Abstracts are due by 15 August 2024 at 5:00 PM ET
Abstract Fee and Author Instructions
All presenters must also register for the meeting.
The 11th Symposium on High Performance Computing for Weather, Water, and Climate is sponsored by the American Meteorological Society and organized by the AMS Board on Environmental Information Processing Technologies.
In 2025, the American Meteorological Society High Performance Computing Symposium for Weather, Water, and Climate is excited to announce it will hold its 11th meeting as a part of the 105th AMS Annual Meeting to be held in New Orleans, LA January 12-16, 2025. This year’s symposium is looking to encourage growth in our relatively small community with 19 total proposed sessions over a large array of topics including new and emerging hardware and algorithms, numerical weather prediction, workflows, artificial intelligence, utilization of GPUs, environment and health, cloud HPC computing, industry trends and innovations, community modeling, and many more. Of these topics, 8 are proposed as joint sessions with relevant sister conferences to help establish or renew ties with the interdisciplinary needs for HPC across the atmospheric, oceanic, and hydrological sciences. We highly encourage those looking to submit abstracts to any of these sessions to please review the topics, their descriptions, and whether or not they are proposed as joint sessions with relevant conferences to help ensure the best audience for your submission. Full proposed session topics and their descriptions are listed below. Abstract submissions are due by August 15, 2024. We look forward to seeing you in New Orleans!
Many application codes running on high performance computing platforms are characterized by their high resource use (nodes, storage, I/O) and their often long run-times. This resource usage leads to high cost, especially when those resources are based in the cloud. Therefore it is important to use the resources efficiently - both to reduce costs and to deliver timely results. Also, application performance is dependent on effective selection and configuration of HPC platforms that suit the application. All of these considerations make it important to have tools and techniques to measure and analyze performance of HPC applications and platforms. This session invites papers on all aspects of this topic, including benchmark suites, tools for performance analysis, performance analysis and benchmarking case studies, and use of these tools and techniques to select HPC platforms and to optimize applications.
The use of artificial intelligence and machine learning algorithms has grown considerably in the past several years, and with that growth has come an increase in demand for faster and larger compute platforms that tailor themselves to quickly and efficiently performing the specialized calculations and data manipulations required. GPUs currently are the true workhorse devices for this task for many reasons, however new processor architectures that are designed with AI in mind from the start are now emerging in the form of IPUs, wafer-scale computers, and other designs touting further performance gains and ease of use. Additionally, new storage systems and memory designs and configurations better tuned to the read and write patterns of AI methods are also coming to market.
This session seeks to explore any of these new AI-centric hardware technologies being put to use in the development of new AI models for weather, water, and climate applications. Submissions are encouraged to not only introduce the innovative AI-centric hardware being used, but also the scientific problem(s) they are being used for, the challenges of using these cutting-edge and oftentimes proprietary technologies in their AI training and evaluation workflows versus more traditional HPC and GPU architectures, and any other useful information that provides insights into the benefits and limitations of these devices.
Both established and emerging companies have long provided products and services which make HPC effective and usable for weather, water and climate applications. The very nature of their business requires that they stay well-informed of customer needs and emerging technologies to ensure they deliver products and solutions most effective in helping solve the often complex and challenging problems HPC users face. Industry perspectives on new architectures and capabilities are therefore often enlightening on what the future holds and where trends of the overall HPC market are heading. To gain insight into these future directions, this session welcomes talks from representatives of industry and research organizations advancing the state of the art in high performance computing, with an emphasis on future directions that will benefit weather, water and climate applications. Talks can address advances in overall system architectures, advances in components such as processors, accelerators, interconnects, memory, and storage, software advances, cloud computing, or other evolving and emerging market-based priorities.
The use of GPUs as general purpose accelerators has continued to grow as data center class GPUs become more powerful, easier to access, and include larger available on-board storage to support larger problems. As of the latest Top500 list, 9 out of the top 10 largest supercomputers in the world now heavily depend upon GPUs for a large proportion of their maximum and peak FLOPs. It is therefore prudent that scientists and engineers developing and maintaining HPC applications continue to consider how best to utilize these accelerators and achieve maximum performance on current and future HPC systems.
This session welcomes submissions of all types exploring or currently exploiting the additional power of GPUs for weather, water, and climate applications. Example topics of interest include, but are not limited to, the benefits, drawbacks, and challenges of developing new or porting existing libraries and applications to GPUs, benchmarking new and existing applications across various vendors or generations of GPUs, profiling of applications on GPUs to realize peak performance, the adoption of performance portable frameworks and vendor-agnostic programming methods, new and innovative algorithms best suited for GPU architectures, and educational efforts to promote or expand the use of GPUs among students, faculty, or researchers at a university, laboratory, or company.
