Information for Mentors

July/25/2012

Dear modeling mentors,

we detected and corrected a couple of typos in the test case document that I would like you to be aware of and double-check your code for test 11 and 20.
It it might be unlikely that it affects you (the Fortran routines were fine), but please take a look. We also had added one more global attribute native_grid and corrected one standard_name entry in Table XX. The changes are listed below where the equation numbers refer to the newly uploaded version
DCMIP-TestCaseDocument_v1.6_26Jul2012
of the test case document (see http://earthsystemcog.org/projects/dcmip-2012/test_cases)

Changes:

• a) Eq. (25), equation for eta_dot: There was an extra sine function around the shape function s(p) which was a remnant from the old formulation. It should only read s(p).
• b) Eq. (27), equation for Delta_p: Same issue as in Eq. (25), it should only read s(p)
• c) Eq. (68), equation for z: The exponent needs to be (Rd Gamma / g). The Fortran program was correct.
• d) Eq. (73), equation for layer-averaged T: a p_0 was missing (thanks, Lucas)
• e) Table XX: the entry for the standard_name of the surface pressure needs to be 'surface_air_pressure' (thanks Markus). The use of the standard_name is optional, please make sure that you provide a long_name and a unit for your variables. The formatting problem in Table XX is also fixed.
• f) Appendix G.1: we added one more global attribute 'native_grid'. In case of interpolated data (encoded in 'grid') native_grid should be set to the actual computational grid like 'cubed', etc. (entries from Table V). In case the data set is on the native grid, then both attributes will be the same. This will allow our search interface to query the data based on their computational grids.
• g) If you model relies in hybrid coefficients or other parameters, like the position of the model top for the Gal-Chen coordinate, a full set of these parameters need to be part of the netcdf file. We also asked for the 3D pressure field in case of non-hydrostatic models. This will allow the reconstruction of the vertical positions, and also allow us to interpolate to pressure levels like U850 via NCL.
• h) The formulation of the optional test 201 (variant of the steady-state at rest with orography according to suggestions by Joe Klemp and the MPAS team) is under development. Stay tuned.
• i) Non CF-compliant error messages will be a problem. We will also help eradicate inconsistencies next week via NCL an NCO. Only compliant data can be uploaded to the NOAA server and linked to the Live-Access-Visualization tool. Therefore, I am so persistent.

Group disk space:

I saw some some groups have started using the project space:
/glade/proj3/mich1098/
Once you create a subdirectory, please open up the permissions for the subdirectory to -rwxrwxr-- and for files -rw-rw-r--. This will allow your whole group to contribute data sets into the same subdirectory and modify the data set. The 'write' permission for the group is not automatically set, but can be set via the chmod command or umask in your Unix setup files like .bashrc, .kshrc etc.

Many thanks for the hard work, it has been a long journey. As always feedback on the test case document and example data sets are greatly appreciated. I still have not heard from some groups. Please send us a short update on your status if you have done so over the last few days.

Best, and see you soon in Boulder,
Christiane

July/21/2012

Dear DCMIP modeling mentors,

DCMIP is only one week away and I hope that you have implemented most if not all of the test cases. Many thanks already for your hard work. I guarantee you that it will pay off if it has not already.
I also would like to welcome a potential new modeling group ASAM under the guidance of Oswald Knoth from the Leibniz Institute for Tropospheric Research in Leipzig, Germany.

I still have not heard from some of you what the status is. Please email me to give us an update on the status of your preparations. It is paramount that your data sets follow the DCMIP data conventions outlined in the test case document (including the choice of the variable names), and that you supply us with example data sets. If you have difficulty creating compliant data sets please let us know since we then might be able to fix some problems offline via the NCO operators.

The main reason for this email is that we have been discussing the design of the advection test cases 1-1 and 1-2 with Thomas Frisius (University of Hamburg), Daniel Reinert (DWD), Marco Giorgetta and Levi Silvers (MPI) and Almut Gassmann (IAP). We very much appreciated your feedback, thank you.
These modeling groups all use the conservation form of the advection equation, and struggled with the fact that the continuity equation was not fulfilled in the original design. We (Paul Ullrich, James Kent and I) therefore saw an urgent need to change this design feature and have been working hard to fix the problem. We apologize for the late notice, but tests 1-1 and 1-2 are so much better and more consistent now, that we now ask you to include these changes before coming to DCMIP. The spirit of the tests is still the same, just some details change (which translates into very minor changes in your codes).

