This wiki page is associated with ticket#630
Hereafter are some guidelines for Statistical Downscaling experts who agreed to review SD Controlled Vocabulary.
What is METAFOR, What Controlled Vocabulary means ?
METAFOR is a European Project that aims to help the Climate modelling community in identification, assessment and use of climate data stored in digital repositories. This is performed through the construction of a standard metadata framework. The core of this metadata framework is the Common Information Model (CIM) that gives the way the information collected is to be organized and linked together. The CIM is a flexible and generic information model that makes it very powerful. It needs to be narrowed-down to be applied to a specific field (e.g. Climate Modelling Intercomparison Projects, CMIPs). This requires to define an accurate set of appropriate terms: the so-called controlled vocabulary (CV) that aims to be a common vocabulary shared and used by the different climate modelling groups.
A specific CV has been built for the GCMs that are running the CMIP5 exercise. It describes the whole modelling system deployed to produce a given data set : GCM themselves detailing each model component (ocean, atmosphere, land surface, sea ice, etc.) and the configuration (initial conditions, forcing, etc.) set-up to run the simulations in the designed framework of a given experiment. A web questionnaire has been developed to collect metadata from the modelling groups (the CMIP5 Questionnaire).
Such standardized climate metadata bank, accessible via discovery and search tools, will reinforce data interoperability and dissemination and will facilitate their interpretation and comparison.
Why a Controlled Vocabulary for Statistical Downscaling ?
Even if METAFOR has been primarily focusing on global climate modelling, it aims at improving data documentation and understandability within all the research fields based on climate data analysis. Regional climate downscaling data can also benefit from the metadata expertise and tools METAFOR has gained. As for Regional Climate Models (dynamical downscaling), they can easily be described through the metadata framework developed for global climate modelling. Statistical Downscaling (SD) requires specific CV. The CIM is sufficiently generic to support metadata for Statistical techniques and a metadata collecting tool is planned to be derived from the existing CMIP5 questionnaire.
In other words, the objective is to apply the metadata building procedure METAFOR used for large-scale Climate Modelling data to Statistical Downscaling data. The end-user communities targeted are the one of the Impact and Adaptation. A standard metadata convention for SD, shared by all downscaled data providers should increase data interoperability and avoid misleading usage of the data.
This task METAFOR is pushing ahead should benefit to other projects like the European project IS-ENES and the WCRP-sponsored CORDEX program. IS-ENES has concerns about downscaled data distribution and will rely on METAFOR expertise and developments for everything related to global and regional climate metadata. The CORDEX coordination program has among its objectives to provide a framework for the evaluation and intercomparison of regional downscaling models and methods. In this context, it is planned to define standards for production and dissemination of downscaled climate data. METAFOR offers here to contribute to the standards for data dissemination.
What is the content of Statistical Downscaling CV?
What we need is to build a structured and hierarchical CV able to describe each Statistical Downscaling method. The granularity of this CV (i.e. the pieces of information to be refined and the level of details to be reached) is something to be agreed on relying on expert knowledge. The questions driving the granularity answer is :
"How much do I need to know about the method that produced this downscaled data set to allow its unambiguous distinction from the others?"
"What are the key-parameters useful to scientifically decipher and analyse these downscaled data?"
The spread in the SD methods is very large: weather typing, weather generator, transfer functions,... some combining several classical methods. That's why we need inputs and opinion from a large number of SD experts.
The draft CV to review
Based on some publications and reports (mainly Wilby et al. 2004, Herrera et al. 2006) we made a first attempt of Controlled Vocabulary content and structure. It is handled in the form of an heuristic map (a mindMap) with some key coding rules detailed in the next paragraph.
The main issues to have in mind when reviewing this CV are:
- "Can I describe the method I am used to with this vocabulary structure?"
- "Is this CV complete enough to cover the wide range of methods?"
- "On the contrary, are some pieces of the CV too much detailed or specific ?"
Reviewers are highly encouraged to criticize, correct, make suggestions and comments and even ask for a full restructuring if the current is not suitable.
