Difference between revisions of "Minutes of the code management group meeting 2017-07-14"

From UKCA
Line 3: Line 3:
 
==Teleconference Details==
 
==Teleconference Details==
   
'''Friday 14th July 2017, 10am - 12noon'''
+
'''Friday 14th July 2017, 10am - 12noon BST'''
   
 
* '''UK Freefone:''' 0800 9171956
 
* '''UK Freefone:''' 0800 9171956

Revision as of 07:19, 14 July 2017

Code management group minutes

Teleconference Details

Friday 14th July 2017, 10am - 12noon BST

Present

Met Office:

Cambridge:

Leeds:

Reading:

Lancaster:

Edinburgh:

UEA:

NIWA:

CSIRO:

Apologies: Nick Savage, Nicolas Bellouin, Alex Archibald, Michael Hollaway, Oliver Wild, Ashok Luhar, Matt Woodhouse, Colin Johnson

Minutes

Actions from the Last Meeting (LA)

Minutes of the code management group meeting 2016-09-23

  1. (ongoing) GM to check 8.4 versus 7.3 (and possibly vn8.4 vs 10.x) GLOMAP-mode code and report any missing bug fixes on the bugfix page.
  2. (ongoing) FOC to incorporate Bug #35 and others into dry deposition work
  3. (ongoing) LA to migrate PUMA UM tickets to MOSRS
  4. (ongoing) UM #368 (RAFEPS value). Error not encountered at UM10.x but LA to consult Oliver Wild regarding background and implications.
  5. (ongoing) LA to check doc for NVOC factor and open ticket to remove it from code
  6. (ongoing) CJ to lodge STASHmaster changes to diagnostics that should make use of UKCA time availability codes
  7. (ongoing) CJ,FOC,AH,AS,LA and John H to decide best way forward for implementing GLOMAP-climatology code and documentation.
  8. (ongoing) JM,BJ,NB to discuss incorporation of Nicolas' RADAER description into UKCA (or separate UMDP). JM to take lead.
  9. (ongoing) FOC to open ticket + Bugfixes Table entry related to 'Unable to calculate tropopause' warnings
  10. (updated) MD to setup Strat+Trop suite as soon as GA7.1 is frozen and also add to Rose-stem. MD set-up/replace current Rose stem job with GA7.1 jobs.
  11. (ongoing) CJ to involve NS in discussions with Dynamics and UM-sys teams regarding movement of UKCA call, for effective implementation of water vapour feedback.
  12. (ongoing) FOC to recommend at next Exec meeting a policy of allowing only CMG to (certify and) distribute any UKCA configurations for science purpose.
  13. (new) MD to chase CMS about MASS archiving
  14. (new) NS to collate all code that could be removed and put on Wiki
  15. (new) LA to do doodle poll for end of Nov/ start of Dec

Standing actions

  1. (ongoing - standing item) CMG to monitor list of bugs and circulate between group for owners.
  2. (ongoing - standing item) At CMG future meetings there should be included an appendix listing the open bugs. If a ticket is included it is not an open bug (as the owner of ticket is person doing lodging).
  3. (ongoing - standing item) JM to continue to link/send powerpoint results (re: GA7.0/GA7.1), as new configurations released and link/put files on UKCA wiki.
  4. (ongoing - standing item) ALL with access to keep monitoring ukca_bugs email list.
  5. (ongoing - standing item) ALL to think about if code changes will change KGO and let systems team know in good time. This should be a standard part of our working practice.
  6. (ongoing - standing item) ALL to review and remove obsolete code

UKCA Tickets from PUMA UKCA trac pages (LA)

