Researchers break new ground in predicting climate’s critical shifts

Global warming is happening faster than ever. Current climate models have not been able to predict the rapid climate changes. These models are based on past climate data and are not accurately predicting extreme events like Arctic ice loss, heatwaves, and floods. This project aims to combine physics-based climate models, data analysis, and mathematical techniques to better understand and predict climate tipping points. Tipping points are critical thresholds where small changes can lead to drastic and often irreversible effects on the climate. A major concern is the potential collapse of the AMOC, a system of ocean currents that affects global climate. Recent studies suggest that it could collapse sooner than previously thought, which would have severe consequences. The project aims to enhance the understanding and prediction of tipping points in climate systems, and seeks to reconcile discrepancies between theoretical predictions, observational data, and climate model predictions.

Professor Susanne Ditlevsen, Department of Mathematical Sciences, University of Copenhagen, and Professor Peter Ditlevsen, Niels Bohr Institute, University of Copenhagen.

DKK 14 million.

CO2 is a primary driver of climate, yet its large variations over the last million years remain a puzzle, which casts doubt on the credibility of our future climate projections. Over the last decade, however, there have been numerous theoretical, observational and numerical breakthroughs, which together promise a better data-driven understanding of the causal mechanisms behind CO2 and climate fluctuations: - novel causal inference methods based on machine learning - a doubling of both resolution and length of the ice-core records - Earth System Models (ESM) that realistically simulate the climate carbon-cycle system The objective of this project is to integrate these developments in a novel framework termed CINCH that can improve our causal explanations of past and predictions of future climate changes. The focus of the present initiation project is to apply CINCH to the well dated and synchronized ice core records from Greenland and Antarctica for the period of 60 to 20,000 years ago to test the following hypothesis: The observed large-scale variability of CO2 is caused by periodic collapses of the North Atlantic marine productivity, and not by changes to Southern Ocean temperature or winds.  

Professor Markus Jochum, Niels Bohr Institute, University of Copenhagen, and Professor Niels Hansen, Department of Mathematical Sciences, University of Copenhagen.

DKK 4.5 million.

Electrocatalysis is a critical element in mitigating climate change by using renewable energy to drive reactions into sustainable fuels and chemicals. Electrocatalytic reaction depends on multiple parameters, which challenge performance metrics i.e. poor selectivity, bad energy efficiency, and the catalyst materials drop in performance over time. The lack of standards towards designing and setting up experiments challenges the creation of new usable knowledge and pushing the field forward.  This project sets out to address this by developing pareto-optimality guided generative model for electrocatalytic CO2 reduction, using a) data streams from scientific research literature, experiments and atomistic simulations, b) hypothesis-driven statistical design-of-experiments to develop the pareto algorithms. The project will predict new and under-explored optimal experiments and conditions, balanced data efficiency, and have expert-in-the-loop with domain-knowledge to improve the model. The potential of this project is to initiate a catalyst informatics platform by offering a new way of developing and improving catalysis, thus, this project is the seed needed for further collaborative funding between the PIs.

Associate Professor Alexander Bagger, DTU Physics, Technical University of Denmark, and Professor Line Clemmensen, Department of Mathematical Sciences, University of Copenhagen.

DKK 4.5 million.

Our main goal is to develop more efficient protein engineering (PE) methods by combining state-of-the-art experimental techniques with novel machine learning (ML) methods that incorporate nonlinear epistatic effects. The domain expert has pioneered high-throughput methods for enzyme engineering, whereas the methodological expert has developed cutting-edge methods in computer science based on Bayesian optimization, a ML algorithm forsolving optimization problems in low-data regimes. This collaboration aims to extend the previous work of both experts by focusing on the challenges of multiparametric optimisation and non-linear epistatic effects that are challenging in PE. As a model system, we will use an NADPH-dependent enzyme called formate dehydrogenase (FDH) that interconverts carbon dioxide (CO2) and formic acid, thus exhibiting potential for recycling biological cofactors in industrial biocatalysis (and in the reverse direction CO2 “capture”). Our computational methods are target agnostic and will be thus transferable to any protein or enzyme to support biotechnnologists optimizing complex systems in which non-linear, epistatic and trade-off effects between multiple properties are wide-spread and difficult to handle.

