Evaluation of WRF-Chem-RTFDDA dust forecasts over the MENA region using in-situ and remote-sensing observations

Abstract

We employed the combined WRF-Chem-RTFDDA model to forecast dust storms in the Middle East and North Africa (MENA). WRF-Chem simulates the emission, transport, mixing, and chemical transformation of trace gases and aerosols simultaneously with the meteorology. RTFDDA continuously assimilates both conventional and nonconventional meteorological observations and provides improved initial conditions for dust analyses and forecasts. WRF-Chem-RTFDDA was run at a horizontal resolution of 9 km using the dust only option without inclusion of anthropogenic aerosols and chemical reactions. The synoptic conditions of the dust events were characterized by a cold front at the low level and an upper-level low-pressure system over the Western Mediterranean. WRF-Chem-RTFDDA was run in continuous assimilation mode, assimilating meteorological observations only, and launching 48-h free forecasts (FF) every 6 h. Two cold starts (CSs) for data assimilation and dust emissions initiation were performed during the study period. NCEP/GFS global analyses and forecasts provided initial and lateral boundary conditions. No global dust model was used for initialization and no dust observations were assimilated. We analyzed the skill of the WRF-Chem-RTFDDA system in reproducing the horizontal and vertical distributions of dust by comparing the FF to Meteosat SEVIRI dust images, MODIS AOD retrievals, CALIPSO extinction coefficients and CAMS aerosols-reanalysis AOD calculations. The skill was analyzed as a function of FF lead time and of the period of time from the CSs. RMSE, bias and correlation between the modeled and CALIPSO measured extinction coefficients were also examined. WRF-Chem-RTFDDA reproduced the main features of the studied dust storms reasonably well. The time distance from the CSs played a more significant role in determining the dust-forecast skill than free-forecast lead time. Since no external dust information was provided to the model, dust emissions and dust spin-up by WRF-Chem played a critical role in dust forecasts. The vertical extent of the CALIPSO extinction coefficients were reasonably well reproduced once model emissions were spun-up. False alarms rates range from 0.03 to 0.26, with many below 0.15, indicating satisfactory performance as a warning system. This study shows the feasibility of dust forecasts using minimal input data over the MENA region.