Ticket Description Owner Milestone MOSRS ticket required? Comments from this meeting
#4 increase items per section to 999 nick No - already done
#6 Merging of MODE developments Initial merge at 7.3 (level 1)
#7 Developments to FastJX code Initial merge at 7.3 (level 1)
#8 Stratospheric Chemistry Scheme luke Initial merge at 7.3 (level 1) No - scheme exists in the trunk but is not currently supported. Implementation of StratCFC scheme in #3243 (Hopefully): lukeabraham
#9 Tropospheric Chemistry with Isoprene Initial merge at 7.3 (level 1) No - scheme exists in the trunk but is not supported. There have been some stability issues reported with this scheme, e.g. to do with rafeps
#10 Chemistry of the Stratosphere and Troposphere (CheST) ata27 Initial merge at 7.3 (level 1) No - already done and is current supported scheme
#11 Tropospheric Aerosol Chemistry colin Initial merge at 7.3 (level 1)
#12 Stratospheric Aerosol Chemistry Initial merge at 7.3 (level 1) #1572(Hopefully). There has been some interest from NIWA to continue this work.
#13 Interface between UKCA and the UM Initial merge at 7.3 (level 1)
#14 Trace Gases and Radiation Initial merge at 7.3 (level 1) No - code in trunk sufficient(?)
#15 Dry Deposition Initial merge at 7.3 (level 1)
#16 Wet Deposition merged branch testing (level 3)
#17 Organic aerosols Complete vn7.3 merge (level 2)
#18 Modal Dust merged branch testing (level 3)
#19 Tropospheric heterogeneous chemistry. colin Complete vn7.3 merge (level 2)
#20 Aerosols and Radiation merged branch at latest release
#21 Indirect effects of aerosol Complete vn7.3 merge (level 2)
#22 Flight track diagnostics Somewhen #1608 (Hopefully): Marcus Koehler
#23 Satellite Diagnostics merged branch testing (level 3)
#24 Code redundency merged branch testing (level 3)
#25 UKCA aerosols for FastJX merged branch testing (level 3) #2689 (UM10.9): Michael Hollaway
#36 Pipework to deliver bVOC emissions from JULES to UKCA MetUM vn9.1
#37 Hard-wiring of land surface tile numbers in UKCA Somewhen
#38 Albedo for FastJ(X) Somewhen #442 (Hopefully)
#41 Improvements to Emissions handling MetUM vn8.6 No - already done (netCDF)
#42 Slow down of AQUM when 8.2 code lodged Somewhen
#45 Offline oxidants for MODE MetUM vn9.1 No - already done and is default scheme in GA7/GA7.1
#46 Improve convective washout of aerosols MetUM vn8.6
#47 RADAER Part 2 MetUM vn8.6 No - already done(?). RADAERv2 is already on the trunk.
#48 Add nitrate aersols to UKCA-GLOMAP-mode Somewhen #827(Hopefully):Steve Rumbold
#49 Modal Dust in UKCA-GLOMAP-mode Somewhen #1497(Hopefully): Stephanie Woodward
#50 FASTJX fix to work with new spectral files Somewhen
#54 Clean up UKCA interface - part 2 MetUM vn8.6
#60 Logicals/ Condition block mismatch for UKCA StratTrop and Het/PSC nick MetUM vn8.6
#62 Changes to make UKCA compatible with UM high resolution/LAM configurations MetUM vn8.6
#63 RCP scenario code MetUM vn8.6 No - already done
#64 Fix array indexing/ dimensioning issues in UKCA MetUM vn8.6

rafeps

rafeps is set in UKCA like this:

sfmin = TINY(1.0)                              [~2e-308]     asad_cinit.F90
sfmin = 10.0**(INT(LOG10(sfmin))+1)            [=1e-306]     asad_cinit.F90
peps  = 1.0e19 * sfmin                         [=1e-287]     asad_cinit.F90
rafeps=SQRT(peps)                              [~3e-144]     asad_spimpmjp.F90

This is rather convoluted! Previously it was found increasing rafeps by ~1.0e50 seemed to improve stability in the TropIsop configuration. There is a ticket open for this (#368).

