Planning and preparation

Owned by Hui Wan
Last updated: Jul 29, 2024
5 min read

 

 

Additional info on this page

Important dates

BER Participant list

BER Breakout Agenda for SciDAC Meeting Site, see details and instructions in the detailed Agenda below : 2024-07 SciDAC5 PI Meeting & BER Breakout

BER Breakout Session Agenda

Wednesday, July 17, 2024

Schedule for the day

  • 8:30 AM ET Partnership Breakout Session 1 (see below)

  • 10:30 AM ET Break (15 min)

  • 10:45 AM ET Partnership Breakout Session 2 (see below)

  • 12:45 PM ET Lunch (Networking)

  • 1:45 PM ET Group Photo session (Location TBD)

  • 2:30 PM ET Poster Session (In Person only)

  • Afternoon Break: set up in poster session area

Session 1 Agenda

8:30-10:30am

Talks/topics

Notes

Talks/topics

Notes

Dorothy' remarks (5 min)

 

Gary’s remarks

 

Xujing overview/intro (10 min)

 

Project updates (70 min)

10 minutes each

Participants: Each project should prepare a 10 min project update. Please upload pdf/ppt next to your project name on the left before the session.

  • NOT a complete overview - assume other participants are familiar with your project

  • Suggested format:

    • 1 slide on high level project overview

    • 3-4 slides of project highlights over the past year - either actual highlight slides submitted to BER or other activities at that level. Focus on the BER/ASCR collaboration part of the activity and less on the scientific details. Can include significant capabilities you have developed that may be of interest to other projects.

    • 1-2 slides on specific challenges, change in scope. It’s ok to admit adjustments - these are high risk projects and we will all learn from how each of us have had to modify our plans.

Q&A (10 Minutes)

 

Discussion (30 min)

Led by Mauro Perego & Shinjae Yoo

  • collaboration between domain scientists and computational scientists

    • Perspective of comp. scientists on how working with Earth system projects compares with other application areas in which they work. What can be learned from other subject areas.

Google Doc for notes and comments

Participants, please add discussion topics/questions here before the start of the meeting:

  • Challenges in starting and maintaining a collaboration:

    • success and failures stories

    • difficulties in keeping people engaged (on both sides)

    • Tension between domain scientists' goals and computational scientists' goals. How to find a good balance

    • What concepts or terminology is confusing or could use more specific definitions?

    • What could be done better?

    • How long does it take to develop a successful collaboration between domain and comp. scientists?

  • What are some quirks or unique ways of doing things that differ from other applications areas?

  • For comp. scientists, what do you like about working with Earth system modelers?

Session 2 Agenda

10:45am-12:45pm

Google Doc for notes and comments

Talks/topics

Notes

Talks/topics

Notes

Presentations about future E3SM Plans (45 min)

10-15 min each with questions

  • Wuyin (atm, coupled)

  • Luke (ocean, sea ice, land ice)

  • Gautam (land, river)

These talks are meant to give SciDAC projects a view to what is coming for E3SM and how their projects fit within expected plans in E3SM.

  • expectations for transition plans for scidac capabilities into E3SM

Discussions (1 hour and 15 minutes)

Participants, please add discussion topics/questions here before the start of the meeting:

  • Collaboration/coordination between SciDAC and E3SM (30 minutes)
    (moderator: Matt Hoffman)

  • How to deal with SciDAC projects working on different components (newer or older) than mainline E3SM?

  • Tensions between high-risk model development in SciDAC and low risk tolerance of production model.

  • AI/ML discussion (30 minutes)
    (moderator: Hui Wan)

  • How are our projects using or developing AI/ML methods? (See simple survey in Google Doc)

  • What ML applications have worked well for your project? What have not?

  • computing resource discussion (15 minutes, time permitting)
    (moderator: Sarat Sreepathi)

  • What machines do you primarily use?

  • Challenges coordinating machine usage between SciDAC and E3SM ?

(Prepare for ~12 min report-back to plenary)

 

Posters

Poster sessions:

  • Tuesday, July 16: 3:15-5:15 pm

  • Wednesday, July 17: 2:30-5:00 pm

Poster guidance

  • Each BER SciDAC project has 2 or 3 poster slots.

    • PAESCAL, IMPACTS = 3 posters

    • All others = 2 posters

  • Unlike last year, where we gave specific guidance on poster topics, this year use your posters however you see fit. A high level project overview poster is not required (but could be used). Some ideas:

    • focus on specific exciting results

    • discuss ASCR/BER partnership successes

    • describe a computational tool that may be of interest to other projects

    • short summaries of progress in various project areas

    • high level project overview

    • etc.

Logistics

Participants, don’t forget to submit poster title and abstract to the meeting site:

Planning notes for BER breakout (participants can ignore)

Thoughts from the first planning meeting on Apr 15, 2024 (Xujing, Matt, and Hui):

Many participants of last year’s meeting and even earlier SciDAC PI meetings have provided the feedback that discussions were more useful than presentations in general. Therefore, we will shorten the “presentation by project” part of the session (especially since comprehensive overviews of the projects were already presented last year) and leave more time for discussions on some topics of special interest.

Possible topics of discussion during BER breakout:

  1. Use of AI/ML.

    1. How are our projects using AI/ML currently? Are there any planned explorations or explorations that we wish to carry out but need help for?

    2. Among the many different research activities that can benefit the development, evaluation, and application of E3SM (e.g., understanding observations, constructing fine-scale process models, developing parameterizations, harmonizing components in large-scale models, tuning uncertain parameters, identifying model deficiencies, etc.), are there sweet spots - or known bad hit places - for the use of AI/ML techniques?

  2. Collaboration between SciDAC projects and the E3SM project.

    1. What are E3SM’s development plans in the next years? We plan to get an overview from SciDAC project team members who are also on E3SM’s leadership team.

      1. CMIP7 scenarios / new model configurations

    2. In the context of E3SM’s planned changes in programming language as well as the anticipated changes in HPC software and hardware, how can we better position SciDAC projects for the integration of their products into E3SM?

  3. Collaboration between domain scientists and computational scientists.

    1. Collaborations between domain scientists and computational scientists are known to be very challenging. Colleagues who have gone through more than 1 SciDAC cycle have had the general impression that fruitful collaborations require commitment from both sides and willingness to make long-term investment. Do we have success stories and lessons to share? Are we facing any specific challenges? Are there ways to further facilitate the interdisciplinary collaborations?

  4. Computing resource discussion

    1. Compare notes on: what machines we run on, what computational budgets we need/use, tips and tricks for maximizing resources (e.g. queue strategies)

PIs and BER project participants, please feel free to add your suggestions or comments here.