Annual CSDMS Group Meetings - What happened?

General
1

Sharing some content here as a record of Group Meetings at the CSDMS Annual Meeting. Starting with the Terrestrial, Geodynamics, Hydrology, and CZ groups, but welcome other groups too!

3 Likes
2

From 2024: “Share a link (group page, profile page, repository, publication) of some work that you are involved in or that inspires you.”

2 Likes
3

Notes from 2025, Terrestrial, Hydrology, Geodynamics, and Critical Zone: (Thanks, Wolfgang!)

Social media: Very diverse set of platforms. People are pretty much everywhere
these days. Empirically, CSDMS has seen the scientific community disappear
from Twitter/X; the community is not as active on Bluesky or anywhere else as
it was.

Code review: Models that are reviewed are listed higher on the CSDMS website.
Models and code can be published via the Journal on Open Source Software
(JOSS). Many respondendents (70+%) would like more resources on reviewing code.

On the use of AI as a coding assistant: About 75% have used them. (Last year:
55%.) Most common tools: ChatGPT, CoPilot. One comment was “It’s ok to use,
but you’re shortcutting your learning, and the resulting code is generally at
best mediocre”. Other comments: “Clearly mark for which assignments AI use is
encouraged/acceptable/forbidden; enforcement is perhaps challenging.”
“Students will use AI as tools, and that’s ok. Let’s make sure students can
explain their code in comments, and also have the skills to write out
algorithms on paper.” “AI is useful.”

Things people are inspired by: Lots of things – concrete links will be posted
on the forum.

Things we could do to help the community: Webinars, building mentoring
connections, updates from the working groups, pointers to new or interesting
papers, co-working sessions, facilitated code reviews. Build your own
community of people who help each other, share experiences, reflect on what
you did and that worked or didn’t work. Long list of other suggestions on
Slido.com.

Resources you wish were available: Job/professional opportunity sites. Places
to learn best practices/software development (like Software Carpentry).

Links:

1 Like
4

Thanks for the notes :slight_smile:

I definitely feel this is true! Anecdotally, it seems like people have left Twitter and fragmented to either Bluesky, LinkedIn, Mastodon, or given up on social media entirely. Roughly speaking, I feel the more “techy” people have gone to Mastodon, the more “industry/professional” people have gone to LinkedIn and the more “academic” people have gone to Bluesky, but that’s completely anecdotal.

5

It’s t minus 2 days to the next annual CSDMS in-person group meetings! What do you wish you could ask your fellow CSDMSers? We’ll ask participants to share links of their work for inspiration again, and have a few repeat topics to look at trends, but are there new questions you would like to introduce? Let me know! @awickert @Castronova (other group leads are not auto-completing! I will find you!) @bangerth, @sfgallen, sagargautam @nicgaspar (alumni :slight_smile: )

1 Like
6

2026 edition:

Share a link (group page, profile page, repository, publication) of some work that you are involved in or that inspires you.

What is something you are trying to accomplish but you haven’t figured out yet?

  • Python/R integration for first time modeling programmers in the field of hydrology
  • Doing cool stuff with repeat LiDAR point cloud data easily. (non-raster) differencing, different returns, uncertainty.
  • Physics of organic rich (but with minerals) soils and sediments to understand their role in landscape evolution. (permafrost soils/tropical peat)
  • Representing a buried surface that changes depth appropriately with erosion + isostatic/other uplift and is associated with properties (e.g., resistance to erosion). Having trouble doing this with LithoLayers
  • Access to the most recent future gridded climate data and tools for downscaling that data to a unique project area
  • More aeolian process components, especially dust/loess transport and downwind grain size/deposition rates
  • Making videos of landscape evolution
  • Facilitated data finding and downloading (multiple types of data)
  • Get landlab to run faster and in parallel, and include nonlinear inversion algorithms to help invert DEMs and analytical data to constrain surface process model parameters and other relevant information (e.g., tectonic history)
  • Familiarity with CSDMS tools for my own research purposes/resources for modeling hydrodynamic problems
  • Are there Landlab models calibrated to “extra”-terrestrial needs?
  • The group discussed better ways to accelerate codes (MPI, parallelism, GPU…)
  • Can we get landlab to pull lithologic data from macrostrat? Can we then map erosive or other processes to certain lithologies or structural features?
  • Is there any sort of guidance available/best practices for applying nested or multiscale grids? For instance, to scale or translate reach-scale physics through a river network?
  • Viability of Coupling Models that work on different spatial and temporal scales.
  • Train statistical surrogates / emulators on heavy physics-based models so we can run many of them quickly
  • It would be nice to have more guidance for writing tests
  • How to utilize AI for model parameterization and/or AI informed models
  • Best practices for coding (commenting, efficiency etc)
  • develop my own model (I am just a grad student), codes to GPU!
  • access to future climate meteorology datasets, gridded at 1 to 5 km resolution
  • tools to downscale coarse gridded meteorology data using established methods like the PRISM (Daly et al 1994) or MALA (abatzoglov & Brown 2012) approaches
  • How to contribute updates to components from Landlab (how do I know if they want my own custom updates/changes)?
  • physics-informed machine learning
  • How to convince my advisor to integrate Landlab into their existing research workflow.
  • High-res meteorologic data (and easier to access); low relief area tools, like agricultural fields with tiles, best practices for organizing larger (?) code practices.
  • Integrated data analysis from webmap portal like for volcanology talk we saw this morning. i.e. define region, pull topo data, condition for analysis, select from common models linked by BMI
  • How to BMI code
  • How to utilize AI for model parameterization and AI-informed (?) models
  • Supplying context files for LLM-assisted coding (to set customized rules, guidelines, etc.)
  • Landlab → depth is missing (soil depth)
  • Better data integration
  • Waterfall versus knickpoint formation / Detection, waterfall face erosion, Connection between microscale waterfall dynamics and reach scale profile response
  • Build a landscape evolution model that centers on the role of organisms, particularly soil fauna.
  • Realistic permafrost soil, hydrology, geomorph on hillslopes

Four additional polls in the carousel below, use arrows to navigate.

Which groups are you associated with?
What social media or other community platforms do you use?
Which of the following applies to you (regarding code review)
What AI Coding assistants have you used?

slido-which-groups-are-you-associated-with
slido-which-groups-are-you-associated-with3040Ă—1852 167 KB

slido-which-social-media-or-other-community-platforms-do-you-use
slido-which-social-media-or-other-community-platforms-do-you-use3040Ă—2068 326 KB

slido-code-review-which-of-the-following-applies-to-you
slido-code-review-which-of-the-following-applies-to-you3040Ă—1852 291 KB

slido-which-ai-coding-assistants-have-you-used-submit-separate-responses
slido-which-ai-coding-assistants-have-you-used-submit-separate-responses3040Ă—2068 315 KB

Links/Resources for Code Review:

Q: Are there tools that use AI to help with code review?
A: There is copilot built into the GitHub interface. Use at own risk.

Q: How can we motivate others/ourselves to do code review?
A: Looking at other people’s code helps us to learn how to code better and teach us new things.

Tip: Getting a code reviewer early will help for them to review smaller chunks of code at a time, instead of the huge monolithic codebase at the end of the project.

Suggestion: Can we make checklists for Landlab-related or CSDMS-relevant code-review, and make them easier to find? (Currently there is a list but you may not see it until you initiate a pull request to see it.)

2 Likes