Minutes of the code management group meeting 19/7/2012
UKCA code management meeting minutes.
Minutes of teleconference meeting held at 14:00 on July 19th 2012.
Those present: Nick Savage (NS, Met.O); Alex Archibald (AA, U.Cam); Paul Telford (PT, U.Cam); Luke Abraham (LA, U.Cam); Graham Mann (GM, U.Leeds); Colin Johnson (CJ, Met.O); Mohit Dalvi (MD, Met.O); Sandip Dhomse (SD, U.Leeds); Zak Kipling (ZK, U.Oxford).
Items of discussion (those in italics refer to the meeting agenda sent by NS on July 18th):
[1] Appoint minute taker for this meeting. I would like to circulate this job around the groups if possible.
Discussion of minute taking led to an agreement that this role should be shared around. GM indicated willingness to take next set of meeting minutes.
[2] (0) Review action items from CMG held in Cambridge: http://www.ukca.ac.uk/wiki/index.php/Minutes_of_the_code_management_group_meeting_15/5/2012
NS: Can everyone please check the last set of minutes to make sure that they are happy with the designation of code expert for routines. (LA later noted that there was a typo/error in the minutes.
ACTION: NS to correct last meeting's minutes and all to check that they are happy with the designation of code experts for selected routines.
[2] (a) Code management guidelines. NS to make requested modifications and publish the new guidelines.
ACTION: Done
[2] (b) To be proposed at next UKCA exec meeting that a living science plan is needed including input from management on future plans. Proposed and underway.
ACTION: NS to advise on progress at next meeting.
[2] (c) Subsection experts - names and responsibilities. Names are agreed. I need to publish and then start classifying routines. Have got agreement from UM systems team that we can include a comment in each subroutine about which UKCA subsection it belongs to.
ACTION: All to check that they are happy with the designation of code experts for selected routines (see minutes of last meeting).
[2] (d) NS to ask if we can have a Trac system on the Collaboration website. Basically the answer is no. We need to decide now if we want to use the PUMA UM trac site (can we get people like Olaf an account on it) or use the UKCA UM trac or some other platform. Philip Stier seemed keen on Redmine, but I have no experience with it [sic].
There was a discussion on the need for a unified system to record ongoing projects/work plans (with tickets, milestones etc). The PUMA UKCA trac system (http://puma.nerc.ac.uk/trac/UKCA) was suggested as being a good place (this was used for the merged branch) to address item 2(d) on the agenda.
ACTION: All code developers to use the PUMA UKCA trac system to address these issues?
[2] (e) NS to set up templates for code review and ticket summary based on typical UM templates. Once the trac system is set up, NS will populate it with Met Office tickets which already exist inside the Met Office and invite NCAS scientists to add their tickets. Still on hold until we have a suitable platform.
ACTION: NS to provide update at next meeting.
[2] (f) Configurations. NS to provide AQUM configuration at 8.2. Not done yet. NS with Fiona O'Connor to provide HadGEM3A plus StratTrop chemistry + ACHEM + GLOMAP-mode + Fastj-X using GA4. Not yet debugged, in progress.
NS opened discussion on item 2(f), configurations of standard release/scientific jobs at 8.2. So far not a great deal of progress. Fiona O’Connor identified a few issues with the FAST-Jx routines lodged at 8.2 (CJ suggested that these are likely bugs that crept in during optimization of the code). CJ noted that there are some “issues” with chemically inert tracers/aerosol precursors between 7.3 and 8.2. In particular differences in the vertical profile of several species (in experiments that used identical lower boundary conditions) was undesirably high. LA suggested to switch of tracer convection to see if this was the cause of the differences. GM suggested that the 8.2 MODE setup run at 7.3 resulted in good agreement in results. NS suggested that someone could look at using the model physics from 7.3 in an 8.2 job. LA noted that he has started running an atmosphere only job at 8.2. CJ noted that the offline MODE code was almost there (needed to introduce diurnal oxidants).
ACTION: Met O. need more work on providing the global model configurations.
[2] (g) We need both the jobs and input data to be available and a scientific validation. LA to discuss with Fiona. See later agenda item on scientific validation [sic].
[2] (h) NS to write minutes and upload to web.
ACTION: Done.
[2] (i) Date of next meeting. NS to set up Doodle poll to co-ordinate.
ACTION: Done.
[3] Items from July Exec meeting.
[3] (a) How to validate supported scientific configurations.
AA suggested that the phrase model validation is not suitable for the first round of comparison of model data to old results and observations. Suggestion to call this processes model evaluation agreeded by all. MD discussed progress on the UKCA model evaluation package. Oxford (Anu Dudhia) had sent data to MD. GM said that he can provide aerosol data.
ACTION: GM to provide MD with data for aerosol evaluation. SD said that he can provide a range of satellite data based on the CCMVal type analysis.
ACTION: SD to send data to MD for stratospheric composition evaluation.
ACTION: All to agree on diagnostics, data sets, etc by the end of August. MD to provide evaluation tool by end of September.
There followed a discussion on how to define when an evaluated model is suitable for release? A suggestion was raised that an expert judgement system could be used defining a series of limits for acceptability of a set of diagnostics. AA suggested that external moderation would be needed in the long run.
LA raised the issue of who assess which processes?
Aerosols to be assessed by Leeds and Oxford (GM to lead).
Strat. chem to be assessed by Cambridge and Leeds (LA to lead).
Trop. chem to be assessed by Cambridge and Met Office (AA to lead, NS to lead all work on limited area modelling i.e. AQ).
Meteorology (including strat. dynamics i.e. age of air) to tbe led by Met O. and Cambridge (who is responsible??).
ACTION: Everyone to go away, think about their prefered metrics for evaluating their prefered process, and think about what would be acceptable.
NS added that the approach to creating branches for use with supported scientific configurations will be based on the approach used for operational NWP jobs. All branches needed to run a specific configuration (bug fixes and new code) will be merged to form a specific branch for that supported configuration.
ACTION: All to think about/start adopting the above method.
ACTION: NS, CJ and LA to make sure that release configurations are ready by mid-November so that new PhD students can make best use of the code improvements.
[3] (b) How to improve communications. Options include email list, Redmine, Trac, Wiki - combinations are possible.
Discussion of item 3(b) led to the idea of using the UKCA JISC email alert system to improve communication of bug alerts/updates.
ACTION: LA to make sure that everyone that is a known user is on the JISC user list and to for all to start using this email list to communicate UKCA code announcements.
[4] Agree strategy for use of 150 tracers and path to increasing numbers. See recent summary I sent around [sic].
NS led discussion on moving away from using the STASH master file as a placeholder for species (discussion of item 4 highlighted that the current method would cause significant problems when transferring to IRIS and so a suitable solution will have to be found).
ACTION: NS to remove “dead” tracers.
LA and NS agreed that the StratCFC branch is a good candidate to go into 8.3.
ACTION: NS/LA to lodge StratCFC.
LA asked can we please sort out field codes for tracers.
ACTION: NS to talk to UMUI team about this.
[5] Date of next meeting. Early September?.
ACTION: NS set up doodle poll. http://www.doodle.com/g9iuzhma8b6e7x5k
[6] AOB.
No other items discussed.
ACTION: AA to upload minutes on the UKCA website. Done
