Cumulus parameterizations in chemical transport models

Abstract

Global three‐dimensional chemical transport models (CTMs) are valuable tools for studying processes controlling the distribution of trace constituents in the atmosphere. A major uncertainty in these models is the subgrid‐scale parametrization of transport by cumulus convection. This study seeks to define the range of behavior of moist convective schemes and point toward more reliable formulations for inclusion in chemical transport models. The emphasis is on deriving convective transport from meteorological data sets (such as those from the forecast centers) which do not routinely include convective mass fluxes. Seven moist convective parameterizations are compared in a column model to examine the sensitivity of the vertical profile of trace gases to the parameterization used in a global chemical transport model. The moist convective schemes examined are the Emanuel scheme [Emanuel, 1991], the Feichter‐Crutzen scheme [Feichter and Crutzen, 1990], the inverse thermodynamic scheme (described in this paper), two versions of a scheme suggested by Hack [Hack, 1994], and two versions of a scheme suggested by Tiedtke (one following the formulation used in the ECMWF (European Centre for Medium‐Range Weather Forecasting) and ECHAM3 (European Centre and Hamburg Max‐Planck‐Institut) models [Tiedtke, 1989], and one formulated as in the TM2 (Transport Model‐2) model (M. Heimann, personal communication, 1992). These convective schemes vary in the closure used to derive the mass fluxes, as well as the cloud model formulation, giving a broad range of results. In addition, two boundary layer schemes are compared: a state‐of‐the‐art nonlocal boundary layer scheme [Holtslag and Boville, 1993] and a simple adiabatic mixing scheme described in this paper. Three tests are used to compare the moist convective schemes against observations. Although the tests conducted here cannot conclusively show that one parameterization is better than the others, the tests are a good measure of the ability of the schemes to transport trace gases in a realistic manner under important meteorological conditions. From these tests two moist convection schemes and one boundary layer scheme are considered as the most promising for use in CTMs such as the National Center for Atmospheric Research's (NCAR) off‐line model. For moist convection the ECHAM3 form of the Tiedtke scheme and the inverse thermodynamic scheme performed well in the tests against observations. For the boundary layer the results of the nonlocal boundary layer scheme were more similar to observations than the simple adiabatic mixing scheme. Our results suggest that the middle and upper troposphere trace constituent profiles may not be sensitive to the boundary layer scheme employed in the model. The results of this one‐dimensional study imply that the conclusions of chemical transport modeling studies can be highly sensitive to the convective parameterization used in the transport model.