Inversion Model of Nonspherical Aerosol Particle Size Distribution Based on Tikhonov’s Regularization With the PSO-GA Algorithm

Frequent dusty weather has led to an increasing interest in dust aerosols, which were observed by lidar as nonspherical particles. To solve the problem of large errors in the previous inversion results of nonspherical aerosol particle size distribution (APSD), an inversion model of nonspherical APSD is proposed in this article. This inversion model is made up of three steps. First, the discrete dipole approximation (DDA) method was used instead of the Lorenz–Mie (Mie) theory to compute the optical properties of nonspherical particles. Second, the traditional generalized cross-validation (GCV) method was replaced by a hybrid algorithm and combined with Tikhonov’s regularization. The hybrid algorithm combines a particle swarm optimization (PSO) with a genetic algorithm (GA). The problem of the instability of the Lagrange multiplier values is solved. Third, a multi wavelength polarization lidar and an aerodynamic particle sizer (APS) were used to detect dust aerosols in Yinchuan. The nonspherical property of the dust aerosol was verified by the depolarization ratio of the 532-nm polarization channel. Overall, the simulation shows that the hybrid algorithm gives more accurate results and that the real part of the complex refractive index has a greater effect on the nonspherical APSD than the imaginary part of the complex refractive index. There is a clear inverse relationship between the variation of dust aerosol quantities at high altitudes and near the ground. Calculations from the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model revealed the reason for this phenomenon. The model shows that dust aerosol number changes were influenced by updrafts from Russian and Afghan regions and Shaanxi.