AI-Driven Estimation of Aerosol Optical Thickness using Remote Sensing and Meteorological Data

Abstract

Accurately estimating Aerosol Optical Thickness (AOT) is essential for understanding atmospheric aerosol dynamics and their climatic effects. In this study, a machine learning model was developed using the Random Forest (RF) algorithm, to estimate AOT at 500 nm. The reference AOT data are obtained from the AERONET network, using data from the Évora Atmospheric Sciences Observatory (Portugal), ensuring high-quality training and validation. The RF model, configured with 160 estimators and a maximum depth of 35, was trained using a dataset of global horizontal radiation, solar zenith angle, precipitable water vapor and meteorological variables. The model achieves a root mean square error (RMSE) of 0.0594, demonstrating its effectiveness in capturing AOT variability. This AI-driven approach offers a promising tool for estimating AOT in regions with limited direct aerosol observations, thereby enhancing atmospheric monitoring and climate research.