Reports of technical meeting ORCHIDEE

7/07/2011

Present: Martial, Fabienne, Sebastiaan, Didier, Nicolas Vio, Marie-Alice, Philippe

Apres un tour de table general, les points qui ressortent sont:

1) DGVM Les divers bug ont ete corriges; les resultats obtenus sont correcte sauf pour certaines regions (notamment Hlat). Les Nicolas realisent encore qqs tests pour ameliorer les H.lat PUIS "Tag" de la version par Martial (date limite au 14 Jui);

Ensuite prise en charge/test par les personnes directement interessee (ex: Masa)

2) Listes de diffusion Besoin d'une liste pour la diffusion des "commit" (differente de la liste projet); liste de developpement.

Au total nous auront 4 liste de diffusion:

• liste global ORCHIDEE@ipsl

• ORCHIDEE_projet (a migrer a ipsl)

• ORCHIDEE developpement: pour tous les mails d'info technique

• ORCHIDEE support: help-orchidee@ipsl

La demande pour la liste help a ete faite; probleme de personnel en vacance pour sa creation imediate (MA s'en occupe)

3) DOCUMENTATION

Martial, Sebastiaan ont regarde l'utilisation du logiciel Doxygen; les fonctionnalites desirees existent pour les prog C mais il faut regarder encore pour f90 ==> Bilan final en Septembre

Ensuite Organisation d'une RETRAITE de 1 Jour dans un lieu "agreable" pour ecrire ensemble la DOC (solution retenue par tous!)

IL existe une pression supplementaire pour la DOC, suite a l'initiative QUALITE engagee par le CEA (Fabienne est responsable cote LSCE); ORC est une composante ciblee, i.e. la direction du CEA sera attentive a nos efforts.

METAPHORE: Ce document en cours de creation doit decrir les composantes du systeme terre ipsl. On pourra utiliser ce document comme premiere base pour la documentation. (a remplir par tous durant l'ete) (Martial peut le faire circuler ?)

2) DIVERS

• Parallelisation de Test_stomate en cours (Martial)

• Poste Benchmark simulations couple ouvert par CEA (Pascal B.): verifier

aupres de Pascal quel support cela peut apporter pour ORC? (Philippe P)

• Mise en place du cours sur l'utilisation du systeme subversion: Martial+Didier

17/11/2010

Presentation of the new versioning system svn by Martial Mancip

1 - Design (see ​https://forge.ipsl.jussieu.fr/orchidee/browser)

• The tag directory was first created by importing the last cvs 1_9_5 tag (AR5 version) including both ORCHIDEE and ORCHIDEE_OL (meaning source code, scripts, diagnostics and documentation).
  
• The trunk directory was then created by copying this first tag. This directory represents the current version under development and is to be tagged when ready.
  
• The branches directory will include separately the major on-going developments (nitrogen, phenology, assimilation, ...). When ready, they will be merged into the trunk directory.
  
• The perso directory will include separately all developments made by students and post-docts.
  
• Note that the release number is globally incremented each time any file is modified (this is a major difference from cvs).
  
• Click on Revtree (upper right) to visualize the Revision Tree and see the model evolution (select options then click on 'Update').
  

2 - Project Management Tools (labels 'See Tickets', 'New Ticket', 'Roadmap', 'Timeline')

• A New Ticket is created each time an action is required (bug to resolve, new branch to create, enhancement required, ...).
  
• Tickets are associated to Milestones, defining the main objectives we are currently trying to reach.
  
• Tickets are listed and discussed every two weeks during the technical meetings.
  
• The Roadmap enables to visualize the different Milestones completion.
  

3 - Authorizations

• Each user will have a wiki login and password.
  
• Already existing groups are adm (Martial Mancip, Nicolas Vuichard, Didier Solyga, Fabienne Maignan), group (orchidee project), users and assimilation. A group can be created for each corresponding branch (e.g. pheno).
• Trunk commits are decided within the orchidee project meetings.
  

4 - Next actions

• Decide which commits are sent by e-mail to which list.
  
• Foresee another technical meeting to declare Milestones, create branches, ...
  

04/11/2010

1 - SVN :
As all peole asked to at the last meeting, Martial install the new version management software : SVN (subversion). SVN will replace the current version management software CVS. More detailed talk about the characteristics and the possibilties of SVN at the next meeting.
You could find the (free) SVN-book (in french) here : ​http://svnbook.red-bean.com/

2 - About the externalisation of the parameters
(see attached files : work doc​ and presentation​)

• We agreed to externalise all the potential parameters
• Concerning the splitting of the run.def : the run.def will be divided in three parts. one part called orchidee_run.def, containing all the options of the run. Two parts containing all the physical parameters of the model called sechiba_param.def and stomate_parameters.def. These two last files will be joined by two files called sechiba_parameters.txt and stomate_parameters.txt containing the list of the default valuesof all the "externalised" parameters.
• When it is possible, we agreed that the config key in the run.def and the name of the parameter in the code will be the same.
• Didier should write a script which will test if a parameter is already defined in one of the run.def

3 - Next Technical Meeting : 18th November 2010 (10h)

Suggested topics : Presentation of SVN by Martial

From Peylin's mail 21/11/10

Discussion sur l'externalisation des parametres et le travail de Didier Solyga :

• Externalisation de tous les parametres possible (regroupement dans 3 fichiers differents)
• Le fichier run.def contient seulement les include pour ces 3 fichiers:
  • orchidee_run.def: contient que les options du run
  • sechiba_param.def: contient param physiques
  • stomate_param.def: contient param biophysiques
• Ces trois fichiers contiendront seulement les parametres que l'on veut modifier (parametres des PFT que l'on redefinit a partir des metaclass du modele)
• On créé aussi 3 fichiers (orchidee_run.txt, sechiba_param.txt, stomate_param.txt) qui contiennent TOUS les parametres possibles avec leur valeur par defaut. Ces fichiers ne sont pas utiliser par ORC mais servent de documentation/reference.

