DO4Models: Testing the Performance of Current Dust Emission Schemes from a Box and Climate Model Perspective

Thursday, 18 December 2014
Karsten Haustein1, Richard Washington1, James King2, Giles Wiggs1, David S G Thomas1 and Laurent Menut3, (1)University of Oxford, Oxford, United Kingdom, (2)Indiana University Bloomington, Bloomington, IN, United States, (3)Lab Meteorologie Dynamique, Palaiseau Cedex, France
Dust emission schemes in climate models are relatively simple and are tuned to represent observed background aerosol concentrations. Representations of dust emission in the models were developed from idealized experiments such as those conducted in wind tunnels. Improvement of current model dust emission schemes is hampered by a paucity of observations from key dust sources. The Dust Observations for Models project (DO4Models) was aiming on gathering data from source regions at a scale appropriate to climate model grid box resolution.

Here we present (1) the results of 1D box model simulations using three commonly used parameterizations for the horizontal and vertical dust emission flux, (2) using a newly proposed stochastic dust emission scheme for turbulent wind conditions, and (3) HadGEM3 regional climate model simulations using the current model setup for dust emissions. We are comparing box and RCM model results with DO4Models field campaign data retrieved over a typical dust source in Botswana during two consecutive dry seasons (2011 and 2012).

The box model performance is further tested using observed soil moisture content, aerodynamic surface stress, shear velocity, and soil size properties. The results suggest that current dust emission schemes do not capture the observed emission flux well. The saltation flux is hugely overestimated, whereas the vertical flux is moderately overestimated. The choice of the sand transport, soil moisture and roughness correction scheme is important but insufficient to bring modeled fluxes into agreement with observed dust fluxes. The stochastic scheme does not suffer from this flux disparity, but cannot be used in cases of strong surface saltation. Potential reasons for the diagnosed mismatch are discussed and the impact of spatial averaging over the 11 field sites within the 12x12km grid is evaluated.

HadGEM3 is tested with regard to its capability to reproduce the observed meteorological conditions. Very good agreement with regard to the magnitude of the diurnal cycle in 10m wind speed, wind gusts and surface shear stress is found. overall, the performance of the model regarding the dynamics is sufficient to focus on the shortcomings of the deployed dust emission scheme which is found to be the largest source of uncertainty in terms of dust emission flux.