Atmosphere Group Meeting Agenda, Nov 2016

Wednesday Nov 9 1:10-3:10, Plenary Room

TimeSpeaker/ModeratorTopic
1:10-1:40Phil and Shaocheng

Organizational Issues

Go to Organization

1:40-2:10Phil

Discussion about new capabilities for V3

Goto V3 Discussion 

2:10-2:40Shaocheng

Understanding the ACME model:

Goto Understanding V1

2:40-3:10

Diagnostics Discussion

ACME_ATM_diags_discussion.pdf

Rapporteur: Susannah Burrows


Organizational Issues

  • name for the model?
  • code review process, see Code Review Process Implementation
  • what tasks are done and should be ramped down and closed as publications are completed?
    • Ozone Hole (A1)
    • CLUBB/UNICON (A2)
    • Polar Mods (A5)
    • Vertical resolution (R1 VR)
  • what tasks should be finished
    • first fully functional version of tier1b diagnostics needs to be completed
  • what tasks are active and need to continue?
    • small team deputized to suggest a re-organization?
    • regionally refined model
    • clouds/aerosols
    • interactions with surface models (more broadly than before, SMB, fluxes, etc)
    • assessment/testing/tuning (should be partitioned with specific leads)
      • short simulation strategies
      • tuning
      • UQ
      • scorecards
      • COSP/Simulators
  • what should we be doing that we are not doing?
    • Convection (replacement for ZM)
      • requirements document
      • my (PJR) inclination is to continue to consider alternatives to CLUBB
    • Non-hydrostatic dycore?
    • different vertical coordinate?
    • SCAM?
  • are we missing partnering with particular research groups or scientists?




Discussion about new capabilities for V3

See /wiki/spaces/ATM/pages/36865199

  • Note the brainstorming breakout that does not involve Atm Leads (7-8PM wednesday, Breakout #5, see Brainstorming on next generation Atm Model)
  • What model features need to be fixed?
    • pressure gradient term
    • sub-column treatment for radiation
      (convective/stratiform partitioning)
    • ?
  • What physics is missing/deficient? (Ideas relevant to ACME science, but not vetted/definitive)
    • elevation classes
      • deep dive for v2 data structures and will intersect w the CMDV software project (Ghan, Goldhaber, Salinger, Taylor, Atm GLs)
      • Intersection with parallel/sequential splitting of physics processes?
      • Surface interactions
    • cloud overlap (correlation length scales)
    • See Talking points on Radiative Transfer + Prather Fast-J → generalized shortwave code
    • consistent treatment of cloud precipitation and aerosol scavenging
    • Improvements to aerosols
      • Prognostic aerosol surface area
      • Stratospheric aerosol treatments
      • Modify MAM aerosols to allow stratospheric aerosols
      • Gas/Particle partitioning (MOSAIC)
      • SOA
      • Brown Carbon
      • Nitrate
      • Speciated Dust
    • What software capabilities are missing?
    • More Atm Chemistry? (from conversation with Ruby Leung)
      • For biogeochemistry, the v2/v3 focus is on the impacts of different carbon reduction pathways (e.g., high fossil fuel vs. high renewable energy penetration) on the carbon sink and associated impacts on water availability. We’ve identified a particular need for improving modeling of the nitrogen cycle as relate to the terrestrial C-N cycle, land use (e.g., land management/fertilizers), and related atmospheric chemistry. This topic fits in well with the biogeochemistry question for v2/v3 and we also recognized the opportunities to collaborate with university teams on topic.
      • We will continue this discussion and include others to further define key atmospheric chemistry modeling needs that are consistent with the v2/v3 science drivers. For example, one topic that still needs to be explored is the methane cycle. 





Understanding the ACME model:

Notes:

Organization:

  • The following people expressed interest in participating discussion about reorganization: Steve Ghan, Po-Lun Ma, Kai Zhang, Hailong Wang, Yun Qian, Rich Neale

  • Shaocheng’s comments:

    • Regarding re-org, Phil and I have discussed this for a long time, but have not started to work on it yet
    • Suggest Phil and I write first draft, then invite comments from others
    • Some of the things being discussed could be proposed to be part of v2 or v3.
  • Peter: why do you need a group to decide how to re-organize?  Is this just scoping / prioritization?

    • Phil: identifying who will work on what?  If we identify a need and don’t have the resources, then we may need to re-allocate resources to do it.  Want to allow conversation to occur about this, and to give an opportunity for input.  Want to play to the strengths of the team, but also to make sure high-priority tasks are addressed.
    • Shaocheng: Since we are approaching the end of the v1 development cycle, people may be wondering what they will be working on next.   Our group has more than 30 or 50 scientists, who can work on different things in parallel.

