Manuscript received October 30, 2024; accepted February15, 2025; published October 27, 2025
Abstract—Aerosol Optical Depth (AOD) is a key parameter
in atmospheric science. Accurate AOD prediction plays a
crucial role for environmental management, air quality
assessment, and understanding the earth climate change. It also
serves as a direct indicator of particulate matter pollution,
which poses significant health risks. This study addresses the
need for precise AOD forecasting by implementing SwinLSTM,
a multi-step spatiotemporal deep learning model. A 6-hour
forecasting horizon was selected to align with practical
applications, with lookback periods of 12 and 24 h. Both
lookback periods yielded comparable model performance, as
evidenced by the following metrics: Mean Absolute Error of
0.04, Root Mean Square Error of 0.08, Structural Similarity
Index of 0.97, and Peak Signal-to-Noise Ratio ranging from 44.4
to 44.95. By enhancing our understanding of AOD and its
contributing factors, we can develop more effective strategies to
mitigate the negative impacts of air pollution and protect
human health and the environment. Additionally, AOD
forecasting can aid in understanding the impact of atmospheric
particles on the Earth’s climate.
Keywords—aerosol optical depth prediction, air pollution,
climate change, spatiotemporal deep learning, SwinLSTM
model
Cite: Chaiyo Churngam, Veerasit Kaewbundit, and Papis Wongchaisuwat, "A Spatiotemporal Aerosol Optical Depth Forecasting in Thailand Using Deep Learning," International Journal of Machine Learning vol. 15, no. 4, pp. 78-82, 2025.
Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).