On estimation of cloudiness characteristics and parameters of DOAS retrieval from spectral measurements using a neural network

Abstract

Light scattering by clouds significantly affects the values associated with the content of NO2, H2CO and other small gases in the lower troposphere, which are obtained by the differential optical absorption spectroscopy (DOAS) technique. Since there are a large databases of optical observations of trace gases by DOAS technique that are not accompanied by other measurements of clouds, the development of approaches to the refinement of scattering characteristics and coefficients linking the DOAS slant column depth with the gas vertical content directly from spectral measurements remains an important task. The paper considers the tasks of determining the coefficient F used for transformation of the DOAS slant column depth of a gas to its vertical column from quantitates obtained from ZDOAS measurements (the O4 slant column, the color index, the absolute intensity, etc.). It was shown in numerical experiments that an algorithm based on a neural network can estimate the coefficient F in cloudy conditions. It looks like the better approach that two step estimation of this parameter using a neural network for estimation of cloud characteristics in the first step with the following radiative transfer simulation at the second step.