Connecting Laboratory Experiments and In-Situ Observations of Depositional Ice Growth

Abstract

Motivation: Depositional ice growth is an important microphysical process impacting ice cloud formation. The habits and growth rates of ice crystals are known to sensitively depend on temperature (T), pressure, and supersaturation. However, the surface effects controlling vapor deposition are complex, and challenging to characterize experimentally. Recent experiments indicate these surface effects may depend on the ice nucleation mechanism, suggesting depositional ice growth cannot be understood purely as phase partitioning between vapor and ice, but should instead be understood as an aerosol-cloud interaction, fundamentally linked to the type and availability of aerosols acting as ice nucleating particles. This lack of clear physical understanding makes it challenging to predict ice crystals habits and their  growth rates in weather and climate models. The sensitivity of colder clouds (T < -20 C) to surface effects  is not well understood or constrained because of the paucity of measurements at low temperatures. Recent progress on modeling surface processes on faceted ice has been made through the development of the Diffusion Surface Kinetics Ice Crystal Evolution (DiSKICE) model, which models crystal habits with two semi-axes  and parameterizes surface growth mechanisms, including possible nucleation-dependent mechanisms. A recent  extension of this model to “budding” bullet rosettes allows for more complex crystal habits to be modeled. The extension of the DiSKICE model to more complex crystals has been developed using single-crystal laboratory experiments, which limit constraints on processes to small ice crystals.

Proposed research:

Here we propose using Bayesian parameter estimation and deep learning techniques to put stronger constraints and reduce structural uncertainty in ice deposition models by combining prior laboratory measurements and observations from recent US Department of Energy Atmospheric Radiation Measurement (DOE-ARM) facility supported field campaigns. Single-crystal laboratory and  cloud chamber measurements will be used with Bayesian techniques to estimate uncertain surface parameters and constrain the growth rates of small ice crystals. Using a large data set of laboratory grown crystals will allow us to determine whether the subsequent growth of the crystals depends on how the crystals were nucleated. Machine learning approaches applied to in-situ data of crystal shapes will be used to determine the common habit characteristics needed to inform microphysical models (such as rosette arm aspect ratio, which determines the crystal density).  These common characteristics, in combination with constraints derived from the lab measurements, will be used to inform Bayesian parameter estimation techniques applied to model simulations of observed cirrus.

Expected outcomes:

This research will provide better constrained parameterizations for the vapor growth of small ice crystals at temperatures below -40C, and dependencies on how the crystals nucleate will also be determined. Parameterizations for the combined influences of surface attachment kinetics, nucleation mechanism, and crystal effective density are expected to be produced by this research. Extrapolations of these parameterizations to larger size crystals will be accomplished as well, through the use of machine learning techniques and Bayesian parameter estimation. The parameterizations will be provided as power-laws so that they can easily be incorporated into traditional and particle-property microphysical schemes.