Location: Room P6 – Peony Jr 4511-2 (Level 4)
Abstract: The provision of weather forecasts via Numerical Weather Prediction (NWP) has traditional been a major application of large-scale supercomputers. Similarly, compute-intense model-based projections of future climate have contributed significantly in recent decades to our understanding and preparations for changes climate in coming decades. This session will review some of the key benefits, computational challenges and new science and technical solutions in the field of weather and climate science. On weather timescales in particular, a revolution is underway through the rapidly expanding use of AI-based approaches trained on the vast numbers of observations archived over recent decades. The use of AI on climate change timescales is also increasing, but frequently takes different forms due to the inherent lack of observations of future climate. This session will include perspectives from universities, the private sector and agencies on the role of HPC/AI in the addressing the challenge of dealing with extreme weather and climate change, both now and in the future.
Track Chair: Prof Dale Barker, Director, Centre for Climate Research Singapore, Meteorological Service Singapore
[Invited Track]
Programme:
| Time | Session |
| 01:30pm – 03:30pm | Opening – Prof Dale Barker, Director (CCRS) |
| Computation and AI for High Performance Climate In this talk, we will explore the role of computation and artificial intelligence (AI) in advancing high-performance climate modeling. We will delve into how computation can simulate and predict climate change, addressing the limitations of current climate simulations. With this problem in mind, we will discuss the onset of a new computational era, highlighted by the rapid growth of AI infrastructure investments. We realize how this development can help us to drive finer-grained climate models and discuss the challenges and computational demands required to achieve higher resolutions. We then dive into ML-based models with the example of diffusion-based models for data assimilation, showcasing opportunities and improvements over traditional methods. Additionally, we will show pitfalls with existing AI-based weather forecasting networks, using case studies such as Storm Ciaran and Typhoon Doksuri to underscore their limitations in predicting extreme events and capturing underlying physics. We will conclude with an overview of the Swiss AI climate initiative, a project aimed at developing a foundational model for climate and weather prediction by integrating more data, advanced models, and higher resolutions. Efforts to democratize access to climate data through compression techniques will also be highlighted. – Prof Torsten Hoeffler, ETH, Switzerland | |
| Developments in Generative AI for Climate and Weather Generative AI (GenAI) has become extremely popular in recent times with the advent of LLMs like ChatGPT and image/video generation tools like Stable Diffusion, DALL-E, etc. GenAI methods have also shown great promise in the Climate and Weather domain. Early work with GANs has been augmented with newer generative methods, such as denoising diffusion models and score-based generative models. This talk discusses the features of these latest models and their potential applications. We will also share recent progress on use-cases of generative models for climate and weather modelling. – Dr Jeff Adie, NVIDIA | |
| Preparations for Next-Generation Weather and Climate Modelling at the Centre for Climate Research Singapore (CCRS) The increasing diversity of large-scale HPC architectures available to run compute-intense numerical weather prediction (NWP) and climate projection simulations requires a fundamental rewrite of the traditional physical models that have underpinned weather and climate science for decades. At the same time, the advent of AI/ML-based approaches to simulate physical and dynamical processes within the atmosphere provides an opportunity to replace aspects of the physical models (or even replace them completely) with more computationally efficient, and often more accurate, AI-based algorithms. This talk will provide an overview of CCRS’ activities to move from the current generation of physical models and data assimilation algorithms based on the Unified Modelling (UM) System. Next-generation weather and climate modelling for the Singapore region has specific requirements, given the unique tropical urban nature of the weather and climate in the region. Despite this, significant progress has been made in recent years to add value to global predictions/projections through regional NWP And climate projections using km-scale models. A brief description of past efforts, plus plans to move the the next-generation SINGV_NG system will be presented. – Prof Dale Barker, Professor, CCRS | |
| Potential for Impact in Modelling and Simulation via AI Innovations in Weather Forecasting and Climate Change Impact Assessment Recent transformative AI advances in short-term weather forecasting and long-term climate projections are an exciting development with major implications for downstream tasks. In this talk, we first illustrate how weather forecasting and climate simulations can enhance numerical models for critical downstream applications such as urban planning and environmental management. For example, accurate weather predictions can support operational decision-making in pollutant and chemical dispersion modeling as well as urban air quality management. Similarly, reliable long-term climate projections can inform sustainable and resilient urban design. The integration of AI with physics-based models enables actionable insights, thereby empowering decision-makers in the cities and industry of tomorrow. – Dr Ooi Chin Chun, A*STAR | |
| 03:30pm – 04:00pm | Tea Break |
| 04:00pm – 06:00pm | Explainable Natural Language Processing for Corporate Sustainability Analysis Sustainability commonly refers to entities, such as individuals, companies, and institutions, having a nondetrimental (or even positive) impact on the environment, society, and the economy. With sustainability becoming a synonym of acceptable and legitimate behaviour, it is being increasingly demanded and regulated. Several frameworks and standards have been proposed to measure the sustainability impact of corporations, including United Nations’ sustainable development goals and the recently introduced global sustainability reporting framework, amongst others. However, the concept of corporate sustainability is complex due to the diverse and intricate nature of firm operations (i.e. geography, size, business activities, interlinks with other stakeholders). As a result, corporate sustainability assessments are plagued by subjectivity both within data that reflect corporate sustainability efforts (i.e. corporate sustainability disclosures) and the analysts evaluating them. This subjectivity can be distilled into distinct challenges, such as incompleteness, ambiguity, unreliability and sophistication on the data dimension, as well as limited resources and potential bias on the analyst dimension. Put together, subjectivity hinders effective cost attribution to entities non-compliant with prevailing sustainability expectations, potentially rendering sustainability efforts and its associated regulations futile. To this end, we argue that Explainable Natural Language Processing (XNLP) can significantly enhance corporate sustainability analysis. Specifically, linguistic understanding algorithms (lexical, semantic, syntactic), integrated with XAI capabilities (interpretability, explainability, faithfulness), can bridge gaps in analyst resources and mitigate subjectivity problems within data. – Prof Erik Cambria, NTU |
| Toward integration of ML/NWP/DA Machine learning (ML)-based models for weather and climate have been evolving rapidly. Most of these ML models are trained on existing atmospheric reanalysis data produced by conventional numerical weather prediction (NWP) and data assimilation (DA) systems, which can also be further improved by ML. At RIKEN, we are investigating various applications of ML in the NWP-DA framework. We will present a precipitation nowcasting system by combining ML and NWP, a ML-based observation operator for satellite radiances, a combination of ML and DA to generate better training data for ML surrogate, and a regional NWP with a ML-based model. – Dr Shigenori Otsuka, RIKEN | |
| Digital Twins of the Earth’s Weather: To and from (X)AI Artificial Intelligence (AI) is transforming weather and climate science, offering new ways to enhance predictions, improve model efficiency, and uncover hidden patterns in complex datasets. In the first part of this talk, I will first present our work that led us to AI, taking the route of physics and dynamical systems theory. I will then show some more recent work that brings us back from AI to physics and nonlinear dynamics, where the key is bridging human and machine knowledge. – Asst Prof Gianmarco Mengaldo, NUS | |
| [Topic TBC] – JAMSTEC | |
| Closing – Prof Dale Barker, Director (CCRS) |
