Sparse sets of weighted particle (stems) accurately represent realistically complex distributions simulated by a particle-resolved model (color densities). Distributions with respect to dry diameter and hygroscopicity are a projection of the size-composition distributions simulated by PartMC-MOSAIC. Whereas the particle-resolved model simulates 10,000 – 100,000 particles, the new sparse-particle framework represents CCN activation of these complex distributions using only 8 sparse particles.

Sparse sets of weighted particle (stems) accurately represent realistically complex distributions simulated by a particle-resolved model (color densities). Distributions with respect to dry diameter and hygroscopicity are a projection of the size-composition distributions simulated by PartMC-MOSAIC. Whereas the particle-resolved model simulates 10,000 – 100,000 particles, the new sparse-particle framework represents CCN activation of these complex distributions using only 8 sparse particles.

Science

A key challenge in simulation of aerosol interactions with clouds is capturing processes and properties across multiple scales. Aerosol impacts on clouds depend on particle-level variation in size and composition, but this small-scale complexity is not easily captured in large-scale atmospheric models. Existing aerosol schemes in large-scale models simplify the representation of the aerosol mixing state, leading to error in the representation of aerosol effects on clouds and radiation. In their 2017 paper, Fierce and McGraw introduce a new framework for representing multivariate aerosol size-composition distributions, which captures the multivariate complexity of aerosol size-composition distributions using only a small set of weighted particles.

Impact

The study is a first step toward a new paradigm in aerosol simulation that will enable large-scale models to accurately and efficiently represent key features of multivariate aerosol distributions. The new framework replaces complex multivariate aerosol distribution with a sparse set of representative particles. Whereas existing aerosol schemes are either too simple to accurately represent climate-relevant aerosol properties or too complex for large-scale simulation, the new sparse-particle representation will enable accurate simulation of particle-level properties in large-scale atmospheric models with minimal computational cost.

Summary

Fierce and McGraw (2017) describe a new technique for constructing sparse representations of realistically complex aerosol populations from distribution moments. The study shows that cloud condensation nuclei activity simulated by a particle-resolved model, which tracks tens to hundreds of thousands of computational particles, is accurately represented using only a few sparse particles. This sparse representation of the aerosol mixing state, designed for use in quadrature-based moment models, is constructed from a linear program constrained by low-order moments and combined with an entropy-inspired cost function. The critical supersaturation at which each sparse particle becomes CCN-active is computed as a function of its size and composition. Constrained maximum entropy distributions are used to construct continuous CCN activation spectra from the sparse set of critical supersaturation values. Unlike reduced representations of the aerosol mixing state that are commonly used in large-scale atmospheric models, such as modal and sectional schemes, the approach described here is not confined to pre-determined size bins or assumed distribution shapes. This study is a first step toward a quadrature-based aerosol scheme that will track multivariate aerosol distributions with both reliable accuracy and sufficient computational efficiency for large-scale simulations.