This December, AGU25 returns to New Orleans with the theme “Where Science Connects Us”.

CMIP community events

CMIP@AGU community meetup will take place directly after the CMIP Town Hall at 14:00 outside of room 272.  Town Hall is on Tuesday,13:00-14:00  (TH23J), Room 272-273

Fresh Eyes on CMIP BYO lunch will take place, 13:30 at the New Orleans Theatre C immediately before the CMIP forcings session which takes place on Wednesday – join the forcings session 14:15-15:45 in the New Orleans Theatre C, further details below.

CMIP sessions and events

(Off-site) Earth System Grid Federation 2 – US Project Meeting and ESGF Tutorial

An opportunity for CMIP community members to get an overview of ESGF developments – the new ESGF NG and the tools the ESGF team are putting together to support analysis and research around CMIP activities. The session include hands on tutorial covering Metagrid, Intake-ESGF and discussion on the new STAC catalogue, as well as an opportunity to directly access technical support from members of the ESGF development team. In person and online registration is essential prior to the event: https://bit.ly/2025_ESGF_Meeting. There is an optional dinner just before the event, at your own cost, that provides an opportunity to meet and network with other participants ahead of the session.

• Monday 15th December
• 18:30- 21:30 CDT  | 23:30 – 02:30 UTC
• New Orleans Marriott Warehouse Arts District 859 Convention Center Boulevard
• Optional dinner will take place 17:30 at Mulate’s, The Original Cajun Restaurant, 201 Julia Street, New Orleans.

A CMIP7 Update: Forcings, Data Request, the REF, and More (TH23J)

Over four decades, the Coupled Model Intercomparison Project (CMIP) has advanced Earth system modeling significantly, addressing a wide range of scientific questions through 24 CMIP6 MIPs. Despite successes, CMIP6 also highlighted challenges for modeling centers, data infrastructure, and users that must be addressed. To meet these needs, the CMIP Panel developed the CMIP Assessment Fast Track—an innovative component of the CMIP7 experimental design. This streamlined set of experiments, designed to meet IPCC Seventh Assessment Report (AR7) and numerous national assessment requirements, prioritizes critical simulations for climate assessment and policy, distributing project burdens on modeling centers and improving computational efficiency. As CMIP7 simulations begin in 2025, several key elements are being finalized: the DECK forcing datasets were published in April 2025, and the final CMIP Assessment Fast Track Data Request (v1.2) by May 2025. Work is underway on the Rapid Evaluation Framework (REF) to support systematic assessment of submitted simulations. This Townhall will update status: forcing datasets, Data Request, REF, and activities of the Fresh Eyes on CMIP group. We welcome the community to join, share feedback, and learn about engagement opportunities for CMIP7 and beyond. Presenters include CMIP Panel, CMIP Task Teams, and Fresh Eyes on CMIP members.

• Tuesday, 16th December 2025
• 13:00-14:00
• 272-273

Link to session on AGU website

Climate Forcing: Quantifying the Roles and Responses of Natural and Anthropogenic Drivers (GC33A/GC34A)

Conveners: Paul James Durack, Vaishali Naik, Yuhan Rao, Thomas Jacques AubryEarth system variability and change arise from shifts in atmospheric composition affecting top-of-atmosphere energy balance, from land use changes altering surface albedo, and from ocean surface temperature and sea ice change among other “forced” perturbations. These natural or anthropogenic change drivers are termed climate forcing agents. This session highlights research quantifying uncertainties in the evolution/impacts of these agents using Earth System Models and observations. We invite contributions across all aspects of forcing research, including the development of historical and future forcing time series, idealized analyses, single- or multi-model frameworks, or observational methods to assess their influence on Earth system dynamics. We are especially interested in studies examining responses to forcing over time, using next-generation (CMIP7), or earlier CMIP phases (CMIP6/CMIP6Plus). Submissions that explore interactions across multiple components of the Earth system—ocean, atmosphere, cryosphere, land surface and subsurface, and biosphere—and provide evidence of linked or coherent forced responses are highly encouraged.

