Verena Kain
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Verena Kain has a total of 225 co-authors. Here are the top co-authors by number of shared publications:
- Brennan Goddard
European Organization for Nuclear Research
6 shared publications - Annika Högner
International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria
4 shared publications - Carl-Friedrich Schleussner
International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria
4 shared publications - Francesco Maria Velotti
CERN
4 shared publications - Matthew Alexander Fraser
CERN, Geneva, Switzerland
4 shared publications - Michelangelo Pari
CERN
4 shared publications
Latest Publications
- Open Access
- 16 November 2026
- Published by Copernicus GmbH
Supplementary material to "FASTMIP-CMIP7: A Coordinated Multi-Emulator Ensemble for Rapid, Spatially Resolved Climate Projections under CMIP7 Scenarios"
- 16 November 2026
- Published by Copernicus GmbH
Open AccessSupplementary material to "FASTMIP-CMIP7: A Coordinated Multi-Emulator Ensemble for Rapid, Spatially Resolved Climate Projections under CMIP7 Scenarios"
- Open Access
- 16 November 2026
- Published by Copernicus GmbH in EGUsphere
FASTMIP-CMIP7: A Coordinated Multi-Emulator Ensemble for Rapid, Spatially Resolved Climate Projections under CMIP7 Scenarios
Earth System Models (ESMs) remain the gold standard for translating socioeconomic scenarios into spatially resolved climate projections, but their computational cost limits the number of scenarios and ensemble members that can be simulated and ties projections to the evidence available when simulations are performed. Climate emulators, which are statistical and machine-learning models trained on ESM output, enable large ensembles and allow projections to be updated as new observations emerge. Here, we present FASTMIP-CMIP7, a coordinated ensemble of seven emulators providing spatially and temporally resolved projections of temperature, precipitation and climate impact indicators for four newly released scenarios from the 7th phase of the Coupled Model Intercomparison Project (CMIP7), spanning very low (VL) to high (H) emissions. Despite scenario ranges having narrowed in CMIP7 comapred to CMIP6, emissions pathways remain the dominant source of uncertainty in regional end-of-century temperature projections, making mitigation the strongest lever for limiting future warming. In 2100, global land temperature reaches 2.3 °C under stringent mitigation (VL), and 3.9 °C under the current-policies scenario (M), approximately 1.4 times the global mean warming. Under M, even the coolest 5 % of years in 2100 are warmer over land than 95 % of present-day years. Precipitation projections are more heterogeneous, with internal variability and regional climate responses often obscuring scenario differences. Nevertheless, robust drying emerges in northern South America and the Mediterranean, where additional drying relative to today is 2.5 to almost four times greater under current policies (M) than under stringent mitigation (VL). Structural differences between emulators and choices in training, testing and validation introduce a new layer of uncertainty to emulated ensembles. Common, streamlined validation experiments are needed to establish confidence in emulated projections and support the broader integration of climate emulators into future climate assessments.
- 16 November 2026
- Published by Copernicus GmbH in EGUsphere
Open AccessFASTMIP-CMIP7: A Coordinated Multi-Emulator Ensemble for Rapid, Spatially Resolved Climate Projections under CMIP7 Scenarios
Earth System Models (ESMs) remain the gold standard for translating socioeconomic scenarios into spatially resolved climate projections, but their computational cost limits the number of scenarios and ensemble members that can be simulated and ties projections to the evidence available when simulations are performed. Climate emulators, which are statistical and machine-learning models trained on ESM output, enable large ensembles and allow projections to be updated as new observations emerge. Here, we present FASTMIP-CMIP7, a coordinated ensemble of seven emulators providing spatially and temporally resolved projections of temperature, precipitation and climate impact indicators for four newly released scenarios from the 7th phase of the Coupled Model Intercomparison Project (CMIP7), spanning very low (VL) to high (H) emissions. Despite scenario ranges having narrowed in CMIP7 comapred to CMIP6, emissions pathways remain the dominant source of uncertainty in regional end-of-century temperature projections, making mitigation the strongest lever for limiting future warming. In 2100, global land temperature reaches 2.3 °C under stringent mitigation (VL), and 3.9 °C under the current-policies scenario (M), approximately 1.4 times the global mean warming. Under M, even the coolest 5 % of years in 2100 are warmer over land than 95 % of present-day years. Precipitation projections are more heterogeneous, with internal variability and regional climate responses often obscuring scenario differences. Nevertheless, robust drying emerges in northern South America and the Mediterranean, where additional drying relative to today is 2.5 to almost four times greater under current policies (M) than under stringent mitigation (VL). Structural differences between emulators and choices in training, testing and validation introduce a new layer of uncertainty to emulated ensembles. Common, streamlined validation experiments are needed to establish confidence in emulated projections and support the broader integration of climate emulators into future climate assessments.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Systematic comparison of emulation techniques for regional climate under temperature overshoot scenarios
