A5076653467- Climate variability and models
- Bayesian Modeling and Causal Inference
- Meteorological Phenomena and Simulations
- Atmospheric and Environmental Gas Dynamics
- Complex Systems and Time Series Analysis
- Time Series Analysis and Forecasting
- Functional Brain Connectivity Studies
- Geochemistry and Geologic Mapping
- Cognitive Science and Mapping
- Neural Networks and Applications
- Complex Network Analysis Techniques
- Mental Health Research Topics
- Atmospheric chemistry and aerosols
- Marine and coastal ecosystems
- Oceanographic and Atmospheric Processes
- Ionosphere and magnetosphere dynamics
- Statistical Methods and Inference
- Atmospheric aerosols and clouds
- Arctic and Antarctic ice dynamics
- Fault Detection and Control Systems
- Opinion Dynamics and Social Influence
- Advanced Graph Neural Networks
- Geomagnetism and Paleomagnetism Studies
- Advanced Computational Techniques and Applications
- Remote Sensing in Agriculture
Deutsches Zentrum für Luft- und Raumfahrt e. V. (DLR)
2017-2025
Technische Universität Berlin
2021-2025
TU Dresden
2024-2025
Center for Scalable Data Analytics and Artificial Intelligence
2024
Potsdam Institute for Climate Impact Research
2011-2020
Imperial College London
2019-2020
Vrije Universiteit Amsterdam
2020
University of Amsterdam
2020
Hochschule Magdeburg-Stendal
2020
Deltares
2020
A novel causal discovery method for estimating nonlinear interdependency networks from large time series datasets.
Causal network reconstruction from time series is an emerging topic in many fields of science. Beyond inferring directionality between two series, the goal causal or discovery to distinguish direct indirect dependencies and common drivers among multiple series. Here, problem networks including lags multivariate recapitulated underlying assumptions practical estimation problems. Each aspect illustrated with simple examples unobserved variables, sampling issues, determinism, stationarity,...
Multivariate transfer entropy (TE) is a model-free approach to detect causalities in multivariate time series. It able distinguish direct from indirect causality and common drivers without assuming any underlying model. But despite these advantages it has mostly been applied bivariate setting as hard estimate reliably high dimensions since its definition involves infinite vectors. To overcome this limitation, we propose embed TE into the framework of graphical models present formula that...
Abstract In recent years, the Northern Hemisphere midlatitudes have suffered from severe winters like extreme 2012/13 winter in eastern United States. These cold spells were linked to a meandering upper-tropospheric jet stream pattern and negative Arctic Oscillation index (AO). However, nature of drivers behind these circulation patterns remains controversial. Various studies proposed different mechanisms related changes Arctic, most them reduction sea ice concentrations or increasing...
Abstract Identifying regions important for spreading and mediating perturbations is crucial to assess the susceptibilities of spatio-temporal complex systems such as Earth’s climate volcanic eruptions, extreme events or geoengineering. Here a data-driven approach introduced based on dimension reduction, causal reconstruction, novel network measures effect theory that go beyond standard tools by distinguishing direct from indirect pathways. Applied data set atmospheric dynamics, method...
Abstract Global climate models are central tools for understanding past and future change. The assessment of model skill, in turn, can benefit from modern data science approaches. Here we apply causal discovery algorithms to sea level pressure a large set simulations and, as proxy observations, meteorological reanalyses. We demonstrate how the resulting networks (fingerprints) offer an objective pathway process-oriented evaluation. Models with fingerprints closer observations better...
Physics is a field of science that has traditionally used the scientific method to answer questions about why natural phenomena occur and make testable models explain phenomena. Discovering equations, laws, principles are invariant, robust, causal been fundamental in physical sciences throughout centuries. Discoveries emerge from observing world and, when possible, performing interventions on system under study. With advent big data data-driven methods, fields equation discovery have...
Abstract Lagged cross-correlation and regression analysis are commonly used to gain insights into interaction mechanisms between climatological processes, in particular assess time delays quantify the strength of a mechanism. Exemplified on temperature anomalies Europe tropical Pacific Atlantic, authors study lagged correlation regressions analytically for simple model system. A strong dependence influence serial dependencies or autocorrelation is demonstrated, which can lead misleading...
While it is an important problem to identify the existence of causal associations between two components a multivariate time series, topic addressed in Runge, Heitzig, Petoukhov, and Kurths [Phys. Rev. Lett. 108, 258701 (2012)], even more assess strength their association meaningful way. In present article we focus on defining coupling using information-theoretic measures demonstrate shortcomings well-known mutual information transfer entropy. Instead, propose certain time-delayed...