18/10/2010

• Work on the externalisation of the I/O of ORCHIDEE. Yann Meurdesoif presents the development of XMLIO-Server aiming at replacing IOIPSL in the future. Before this, it can serve as an interface between a code (such as ORCHIDEE) and IOIPSL. It is based on a XML file and a library (initially written in Fortran -> move to C). (see attached file XMLIO_SERVER_ORCHIDEE.pdf​)
  
  • We can expect to have a version of ORCHIDEE using XMLIO-Server for February 2011.
    
  • We can already work on the groups of variables that have to be defined in the XML file.
    
• Point on the SVN server: We agree to move now from CVS to SVN. We will maintain the CVS system as long as needed for the CMIP5 exercise (normally just one update has to be done when Patricia will be back).
• Next technical meeting : 4th November 2010 10pm
  Suggested topics :
  1. report of Didier on the externalization

30/09/2010

• Bug report on interpolation of a 6-hourly climate dataset (see attached file 6_hourly_interpolation.pdf​)
  The routine 'forcing_read_interpol' in readdim2.f90 file works fine with climate datasets such as NCEP/NCAR in which:
  
  • State variables are forecasts valid 6 hours after the reference time
    
  • Flux variables are 6 hour averages starting at the reference time
    For dataset such as ERA-interim, there are different time lags for state variables and flux variables and a specific problem
    for the interpolation of SWdown. Cédric Bacour already fixed this last bug.
    To be done
  • To define different flags helping to characterize the temporal structure of a climate dataset:
    Instanteous vs Average variables, Centered vs shifted(+1 timestep) value, GMT vs local-time data
  • To auto-document any Netcdf dataset used by ORCHIDEE with these flags. This will prevent of using datasets without knowing these characteristics
  • To modify the 'forcing_read_interpol' routine in order to account for these flags
    

• Output variables
  In order to reduce the number of output variables (ie avoid 2 variables for a same physical entity), we (N. Vuichard and M. Mancip) will circulate the ALMA and CMIP5 standard and the list of output variables that are not in these 2 standards.
  Specific request :
  • to modify the standard driver for being able to specify as a list, the variables that a user wants as outputs
  • modification of the call to histdef function : loop over the outputs variables stored in a text file (like in viovy's driver)
    
• Next technical meeting : 14th October 2010 10pm
  Suggested topics :
  1. report of Didier on the externalization
  2. work on the output variables (to continue)

16/09/2010

• introduction on forge ORCHIDEE wiki : create this page
• short words from Phillipe on the organization of those meetings
  • every two weeks
  • organized by Martial, supplied by Nicolas Vuichard
  • reports are done during the meeting or just after
    and all participants have to read and perhaps correct the report
• presentation of work on externalization of parameters (see attached file ​pdf)
  • definition of readding PFTs versus Meta-Classes : discution of idea of Didier versus Sebastiaan.
    We don't keep idea that definition of PFT are in MTC sets (Sebastiaan)
    But for clearing the def file, Nicolas has proposed that we give number of MTC for each PFT per line. Accepted.
  • simplification of all parameters in code : regroup all constant modules (MTC parameters) in src_parameters and stomate_constants (all participant are agreed ).
    Discussion on initialization functions
    We have to talk to all ORCHIDEE developpers before this fusion. Two steps
    
    1. put all STOMATE constant in src_parameter
    2. fusion if needed all constant files. Some SAVE values are not parameters and lot of MTC array are SAVE and will be PARAMETER.
  • Possible flags related to the externalization
    
    1. flag CONFIG_AR5 in order to get back to the AR5 config (13 PFT with default values for all PFT parametes)
       This corresponds to only define nvm=13 and pft_to_mtc = 1 2 3 4 5 6 7 8 9 10 11 12 13
       We wont' implement the CONFIG_AR5 flag
       
    2. flag IMPOSE_PARAM : a user has developed a orchidee.def with for instance 16 PFT and has modified some parameter values
       In order to be able to use this .def file but for a run with default values (eg without removing the lines in the .def that redefined the parameter values)
       we set IMPOSE_PARAM to FALSE else we set it to TRUE.
  • All vegetation maps for variable number of PFTs will be land use like maps (result of pre-process in the right number of types).
• Spacialization of variables : Didier has asked for a list of spatialization parameters to begin change in the code.
  We decide to begin the work with vcmax_opt. Nicolas Viovy has proposed to build a map by PFTs.
  
• Next technical meeting : 30th September 2010 2pm
  Suggested topics (to be confirmed) :
  1. interpolation of 6-hourly climatology
  2. bug reported by Tea/Pierre? on water stress