How regional climate change evolves in overshoot scenarios, in particular after the global mean temperature (GMT) peak, is not well understood. To investigate regional changes under overshoot, we develop an emulator that predicts trends in regional climate variables at the spatial level of IPCC regions from GMT time series, with applicability both before and after overshoot.A commonly used approach to relate regional climate change to GMT is pattern scaling, which assumes a linear relationship between GMT and regional climate variables. Previous studies indicate limitations in applying pattern scaling under post-overshoot conditions, a finding that is also reflected in results produced as part of our emulator development.We therefore apply a range of alternative techniques to solve the regional climate trend emulation problem. These include approaches based on the existing literature, such as impulse response functions and operator approximation, as well as machine-learning-based methods, including Gaussian process regression, random forests, XGBoost, state space models, and pre-trained deep-learning-based time series prediction techniques. All methods are trained on overshoot and non-overshoot simulations from CMIP6, Flat10MIP, and additional model experiments available in the literature.We assess the performance of each approach under overshoot scenarios and compare them with simple pattern scaling used as a baseline to assess approach performance. We introduce an evaluation framework for emulations under long-term stabilisation and overshoot pathways that accounts for whether regional climate signals are reversible or irreversible and enables robust detection of overshoot and stabilisation dynamics.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessSystematic comparison of emulation techniques for regional climate under temperature overshoot scenarios
How regional climate change evolves in overshoot scenarios, in particular after the global mean temperature (GMT) peak, is not well understood. To investigate regional changes under overshoot, we develop an emulator that predicts trends in regional climate variables at the spatial level of IPCC regions from GMT time series, with applicability both before and after overshoot.A commonly used approach to relate regional climate change to GMT is pattern scaling, which assumes a linear relationship between GMT and regional climate variables. Previous studies indicate limitations in applying pattern scaling under post-overshoot conditions, a finding that is also reflected in results produced as part of our emulator development.We therefore apply a range of alternative techniques to solve the regional climate trend emulation problem. These include approaches based on the existing literature, such as impulse response functions and operator approximation, as well as machine-learning-based methods, including Gaussian process regression, random forests, XGBoost, state space models, and pre-trained deep-learning-based time series prediction techniques. All methods are trained on overshoot and non-overshoot simulations from CMIP6, Flat10MIP, and additional model experiments available in the literature.We assess the performance of each approach under overshoot scenarios and compare them with simple pattern scaling used as a baseline to assess approach performance. We introduce an evaluation framework for emulations under long-term stabilisation and overshoot pathways that accounts for whether regional climate signals are reversible or irreversible and enables robust detection of overshoot and stabilisation dynamics.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Cascaded score-based emulation of Earth system models for impact evaluation with SCALES-MESH
Today's climate adaptation and mitigation planning tasks require rapid access to large ensembles of climate projections for a wide range of emissions scenarios, including overshoot scenarios. While Earth system models (ESMs) provide physically consistent projections, their high computational cost limits scenario exploration. Climate emulators - statistical or machine-learning-based models trained on ESM data to generate data replicating the ESMs behaviour for a multitude of emissions scenarios - are therefore proposed to deliver these projections efficiently. Here we present the novel modular SCALES–MESH emulator framework, combining physics-based regional projections with AI downscaling capabilities. The SCALES module translates projections of global mean surface air temperature into regional surface air temperature projections aggregated over the AR6-IPCC regions, while the MESH module performs spatio-temporal downscaling to gridded fields using a conditional score-based generative model. MESH is trained on multiple datasets and evaluated against parent ESMs using spatial, temporal, and distributional diagnostics. Results show that the emulator captures regional patterns, temporal variability, and probability distributions of emulated climate variables, including during warming and cooling phases of overshoot scenarios. We further demonstrate the potential for transfer learning across ESMs, pointing toward scalable multi-model and resolution-agnostic emulation. Together, SCALES–MESH enables rapid, flexible, and physically grounded exploration of climate futures, supporting decision-relevant climate risk assessment at unprecedented scope.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessCascaded score-based emulation of Earth system models for impact evaluation with SCALES-MESH