Abstract The stratospheric polar vortex can influence the tropospheric circulation and thereby winter weather in mid-latitudes. Weak states, often associated with sudden warmings (SSW), have been shown to increase risk of cold-spells especially over Eurasia, but its role for North American winters is less clear. Using cluster analysis, we show that there are two dominant patterns increased cap heights lower stratosphere. Both represent a weak they different wave mechanisms regional impacts....
Across geosciences, many investigated phenomena relate to specific complex systems consisting of intricately intertwined interacting subsystems. Such dynamical can be represented by a directed graph, where each link denotes an existence causal relation, or information exchange between the nodes. For geophysical such as global climate, these relations are commonly not theoretically known but estimated from recorded data using causality analysis methods. These include bivariate nonlinear...
This review provides a summary of methods originated in (non-equilibrium) statistical mechanics and information theory, which have recently found successful applications to quantitatively studying complexity various components the complex system Earth. Specifically, we discuss two classes methods: (i) entropies different kinds (e.g., on one hand classical Shannon R´enyi entropies, as well non-extensive Tsallis entropy based symbolic dynamics techniques and, other hand, approximate entropy,...
Network theory, as emerging from complex systems science, can provide critical predictive power for mitigating the global warming crisis and other societal challenges. Here we discuss main differences of this approach to classical numerical modeling highlight several cases where network substantially improved prediction high-impact phenomena: 1) El Niño events, 2) droughts in central Amazon, 3) extreme rainfall eastern Central Andes, 4) Indian summer monsoon, 5) stratospheric polar vortex...
Abstract Climate models are essential to understand and project climate change, yet long‐standing biases uncertainties in their projections remain. This is largely associated with the representation of subgrid‐scale processes, particularly clouds convection. Deep learning can learn these processes from computationally expensive storm‐resolving while retaining many features at a fraction computational cost. Yet, simulations embedded neural network parameterizations still challenging highly...
We propose a method to analyze couplings between two simultaneously measured time series. Our approach is based on conditional mutual sorting information. It related other concepts for detecting coupling directions: the old idea of Marko directed information and more recent concept Schreiber's transfer entropy. By setting suitable conditions we first all consider momentary in both This enables detection not only directions but also delays. Sorting refers ordinal properties series, which...
Complex network theory provides a powerful toolbox for studying the structure of statistical interrelationships between multiple time series in various scientific disciplines. In this work, we apply recently proposed climate approach characterizing evolving correlation Earth's system based on reanalysis data surface air temperatures. We provide detailed study temporal variability several global characteristics. Based simple conceptual view red networks (i.e., with comparably low number...
Abstract Variability in the stratospheric polar vortex (SPV) can influence tropospheric circulation and thereby winter weather. Early predictions of extreme SPV states are thus important to improve forecasts weather including cold spells. However, dynamical models usually restricted lead time because they poorly capture low‐frequency processes. Empirical often suffer from overfitting problems as relevant physical processes lags not well understood. Here we introduce a novel empirical...
Conditional independence testing is a fundamental problem underlying causal discovery and particularly challenging task in the presence of nonlinear high-dimensional dependencies. Here fully non-parametric test for continuous data based on conditional mutual information combined with local permutation scheme presented. Through nearest neighbor approach, efficiently adapts also to non-smooth distributions due strongly Numerical experiments demonstrate that reliably simulates null distribution...
We introduce the \texttt{pyunicorn} (Pythonic unified complex network and recurrence analysis toolbox) open source software package for applying combining modern methods of data modeling from theory nonlinear time series analysis. is a fully object-oriented easily parallelizable written in language Python. It allows construction functional networks such as climate climatology or brain neuroscience representing structure statistical interrelationships large sets and, subsequently,...
Measures of information transfer have become a popular approach to analyze interactions in complex systems such as the Earth or human brain from measured time series. Recent work has focused on causal definitions aimed at decompositions predictive about target variable, while excluding effects common drivers and indirect influences. While clearly constitute spurious causality, aim present article is develop measures quantifying different notions strength along paths, based first...
Abstract. The dynamics of biochemical processes in terrestrial ecosystems are tightly coupled to local meteorological conditions. Understanding these interactions is an essential prerequisite for predicting, e.g. the response carbon cycle climate change. However, many empirical studies this field rely on correlative approaches and only very few apply causal discovery methods. Here we explore potential a recently proposed graph algorithm reconstruct dependency structure underlying...