An Improved Parameterization of Dust Dry Deposition Velocity for Climate Models

Thursday, 18 December 2014
Marcelo Chamecki and Livia Souza Freire, Pennsylvania State University, Department of Meteorology, University Park, PA, United States
The parameterization of dry deposition fluxes in climate models is an important component of the dust model, since it is one of the main processes responsible for the removal of aerosol particles from the atmosphere. The parameterization currently used in most climate models is obtained from a two-layer approach in which the physical domain between the first grid point and the ground surface is divided into an upper and a lower layer. The lower layer is the region where Brownian diffusion is relevant, and where particles interact with the surface roughness elements. In this layer important processes such as impaction, interception, and rebound of particles on the surface are parameterized. In the upper layer turbulent transport and gravitational settling determine the particle flux, and the effects of atmospheric stability become relevant. In the usual two-layer model, mean vertical gradients of particle concentration are approximated using a first order finite-difference approach, resulting in a set of two algebraic equations. The solution of these equations yields the desired parameterization for the deposition velocity. We argue that in the upper model, the finite-difference approximation is the largest source of error. In addition, the error introduced by this approximation is grid-dependent, increasing with increasing distance between the first grid point and the ground. We develop a new model in which the finite-difference approximation is not invoked. Instead, the analytical solution for the differential equation is obtained. This solution can be seen as a generalization of the Monin-Obukhov similarity theory to concentration of settling particles. We use large-eddy simulations to show the validity of the analytical solution for a range of particle sizes. The final expression obtained for the deposition velocity is as simple as the one currently used, and the parameterizations of impaction, interception, and rebound in the lower layer can be introduced in the same way. Compared to the proposed model, the deposition velocity parameterization currently used in climate models over predicts fluxes by as much as 30%. The errors tend to be largest in the range of particle sizes between one and ten micrometers, a range of critical importance for dust modeling.