wiki:access/UMTIPProgressJan2017

ACCESS UMTIP Dec-Jan 2017 Highlight Reports

Revision status

[Michael Naughton 09/02/2017] Work in progress: This page is currently being updated. Information below is subject to change.

Highlight reports for UM Partners activities were requested to be provided. Here are ACCESS responses for BoM, CSIRO, ARCCSS and NCI contributions, and additional ACCESS-related comments collated from the individual highlight report pages in https://code.metoffice.gov.uk/trac/TI/wiki/TIAG/TAG/highlightReports/0217.

Please advise me if there's any items missed here.

TIAG: Advisory Group

  • [Michael Naughton: updated 10/02/2017]
  • Peter May represented ACCESS in UM Partner Board meetings and discussions.
  • Michael Naughton, Martin Dix and Scott Wales participated in TAG meetings and discussions.
  • BoM hosted CAWCR Workshop and Reanalysis Workshops at start of December, with large number of Australian and overseas UM Partnership participants.
  • Keir Bovis visited BoM from 28 November to 9 December for the Reanalysis and DA Workshops and UMTIP discussions.
  • Peter May visited Met Office in January for collaboration discussions.
  • Tony Hirst, Oscar Alves and Harry Hendon will attend Wellington GMED meeting in February.

TIAN: Ancillary files software

TIAV: Data processing and visulalisation

TICD: Collaboration facilities

  • [10/02/2017: To be updated]
  • MOSRS TICD highlights link
  • BoM
    • Jin Lee, Martin Dix and Michael Naughton attended November TICD video-conference.
    • Intermittent connectivity problems from Australia and New Zealand were discussed.
    • Use of IP address range white-listing to be expanded at partner sites to allow Trac page viewing without requiring authenticated sign-in is under discussion for BoM.

TICT: UM versions and configuration testing

TIDA: Data assimilation

  • [10/02/2017: To be updated]
  • MOSRS TIDA highlights link
  • Met Office
    • TIDA videoconference took place in November.
      • Focus of discussion centred on paper outlining scope for the development of an observation capability covering:
        • ODB based monitoring - Have ability to utilise ODB data but need to undertake more work to exploit the capability further.
        • Improving the way we ingest BURF data
        • Improved ODB Infrastructure - ODB-1 to ODB-2 conversion
        • Improved OPS infrastructure
      • Paper to be presented at next TAG meeting
  • BoM
    • BoM progress to-date can be found found here
    • Also endeavouring to arrange secondment to provide extra assistance and come up with a more complete solution

TIHP: HPC benchmarking and optimisation

  • [Michael Naughton: BoM section updated 10/02/2017]
  • MOSRS TIHP highlights link
  • BoM
    • Ilia's work
      • Continued work on the OpenMP coverage improvement in the UM and JULES sources; see following tickets:
        • um (15):UM: 1960, 2023, 2370, 2404, 2512, 2516, 2540, 2551, 2595, 2614, 2617, 2620, 2630, 2641, 2683
        • jules: 433
      • Successfully ran 10km PS39 job on NCI Raijin machine (Fujitsu), ported to BoM Terra machine (Cray) using Cray compiler.
  • NCI
    • Ben Evans has completed ACCESS-Opt project Apr-Sep 2016 progress report (to be added)
    • Ben Evans and Rui Yang (?) attended ECMWF HPC Workshop

TILS: Land surface modelling

  • [10/02/2017: To be updated]
  • MOSRS TILS highlights link
    • CSIRO
      • The interoperability between CABLE and JULES Land-Surface Schemes (LSSs) was introduced in the JULES(35) and the UM(264) ticketing, management systems . A staged approach has been adopted to implement the use of CABLE, via coupling to JULES, as an alternative Land-Surface Scheme. The above tickets have been updated to reflect this staged approach. JULES(393) and UM(2433) tickets have been submitted to specifically address the implementation of stage 1.
      • In addition to the more official, ticketed, development of the CABLE-JULES interoperability, we have also coupled CABLE to the UM via JULES for UM8.5 and UM10.1. We have a long coupled (atmosphere/ocean) run with UM8.5 now and are reasonably happy with the results.
      • Ian Harman (CSIRO) has been reviewing our coupling strategy from a science perspective and has suggested some improvements which have been tested in ACCESS1.4 (UM7.3) and will now be ported to versions running with ENDGAME dynamics (UM8.5).
      • CABLE has been coupled via JULES to the UM10.1 and standard AMIP runs of 10 years completed, although further assessment has not been done. We are currently porting CABLE via JULES to the UM10.5 for intended use in CMIP6 experiments.
      • Collaborative meetings were held via video conference in August and September to update on current status and discuss future direction of project. Agenda, participants, actionable items resulting from these meetings are described in the links.
      • Ian Harman (CSIRO) visiting the Met Office for 1 week during November 2016