Comment from Oliver Wild:

The intention is that peps is the underflow limit, and that rafeps is a
minimum of one order of magnitude more than this, just to avoid problems.
However, this is a bit slack, as the value of f may be multiplied by numbers
quite a bit less than one later in the code. So it's better to crank it up
a bit - the sqrt halves the exponent, so should be sufficient. I hardwire
peps to 1.d-200 in my code (substantially more than the underflow limit),
and rafeps is just peps*10.d0. Your value of 3.e-144 is fine.

My feeling is that as long as rafeps is "small", i.e., considerably below
any expected mixing ratios in the atmosphere, then all should be fine. By 
setting zeros to rafeps, you may be adding material to the atmosphere, so
you don't want to let rafeps get too large - but the principal issue is
avoiding underflows, or overflows when dividing. I don't think 1.e-100 would
be any problem. Even at the 1.e-50 level you may still be OK, although this
scale starts to makes me uncomfortable as errors can build up quickly if left 
unchecked.

I would suggest that altering the value of rafeps in itself is fine, although I (LA) would personally not want it to go smaller than 1.0e-100.

Redundant Code (LA for NS)

NS has created the following page for review: Redundant code

Code Restructuring

OM has been considering how best to restructure the chemistry. This proposal will remove all the existing ukca_chem_scheme.F90 routines, and replace them with a single ukca_chem_master.F90 routine that contains all the species, reactions, & deposition from all the current schemes, as well as a routine to construct the correct arrays for each scheme in an intelligent way.

Define the chemistry types
These are used to set a new last entry of the derived types used to define the chemistry scheme arrays, called scheme
! Define chemistry types
INTEGER, PARAMETER, PUBLIC :: ST = 1   ! stratosphere-troposphere scheme
INTEGER, PARAMETER, PUBLIC :: T = 2    ! troposphere scheme
INTEGER, PARAMETER, PUBLIC :: S = 4    ! stratosphere scheme
INTEGER, PARAMETER, PUBLIC :: R = 8    ! RAQ scheme
INTEGER, PARAMETER, PUBLIC :: STA = 16 ! strat-trop-aerosol scheme
INTEGER, PARAMETER, PUBLIC :: OL = 32  ! offline scheme
Define the list of species for chemistry, e.g.
!   3 DD: 1,
chch_t(  3,'O3        ',  1,'TR        ','Ox        ',  1,  0,  0, S+T+ST),  &
!   4  Wet deposition of ozone was introduced for STA 
chch_t(  3,'O3        ',  1,'TR        ','Ox        ',  1,  1,  0, STA),  &
!   4
chch_t(  3,'O3        ',  1,'CF        ','Ox        ',  0,  1,  0, OL),  &
Define the heterogeneous chemistry, e.g.
rath_t('ClONO2    ','H2O       ','HOCl      ','HONO2     ','          ', &
'          ', 0.000, 0.000, 0.000, 0.000, S+ST+STA), &
!HSO3+H2O2(aq)
rath_t('SO2       ','H2O2      ','NULL0     ','          ','          ', &
'          ', 0.000, 0.000, 0.000, 0.000, STA+OL),                          &
Define photolysis, e.g.
! 56
ratj_t('HO2NO2    ','PHOTON    ','OH        ','NO3       ','          ', &
     '          ',    0.0,   0.0,   0.0,   0.0, 100.000,'jpna33    ',T+S+ST+STA), &
Define termolecular reactions, e.g.
ratt_t('SO2       ','OH        ','SO3       ','HO2       ',     0.6,    &
  3.00e-31, -3.30,     0.00,  1.50e-12,  0.00,     0.00, 0.000, 0.000, STA), &
ratt_t('SO2       ','OH        ','H2SO4     ','          ',     0.6,    &
  3.00e-31, -3.30,     0.00,  1.50e-12,  0.00,     0.00, 0.000, 0.000, OL)         &
Define 2D dry-deposition. e.g.
depvel('H2O2      ', &
(/1.00,  1.00,  1.00,  1.00,  1.00,  1.00,  & !      8.1
  1.25,  0.16,  0.71,  0.28,  0.12,  0.20,  & !      8.2
  1.25,  0.53,  0.89,  0.83,  0.78,  0.81,  & !      8.3
  0.26,  0.26,  0.26,  0.26,  0.26,  0.26,  & !      8.4
  0.32,  0.32,  0.32,  0.32,  0.32,  0.32/),  & !      8.5
  T+S+ST+STA+OL), &
Define wet-deposition. e.g.
!    1  
wetdep('NO3       ',(/0.2000e+01, 0.2000e+04, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00/),S+T+ST+STA),&
Define bimolecular reactions, e.g.
! B051 JPL2011 see also asad_bimol
ratb_t('HO2       ','HO2       ','H2O2      ','          ','          ',&
'          ',  3.00e-13,  0.00,   -460.00, 0.000, 0.000, 0.000, 0.000,S+ST+STA), &
ratb_t('HO2       ','HO2       ','H2O2      ','          ','          ',&
'          ',  2.20e-13,  0.00,   -600.00, 0.000, 0.000, 0.000, 0.000,T+OL), &
Determine whether or not species/reaction/deposition should be included in the selected scheme, e.g.
DO i=1,n_ratb_master
  IF ((ratb_def_master(i).scheme MOD (2*chem_scheme))/chem_scheme .EQ. 1) THEN
    jpbk = jpbk + 1
  ENDIF
ENDDO