Senior Researcher Carlos Acevedo-Rocha, Novo Nordisk Foundation Center for Biosustainability (DTU Biosustain), Technical University of Denmark, and Professor Søren Hauberg, Department of Mathematics and Computer Science, Technical University of Denmark.

DKK 4.5 million

Quantum advantage, where classical supercomputers fail to simulate quantum computations, is a milestone on the path to practical quantum computation. Gaussian boson sampling is a leading task, but optical losses and improved classical algorithms make its quantum advantage debatable. A promising recent development is adding non-linear optical elements, moving beyond the Gaussian regime. This proposal aims to identify the classical-quantum boundary of non-linear boson sampling by (i) designing tensor network algorithms to test classical limits and (ii) characterizing experimental setups that maximize quantum advantage. This research has practical applications, particularly in simulating vibrational quantum dynamics of correlated molecules. Success would both determine a route to quantum advantage in a real problem and develop methods that could lead to better classical simulation algorithms for molecular vibronics.

Associate Professor Michael Kastoryano, Department of Computer Science, University of Copenhagen, and Associate Professor Stefano Paesani, Niels Bohr Institute, University of Copenhagen.

DKK 4.5 million.

Biomolecules interact and function in a complex ‘dance’. Single molecule microscopy imaging enables the direct observation of individual biomolecules in real-time. This provides unprecedented insights into their behavior and interactions, and brings us closer to a dynamic understanding of biological processes – the molecular dance. However, strong data analysis approaches are required to harvest the molecular information, and the complexity of biological processes and single molecule data calls for new data-driven mathematical models and their experimental validation. Mathematicians and biophysical chemists aim together in this synergy project to bring new analysis tools based on advances in hidden Markov modelling to decode real-time single molecule imaging data.     

Associate Professor Victoria Birkedal, Department of Chemistry, Aarhus University, and Professor Asger Hobolth, Department of Mathematics, Aarhus University.

DKK 4.5 million. 

Single-photon sources are crucial for quantum technology, enabling secure communications, precise measurements, and efficient quantum computing. However, current technologies for generating single-photon sources suffer from low efficiency, multi-photon noise, and inhomogeneous broadening. This project aims to transform quantum technologies by creating a new on-chip quantum light source for integrated quantum devices on a single chip. Using 2D superlattice, the project aims to generate high-efficiency, high-indistinguishability quantum emitters. We will employ machine learning (ML) to rapidly identify monolayer 2D materials, structural optimization and quantum measurement. The project combines the expertise of AAU's ML specialists and DTU Electro's 2D photonic experts to achieve its ambitious goals.

Associate Professor Sanshui Xiao, DTU Electro, Technical University of Denmark, and Professor Zheng-Hua Tan, Faculty of IT and Design, Aalborg University.

DKK 4.5 million.  

Batteries are vital for the green transition, enabling electrification of transportation and integration of renewable energy. Despite falling costs, uncertainties about battery lifetime challenge sustainability. Accurate lifetime prediction is key for effective management, performance, and preventing failures such as fires. Current prediction methods rely on extensive lab testing and post-mortem analyses, which are complex and energy-intensive. To address these issues, a cross-disciplinary collaboration is dedicated, bringing together experts in battery technology, statistical signal processing, and machine learning. The goal is to apply advanced data science techniques for online, non-invasive battery lifetime prediction through acoustic emission detection. This approach models the battery as a 3D acoustic landscape, tracking its changes over time. The expected outcomes are groundbreaking for battery management and will advance explainable, data-driven prediction for time series data.

Associate Professor Daniel-Ioan Stroe, AAU Energy, Aalborg University, and Professor Jan Østergaard, Department of Electronic Systems, Aalborg University.

DKK 4.5 million.

The need to secure information permeates all layers of today’s digitalized society, individuals, companies, and governmental institutions. Goals are, e.g., communication privacy, secure transactions, and countering censorship and disinformation. Quantum cryptography (QC) offers highly secure solutions, for, e.g., randomness generation and secret key exchange. Quantum correlations enable security under minimal assumptions, even with untrusted devices ( device independent (DI) QC). Such ultra-high assurance comes at a cost: DI-QC is technologically challenging and inefficient. Semi-device-independent quantum cryptography (SDI-QC) strikes a balance between minimal trust and efficiency. We will take on two bottlenecks of SDI-QC, a feasibility barrier and an efficiency bottleneck.  We aim to overcome both barriers using representation theory . Especially in the multi-party setting, we will exploit symmetries to reduce complexity and thus enable feasible and efficient analysis.    