Plans for V3:

  • model currently has large aerosol indirect effects, this is partly due to radiative transfer treating all clouds (including both stratiform and deep convective clouds) the same.  Should modify to allow for separate subcolumns and separate properties for stratiform and deep convective clouds.
  • Comments from the floor:
    • Philip: Most of the chemistry that is listed is actually representation of aerosols, not chemistry.  Maybe some more chemistry should be discussed later.
    • Yun: You listed a lot of ideas, but we will not be able to do all of them.  We will have to determine priorities.  These should depend on the science goals, how large the impacts are, whether we have the necessary expertise on the team, and the computational cost.
    • Dorothy: question about tracers and performance.  Effort was underway to make tracer transport more efficient, improve parallelization.  What is the outcome of this?
      • Phil: I'm not aware of anyone working on this on ACME right now, could you connect us?
      • Dorothy: Mark Taylor and Phil Jones are working on algorithms for this.
    • Julio: Surface drag and other sorts of drag processes ought to be somewhere on the radar screen
    • Shaocheng: Phil, you listed a number of things, do you have a detailed plan on these? Phil: I know how I would do it.  Shaocheng: we should also coordinate with CMDV.
  • Hailong: added volcanic aerosol treatment, which is very important.  Dick Easter and Steve Ghan worked on implementing this in CCSM, so we have the expertise.  Phil: This would require some justification.
  • Rich: How do you envision the division between v2 and v3: will we work in parallel on efforts for v2 and v3, or is it more linear?
    • Phil: several possible ways to work moving forward
      • learning about the model we have, documenting it in publications
      • if we learn about problems, devote human resources to fix them as quickly as possible
      • if we identify known deficiencies, or features that are important for the next-generation science problems, start working on those today
      • work with coupled team and other teams (ocean, sea ice) to identify opportunities
  • Scott: for v2 model, is there a target resolution?  A lot of processes are scale-dependent, so developments might depend on this.
    • Ruby: v3 is quite a few years out, so higher-resolution is possible as target
    • Phil: we will likely be doing exploratory high-resolution simulations at higher resolution before v3.
  • Anshu: as part of CMDV, I'm responsible for testing, and building a culture of testing.  We are tasked with making the model development process easier through testing.  Some of us are meeting tomorrow at lunchtime, all are welcome to join.  I have a history of working with large physical development codes that have seen clear benefits from testing, so I urge this project to adopt testing practices as well.
    • Phil: as we move forward, we will need to adapt to higher bar for testing set by technical teams
    • Anshu: I don't want to impose, I want to persuade you that this is needed
  •  Po-Lun: in v1 we had very strong AIE, and tried to reduce these by tuning, but we know that some process improvements are needed to improve aerosol-cloud interactions.
    • Phil: thanks, that is an oversight.  We will want to do better on cloud treatments including cloud microphysical treatments
  • Philip C-S: What do you see as the importance and relevance of the stratosphere?
    • Phil: I lobbied for raising the model top because I believe the stratosphere influences troposphere, particularly the variability at high latitudes.  Stratosphere also needed for representation of volcanic eruptions, optical properties of aerosol that we're using are not the best.  It may eventually also be interesting to look at geoengineering with stratospheric aerosols.

Understanding The ACME model:

Tier 1b diagnostics (Xie_ACME_Atmos_Diagnostics.pptx)

  • During tuning, we used AMWG for diagnostics.  Now that tuning is done, we should do more diagnostics and analysis of the model when v1 is complete.  We should revisit the Tier 1a, Tier 1b diagnostics and produce a package that everyone can run, not just the PIs who wrote the code.
    • Tier 1a is done.
    • Reviewing the Tier 1b diagnostic collections:
      • Some PIs left. (Jin-Ho Yoon, Tianyu Jiang)
      • Some are listed but don't have anyone working on them
      • If someone is listed as working on a diagnostic collection but doesn't have time to do it, please let team leads know.
    • Phil: we originally identified these as particularly important for ACME science and planned to use them for tuning the model.  If we get these right, then the next time we build a model, we will optimize for these things.  Right now, we're still doing the tuning using the old strategies and not using the new approaches.  But there is no way to use the new approach until these things work out-of-the-box.  Once these work, we could use these to develop objective metrics for model tuning.
    • Shaocheng: we did use some of this (e.g. Tier 1a) when doing the convection assessment.  We just need a little effort to get these things done.  Once v1 tuning is complete, you can expect that Phil and I will push you to really get them done.
    • Scott: Wouldn't the idea be to build these tests to be run automatically, similarly to unit tests?
      • Phil: the unit tests are checking to be sure that the software is operating correctly.  This exercise is to tell us about the fidelity of the simulations. IN the ideal world, we would have a workflow that not only runs the model, but runs the diagnostics in a fashion that is as automated as possible.
      • Scott: in addition to a human looking at it, building really objective diagnostics would be valuable, e.g., what is an acceptable error in precipitation over the Amazon, some overall performance score, where some statistically clever person has looked at how to diagnose what is an acceptable performance.
    • Shaocheng: next step is that we need to make sure that Tier 1b diagnostics that exist already can be run by any team member, not just the PI.  Peter and Kate will discuss this further.