TISD: Suite Design

  • [Yi Xiao, 11/02/2017]
  • MOSRS TISD highlights link
  • Dec. 2016 to Feb. 2017
  • The focus in the past three months is implementing PS38 suites, global, UKV and MOGREPS-UK at NCI, replacing UK STASH with BOM STASH, and associated archiving tasks. We also worked with Marek Wlasak on the CVT suite, and it is now can support both sites, UK and NCI.
  1. Global Suite - Completed upgrade to PS38, trailed at N320/N216 for about 6 weeks at NCI. In the process of moving the suite to Terra, BOM small Cray machine, where UM will be upgraded to 10.6. Operational implementation is being planned. It is recommended that our global suite will be running with deterministic at 12km. The planed ensemble resolution was N400. However, it is not finalised. The operational implementation will soon start on our Cray HPC in the Bureau.
  2. City and City Ensemble suite - It is proven rather challenging to generate correct ancillary files for high resolution city systems. Much of time has been spent on this activity, and we are now much better in generating ancillaries. However not fully resolved yet. The city suite is based on UKV-PS38, with updated OPS (to version 2016.11.0), to accommodate our radar observations, and UM to 10.6. The suite is tested on our APS3 Sydney domain. Domain grids are handled via a dictionary. The city ensemble suite is based on PS38 MOGREPS-UK. The suite is adapted to one version of Sydney domain, which was used in a forecast demonstration project in 2013. It has run a couple of cycles, with dummy VER and archiving steps. Work is currently underway to move the domain to the APS3 ACCESS-SYE domain. We have had an issue with the Stomatal Conductance field that has recently been solved, which resulted from the Bureau attempting to run the single tile urban scheme rather than the 2-tile urban scheme. Some preliminary work has begun to look at the potential to upgrade to UM10.6. The plan is to extend testing to the rest city domains. We will start to work on generating covariance files, and other control data, such as stationlists.
  3. Other activities
    • Early this month, the Bureau has invited Dr. Hilary Oliver to attend a cylc workshop, which aims to gather requirements on cylc -- a potential operational scheduler.
    • The Bureau will invite Dr. Mike Thurlow to working in the Bureau for four weeks, assisting the operational global suite implementation.
  4. Suite Portability - It is great to hear that the portability will be formally addressed across UM partners. We are very keen to contribute. We believe that Mike’s visit will boost the progress on this important activity, through our operational implementation. Our experience and approaches to the portability is not entirely consistency, and we are looking forward to some assistance from Mike’s visit. Here are a few examples:
    • Research global suite We adopted a two-stage approach to producing our suite to run at NCI.
      • The first stage was to run the suite with minimal changes, by adding extra NCI config files etc. to the suite. This allows us to more-or-less run the suite in an identical manner to the Met Office. The first stage suite is u-ag312.
      • The second stage involved implementing BoM components including STASH and new archiving tasks. The second stage suite is u-ah862.
    • Research City suites - We factored out site dependent elements from the runtime.rc file, such as HPC directives and resources, they are put into separate include files, and get included back into the suite.rc file. We also replaced hard-wired domain in apps with domain specific variables, these domain specific variables are defined in a separate include file, or through a dictionary variable.
    • The CVT suite - A branch is created from the suite owner’s suite, and modifications are explained, reviewed, and adjusted, eventually changes are merged back into the suite owner’s trunk after some testing.
    • Operational Implementation - Approach is like our city suite approach, but we have gone a bit further to address additional dependencies between different operational systems. The work is at its early stage; Mike’s visit will help us to finalize it, to create a truly portable global suite.
  5. Further comments
    • Convective scale: hopefully, portability design aspects will be included in the new regional NWP trial suites (being developed for PS40?).
    • Modular suites: we're interested in planning for using modular Cylc functionality in suite deign, to increase flexibility in our internal suites, and between research and operations, and between partners.

TISU: Small utilities for model diagnosis, monitoring, optimisation

TIVE: Verification

Last modified 3 years ago Last modified on Feb 14, 2017 6:47:03 PM