Bugs/Bugfixes (LA)

Open Bugs

# Description Ticket Owner Status/ Updates Comments from previous meeting Comments from this meeting
31 FP exceptions in asad_sparse_vars
or asad_spimjmp
#368 (MOSRS) Nick Savage leave open. vn10.x seems to be better
35 Error in labelling: Surface resistance <--> stomatal resistance ? MOSRS #1409 Fiona O'Connor open - folded into DD. FOC to update
50 ukca_solve_coagnucl failures #319 (MOSRS) Mohit Dalvi/Graham Mann? Not encountered any more ? dissapeared at vn10.x
57 Incorrect logic for DMS anciliaries required #756 (MOSRS) Nick Savage trivial. Might close as wontfix, but might want to do in RAQ
68 HO2 heterogeneous loss rate 1st order(?) No ticket Nick Savage NS to look at that. Leave open to review
72 UKCA Dry Deposition: Surface resistance values
not set for some species
#1250 Alan Hewitt? discussed above, have way forward
74 Bug in ukca_tropopause with definition of PV tropopause #1283 Steven Hardiman Change ownership? discussed above, wrap together issues. Fix is separate from #75
75 Bug in STE fluxes #1367 Fiona O'Connor could be endgame. Needs further work
80 H2O feedback broken under ENDGAME UM#1646 Colin Johnson discussed above, keep open
87 Possible issues with YSTORE in UKCA_CHEMISTRY_CTL. UM#3245 Colin Johnson/
Luke Abraham
88 Potential FastJX Scattering Bug UM#3261 David Wade/
Luke Abraham

Standard Release jobs: What is required now to move UKCA recommended jobs up to GA7.1? (LA/CJ)

Strategy for next UKCA Release Job

Highlights from vn10.8 (LA)

vn10.8 UKCA tickets
UMPB May 2017 Highlight Report

  • CJ: Water vapour feedback - This includes flux calculations of water production and loss available in stash, as well as a diagnostic which gives the change to q calculated at each timestep. (#1646 & #3123)
  • LA: Quasi-Newton steps in Newton-Raphson solver - speeds up the model by ~2-3%. Will not bit-compare with runs without it however. (#2819)
  • MR: OpenMP in UKCA_AERO_CTL - alternative method for passing array slices to sub processes, e.g. a(:,:,:,mode) had to be re-written a(1,1,1,mode) for the combination of Linux system SPICE and Intel ifort 12. (#1286 )
  • MR: OpenMP in UKCA_AERO_CTL - a bespoke workshare had to be written due to ONE failure in rose stem. The formal method that used the default blocking of data failed. Again Linux and Intel 12. (#1286 )
  • LA: 3 new UKCA rose-stem tests were added to the MetO XC40. This tests mirror the ones available on the VM, but run incredibly quickly. One has been added to the xc40_developer group - meto_xc40_n48_ukca_eg_omp_noios_gnu. (#2714)
    • Note: On the XC40 the _omp and _noomp jobs do not bit-compare, whereas they do on the VM. The version of the GNU compiler is different between these two sites - the XC40 uses GNU Fortran (GCC) 4.9.1 20140716 (Cray Inc.), whereas the VM uses GNU Fortran (Ubuntu 5.2.1-22ubuntu2) 5.2.1 20151010. The domain decomposition is different as well (4x4 on the XC40 vs. 1x{{MPI_TASKS}} on the VM). #3127 has been opened to investigate this.

