Pushing the frontiers in climate modelling and analysis with machine learning

Nature Climate Change, Год журнала: 2024, Номер 14(9), С. 916 - 928

Опубликована: Авг. 23, 2024

Язык: Английский

---

A Digital Twin of the terrestrial water cycle: a glimpse into the future through high-resolution Earth observations

Frontiers in Science, Год журнала: 2024, Номер 1

Опубликована: Март 5, 2024

Climate change is profoundly affecting the global water cycle, increasing likelihood and severity of extreme water-related events. Better decision-support systems are vital to accurately predict monitor environmental disasters optimally manage resources. These must integrate advances in remote sensing, situ , citizen observations with high-resolution Earth system modeling, artificial intelligence (AI), information communication technologies, high-performance computing. Digital Twin (DTE) models a ground-breaking solution offering digital replicas simulate processes unprecedented spatiotemporal resolution. Advances observation (EO) satellite technology pivotal, here we provide roadmap for exploitation these methods DTE hydrology. The 4-dimensional Hydrology datacube now fuses EO data advanced modeling soil moisture, precipitation, evaporation, river discharge, report latest validation Mediterranean Basin. This can be explored forecast flooding landslides irrigation precision agriculture. Large-scale implementation such will require further assess products across different regions climates; create compatible multidimensional datacubes, retrieval algorithms, that suitable multiple scales; uncertainty both models; enhance computational capacity via an interoperable, cloud-based processing environment embodying open principles; harness AI/machine learning. We outline how various planned missions facilitate hydrology toward benefit if scientific technological challenges identify addressed.

Язык: Английский

---

Do AI models produce better weather forecasts than physics-based models? A quantitative evaluation case study of Storm Ciarán

npj Climate and Atmospheric Science, Год журнала: 2024, Номер 7(1)

Опубликована: Апрель 22, 2024

Abstract There has been huge recent interest in the potential of making operational weather forecasts using machine learning techniques. As they become a part forecasting toolbox, there is pressing need to understand how well current models can simulate high-impact events. We compare short medium-range Storm Ciarán, European windstorm that caused sixteen deaths and extensive damage Northern Europe, made by numerical prediction models. The four considered (FourCastNet, Pangu-Weather, GraphCast FourCastNet-v2) produce accurately capture synoptic-scale structure cyclone including position cloud head, shape warm sector location conveyor belt jet, large-scale dynamical drivers important for rapid storm development such as relative upper-level jet exit. However, their ability resolve more detailed structures issuing warnings mixed. All underestimate peak amplitude winds associated with storm, only some core seclusion none sharp bent-back frontal gradient. Our study shows great deal about performance properties be derived from case studies events Ciarán.

Язык: Английский

---

Artificial intelligence for modeling and understanding extreme weather and climate events

Nature Communications, Год журнала: 2025, Номер 16(1)

Опубликована: Фев. 24, 2025

In recent years, artificial intelligence (AI) has deeply impacted various fields, including Earth system sciences, by improving weather forecasting, model emulation, parameter estimation, and the prediction of extreme events. The latter comes with specific challenges, such as developing accurate predictors from noisy, heterogeneous, small sample sizes data limited annotations. This paper reviews how AI is being used to analyze climate events (like floods, droughts, wildfires, heatwaves), highlighting importance creating accurate, transparent, reliable models. We discuss hurdles dealing data, integrating real-time information, deploying understandable models, all crucial steps for gaining stakeholder trust meeting regulatory needs. provide an overview can help identify explain more effectively, disaster response communication. emphasize need collaboration across different fields create solutions that are practical, understandable, trustworthy enhance readiness risk reduction. Artificial Intelligence transforming study like helping overcome challenges integration. review article highlights models improve response, communication trust.

Язык: Английский

---

AI-empowered next-generation multiscale climate modelling for mitigation and adaptation

Nature Geoscience, Год журнала: 2024, Номер 17(10), С. 963 - 971

Опубликована: Сен. 25, 2024

Язык: Английский

---

Opinion: Why all emergent constraints are wrong but some are useful – a machine learning perspective

Atmospheric chemistry and physics, Год журнала: 2025, Номер 25(4), С. 2365 - 2384