Associate Professor Christian Majenz, DTU Compute, Technical University of Denmark, and Associate Professor Jonatan Bohr Brask, DTU Physics, Technical University of Denmark.

DKK 4.5 million.

District heating systems are a clean and affordable solution to heat thousands of buildings in Denmark and Europe. Their sustainability depends on their efficiency and on their ability to decarbonize heat generation. Integrating excess heat sources is key to achieving this, but they should be economically viable and reliable, possibly requiring investments in energy storage. However, their temporal variability together with the spatial arrangement of the network's elements makes planning these improvements challenging. Current engineering methods struggle with this complexity and scale. This project will bridge this gap by employing techniques designed for large-scale networks, such as social networks, to this problem. Using data from the Aarhus district heating system we will construct a temporal graph from the data, detect the parts of the network to enhance, and enable engineering-informed optimization for accurate identification of exploitable CO2 reduction opportunities. 

Associate Professor Davide Mottin, Department of Computer Science, Aarhus University, and Massimo Fiorentini, Department of Civil and Architectural Engineering, Aarhus University.

DKK 4.5 million. 

Explaining the decisions of complex AI models, like deep neural networks, is very hard. This is a key challenge for ethically responsible AI. Computer science develops technical solutions to this challenge within the field of Explainable AI. However, rigorous methods for evaluating the explainability of AI are currently lacking. Computer scientists use data-driven metrics. While they provide some insights for developers, it is unclear whether such metrics track ethically relevant properties. Conversely, AI ethics uses normative criteria. While based on principled ethical analyses of explainability, they are too abstract and provide little practical guidance. REMAX will bridge this gap by establishing a close collaboration between computer science and AI ethics. We will study how principled normative criteria can be translated into novel metrics of AI explainability, and how closer alignment with AI development practices can make these normative criteria more practically usable.

Associate Professor Rune Nyrup, Department of Mathematics, Aarhus University, and Professor Ira Assent, Department of Computer Science, Aarhus University.

DKK 4.5 million.
 

DK-Future will develop novel geographical machine learning models, termed probabilistic GeoML, designed to predict future land use changes by integrating probabilistic approaches that incorporate geographical data and accounting for uncertainties in climate change scenarios. These models will leverage historical Earth observation and climate projection data to forecast land use changes in Denmark under compound climate impacts. With its low-lying terrain and long coastline, Denmark is highly vulnerable to climate-induced land use changes, highlighting the need for forecasting tools to support proactive land use management with the associated uncertainties. This effort requires advances in probabilistic modeling of complex spatio-temporal processes, in close synergy with the geographical sciences, incorporating concepts like spatial proximity and autocorrelation into ML models. This blend of probabilistic ML (PI1) and land use and climate change dynamics (PI2) will open new research avenues for modeling spatio-temporal processes and expand our understanding of how climate change impacts land use, under uncertain climate projections. This research is essential for decision-makers, providing insights into extreme events and their likelihood.

Lektor Andres Masegosa, Institut for Datalogi, Aalborg Universitet og professor Jamal Jokar Arsanjani, Institut for Bæredygtighed og Planlægning, Aalborg Universitet.

DKK 4.5 million.

Genetic exchange across species boundaries is a much more common phenomenon than previously assumed. Understanding how different species are connected in reticulated evolutionary networks is of fundamental interest in Biology. However, currently available methods to infer evolutionary relationships are based on simplified bifurcating models of species and population divergence. We propose to develop a new type of Admixture Graph Models, to disentangle the evolutionary network of wild and domestic cattle, with a special emphasis on detecting selection acting on genes that have crossed species boundaries. This will help us understand fundamental aspects of the evolutionary process, in particular speciation. It will also be highly valuable for identifying the types of genes that repeatedly transcend species boundaries. The two PIs have extensive experience in graph theory and applied mathematics (PI 1), and wildlife genetics and biology (PI 2), respectively. Therefore, the project holds great promise for integrating disciplines that are not usually applied to the same research questions. 

Associate Professor Rasmus Heller, Department of Biology, University of Copenhagen, and Professor Carsten Wiuf, Department of Mathematical Sciences, University of Copenhagen.

DKK 4.5 million.