A071-06
Application of Machine Learning Techniques for the retrieval of Cloud Properties from Copernicus Satellites Sentinel-4 (S4) and TROPOMI/Sentinel-5 Precursor (S5P)

Wednesday, 9 December 2020: 07:20
Virtual
Fabian Romahn1, Athina Argyrouli2, Ronny Lutz3, Víctor Molina García3 and Diego G Loyola3, (1)German Aerospace Center (DLR), The Remote Sensing Technology Institute (IMF), Oberpfaffenhofen, Germany, (2)Technical University of Munich (TUM), Department of Civil, Geo and Environmental Engineering, Chair of Remote Sensing Technology, Munich, Germany, (3)German Aerospace Center (DLR), Remote Sensing Technology Institute (IMF), Oberpfaffenhofen, Germany
Recorded Presentation
Abstract:
In order to solve the inverse problems that arise in atmospheric remote sensing, usually complex radiative transfer models (RTMs) are used. These are very accurate, however also computationally very expensive and therefore often not feasible in combination with the strict time requirements of operational satellite products, where near-real-time processing is necessary. With the recent significant breakthroughs in machine learning, easier application through better software and more powerful hardware, the methods of this field have become very interesting as a way to improve the classical remote sensing retrieval algorithms.

In this presentation we show a general approach to replace RTMs used in inversion algorithms with artificial neural networks (ANNs) with sufficient accuracy while at the same time increasing the processing performance by several orders of magnitude. The several steps, smart sampling and scaling of the training data, the selection of the ANN architecture and the training itself, are explained in detail. This is then demonstrated at the example of the ROCINN (Retrieval of cloud information using neural networks) algorithm currently used for the operational cloud product of S5P. Furthermore, it is also shown how this approach will be applied for the operational cloud product in the upcoming S4 mission and how the current algorithm for S5P can be improved by adding new physical models (e.g. for ice clouds) in the form of ANNs.

The procedure has been continuously developed and evaluated over time and the most important results, especially in terms of smart sampling, architecture selection and training parameters, are presented.

Finally, the huge performance benefits of using an ANN instead of the original RTM also allow for improvements in the inversion algorithm. Several ideas regarding this, e.g. global optimization techniques, are shown.

<< Previous Abstract | Next Abstract >>