Dust storms are common in the Middle East and can seriously affect people’s health. However, it is difficult to measure exactly how much dust people are exposed to because there are very few ground-based sensors in the region. To fill this gap, we use satellite-based data from NASA’s MERRA-2, which estimates dust in the atmosphere over time. But MERRA-2 data is spatially coarse, with each data point covering an area the size of a small city. This makes it difficult to use for local health studies. In our study, we used AI to take the coarse MERRA-2 data and join it with a higher resolution NASA model called G5NR. This process, known as probabilistic diffusion model downscaling, helped us create much finer dust maps at a 7 km resolution.

We then linked these refined dust estimates to over 20 years of hospital data in Kuwait, and also studied exposure among U.S. Veterans who served in the Middle East. Our approach shows how combining NASA data and AI can help protect public health in dusty, data-poor regions.