Modeling dust in East Asia by CESM and sources of biases

Mineral dust plays an important role in the Earth's climate system due to its effects on radiation budgets, clouds, chemistry and biosphere. However, modeled dust aerosol is not well constrained and large uncertainties exist in the modeled dust life cycle. We evaluate the dust spatial and temporal distributions simulated by the NCAR Community Earth System Model (CESM) with dust extinction retrievals (Luo et al., 2015a, b) against CALIPSO and CloudSat measurement, standard CALIPSO, MODIS, MISR, AERONET, and surface concentrations, with special focus on the Asian dust source regions. It is shown that CESM overestimates the dust extinction by 1-2 order of magnitude in the upper troposphere over Taklamakan and Gobi. The model produces too high dust optical depth (DOD) in summer but too low DOD in winter which may be caused by the wrong seasonality of dust emission in source regions. When compared with dust surface concentrations derived from the meteorological records of visibility, the model underestimates the dust surface concentrations over source regions during sand and dust storm events. The comparison of probability density function (PDF) of the dust surface concentration show that the model produces lower frequency of high dust concentration, which is further related to surface wind. The comparison of surface wind PDFs with ground observations shows that the model fails to capture the strong wind gusts during the sand and dust storm (SDS) events. Sensitivity experiments that apply dust emission schemes by Kok et al. (2014a, b) (K14) and Ginoux et al. (2012) (G01) and consider sub-grid surface wind variability by Zhang et al. (2016) (SGV) have been conducted to better understand the modeling issue in East Asia source regions. K14 and G01 further decreases modeled dust surface concentration during SDS events, while SGV increases the dust surface concentration. The model was also ran at different resolutions to understand the simulation of surface wind.