Search Results
Complex systems approaches for Earth system data analysis
2021, Boers, Niklas, Kurths, Jürgen, Marwan, Norbert
Complex systems can, to a first approximation, be characterized by the fact that their dynamics emerging at the macroscopic level cannot be easily explained from the microscopic dynamics of the individual constituents of the system. This property of complex systems can be identified in virtually all natural systems surrounding us, but also in many social, economic, and technological systems. The defining characteristics of complex systems imply that their dynamics can often only be captured from the analysis of simulated or observed data. Here, we summarize recent advances in nonlinear data analysis of both simulated and real-world complex systems, with a focus on recurrence analysis for the investigation of individual or small sets of time series, and complex networks for the analysis of possibly very large, spatiotemporal datasets. We review and explain the recent success of these two key concepts of complexity science with an emphasis on applications for the analysis of geoscientific and in particular (palaeo-) climate data. In particular, we present several prominent examples where challenging problems in Earth system and climate science have been successfully addressed using recurrence analysis and complex networks. We outline several open questions for future lines of research in the direction of data-based complex system analysis, again with a focus on applications in the Earth sciences, and suggest possible combinations with suitable machine learning approaches. Beyond Earth system analysis, these methods have proven valuable also in many other scientific disciplines, such as neuroscience, physiology, epidemics, or engineering.
Neural partial differential equations for chaotic systems
2021, Gelbrecht, Maximilian, Boers, Niklas, Kurths, Jürgen
When predicting complex systems one typically relies on differential equation which can often be incomplete, missing unknown influences or higher order effects. By augmenting the equations with artificial neural networks we can compensate these deficiencies. We show that this can be used to predict paradigmatic, high-dimensional chaotic partial differential equations even when only short and incomplete datasets are available. The forecast horizon for these high dimensional systems is about an order of magnitude larger than the length of the training data.
Impact of an AMOC weakening on the stability of the southern Amazon rainforest
2021, Ciemer, Catrin, Winkelmann, Ricarda, Kurths, Jürgen, Boers, Niklas
The Atlantic Meridional Overturning Circulation (AMOC) and the Amazon rainforest are potential tipping elements of the Earth system, i.e., they may respond with abrupt and potentially irreversible state transitions to a gradual change in forcing once a critical forcing threshold is crossed. With progressing global warming, it becomes more likely that the Amazon will reach such a critical threshold, due to projected reductions of precipitation in tropical South America, which would in turn trigger vegetation transitions from tropical forest to savanna. At the same time, global warming has likely already contributed to a weakening of the AMOC, which induces changes in tropical Atlantic sea-surface temperature (SST) patterns that in turn affect rainfall patterns in the Amazon. A large-scale decline or even dieback of the Amazon rainforest would imply the loss of the largest terrestrial carbon sink, and thereby have drastic consequences for the global climate. Here, we assess the direct impact of greenhouse gas-driven warming of the tropical Atlantic ocean on Amazon rainfall. In addition, we estimate the effect of an AMOC slowdown or collapse, e. g. induced by freshwater flux into the North Atlantic due to melting of the Greenland Ice Sheet, on Amazon rainfall. In order to provide a clear explanation of the underlying dynamics, we use a simple, but robust mathematical approach (based on the classical Stommel two-box model), ensuring consistency with a comprehensive general circulation model (HadGEM3). We find that these two processes, both caused by global warming, are likely to have competing impacts on the rainfall sum in the Amazon, and hence on the stability of the Amazon rainforest. A future AMOC decline may thus counteract direct global-warming-induced rainfall reductions. Tipping of the AMOC from the strong to the weak mode may therefore have a stabilizing effect on the Amazon rainforest.
Reconstructing complex system dynamics from time series: a method comparison
2020, Hassanibesheli, Forough, Boers, Niklas, Kurths, Jürgen
Modeling complex systems with large numbers of degrees of freedom has become a grand challenge over the past decades. In many situations, only a few variables are actually observed in terms of measured time series, while the majority of variables - which potentially interact with the observed ones - remain hidden. A typical approach is then to focus on the comparably few observed, macroscopic variables, assuming that they determine the key dynamics of the system, while the remaining ones are represented by noise. This naturally leads to an approximate, inverse modeling of such systems in terms of stochastic differential equations (SDEs), with great potential for applications from biology to finance and Earth system dynamics. A well-known approach to retrieve such SDEs from small sets of observed time series is to reconstruct the drift and diffusion terms of a Langevin equation from the data-derived Kramers-Moyal (KM) coefficients. For systems where interactions between the observed and the unobserved variables are crucial, the Mori-Zwanzig formalism (MZ) allows to derive generalized Langevin equations that contain non-Markovian terms representing these interactions. In a similar spirit, the empirical model reduction (EMR) approach has more recently been introduced. In this work we attempt to reconstruct the dynamical equations of motion of both synthetical and real-world processes, by comparing these three approaches in terms of their capability to reconstruct the dynamics and statistics of the underlying systems. Through rigorous investigation of several synthetical and real-world systems, we confirm that the performance of the three methods strongly depends on the intrinsic dynamics of the system at hand. For instance, statistical properties of systems exhibiting weak history-dependence but strong state-dependence of the noise forcing, can be approximated better by the KM method than by the MZ and EMR approaches. In such situations, the KM method is of a considerable advantage since it can directly approximate the state-dependent noise. However, limitations of the KM approximation arise in cases where non-Markovian effects are crucial in the dynamics of the system. In these situations, our numerical results indicate that methods that take into account interactions between observed and unobserved variables in terms of non-Markovian closure terms (i.e., the MZ and EMR approaches), perform comparatively better. © 2020 The Author(s). Published by IOP Publishing Ltd on behalf of the Institute of Physics and Deutsche Physikalische Gesellschaft
Analysis of a bistable climate toy model with physics-based machine learning methods
2021, Gelbrecht, Maximilian, Lucarini, Valerio, Boers, Niklas, Kurths, Jürgen
We propose a comprehensive framework able to address both the predictability of the first and of the second kind for high-dimensional chaotic models. For this purpose, we analyse the properties of a newly introduced multistable climate toy model constructed by coupling the Lorenz ’96 model with a zero-dimensional energy balance model. First, the attractors of the system are identified with Monte Carlo Basin Bifurcation Analysis. Additionally, we are able to detect the Melancholia state separating the two attractors. Then, Neural Ordinary Differential Equations are applied to predict the future state of the system in both of the identified attractors.