In this study, comparative analysis and step-by-step evaluation of existing wind-blown dust models were performed to improve the model accuracy. The models used in the evaluation processes are CMAQ-DUST model, which is used as current working model of CMAQ, SCNU model, which accounts for the effects of soil texture in detail, and KOK model, which elaborately takes into account the effects of soil erodibility.
Two methods were used to evaluate the models: a direct method that compares the calculated model outputs with observations and an indirect method that compares the aerosol optical depth (AOD) or the aerosol concentration predicted by a three-dimensional chemical transport model, employing the wind-blown dust modules in it, with observations.
The direct verification used the results of the Japan-Australia Dust Experiment (JADE) project conducted in Australia in 2006. The JADE project is the only project that provided observation data for each step of dust emission calculation procedure, i.e., threshold friction velocity, saltation flux, and sandblasting flux. The threshold friction velocity was predicted very accurately when accurate input data (land cover, vegetation fraction, etc.) were used. In addition, when the threshold friction velocity was accurate, the KOK model predicted the saltation flux the best within the range of measured friction velocities. As for the saltation flux prediction, the differences among the models were not remarkable. The differences in saltion flux schemes appeared not very important because the saltation flux schemes can be tuned simply by changing the coefficients. In the case of sandblasting flux, which means the amount of vertical mass flux of dust, all models showed considerable underestimation, indicating that more thorough verification of the sandblasting flux scheme is required.
One limitation of the JADE project was that the surface of the site where the experiments were conducted consisted of only one soil texture type called loam. Therefore, to understand the influence of soil texture, different soil texture types were used as input in each model and the results were compared. CMAQ-DUST model predicted high dust emission flux for soil textures with high silt fraction, while SCNU and KOK models predicted high dust emission flux for soil textures with high clay fraction.
The predictions of the wind-blown dust models for East Asia were quite different from those for JADE; the dust concentrations around the Gobi Desert were overestimated or underestimated depending on the model used and the episode modeled. SCNU overestimated the dust concentrations consistently in all episodes, whereas CMAQ and KOK overestimated dust concentrations during severe dust events but underestimated during weak dust events.
The KOK model showed very strong dependency on wind speed, resulting in overestimation of dust concentration during severe dust events and underestimation during weak events. The reason of this characteristic is α, which is a parameter accounting for the soil erodibility used in in the KOK scheme for calculating the amount of dust emission. The KOK model is designed to respond more sensitively to friction velocity for the soils with smaller erodability; the value of α is higher for a larger threshold friction velocity. In order to reduce the sensitivity to it, C⍺, an empirical coefficient used in the calculation of α, was adjusted to a minimum value. The simulation using the adjusted parameter reduced the overestimation of dust concentration remarkably during strong dust events, while the dust emissions during weak dust events were not changed a lot, improving the model performance significantly. This result implies that the effect of soil erodibility was lower than expected in East Asia. The second East Asia dust episode (EP2) showed similar environmental conditions (soil moisture content, friction velocity, roughness length) with those of JADE. Therefore, the results of EP2 were compared with those of the JADE project. The differences in model-predicted sandblasting fluxes were within a factor of two. However, the models severely underestimated the dust emission for the JADE project and slightly overestimated for EP2. This indicates that the current wind-blown dust models do not adequately account for all the phenomena occurring during the wind erosion process. Further fundamental research on physical processes of wind erosion is warranted.