Cloud computing has become pervasive across computational domains, including for providing HPC resources to the Weather, Water and Climate community. The cloud offers benefits in cost, ease of use, ease of access, and flexible provisioning, but there are challenges to gaining these benefits and the cloud will not be appropriate for all workloads. This session solicits papers presenting experiences in using HPC in the cloud for Weather, Water and Climate applications as well as analyses addressing suitability of cloud platforms for delivering HPC to Weather, Water and Climate applications. Abstracts are invited that address challenges and lessons learned from the experience, including issues of performance/optimization, availability, scalability, portability, data storage, data transfer and security and are also invited to address tools and techniques for HPC in the cloud. Abstracts are invited from users and providers of HPC in the cloud.
Python has become pervasive in scientific computing across academia, government, and industry. This session will focus on the use of Python as a language for development, implementation, and execution of applications on high performance computing platforms. Papers are invited addressing all aspects of the topic including success stories and lessons learned, tools and libraries, experience with on-premises and cloud platforms, application performance, and application portability.
The computational resources required to perform global weather and climate modeling have always been demanding. Given the need for kilometer scale resolution, enhanced physics, O(100) ensemble members, multiple forecast horizons (i.e. short-range, medium range and long-range) etc., this computational demand can be perceived as infinitely large. A multidisciplinary, cross-industry collaboration between software, silicon, system manufacturers, providers, academia etc. is required to meet ever increasing demand promptly and efficiently. This session invites abstracts on any innovative/emerging technology, innovation, service, etc. that will be useful in shaping and enabling the future of computing for weather, water, and climate modeling.
This session will feature talks addressing advances in algorithms and computational techniques that lead to advances in the state of the art for High Performance Computing applications in Weather, Water, and Climate. Benefits of emerging novel/innovative techniques may include increased performance, enhanced scalability, reduced cost and others. Approaches addressed could include, but are not limited to, exploiting mixed-precision arithmetic, new solvers, novel architectures and many others and can be hardware, software, or firmware solutions.
Quantum computing holds great promise for dramatic increases in computing speed for a wide range of applications. While this is still an emerging technology, progress is being demonstrated in many areas. This session invites papers on the application of quantum computing in weather, water, and climate applications. Papers are invited on specific applications as well as on foundational work in underlying algorithms and techniques.
Accessing multiple HPC systems on-prem, remotely at partner institutions, and/or in the cloud for a single project is much more common now and opens up the ability for researchers to utilize those resources most suited to their work rather than those already readily accessible. However, the constraints on hardware, data storage resources, and software choices at each unique site and system can cause additional burdens to completing work efficiently. Federated computing seeks to alleviate these issues by creating more automated means by which to share credentials, handle data movement, track software dependencies and environments, and access unique hardware more easily independent of location by abstracting the machine and system specifics behind a user-defined resource requirement specification used to orchestrate the workflow desired. The overall goal is to simplify how HPC applications operate between on-premises, remote, and cloud systems as needed based on the requirements of the workflow to be executed. In this session, submissions will focus on efforts to develop, use, or improve workflows or management systems for federated use of on-prem, remote, and/or cloud HPC systems. Submissions are encouraged to highlight the many challenges involved with federated HPC systems and workflows, and the solutions proposed or employed to overcome them.
This session is designed to allow researchers with innovative work utilizing High Performance Computing for Weather, Water, or Climate Research a place to showcase that work, even if it does not fit into any of the pre-defined sessions already listed. To truly grow our community, we need to see what that community has accomplished, what they are currently working on, or what they feel they want to discuss with the community for potential new research efforts related to building new or utilizing current HPC resources. Abstracts submitted under this general heading will be grouped when possible with submissions on similar topics during the review process.
Joint with 33rd Conference on Weather Analysis and Forecasting/29th Conference on Numerical Weather Prediction
High performance computing continues to advance with the incorporation of powerful accelerators and larger and faster storage systems. These increased computational capabilities support the nesting and coupling of combinations of climate, global, mesoscale, and turbulence-resolving models and allow for investigations of flow features and their influence across conventional model scales. These performance increases may be further augmented by AI models which have been shown to be able to further enhance model capabilities. Therefore, this session seeks submissions which focus on the utilization or enhancement of multiscale modeling frameworks to study the inherent relationships that exist across the scales of atmospheric motion as well as the benefits that trends in high performance computing have had or may have in the future towards advancing the use of multiscale modeling approaches.
Joint with 33rd Conference on Weather Analysis and Forecasting/29th Conference on Numerical Weather Prediction
Numerical weather prediction and high performance computing have a long history of working synergistically to push the boundaries of weather forecasting. As HPC resources grew and advanced, NWP methods and algorithms were able to take advantage of those enhanced capabilities, which further informed what HPC resources were needed for future advances in NWP. To continue this mutually beneficial relationship between these two fields, this session seeks submissions regarding the following topics that are relevant to modern advances in both HPC and NWP. In addition to the specific topics listed below, general submissions that include significant NWP and HPC crossover are welcome.
