Articles | Volume 14, issue 6
https://doi.org/10.5194/amt-14-4083-2021
https://doi.org/10.5194/amt-14-4083-2021
Research article
 | 
04 Jun 2021
Research article |  | 04 Jun 2021

Efficient multi-angle polarimetric inversion of aerosols and ocean color powered by a deep neural network forward model

Meng Gao, Bryan A. Franz, Kirk Knobelspiesse, Peng-Wang Zhai, Vanderlei Martins, Sharon Burton, Brian Cairns, Richard Ferrare, Joel Gales, Otto Hasekamp, Yongxiang Hu, Amir Ibrahim, Brent McBride, Anin Puthukkudy, P. Jeremy Werdell, and Xiaoguang Xu

Data sets

Aerosol Characterization from Polarimeter and Lidar ACEPOL Science Team https://doi.org/10.5067/SUBORBITAL/ACEPOL2017/DATA001

FastMAPOL AirHARP Level 2 data product M. Gao https://data.nasa.gov/Earth-Science/FastMAPOL_ACEPOL_AIRHARP_L2/8b9y-7rgh

Short summary
Multi-angle polarimetric measurements can retrieve accurate aerosol properties over complex atmosphere and ocean systems; however, most retrieval algorithms require high computational costs. We propose a deep neural network (NN) forward model to represent the radiative transfer simulation of coupled atmosphere and ocean systems and then conduct simultaneous aerosol and ocean color retrievals on AirHARP measurements. The computational acceleration is 103 times with CPU or 104 times with GPU.