Changes in the test case document version 1.6:

• I just uploaded version 1.6 of the test case document to the DCMIP web page (also appended). The revised Fortran code is also uploaded (and appended). The changes are minor and hopefully very quick to include:
  
  • slighly revised zonal and vertical velocity in test 1-1, removal of the altitude cutoff for the tracers (reversal of the change in earlier version)
  • slightly revised meridional and vertical velocity in test 1-2, modified parameters w0, z1 and z2.

• Yesterday, Joseph Klemp and the MPAS team also gave feedback concerning the design of the optional test 2-0 (steady-state at rest with mountain). Many thanks. We decided to include his suggestions as an additional optional test 2-0-1 and work out the scientific description right now (not yet included in the test case document 1.6). This necessitated a change of test case number 2-0 to 2-0-0 which is reflected in the document. Stay tuned.
• Please also note that we slightly updated the Fortran initialization routine for test 4 (on DCMIP page). A new if-construct in the computation of the relative vorticity now prevents the possible division-by-zero issue that had the potential to appear in models with latitude-longitude grids, or models that include the pole points in the grid.

Computing-related updates:

• Submission queue:
  Starting on Monday (7/23) DCMIP will have access to a dedicated high-priority queue on Bluefire and Lynx. The name of the queue is 'workshop'. Please use this queue in your jobscripts. The priority of 'workshop' will be elevated from 7/30 onwards.
  
  
• Disk space:
  DCMIP has now access to project disk space in the directory: /glade/proj3/mich1098/
  This space will be available until some time in September and will not be purged. You can create your own subdirectory. It will be important that the modeling teams open up the access rights (rw) for the group members of group mich1098 so that the teams can share the space. This will need to be done manually or via your umask Unix setting.
  
  Do not write anything into the directory
  /glade/proj3/mich1098/noaa_transfer
  This is a staging area for data that are quality controlled (compliant with our DCMIP data format and CF compliant). Recall the the location /glade/scratch/<user_name> that you might have been using until now is scratch space, and files are purged after about 25 days.
  
  
• Data and NetCDF CF-compliance:
  Please check the CF compliance of your data sets yourself and fix potential problems. We provide instructions on how to use the CF compliance checker installed on the NCAR machine mirage0.ucar.edu on the DCMIP page: http://earthsystemcog.org/projects/dcmip-2012/netcdf
  It is paramount that your data sets follow the DCMIP data conventions outlined in the test case document (including the choice of the variable names), and that you supply us with example data sets. Otherwise, our analysis tools will likely break.
  
  
• Laptops:
  We assume that you all bring your own laptops. If this is not the case and you need an NCAR laptop please let us know.

Presentations by mentors: Open discussion of the pros and cons of your model designs

• We scheduled all modeling groups to give a 20-minute overview of the pros and cons of your model design choices. Please take a look at the schedule on the page and your time slot:
  
  http://earthsystemcog.org/projects/dcmip-2012/lectures
  
  These presentations should not focus on details, but rather on the big picture and should provide insight into the physical reasons or motivation why a design choice was made, and whether you think about changing things in the future.

Strategic meeting about the vision for DCMIP:

• We would like to invite all mentors to a strategic meeting to think about DCMIP and its vision to create a virtual dynamical core modeling community which is supported by cyberinfrastructure. Most likely, we will try to call this meeting on Thursday afternoon (8/2) during the hands-on session when we leave our students by themselves for a little while. Please stay tuned.

Informal get-together on Sunday evening 7/29/2012:

• Please be reminded that we arranged for an informal get-together on the first Sunday evening at the Baker Street Pub & Grill near the hotel, 1729 28th St., Boulder, CO 80301, starting at 6pm. Find us on the patio if you like.

As always, feedback is very much appreciated,

Christiane

---

July/8/2012

Dear DCMIP modeling mentors,

we would like to give you updated information on DCMIP and point you to some important revisions of the test case document and the associated Fortran routines.

1. Test cases
   
   We finalized the description of test 1-3 (horizontal advection in the presence of orography) and tested it with two models MCORE and CAM-FV.  The design of this test promises to be very exciting and insightful. We are sure that all teams will learn something interesting about the characteristics of their advection schemes. Two sneak previews from CAM-FV and MCORE are featured on the page
   http://earthsystemcog.org/projects/dcmip-2012/example_results/test-13
   that shows how initially shallow cloud-like tracer distributions behave when they are horizontally transported over a mountain range (e.g. at day 6). The horizontal transport refers to a transport with respect to a constant height above the mean sea level, which invokes vertical transport in case of orography-following coordinates. We hope we will discuss the consequences of such merging tracers on physical parameterizations during DCMIP.
   