1.) Harnessing HPC for Urban and Convective Scale Modeling
2.) Acceleration of NWP Models and Parameterizations with GPUs
3.) Profiling and Addressing Problems of Scale to Prepare NWP Models for Exascale Systems
4.) Data Management of NWP “Big Data” Inputs and Outputs for Processing, Transport, Storage, and Archiving
Joint with 24th Conference on Artificial Intelligence for Environmental Science
The size and scale of AI models has grown quickly with available computing power and storage resources, and the beginning of the exascale era of supercomputing (massive HPC systems capable of 1018 floating point operations per second) further expands the boundaries for how large AI models can be. With resources and systems this large, the prospects of data movement, storage, and analysis at this scale provide both opportunities and challenges. This session will focus on AI development projects targeting, actively preparing to scale up to, or currently using these largest HPC systems for weather, water, or climate applications. Submissions are encouraged to address not only their scientific goals, methods, and achievements, but also the computational, data storage and management, and/or performance scaling challenges they encountered or expect to encounter when using exascale resources to achieve their scientific goals.
Joint with 16th Conference on Environment and Health
Numerical predictions and projections of weather and climate can allow public health professionals the ability to prepare for changes in weather-related illnesses such as those caused by extreme heat, changes in air quality, and the likelihood of environmental conditions which promote the transmission of vector borne diseases. Continual improvements in high performance computing resources and advances in conventional weather, climate, ocean, and hydrological models as well as newer data-driven AI modeling approaches allow projections to be more accurate over longer periods of time and with enhanced detail, helping to better target limited public health resources to areas of greatest need. This session seeks submissions that depend upon or further improve the relationship between HPC resources, conventional and AI data driven models, and assessments of public health impacts and resource needs. Example submissions might include, but are not limited to, utilizing GPUs and large HPC storage resources to develop new AI data driven forecasts, scaling up ensembles for clearer probabilistic guidance on future regional or local health impacts, or developing new models or analysis tools that require HPC to quickly and efficiently process datasets of weather, climate, and/or public health information.
Joint with Fourth Symposium of Community Modeling and Innovation
Community modeling has emerged as a major strategy for developing and exploiting earth system models. Initiatives such as NOAA’s Earth Prediction Innovation Center strive to foster the development of tools and infrastructure to support and enable community modeling. This session solicits abstracts on innovative computing approaches that will further support both developers and users of community earth system models, including high performance computing, use of AI/ML, exploiting GPUs and other accelerators, software tools, and other new and emerging technologies.
Joint with 13th Symposium on the Weather, Water, and Climate Enterprise and 3rd Symposium on Environmental Security
Recent extreme weather events have increased awareness of the impact of climate change on the earth’s ecosystems – and the resulting impacts on society as ecosystem services such as food, water, security, and shelter are under increasing stress. This session invites abstracts on developments in ecosystems modeling and ecological forecasting that are enabled or facilitated by high performance computing and other technologies that enable these applications to provide insight into changes in the natural environment and the impact of these changes on society. Abstracts are invited on topics such as infrastructure and computational support, development and use of digital twins, use of cloud computing resources in ecological forecasting and for ecosystem modeling, and software tools to facilitate development of these models, as well as other emerging tools and techniques.
Joint with 24th Conference on Artificial Intelligence for Environmental Science
With the rapid pace of AI model testing and development and the option to run large ensembles of models to find the best combinations of model super-parameters, enhancing performance and utilization of high performance computing resources by the AI model training and testing processes can greatly reduce the time necessary to find the “best” model. Therefore, this session will look at efforts to investigate the tools and libraries used for training and evaluating AI models with an emphasis on identifying inefficiencies or introducing new algorithms that could lead to further decreases in time to solution or increase efficient utilization of HPC resources for greater scalability and higher data throughput. Examples of submissions would include but not be limited to enhanced utilization of computational resources such as CPUs and GPUs, optimal use of high performance storage and file systems to increase data throughput, fully utilizing available bandwidth of or optimizing communication patterns across interconnect networks, and finding areas to overlap communication, storage, and/or computation to reduce process idle time.
Joint with 33rd Conference on Weather Analysis and Forecasting/29th Conference on Numerical Weather Prediction and 24th Conference on Artificial Intelligence for Environmental Science
Interest in artificial intelligence or data-driven weather forecasting models has seen impressive growth recently in light of results showing forecasting skill on par with traditional prediction models, but with much shorter runtimes. This development has opened prospects for much larger forecast ensembles, more rapidly updating forecast cycles, and higher resolution forecasts over larger forecast domains than currently possible with traditional NWP models on many current HPC systems with much fewer resources.
However, with these prospects come the potential HPC challenges of shifting more compute resources to accelerator-based designs as well as managing, post-processing, and disseminating the much greater levels of input and output data expected in much shorter periods of time, potentially straining HPC storage and data transport network systems. Therefore, this session seeks to explore efforts to implement AI NWP models in current HPC environments. Submissions are encouraged to give insight into the model selection process, the development of workflows built around these methods that handle the data management process, and/or the challenges that were encountered while developing and implementing the end-to-end forecast process.
For additional information, please contact the program chair: Marc Cotnoir ([email protected]).