Hyperspectral A-band retrievals of cloud droplet number concentration from OCO-2

NASA’s Orbiting Carbon Observatory-2 (OCO-2) includes a hyperspectral (Dl~0.02 nm) oxygen A-band sensor, and the depth of its absorption features is related to the photon path length. Photon path length increases above a cloud if it is lower (i.e. higher P_top), and within a cloud if its droplets are farther apart (i.e. lower N­_d). This is a novel approach for retrieving N_d that is independent of MODIS-like retrievals, which take an a priori vertical cloud structure and assume that non-adiabatic processes such as precipitation or entrainment affect clouds uniformly. Our last product, OCO2CLD-LIDAR-AUX, used CALIPSO P_top to help separate the above- and within-cloud path length. Here we show progress in an updated OCO-2 only retrieval of marine boundary layer clouds, including using neural networks for cloud identification and phase classification, additional retrieval of r_e, and how cloud vertical structure can bias retrieved P_top and N_d. Successfully addressing this bias would provide a new and independent N_d retrieval that should capture changes due to non-adiabatic processes, and therefore provide a new test of aerosol cloud effects.