Atmospheric Water Cycle and Remote Sensing

General Introduction

The Institute for Geophysics and Meteorology at the University of Cologne is involved in geophysical exploration (from solar system, extrasolar planets, and moons) to our earth system’s atmosphere (weather and climate). Here in our research group, it’s all about Atmospheric Water Cycle and Remote Sensing (AWARES).

AWARES research activities

Our present group consists a combined strength of 21 researchers and supporting staff. Our interests divide us into four active subgroups –

  1. Arctic subgroup where researchers gain insight into water vapor, clouds, precipitation, and radiative effects in the Arctic. They exploit the ground-based and remote sensing observations from long-term measurements and campaigns such as Ny-Ålesund and MOSAiC.
  2. Eurec4a or the tropical cloud subgroup are researching cloud dynamics using the unique multiscale dataset collected during the EUREC4A campaign.
  3. Satellite innovations subgroup involves extracting relevant information from advanced generation satellites using artificial intelligence and deep learning techniques.
  4. The airborne subgroup is interested in taking a closer look at the clouds, precipitation using instruments such as microwave radar, Aerosol Lidar onboard the Polar 5 research aircraft.

AWARES – KISTE Overlap of Interest

Our group is involved with the KI:STE consortium in terms of satellite innovations for clouds and solar energy. We are currently trying to understand how we can smartly use deep learning algorithms to understand cloud distribution and patterns for the use of solar energy. We are interested in self-supervised learning and gaining physical insights from the novel high spatiotemporal Meteosat third-generation satellite data.

What makes our task more challenging

If clouds are labelled by a person then we are bringing the biasness inherently into the learning stage of AI. In Earth science domain we want to understand how objects are interacting or how physics plays a role into the task. Therefore, in order to understand the physical information from satellite data such as ‘clouds’ we want the AI to have a more deeper sense of understanding beyond what is specified in manually labelled dataset.

by Dwaipayan Chatterjee