K-nearest neighbors approximation for the multiwavelength lidar retrieval of bimodal aerosol microphysical properties

Le résumé fourni par la source

Aerosol particle size distribution and complex refractive index are key microphysical parameters for assessing radiative and climate effects. However, bimodal aerosol systems with fine and coarse modes present strong ill-posedness and high computational demands, limiting the performance of conventional lookup-table approaches. We propose a prior-constrained K-nearest neighbors (KNN) approximation algorithm that estimates the modal contributions from multiwavelength extinction ratios, enabling optical property decomposition and independent retrieval of microphysical parameters for each mode. Simulations demonstrate that this method improves the accuracy of median radius and complex refractive index compared with unimodal lookup tables, while more effectively capturing the heterogeneity of bimodal aerosols. Under realistic Gaussian noise, retrievals remain robust, with 75th-percentile uncertainties of median radius and logarithmic standard deviation within 14% and 12%. These results indicate that the prior-constrained KNN approach effectively mitigates the ill-posedness inherent in bimodal inversions and offers a feasible approach for efficient, quantitative aerosol characterization.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.