Today's climate adaptation and mitigation planning tasks require rapid access to large ensembles of climate projections for a wide range of emissions scenarios, including overshoot scenarios. While Earth system models (ESMs) provide physically consistent projections, their high computational cost limits scenario exploration. Climate emulators - statistical or machine-learning-based models trained on ESM data to generate data replicating the ESMs behaviour for a multitude of emissions scenarios - are therefore proposed to deliver these projections efficiently. Here we present the novel modular SCALES–MESH emulator framework, combining physics-based regional projections with AI downscaling capabilities. The SCALES module translates projections of global mean surface air temperature into regional surface air temperature projections aggregated over the AR6-IPCC regions, while the MESH module performs spatio-temporal downscaling to gridded fields using a conditional score-based generative model. MESH is trained on multiple datasets and evaluated against parent ESMs using spatial, temporal, and distributional diagnostics. Results show that the emulator captures regional patterns, temporal variability, and probability distributions of emulated climate variables, including during warming and cooling phases of overshoot scenarios. We further demonstrate the potential for transfer learning across ESMs, pointing toward scalable multi-model and resolution-agnostic emulation. Together, SCALES–MESH enables rapid, flexible, and physically grounded exploration of climate futures, supporting decision-relevant climate risk assessment at unprecedented scope.
- Open Access
Quantum Boltzmann Machines for Sample-Efficient Reinforcement Learning
We introduce theoretically grounded Continuous Semi-Quantum Boltzmann Machines (CSQBMs) that supports continuous-action reinforcement learning. By combining exponential-family priors over visible units with quantum Boltzmann distributions over hidden units, CSQBMs yield a hybrid quantum-classical model that reduces qubit requirements while retaining strong expressiveness. Crucially, gradients with respect to continuous variables can be computed analytically, enabling direct integration into Actor-Critic algorithms. Building on this, we propose a continuous Q-learning framework that replaces global maximization by efficient sampling from the CSQBM distribution, thereby overcoming instability issues in continuous control.
Open AccessQuantum Boltzmann Machines for Sample-Efficient Reinforcement Learning
We introduce theoretically grounded Continuous Semi-Quantum Boltzmann Machines (CSQBMs) that supports continuous-action reinforcement learning. By combining exponential-family priors over visible units with quantum Boltzmann distributions over hidden units, CSQBMs yield a hybrid quantum-classical model that reduces qubit requirements while retaining strong expressiveness. Crucially, gradients with respect to continuous variables can be computed analytically, enabling direct integration into Actor-Critic algorithms. Building on this, we propose a continuous Q-learning framework that replaces global maximization by efficient sampling from the CSQBM distribution, thereby overcoming instability issues in continuous control.
- Top Cited PapersOpen Access
- 6 August 2024
- Published by American Physical Society (APS) in Physical Review Accelerators and Beams
- Vol. 27 (8) , 084801
Bayesian optimization algorithms for accelerator physics
Accelerator physics relies on numerical algorithms to solve optimization problems in online accelerator control and tasks such as experimental design and model calibration in simulations. The effectiveness of optimization algorithms in discovering ideal solutions for complex challenges with limited resources often determines the problem complexity these methods can address. The accelerator physics community has recognized the advantages of Bayesian optimization algorithms, which leverage statistical surrogate models of objective functions to effectively address complex optimization challenges, especially in the presence of noise during accelerator operation and in resource-intensive physics simulations. In this review article, we offer a conceptual overview of applying Bayesian optimization techniques toward solving optimization problems in accelerator physics. We begin by providing a straightforward explanation of the essential components that make up Bayesian optimization techniques. We then give an overview of current and previous work applying and modifying these techniques to solve accelerator physics challenges. Finally, we explore practical implementation strategies for Bayesian optimization algorithms to maximize their performance, enabling users to effectively address complex optimization challenges in real-time beam control and accelerator design. Published by the American Physical Society 2024
- 6 August 2024
- Published by American Physical Society (APS) in Physical Review Accelerators and Beams
- Vol. 27 (8) , 084801
Top Cited PapersOpen AccessBayesian optimization algorithms for accelerator physics
