PhD position - Machine learning based assimilation of satellite data to improve air quality predictions (HDS-LEE graduate school)

Remote sensing data from satellites are an extremely valuable information source to improve air quality predictions. They monitor aerosol and trace gases often with global coverage, which is far beyond in-situ observational networks. However, the accuracy of satellite retrievals depends on several assumptions leading to biased observations.

• The aim of the PhD-project is to overcome this limitation in the use of satellite observations by make a direct use of radiance observations retrieved by satellites using machine learning without the need of radiative transfer calculations. The new model will be developed using the example of the novel Earth Explorer EarthCARE and will be integrated as observation operator to the sophisticated data assimilation system of the EURopean Air pollution Dispersion - Inverse Model (EURAD-IM).

The PhD project will be conducted in the atmospheric modeling group of the Institute of Climate and Energy Systems (ICE-3) at Forschungszentrum Jülich and the Meteorological Institute at the University of Cologne. At ICE-3 detailed investigations of the emission of atmospheric constituents are performed using the 4-dimensional variational (4D-var) data assimilation method implemented into the EURAD-IM. In atmospheric chemistry modeling, the 4D-var method is a powerful tool to assess the state of the atmosphere and the corresponding emissions that are in compliance with observations. The AWARES group at the University of Cologne provides extensive expertise in remote sensing and radiative effects in the atmosphere. To include the observations in the assimilation procedure, so called observation operators, mapping the model state to the observation space, are essential.

• Within the project, the direct use of radiance measurements from EarthCARE will be explored by developing a machine learning model that directly maps the radiances on the aerosol properties. Thus, the modelling framework of the EURAD-IM will be enabled to use radiance assimilation for operational services.

The innovative application of machine learning to exploit satellite data within numerical modelling links to other scientific applications beyond those of the atmosphere. The project builds on strong cooperation between ICE-3, the University of Cologne and other partners within and outside Forschungszentrum Jülich, which provides an ideal fundament to combine competences across different disciplines (meteorology, environmental science, high performance computing, software development and data science).

• M. Sc. degree in meteorology, physics, mathematics, or a related field
• Good knowledge in data handling and machine learning
• Good knowledge in software development and data processing and visualization with Python
• Strong interest in atmospheric physics and chemistry
• Outstanding organizational skills and the ability to work independently
• Very good cooperation and communication skills and ability to work as part of a team in an international and interdisciplinary environment

Conducting research for a changing society: This is what drives us at Forschungszentrum Jülich. As a member of the Helmholtz Association, we aim to tackle the grand challenges of our times. How can we make a success of the energy transition and mitigate the effects of climate change? What challenges are emerging due to the increasing digitization of our society? Will we succeed in understanding the human brain? And how can we facilitate the transition to a bio-based sustainable economy? Come and work with us at our scientific institutes, in our technical or administrative infrastructure, or in research management alongside roughly 6,800 colleagues at one of Europe’s biggest research centres and help make a contribution to solving societal challenges. Help us to shape change!