Land Group Agenda, November 2016 all-hands

Owned by Peter Thornton
Last updated: Nov 10, 2016 by Daniel Ricciuto
3 min read
  • Discuss coupled model metrics (ILAMB - 30 minutes, others)
    • Suggest that all the groups do something similar.
    • New diagnostics on the human/ energy side?
  • Review our planned papers (10 min)
  • Live demo from Renata on quarterly reporting updates (10 min)
  • Update on compsets (5 min)
  • V2/v3 roadmap update discussion (10 min?)
  • Planning for Ruby's session on papers
    • 8 min (Thurs, 10 Nov, 10:49-10:57am)
  • Update v2/v3 roadmaps (For session on Thurs, Nov 10, 3:15-3:45pm)
  • Known topics for speed-dating?
    • Add Coupled Group, discuss diagnostics, metrics.
    • Workflow?
    • CIME?
  • CMDV-Land - 10 min ()


Discussion notes 

  • BGC experiment 
    • V1 deliverables are water cycle experiment and DECK, BGC is lower priority.  High computational demand.
    • Timeline is still unclear.  We need some water cycle output to begin.  ALCC resources are not likely to be available this year, next cycle starts in July.
    • Need to investigate how to to bundling on Mira and Titan.  OLCF supported tools and training are available.
    • We can do offline land experiments/coupled testing in the meantime on institutional clusters, Eos.
  • ILAMB diagnostics (presentation by Forrest M. Hoffman (Unlicensed))
    • Several versions of ACME have been run through ILAMB, including early versions of ALM (without P), and the water cycle experiment.
    • Different forcings evaluated (CRU-NCEP and GSWP3)
    • Scoring is based upon weighting for different datasets, but obs uncertainty is not yet considered.  We rarely have real characterization of uncertainty (often, it is only methodological).  However, considering uncertainty is important to avoid overtuning.
    • Can compare all models using different metrics on the same page.  Works on mobile devices!
    • Pre-defined regions available for analysis
    • Works on Edison, rhea, and package is available to install anywhere
    • This could be useful as a sanity check for model commits in the development phase as well as rigorous scientific evaluation.  We could have tiered levels of testing.
    • Demography-relevant variables?  Biomass, burned area.  New fields can be easily added, if units are reasonably CF compliant.
    • Building tie-ins to Obs4MIPS, ESGF.  One idea, if resources permit, is to build automatic capability to run diagnostics on CMIP6 models as they are submitted.
    • Observation products often change.  Considering long-lived versions of observed datasets.
    • ILAMB workshop report is nearly finished and includes many ACME-relevant ideas (e.g. testbed, new datasets, UQ).  
    • ACME effort on ILAMB should focus on testing features that are new/unique to ACME.  
    • We could use ILAMB with perturbation experiments to evaluate responses.  Protocols (e.g. FACE) could be integrated.
    • Action item - see benchmarking datasets /wiki/spaces/LND/pages/2523374, prioritize these and add others as necessary (Especially human/energy related.  Crop yield, stream flow/temperature, flood frequency)
  • Quarterly reporting demo (presentation by Renata McCoy)
    • Meeting page automatically links information from the quarterly report page.
    • Each person could enter a short summary in "excerpt" box on quarterly report page - this is linked to a Macro that automatically dates and populates the meeting notes page.
    • The box can be cleared before each meeting and archived under "summary of progress".
    • There will be a new meeting notes template that incorporates this capability.  It can be copied for each new meeting and the date can be updated.
    • Replace macro with text from previous meeting.
  • Bill R. suggests 1-2 10 minute science talks on each land group telecon.  Could be a new result, upcoming pub, conference presentation.
  • Planned papers
    • We had a list of 40 last time - too many.  Linked under last year's agenda.
    • Focus on highest priority ACME needs.  
    • ~3 from each group/lab, more if papers underway. 
    • Generally restrict to something reasonably achievable within 1 year, which means more than just an idea/outline.
    • We do need to make sure we're developing a plan/coordination for the critical papers from high-profile experiments.
    • Make sure highlights are being submitted to Dorothy's program page.
  • V2/V3 roadmap
    • V2 development tasks put together about a year ago, but much of this is science development.
    • V2 will focus on architecture/software, so much of this proposed work will go into V3.
    • June/July for initial V2 code review, PRs November, integration by Jan. 2018 (timeline uncertain)
    • Long-lived physics branches could be an issue, topic for SE speed dating?
    • Action item:  Before 11/22, Review V2 tasks, institutional leads make sure what we are working on is listed, and mark tasks as V3 as appropriate.
  • Potential speed-dating topics/issues
    • Ocean biogeochemistry/performance
    • Slowdown (up to 10x) of ACME/CESM on NERSC
    • Atm/land subgrid
    • Dynamic landunits (high priority - can SE help with this?)
      • 6 months+ of effort
      • What is being blocked by this and what are the workarounds?
      • Focus first on I compsets
      • Should also think about how this will interact with FATES, new belowground developments
      • Other land use types (urban, glacier) complicate the balance of C/N since they don't track this.
      • Short-term solution - run crops separately (offline) to give baselines. 
      • Program resources to support a dedicated software engineer?
    • GCAM coupling (using the coupler)