High-Resolution Daily PM2.5 Dataset for the Contiguous US (2005–2021)

Overview This dataset provides high-resolution (1 km) daily surface-level PM2.5 concentration estimates for the contiguous United States from 2005 to 2021. The data is organized into annual archives containing the monthly files. Methodology The estimates were generated using a Bidirectional Long Short-Term Memory network with an Attention mechanism. To improve estimation precision, particularly on days with extreme pollution events, the deep learning model integrates multiple data sources. These sources include in situ measurements, remotely sensed data, and wildfire smoke density observations. By leveraging the temporal dynamics of air pollution, this model demonstrates a significant improvement on high-concentration days compared to existing products. Associated Publication & Citation If you use this dataset in your research, please cite the corresponding methodology paper: Wang, Z., Crooks, J. L., Regan, E. A., & Karimzadeh, M. (2025). High-Resolution Estimation of Daily PM2.5 Levels in the Contiguous US Using Bi-LSTM with Attention. Remote Sensing, 17(1), 126. https://doi.org/10.3390/rs17010126