Satellite retrieval of aerosol microphysical and optical parameters using neural networks: a new methodology applied to the Sahara desert dust peak

• Published in: Atmospheric Measurement Techniques
• Main Authors: M. Taylor, S. Kazadzis, A. Tsekeri, A. Gkikas, V. Amiridis
• Format: Article in Journal/Newspaper
• Language: English
• Published: Copernicus Publications 2014
• Subjects: Environmental engineering; TA170-171; Earthwork. Foundations; TA715-787; Aerosol Robotic Network
• Online Access: https://doi.org/10.5194/amt-7-3151-2014; https://doaj.org/article/8a42c0b236ac465d98bd7b897136fec4

In order to exploit the full-earth viewing potential of satellite instruments to globally characterise aerosols, new algorithms are required to deduce key microphysical parameters like the particle size distribution and optical parameters associated with scattering and absorption from space remote sensing data. Here, a methodology based on neural networks is developed to retrieve such parameters from satellite inputs and to validate them with ground-based remote sensing data. For key combinations of input variables available from the MODerate resolution Imaging Spectro-radiometer (MODIS) and the Ozone Measuring Instrument (OMI) Level 3 data sets, a grid of 100 feed-forward neural network architectures is produced, each having a different number of neurons and training proportion. The networks are trained with principal components accounting for 98% of the variance of the inputs together with principal components formed from 38 AErosol RObotic NETwork (AERONET) Level 2.0 (Version 2) retrieved parameters as outputs. Daily averaged, co-located and synchronous data drawn from a cluster of AERONET sites centred on the peak of dust extinction in Northern Africa is used for network training and validation, and the optimal network architecture for each input parameter combination is identified with reference to the lowest mean squared error. The trained networks are then fed with unseen data at the coastal dust site Dakar to test their simulation performance. A neural network (NN), trained with co-located and synchronous satellite inputs comprising three aerosol optical depth measurements at 470, 550 and 660 nm, plus the columnar water vapour (from MODIS) and the modelled absorption aerosol optical depth at 500 nm (from OMI), was able to simultaneously retrieve the daily averaged size distribution, the coarse mode volume, the imaginary part of the complex refractive index, and the spectral single scattering albedo – with moderate precision: correlation coefficients in the range 0.368 ≤ R ≤ 0.514. The network failed to ...