Using the ACME model to further ACME science that does not involve the coupled simulations ( Ruby_ACME_WaterCycle_Science.pptx )

  • Some science questions (e.g. aerosol indirect effect) don't require coupled simulations.  What science can we do now / what papers can we write?
  • Ruby: effect of high resolution on simulations?  What does high resolution buy us?
  • Hailong: quick comment about diagnostics – when Po-Lun and I were tuning the aerosols, we used the new AOD diagnostics all the time
  • Yun: Recently did some AMAZON work, so I volunteer to replace Jin-Ho on leading Amazon diagnostics
  • Rich Neale: for high resolution simulations we really should be looking at time-step-by-time-step analysis to make sure the model is doing what it should be doing.
  • Yun: aside from the spatial/horizontal resolution, the vertical resolution is important. Po-Lun might want to lead a paper on that.
  • Shaocheng: think about what papers you want to write and the science that would drive those papers.

Where would Atm Team like to take a lead in investigating coupled features?

  • Phil: I tried to identify some important topics that involve interactions with other model components, but in which the atmosphere team would need to play a critical role.
    • ENSO?
    • Surface Mass Balance over Ice Sheets?

Publications (Xie_ACME_Atmos_Publications.pptx)

  • Please send Shaocheng a tentative title for each paper listed here.
  • Think about papers that you could write, especially papers focussed on ACME v1 model developments.
  • Brief explanations of each paper:
    • Hui, Kate, and Chris Terai have papers already submitted
    • Papers in preparation:
    • Po-Lun Ma:
      • Writing paper on aerosol indirect effects, and observational constraints from satellite simulator
      • Tuning papers: I can't write all these papers by myself.  If anyone is interested in any of the tuning experiments and wants to write it up, please come get the climo files from me.
    • Hailong: almost ready to submit paper on ACME aerosol radiative forcing, indirect effect
    • Susannah: writing paper on impact of marine organic emissions in ACME
    • Kai: working on paper on water conservation issues, would like to submit as soon as possible.
    • Yuying: working on paper on COSP simulator.  Plan to submit it in a month or two.
    • Salil: Looking at impact of NAO on extremes, would like to submit in GRL before end of year.
    • Wuyin: Paper focussed on tuning exercise, practical realities of tuning with limited resources.  How to use CAPT to tune the physics and quickly tune the physics response.
    • Yun: Hui and I have done the PPE simulations and will write a paper documenting the framework for UQ analysis and optimization.
    • Regionally-refined model paper was left off the list – it is in development but there were some issues with the v0.3 model, need to update thte model version that we use.
  • Shaocheng, a lot of papers based on v0 model.  Will we repeat these for v1?
    • Phil: sometimes we might write a paper about a methodology, other times we might want to write papers documenting the differences between the old and new models
  • Rich Neale: for v1, is it worth drawing a line and choose a version to run with that can be used to do science, with high-resolution output.
    • Phil: These are the papers that were already started.  There are a slew of papers that I think will be written about the v1 model.
  • Renata: add new planned papers to this list (or update them): /wiki/spaces/ATM/pages/72450145


Diagnostics Discussion ACME_ATM_diags_discussion.pdf

Kate Evans (Unlicensed)

ACME Atmosphere priority metrics

  • Coupled team had a severe need for a package to produce "top-ten" diagnostics.  We produced this tool in response to their request. 
  • We designed it to be familiar to you, look-and-feel is similar to AMWG.  But some improvements:
    • Unlike AMWG, this package is python-based and modular.
    • You can more easily add new plots, code to produce these can be written in python, but other languages can also be used if package can call/link to them.
      • currently using matplotlib (in python)
      • can use other things such as NCL if willing to maintain these on an "opt-in basis"
    • Extensible to include tier 1B
    • quick and easy to run, has validation, reproducibility
    • conda-based (runs in conda python environment)
  • Questions?
    • Rich: Do you envisage this as a single script that could control all the Tier 1b collections?  Could see some benefits, but also some drawbacks to doing it that way.
      • Kate: could use this as a top-tier script that calls other things, or the other way around, or run it within a larger outer shell as one of a number of other things.
      • Kate: hopefully people will work on improving components (e.g. the web page)
    • Rich: graphics: how good are they at doing non-standard plots?  Is the functionality really there in python?
      • Answer: yes.  Matplotlib is very powerful, lots of other community-contributed libraries and packages exist.  PyNGL allows 
    • Scott: will you use this for doing regridding?
      • Kate: Other ACME folks have developed regridding scripts which we are using.
    • Philip: Will you have a set of automatically calculated metrics?
      • Kate: yes we can add that.
    • Todd: We could save these diagnostics in a searchable database.  We should keep that forever.  This is a form of provenance, then people would come back and look at it.
      • Phil: we have already thought about some of these issues, so maybe we can make connections between us.
      • Susannah: Bibi Raju from the workflow team is based at PNNL and is working on database technologies for provenance, and we have been discussing that with them.
      • Phil: Peter and Susannah and I also came up with some ideas about provenance that we would like to track, and some basic mechanisms for doing that which could set a "low bar" for provenance capture.
    • Peter: I'm hearing ideas about different cool things we could do with diagnostics, keep thinking about that.  Also we should think about what requirements we would have.  I have a list of ideas (in the slides), but are there things missing?