   In addition, we added test 2-0 that evaluates the accuracy of the pressure-gradient calculation in a non-rotating atmosphere at rest in the presence of mountains. It is a clean simple assessment, especially for models with orography following coordinates and has been very insightful for CAM-FV. It is also quick to run. Our suggestion is to run this test as an optional test, but we urge you to include it in your test suite. This is again an opportunity for all mentors to learn more about their models.
   
   The new test case document (version 1.5) is available at
   http://earthsystemcog.org/projects/dcmip-2012/test_cases
   
   The updated Fortran routines and a detailed change log are available at:
   http://earthsystemcog.org/projects/dcmip-2012/fortran_routines
   
   Since we also fixed bugs in the revised versions of the Fortran routines, please pay close attention to the changes listed in the change log. They are important to include in your models. We suggest that you use the newest Fortran routines as a replacement for earlier versions. In particular, we had two bug fixes in the EPV tracer in test 41X, needed to introduce slight changes to the tracers in test 11, and had two bug fixes in the simple-physics routine.
   
   Please complete all experiment of the whole test suite first, but feel free to try out additional experiments (e.g. changed diffusion coefficients) or model configurations (variable resolution grids, hydrostatic versus non-hydrostatic options, etc.). If you model supports both a hydrostatic and non-hydrostatic option, please make the non-hydrostatic option the default.
   
   Plots of the initial data are provided on the page:
   http://earthsystemcog.org/projects/dcmip-2012/initial_data_plots
   
   
2. NCAR computing
   
   DCMIP is only three weeks away, and by now your model preparations are most likely well under way. We hope you all installed your model on the NCAR machines and tested it on the NCAR IBM 'Bluefire'. You can also have access to a CRAY machine Lynx, but in an earlier questionnaire only two groups (MPAS and MCORE) indicated potential interest. Let us know whether you changed your mind. The advantage of the Cray is that we have it almost to ourselves. The page
   http://earthsystemcog.org/projects/dcmip-2012/computing
   lists information about the NCAR computing environment including our DCMIP project number 35201098 for the GAU computing charges.
   During DCMIP we will have a special submission queue with high priority. Shortly, we will have access to a project disk space (details to follow). So far, keep using /glade/scratch/login_name
   
   Help is provided through the NCAR consulting group (see the web page).
   
   
3. Lecture schedule and lectures by mentors
   
   The lecture schedule is available on the page:
   http://earthsystemcog.org/projects/dcmip-2012/lectures
   
   As you see in the agenda we reserved 20-minute time slots for all modeling groups. We would like to ask one mentor per modeling group to provide insight into the high-level design decisions in your dynamical core.
   
   This is not intended to be a condensed talk where all details are presented. Rather, we would like to learn about the physical motivations behind your design choices, and their pros and cons as you look back. Were there hurdles when putting the model together? Are there things you would do differently in the future? What are the lessons learned or surprises concerning the model design? E.g. please motivate why you or your institution chose
   
   • the equation set and prognostic variables
   • a particular computational grid
   • the grid staggering
   • the spirit of the numerical method (e.g. spectral, finite difference, finite volume, etc.)
   • a particular formal accuracy of the numerical method 
   • the conservation properties
   • dissipation mechanisms
   
   Your presentation should reveal the thought and development process, and also encourage questions. I will schedule all groups. In case you do not want to talk about your model, please let us know.
   
   
4. Daily agenda and your role as a mentor:
   
   Our daily agenda is available on the page:
   http://earthsystemcog.org/projects/dcmip-2012/agenda
   
   We will assign a small group of students (2-3) to each modeling team, and these team will work together for the complete 2 weeks during the afternoon hands-on sessions.
   
   Our days will be as follows: Morning lectures until 12:30pm, and hands-on afternoon sessions (1:30-5pm) in small groups in breakout rooms. The task of the modeling mentor will be to introduce the students to the principles of your model, and to teach them how to run it (e.g. namelist settings, run scripts, settings of empirical coefficients). Therefore, it is important that your model has been set up for the test cases. We ask you and your students to run the test suite on the NCAR machine together during the two weeks, and to analyze the results. We provide NCL analysis scripts. Basically, we grow a database with all model results during the summer school.
   