Feedbacks can be provided by e-mail to moine<at>cerfacs<dot>fr.
Some rules to read statistical downscaling CV mindmap
The mindMap is written according to some visual rules (colors, fonts, icons) we fixed in order to apply automatic tools for ingestion into a web-questionnaire. These rules give meaning to the elements of the chart.
Fonts are for typing the different elements of the structure:
- Brown = leaf parameter : a single parameter expecting a value.
- Purple = parameter group : grouping together leaf parameters under a single concept.
- Black = component : gathering purple parameter groups in an "editable" container called component. Editable means that user can :
- activate/inactivate the component;
- define new leaf parameters within a given component;
- attach references to the component;
- add new sibling components.
Icons are for specifying the type of values expected for a given leaf parameter:
- pencil icon = free-text.
- number-1 icon = numerical value.
- red cross = exclusive choice among a proposed list of values.
- green tick mark = possibly multiple choice among a proposed list of values.
Construction marks:
- italic font = just comments (query marks is for issues to be solved or remaining question)
- warning signal = elements to be removed or not finished.
- light signal = modified following recommendations from reviewers.
Contributors
Here is reproduced the list of interviews we have to do to capture feedbacks from the statistical downscaling expert on the initial CV draft.
A compilation of expert feedbacks is available on the SD feedbacks page.
| Method | Downcaling expert | Metafor person in charge of the contact | Status (date:MM/DD/YY)' |
| DSCLIM (weather typing) | Christian Page (CERFACS, France) page<at>cerfacs<dot>fr | Marie-Pierre | has provided feedbacks (11/29/10) not yet included in MindMap... |
| transfer function method (derived from M. Dequé) | Michel Kolasinski (CLIMPACT, France) ko<at>climpact<dot>com Paul-Antoine Michelangeli (CLIMPACT, France) pam<at>climpact<dot>com | Michel | have provided feedbacks (09/10/10) included in MindMap - Done. |
| extreme downscaling | Michel Kolasinski (CLIMPACT, France) ko<at>climpact<dot>com Paul-Antoine Michelangeli (CLIMPACT, France pam<at>climpact<dot>com | Michel | have provided feedbacks (09/10/10) included in MindMap - Done. |
| anomalies method | Michel Dequé (CNRM, France) michel<dot>deque<at>meteo<dot>fr | Marie-Pierre | has provided feedbacks (09/07/10) included in MindMap - Done. |
| hybrid statistico-dynamical (wind downscaling) | Julien Najac (EDF, France) julien<dot>najac<at>edf<dot>fr | Marie-Pierre | contacted (11/18/10) agreed to review (12/09/10) waiting for feedbacks... |
| methods available on the downscaling portal | Antonio S. Cofino (univ. of Cantabria, Spain) antonio<dot>cofino<at>unican<dot>es | Marie-Pierre | agreed to review re-contacted (11/18/10) waiting for feefbacks... |
| weather generator | C. Flécher (CLIMATOR project, France) | Marie-Pierre (via Christian) | not contacted yet... |
| other methods | Mathieu Vrac (LSCE/IPSL, France) mathieu<dot>vrac<at>lsce.ipsl<dot>fr | Eric | contacted (09/06/10) agreed to review(09/06/10) re-contacted (01/05/11) no news... |
| statistical downscaling over the Jordan river basin | Imad Khatib (PAST, Palestine) ikhatib<at>palestineacademy<dot>org | Marie-Pierre | agreed to review (09/30/10) contacted (11/18/10) no news... |
| other methods | Markku RUMMUKAINEN (SMHI, Sweden) markku<dot>rummukainen<at>smhi<dot>se | Marie-Pierre | contacted (11/18/10) redirected us to C. GOODESS (12/04/10) - Done. |
| other methods | Clare GOODESS (UEA, UK) | Marie-Pierre | contacted (01/04/11) no news for the moment... |
| other methods | Fatima DRIOUECH(Marrocan Met Services, Morocco) driouechfatima<at>yahoo<dot>fr | Marie-Pierre | contacted (11/18/10) no news... |
