Jacobians of single-scattering optical properties of super-spheroids computed using neural networks

Abstract

In atmospheric aerosol remote sensing and data assimilation studies, the Jacobians of the optical properties of non-spherical aerosol particles are required. Specifically, the partial derivatives of the extinction efficiency factor, single-scattering albedo, asymmetry factor, and scattering matrix should be obtained with respect to microphysical parameters, such as complex refractive indices, shape parameters and size parameters. When a look-up table (LUT) of optical properties of particles is available, the Jacobians traditionally can be calculated using the finite difference method (FDM), but the accuracy of the process depends on the resolution of microphysical parameters. In this paper, a deep learning scheme was proposed for computing Jacobians of the optical properties of super-spheroids, which is a flexible model of non-spherical atmospheric particles. Using the neural networks (NN), the error of the Jacobians in the FDM can be reduced by more than 60%, and the error reduction rate of the Jacobians of the scattering matrix elements can be more than 90%. We also tested the efficiency of the NN for computing the Jacobians. The computation takes 30 seconds for one million samples on a host with an NVIDIA GeForce RTX 3070 GPU. The accuracy and efficiency of the present NN scheme proves it is promising for applications in remote sensing and data assimilation studies.

Single‐Scattering Properties of Encapsulated Fractal Black Carbon Particles Computed Using the Invariant Imbedding T‐Matrix Method and Deep Learning Approaches

“…The DNN method has emerged as a more viable solution, offering efficient computations and significantly smaller storage requirements for network parameters than the original database. For example, deep learning methods have already been applied to predict the single‐scattering properties and derivatives of dust aerosols (Chen et al., 2022; Yu et al., 2022). In the case of bare BC aggregates, a support vector machine method has been utilized for direct parameterization, establishing a relationship between optical properties and particle morphology (Luo, Zhang, Wang, et al., 2018).…”

Linearized Single-Scattering Property Database for Hexagonal Prism Ice Particles

“…Scattering for a non-spherical particle of large size can be solved by approximate methods, such as the physical-geometric optics method (PGOM), that require fewer computational resources than rigorous methods [20]. The fundamental integral and differential single scattering properties required for radiative transfer computation can be generated using these algorithms or obtained using deep learning approaches [21,22]. Plenty of single scattering property databases for application in remote sensing and the simulation of radiative transfer have been developed to satisfy spectral and scale continuity, and they employ a variety of algorithms, such as those of the Lorenz-Mie theory [23][24][25][26], the discrete dipole approximation [26][27][28], the T-matrix methods [25,[28][29][30][31][32], the geometric optics method [33][34][35][36][37], the finite-difference time-domain method [35,38], etc.…”