Lightweight algorithm for the rapid prediction of tri-axial ellipsoidal dust aerosol optical properties

Authors:
Publisher website
Google Scholar
Abstract
The single-scattering properties of dust aerosols form the fundamental basis for atmospheric radiative transfer simulations and climate model construction. However, the actual tri-axial ellipsoidal morphology of dust deviates from traditional approximations. Furthermore, the low computational efficiency of traditional electromagnetic scattering calculations and the high storage costs of optical property databases severely constrain their engineering applications. In this study, a prediction model for the optical properties of tri-axial ellipsoidal dust aerosols based on the extreme gradient boosting algorithm is proposed. Utilizing single-scattering property data of tri-axial ellipsoidal dust calculated via traditional methods as a benchmark, this model achieves the rapid prediction of optical parameters, including the extinction efficiency factor, scattering efficiency factor, single scattering albedo, and asymmetry factor. The results demonstrate that the coefficients of determination for the four optical properties are larger than 0.998, with root mean square error values below 0.05. The prediction errors of the model exhibit a significant scale dependence, where the errors for large size parameters X are substantially lower than those for small to medium size parameters X . This pattern is highly consistent with the physical mechanism wherein the influence of non-spherical details on macroscopic optical properties weakens during the scattering process of large particles, and the dependence of optical properties on size tends to flatten. Simultaneously, the model demonstrates robust extrapolation capabilities for combinations of size parameters and axis ratios not covered by the training database, yielding error distributions that closely approximate those of the covered parameters. By substituting the original 8.52 GB optical property database with a mere 916 KB model file, this study not only breaks through the efficiency and storage bottlenecks of traditional methods but also provides a highly efficient and accurate optical property calculation scheme for applications such as atmospheric radiative transfer models and aerosol remote sensing retrievals. This is of significant importance for improving the estimation accuracy of the climate effects of dust aerosols.
Funding Information
  • Cultivation Project of Major Research Program of the National Natural Science Foundation of China (92052106)
  • National Natural Science Foundation of China (61771385)
  • Innovation Team of Higher Education Institutions in Shaanxi Province, the Key Core Technology Tackling Project of Key Industrial Chain of Xi’an Municipality (103-433023062)
  • Key Research and Development Program of Xianyang Municipality (L2023- ZDYF-QYCX-025)
  • Key Laboratory Fund of Visible Optical Communication in Henan Province (HKLVLC2023-B05)

This publication has 39 references indexed in Scilit: