Sensitivity, Uncertainty, Information Content, and Channel Selection for Hyperspectral Reflective Solar Retrievals of Cirrus Cloud Properties

Inferring cirrus cloud properties from spaceborne observations inevitably involves uncertainties from both forward models and measurements. These uncertainties are present in the retrieval products and propagate to downstream data-ingesting applications in atmospheric and climate modeling. Satellite reflective solar hyperspectral instruments are highly information dense and offer a means to reduce uncertainties in cirrus cloud property retrieval products. However, hyperspectral instruments contain a large number of channels imposing large computational expenses both with measurement data and radiative transfer modeling needed to perform cirrus cloud property retrievals. While using all the channels will maximize the information content, the computational expense may not be acceptable. The purpose of this study is to objectively identify channel subsets in the spectral coverage of the Climate Absolute Radiance and Refractivity Observatory Pathfinder (CLARREO-PF) instrument that, when used as reflectance inputs to a cirrus radiative and microphysical property retrieval, will minimize uncertainties in the retrieval products. An ensemble of 216 cloud/atmosphere cases are used with an information content iterative routine and a clustering algorithm. Water vapor channels are identified as highly informative for cirrus cloud optical depth and cloud top pressure, away from window regions normally used in reflective solar ice cloud property retrievals. The ice-absorbing channels are most informative for effective particle size. The extractable independent information content increases with the increasing number of CLARREO-PF channels that are added to the retrieval system. However, with each channel added to the retrieval selection, the independent information content of all the remaining unselected channels decreases.