Targets for vn10.9 (and beyond) (LA)

Tickets currently targeted for UM10.9 in Trac

Ticket Summary Owner Comments from this meeting
#100 PM10 and PM2.5 diagnostics for MODE johnhemmings
#1250 UKCA dry deposition: surface resistance values not set for some species alanjhewitt
#1703 Update appropriate UMDPs with regard UKCA-mode aerosol and interactions janemulcahy
#2086 Time updating for glomap climatology alanjhewitt
#2301 Improve the handling of emissions from explosive volcanoes in UKCA/GLOMAP --
#2382 rose stem app for Section 54 alanjhewitt
#2405 Test and implement sea-salt density change colinjohnson
#2532 Review of option codes associated with UKCA burden diagnostics when running with BE solver fionaoconnor
#2587 Adding solar variability to photolysis scheme fraserdennison
#2689 Enable coupling of GLOMAP-MODE Aerosol to FAST-JX photolysis. michaelhollaway
#2741 Addition of new diagnostic for the dry deposition flux of nitrogen fionaoconnor
#2742 Addition of new diagnostic for the wet deposition flux of nitrogen fionaoconnor
#2814 Allow UKCA Chemistry to work with aerosol climatologies lukeabraham
#2898 Input CO2 concentrations into UKCA when interactive CO2 is operational colinjohnson
#3025 Add functionality to use NetCDF GLOMAP-mode climatology fields to generate fields required by ACTIVATE alanjhewitt
#3028 Improve STASH section references related to UKCA components. mohitdalvi
#3127 Investigate KGO differences in um_n48_ukca_eg app using GNU on the Cray lukeabraham
#3209 column-call UKCA Newton-Raphson solver lukeabraham
#3267 Fix UKCA NRUNvsCRUN test
(CRUN running full 2-days)
mohitdalvi
  • Michael Hollaway: I currently have a test version of the code working that uses RADAER to pass the aerosol optical properties to FAST-JX. This runs at UM-UKCA 10.6.1. I am currently working on testing the code with different MODE setups and then will work on going through the required steps to get the code into the trunk at 10.9 (sci/tech review, code review, etc). All being well the updates should hopefully be ready for 10.9 (fingers crossed).

Tickets currently targeted as Hopefully

Highlights/Targets from CSIRO

  1. Matt W. has incorporated a primary marine organic aerosol emission parameterisation into GLOMAP (included in the trunk from UMvn10.6).
  2. We are close to finishing further work on parameterisation of ozone dry deposition to the ocean (work so far reported in the ACP paper, 17, 3749-3767, 2017). I have communicated with Gerd Folberth (MO) as to the process to put the new parameterisation into the trunk once our studies are finished.
  3. A CSIRO project with Melbourne University has started which aims to have an ACCESS coupled model with UKCA CheST configuration at UM10.6 and GA7.1 (main ACCESS components are UM atmos + MOM ocean + sea ice + CABLE land surface + UKCA).
  4. ACCESS-CM2 coupled model is being tested for aerosols (including dust, at GA7.1).

Virtual Machine (LA)

UM Virtual Machine Guide - Note: this will be superseded by UMPB X10 from vn10.8 onwards.

Training (LA)

UKCA Training January 2018

AOB/DNM (LA)

Date of next meeting (LA)