The Information Content in Cloud Thermodynamic Phase from Shortwave-only Reflectance Measurements by MODIS, VIIRS, and PACE

Information content methodologies quantify the information inherent in a measurement and can be especially useful when distinct physical properties are not uniquely discriminated by the features of the measurement. Challenges to attribution can arise due to interference from competing signals in time, space, or wavelength, from instrument effects, or from intermittent sampling of naturally varying fields. Such is often times the case with the retrieval of thermodynamic phase from remotely sensed cloud observations, where passive measurements in one or more discrete spectral channels identify water absorbing radiation differently, but not uniquely, for liquid and ice phases. The challenge for cloud remote sensing is multiplied because useful retrievals of cloud optical thickness and effective particle size first require accurate phase retrievals.

Most of the satellite imagers contributing to global cloud records, such as the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Visible Infrared Imaging Radiometer Suite (VIIRS), combine infrared channels along with visible and near-infrared channels to derive cloud properties. A future imager, the Ocean Color Instrument (OCI) of the Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission will also generate global cloud properties but it will not have infrared channels. Instead, OCI will have two channels in the 2 μm window (vs. single channels for MODIS and VIIRS) where water absorption is relatively strong and particle size retrievals are most sensitive.

In this work, we quantitatively assess the probability of correct cloud thermodynamic phase discrimination using shortwave-only channels specific to the MODIS, VIIRS, and future PACE/OCI imagers and evaluate whether two channels near 2 μm provide more information about phase than a single channel. To achieve our goal, we apply the GEneralized Nonlinear Retrieval Analysis (GENRA) approach, which utilizes the Shannon information content and maximum posteriori estimate as diagnostics to quantify the information given a measurement with its associated uncertainty and a set of simulated solutions from cloud radiation models with their own set of associated uncertainties. Then, by enhancing the GENRA capabilities to mutual and conditional information content metrics, we quantify how, and to what degree, cloud radiation models show a given reflectance measurement providing information about both cloud optical thickness and effective particle size. The combination of Shannon, conditional and mutual information contents offer future directions for quantifying the information in a measurement from single or multiple platforms and for visualizing dependencies between physical parameters such as cloud or aerosol optical properties.