Retrieval of Vertical Cloud Droplet Profiles and Above-Cloud Integrated Water Vapor from Hyperspectral Measurements: Reducing Liquid Water Path Retrieval Bias with Application to EMIT

Accurate liquid water path estimates derived from backscattered solar radiation require knowledge of the vertical structure of cloud droplet effective radius, yet standard bispectral retrievals assume a vertically homogeneous cloud and overestimate liquid water path by up to 45% compared to in situ measurements. We developed a Gauss–Newton optimal estimation retrieval that simultaneously estimates vertical profiles of cloud droplet effective radius, cloud optical thickness, and above-cloud integrated water vapor from hyperspectral solar backscatter measurements in the visible and shortwave infrared. The retrieval solves for effective radius at cloud top and base, cloud optical thickness, and above-cloud integrated water vapor in logarithmic space, using an a priori covariance matrix with off-diagonal elements derived from VOCALS-REx in situ measurements, and incorporating forward model uncertainty with a forward model Jacobian. Tested on 69 simulated HySICS reflectance spectra constructed from in situ cloud microphysics, the hyperspectral retrieval reduces the average liquid water path error to 26.9%, compared to 45.2% for the standard bispectral method. Applied to 3695 EMIT hyperspectral measurements over the southeast Pacific, MODIS-retrieved liquid water path exceeds the hyperspectral estimate by 19% on average. These results demonstrate that simultaneous retrieval of the integrated water vapor above-cloud is necessary for accurate droplet profile retrievals, and that the upcoming CLARREO Pathfinder instrument, with its 0.3% radiometric uncertainty, should enable routine vertical profiling of cloud droplet size.