1. Oral session 1: Wednesday 17 December, 14:15 – 15:45, New Orleans Theater C
2. Oral session 2: Wednesday 17 December, 16:15 – 17:45, New Orleans Theater C
3. Poster session: Thursday 18 December,  08:30 – 12:00, Hall EFG

Confronting Earth System Model Trends with Observations (GC43B)

Conveners: Tiffany Shaw, Isla Simpson, Robert Wills, Stephan Po-Chedley

Abstract: Anthropogenically forced climate change signals are emerging from the noise of internal variability in observations, and the impacts on society are growing. For decades, Climate or Earth System Models have been predicting how these climate change signals will unfold. The climate science community is now in a position to confront the signals, as represented by historical trends, in models with observations. We welcome contributions that take stock of the ability of models to capture recent trends across all Earth system components. Contributions highlighting successes and discrepancies, robust procedures for confronting observed and modeled trends, moving beyond quantification into understanding the origins of historical trends in the observational record and/or models, and cutting-edge methods (e.g. Machine Learning, kilometer scale models, hindcasts) for identifying sources of discrepancies and separating forced signals from internal variability are all welcome.

• Oral session 1: Thursday 18 December, 14:15 – 15:45, New Orleans Theater C
• Oral session 2: Thursday 18 December, 16:15 – 17:45, New Orleans Theater C
• Poster session: Friday 19 December,  08:30 – 12:00, Hall EFG

CMIP presentations

An evolving Coupled Model Intercomparison Project phase 7 (CMIP7) and Fast Track in support of future climate assessment (GC34A-01)

Authors: John P Dunne, Helene Hewitt, Julie Arblaster, Frédéric Bonou, Olivier Boucher, Tereza Cavazos, Beth Dingley, Paul James Durack, Birgit Hassler, Martin Juckes, Tomoki Miyakawa, Matthew Mizielinski, Dr. Vaishali Naik, Zebedee Nicholls, Eleanor O’Rourke, Robert Pincus, Benjamin M Sanderson, Isla Simpson and Karl E. Taylor

The Coupled Model Intercomparison Project (CMIP) coordinates community-based efforts to answer key and timely climate science questions, facilitate delivery of relevant multi-model simulations through shared infrastructure and support national and international climate assessments. Generations of CMIP have evolved through extensive community engagement from punctuated phasing into more continuous support for the design of experimental protocols, infrastructure for data publication and access, and public delivery of climate information. We identify four fundamental research questions motivating a seventh phase of coupled model intercomparison relating to: patterns of sea surface temperature change, changing weather, the water-carbon-climate nexus, and tipping points. Key CMIP7 advances include: expansion of baseline experiments; focus on CO2-emissions-driven experiments; sustained support for community MIPs; periodic updating of historical forcings and diagnostics requests; and a collection of prioritized experiments, or “Assessment Fast Track”, drawn from community MIPs to support climate research, assessment, and services goals across prediction and projection, characterization, attribution, and process understanding.

Oral presentation, Wednesday 17th December, 16:15-16:25, New Orleans Theatre C

Link to presentation

Rapid Evaluation Framework for the CMIP7 Assessment Fast Track (A24D-08)

Authors: Forrest Hoffman, Birgit Hassler, Ranjini Swaminathan

As Earth system models (ESMs) grow in complexity, there is an increasing need for rapid and comprehensive evaluation of their scientific performance. The upcoming Assessment Fast Track for the Seventh Phase of the Coupled Model Intercomparison Project (CMIP7) will require a fast response for model analyses designed to inform and drive integrated Earth system assessments. To meet this challenge, the Rapid Evaluation Framework (REF), a community-driven platform for evaluation, benchmarking and performance assessment of ESMs, was designed and developed. The initial implementation of the REF builds upon existing community evaluation and benchmarking tools and is aimed at generating and organizing diagnostics covering a variety of model variables. It runs within a containerized workflow for portability and reproducibility and to encourage widespread adoption in the community. The REF leverages best-available, CC-BY-4.0 or equivalent license, obs4MIPS compliant, observational datasets available on the Earth System Grid Federation (ESGF), to provide assessments of model fidelity across a collection of diagnostics identified and selected with community involvement. Operational integration with the ESGF permits automated execution of the REF for specific diagnostics as soon as model data are published on ESGF by participating modeling centers. The REF is designed to be portable across a range of current computational platforms to facilitate use by modeling centers for assessing the evolution of model versions or gauging the relative performance of CMIP simulations before being published on ESGF. When integrated into production simulation workflows, results from the REF provide immediate quantitative feedback that allows model developers and scientists to quickly identify model biases and performance issues. Expansion of the REF and its subsequent development and support will be prioritized by an international consortium of scientists and engineers, enabling a broader impact across Earth science disciplines.