Опубликована: Фев. 21, 2025

Abstract. Global climate change projections are subject to substantial modelling uncertainties. A variety of emergent constraints, as well several other statistical model evaluation approaches, have been suggested address these However, they remain heavily debated in the science community. Still, central idea relate future already observable quantities has no real substitute. Here, we highlight validation perspective predictive skill machine learning community a promising alternative viewpoint. Specifically, argue for quantitative approaches which each constraining relationship can be evaluated comprehensively based on out-of-sample test data – top qualitative physical plausibility arguments that commonplace justification new constraints. Building this perspective, review ideas types controlling-factor analyses (CFAs). The principal behind CFAs is use find climate-invariant relationships historical hold approximately under strong scenarios. On basis existing archives, validated perfect-climate-model frameworks. From such three reasons: (a) objectively both past and data, (b) provide more direct and, by design, physically plausible links between observations potential climates, (c) take high-dimensional complex into account functions learned constrain response. We demonstrate advantages two recently published CFA examples form constraints feedback mechanisms (clouds, stratospheric water vapour) discuss further challenges opportunities using example rapid adjustment mechanism (aerosol–cloud interactions). avenues work, including strategies non-linearity, tackle blind spots ensembles, integrate helpful priors Bayesian methods, leverage physics-informed learning, enhance robustness through causal discovery inference.

Язык: Английский

---

Equilibrium in the Computing Continuum through Active Inference

Future Generation Computer Systems, Год журнала: 2024, Номер 160, С. 92 - 108

Опубликована: Май 30, 2024

Computing Continuum (CC) systems are challenged to ensure the intricate requirements of each computational tier. Given system's scale, Service Level Objectives (SLOs), which expressed as these requirements, must be disaggregated into smaller parts that can decentralized. We present our framework for collaborative edge intelligence, enabling individual devices (1) develop a causal understanding how enforce their SLOs and (2) transfer knowledge speed up onboarding heterogeneous devices. Through collaboration, they (3) increase scope SLO fulfillment. implemented evaluated use case in CC system is responsible ensuring Quality (QoS) Experience (QoE) during video streaming. Our results showed required only ten training rounds four SLOs; furthermore, underlying structures were also rationally explainable. The addition new types done posteriori; allowed them reuse existing models, even though device type had been unknown. Finally, rebalancing load within cluster recover compliance after network failure from 22% 89%.

Язык: Английский

---

Causal connectivity measures for pulse-output network reconstruction: Analysis and applications

Proceedings of the National Academy of Sciences, Год журнала: 2024, Номер 121(14)

Опубликована: Март 29, 2024

The causal connectivity of a network is often inferred to understand function. It arguably acknowledged that the relies on causality measure one applies, and it may differ from network’s underlying structural connectivity. However, interpretation remains be fully clarified, in particular, how depends measures relates Here, we focus nonlinear networks with pulse signals as measured output, e.g., neural spike address above issues based four commonly utilized measures, i.e., time-delayed correlation coefficient, mutual information, Granger causality, transfer entropy. We theoretically show these are related another when applied signals. Taking simulated Hodgkin–Huxley real mouse brain two illustrative examples, further verify quantitative relations among demonstrate by any well coincides connectivity, therefore illustrating direct link between stress pulse-output can reconstructed pairwise without conditioning global information all other nodes network, thus circumventing curse dimensionality. Our framework provides practical effective approach for reconstruction.

Язык: Английский

---

Digital twins of the Earth with and for humans

Communications Earth & Environment, Год журнала: 2024, Номер 5(1)

Опубликована: Авг. 27, 2024

Digital twins of the Earth are digital representations system, spanning scales and domains. Their purpose is to monitor, forecast assess system consequences human interventions on system. Providing users with capability interact interrogate decision support systems for addressing environmental challenges. By informing humans their impact aspire promote new pathways moving forward. answering causal queries through intervention analysis, they can enhance evidence-based policy making. Existing primarily technological information that represent physical world. However, as social worlds intrinsically interconnected, we argue must be accounted both within outside Earth: Within impacts responses integral system; govern access development guide responsible use acquired from twins. Incorporating interactions in represents a transformative frontier, promising unparalleled insights into dynamics empower action. Humans represented Earth, but also play role usage, argues perspective based interdisciplinary scientific expert viewpoints.

Язык: Английский

---

An iterative crack tip correction algorithm discovered by physical deep symbolic regression

International Journal of Fatigue, Год журнала: 2024, Номер 187, С. 108432 - 108432

Опубликована: Июнь 12, 2024

Digital image correlation is a widely used technique in the field of experimental mechanics. In fracture mechanics, determining precise location crack tip crucial. this paper, we introduce novel detection algorithm based on displacement and strain fields obtained by digital correlation. Iterative correction formulas are discovered applying deep symbolic regression guided physical unit constraints to dataset simulated cracks under mode I, II mixed-mode conditions with variable T-stress. For training dataset, fit Williams series expansion super-singular terms at randomly chosen origins around actual tip. We analyse apply most promising one data from uniaxial biaxial fatigue growth experiments AA2024-T3 sheet material. Throughout experiments, positions reliably detected leading improved stability propagation curves.

Язык: Английский

---