Changes between Version 10 and Version 11 of GroupActivities/Meetings/Retreat20190527

Timestamp:

2019-06-12T14:22:13+02:00 (5 years ago)

Author:

peylin

Comment:

--

GroupActivities/Meetings/Retreat20190527

@@ -143 +143 @@
 '''Organisation:''' 
 
- - Overview slides to show how the different parts of this session fit together (MATT, 5 mins).
-Why have we decided to organize the session this way? Logical ties between presentations
+ - Overview slides to show how the different parts of this session fit together (MATT, 5 mins)[attachment:Mcgrath_ORCHIDEE_retreat_parameters.pdf​]. Why have we decided to organize the session this way? Logical ties between presentations
 
-Current situation of parameter optimization (15 min + 5 min for urgent questions)
-Problem with creating uncertainty in parameters due to optimizing against only one or two datasets and implicitly incorporating biases (TASH, 5 min)
-The needs for a Bayesian calibration to define always physical range of variations and parameter uncertainties. (PHILIPPE, 5 min)
-Is there a preferred spatial and/or temporal scale to optimize a specific parameter (GUILLAUME, 5 min)
-Relatively easy enhancements on top of what we have now (30 min + 5 min for urgent questions
-Outline of a database to enhance the traceability of ORCHIDEE’s model parameters (SEBASTIAAN, 5 min)
-Technical considerations about managing parameters in ORCHIDEE (JAN, 10 min)
+ - Current situation of parameter optimization:  The needs for a Bayesian calibration to define always physical range of variations and parameter uncertainties. (PHILIPPE, 5 min) [attachment:Peylin_Data_assimilation.pdf]; Problem with creating uncertainty in parameters due to optimizing against only one or two datasets and implicitly incorporating biases (TASH, 5 min)
+
+ - Is there a preferred spatial and/or temporal scale to optimize a specific parameter (GUILLAUME, 5 min)[attachment:Marie_ORCHIDEE_parameters.pdf]
+
+ - Relatively easy enhancements on top of what we have now (30 min + 5 min for urgent questions Outline of a database to enhance the traceability of ORCHIDEE’s model parameters (SEBASTIAAN, 5 min) [attachment:Luyssaert_Parameters.pdf]
+
+ - Technical considerations about managing parameters in ORCHIDEE (JAN, 10 min) [attachment:Polcher_RunDef_issues.pdf]
+
 Changes to speed-up the model to enhance parameterization (20 min + 5 min for urgent questions)
 How can we speed-up ORCHIDEE? (MATT, 5 min)
@@ -164 +164 @@
 '''Short summary of the session:'''
 
-The session was divided up into two main sections: presentations and discussion.  The presentations were divided up into three main sections: what is currently being done, ideas for short-term items which could improve parameter documentation and transparency, and a few long-term items related to parameters which will require significant effort and funding.  The idea was to focus discussion around the short-term items.
+ - The session was divided up into two main sections: presentations and discussion.  The presentations were divided up into three main sections: what is currently being done, ideas for short-term items which could improve parameter documentation and transparency, and a few long-term items related to parameters which will require significant effort and funding.  The idea was to focus discussion around the short-term items.
 
-Philippe, Tash, and Guillaume presented on efforts currently being made regarding parameter optimization, as well as raising questions to be considered when optimizing.  
+ - Philippe, Tash, and Guillaume presented on efforts currently being made regarding parameter optimization, as well as raising questions to be considered when optimizing.  
 
-Sebastiaan and Jan presented two ideas to improve parameter documentation, touching on two different aspects.  Sebastiaan discussed the need for documenting the origins of parameters, developing some kind of database that stores parameter values with metadata indicating the origin and scripts which update parameter values from existing databases as those databases themselves are updated.  Such a database could be converted into a run.def through the use of an additional script, and becomes useful when multiple people are attempting to parameterize simultaneously.  Jan introduced a method using XML files that could be formatted as run.defs. Such script would allow to do the quality control on the parameters before they enter the model. The approach Jan proposed would be a very nice last step of the traceability Chain presented by SL.
+ - Sebastiaan and Jan presented two ideas to improve parameter documentation, touching on two different aspects.  Sebastiaan discussed the need for documenting the origins of parameters, developing some kind of database that stores parameter values with metadata indicating the origin and scripts which update parameter values from existing databases as those databases themselves are updated.  Such a database could be converted into a run.def through the use of an additional script, and becomes useful when multiple people are attempting to parameterize simultaneously.  Jan introduced a method using XML files that could be formatted as run.defs. Such script would allow to do the quality control on the parameters before they enter the model. The approach Jan proposed would be a very nice last step of the traceability Chain presented by SL.
 
-Matt introduced thinking about ways parameters could be improved if the model speed was significantly increased.  For example, if emulation of certain parts or the entire model include parameters as inputs.  Such speed-up would allow for more thorough sampling of parameter space, in part because the nutrient pool spin-ups could be carried out more efficiently. 
+ - Matt introduced thinking about ways parameters could be improved if the model speed was significantly increased.  For example, if emulation of certain parts or the entire model include parameters as inputs.  Such speed-up would allow for more thorough sampling of parameter space, in part because the nutrient pool spin-ups could be carried out more efficiently. 
 Aude and March showed how some of the model parameters could be replaced by traits which could help to make use of trait-trait relationships and the coordination theory for photosynthesis as well as for phenology. 
 
 '''List of actions / follow up:'''
 
-The presenters had the impression that the ORCHIDEE group does not experience a sense of urgency related to the ORCHIDEE parameters. No immediate actions will be taken. 
-Matt submitted a parameter-related proposal, Aude and Marc are preparing parameter-related proposals. Philippe, Vlad and Tash will continue their work on parameters
-Tash is setting up a sensitivity analysis of the ORCHIDEE model
-Sebastiaan will look for a way to start developing parts of parameter traceability chain.
+ - The presenters had the impression that the ORCHIDEE group does not experience a sense of urgency related to the ORCHIDEE parameters. No immediate actions will be taken. 
+ - Matt submitted a parameter-related proposal, Aude and Marc are preparing parameter-related proposals. Philippe, Vlad and Tash will continue their work on parameters optimisation.
+ - Tash is setting up a sensitivity analysis of the ORCHIDEE model
+ - Sebastiaan will look for a way to start developing parts of parameter traceability chain.