Oral presentation, Tuesday 16th December, 17:30-17:40, Room 271

Link to presentation

The Next Generation Earth System Grid Federation (ESGF) Distributed Data Infrastructure

Authors: Forrest M. Hoffman, Philip Kershaw, Sasha Ames, Rachana Ananthakrishnan, Laura Carriere, Stephan Kindermann, Christian Page, Aparna Radhakrishnan, and Andrew Robinson

The Earth System Grid Federation (ESGF) is an international consortium that develops, deploys, and maintains software infrastructure and is the global network of enterprise data systems that employ the software for the management, dissemination, and analysis of Earth system model output and related forcing, reanalysis, downscaled, and observational data. Constructed and operated primarily in support of the WCRP Working Group on Coupled Modelling’s (WGCM’s) Coupled Model Intercomparison Project (CMIP), ESGF infrastructure catalogs, stores and delivers Earth system model output to the scientific community for research and analyses of Earth system interactions. The great majority of the data, in excess of 10 petabytes for the recent CMIP phase 6 (CMIP6), are open and freely accessible to stakeholders, industry, and the general public. To better serve the community and in preparation for CMIP7 Assessment Fast Track, the ESGF consortium has modernized the software architecture for improved performance and resilience, and is incrementally deploying new software tools and capabilities on expanding storage and analysis hardware infrastructure. System enhancements include a new core architecture based on synchronized indexes that use the SpatioTemporal Asset Catalogs (STAC) specification; a messaging queue system to manage data publication, replication, and retraction; more user authentication options; additional data access and transfer technologies; server-side data subsetting and summary product generation; and, at some regional data centers, user computing platforms that enable data-proximate computing from JupyterHub or custom-developed analysis environments. Working closely with the WGCM Infrastructure Panel and the CMIP Task Teams, ESGF expects to build out additional resources and capabilities to support research community needs for CMIP7 and beyond.

Poster presentation, Tuesday 16th December, 08:30-12:00, Hall EFG

Link to presentation

Navigating ESGF with AI: Domain-Specific LLM for Earth System Data

Authors: Daniel Saedi Nia, Elias Massoud, Bharat Sharma, Forrest Hoffman, Jitendra Kumar, Nathan Collier and Sheryl Arya

The Earth System Grid Federation (ESGF) is a globally distributed network of data centers that hosts more than 50 petabytes of model simulations and observational data critical for Earth system research. While ESGF is an essential resource, many users face challenges when attempting to navigate its interface, understanding the structure of its datasets, or retrieving data efficiently. These difficulties are especially common among users who are less familiar with programming or with ESGF’s metadata and workflows. To address these challenges, the ESGF2-US team is developing a domain-specific Large Language Model (LLM) that enables intuitive, natural language interaction with ESGF data and services. Built on an open-source foundation model (currently Meta’s Llama 3.1 8B), the LLM has been fine-tuned using custom curated ESGF-specific instruction and response pairs derived from real user-workflows, curated synthetic examples and documentation. In addition, we have implemented retrieval-augmented generation (RAG), which allows the model to access ESGF metadata and supporting documents at inference time, improving accuracy and reducing hallucinated responses. This system aims to streamline three main tasks: dataset discovery, data access, and analysis assistance. For example, users can ask what data are available for a particular variable or model scenario, generate code to search and, in future iterations, download data, or request assistance with data analysis. Preliminary evaluations show that the fine-tuned model performs significantly better than general-purpose LLMs in ESGF-specific scenarios, with improved accuracy and contextual relevance. Future work will expand the training set to encompass a wider range of ESGF user scenarios, including upcoming datasets such as CMIP7 and high-resolution observations, and will further refine integration with ESGF infrastructure. Our ultimate goal is to create a scalable and accessible interface that empowers researchers to engage more efficiently with Earth system data.

Poster presentation, Tuesday 16th December, 08:30-12:00, Hall EFG

Link to presentation