, the largest uncertainty in climate forcing, is the vertical velocity of cloud-scale updrafts. Model-based studies of climate sensitivity indicate that convective entrainment, which is closely related to updraft speeds, is an important control on climate sensitivity. Updraft vertical velocities also drive many physical processes essential to numerical weather prediction. Vertical velocities and their role in atmospheric physical processes have been given very limited attention in models for climate and numerical weather prediction. The relevant physical scales range down to tens of meters and are thus frequently sub-grid and require parameterization. Many state-of-science convection parameterizations provide mass fluxes without specifying vertical velocities, and parameterizations which do provide vertical velocities have been subject to limited evaluation against what have until recently been scant observations. Atmospheric observations imply that the distribution of vertical velocities depends on the areas over which the vertical velocities are averaged. Distributions of vertical velocities in climate models may capture this behavior, but it has not been accounted for when parameterizing cloud and precipitation processes in current models. New observations of convective vertical velocities offer a potentially promising path toward developing process-level cloud models and parameterizations for climate and numerical weather prediction. Taking account of scale-dependence of resolved vertical velocities offers a path to matching cloud-scale physical processes and their driving dynamics more realistically, with a prospect of reduced uncertainty in both climate forcing and sensitivity.

Transport of radon-222 and methyl iodide by deep convection is analyzed in the Geophysical Fluid Dynamics Laboratory (GFDL) Atmospheric Model 2 (AM2) using two parameterizations for deep convection. One of these parameterizations represents deep convection as an ensemble of entraining plumes; the other represents deep convection as an ensemble of entraining plumes with associated mesoscale updrafts and downdrafts. Although precipitation patterns are generally similar in AM2 with both parameterizations, the deep convective mass fluxes are more than three times larger in the middle- to upper troposphere for the parameterization consisting only of entraining plumes, but do not extend across the tropopause, unlike the parameterization including mesoscale circulations. The differences in mass fluxes result mainly from a different partitioning between convective and stratiform precipitation; the parameterization including mesoscale circulations detrains considerably more water vapor in the middle troposphere and is associated with more stratiform rain. The distributions of both radon-222 and methyl iodide reflect the different mass fluxes. Relative to observations (limited by infrequent spatial and temporal sampling), AM2 tends to simulate lower concentrations of radon-222 and methyl iodide in the planetary boundary layer, producing a negative model bias through much of the troposphere, with both cumulus parameterizations. The shapes of the observed profiles suggest that the larger deep convective mass fluxes and associated transport in the parameterization lacking a mesoscale component are less realistic. Copyright 2007 by the American Geophysical Union.