Characterising Surface Roughness in Ice Clouds Using In-Situ Measurements of Frozen Droplets and Bayesian-Optimised Physical Optics Simulations

Ice crystal surface roughness influences the global shortwave cloud radiative effect by an estimated 1-2 Wm^-2 and affects backscattering properties required for lidar retrievals, yet the microscale structure of atmospheric ice particles remains poorly constrained. In-situ observations indicate that rough and irregular surfaces are common, but insufficient measurements linking particle imagery to angular scattering data limit the development of representative shape models.

During a flight of the CIRRUS-HL campaign in summer 2021, an unusually large proportion of quasi-spherical ice particles resembling frozen droplets were observed by the Particle Habit Imaging and Polar Scattering probe (PHIPS). We use this dataset as a case study to constrain surface roughness in ice clouds. PHIPS provides particle imagery from two viewing angles alongside simultaneous scattering measurements from 18 to 170°. Several thousand single, chain, and aggregated frozen droplets were identified, with mean radius ~14 μm (size parameter X ≈ 200).

We model these particles using the droxtal geometry, and compute scattering properties using a beam tracing physical optics method. Preliminary results indicate pristine droxtals are insufficient to reproduce observed scattering, suggesting that surface roughness cannot be ignored.

The surface roughness implementation is characterised by a mesh edge length and vertex displacement amplitude. Ensembles of roughened droxtals with radii sampled from the measured size distribution are compared against PHIPS measurements using a novel Bayesian optimisation implementation to efficiently explore the 2D roughness parameter space. We present results and discuss implications for constraining ice crystal roughness from in-situ measurements.