Optimal sparse-particle representations for efficient modeling of aerosols and aerosol-cloud interactions

Description

Sparse-particle aerosol models are an attractive approach for representation of complex, generally mixed particle populations. In the quadrature method of moments (QMOM) a small set of abscissas and weights, determined from distributional moments, provides the sparse set. New applications of linear and nonlinear optimization methods yield a generalization of the QMOM that is especially useful for sparse particle selection. Here we illustrate the new approach by obtaining rigorous and nested upper and lower bounds to aerosol optical properties and CCN activity in terms of prescribed Bayesian-like sequences of model and/or simulated measurement constraints. Examples of measurement constraints include remotely-sensed light extinction at different wavelengths, measurements particulate mass and composition either in bulk or on a particle by particle basis, etc. Successive reduction in bound separation with each new constraint gives an information-based measure of the value of each such measurement or model constraint. In addition to univariate populations, the present study looks towards the development of a new aerosol microphysics algorithm for tracking the properties of multivariate particle populations with sufficient computational efficiency for large-scale simulation. The new approach is illustrated through sparse representation of bivariate distributions with respect to particle dry diameter and hygroscopicity. CCN concentrations from the sparse representations are benchmarked against particle-resolved model output from PartMC-MOSAIC.