[GO-ESSP] Metadata for Climate Models Development (2bd attempt)

[K. A. Bouton]

Sorry go-essp - bounced - trying again

-----Original Message-----
From: K. A. Bouton [mailto:k.a.bouton at rdg.ac.uk] 
Sent: Wednesday, February 02, 2005 4:17 PM
To: 'taylor13 at llnl.gov'; 'Ronald.Stouffer at noaa.gov'
Cc: 'Chris.Kerr at noaa.gov'; Lois Steenman-Clark
(L.Steenman-Clark at reading.ac.uk); 'climate-model-doc at tornado.badc.rl.ac.uk';
'go-essp at ucar.edu.'; 'jordan.alpert at noaa.gov'
Subject: Metadata for Climate Models Development


This is sent to the GO-ESSP and the BADC mailing lists as trying to expand
the discussion. Sorry for the multiple hits.

Dear Karl and all other interested cc'd people


Glad to see that you and Chris Kerr (GFDL) are talking again about the model
metadata. I thought maybe I could help with some of your comments, given
that we at U Reading are the lead in developing this model metadata
standard.

The development of the schema has progressed quite a bit in the 6 months
since we last corresponded - and it still has a ways to go.  But I think
many of your questions to Chris below are in fact being discussed and
addressed.

When I look at your questionnaire to IPCC modellers, I think our proposed
schema would work very well for you.  The standard could answers most of the
questions you pose - and provide a structure to the answers such that they
could be more easily compared and contrasted.

> My impression is that some of your metadata is directed more
> toward a user interested in applying the model, especially 
> within an earth system model "framework" (e.g., technical 
> description, interface requirements, and input/output 
> requirements).  Some of the information we asked for is 
> redundant in that the files containing the data already 
> include this information (e.g., resolution).  A model can be 
> run at various resolutions, so this is not really a 
> characteristic of the model, but of the simulation.  In the 
> case of the IPCC simulations the same model run at different 
> resolutions is treated as a "different" model.

The 'complete suite' would be made up of a hierachy of schemas - 1.the model
metadata will describe the model, 
  2. all of its components (atmosphere, ocean, land etc) 
    3. all the simulations done using them. 

We leave the description of the output data itself to something like CF

I agree that the example that Chris sent you was for what we would define as
a simulation. His documentation was for a simulation (AM2) using the
atmospheric component of the GFDL Flexible Modelling System (FMS).  To see
how we would implement that using our schema, follow the links below.

Our envisioned user is, at the moment, the more experienced scientist who is
pretty savvy as to what is what - and who could be someone interested in
coupling the model, but also interested in just what was done inside the
model to create a simulation - ie what gravity wave drag scheme and settings
were used.

> There are three types of information that don't be appear to
> be part of your example, but which are of interest to those 
> performing analysis of model output:
> 
> 1) Information providing a higher level description of the
> model (i.e., the "coupled model" itself, as opposed to the 
> individual components, which you have described).  Of 
> particular interest are the "coupling details" (see section 
> IV-E of the attachment).

As mentione above, this is indeed part of the schema we are developing. The
highest level of the schema would describe the model itself, and its
components.  We envision it as something written (once for each version) by
the model developers who can best describe the code, detailing all the
possibilities - ie it CAN run atmosphere, it CAN run a) GWD scheme 1 or b)
GWD scheme 2.

Then the schema for the simulation/experiment itself is just an instance of
that defined model - ie which of the possiblilities were actually
'set'/used. Eg my simulation used model A with compnent atm B and GWD 1.

 
> 2)  Information about which "benchmark" tests have been
> performed with the model (e.g., AMIP, CMIP, PILPS, etc.).  
> These could either be simply listed or links could be 
> provided to web pages describing the results.

Good idea. We had not thought of this. 
> 
> 3)  Information about experiments performed with the model
> (e.g., IPCC climate of the 20th Century simulations).  See 
> section V of our questionnaire for some aspects of particular 
> interest.

Again as mentioned above, this is the third part of the suite of our schema
- that which defines the simulation. In fact this will probably be the most
used part of the suite.

> Getting to your question concerning how to proceed, I think
> we should first refine (and perhaps generalize to other model 
> components) what you've done.  

I have taken Chris' AM2 simulation and put it into the model metadata
schema.

You can see how the suite or hierachy works for the documentation he sent by
following the links below through the description of the model, its
components and the experiments done using them.

First go to http://ugamp.nerc.ac.uk/bouton/model_metadata/implementation.php
Select list all model and their components

Here you will see the FMS and the atmosphere component listed below it.
Select view details.
http://ugamp.nerc.ac.uk/bouton/model_metadata/list_component_details.php?com
ponent_ID=18
This lists the possibilities of the component (based onwhat he sent me so it
reaaly is not the complete set of possibilities for the atmosphere component
on the FMS)

Then to see his simulation - click on related experiments AM2 simulation
just above technical properties
http://ugamp.nerc.ac.uk/bouton/model_metadata/list_experiment_details.php?ex
periment_ID=12

This provides the metadata for the experiment.

All of this is also available as XML which is how we imagine this being
passed around - allowing for any sort of GUI etc to be used.

You can also look at the hiearchy graphically
http://ugamp.nerc.ac.uk/bouton/model_metadata/xml/graphical_representation/E
arleySuite.php

You can see how it doesn't quite fit because his physical paramters for
example are quite descriptive, rather than an on off setting. 


The next step as we see it involves several things
1) How to combine the descriptive type of info sent by Chris, with the type
of info we get out ot the Unified Model?

2) flushing out the numerical properties.  We are very weak on describing
the vertical, horizontal and time representations. Belaji et al are working
hard on this and we want to see how we can use the work they are doing.

3) working as a community to define standard names for the physical and
dynamical properties

4) Trying to apply the schema to other models.  We really need people to
start looking at and trying this out!

This is where Chris' work has been great. We have been able to plug a model
other than our own in and seen where the schema works and doesn't work.

I talked to Dean Williams at AMS who mentioned either Bob Drach or their new
hire may begin to work with us in applying this to the IPCC models but not
until after the data flood ends after March 2005.  This would be perfect to
show where the holes are that need filling.

Also at the AMS meeting in January, Jordan Alpert has also expressed some
interest in getting his NOAA/NOMADS project involved, and trying to use the
schema to describe his models.

5. Developing tools to automate the process.

> Then we might discuss it at the GO-ESSP workshop in
> June.  Finally, we could present it to the WGCM (Ron knows
> about this) and see if we can develop support there to make 
> collection of this information an integral part of the model 
> intercomparison projects they oversee.

It is in the works to discuss this further at the next GO-ESP mtg in June as
they have adopted the project as one of interest to them.

I think at this point it is to early to be talking WGCM, but certainly down
the line when this gets further developed, there should be some governing
body to continue this what is now a grassroots effort.

I know this kind of thing is hard to look at and understand the flow of it -
a whole new language. So if anyone has any questions, comments - feel free.

Regards

Katherine

**************************************************
Dr Katherine A Bouton
Centre for Global Atmospheric Modelling (CGAM)
Department of Meteorology
University of Reading
Earley Gate, PO Box 243
Reading RG6 6BB
UK
Tel: 44 (0)118 378-7841