Retrieval of dust storm aerosols using an integrated Neural Network model

Dust storms are known to have adverse effects on public health. Atmospheric dust loading is also one of the major uncertainties in global climatic modeling as it is known to have a significant impact on the radiation budget and atmospheric stability. This study develops an integrated model for dust storm detection and retrieval based on the combination of geostationary satellite images and forward trajectory model. The proposed model consists of three components: (i) a Neural Network (NN) model for near real-time detection of dust storms; (ii) a NN model for dust Aerosol Optical Thickness (AOT) retrieval; and (iii) the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model to analyze the transports of dust storms. These three components are combined using an event-driven active geo-processing workflow technique. The NN models were trained for the dust detection and validated using sunphotometer measurements from the AErosol RObotic NETwork (AERONET). The HYSPLIT model was applied in the regions with high probabilities of dust locations, and simulated the transport pathways of dust storms. This newly automated hybrid method can be used to give advance near real-time warning of dust storms, for both environmental authorities and public. The proposed methodology can be applied on early warning of adverse air quality conditions, and prediction of low visibility associated with dust storm events for port and airport authorities.