loading page

A neural network approach to polarimetric observations of aerosols above clouds - design, demonstration, and comparison to existing algorithms
  • Daniel Miller,
  • Michal Segal-Rozenhaimer,
  • Kirk Knobelspiesse
Daniel Miller
NASA Goddard Space Flight Center

Corresponding Author:[email protected]

Author Profile
Michal Segal-Rozenhaimer
NASA Ames Research Center
Author Profile
Kirk Knobelspiesse
NASA Goddard Space Flight Center
Author Profile


We present a neural network (NN) based algorithm for the retrieval of cloud and aerosol properties from above cloud aerosol (ACA) scenes. The large state space explored in ACA scenes causes traditional retrieval approaches slow and complicated. This is especially true for optimal inversion retrieval approaches, where a growth in the number of dependent variables can drastically complicate and slow the retrieval search. Our NN retrieval is applied to data from the airborne Research Scanning Polarimeter (RSP), which measures both polarized and total reflectance in the spectral range of 410 to 2260 nm, scanning along the flight track at ~150 viewing zenith angles spanning the angular range between -60˚ to 60˚. We apply this algorithm to field campaign data from the ObseRvations of Aerosols above CLouds and their intEractionS (ORACLES) 2016 and 2017 campaigns and compare to results obtained from other algorithms.