Identifying dust emission sources accurately is essential for understanding their effects on air quality and developing effective control strategies. However, traditional methods using single-source satellite data often face challenges due to limitations in spatial resolution, temporal coverage, or spectral sensitivity. To address these issues, we introduce a multi-sensor approach that integrates data from SEVIRI, MODIS, and Sentinel-2 satellites. This approach enhances the precision and reliability of dust source detection. Sentinel-2\\\'s high spatial resolution (10–20 m) enables precise mapping of localized dust hotspots, while MODIS offers daily observations to monitor transient dust events. By combining spectral indices, such as the brightness temperature adjusted dust index (BDI), with thermal infrared bands, we can effectively differentiate dust-laden areas from bare soil or cloud cover. Machine learning algorithms, trained on multispectral and temporal features, further boost classification accuracy by minimizing false positives in complex terrains. The inclusion of SEVIRI data improves regional specificity at high temporal resolution. Validation with ground-based measurements demonstrates a significant increase in detection accuracy compared to methods using single-sensor data.

Keywords