Accelerator physics relies on numerical algorithms to solve optimization problems in online accelerator control and tasks such as experimental design and model calibration in simulations. The effectiveness of optimization algorithms in discovering ideal solutions for complex challenges with limited resources often determines the problem complexity these methods can address. The accelerator physics community has recognized the advantages of Bayesian optimization algorithms, which leverage statistical surrogate models of objective functions to effectively address complex optimization challenges, especially in the presence of noise during accelerator operation and in resource-intensive physics simulations. In this review article, we offer a conceptual overview of applying Bayesian optimization techniques toward solving optimization problems in accelerator physics. We begin by providing a straightforward explanation of the essential components that make up Bayesian optimization techniques. We then give an overview of current and previous work applying and modifying these techniques to solve accelerator physics challenges. Finally, we explore practical implementation strategies for Bayesian optimization algorithms to maximize their performance, enabling users to effectively address complex optimization challenges in real-time beam control and accelerator design. Published by the American Physical Society 2024
- Open Access
- 9 July 2024
- Published by American Physical Society (APS) in Physical Review Accelerators and Beams
- Vol. 27 (7) , 074001
Empty-bucket techniques for spill-quality improvement at the CERN Super Proton Synchrotron
Synchrotrons can provide long spills of particles by employing resonant extraction where the circulating beam is slowly ejected over thousands to millions of turns by exploiting the amplitude growth caused by a transverse resonance. In the CERN Super Proton Synchrotron (SPS), this method is used to satisfy the experimental requests of the North Area. However, the extracted particle flux is modulated by power-converter ripple, an issue shared across all sychrotrons that perform resonant extraction. In order to suppress such modulations, empty-bucket techniques can be employed, which take advantage of chromaticity to quickly accelerate particles into resonant motion by using a longitudinal rf system. This paper explores empty-bucket techniques via theory, simulation, and measurement, providing a systematic characterization with general applicability to any machine. Additionally, the operational implementation in the SPS is detailed, where the impact on the beam profile and extracted intensity is addressed. Published by the American Physical Society 2024
- 9 July 2024
- Published by American Physical Society (APS) in Physical Review Accelerators and Beams
- Vol. 27 (7) , 074001
Open AccessEmpty-bucket techniques for spill-quality improvement at the CERN Super Proton Synchrotron
Synchrotrons can provide long spills of particles by employing resonant extraction where the circulating beam is slowly ejected over thousands to millions of turns by exploiting the amplitude growth caused by a transverse resonance. In the CERN Super Proton Synchrotron (SPS), this method is used to satisfy the experimental requests of the North Area. However, the extracted particle flux is modulated by power-converter ripple, an issue shared across all sychrotrons that perform resonant extraction. In order to suppress such modulations, empty-bucket techniques can be employed, which take advantage of chromaticity to quickly accelerate particles into resonant motion by using a longitudinal rf system. This paper explores empty-bucket techniques via theory, simulation, and measurement, providing a systematic characterization with general applicability to any machine. Additionally, the operational implementation in the SPS is detailed, where the impact on the beam profile and extracted intensity is addressed. Published by the American Physical Society 2024
- Open Access
- 21 February 2024
- Published by IOP Publishing in Quantum Science and Technology
- Vol. 9 (2) , 025012
Hybrid actor-critic algorithm for quantum reinforcement learning at CERN beam lines
Free energy-based reinforcement learning (FERL) with clamped quantum Boltzmann machines (QBM) was shown to significantly improve the learning efficiency compared to classical Q-learning with the restriction, however, to discrete state-action space environments. In this paper, the FERL approach is extended to multi-dimensional continuous state-action space environments to open the doors for a broader range of real-world applications. First, free energy-based Q-learning is studied for discrete action spaces, but continuous state spaces and the impact of experience replay on sample efficiency is assessed. In a second step, a hybrid actor-critic (A-C) scheme for continuous state-action spaces is developed based on the deep deterministic policy gradient algorithm combining a classical actor network with a QBM-based critic. The results obtained with quantum annealing (QA), both simulated and with D-Wave QA hardware, are discussed, and the performance is compared to classical reinforcement learning methods. The environments used throughout represent existing particle accelerator beam lines at the European Organisation for Nuclear Research. Among others, the hybrid A-C agent is evaluated on the actual electron beam line of the Advanced Wakefield Experiment (AWAKE).
