Dimensionality reduction of optical data: application to total ozone column retrieval

Abstract

The new generation of atmospheric composition sensors such as TROPOMI, deliver a great amount of data, which is recognized as Big Data. To process the challenging data volumes of spectral information, fast radiative transfer models (RTMs) are required. However, the bottleneck in remote sensing retrieval problems is the computation of the radiative transfer. Thus, the operational processing of remote sensing data requires high-performance RTMs for simulating spectral radiances (level-1 data). In particular, ozone total column retrieval algorithms use the level-1 data in the Huggins band (325-335 nm). Hence, accurate simulation of this absorption band may require several hundreds of monochromatic computations. However, hyper-spectral input data for RTMs has a redundant information, which can be excluded by using the dimensionality reduction techniques. In addition, there is a strong correlation between the input optical data for RTMs and output radiances. Such statistical dependency can be taken into account for accelerating level-1 data simulations using principal component analysis (PCA), thereby providing the performance enhancement for the whole processing chain. In this paper we analyze the efficiency and potential benefits of the optical data dimensionality reduction schemes for simulating the Huggins band and discuss several modifications of this approach.