Evaluation of mineral dust aerosol optical depth and related components from the CHIMERE-DUST model using satellite remote sensing and ground-based observations

Dust aerosol optical depth (AOD) simulated by the CHIMERE-DUST model over North Africa during 2006-2008 was evaluated by the SeaWiFS Deep Blue satellite data and AERONET ground based remote sensing measurements. The simulations were conducted using two meteorological inputs, ECMWF and CFSR forecast data, downscaled by the WRF model. The two simulations generated similar AODs and PM10 surface concentrations, which are highly correlated to dust emissions. The comparison of the simulated daily value of AOD with the satellite data showed good correlations (0.6-0.8) between them. A slightly better correlation was achieved using WRF-CFSR data. The regression coefficients fluctuated around 1.0 in cold months and decreased in summer, with minimums in August. The biases had similar seasonal cycles. This is consistent with the comparison against AERONET hourly AOD data that the model underestimated AODs in warm months at most stations. The comparison of PM10 surface concentration and dry deposition also showed negative biases in warm months at stations in the Sahel. The significant underestimation of wet deposition was found in monsoon season due to the unrealistic simulation of mesoscale convective system. It indicates that the low dust emission in August is partly associated with cold pools generated by the convective precipitation. Frequent overestimations of AOD were found in the northwest mostly because of the high dust emission produced by the overestimated winds near the Atlas Mountains. The dust emissions resulted from the two simulations were similar in terms of intra annual variations but differed in terms of annul dust emission. The estimated annual emission using WRF-CFSR data could be about 200 Tg more than the estimation using WRF-ECMWF data.