- 21 February 2024
- Published by IOP Publishing in Quantum Science and Technology
- Vol. 9 (2) , 025012
Open AccessHybrid actor-critic algorithm for quantum reinforcement learning at CERN beam lines
Free energy-based reinforcement learning (FERL) with clamped quantum Boltzmann machines (QBM) was shown to significantly improve the learning efficiency compared to classical Q-learning with the restriction, however, to discrete state-action space environments. In this paper, the FERL approach is extended to multi-dimensional continuous state-action space environments to open the doors for a broader range of real-world applications. First, free energy-based Q-learning is studied for discrete action spaces, but continuous state spaces and the impact of experience replay on sample efficiency is assessed. In a second step, a hybrid actor-critic (A-C) scheme for continuous state-action spaces is developed based on the deep deterministic policy gradient algorithm combining a classical actor network with a QBM-based critic. The results obtained with quantum annealing (QA), both simulated and with D-Wave QA hardware, are discussed, and the performance is compared to classical reinforcement learning methods. The environments used throughout represent existing particle accelerator beam lines at the European Organisation for Nuclear Research. Among others, the hybrid A-C agent is evaluated on the actual electron beam line of the Advanced Wakefield Experiment (AWAKE).
- Open Access
- 19 October 2023
- Published by MDPI AG in Physical Sciences Forum
- Vol. 8 (1) , 65
The Design of the ENUBET Beamline
The ENUBET project aims to reduce the flux-related systematics to 1% on a narrow band neutrino beam through monitoring the associated charged leptons in an instrumented decay tunnel. A key element of the project is the design of a meson transfer line with conventional magnets that maximize the yield of K and while minimizing the total length to reduce meson decay outside the instrumented region. In order to limit particle rates in the tunnel instrumentation, a high level of beam collimation is needed, thus allowing non-decayed mesons to reach the end of the tunnel. At the same time, fine-tuning of the shielding and the collimators is required to minimize any beam-induced background in the decay region. The magnetic lattice is optimized with TRANSPORT. The focusing of mesons from the target is performed with a static (quadrupole-based) system that, coupled with a slow proton extraction scheme, allows for a significant pile-up reduction at the tunnel instrumentation while retaining a particle yield large enough for high-precision neutrino cross-section measurements on a 3 year time scale. Charge and momentum selection in an 8.5GeV ± 10% momentum bite is performed by a double dipole system. Shielding elements are optimized with full simulation of the facility in Geant4. In particular, a powerful genetic algorithm is used to scan the parameter space of the collimators automatically in order to find a configuration that minimizes the halo background in the decay tunnel while preserving a large meson yield. This contribution will report the results of the optimization studies and the final design of the ENUBET beamline, together with dose estimation through a FLUKA simulation. The design of an alternative secondary beamline with a broad momentum range (4, 6, and 8.5 GeV/c) that could enhance the physics reach of the facility is additionally discussed.- 19 October 2023
- Published by MDPI AG in Physical Sciences Forum
- Vol. 8 (1) , 65
Open AccessThe Design of the ENUBET Beamline
The ENUBET project aims to reduce the flux-related systematics to 1% on a narrow band neutrino beam through monitoring the associated charged leptons in an instrumented decay tunnel. A key element of the project is the design of a meson transfer line with conventional magnets that maximize the yield of K and while minimizing the total length to reduce meson decay outside the instrumented region. In order to limit particle rates in the tunnel instrumentation, a high level of beam collimation is needed, thus allowing non-decayed mesons to reach the end of the tunnel. At the same time, fine-tuning of the shielding and the collimators is required to minimize any beam-induced background in the decay region. The magnetic lattice is optimized with TRANSPORT. The focusing of mesons from the target is performed with a static (quadrupole-based) system that, coupled with a slow proton extraction scheme, allows for a significant pile-up reduction at the tunnel instrumentation while retaining a particle yield large enough for high-precision neutrino cross-section measurements on a 3 year time scale. Charge and momentum selection in an 8.5GeV ± 10% momentum bite is performed by a double dipole system. Shielding elements are optimized with full simulation of the facility in Geant4. In particular, a powerful genetic algorithm is used to scan the parameter space of the collimators automatically in order to find a configuration that minimizes the halo background in the decay tunnel while preserving a large meson yield. This contribution will report the results of the optimization studies and the final design of the ENUBET beamline, together with dose estimation through a FLUKA simulation. The design of an alternative secondary beamline with a broad momentum range (4, 6, and 8.5 GeV/c) that could enhance the physics reach of the facility is additionally discussed.