The research group on Climate Informatics develops innovative data science methods to advance our understanding of the climate system and address climate change topics of critical importance for the society.
The Earth system is one of the best-observed complex dynamical systems with satellite observations and weather stations providing almost global coverage for the past decades to centuries. These datasets are complemented by the output of global climate models that simulate the basic physical laws underlying the atmosphere and oceans and can give us projections of how future climate looks like - given different scenarios of anthropogenic influence.
Yet, our tools to analyze and understand Earth system data are still in their infancy due to several challenges. Firstly, the climate system is nonlinear and processes interact on vastly different time scales. Furthermore, the measured datasets often do not represent all relevant processes and datasets contain missing values and uncertainties. Last, the climate system is very high-dimensional and datasets are easily on the Tera- and Petabyte scale.p>
The goal and mission of the Climate Informatics group is to develop data science methods that handle these challenges based on causal inference, machine learning and in particular deep learning, and nonlinear time series analysis.