Toward a better ice cloud product for the next generation spaceborne radiometers

A handful of new generation spaceborne radiometers (e.g., NASA's AOS radiometers, Earth Venture Instrument PolSIR, and ESA's AWS and ICI missions) will include new polarized sub-millimeter (sub-mm) channels around 325 and/or 664 GHz. Both the new frequencies and the polarization measurements contain ample ice microphysics information. In this presentation, I will present two ongoing efforts leveraging the airborne sub-mm radiometer measurements and artificial intelligence/machine learning (AI/ML) techniques, which pave toward a promising future of bringing new cloud ice retrieval products for the next generation sub-mm and MW radiometers.

The first work is on exploring the new ice microphysics information embedded in the sub-mm polarization measurements. We will use airborne measurements from two sub-mm radiometers (Met Office's ISMAR and NASA's CoSSIR) and collocated microphysics "truth" measurements to demonstrate the sensitivity of 325 and 664 GHz polarization difference to differentiate supercooled liquid, to identify the aspect ratio of ice particles, and to even further infer the turbulence condition within the cloud layer.

The second work is on exploring the ice water content (IWC) profiling capability using the existing spaceborne radiometer channels. With collocated ATMS-CloudSat measurements, a multi-task ML retrieval framework is designed to retrieve ice cloud flag and IWC simultaneously and consistently. The ML loss function is customized to prioritize cloud vertical continuity, cloud top height (important radiatively) and cloud thickness (important hydrologically). The new framework is demonstrated its robustness through some preliminary comparison with campaign measurements.