   On the last day (Friday 8/10) we will spend the whole day discussing the model results. The students will then present highlights of their groups and encourage discussions.
   
   
5. Example netcdf data sets
   
   In order to get our NCL analysis scripts and visualization tools ready, we now need example data sets from you. I ask you to point me to example netcdf data on a latitude-longitude grid and native computational grids for
   
   • 11.medium.L60
   • 21.medium.L60
   • 31.medium.L10
   • 410.medium.L30
   • 42.medium.L30 
   • 51.medium.L30
   
   on the NCAR machine, e.g on /glade/scratch/login_name server. If you have tested the model on your home machine so far, please transfer these data sets to the IBM in the next 1-2 days. It is important that we know your data sets as NCL training data. Otherwise, our tools might likely break.
   
   It is paramount that your netcdf data are CF-compliant, and adhere to the standards listed in the test case document. It also needs to be clear how to compute the pressure at the model levels, in case the 3D pressure field is not provided as an output quantity. The other important factor is whether your model levels are stored from the bottom up or the top down, encoded in the netcdf metadata attribute positive = "down" or positive = "up" (see sections in the Appendix G.3 and G.5)
   
   For models on non-latitude-longitude grids, please also provide the data on an interpolated latitude-longitude grid. If this creates difficulties please let us know.
   
   
6. Travel logistics:
   
   For the non-local (non-Colorado) mentors we booked all your hotel rooms and the check-in and check-out dates are listed on the page:
   http://earthsystemcog.org/projects/dcmip-2012/lodging
   Shortly, you will receive an email from the Best Western confirming your reservation details.
   
   The wep page also provides information about the local transportation. We fully sponsor (flight, hotel, allowance for food, local transportation, fees, etc) one mentor per modeling team, except the local mentors from Colorado. Mentors from CSU receive reimbursements for the mileage, and food. If you no longer need this hotel room for some reason, please let us know immediately.
   We are happy to report, that we will be able to sponsor the Best Western hotel rooms of all self-funded mentors. This is a small thank-you for the hard work you all have been doing. The lodging costs will be directly billed to the University of Michigan (UM). You might still be asked to provide a credit card for incidentals. The hotel provides free internet. It is beneficial to bring your own ethernet cable since a most rooms are only wired (no wireless).
   
   Concerning the food, local transportation and fee reimbursements of the fully sponsored mentors, Barb Lupi (blupi@umich.edu) from UM will handle them. She will ask you for your home address and an ID number (for US mentors this is the SSN), and a check will be sent to this address after the DCMIP summer school. We will reimburse a total of $750 for all food, local transportation and fee expenses (no receipts, this is a flat rate).
   
   
7. What to bring?
   
   Please bring a laptop and a mouse. If you do not have a laptop, please let us know and we will provide you with an NCAR laptop during DCMIP.
   
   
8. DCMIP virtual workspace
   
   DCMIP is supported by a newly created virtual workspace hosted at
   http://earthsystemcog.org/projects/dcmip-2012
   
   This is a Wiki-supported site that will also allow access to the data and a Live Access Visualization (therefore we need your examples to test this further).
   A few of you already created an account and became members of the DCMIP-2012 project. Instructions are included in the test case document (section 0.7).
   
   The mentors who have not done this, please become a member early this week. We ask you to provide some basic information about your model on the already created page for your model (via the Wiki functionality) this week. If you need instructions about the virtual workspace, I suggest you watch the 21-minute demonstration linked to the page
   http://earthsystemcog.org/projects/dcmip-2012/how-to
   
   We also uploaded your name, basic contact information and your online photo (if available) to the page
   http://earthsystemcog.org/projects/dcmip-2012/participants
   
   Feel free to change or complete your information via the Wiki.
   
   
9. Virtual participation, outreach and broad dissemination:
   
   We have two modeling groups (GEM and PUMA) that participate remotely. We would like to give you the opportunity to present the highlights of your model results to the whole DCMIP community on the last day of DCMIP via screen-sharing technology. We will approach you individually to test this technology.
   
   We will also set up a live webcast of all lectures, and record all lectures, including the talks by the mentors. These recordings will be linked to the DCMIP virtual workspace, and we will also ask for your slides in order to post them.
   
   
10. Social events:
    
    We plan several social events. Please take a look at
    http://earthsystemcog.org/projects/dcmip-2012/social_events