18/08/2026
Participants: 11
10-year roadmap for ACCESS
(AI generated notes)
Quick recap
The meeting focused on developing a 10-year roadmap for ACCESS with respect to land modelling research infrastructure, with @alexnorton leading the discussion as part of the Scientific Advisory Committee for ACCESS-NRI. The group discussed strategic priorities including scientific success metrics, model performance versus breadth, technical improvements like machine learning integration, community building, and the need for better organization of the CABLE model codebase. Key considerations included balancing Australian-specific applications in water and carbon cycling with broader scientific needs, coordinating with other modelling communities (JULES, LPJ), and addressing structural challenges with the CABLE model’s multiple versions and complexity.
Summary
ACCESS-NRI Roadmap Planning
Alex introduced a 10-year roadmap project for ACCESS-NRI, seeking input from the land modelling community ahead of a September meeting in Canberra. Alex likes the structure of the current strategic plan with goals, strategies, and outcomes, with considerations including scientific success, process complexity, spatial resolution, community building, and technical advancements. Alex plans to create a survey with multiple choice questions about priorities and limitations in land modelling, which will be shared on the Hive forum for community input.
Performance-Focused Land Surface Modeling
Gab discussed focusing on performance rather than breadth in land surface modeling, arguing that the community could leverage their benchmarking capabilities and machine learning expertise to develop a high-performing model similar to ECMWF’s approach. He emphasized the importance of using evaluation systems that can identify specific improvements and distinguish between observational and model errors. Gab mentioned he would prepare a talk for an upcoming workshop to further elaborate on this approach.
The group discussed science priorities for Australia, with Anna and Gab highlighting water and carbon as key areas due to good observational constraints. Claire emphasized the need for a unified and better-organized version of the cable model, suggesting a significant structural rethink rather than just small improvements.
The group discussed challenges in applying machine learning models across different geographical regions, with Chermelle expressing concerns about using Australian sites for broader applications. Gab suggested testing models using out-of-sample approaches, citing examples of successful cross-regional performance. He explained how high-resolution temporal data could effectively cover large areas despite limited spatial sites. Ian highlighted the need to balance the community’s own interests with broader ACCESS community demands, noting potential conflicts with other models like CCAM and emphasizing the importance of maintaining inter-model consistency.
Australian Land Modeling Collaboration
The group discussed collaboration between different land modeling communities in Australia, including users of CABLE, JULES, and LPJ-GUESS models. Claire suggested bringing these communities together, while Ian noted the importance of not discounting JULES users since they receive it for free as part of atmospheric coupling. Anna mentioned plans for an Australia-wide evaluation across CABLE and JULES, with potential to include LPJ-GUESS users, and Gab suggested running all models through the benchmarking suite.
Alison offered to provide documentation of previous working group ideas from past meetings.
Land Modeling Survey Planning
The team discussed planning for a land modeling survey and roadmap, with Alex leading efforts to capture existing priorities and processes from previous years’ documentation. Ian emphasized the need to align future priorities with broader CSIRO strategic directions and avoid potential conflicts with other parts of the organization.
The group also reviewed current working group projects, including discussions about ancillary data needs for CABLE and JULES models, particularly regarding nitrogen and phosphorus sources.
Next steps
- Alex: Put together a survey with limited questions about land modelling priorities and share it with the community in the next couple of days
- Alex: Share the synthesized survey results on the Hive after collecting inputs
- Juergen: Point Lachlan to the new BNF product that should replace the existing one in the gridinfo